From 218ffcb12362aec2bd207c613f3842cda5c6f6a8 Mon Sep 17 00:00:00 2001
From: ModelHub XC <noreply@modelhub.org.cn>
Date: Thu, 23 Jul 2026 04:18:09 +0800
Subject: [PATCH] =?UTF-8?q?=E5=88=9D=E5=A7=8B=E5=8C=96=E9=A1=B9=E7=9B=AE?=
 =?UTF-8?q?=EF=BC=8C=E7=94=B1ModelHub=20XC=E7=A4=BE=E5=8C=BA=E6=8F=90?=
 =?UTF-8?q?=E4=BE=9B=E6=A8=A1=E5=9E=8B?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Model: heyalexchoi/qwen3-1.7b-math-sft-v3a
Source: Original Platform
---
 .gitattributes                         |   36 +
 README.md                              |   58 +
 chat_template.jinja                    |   85 ++
 config.json                            |   63 +
 generation_config.json                 |   10 +
 last-checkpoint/chat_template.jinja    |   85 ++
 last-checkpoint/config.json            |   63 +
 last-checkpoint/generation_config.json |   10 +
 last-checkpoint/model.safetensors      |    3 +
 last-checkpoint/optimizer.pt           |    3 +
 last-checkpoint/rng_state.pth          |    3 +
 last-checkpoint/scheduler.pt           |    3 +
 last-checkpoint/tokenizer.json         |    3 +
 last-checkpoint/tokenizer_config.json  |   30 +
 last-checkpoint/trainer_state.json     | 1488 ++++++++++++++++++++++++
 last-checkpoint/training_args.bin      |    3 +
 model.safetensors                      |    3 +
 tokenizer.json                         |    3 +
 tokenizer_config.json                  |   30 +
 training_args.bin                      |    3 +
 20 files changed, 1985 insertions(+)
 create mode 100644 .gitattributes
 create mode 100644 README.md
 create mode 100644 chat_template.jinja
 create mode 100644 config.json
 create mode 100644 generation_config.json
 create mode 100644 last-checkpoint/chat_template.jinja
 create mode 100644 last-checkpoint/config.json
 create mode 100644 last-checkpoint/generation_config.json
 create mode 100644 last-checkpoint/model.safetensors
 create mode 100644 last-checkpoint/optimizer.pt
 create mode 100644 last-checkpoint/rng_state.pth
 create mode 100644 last-checkpoint/scheduler.pt
 create mode 100644 last-checkpoint/tokenizer.json
 create mode 100644 last-checkpoint/tokenizer_config.json
 create mode 100644 last-checkpoint/trainer_state.json
 create mode 100644 last-checkpoint/training_args.bin
 create mode 100644 model.safetensors
 create mode 100644 tokenizer.json
 create mode 100644 tokenizer_config.json
 create mode 100644 training_args.bin

diff --git a/.gitattributes b/.gitattributes
new file mode 100644
index 0000000..52373fe
--- /dev/null
+++ b/.gitattributes
@@ -0,0 +1,36 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text
diff --git a/README.md b/README.md
new file mode 100644
index 0000000..ba015a7
--- /dev/null
+++ b/README.md
@@ -0,0 +1,58 @@
+---
+base_model: Qwen/Qwen3-1.7B-Base
+library_name: transformers
+model_name: qwen3-1.7b-math-sft-v3a
+tags:
+- generated_from_trainer
+- trl
+- sft
+licence: license
+---
+
+# Model Card for qwen3-1.7b-math-sft-v3a
+
+This model is a fine-tuned version of [Qwen/Qwen3-1.7B-Base](https://huggingface.co/Qwen/Qwen3-1.7B-Base).
+It has been trained using [TRL](https://github.com/huggingface/trl).
+
+## Quick start
+
+```python
+from transformers import pipeline
+
+question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
+generator = pipeline("text-generation", model="heyalexchoi/qwen3-1.7b-math-sft-v3a", device="cuda")
+output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
+print(output["generated_text"])
+```
+
+## Training procedure
+
+[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/heyalexchoi/qwen3-math-rlvr/runs/y516favu) 
+
+
+
+This model was trained with SFT.
+
+### Framework versions
+
+- TRL: 1.7.0
+- Transformers: 5.10.2
+- Pytorch: 2.11.0+cu129
+- Datasets: 5.0.0
+- Tokenizers: 0.22.2
+
+## Citations
+
+
+
+Cite TRL as:
+    
+```bibtex
+@software{vonwerra2020trl,
+  title   = {{TRL: Transformers Reinforcement Learning}},
+  author  = {von Werra, Leandro and Belkada, Younes and Tunstall, Lewis and Beeching, Edward and Thrush, Tristan and Lambert, Nathan and Huang, Shengyi and Rasul, Kashif and Gallouédec, Quentin},
+  license = {Apache-2.0},
+  url     = {https://github.com/huggingface/trl},
+  year    = {2020}
+}
+```
\ No newline at end of file
diff --git a/chat_template.jinja b/chat_template.jinja
new file mode 100644
index 0000000..699ff8d
--- /dev/null
+++ b/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/config.json b/config.json
new file mode 100644
index 0000000..4cab803
--- /dev/null
+++ b/config.json
@@ -0,0 +1,63 @@
+{
+  "architectures": [
+    "Qwen3ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": null,
+  "dtype": "bfloat16",
+  "eos_token_id": 151645,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 2048,
+  "initializer_range": 0.02,
+  "intermediate_size": 6144,
+  "layer_types": [
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention"
+  ],
+  "max_position_embeddings": 32768,
+  "max_window_layers": 28,
+  "model_type": "qwen3",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 8,
+  "pad_token_id": 151643,
+  "rms_norm_eps": 1e-06,
+  "rope_parameters": {
+    "rope_theta": 1000000,
+    "rope_type": "default"
+  },
+  "sliding_window": null,
+  "tie_word_embeddings": true,
+  "transformers_version": "5.10.2",
+  "use_cache": false,
+  "use_sliding_window": false,
+  "vocab_size": 151936
+}
diff --git a/generation_config.json b/generation_config.json
new file mode 100644
index 0000000..fbf2077
--- /dev/null
+++ b/generation_config.json
@@ -0,0 +1,10 @@
+{
+  "do_sample": false,
+  "eos_token_id": [
+    151645,
+    151643
+  ],
+  "max_new_tokens": 2048,
+  "pad_token_id": 151643,
+  "transformers_version": "5.10.2"
+}
diff --git a/last-checkpoint/chat_template.jinja b/last-checkpoint/chat_template.jinja
new file mode 100644
index 0000000..699ff8d
--- /dev/null
+++ b/last-checkpoint/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set content = message.content %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in message.content %}
+                {%- set content = message.content.split('</think>')[-1].lstrip('\n') %}
+                {%- set reasoning_content = message.content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- message.content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/last-checkpoint/config.json b/last-checkpoint/config.json
new file mode 100644
index 0000000..4cab803
--- /dev/null
+++ b/last-checkpoint/config.json
@@ -0,0 +1,63 @@
+{
+  "architectures": [
+    "Qwen3ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": null,
+  "dtype": "bfloat16",
+  "eos_token_id": 151645,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 2048,
+  "initializer_range": 0.02,
+  "intermediate_size": 6144,
+  "layer_types": [
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention"
+  ],
+  "max_position_embeddings": 32768,
+  "max_window_layers": 28,
+  "model_type": "qwen3",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 8,
+  "pad_token_id": 151643,
