Add files using upload-large-folder tool

.gitattributes

@@ -43,3 +43,7 @@ wandb/run-20251226_180808-ny9q48hd/run-ny9q48hd.wandb filter=lfs diff=lfs merge=
 wandb/run-20251226_180702-oordmylf/run-oordmylf.wandb filter=lfs diff=lfs merge=lfs -text
 wandb/run-20251226_180613-i1cmzyri/run-i1cmzyri.wandb filter=lfs diff=lfs merge=lfs -text
 wandb/run-20251226_181544-upub1jan/run-upub1jan.wandb filter=lfs diff=lfs merge=lfs -text
+best_adapter/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+checkpoints/checkpoint-500/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+checkpoints/checkpoint-400/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+wandb/run-20251227_194423-jz7bptqa/run-jz7bptqa.wandb filter=lfs diff=lfs merge=lfs -text

best_adapter/README.md

@@ -1,9 +1,9 @@
 ---
-base_model: Models/Qwen2.5-Coder-14B-CPT
+base_model: /workspace/Models/Qwen2.5-Coder-14B-CPT-SFT_v2
 library_name: peft
 pipeline_tag: text-generation
 tags:
-- base_model:adapter:Models/Qwen2.5-Coder-14B-CPT
+- base_model:adapter:/workspace/Models/Qwen2.5-Coder-14B-CPT-SFT_v2
 - lora
 - transformers
 ---

best_adapter/adapter_config.json

@@ -3,7 +3,7 @@
   "alpha_pattern": {},
   "arrow_config": null,
   "auto_mapping": null,
-  "base_model_name_or_path": "Models/Qwen2.5-Coder-14B-CPT",
+  "base_model_name_or_path": "/workspace/Models/Qwen2.5-Coder-14B-CPT-SFT_v2",
   "bias": "none",
   "corda_config": null,
   "ensure_weight_tying": false,
@@ -16,7 +16,7 @@
   "layers_pattern": null,
   "layers_to_transform": null,
   "loftq_config": {},
-  "lora_alpha": 64,
+  "lora_alpha": 32,
   "lora_bias": false,
   "lora_dropout": 0.05,
   "megatron_config": null,
@@ -25,14 +25,17 @@
   "peft_type": "LORA",
   "peft_version": "0.18.0",
   "qalora_group_size": 16,
-  "r": 32,
+  "r": 16,
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
+    "q_proj",
     "o_proj",
     "v_proj",
-    "k_proj",
-    "q_proj"
+    "up_proj",
+    "gate_proj",
+    "down_proj",
+    "k_proj"
   ],
   "target_parameters": null,
   "task_type": "CAUSAL_LM",

best_adapter/added_tokens.json

@@ -0,0 +1,24 @@
+{
+  "</tool_call>": 151658,
+  "<tool_call>": 151657,
+  "<|box_end|>": 151649,
+  "<|box_start|>": 151648,
+  "<|endoftext|>": 151643,
+  "<|file_sep|>": 151664,
+  "<|fim_middle|>": 151660,
+  "<|fim_pad|>": 151662,
+  "<|fim_prefix|>": 151659,
+  "<|fim_suffix|>": 151661,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644,
+  "<|image_pad|>": 151655,
+  "<|object_ref_end|>": 151647,
+  "<|object_ref_start|>": 151646,
+  "<|quad_end|>": 151651,
+  "<|quad_start|>": 151650,
+  "<|repo_name|>": 151663,
+  "<|video_pad|>": 151656,
+  "<|vision_end|>": 151653,
+  "<|vision_pad|>": 151654,
+  "<|vision_start|>": 151652
+}

best_adapter/chat_template.jinja

@@ -0,0 +1,54 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0]['role'] == 'system' %}
+        {{- messages[0]['content'] }}
+    {%- else %}
+        {{- 'You are a helpful assistant.' }}
+    {%- endif %}
+    {{- "\n\n# 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' }}
+    {%- else %}
+        {{- '<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) or (message.role == "assistant" and not message.tool_calls) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {{- '<|im_start|>' + message.role }}
+        {%- if message.content %}
+            {{- '\n' + message.content }}
+        {%- endif %}
+        {%- for tool_call in message.tool_calls %}
+            {%- if tool_call.function is defined %}
+                {%- set tool_call = tool_call.function %}
+            {%- endif %}
+            {{- '\n<tool_call>\n{"name": "' }}
+            {{- tool_call.name }}
+            {{- '", "arguments": ' }}
+            {{- tool_call.arguments | tojson }}
+            {{- '}\n</tool_call>' }}
+        {%- endfor %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if (loop.index0 == 0) 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' }}
+{%- endif %}

best_adapter/special_tokens_map.json

@@ -0,0 +1,31 @@
+{
+  "additional_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|>"
+  ],
+  "eos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

best_adapter/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9c5ae00e602b8860cbd784ba82a8aa14e8feecec692e7076590d014d7b7fdafa
+size 11421896

best_adapter/tokenizer_config.json

@@ -0,0 +1,207 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "additional_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|>"
+  ],
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": {},
+  "model_max_length": 32768,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

best_adapter/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:afae4a8ce1391f149eb00b14eed8f891c715e892ea546bf754d22db2c2bc5969
-size 4792
+oid sha256:a4a4e48ed61b7c96f3bd2836ac828013a311834ab8a9542ea461fe1ff953396b
+size 5496

checkpoints/checkpoint-400/README.md

@@ -0,0 +1,207 @@
+---
+base_model: /workspace/Models/Qwen2.5-Coder-14B-CPT-SFT_v2
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:/workspace/Models/Qwen2.5-Coder-14B-CPT-SFT_v2
+- lora
+- transformers
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.0

checkpoints/checkpoint-400/adapter_config.json

@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "/workspace/Models/Qwen2.5-Coder-14B-CPT-SFT_v2",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 32,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.0",
+  "qalora_group_size": 16,
+  "r": 16,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "up_proj",
+    "gate_proj",
+    "down_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

checkpoints/checkpoint-400/added_tokens.json

@@ -0,0 +1,24 @@
+{
+  "</tool_call>": 151658,
+  "<tool_call>": 151657,
+  "<|box_end|>": 151649,
+  "<|box_start|>": 151648,
+  "<|endoftext|>": 151643,
+  "<|file_sep|>": 151664,
+  "<|fim_middle|>": 151660,
+  "<|fim_pad|>": 151662,
+  "<|fim_prefix|>": 151659,
+  "<|fim_suffix|>": 151661,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644,
+  "<|image_pad|>": 151655,
+  "<|object_ref_end|>": 151647,
+  "<|object_ref_start|>": 151646,
+  "<|quad_end|>": 151651,
+  "<|quad_start|>": 151650,
+  "<|repo_name|>": 151663,
+  "<|video_pad|>": 151656,
+  "<|vision_end|>": 151653,
+  "<|vision_pad|>": 151654,
+  "<|vision_start|>": 151652
+}

checkpoints/checkpoint-400/chat_template.jinja

@@ -0,0 +1,54 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0]['role'] == 'system' %}
+        {{- messages[0]['content'] }}
+    {%- else %}
+        {{- 'You are a helpful assistant.' }}
+    {%- endif %}
+    {{- "\n\n# 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' }}
+    {%- else %}
+        {{- '<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) or (message.role == "assistant" and not message.tool_calls) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {{- '<|im_start|>' + message.role }}
+        {%- if message.content %}
+            {{- '\n' + message.content }}
+        {%- endif %}
+        {%- for tool_call in message.tool_calls %}
+            {%- if tool_call.function is defined %}
+                {%- set tool_call = tool_call.function %}
+            {%- endif %}
+            {{- '\n<tool_call>\n{"name": "' }}
+            {{- tool_call.name }}
+            {{- '", "arguments": ' }}
+            {{- tool_call.arguments | tojson }}
+            {{- '}\n</tool_call>' }}
+        {%- endfor %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if (loop.index0 == 0) 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' }}
+{%- endif %}