+  "rms_norm_eps": 1e-06,
+  "rope_parameters": {
+    "rope_theta": 1000000,
+    "rope_type": "default"
+  },
+  "sliding_window": null,
+  "tie_word_embeddings": true,
+  "transformers_version": "5.10.2",
+  "use_cache": false,
+  "use_sliding_window": false,
+  "vocab_size": 151936
+}
diff --git a/last-checkpoint/generation_config.json b/last-checkpoint/generation_config.json
new file mode 100644
index 0000000..fbf2077
--- /dev/null
+++ b/last-checkpoint/generation_config.json
@@ -0,0 +1,10 @@
+{
+  "do_sample": false,
+  "eos_token_id": [
+    151645,
+    151643
+  ],
+  "max_new_tokens": 2048,
+  "pad_token_id": 151643,
+  "transformers_version": "5.10.2"
+}
diff --git a/last-checkpoint/model.safetensors b/last-checkpoint/model.safetensors
new file mode 100644
index 0000000..e11d4d8
--- /dev/null
+++ b/last-checkpoint/model.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:682a844772c7f371440abfa4d2b2af7a2cf85cf38e5a014643aaa64dbf9ab40d
+size 3441185608
diff --git a/last-checkpoint/optimizer.pt b/last-checkpoint/optimizer.pt
new file mode 100644
index 0000000..82d31c3
--- /dev/null
+++ b/last-checkpoint/optimizer.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8e0bae88c35d0176de22e1a568b3571e3a2049b1adb2bc84a6ccea1e747c92fb
+size 6882572207
diff --git a/last-checkpoint/rng_state.pth b/last-checkpoint/rng_state.pth
new file mode 100644
index 0000000..7cd3450
--- /dev/null
+++ b/last-checkpoint/rng_state.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b928d2d8033ac6bd87c58a39b741faace8bd1c6b0d070b7fad23c19520ff9f1a
+size 14645
diff --git a/last-checkpoint/scheduler.pt b/last-checkpoint/scheduler.pt
new file mode 100644
index 0000000..1571e25
--- /dev/null
+++ b/last-checkpoint/scheduler.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:17ad2e2f6666a5b110f084d4e012249bd44b831639fad091b11d78bc62da9d5b
+size 1465
diff --git a/last-checkpoint/tokenizer.json b/last-checkpoint/tokenizer.json
new file mode 100644
index 0000000..c7afbed
--- /dev/null
+++ b/last-checkpoint/tokenizer.json
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:be75606093db2094d7cd20f3c2f385c212750648bd6ea4fb2bf507a6a4c55506
+size 11422650
diff --git a/last-checkpoint/tokenizer_config.json b/last-checkpoint/tokenizer_config.json
new file mode 100644
index 0000000..770e41d
--- /dev/null
+++ b/last-checkpoint/tokenizer_config.json
@@ -0,0 +1,30 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "local_files_only": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/last-checkpoint/trainer_state.json b/last-checkpoint/trainer_state.json
new file mode 100644
index 0000000..ed4ca23
--- /dev/null
+++ b/last-checkpoint/trainer_state.json
@@ -0,0 +1,1488 @@
+{
+  "best_global_step": 800,
+  "best_metric": 0.34141287207603455,
+  "best_model_checkpoint": "/workspace/qwen3-math-rlvr/outputs/sft_v3a_checkpoint/checkpoint-800",
+  "epoch": 3.0,
+  "eval_steps": 100,
+  "global_step": 1308,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 0.8206211909651756,
+      "epoch": 0.022935779816513763,
+      "grad_norm": 5.90625,
+      "learning_rate": 1.3740458015267178e-06,
+      "loss": 0.8646629333496094,
+      "mean_token_accuracy": 0.8037635639309884,
+      "num_tokens": 102762.0,
+      "step": 10
+    },
+    {
+      "entropy": 0.8097452089190483,
+      "epoch": 0.045871559633027525,
+      "grad_norm": 6.6875,
+      "learning_rate": 2.900763358778626e-06,
+      "loss": 0.8560400009155273,
+      "mean_token_accuracy": 0.8035074040293694,
+      "num_tokens": 207338.0,
+      "step": 20
+    },
+    {
+      "entropy": 0.7938545405864715,
+      "epoch": 0.06880733944954129,
+      "grad_norm": 4.96875,
+      "learning_rate": 4.427480916030535e-06,
+      "loss": 0.8260902404785156,
+      "mean_token_accuracy": 0.808563794195652,
+      "num_tokens": 309467.0,
+      "step": 30
+    },
+    {
+      "entropy": 0.7353588879108429,
+      "epoch": 0.09174311926605505,
+      "grad_norm": 3.6875,
+      "learning_rate": 5.9541984732824435e-06,
+      "loss": 0.7422486305236816,
+      "mean_token_accuracy": 0.8224825605750083,
+      "num_tokens": 417166.0,
+      "step": 40
+    },
+    {
+      "entropy": 0.6841610729694366,
+      "epoch": 0.11467889908256881,
+      "grad_norm": 3.375,
+      "learning_rate": 7.480916030534352e-06,
+      "loss": 0.700370454788208,
+      "mean_token_accuracy": 0.8279992192983627,
+      "num_tokens": 516185.0,
+      "step": 50
+    },
+    {
+      "entropy": 0.5947812639176846,
+      "epoch": 0.13761467889908258,
+      "grad_norm": 2.359375,
+      "learning_rate": 9.007633587786259e-06,
+      "loss": 0.5929324626922607,
+      "mean_token_accuracy": 0.8471397697925568,
+      "num_tokens": 624710.0,
+      "step": 60
+    },
+    {
+      "entropy": 0.5556295536458492,
+      "epoch": 0.16055045871559634,
+      "grad_norm": 2.546875,
+      "learning_rate": 1.0534351145038168e-05,
+      "loss": 0.5473729133605957,
+      "mean_token_accuracy": 0.8533728003501893,
+      "num_tokens": 733444.0,
+      "step": 70
+    },
+    {
+      "entropy": 0.4859010189771652,
+      "epoch": 0.1834862385321101,
+      "grad_norm": 2.296875,
+      "learning_rate": 1.2061068702290078e-05,
+      "loss": 0.477904224395752,
+      "mean_token_accuracy": 0.8700864613056183,
+      "num_tokens": 839848.0,
+      "step": 80
+    },
+    {
+      "entropy": 0.4738590016961098,
+      "epoch": 0.20642201834862386,
+      "grad_norm": 2.109375,
+      "learning_rate": 1.3587786259541987e-05,
+      "loss": 0.46568708419799804,
+      "mean_token_accuracy": 0.8690521270036697,
+      "num_tokens": 946891.0,
+      "step": 90
+    },
+    {
+      "entropy": 0.46080359891057016,
+      "epoch": 0.22935779816513763,
+      "grad_norm": 2.1875,
+      "learning_rate": 1.5114503816793895e-05,
+      "loss": 0.43770489692687986,
+      "mean_token_accuracy": 0.8755019798874855,
+      "num_tokens": 1051235.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.22935779816513763,
+      "eval_entropy": 0.44482566804989526,
+      "eval_loss": 0.44241398572921753,
+      "eval_mean_token_accuracy": 0.874003436902295,
+      "eval_num_tokens": 1051235.0,
+      "eval_runtime": 30.3936,
+      "eval_samples_per_second": 12.075,
+      "eval_steps_per_second": 3.027,
+      "step": 100
+    },
+    {
+      "entropy": 0.4550655007362366,
+      "epoch": 0.25229357798165136,
+      "grad_norm": 2.078125,
+      "learning_rate": 1.6641221374045802e-05,
+      "loss": 0.4378714084625244,
+      "mean_token_accuracy": 0.8735761269927025,
+      "num_tokens": 1154595.0,
+      "step": 110
+    },
+    {
+      "entropy": 0.47160537242889405,
+      "epoch": 0.27522935779816515,
+      "grad_norm": 2.15625,
+      "learning_rate": 1.816793893129771e-05,
+      "loss": 0.43929414749145507,
+      "mean_token_accuracy": 0.874704971909523,
+      "num_tokens": 1255238.0,
+      "step": 120
+    },
+    {
+      "entropy": 0.45139552727341653,
+      "epoch": 0.2981651376146789,