checkpoints/checkpoint-400/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:df773721583080a5937ae757c6cabb41d190e187d9dcba072940a8fb11a22714
+size 14244

checkpoints/checkpoint-400/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f06565d55295681785448bb88140fb9d768cda54cb2f3e5bfc6f0d216a81cfc8
+size 1064

checkpoints/checkpoint-400/special_tokens_map.json

@@ -0,0 +1,31 @@
+{
+  "additional_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|>"
+  ],
+  "eos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

checkpoints/checkpoint-400/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9c5ae00e602b8860cbd784ba82a8aa14e8feecec692e7076590d014d7b7fdafa
+size 11421896

checkpoints/checkpoint-400/tokenizer_config.json

@@ -0,0 +1,207 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "additional_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|>"
+  ],
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": {},
+  "model_max_length": 32768,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

checkpoints/checkpoint-400/trainer_state.json

@@ -0,0 +1,914 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.6006006006006006,
+  "eval_steps": 50,
+  "global_step": 400,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.02002002002002002,
+      "grad_norm": 0.05460292845964432,
+      "grpo_mean_advantage": -1.3560057254835556e-07,
+      "grpo_mean_group_score": 0.5922331809997559,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 3.0318567496578908e-06,
+      "learning_rate": 8.000000000000001e-07,
+      "loss": 0.007,
+      "step": 5
+    },
+    {
+      "epoch": 0.04004004004004004,
+      "grad_norm": 0.0679207444190979,
+      "grpo_mean_advantage": 3.6619603633880615e-06,
+      "grpo_mean_group_score": 0.5561589002609253,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.6246918676188216e-05,
+      "learning_rate": 1.8000000000000001e-06,
+      "loss": 0.0107,
+      "step": 10
+    },
+    {
+      "epoch": 0.06006006006006006,
+      "grad_norm": 0.05788416787981987,
+      "grpo_mean_advantage": -1.0654330395709621e-07,
+      "grpo_mean_group_score": 0.5759152173995972,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 5.399440965447866e-07,
+      "learning_rate": 2.8000000000000003e-06,
+      "loss": 0.007,
+      "step": 15
+    },
+    {
+      "epoch": 0.08008008008008008,
+      "grad_norm": 0.0746568813920021,
+      "grpo_mean_advantage": -5.871057737749652e-07,
+      "grpo_mean_group_score": 0.5127314329147339,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 2.6951597646984737e-06,
+      "learning_rate": 3.8000000000000005e-06,
+      "loss": 0.0246,
+      "step": 20
+    },
+    {
+      "epoch": 0.1001001001001001,
+      "grad_norm": 0.11442846059799194,
+      "grpo_mean_advantage": 6.370246410369873e-07,
+      "grpo_mean_group_score": 0.539706826210022,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 2.8908377771585947e-06,
+      "learning_rate": 4.800000000000001e-06,
+      "loss": 0.0337,
+      "step": 25
+    },
+    {
+      "epoch": 0.12012012012012012,
+      "grad_norm": 0.05778791010379791,
+      "grpo_mean_advantage": 6.705522359595761e-09,
+      "grpo_mean_group_score": 0.5812538862228394,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 6.189450800775376e-07,
+      "learning_rate": 4.999125183044924e-06,
+      "loss": 0.0171,
+      "step": 30
+    },
+    {
+      "epoch": 0.14014014014014015,
+      "grad_norm": 0.05819695070385933,
+      "grpo_mean_advantage": 3.859400692363124e-07,
+      "grpo_mean_group_score": 0.5909844636917114,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.6833292875162442e-06,
+      "learning_rate": 4.995572288443412e-06,
+      "loss": 0.0145,
+      "step": 35
+    },
+    {
+      "epoch": 0.16016016016016016,
+      "grad_norm": 0.07968433201313019,
+      "grpo_mean_advantage": 2.600252742013254e-07,
+      "grpo_mean_group_score": 0.5630953907966614,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.4095899132371414e-06,
+      "learning_rate": 4.98929052218411e-06,
+      "loss": 0.0196,
+      "step": 40
+    },
+    {
+      "epoch": 0.18018018018018017,
+      "grad_norm": 0.0733402892947197,
+      "grpo_mean_advantage": -1.2591480924584175e-07,
+      "grpo_mean_group_score": 0.5604403614997864,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.0309080380466185e-06,
+      "learning_rate": 4.980286753286196e-06,
+      "loss": 0.0186,
+      "step": 45
+    },
+    {
+      "epoch": 0.2002002002002002,
+      "grad_norm": 0.07136482000350952,
+      "grpo_mean_advantage": -2.808868941883702e-07,
+      "grpo_mean_group_score": 0.5971035957336426,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.5696078889959608e-06,
+      "learning_rate": 4.9685708272387645e-06,
+      "loss": 0.0286,
+      "step": 50
+    },
+    {
+      "epoch": 0.22022022022022023,
+      "grad_norm": 0.08851475268602371,
+      "grpo_mean_advantage": 2.6822089438383045e-08,
+      "grpo_mean_group_score": 0.5892971754074097,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 3.7878271541558206e-07,
+      "learning_rate": 4.9541555552349404e-06,
+      "loss": 0.0054,
+      "step": 55
+    },
+    {
+      "epoch": 0.24024024024024024,
+      "grad_norm": 0.07778509706258774,
+      "grpo_mean_advantage": -5.662441182607836e-08,
+      "grpo_mean_group_score": 0.564322292804718,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 6.128998393251095e-07,
+      "learning_rate": 4.9370567001630155e-06,
+      "loss": -0.0074,
+      "step": 60
+    },
+    {
+      "epoch": 0.2602602602602603,
+      "grad_norm": 0.08740051090717316,
+      "grpo_mean_advantage": -1.5944242193199898e-07,
+      "grpo_mean_group_score": 0.562497615814209,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.6374274309782777e-06,
+      "learning_rate": 4.917292959369968e-06,
+      "loss": 0.0145,
+      "step": 65
+    },
+    {
+      "epoch": 0.2802802802802803,
+      "grad_norm": 0.19070060551166534,
+      "grpo_mean_advantage": 1.6838312433264946e-07,
+      "grpo_mean_group_score": 0.5904761552810669,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 8.536571272088622e-07,
+      "learning_rate": 4.8948859442161876e-06,
+      "loss": 0.0257,
+      "step": 70
+    },
+    {
+      "epoch": 0.3003003003003003,
+      "grad_norm": 0.07321271300315857,
+      "grpo_mean_advantage": 1.1175870895385742e-07,
+      "grpo_mean_group_score": 0.5765624046325684,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 6.451961667153228e-07,
+      "learning_rate": 4.869860156443768e-06,
+      "loss": 0.0024,
+      "step": 75
+    },
+    {
+      "epoch": 0.3203203203203203,
+      "grad_norm": 0.07126748561859131,
+      "grpo_mean_advantage": -1.4603138254187797e-07,
+      "grpo_mean_group_score": 0.5858271718025208,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.1309343790344428e-06,
+      "learning_rate": 4.842242961384211e-06,
+      "loss": 0.0277,
+      "step": 80
+    },
+    {
+      "epoch": 0.34034034034034033,
+      "grad_norm": 0.08629189431667328,
+      "grpo_mean_advantage": -1.817941665649414e-06,
+      "grpo_mean_group_score": 0.5871662497520447,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.1141768482048064e-05,
+      "learning_rate": 4.812064558034847e-06,
+      "loss": 0.0246,
+      "step": 85
+    },
+    {
+      "epoch": 0.36036036036036034,
+      "grad_norm": 0.0998779758810997,
+      "grpo_mean_advantage": 1.8179416372277046e-07,
+      "grpo_mean_group_score": 0.5330992937088013,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 6.210335072864837e-07,
+      "learning_rate": 4.779357946036662e-06,
+      "loss": 0.0056,
+      "step": 90
+    },
+    {
+      "epoch": 0.38038038038038036,
+      "grad_norm": 0.10614689439535141,
+      "grpo_mean_advantage": -2.972781771859445e-07,
+      "grpo_mean_group_score": 0.5265295505523682,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 3.1582342217006953e-06,
+      "learning_rate": 4.74415888958968e-06,
+      "loss": 0.0053,
+      "step": 95
+    },
+    {
+      "epoch": 0.4004004004004004,
+      "grad_norm": 0.10345634073019028,
+      "grpo_mean_advantage": -7.033348197182931e-07,
+      "grpo_mean_group_score": 0.5660771131515503,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 4.245831405569334e-06,
+      "learning_rate": 4.706505878345343e-06,
+      "loss": 0.0134,
+      "step": 100
+    },
+    {
+      "epoch": 0.42042042042042044,
+      "grad_norm": 0.10077933222055435,
+      "grpo_mean_advantage": 1.1920928955078125e-07,
+      "grpo_mean_group_score": 0.57631915807724,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 3.2809634831210133e-07,
+      "learning_rate": 4.666440085318626e-06,
+      "loss": 0.0004,
+      "step": 105
+    },
+    {
+      "epoch": 0.44044044044044045,
+      "grad_norm": 0.09548182785511017,
+      "grpo_mean_advantage": -4.0978193283081055e-07,
+      "grpo_mean_group_score": 0.546563982963562,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 6.0397578636184335e-06,
+      "learning_rate": 4.624005321865968e-06,
+      "loss": 0.0033,
+      "step": 110
+    },
+    {
+      "epoch": 0.46046046046046046,
+      "grad_norm": 0.09417816251516342,
+      "grpo_mean_advantage": -1.467764434437413e-07,
+      "grpo_mean_group_score": 0.5519219636917114,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 2.2689375782647403e-06,
+      "learning_rate": 4.57924798977818e-06,
+      "loss": 0.0095,
+      "step": 115
+    },
+    {
+      "epoch": 0.4804804804804805,
+      "grad_norm": 0.10022275149822235,
+      "grpo_mean_advantage": -5.215406329028838e-09,
+      "grpo_mean_group_score": 0.5490407943725586,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 7.929010621410271e-07,
+      "learning_rate": 4.532217030540781e-06,
+      "loss": 0.0006,
+      "step": 120
+    },
+    {
+      "epoch": 0.5005005005005005,
+      "grad_norm": 0.14057794213294983,
+      "grpo_mean_advantage": -5.7369469175228005e-08,
+      "grpo_mean_group_score": 0.5646580457687378,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.2823379620385822e-06,
+      "learning_rate": 4.482963871817195e-06,
+      "loss": -0.0046,
+      "step": 125
+    },
+    {
+      "epoch": 0.5205205205205206,
+      "grad_norm": 0.12420658767223358,
+      "grpo_mean_advantage": 2.9876827056796174e-07,
+      "grpo_mean_group_score": 0.6111599802970886,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.0496698905626545e-06,
+      "learning_rate": 4.4315423712133595e-06,
+      "loss": -0.003,
+      "step": 130
+    },
+    {
+      "epoch": 0.5405405405405406,
+      "grad_norm": 0.14342808723449707,
+      "grpo_mean_advantage": 1.5869736103013565e-07,
+      "grpo_mean_group_score": 0.5619662404060364,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.2748531617035042e-06,
+      "learning_rate": 4.378008757385222e-06,
+      "loss": 0.0154,
+      "step": 135
+    },
+    {
+      "epoch": 0.5605605605605606,
+      "grad_norm": 0.14729444682598114,
+      "grpo_mean_advantage": 3.0100346748440643e-07,
+      "grpo_mean_group_score": 0.5795454978942871,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 2.4499684059264837e-06,
+      "learning_rate": 4.322421568553529e-06,
+      "loss": -0.0262,
+      "step": 140
+    },
+    {
+      "epoch": 0.5805805805805806,
+      "grad_norm": 0.15249410271644592,
+      "grpo_mean_advantage": -3.233552092751779e-07,
+      "grpo_mean_group_score": 0.5804953575134277,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.248456669600273e-06,
+      "learning_rate": 4.2648415884931476e-06,
+      "loss": 0.0018,
+      "step": 145
+    },
+    {
+      "epoch": 0.6006006006006006,
+      "grad_norm": 0.1841023564338684,
+      "grpo_mean_advantage": 3.2261013416245987e-07,
+      "grpo_mean_group_score": 0.5628539323806763,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.4773489738217904e-06,
+      "learning_rate": 4.205331780066892e-06,
+      "loss": -0.017,
+      "step": 150
+    },
+    {
+      "epoch": 0.6206206206206206,
+      "grad_norm": 0.18597163259983063,
+      "grpo_mean_advantage": -2.5331974029541016e-07,
+      "grpo_mean_group_score": 0.5727725625038147,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.5092309695319273e-06,
+      "learning_rate": 4.1439572163765615e-06,
+      "loss": 0.0044,
+      "step": 155
+    },
+    {
+      "epoch": 0.6406406406406406,
+      "grad_norm": 0.18310388922691345,
+      "grpo_mean_advantage": -6.780028627417778e-08,
+      "grpo_mean_group_score": 0.5833909511566162,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 8.550978805033083e-07,
+      "learning_rate": 4.0807850096064605e-06,
+      "loss": -0.005,
+      "step": 160
+    },
+    {
+      "epoch": 0.6606606606606606,
+      "grad_norm": 0.2192923128604889,
+      "grpo_mean_advantage": -5.587935447692871e-08,
+      "grpo_mean_group_score": 0.5742615461349487,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 3.564579174053506e-07,
+      "learning_rate": 4.015884237637206e-06,
+      "loss": -0.015,
+      "step": 165
+    },
+    {
+      "epoch": 0.6806806806806807,
+      "grad_norm": 0.16708803176879883,
+      "grpo_mean_advantage": -5.327165126800537e-07,
+      "grpo_mean_group_score": 0.5758188962936401,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 2.309018327650847e-06,
+      "learning_rate": 3.949325868510083e-06,
+      "loss": -0.0314,
+      "step": 170
+    },
+    {
+      "epoch": 0.7007007007007007,
+      "grad_norm": 0.3401262164115906,
+      "grpo_mean_advantage": 5.863606702405377e-07,
+      "grpo_mean_group_score": 0.5767683982849121,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 2.4449204829579685e-06,
+      "learning_rate": 3.881182682824534e-06,
+      "loss": -0.0441,
+      "step": 175
+    },
+    {
+      "epoch": 0.7207207207207207,
+      "grad_norm": 0.1931898146867752,
+      "grpo_mean_advantage": 3.2186508747145126e-07,
+      "grpo_mean_group_score": 0.586772084236145,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 2.293551688126172e-06,
+      "learning_rate": 3.811529194153635e-06,
+      "loss": -0.0162,
+      "step": 180
+    },
+    {
+      "epoch": 0.7407407407407407,
+      "grad_norm": 0.2537969648838043,
+      "grpo_mean_advantage": -4.470348358154297e-08,
+      "grpo_mean_group_score": 0.549396276473999,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 3.7067667335577426e-07,
+      "learning_rate": 3.7404415675646054e-06,
+      "loss": -0.0386,
+      "step": 185
+    },
+    {
+      "epoch": 0.7607607607607607,
+      "grad_norm": 0.20326584577560425,
+      "grpo_mean_advantage": -2.1010637851759384e-07,
+      "grpo_mean_group_score": 0.5798425078392029,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.1695076409523608e-06,
+      "learning_rate": 3.667997536333424e-06,
+      "loss": -0.037,
+      "step": 190
+    },
+    {
+      "epoch": 0.7807807807807807,
+      "grad_norm": 0.25048357248306274,
+      "grpo_mean_advantage": 1.765787658314366e-07,
+      "grpo_mean_group_score": 0.5584167838096619,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 2.429934738756856e-06,