+      "grad_norm": 1.9765625,
+      "learning_rate": 1.969465648854962e-05,
+      "loss": 0.4124535083770752,
+      "mean_token_accuracy": 0.8811365008354187,
+      "num_tokens": 1364980.0,
+      "step": 130
+    },
+    {
+      "entropy": 0.44258273467421533,
+      "epoch": 0.3211009174311927,
+      "grad_norm": 2.375,
+      "learning_rate": 1.9997720286367865e-05,
+      "loss": 0.41561241149902345,
+      "mean_token_accuracy": 0.8787958458065986,
+      "num_tokens": 1462504.0,
+      "step": 140
+    },
+    {
+      "entropy": 0.4124230518937111,
+      "epoch": 0.3440366972477064,
+      "grad_norm": 1.984375,
+      "learning_rate": 1.9988460731134392e-05,
+      "loss": 0.3839062213897705,
+      "mean_token_accuracy": 0.8872552439570427,
+      "num_tokens": 1567430.0,
+      "step": 150
+    },
+    {
+      "entropy": 0.4063340038061142,
+      "epoch": 0.3669724770642202,
+      "grad_norm": 1.921875,
+      "learning_rate": 1.9972085443585466e-05,
+      "loss": 0.3948183059692383,
+      "mean_token_accuracy": 0.8833665549755096,
+      "num_tokens": 1673064.0,
+      "step": 160
+    },
+    {
+      "entropy": 0.4096576377749443,
+      "epoch": 0.38990825688073394,
+      "grad_norm": 2.1875,
+      "learning_rate": 1.99486060893987e-05,
+      "loss": 0.38925724029541015,
+      "mean_token_accuracy": 0.8824431821703911,
+      "num_tokens": 1773169.0,
+      "step": 170
+    },
+    {
+      "entropy": 0.386829512566328,
+      "epoch": 0.41284403669724773,
+      "grad_norm": 2.609375,
+      "learning_rate": 1.9918039395155278e-05,
+      "loss": 0.37414252758026123,
+      "mean_token_accuracy": 0.8875698506832123,
+      "num_tokens": 1871532.0,
+      "step": 180
+    },
+    {
+      "entropy": 0.39463941305875777,
+      "epoch": 0.43577981651376146,
+      "grad_norm": 1.9609375,
+      "learning_rate": 1.988040713642402e-05,
+      "loss": 0.37338783740997317,
+      "mean_token_accuracy": 0.8848363608121872,
+      "num_tokens": 1977596.0,
+      "step": 190
+    },
+    {
+      "entropy": 0.3883584551513195,
+      "epoch": 0.45871559633027525,
+      "grad_norm": 2.109375,
+      "learning_rate": 1.9835736122248558e-05,
+      "loss": 0.3809430837631226,
+      "mean_token_accuracy": 0.8834147050976753,
+      "num_tokens": 2079498.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.45871559633027525,
+      "eval_entropy": 0.37745926464381424,
+      "eval_loss": 0.3726924657821655,
+      "eval_mean_token_accuracy": 0.8875944420047428,
+      "eval_num_tokens": 2079498.0,
+      "eval_runtime": 30.5955,
+      "eval_samples_per_second": 11.995,
+      "eval_steps_per_second": 3.007,
+      "step": 200
+    },
+    {
+      "entropy": 0.39538886621594427,
+      "epoch": 0.481651376146789,
+      "grad_norm": 2.21875,
+      "learning_rate": 1.97840581760487e-05,
+      "loss": 0.383013653755188,
+      "mean_token_accuracy": 0.8827495723962784,
+      "num_tokens": 2188160.0,
+      "step": 210
+    },
+    {
+      "entropy": 0.3952382251620293,
+      "epoch": 0.5045871559633027,
+      "grad_norm": 2.125,
+      "learning_rate": 1.972541011294959e-05,
+      "loss": 0.37931249141693113,
+      "mean_token_accuracy": 0.8842660069465638,
+      "num_tokens": 2294740.0,
+      "step": 220
+    },
+    {
+      "entropy": 0.39509702771902083,
+      "epoch": 0.5275229357798165,
+      "grad_norm": 2.109375,
+      "learning_rate": 1.9659833713554773e-05,
+      "loss": 0.3826189756393433,
+      "mean_token_accuracy": 0.883507713675499,
+      "num_tokens": 2406600.0,
+      "step": 230
+    },
+    {
+      "entropy": 0.3746883951127529,
+      "epoch": 0.5504587155963303,
+      "grad_norm": 2.140625,
+      "learning_rate": 1.958737569418189e-05,
+      "loss": 0.36290385723114016,
+      "mean_token_accuracy": 0.8878081619739533,
+      "num_tokens": 2511933.0,
+      "step": 240
+    },
+    {
+      "entropy": 0.3867727659642696,
+      "epoch": 0.573394495412844,
+      "grad_norm": 1.7421875,
+      "learning_rate": 1.9508087673582204e-05,
+      "loss": 0.37698163986206057,
+      "mean_token_accuracy": 0.8852298274636269,
+      "num_tokens": 2619847.0,
+      "step": 250
+    },
+    {
+      "entropy": 0.37598335891962054,
+      "epoch": 0.5963302752293578,
+      "grad_norm": 1.96875,
+      "learning_rate": 1.9422026136167644e-05,
+      "loss": 0.3573493242263794,
+      "mean_token_accuracy": 0.8899750009179115,
+      "num_tokens": 2723132.0,
+      "step": 260
+    },
+    {
+      "entropy": 0.34912080466747286,
+      "epoch": 0.6192660550458715,
+      "grad_norm": 1.78125,
+      "learning_rate": 1.9329252391771556e-05,
+      "loss": 0.34246680736541746,
+      "mean_token_accuracy": 0.8929290145635604,
+      "num_tokens": 2832888.0,
+      "step": 270
+    },
+    {
+      "entropy": 0.3983289562165737,
+      "epoch": 0.6422018348623854,
+      "grad_norm": 2.265625,
+      "learning_rate": 1.9229832531971895e-05,
+      "loss": 0.3874169588088989,
+      "mean_token_accuracy": 0.8818993121385574,
+      "num_tokens": 2934344.0,
+      "step": 280
+    },
+    {
+      "entropy": 0.3607022054493427,
+      "epoch": 0.6651376146788991,
+      "grad_norm": 2.5,
+      "learning_rate": 1.9123837383007885e-05,
+      "loss": 0.3489506244659424,
+      "mean_token_accuracy": 0.8915312319993973,
+      "num_tokens": 3040953.0,
+      "step": 290
+    },
+    {
+      "entropy": 0.35854368582367896,
+      "epoch": 0.6880733944954128,
+      "grad_norm": 2.171875,
+      "learning_rate": 1.9011342455323723e-05,
+      "loss": 0.3501297950744629,
+      "mean_token_accuracy": 0.8913966670632363,
+      "num_tokens": 3151335.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.6880733944954128,
+      "eval_entropy": 0.3628946772736052,
+      "eval_loss": 0.35798636078834534,
+      "eval_mean_token_accuracy": 0.8898697566726933,
+      "eval_num_tokens": 3151335.0,
+      "eval_runtime": 30.6951,
+      "eval_samples_per_second": 11.956,
+      "eval_steps_per_second": 2.997,
+      "step": 300
+    },
+    {
+      "entropy": 0.3661672294139862,
+      "epoch": 0.7110091743119266,
+      "grad_norm": 1.7890625,
+      "learning_rate": 1.8892427889775334e-05,
+      "loss": 0.35704004764556885,
+      "mean_token_accuracy": 0.8901202246546746,
+      "num_tokens": 3259905.0,
+      "step": 310
+    },
+    {
+      "entropy": 0.34835164099931715,
+      "epoch": 0.7339449541284404,
+      "grad_norm": 1.9296875,
+      "learning_rate": 1.8767178400538362e-05,
+      "loss": 0.3407481908798218,
+      "mean_token_accuracy": 0.8939173936843872,
+      "num_tokens": 3371657.0,
+      "step": 320
+    },
+    {
+      "entropy": 0.35573601722717285,
+      "epoch": 0.7568807339449541,
+      "grad_norm": 1.859375,
+      "learning_rate": 1.8635683214758213e-05,
+      "loss": 0.34518754482269287,
+      "mean_token_accuracy": 0.8920813903212548,
+      "num_tokens": 3477691.0,