+      "learning_rate": 3.59427631694463e-06,
+      "loss": -0.0292,
+      "step": 195
+    },
+    {
+      "epoch": 0.8008008008008008,
+      "grad_norm": 0.2687569260597229,
+      "grpo_mean_advantage": 1.6540289493605087e-07,
+      "grpo_mean_group_score": 0.5676193237304688,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 2.6342788714828203e-06,
+      "learning_rate": 3.5193585224692595e-06,
+      "loss": -0.0454,
+      "step": 200
+    },
+    {
+      "epoch": 0.8208208208208209,
+      "grad_norm": 0.22301620244979858,
+      "grpo_mean_advantage": -1.0944902442133753e-06,
+      "grpo_mean_group_score": 0.5669739842414856,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 5.346942998585291e-06,
+      "learning_rate": 3.44332607441564e-06,
+      "loss": -0.0423,
+      "step": 205
+    },
+    {
+      "epoch": 0.8408408408408409,
+      "grad_norm": 0.3040211498737335,
+      "grpo_mean_advantage": 2.4065374759629776e-07,
+      "grpo_mean_group_score": 0.5922158360481262,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.6327536513927043e-06,
+      "learning_rate": 3.3662621131494204e-06,
+      "loss": -0.0857,
+      "step": 210
+    },
+    {
+      "epoch": 0.8608608608608609,
+      "grad_norm": 0.27231141924858093,
+      "grpo_mean_advantage": -5.21540641784668e-08,
+      "grpo_mean_group_score": 0.5473950505256653,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 5.847922466273303e-07,
+      "learning_rate": 3.2882509069808044e-06,
+      "loss": -0.0278,
+      "step": 215
+    },
+    {
+      "epoch": 0.8808808808808809,
+      "grad_norm": 0.3571636378765106,
+      "grpo_mean_advantage": 6.541609991472797e-07,
+      "grpo_mean_group_score": 0.5880032777786255,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 4.072162937518442e-06,
+      "learning_rate": 3.2093777600183873e-06,
+      "loss": -0.0727,
+      "step": 220
+    },
+    {
+      "epoch": 0.9009009009009009,
+      "grad_norm": 0.306273490190506,
+      "grpo_mean_advantage": -1.2218951894737984e-07,
+      "grpo_mean_group_score": 0.5835092663764954,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 4.386006935419573e-07,
+      "learning_rate": 3.1297289188903705e-06,
+      "loss": -0.0464,
+      "step": 225
+    },
+    {
+      "epoch": 0.9209209209209209,
+      "grad_norm": 0.2700377106666565,
+      "grpo_mean_advantage": 1.7605722177904681e-06,
+      "grpo_mean_group_score": 0.5394966006278992,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 8.007580618141219e-06,
+      "learning_rate": 3.049391478435133e-06,
+      "loss": -0.0295,
+      "step": 230
+    },
+    {
+      "epoch": 0.9409409409409409,
+      "grad_norm": 0.39531761407852173,
+      "grpo_mean_advantage": -3.3080578987210174e-07,
+      "grpo_mean_group_score": 0.5687432289123535,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.551636614749441e-06,
+      "learning_rate": 2.9684532864643123e-06,
+      "loss": -0.031,
+      "step": 235
+    },
+    {
+      "epoch": 0.960960960960961,
+      "grad_norm": 0.5987040996551514,
+      "grpo_mean_advantage": 2.712011450967111e-07,
+      "grpo_mean_group_score": 0.5550583600997925,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.4400844747797237e-06,
+      "learning_rate": 2.887002847702504e-06,
+      "loss": -0.0789,
+      "step": 240
+    },
+    {
+      "epoch": 0.980980980980981,
+      "grad_norm": 0.5680716037750244,
+      "grpo_mean_advantage": -3.2857059295565705e-07,
+      "grpo_mean_group_score": 0.558111310005188,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 2.105091425619321e-06,
+      "learning_rate": 2.8051292270086506e-06,
+      "loss": -0.1131,
+      "step": 245
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 0.6204046010971069,
+      "grpo_mean_advantage": 4.470348358154297e-08,
+      "grpo_mean_group_score": 0.6196198463439941,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 5.315724820320611e-07,
+      "learning_rate": 2.722921951984927e-06,
+      "loss": -0.2232,
+      "step": 250
+    },
+    {
+      "epoch": 1.02002002002002,
+      "grad_norm": 0.8389026522636414,
+      "grpo_mean_advantage": 9.290873776990338e-07,
+      "grpo_mean_group_score": 0.582168459892273,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 4.219644324621186e-06,
+      "learning_rate": 2.640470915079614e-06,
+      "loss": -0.1363,
+      "step": 255
+    },
+    {
+      "epoch": 1.04004004004004,
+      "grad_norm": 0.9067686796188354,
+      "grpo_mean_advantage": 2.533197474008375e-08,
+      "grpo_mean_group_score": 0.5551307797431946,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.6600588992332632e-07,
+      "learning_rate": 2.557866275291035e-06,
+      "loss": -0.1868,
+      "step": 260
+    },
+    {
+      "epoch": 1.06006006006006,
+      "grad_norm": 0.9277902841567993,
+      "grpo_mean_advantage": -5.662441182607836e-08,
+      "grpo_mean_group_score": 0.535040020942688,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.0909400316450046e-06,
+      "learning_rate": 2.4751983595800093e-06,
+      "loss": -0.1792,
+      "step": 265
+    },
+    {
+      "epoch": 1.08008008008008,
+      "grad_norm": 1.0715463161468506,
+      "grpo_mean_advantage": -9.536743306171047e-08,
+      "grpo_mean_group_score": 0.5673571825027466,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 5.838213610331877e-07,
+      "learning_rate": 2.392557564098649e-06,
+      "loss": -0.1691,
+      "step": 270
+    },
+    {
+      "epoch": 1.1001001001001,
+      "grad_norm": 0.7759184837341309,
+      "grpo_mean_advantage": 3.278255533700758e-08,
+      "grpo_mean_group_score": 0.5874732732772827,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 9.317170679423725e-07,
+      "learning_rate": 2.3100342553434924e-06,
+      "loss": -0.1655,
+      "step": 275
+    },
+    {
+      "epoch": 1.12012012012012,
+      "grad_norm": 0.9387398958206177,
+      "grpo_mean_advantage": -1.206994113545079e-07,
+      "grpo_mean_group_score": 0.5569106340408325,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 6.201085511747806e-07,
+      "learning_rate": 2.2277186713410688e-06,
+      "loss": -0.1821,
+      "step": 280
+    },
+    {
+      "epoch": 1.14014014014014,
+      "grad_norm": 1.6132302284240723,
+      "grpo_mean_advantage": 4.470348358154297e-08,
+      "grpo_mean_group_score": 0.5578873157501221,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 6.115651558502577e-07,
+      "learning_rate": 2.1457008229739395e-06,
+      "loss": -0.2102,
+      "step": 285
+    },
+    {
+      "epoch": 1.16016016016016,
+      "grad_norm": 0.8679026961326599,
+      "grpo_mean_advantage": -3.3453108017056365e-07,
+      "grpo_mean_group_score": 0.5735999345779419,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 3.5326345368957845e-06,
+      "learning_rate": 2.0640703955551214e-06,
+      "loss": -0.2937,
+      "step": 290
+    },
+    {
+      "epoch": 1.1801801801801801,
+      "grad_norm": 1.0550166368484497,
+      "grpo_mean_advantage": -1.110136480519941e-07,
+      "grpo_mean_group_score": 0.5626259446144104,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 4.731904823529476e-07,
+      "learning_rate": 1.9829166507585084e-06,
+      "loss": -0.2598,
+      "step": 295
+    },
+    {
+      "epoch": 1.2002002002002001,
+      "grad_norm": 1.2819372415542603,
+      "grpo_mean_advantage": -5.08874677507265e-07,
+      "grpo_mean_group_score": 0.5463050603866577,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.840126174101897e-06,
+      "learning_rate": 1.90232832901255e-06,
+      "loss": -0.2546,
+      "step": 300
+    },
+    {
+      "epoch": 1.2202202202202201,
+      "grad_norm": 1.0188143253326416,
+      "grpo_mean_advantage": 1.01327898960335e-07,
+      "grpo_mean_group_score": 0.5352144241333008,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 7.798533943059738e-07,
+      "learning_rate": 1.82239355246389e-06,
+      "loss": -0.1809,
+      "step": 305
+    },
+    {
+      "epoch": 1.2402402402402402,
+      "grad_norm": 2.0709052085876465,
+      "grpo_mean_advantage": 1.341104507446289e-07,
+      "grpo_mean_group_score": 0.5547868013381958,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 7.821902840987605e-07,
+      "learning_rate": 1.7431997286170923e-06,
+      "loss": -0.3559,
+      "step": 310
+    },
+    {
+      "epoch": 1.2602602602602602,
+      "grad_norm": 1.8516215085983276,
+      "grpo_mean_advantage": 9.015202806494926e-08,
+      "grpo_mean_group_score": 0.5859472751617432,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.0693488547985908e-06,
+      "learning_rate": 1.6648334547558227e-06,
+      "loss": -0.3874,
+      "step": 315
+    },
+    {
+      "epoch": 1.2802802802802802,
+      "grad_norm": 1.283104419708252,
+      "grpo_mean_advantage": -2.443790378947597e-07,
+      "grpo_mean_group_score": 0.5751550793647766,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.183122208203713e-06,
+      "learning_rate": 1.5873804232499862e-06,
+      "loss": -0.3467,
+      "step": 320
+    },
+    {
+      "epoch": 1.3003003003003002,
+      "grad_norm": 1.4108576774597168,
+      "grpo_mean_advantage": -6.705522537231445e-08,
+      "grpo_mean_group_score": 0.5497723817825317,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 6.109748937888071e-07,
+      "learning_rate": 1.51092532785238e-06,
+      "loss": -0.1703,
+      "step": 325
+    },
+    {
+      "epoch": 1.3203203203203202,
+      "grad_norm": 1.0421361923217773,
+      "grpo_mean_advantage": -1.639127766850379e-08,
+      "grpo_mean_group_score": 0.55989670753479,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 5.529495297196263e-07,
+      "learning_rate": 1.4355517710873184e-06,
+      "loss": -0.2918,
+      "step": 330
+    },
+    {
+      "epoch": 1.3403403403403402,
+      "grad_norm": 1.3465828895568848,
+      "grpo_mean_advantage": 4.418194237132411e-07,
+      "grpo_mean_group_score": 0.5809233784675598,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 2.9275292945385445e-06,
+      "learning_rate": 1.361342172832502e-06,
+      "loss": -0.3069,
+      "step": 335
+    },
+    {
+      "epoch": 1.3603603603603602,
+      "grad_norm": 1.1959459781646729,
+      "grpo_mean_advantage": 9.685754776000977e-08,
+      "grpo_mean_group_score": 0.5568087100982666,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 3.754235251562932e-07,
+      "learning_rate": 1.2883776801940884e-06,
+      "loss": -0.5594,
+      "step": 340
+    },
+    {
+      "epoch": 1.3803803803803802,
+      "grad_norm": 1.8967422246932983,
+      "grpo_mean_advantage": -2.384185791015625e-07,
+      "grpo_mean_group_score": 0.5655568838119507,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 6.821086913078034e-07,
+      "learning_rate": 1.216738078773522e-06,
+      "loss": -0.4102,
+      "step": 345
+    },
+    {
+      "epoch": 1.4004004004004005,
+      "grad_norm": 2.221132755279541,
+      "grpo_mean_advantage": -8.717179156292332e-08,
+      "grpo_mean_group_score": 0.6089578866958618,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 2.500940354366321e-06,
+      "learning_rate": 1.146501705423155e-06,
+      "loss": -0.338,
+      "step": 350
+    },
+    {
+      "epoch": 1.4204204204204205,
+      "grad_norm": 2.3640377521514893,
+      "grpo_mean_advantage": 2.1606683731079102e-07,
+      "grpo_mean_group_score": 0.6129671335220337,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.4568390724889468e-06,
+      "learning_rate": 1.0777453625860474e-06,
+      "loss": -0.4985,
+      "step": 355
+    },
+    {
+      "epoch": 1.4404404404404405,
+      "grad_norm": 1.9084734916687012,
+      "grpo_mean_advantage": -3.725290298461914e-09,
+      "grpo_mean_group_score": 0.5562310814857483,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 2.965894054796081e-06,
+      "learning_rate": 1.0105442343136184e-06,
+      "loss": -0.4347,
+      "step": 360
+    },
+    {
+      "epoch": 1.4604604604604605,
+      "grad_norm": 1.6063904762268066,
+      "grpo_mean_advantage": 4.313886279305734e-07,
+      "grpo_mean_group_score": 0.5884170532226562,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.9621948013082147e-06,
+      "learning_rate": 9.449718040529987e-07,
+      "loss": -0.6217,
+      "step": 365
+    },
+    {
+      "epoch": 1.4804804804804805,
+      "grad_norm": 2.114664077758789,
+      "grpo_mean_advantage": 2.0489096641540527e-07,
+      "grpo_mean_group_score": 0.5795440673828125,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.0235522722723545e-06,
+      "learning_rate": 8.810997742939531e-07,
+      "loss": -0.5364,
+      "step": 370
+    },
+    {
+      "epoch": 1.5005005005005005,
+      "grad_norm": 1.8450465202331543,
+      "grpo_mean_advantage": -1.4185905001795618e-06,
+      "grpo_mean_group_score": 0.5607603788375854,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.0947338523692451e-05,
+      "learning_rate": 8.189979881632634e-07,
+      "loss": -0.4798,
+      "step": 375
+    },
+    {
+      "epoch": 1.5205205205205206,
+      "grad_norm": 2.673438787460327,
+      "grpo_mean_advantage": -1.758337049295733e-07,
+      "grpo_mean_group_score": 0.5381432771682739,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 9.663675655247062e-07,
+      "learning_rate": 7.587343530522945e-07,
+      "loss": -0.4805,
+      "step": 380
+    },
+    {
+      "epoch": 1.5405405405405406,
+      "grad_norm": 2.2263550758361816,
+      "grpo_mean_advantage": -6.973743325033865e-07,
+      "grpo_mean_group_score": 0.5528443455696106,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 4.341973180999048e-06,
+      "learning_rate": 7.003747663612581e-07,
+      "loss": -0.433,
+      "step": 385
+    },
+    {
+      "epoch": 1.5605605605605606,
+      "grad_norm": 2.3657093048095703,
+      "grpo_mean_advantage": 1.7881394143159923e-08,
+      "grpo_mean_group_score": 0.6091476678848267,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.3004198251564958e-07,
+      "learning_rate": 6.439830434413754e-07,
+      "loss": -0.6021,
+      "step": 390
+    },
+    {
+      "epoch": 1.5805805805805806,
+      "grad_norm": 1.9847129583358765,
+      "grpo_mean_advantage": 3.4868716625169327e-07,
+      "grpo_mean_group_score": 0.5397372245788574,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 2.059372718576924e-06,
+      "learning_rate": 5.896208478137222e-07,
+      "loss": -0.5595,
+      "step": 395
+    },
+    {
+      "epoch": 1.6006006006006006,
+      "grad_norm": 2.922114133834839,
+      "grpo_mean_advantage": -2.1636485598719446e-06,
+      "grpo_mean_group_score": 0.5873125195503235,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 9.725940799398813e-06,
+      "learning_rate": 5.373476237410808e-07,
+      "loss": -0.5592,
+      "step": 400
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 500,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 100,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 0.0,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoints/checkpoint-400/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a4a4e48ed61b7c96f3bd2836ac828013a311834ab8a9542ea461fe1ff953396b
+size 5496