+      "step": 330
+    },
+    {
+      "entropy": 0.3739293247461319,
+      "epoch": 0.7798165137614679,
+      "grad_norm": 1.7109375,
+      "learning_rate": 1.8498036008985058e-05,
+      "loss": 0.3678521394729614,
+      "mean_token_accuracy": 0.8885598212480545,
+      "num_tokens": 3583628.0,
+      "step": 340
+    },
+    {
+      "entropy": 0.35215074978768823,
+      "epoch": 0.8027522935779816,
+      "grad_norm": 1.9140625,
+      "learning_rate": 1.8354334842439096e-05,
+      "loss": 0.34316134452819824,
+      "mean_token_accuracy": 0.8933907493948936,
+      "num_tokens": 3689504.0,
+      "step": 350
+    },
+    {
+      "entropy": 0.3545384183526039,
+      "epoch": 0.8256880733944955,
+      "grad_norm": 2.015625,
+      "learning_rate": 1.820468208715362e-05,
+      "loss": 0.34501280784606936,
+      "mean_token_accuracy": 0.8933851882815361,
+      "num_tokens": 3794695.0,
+      "step": 360
+    },
+    {
+      "entropy": 0.3516731861978769,
+      "epoch": 0.8486238532110092,
+      "grad_norm": 2.078125,
+      "learning_rate": 1.8049184355045677e-05,
+      "loss": 0.34716789722442626,
+      "mean_token_accuracy": 0.8922320544719696,
+      "num_tokens": 3899779.0,
+      "step": 370
+    },
+    {
+      "entropy": 0.348944878578186,
+      "epoch": 0.8715596330275229,
+      "grad_norm": 1.8203125,
+      "learning_rate": 1.788795242196623e-05,
+      "loss": 0.3417558431625366,
+      "mean_token_accuracy": 0.8929858908057213,
+      "num_tokens": 4011147.0,
+      "step": 380
+    },
+    {
+      "entropy": 0.37281630262732507,
+      "epoch": 0.8944954128440367,
+      "grad_norm": 2.0,
+      "learning_rate": 1.7721101148783985e-05,
+      "loss": 0.36486363410949707,
+      "mean_token_accuracy": 0.8859157666563988,
+      "num_tokens": 4111015.0,
+      "step": 390
+    },
+    {
+      "entropy": 0.35566350184381007,
+      "epoch": 0.9174311926605505,
+      "grad_norm": 1.9375,
+      "learning_rate": 1.754874939955901e-05,
+      "loss": 0.3445189714431763,
+      "mean_token_accuracy": 0.8937858313322067,
+      "num_tokens": 4220197.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9174311926605505,
+      "eval_entropy": 0.34688637859147525,
+      "eval_loss": 0.3506438434123993,
+      "eval_mean_token_accuracy": 0.8918481352536575,
+      "eval_num_tokens": 4220197.0,
+      "eval_runtime": 30.4467,
+      "eval_samples_per_second": 12.054,
+      "eval_steps_per_second": 3.022,
+      "step": 400
+    },
+    {
+      "entropy": 0.3542441871017218,
+      "epoch": 0.9403669724770642,
+      "grad_norm": 2.046875,
+      "learning_rate": 1.7371019956864584e-05,
+      "loss": 0.35869665145874025,
+      "mean_token_accuracy": 0.889110067486763,
+      "num_tokens": 4328050.0,
+      "step": 410
+    },
+    {
+      "entropy": 0.34873349256813524,
+      "epoch": 0.963302752293578,
+      "grad_norm": 1.9921875,
+      "learning_rate": 1.7188039434317437e-05,
+      "loss": 0.3368797779083252,
+      "mean_token_accuracy": 0.8951233714818955,
+      "num_tokens": 4432381.0,
+      "step": 420
+    },
+    {
+      "entropy": 0.33490167558193207,
+      "epoch": 0.9862385321100917,
+      "grad_norm": 2.015625,
+      "learning_rate": 1.6999938186378826e-05,
+      "loss": 0.3344027042388916,
+      "mean_token_accuracy": 0.8973124548792839,
+      "num_tokens": 4540406.0,
+      "step": 430
+    },
+    {
+      "entropy": 0.34897951520979403,
+      "epoch": 1.0091743119266054,
+      "grad_norm": 1.8515625,
+      "learning_rate": 1.680685021549063e-05,
+      "loss": 0.33591907024383544,
+      "mean_token_accuracy": 0.8952021867036819,
+      "num_tokens": 4645377.0,
+      "step": 440
+    },
+    {
+      "entropy": 0.32901966348290446,
+      "epoch": 1.0321100917431192,
+      "grad_norm": 2.171875,
+      "learning_rate": 1.6608913076612628e-05,
+      "loss": 0.3125960111618042,
+      "mean_token_accuracy": 0.9004640519618988,
+      "num_tokens": 4754008.0,
+      "step": 450
+    },
+    {
+      "entropy": 0.3282691590487957,
+      "epoch": 1.0550458715596331,
+      "grad_norm": 1.8984375,
+      "learning_rate": 1.640626777922901e-05,
+      "loss": 0.3234114170074463,
+      "mean_token_accuracy": 0.895183089375496,
+      "num_tokens": 4858923.0,
+      "step": 460
+    },
+    {
+      "entropy": 0.31950610168278215,
+      "epoch": 1.0779816513761469,
+      "grad_norm": 1.65625,
+      "learning_rate": 1.619905868689388e-05,
+      "loss": 0.31262228488922117,
+      "mean_token_accuracy": 0.9011708259582519,
+      "num_tokens": 4970628.0,
+      "step": 470
+    },
+    {
+      "entropy": 0.3411460068076849,
+      "epoch": 1.1009174311926606,
+      "grad_norm": 2.09375,
+      "learning_rate": 1.598743341438734e-05,
+      "loss": 0.33199501037597656,
+      "mean_token_accuracy": 0.8963311001658439,
+      "num_tokens": 5071949.0,
+      "step": 480
+    },
+    {
+      "entropy": 0.3385404385626316,
+      "epoch": 1.1238532110091743,
+      "grad_norm": 1.9921875,
+      "learning_rate": 1.5771542722555423e-05,
+      "loss": 0.33367342948913575,
+      "mean_token_accuracy": 0.8964887470006943,
+      "num_tokens": 5176802.0,
+      "step": 490
+    },
+    {
+      "entropy": 0.34455578960478306,
+      "epoch": 1.146788990825688,
+      "grad_norm": 2.140625,
+      "learning_rate": 1.555154041090876e-05,
+      "loss": 0.332703685760498,
+      "mean_token_accuracy": 0.8943984031677246,
+      "num_tokens": 5278787.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.146788990825688,
+      "eval_entropy": 0.3309480096011058,
+      "eval_loss": 0.34800779819488525,
+      "eval_mean_token_accuracy": 0.8921307053254999,
+      "eval_num_tokens": 5278787.0,
+      "eval_runtime": 30.5968,
+      "eval_samples_per_second": 11.995,
+      "eval_steps_per_second": 3.007,
+      "step": 500
+    },
+    {
+      "entropy": 0.3380110088735819,
+      "epoch": 1.1697247706422018,
+      "grad_norm": 1.984375,
+      "learning_rate": 1.5327583208056553e-05,
+      "loss": 0.3292912483215332,
+      "mean_token_accuracy": 0.896933126449585,
+      "num_tokens": 5386283.0,
+      "step": 510
+    },
+    {
+      "entropy": 0.32564901672303675,
+      "epoch": 1.1926605504587156,
+      "grad_norm": 2.265625,
+      "learning_rate": 1.5099830660053849e-05,
+      "loss": 0.3155273199081421,
+      "mean_token_accuracy": 0.9006841659545899,
+      "num_tokens": 5492232.0,
+      "step": 520
+    },
+    {
+      "entropy": 0.3386135693639517,
+      "epoch": 1.2155963302752293,
+      "grad_norm": 2.453125,
+      "learning_rate": 1.4868445016741677e-05,
+      "loss": 0.3391716003417969,
+      "mean_token_accuracy": 0.8942732483148574,
+      "num_tokens": 5601457.0,
+      "step": 530
+    },
+    {
+      "entropy": 0.32902440130710603,
+      "epoch": 1.238532110091743,
+      "grad_norm": 2.25,
+      "learning_rate": 1.463359111616103e-05,
+      "loss": 0.3173969268798828,