checkpoints/checkpoint-500/README.md

@@ -1,9 +1,9 @@
 ---
-base_model: Models/Qwen2.5-Coder-14B-CPT
+base_model: /workspace/Models/Qwen2.5-Coder-14B-CPT-SFT_v2
 library_name: peft
 pipeline_tag: text-generation
 tags:
-- base_model:adapter:Models/Qwen2.5-Coder-14B-CPT
+- base_model:adapter:/workspace/Models/Qwen2.5-Coder-14B-CPT-SFT_v2
 - lora
 - transformers
 ---

checkpoints/checkpoint-500/adapter_config.json

@@ -3,7 +3,7 @@
   "alpha_pattern": {},
   "arrow_config": null,
   "auto_mapping": null,
-  "base_model_name_or_path": "Models/Qwen2.5-Coder-14B-CPT",
+  "base_model_name_or_path": "/workspace/Models/Qwen2.5-Coder-14B-CPT-SFT_v2",
   "bias": "none",
   "corda_config": null,
   "ensure_weight_tying": false,
@@ -16,7 +16,7 @@
   "layers_pattern": null,
   "layers_to_transform": null,
   "loftq_config": {},
-  "lora_alpha": 64,
+  "lora_alpha": 32,
   "lora_bias": false,
   "lora_dropout": 0.05,
   "megatron_config": null,
@@ -25,14 +25,17 @@
   "peft_type": "LORA",
   "peft_version": "0.18.0",
   "qalora_group_size": 16,
-  "r": 32,
+  "r": 16,
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
+    "q_proj",
     "o_proj",
     "v_proj",
-    "k_proj",
-    "q_proj"
+    "up_proj",
+    "gate_proj",
+    "down_proj",
+    "k_proj"
   ],
   "target_parameters": null,
   "task_type": "CAUSAL_LM",

checkpoints/checkpoint-500/added_tokens.json

@@ -0,0 +1,24 @@
+{
+  "</tool_call>": 151658,
+  "<tool_call>": 151657,
+  "<|box_end|>": 151649,
+  "<|box_start|>": 151648,
+  "<|endoftext|>": 151643,
+  "<|file_sep|>": 151664,
+  "<|fim_middle|>": 151660,
+  "<|fim_pad|>": 151662,
+  "<|fim_prefix|>": 151659,
+  "<|fim_suffix|>": 151661,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644,
+  "<|image_pad|>": 151655,
+  "<|object_ref_end|>": 151647,
+  "<|object_ref_start|>": 151646,
+  "<|quad_end|>": 151651,
+  "<|quad_start|>": 151650,
+  "<|repo_name|>": 151663,
+  "<|video_pad|>": 151656,
+  "<|vision_end|>": 151653,
+  "<|vision_pad|>": 151654,
+  "<|vision_start|>": 151652
+}

checkpoints/checkpoint-500/chat_template.jinja

@@ -0,0 +1,54 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0]['role'] == 'system' %}
+        {{- messages[0]['content'] }}
+    {%- else %}
+        {{- 'You are a helpful assistant.' }}
+    {%- endif %}
+    {{- "\n\n# 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' }}
+    {%- else %}
+        {{- '<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- for message in messages %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) or (message.role == "assistant" and not message.tool_calls) %}
+        {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {{- '<|im_start|>' + message.role }}
+        {%- if message.content %}
+            {{- '\n' + message.content }}
+        {%- endif %}
+        {%- for tool_call in message.tool_calls %}
+            {%- if tool_call.function is defined %}
+                {%- set tool_call = tool_call.function %}
+            {%- endif %}
+            {{- '\n<tool_call>\n{"name": "' }}
+            {{- tool_call.name }}
+            {{- '", "arguments": ' }}
+            {{- tool_call.arguments | tojson }}
+            {{- '}\n</tool_call>' }}
+        {%- endfor %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if (loop.index0 == 0) 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' }}
+{%- endif %}

checkpoints/checkpoint-500/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7554aa7bc01964a4c12187b3026a2cce7c932c5c0fd522183b3a9f9414ab4d39
+oid sha256:140a5b8faae8f6aa3d7c8abd94812eb021a794706417ce578c4228067f08f646
 size 14244

checkpoints/checkpoint-500/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b9ab227797448d6a0b690540001676d252024d08b5dd438fce4fd8a3b96a64c5
+oid sha256:622e95a06e3b8ce556346f153f9b954936262c4f61ed83e89373f1300da67c29
 size 1064

checkpoints/checkpoint-500/special_tokens_map.json

@@ -0,0 +1,31 @@
+{
+  "additional_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|>"
+  ],
+  "eos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

checkpoints/checkpoint-500/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9c5ae00e602b8860cbd784ba82a8aa14e8feecec692e7076590d014d7b7fdafa
+size 11421896

checkpoints/checkpoint-500/tokenizer_config.json

@@ -0,0 +1,207 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "additional_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|>"
+  ],
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": {},
+  "model_max_length": 32768,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