+      "mean_token_accuracy": 0.8995897427201272,
+      "num_tokens": 5709684.0,
+      "step": 540
+    },
+    {
+      "entropy": 0.3364047706127167,
+      "epoch": 1.261467889908257,
+      "grad_norm": 1.6875,
+      "learning_rate": 1.4395436267123017e-05,
+      "loss": 0.3304882049560547,
+      "mean_token_accuracy": 0.8970319986343384,
+      "num_tokens": 5819673.0,
+      "step": 550
+    },
+    {
+      "entropy": 0.3300769329071045,
+      "epoch": 1.2844036697247707,
+      "grad_norm": 2.078125,
+      "learning_rate": 1.4154150130018867e-05,
+      "loss": 0.3214289665222168,
+      "mean_token_accuracy": 0.8994089663028717,
+      "num_tokens": 5922492.0,
+      "step": 560
+    },
+    {
+      "entropy": 0.32826707139611244,
+      "epoch": 1.3073394495412844,
+      "grad_norm": 2.3125,
+      "learning_rate": 1.3909904595954651e-05,
+      "loss": 0.32586624622344973,
+      "mean_token_accuracy": 0.8974115282297135,
+      "num_tokens": 6019243.0,
+      "step": 570
+    },
+    {
+      "entropy": 0.34469916522502897,
+      "epoch": 1.3302752293577982,
+      "grad_norm": 1.8515625,
+      "learning_rate": 1.3662873664296885e-05,
+      "loss": 0.33226957321166994,
+      "mean_token_accuracy": 0.8959178075194358,
+      "num_tokens": 6123675.0,
+      "step": 580
+    },
+    {
+      "entropy": 0.31974404454231264,
+      "epoch": 1.353211009174312,
+      "grad_norm": 2.015625,
+      "learning_rate": 1.3413233318716188e-05,
+      "loss": 0.3171189785003662,
+      "mean_token_accuracy": 0.9018555253744125,
+      "num_tokens": 6232954.0,
+      "step": 590
+    },
+    {
+      "entropy": 0.3314074445515871,
+      "epoch": 1.3761467889908257,
+      "grad_norm": 2.09375,
+      "learning_rate": 1.3161161401817353e-05,
+      "loss": 0.3289859056472778,
+      "mean_token_accuracy": 0.8955597743391991,
+      "num_tokens": 6337476.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.3761467889908257,
+      "eval_entropy": 0.326329352091188,
+      "eval_loss": 0.3443230092525482,
+      "eval_mean_token_accuracy": 0.8931592741738195,
+      "eval_num_tokens": 6337476.0,
+      "eval_runtime": 30.3343,
+      "eval_samples_per_second": 12.099,
+      "eval_steps_per_second": 3.033,
+      "step": 600
+    },
+    {
+      "entropy": 0.3287132531404495,
+      "epoch": 1.3990825688073394,
+      "grad_norm": 2.15625,
+      "learning_rate": 1.2906837488445114e-05,
+      "loss": 0.32199873924255373,
+      "mean_token_accuracy": 0.8973362505435943,
+      "num_tokens": 6438212.0,
+      "step": 610
+    },
+    {
+      "entropy": 0.321663286164403,
+      "epoch": 1.4220183486238533,
+      "grad_norm": 1.8203125,
+      "learning_rate": 1.265044275775586e-05,
+      "loss": 0.31311841011047364,
+      "mean_token_accuracy": 0.8999396815896035,
+      "num_tokens": 6537652.0,
+      "step": 620
+    },
+    {
+      "entropy": 0.33257763721048833,
+      "epoch": 1.4449541284403669,
+      "grad_norm": 2.53125,
+      "learning_rate": 1.2392159864146483e-05,
+      "loss": 0.32827396392822267,
+      "mean_token_accuracy": 0.8963763967156411,
+      "num_tokens": 6644479.0,
+      "step": 630
+    },
+    {
+      "entropy": 0.3169947315007448,
+      "epoch": 1.4678899082568808,
+      "grad_norm": 2.328125,
+      "learning_rate": 1.2132172807132233e-05,
+      "loss": 0.31139681339263914,
+      "mean_token_accuracy": 0.9024187490344048,
+      "num_tokens": 6754776.0,
+      "step": 640
+    },
+    {
+      "entropy": 0.3225189931690693,
+      "epoch": 1.4908256880733946,
+      "grad_norm": 2.0,
+      "learning_rate": 1.1870666800266322e-05,
+      "loss": 0.3168517589569092,
+      "mean_token_accuracy": 0.898341266810894,
+      "num_tokens": 6860996.0,
+      "step": 650
+    },
+    {
+      "entropy": 0.32711520195007326,
+      "epoch": 1.5137614678899083,
+      "grad_norm": 1.828125,
+      "learning_rate": 1.1607828139194683e-05,
+      "loss": 0.319119930267334,
+      "mean_token_accuracy": 0.9000087305903435,
+      "num_tokens": 6965254.0,
+      "step": 660
+    },
+    {
+      "entropy": 0.32906078547239304,
+      "epoch": 1.536697247706422,
+      "grad_norm": 2.015625,
+      "learning_rate": 1.1343844068939815e-05,
+      "loss": 0.32240869998931887,
+      "mean_token_accuracy": 0.898229330778122,
+      "num_tokens": 7072487.0,
+      "step": 670
+    },
+    {
+      "entropy": 0.3303819328546524,
+      "epoch": 1.5596330275229358,
+      "grad_norm": 2.015625,
+      "learning_rate": 1.1078902650508319e-05,
+      "loss": 0.3209313631057739,
+      "mean_token_accuracy": 0.8985160902142525,
+      "num_tokens": 7176921.0,
+      "step": 680
+    },
+    {
+      "entropy": 0.32761351503431796,
+      "epoch": 1.5825688073394495,
+      "grad_norm": 2.046875,
+      "learning_rate": 1.0813192626917116e-05,
+      "loss": 0.3219845533370972,
+      "mean_token_accuracy": 0.8978465691208839,
+      "num_tokens": 7283491.0,
+      "step": 690
+    },
+    {
+      "entropy": 0.3228471975773573,
+      "epoch": 1.6055045871559632,
+      "grad_norm": 2.109375,
+      "learning_rate": 1.0546903288733827e-05,
+      "loss": 0.31817333698272704,
+      "mean_token_accuracy": 0.8989653572440147,
+      "num_tokens": 7391144.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.6055045871559632,
+      "eval_entropy": 0.32780669856330624,
+      "eval_loss": 0.34249600768089294,
+      "eval_mean_token_accuracy": 0.8938729912042618,
+      "eval_num_tokens": 7391144.0,
+      "eval_runtime": 30.7947,
+      "eval_samples_per_second": 11.918,
+      "eval_steps_per_second": 2.988,
+      "step": 700
+    },
+    {
+      "entropy": 0.31083628684282305,
+      "epoch": 1.6284403669724772,
+      "grad_norm": 1.890625,
+      "learning_rate": 1.0280224339227047e-05,
+      "loss": 0.30303895473480225,
+      "mean_token_accuracy": 0.9043862655758858,
+      "num_tokens": 7499840.0,
+      "step": 710
+    },
+    {
+      "entropy": 0.3250539544969797,
+      "epoch": 1.6513761467889907,
+      "grad_norm": 1.6953125,
+      "learning_rate": 1.001334575922265e-05,
+      "loss": 0.31643564701080323,
+      "mean_token_accuracy": 0.8997841477394104,
+      "num_tokens": 7607474.0,
+      "step": 720
+    },
+    {
+      "entropy": 0.3169710818678141,
+      "epoch": 1.6743119266055047,
+      "grad_norm": 1.90625,
+      "learning_rate": 9.746457671762325e-06,
+      "loss": 0.3112821340560913,
+      "mean_token_accuracy": 0.9023401409387588,
+      "num_tokens": 7710747.0,
+      "step": 730
+    },
+    {
+      "entropy": 0.31221029385924337,
+      "epoch": 1.6972477064220184,
+      "grad_norm": 2.015625,
+      "learning_rate": 9.479750206660832e-06,
+      "loss": 0.3069144725799561,
+      "mean_token_accuracy": 0.9038597375154496,
+      "num_tokens": 7819850.0,
+      "step": 740
+    },
+    {
+      "entropy": 0.32356046326458454,
+      "epoch": 1.7201834862385321,
+      "grad_norm": 2.203125,