checkpoints/checkpoint-500/trainer_state.json

@@ -1,1832 +1,1133 @@
 {
-  "best_global_step": 500,
-  "best_metric": 1.042096495628357,
-  "best_model_checkpoint": "task2file/sft_qwen_14B_v2/checkpoints/checkpoint-500",
-  "epoch": 0.2109704641350211,
-  "eval_steps": 100,
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.0,
+  "eval_steps": 50,
   "global_step": 500,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.0008438818565400844,
-      "grad_norm": 0.5386583805084229,
-      "learning_rate": 1.7574692442882248e-07,
-      "loss": 1.6941628456115723,
-      "step": 2
-    },
-    {
-      "epoch": 0.0016877637130801688,
-      "grad_norm": 0.5477277636528015,
-      "learning_rate": 5.272407732864675e-07,
-      "loss": 1.7132279872894287,
-      "step": 4
-    },
-    {
-      "epoch": 0.002531645569620253,
-      "grad_norm": 0.5390765070915222,
-      "learning_rate": 8.787346221441126e-07,
-      "loss": 1.641180396080017,
-      "step": 6
-    },
-    {
-      "epoch": 0.0033755274261603376,
-      "grad_norm": 0.5023683905601501,
-      "learning_rate": 1.2302284710017575e-06,
-      "loss": 1.5616240501403809,
-      "step": 8
-    },
-    {
-      "epoch": 0.004219409282700422,
-      "grad_norm": 0.4899154603481293,
-      "learning_rate": 1.5817223198594026e-06,
-      "loss": 1.572033405303955,
+      "epoch": 0.02002002002002002,
+      "grad_norm": 0.05460292845964432,
+      "grpo_mean_advantage": -1.3560057254835556e-07,
+      "grpo_mean_group_score": 0.5922331809997559,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 3.0318567496578908e-06,
+      "learning_rate": 8.000000000000001e-07,
+      "loss": 0.007,
+      "step": 5
+    },
+    {
+      "epoch": 0.04004004004004004,
+      "grad_norm": 0.0679207444190979,
+      "grpo_mean_advantage": 3.6619603633880615e-06,
+      "grpo_mean_group_score": 0.5561589002609253,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.6246918676188216e-05,
+      "learning_rate": 1.8000000000000001e-06,
+      "loss": 0.0107,
       "step": 10
     },
     {
-      "epoch": 0.005063291139240506,
-      "grad_norm": 0.5239788293838501,
-      "learning_rate": 1.9332161687170474e-06,
-      "loss": 1.6242921352386475,
-      "step": 12
-    },
-    {
-      "epoch": 0.00590717299578059,
-      "grad_norm": 0.5172926783561707,
-      "learning_rate": 2.2847100175746925e-06,
-      "loss": 1.6800041198730469,
-      "step": 14
-    },
-    {
-      "epoch": 0.006751054852320675,
-      "grad_norm": 0.5539224743843079,
-      "learning_rate": 2.6362038664323376e-06,
-      "loss": 1.6450834274291992,
-      "step": 16
-    },
-    {
-      "epoch": 0.007594936708860759,
-      "grad_norm": 0.5255337953567505,
-      "learning_rate": 2.9876977152899827e-06,
-      "loss": 1.6673263311386108,
-      "step": 18
-    },
-    {
-      "epoch": 0.008438818565400843,
-      "grad_norm": 0.5074548721313477,
-      "learning_rate": 3.3391915641476277e-06,
-      "loss": 1.531802773475647,
+      "epoch": 0.06006006006006006,
+      "grad_norm": 0.05788416787981987,
+      "grpo_mean_advantage": -1.0654330395709621e-07,
+      "grpo_mean_group_score": 0.5759152173995972,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 5.399440965447866e-07,
+      "learning_rate": 2.8000000000000003e-06,
+      "loss": 0.007,
+      "step": 15
+    },
+    {
+      "epoch": 0.08008008008008008,
+      "grad_norm": 0.0746568813920021,
+      "grpo_mean_advantage": -5.871057737749652e-07,
+      "grpo_mean_group_score": 0.5127314329147339,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 2.6951597646984737e-06,
+      "learning_rate": 3.8000000000000005e-06,
+      "loss": 0.0246,
       "step": 20
     },
     {
-      "epoch": 0.009282700421940928,
-      "grad_norm": 0.4160279333591461,
-      "learning_rate": 3.6906854130052724e-06,
-      "loss": 1.599354863166809,
-      "step": 22
-    },
-    {
-      "epoch": 0.010126582278481013,
-      "grad_norm": 0.5716474652290344,
-      "learning_rate": 4.0421792618629174e-06,
-      "loss": 1.6700962781906128,
-      "step": 24
-    },
-    {
-      "epoch": 0.010970464135021098,
-      "grad_norm": 0.5148899555206299,
-      "learning_rate": 4.3936731107205625e-06,
-      "loss": 1.66217839717865,
-      "step": 26
-    },
-    {
-      "epoch": 0.01181434599156118,
-      "grad_norm": 0.575722336769104,
-      "learning_rate": 4.7451669595782076e-06,
-      "loss": 1.6692266464233398,
-      "step": 28
-    },
-    {
-      "epoch": 0.012658227848101266,
-      "grad_norm": 0.5345953106880188,
-      "learning_rate": 5.096660808435853e-06,
-      "loss": 1.5518689155578613,
+      "epoch": 0.1001001001001001,
+      "grad_norm": 0.11442846059799194,
+      "grpo_mean_advantage": 6.370246410369873e-07,
+      "grpo_mean_group_score": 0.539706826210022,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 2.8908377771585947e-06,
+      "learning_rate": 4.800000000000001e-06,
+      "loss": 0.0337,
+      "step": 25
+    },
+    {
+      "epoch": 0.12012012012012012,
+      "grad_norm": 0.05778791010379791,
+      "grpo_mean_advantage": 6.705522359595761e-09,
+      "grpo_mean_group_score": 0.5812538862228394,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 6.189450800775376e-07,
+      "learning_rate": 4.999125183044924e-06,
+      "loss": 0.0171,
       "step": 30
     },
     {
-      "epoch": 0.01350210970464135,
-      "grad_norm": 0.4462043344974518,
-      "learning_rate": 5.448154657293498e-06,
-      "loss": 1.5930007696151733,
-      "step": 32
-    },
-    {
-      "epoch": 0.014345991561181435,
-      "grad_norm": 0.5119605660438538,
-      "learning_rate": 5.799648506151143e-06,
-      "loss": 1.6069684028625488,
-      "step": 34
-    },
-    {
-      "epoch": 0.015189873417721518,
-      "grad_norm": 0.5328608751296997,
-      "learning_rate": 6.151142355008788e-06,
-      "loss": 1.5838109254837036,
-      "step": 36
-    },
-    {
-      "epoch": 0.016033755274261603,
-      "grad_norm": 0.5065920352935791,
-      "learning_rate": 6.502636203866433e-06,
-      "loss": 1.608130931854248,
-      "step": 38
-    },
-    {
-      "epoch": 0.016877637130801686,
-      "grad_norm": 0.4479359984397888,
-      "learning_rate": 6.854130052724078e-06,
-      "loss": 1.5942182540893555,
+      "epoch": 0.14014014014014015,
+      "grad_norm": 0.05819695070385933,
+      "grpo_mean_advantage": 3.859400692363124e-07,
+      "grpo_mean_group_score": 0.5909844636917114,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.6833292875162442e-06,
+      "learning_rate": 4.995572288443412e-06,
+      "loss": 0.0145,
+      "step": 35
+    },
+    {
+      "epoch": 0.16016016016016016,