+      "learning_rate": 9.213413365058418e-06,
+      "loss": 0.3189092636108398,
+      "mean_token_accuracy": 0.899696996808052,
+      "num_tokens": 7928323.0,
+      "step": 750
+    },
+    {
+      "entropy": 0.32968592271208763,
+      "epoch": 1.7431192660550459,
+      "grad_norm": 1.984375,
+      "learning_rate": 8.947636884064895e-06,
+      "loss": 0.3251533269882202,
+      "mean_token_accuracy": 0.8982710152864456,
+      "num_tokens": 8032960.0,
+      "step": 760
+    },
+    {
+      "entropy": 0.32864366583526133,
+      "epoch": 1.7660550458715596,
+      "grad_norm": 1.8359375,
+      "learning_rate": 8.682610101591813e-06,
+      "loss": 0.32470901012420655,
+      "mean_token_accuracy": 0.8994028642773628,
+      "num_tokens": 8139955.0,
+      "step": 770
+    },
+    {
+      "entropy": 0.31197240948677063,
+      "epoch": 1.7889908256880735,
+      "grad_norm": 1.921875,
+      "learning_rate": 8.418521821469044e-06,
+      "loss": 0.30414133071899413,
+      "mean_token_accuracy": 0.9039060071110725,
+      "num_tokens": 8246897.0,
+      "step": 780
+    },
+    {
+      "entropy": 0.31388189047575,
+      "epoch": 1.811926605504587,
+      "grad_norm": 1.9296875,
+      "learning_rate": 8.155560178941805e-06,
+      "loss": 0.30535738468170165,
+      "mean_token_accuracy": 0.9023241758346557,
+      "num_tokens": 8355182.0,
+      "step": 790
+    },
+    {
+      "entropy": 0.3281553216278553,
+      "epoch": 1.834862385321101,
+      "grad_norm": 1.8203125,
+      "learning_rate": 7.893912506644003e-06,
+      "loss": 0.3181729793548584,
+      "mean_token_accuracy": 0.8985237941145897,
+      "num_tokens": 8461712.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.834862385321101,
+      "eval_entropy": 0.32175023092523863,
+      "eval_loss": 0.34141287207603455,
+      "eval_mean_token_accuracy": 0.8936791387589081,
+      "eval_num_tokens": 8461712.0,
+      "eval_runtime": 30.5224,
+      "eval_samples_per_second": 12.024,
+      "eval_steps_per_second": 3.014,
+      "step": 800
+    },
+    {
+      "entropy": 0.31686329916119577,
+      "epoch": 1.8577981651376145,
+      "grad_norm": 1.9375,
+      "learning_rate": 7.63376520114335e-06,
+      "loss": 0.3101612091064453,
+      "mean_token_accuracy": 0.9004829794168472,
+      "num_tokens": 8564359.0,
+      "step": 810
+    },
+    {
+      "entropy": 0.3207144971936941,
+      "epoch": 1.8807339449541285,
+      "grad_norm": 1.8671875,
+      "learning_rate": 7.375303590153335e-06,
+      "loss": 0.3119115591049194,
+      "mean_token_accuracy": 0.9005504950881005,
+      "num_tokens": 8667964.0,
+      "step": 820
+    },
+    {
+      "entropy": 0.32959680669009683,
+      "epoch": 1.9036697247706422,
+      "grad_norm": 1.953125,
+      "learning_rate": 7.118711800506625e-06,
+      "loss": 0.327376127243042,
+      "mean_token_accuracy": 0.8956960454583168,
+      "num_tokens": 8766422.0,
+      "step": 830
+    },
+    {
+      "entropy": 0.32939605973660946,
+      "epoch": 1.926605504587156,
+      "grad_norm": 2.28125,
+      "learning_rate": 6.8641726269839735e-06,
+      "loss": 0.3243770360946655,
+      "mean_token_accuracy": 0.8972024530172348,
+      "num_tokens": 8870619.0,
+      "step": 840
+    },
+    {
+      "entropy": 0.3171665370464325,
+      "epoch": 1.9495412844036697,
+      "grad_norm": 1.890625,
+      "learning_rate": 6.611867402092096e-06,
+      "loss": 0.314391565322876,
+      "mean_token_accuracy": 0.9010873824357987,
+      "num_tokens": 8976446.0,
+      "step": 850
+    },
+    {
+      "entropy": 0.3226952590048313,
+      "epoch": 1.9724770642201834,
+      "grad_norm": 2.03125,
+      "learning_rate": 6.36197586688324e-06,
+      "loss": 0.31484920978546144,
+      "mean_token_accuracy": 0.8996226698160171,
+      "num_tokens": 9079844.0,
+      "step": 860
+    },
+    {
+      "entropy": 0.3094219394028187,
+      "epoch": 1.9954128440366974,
+      "grad_norm": 2.015625,
+      "learning_rate": 6.114676042908501e-06,
+      "loss": 0.30544233322143555,
+      "mean_token_accuracy": 0.9039633944630623,
+      "num_tokens": 9185589.0,
+      "step": 870
+    },
+    {
+      "entropy": 0.3221705980598927,
+      "epoch": 2.018348623853211,
+      "grad_norm": 1.7265625,
+      "learning_rate": 5.8701441053961185e-06,
+      "loss": 0.30669331550598145,
+      "mean_token_accuracy": 0.9029042199254036,
+      "num_tokens": 9287606.0,
+      "step": 880
+    },
+    {
+      "entropy": 0.31949374340474607,
+      "epoch": 2.041284403669725,
+      "grad_norm": 1.7421875,
+      "learning_rate": 5.628554257745069e-06,
+      "loss": 0.30752108097076414,
+      "mean_token_accuracy": 0.9037730067968368,
+      "num_tokens": 9393753.0,
+      "step": 890
+    },
+    {
+      "entropy": 0.3091618109494448,
+      "epoch": 2.0642201834862384,
+      "grad_norm": 2.0,
+      "learning_rate": 5.390078607423392e-06,
+      "loss": 0.3008283615112305,
+      "mean_token_accuracy": 0.9035891696810723,
+      "num_tokens": 9502417.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.0642201834862384,
+      "eval_entropy": 0.3158419154908346,
+      "eval_loss": 0.34156936407089233,
+      "eval_mean_token_accuracy": 0.8940030284549879,
+      "eval_num_tokens": 9502417.0,
+      "eval_runtime": 30.4897,
+      "eval_samples_per_second": 12.037,
+      "eval_steps_per_second": 3.017,
+      "step": 900
+    },
+    {
+      "entropy": 0.3212772488594055,
+      "epoch": 2.0871559633027523,
+      "grad_norm": 2.4375,
+      "learning_rate": 5.154887043359633e-06,
+      "loss": 0.31162261962890625,
+      "mean_token_accuracy": 0.9024444729089737,
+      "num_tokens": 9605737.0,
+      "step": 910
+    },
+    {
+      "entropy": 0.30515239015221596,
+      "epoch": 2.1100917431192663,
+      "grad_norm": 1.9453125,
+      "learning_rate": 4.923147114914777e-06,
+      "loss": 0.3026602029800415,
+      "mean_token_accuracy": 0.9041713520884513,
+      "num_tokens": 9710276.0,
+      "step": 920
+    },
+    {
+      "entropy": 0.3088552877306938,
+      "epoch": 2.13302752293578,
+      "grad_norm": 2.140625,
+      "learning_rate": 4.695023912520874e-06,
+      "loss": 0.30738000869750975,
+      "mean_token_accuracy": 0.9026739463210106,
+      "num_tokens": 9820419.0,
+      "step": 930
+    },
+    {
+      "entropy": 0.3046286415308714,
+      "epoch": 2.1559633027522938,
+      "grad_norm": 1.671875,
+      "learning_rate": 4.470679950071394e-06,
+      "loss": 0.295642352104187,
+      "mean_token_accuracy": 0.9056263417005539,
+      "num_tokens": 9929776.0,
+      "step": 940
+    },
+    {
+      "entropy": 0.3212243374437094,
+      "epoch": 2.1788990825688073,
+      "grad_norm": 2.296875,
+      "learning_rate": 4.2502750491470926e-06,
+      "loss": 0.3080165863037109,
+      "mean_token_accuracy": 0.9025693088769913,
+      "num_tokens": 10031729.0,
+      "step": 950
+    },
+    {
+      "entropy": 0.3244902342557907,
+      "epoch": 2.2018348623853212,
+      "grad_norm": 2.375,
+      "learning_rate": 4.033966225159875e-06,
+      "loss": 0.3196781873703003,
+      "mean_token_accuracy": 0.8980901971459389,
+      "num_tokens": 10128604.0,
+      "step": 960
+    },
+    {
+      "entropy": 0.3001674834638834,
+      "epoch": 2.2247706422018347,
+      "grad_norm": 1.9296875,
+      "learning_rate": 3.821907575495755e-06,
+      "loss": 0.29091734886169435,
+      "mean_token_accuracy": 0.9066133707761764,
+      "num_tokens": 10235735.0,
+      "step": 970
+    },
+    {
+      "entropy": 0.3224890284240246,
+      "epoch": 2.2477064220183487,
+      "grad_norm": 1.9921875,
+      "learning_rate": 3.614250169736603e-06,
+      "loss": 0.31416850090026854,
+      "mean_token_accuracy": 0.8999267861247062,
+      "num_tokens": 10339995.0,
+      "step": 980
+    },
+    {
+      "entropy": 0.31570393070578573,
+      "epoch": 2.270642201834862,
+      "grad_norm": 1.8671875,
+      "learning_rate": 3.4111419420388904e-06,
+      "loss": 0.3096719980239868,
+      "mean_token_accuracy": 0.9028396472334862,
+      "num_tokens": 10446651.0,
+      "step": 990
+    },
+    {
+      "entropy": 0.3116422988474369,
+      "epoch": 2.293577981651376,
+      "grad_norm": 2.0625,
+      "learning_rate": 3.2127275857460914e-06,
+      "loss": 0.3034141778945923,
+      "mean_token_accuracy": 0.9035742565989494,
+      "num_tokens": 10551686.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.293577981651376,
+      "eval_entropy": 0.31458691961091495,
+      "eval_loss": 0.3416135013103485,
+      "eval_mean_token_accuracy": 0.8941712055517279,
+      "eval_num_tokens": 10551686.0,
+      "eval_runtime": 30.9012,
+      "eval_samples_per_second": 11.877,
+      "eval_steps_per_second": 2.977,
+      "step": 1000
+    },
+    {
+      "entropy": 0.30370271801948545,
+      "epoch": 2.31651376146789,
+      "grad_norm": 1.75,
+      "learning_rate": 3.0191484503098335e-06,
+      "loss": 0.2965387344360352,
+      "mean_token_accuracy": 0.9055161878466607,
+      "num_tokens": 10658763.0,
+      "step": 1010
+    },
+    {
+      "entropy": 0.30786469504237174,
+      "epoch": 2.3394495412844036,
+      "grad_norm": 1.984375,
+      "learning_rate": 2.830542440593206e-06,
+      "loss": 0.2988230466842651,
+      "mean_token_accuracy": 0.9048854574561119,
+      "num_tokens": 10758819.0,
+      "step": 1020
+    },
+    {
+      "entropy": 0.30542101748287676,
+      "epoch": 2.3623853211009176,
+      "grad_norm": 2.03125,
+      "learning_rate": 2.647043918627994e-06,
+      "loss": 0.29598278999328614,
+      "mean_token_accuracy": 0.9052035048604011,
+      "num_tokens": 10865788.0,
+      "step": 1030
+    },
+    {
+      "entropy": 0.31079764701426027,
+      "epoch": 2.385321100917431,
+      "grad_norm": 2.1875,
+      "learning_rate": 2.468783607895796e-06,
+      "loss": 0.30441627502441404,
+      "mean_token_accuracy": 0.9034854844212532,
+      "num_tokens": 10967177.0,
+      "step": 1040
+    },
+    {
+      "entropy": 0.31058608889579775,
+      "epoch": 2.408256880733945,
+      "grad_norm": 2.0,
+      "learning_rate": 2.295888500201233e-06,
+      "loss": 0.3008920669555664,
+      "mean_token_accuracy": 0.9041848033666611,
+      "num_tokens": 11071623.0,
+      "step": 1050
+    },
+    {
+      "entropy": 0.32119043692946436,
+      "epoch": 2.4311926605504586,
+      "grad_norm": 2.171875,
+      "learning_rate": 2.1284817652035893e-06,
+      "loss": 0.3111232280731201,
+      "mean_token_accuracy": 0.8993611231446266,
+      "num_tokens": 11174222.0,
+      "step": 1060
+    },
+    {
+      "entropy": 0.31053728312253953,
+      "epoch": 2.4541284403669725,
+      "grad_norm": 1.8984375,
+      "learning_rate": 1.9666826626713208e-06,
+      "loss": 0.3087751865386963,
+      "mean_token_accuracy": 0.9034694701433181,
+      "num_tokens": 11283325.0,
+      "step": 1070
+    },
+    {
+      "entropy": 0.31755390875041484,
+      "epoch": 2.477064220183486,
+      "grad_norm": 1.9296875,
+      "learning_rate": 1.810606457521965e-06,
+      "loss": 0.30810863971710206,
+      "mean_token_accuracy": 0.9028798103332519,
+      "num_tokens": 11388674.0,
+      "step": 1080
+    },
+    {
+      "entropy": 0.31256188340485097,
+      "epoch": 2.5,
+      "grad_norm": 2.484375,
+      "learning_rate": 1.660364337707957e-06,
+      "loss": 0.3072192668914795,
+      "mean_token_accuracy": 0.9016716808080674,
+      "num_tokens": 11495818.0,
+      "step": 1090
+    },
+    {
+      "entropy": 0.3098844654858112,
+      "epoch": 2.522935779816514,
+      "grad_norm": 2.21875,
+      "learning_rate": 1.516063335006851e-06,
+      "loss": 0.30339553356170657,
+      "mean_token_accuracy": 0.903311139345169,
+      "num_tokens": 11600053.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.522935779816514,
+      "eval_entropy": 0.3133478334740452,
+      "eval_loss": 0.3417088985443115,
+      "eval_mean_token_accuracy": 0.8939231673012609,
+      "eval_num_tokens": 11600053.0,
+      "eval_runtime": 30.6145,
+      "eval_samples_per_second": 11.988,
+      "eval_steps_per_second": 3.005,
+      "step": 1100
+    },
+    {
+      "entropy": 0.3077207863330841,
+      "epoch": 2.5458715596330275,
+      "grad_norm": 2.09375,
+      "learning_rate": 1.3778062487723964e-06,
+      "loss": 0.29884703159332277,
+      "mean_token_accuracy": 0.9049400717020035,
+      "num_tokens": 11708985.0,
+      "step": 1110
+    },
+    {
+      "entropy": 0.31691030226647854,
+      "epoch": 2.5688073394495414,
+      "grad_norm": 2.09375,
+      "learning_rate": 1.2456915727007446e-06,
+      "loss": 0.3085756540298462,
+      "mean_token_accuracy": 0.9018509045243264,
+      "num_tokens": 11815323.0,
+      "step": 1120
+    },
+    {
+      "entropy": 0.3215593289583921,
+      "epoch": 2.591743119266055,
+      "grad_norm": 1.8359375,
+      "learning_rate": 1.1198134246640347e-06,
+      "loss": 0.31881897449493407,
+      "mean_token_accuracy": 0.8997952118515968,
+      "num_tokens": 11920222.0,
+      "step": 1130
+    },
+    {
+      "entropy": 0.32174803763628007,
+      "epoch": 2.614678899082569,
+      "grad_norm": 1.828125,
+      "learning_rate": 1.0002614796612343e-06,
+      "loss": 0.3117291212081909,
+      "mean_token_accuracy": 0.8999653860926629,
+      "num_tokens": 12024767.0,
+      "step": 1140
+    },
+    {
+      "entropy": 0.3095819726586342,
+      "epoch": 2.6376146788990824,
+      "grad_norm": 2.3125,
+      "learning_rate": 8.871209059341135e-07,
+      "loss": 0.30310988426208496,
+      "mean_token_accuracy": 0.9035083040595054,
+      "num_tokens": 12133161.0,
+      "step": 1150
+    },
+    {
+      "entropy": 0.2892547409981489,
+      "epoch": 2.6605504587155964,
+      "grad_norm": 1.671875,
+      "learning_rate": 7.804723042937756e-07,
+      "loss": 0.2832754850387573,
+      "mean_token_accuracy": 0.909184193611145,
+      "num_tokens": 12246860.0,
+      "step": 1160
+    },
+    {
+      "entropy": 0.29976772107183935,
+      "epoch": 2.68348623853211,
+      "grad_norm": 1.828125,
+      "learning_rate": 6.803916507010388e-07,
+      "loss": 0.29517574310302735,
+      "mean_token_accuracy": 0.9073081210255622,
+      "num_tokens": 12353437.0,
+      "step": 1170
+    },
+    {
+      "entropy": 0.31047735884785654,
+      "epoch": 2.706422018348624,
+      "grad_norm": 1.7421875,
+      "learning_rate": 5.869502421415041e-07,
+      "loss": 0.30287339687347414,
+      "mean_token_accuracy": 0.9031920969486237,
+      "num_tokens": 12462287.0,
+      "step": 1180
+    },
+    {
+      "entropy": 0.3041377454996109,
+      "epoch": 2.729357798165138,
+      "grad_norm": 2.015625,
+      "learning_rate": 5.002146458339185e-07,
+      "loss": 0.2944758176803589,
+      "mean_token_accuracy": 0.9054106563329697,
+      "num_tokens": 12566215.0,
+      "step": 1190
+    },
+    {
+      "entropy": 0.31409593895077703,
+      "epoch": 2.7522935779816513,
+      "grad_norm": 1.7265625,
+      "learning_rate": 4.202466518080184e-07,
+      "loss": 0.299582839012146,
+      "mean_token_accuracy": 0.9041049733757973,
+      "num_tokens": 12673542.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.7522935779816513,
+      "eval_entropy": 0.3133090068788632,
+      "eval_loss": 0.3416964113712311,
+      "eval_mean_token_accuracy": 0.8939006931107977,
+      "eval_num_tokens": 12673542.0,
+      "eval_runtime": 30.4519,
+      "eval_samples_per_second": 12.052,
+      "eval_steps_per_second": 3.021,
+      "step": 1200
+    },
+    {
+      "entropy": 0.30692093782126906,
+      "epoch": 2.7752293577981653,
+      "grad_norm": 1.8359375,
+      "learning_rate": 3.471032288855869e-07,
+      "loss": 0.2946175575256348,
+      "mean_token_accuracy": 0.9054166913032532,
+      "num_tokens": 12778933.0,
+      "step": 1210
+    },
+    {
+      "entropy": 0.31899234019219874,
+      "epoch": 2.7981651376146788,
+      "grad_norm": 2.1875,
+      "learning_rate": 2.8083648409614197e-07,
+      "loss": 0.3127703905105591,
+      "mean_token_accuracy": 0.9000537022948265,
+      "num_tokens": 12879669.0,
+      "step": 1220
+    },
+    {
+      "entropy": 0.29921196661889554,
+      "epoch": 2.8211009174311927,
+      "grad_norm": 1.6640625,
+      "learning_rate": 2.2149362555612951e-07,
+      "loss": 0.29253277778625486,
+      "mean_token_accuracy": 0.9067902445793152,
+      "num_tokens": 12989616.0,
+      "step": 1230
+    },
+    {
+      "entropy": 0.311046426743269,
+      "epoch": 2.8440366972477067,
+      "grad_norm": 2.0,
+      "learning_rate": 1.6911692883809338e-07,
+      "loss": 0.3050737142562866,
+      "mean_token_accuracy": 0.9037070229649544,
+      "num_tokens": 13092246.0,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3184209030121565,
+      "epoch": 2.86697247706422,
+      "grad_norm": 2.359375,
+      "learning_rate": 1.2374370685374504e-07,
+      "loss": 0.31279401779174804,
+      "mean_token_accuracy": 0.9010333761572837,
+      "num_tokens": 13196482.0,
+      "step": 1250
+    },
+    {
+      "entropy": 0.30298629961907864,
+      "epoch": 2.8899082568807337,
+      "grad_norm": 1.7578125,
+      "learning_rate": 8.540628327242894e-08,
+      "loss": 0.29525203704833985,
+      "mean_token_accuracy": 0.9062755778431892,
+      "num_tokens": 13296319.0,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3114538390189409,
+      "epoch": 2.9128440366972477,
+      "grad_norm": 2.125,
+      "learning_rate": 5.41319694938891e-08,
+      "loss": 0.31121370792388914,
+      "mean_token_accuracy": 0.902065885066986,
+      "num_tokens": 13401177.0,
+      "step": 1270
+    },
+    {
+      "entropy": 0.30996891260147097,
+      "epoch": 2.9357798165137616,
+      "grad_norm": 2.03125,
+      "learning_rate": 2.994304519176261e-08,
+      "loss": 0.3023967742919922,
+      "mean_token_accuracy": 0.9039678320288658,
+      "num_tokens": 13506546.0,
+      "step": 1280
+    },
+    {
+      "entropy": 0.3166234001517296,
+      "epoch": 2.958715596330275,
+      "grad_norm": 2.0625,
+      "learning_rate": 1.2856742441637037e-08,
+      "loss": 0.31199233531951903,
+      "mean_token_accuracy": 0.90199134349823,
+      "num_tokens": 13612665.0,
+      "step": 1290
+    },
+    {
+      "entropy": 0.3202438112348318,
+      "epoch": 2.981651376146789,
+      "grad_norm": 1.796875,
+      "learning_rate": 2.885233445004687e-09,
+      "loss": 0.32076048851013184,
+      "mean_token_accuracy": 0.8987957209348678,
+      "num_tokens": 13719965.0,
+      "step": 1300
+    },
+    {
+      "epoch": 2.981651376146789,
+      "eval_entropy": 0.31330409516458924,
+      "eval_loss": 0.34160733222961426,
+      "eval_mean_token_accuracy": 0.8939757956110913,
+      "eval_num_tokens": 13719965.0,
+      "eval_runtime": 30.3526,
+      "eval_samples_per_second": 12.091,
+      "eval_steps_per_second": 3.031,
+      "step": 1300
+    },
+    {
+      "epoch": 3.0,
+      "eval_entropy": 0.3133045512696971,
+      "eval_loss": 0.34170931577682495,
+      "eval_mean_token_accuracy": 0.8939821726602056,
+      "eval_num_tokens": 13804929.0,
+      "eval_runtime": 30.8394,
+      "eval_samples_per_second": 11.9,
+      "eval_steps_per_second": 2.983,
+      "step": 1308
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 1308,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 3,
+  "save_steps": 100,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.6812284470342656e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/last-checkpoint/training_args.bin b/last-checkpoint/training_args.bin
new file mode 100644
index 0000000..d73d64c
--- /dev/null
+++ b/last-checkpoint/training_args.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:41d016ec881fc98db150a1b86f6aee9e58d45c74d42e04ea3a1d58d824b4fc43
+size 5841
diff --git a/model.safetensors b/model.safetensors
new file mode 100644
index 0000000..262953d
--- /dev/null
+++ b/model.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:523bdabd1b10000de1c2ac13171baa0f01bbc4e7b40e4dc47799d2b5812bd1c1
+size 3441185608
diff --git a/tokenizer.json b/tokenizer.json
new file mode 100644
index 0000000..c7afbed
--- /dev/null
+++ b/tokenizer.json
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:be75606093db2094d7cd20f3c2f385c212750648bd6ea4fb2bf507a6a4c55506
+size 11422650
diff --git a/tokenizer_config.json b/tokenizer_config.json
new file mode 100644
index 0000000..770e41d
--- /dev/null
+++ b/tokenizer_config.json
@@ -0,0 +1,30 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": false,
+  "local_files_only": false,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}
diff --git a/training_args.bin b/training_args.bin
new file mode 100644
index 0000000..d73d64c
--- /dev/null
+++ b/training_args.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:41d016ec881fc98db150a1b86f6aee9e58d45c74d42e04ea3a1d58d824b4fc43
+size 5841