+      "grad_norm": 0.07968433201313019,
+      "grpo_mean_advantage": 2.600252742013254e-07,
+      "grpo_mean_group_score": 0.5630953907966614,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.4095899132371414e-06,
+      "learning_rate": 4.98929052218411e-06,
+      "loss": 0.0196,
       "step": 40
     },
     {
-      "epoch": 0.017721518987341773,
-      "grad_norm": 0.42844903469085693,
-      "learning_rate": 7.205623901581722e-06,
-      "loss": 1.6441553831100464,
-      "step": 42
-    },
-    {
-      "epoch": 0.018565400843881856,
-      "grad_norm": 0.476630836725235,
-      "learning_rate": 7.557117750439367e-06,
-      "loss": 1.6068111658096313,
-      "step": 44
-    },
-    {
-      "epoch": 0.019409282700421943,
-      "grad_norm": 0.4532654881477356,
-      "learning_rate": 7.908611599297012e-06,
-      "loss": 1.6618021726608276,
-      "step": 46
-    },
-    {
-      "epoch": 0.020253164556962026,
-      "grad_norm": 0.3701118230819702,
-      "learning_rate": 8.260105448154657e-06,
-      "loss": 1.4730033874511719,
-      "step": 48
-    },
-    {
-      "epoch": 0.02109704641350211,
-      "grad_norm": 0.38471561670303345,
-      "learning_rate": 8.611599297012302e-06,
-      "loss": 1.4828267097473145,
+      "epoch": 0.18018018018018017,
+      "grad_norm": 0.0733402892947197,
+      "grpo_mean_advantage": -1.2591480924584175e-07,
+      "grpo_mean_group_score": 0.5604403614997864,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.0309080380466185e-06,
+      "learning_rate": 4.980286753286196e-06,
+      "loss": 0.0186,
+      "step": 45
+    },
+    {
+      "epoch": 0.2002002002002002,
+      "grad_norm": 0.07136482000350952,
+      "grpo_mean_advantage": -2.808868941883702e-07,
+      "grpo_mean_group_score": 0.5971035957336426,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.5696078889959608e-06,
+      "learning_rate": 4.9685708272387645e-06,
+      "loss": 0.0286,
       "step": 50
     },
     {
-      "epoch": 0.021940928270042195,
-      "grad_norm": 0.3602336347103119,
-      "learning_rate": 8.963093145869948e-06,
-      "loss": 1.3877452611923218,
-      "step": 52
-    },
-    {
-      "epoch": 0.02278481012658228,
-      "grad_norm": 0.40318572521209717,
-      "learning_rate": 9.314586994727593e-06,
-      "loss": 1.49052894115448,
-      "step": 54
-    },
-    {
-      "epoch": 0.02362869198312236,
-      "grad_norm": 0.3223826587200165,
-      "learning_rate": 9.666080843585238e-06,
-      "loss": 1.4912524223327637,
-      "step": 56
-    },
-    {
-      "epoch": 0.024472573839662448,
-      "grad_norm": 0.3873065114021301,
-      "learning_rate": 1.0017574692442883e-05,
-      "loss": 1.526674509048462,
-      "step": 58
-    },
-    {
-      "epoch": 0.02531645569620253,
-      "grad_norm": 0.410159707069397,
-      "learning_rate": 1.0369068541300528e-05,
-      "loss": 1.4480271339416504,
+      "epoch": 0.22022022022022023,
+      "grad_norm": 0.08851475268602371,
+      "grpo_mean_advantage": 2.6822089438383045e-08,
+      "grpo_mean_group_score": 0.5892971754074097,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 3.7878271541558206e-07,
+      "learning_rate": 4.9541555552349404e-06,
+      "loss": 0.0054,
+      "step": 55
+    },
+    {
+      "epoch": 0.24024024024024024,
+      "grad_norm": 0.07778509706258774,
+      "grpo_mean_advantage": -5.662441182607836e-08,
+      "grpo_mean_group_score": 0.564322292804718,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 6.128998393251095e-07,
+      "learning_rate": 4.9370567001630155e-06,
+      "loss": -0.0074,
       "step": 60
     },
     {
-      "epoch": 0.026160337552742614,
-      "grad_norm": 0.3632003962993622,
-      "learning_rate": 1.0720562390158173e-05,
-      "loss": 1.4222990274429321,
-      "step": 62
-    },
-    {
-      "epoch": 0.0270042194092827,
-      "grad_norm": 0.33118435740470886,
-      "learning_rate": 1.1072056239015818e-05,
-      "loss": 1.387171745300293,
-      "step": 64
-    },
-    {
-      "epoch": 0.027848101265822784,
-      "grad_norm": 0.3301764726638794,
-      "learning_rate": 1.1423550087873463e-05,
-      "loss": 1.3523777723312378,
-      "step": 66
-    },
-    {
-      "epoch": 0.02869198312236287,
-      "grad_norm": 0.34342435002326965,
-      "learning_rate": 1.1775043936731108e-05,
-      "loss": 1.4515162706375122,
-      "step": 68
-    },
-    {
-      "epoch": 0.029535864978902954,
-      "grad_norm": 0.3243122100830078,
-      "learning_rate": 1.2126537785588753e-05,
-      "loss": 1.3509243726730347,
+      "epoch": 0.2602602602602603,
+      "grad_norm": 0.08740051090717316,
+      "grpo_mean_advantage": -1.5944242193199898e-07,
+      "grpo_mean_group_score": 0.562497615814209,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.6374274309782777e-06,
+      "learning_rate": 4.917292959369968e-06,
+      "loss": 0.0145,
+      "step": 65
+    },
+    {
+      "epoch": 0.2802802802802803,
+      "grad_norm": 0.19070060551166534,
+      "grpo_mean_advantage": 1.6838312433264946e-07,
+      "grpo_mean_group_score": 0.5904761552810669,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 8.536571272088622e-07,
+      "learning_rate": 4.8948859442161876e-06,
+      "loss": 0.0257,
       "step": 70
     },
     {
-      "epoch": 0.030379746835443037,
-      "grad_norm": 0.3450150787830353,
-      "learning_rate": 1.2478031634446398e-05,
-      "loss": 1.4936245679855347,
-      "step": 72
-    },
-    {
-      "epoch": 0.031223628691983123,
-      "grad_norm": 0.38912028074264526,
-      "learning_rate": 1.2829525483304042e-05,
-      "loss": 1.3419109582901,
-      "step": 74
-    },
-    {
-      "epoch": 0.032067510548523206,
-      "grad_norm": 0.3019310235977173,
-      "learning_rate": 1.3181019332161687e-05,
-      "loss": 1.4284154176712036,
-      "step": 76
-    },
-    {
-      "epoch": 0.03291139240506329,
-      "grad_norm": 0.37803682684898376,
-      "learning_rate": 1.3532513181019332e-05,
-      "loss": 1.4256561994552612,
-      "step": 78
-    },
-    {
-      "epoch": 0.03375527426160337,
-      "grad_norm": 0.34191736578941345,
-      "learning_rate": 1.3884007029876977e-05,
-      "loss": 1.3256909847259521,
+      "epoch": 0.3003003003003003,
+      "grad_norm": 0.07321271300315857,
+      "grpo_mean_advantage": 1.1175870895385742e-07,
+      "grpo_mean_group_score": 0.5765624046325684,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 6.451961667153228e-07,
+      "learning_rate": 4.869860156443768e-06,
+      "loss": 0.0024,
+      "step": 75
+    },
+    {
+      "epoch": 0.3203203203203203,
+      "grad_norm": 0.07126748561859131,
+      "grpo_mean_advantage": -1.4603138254187797e-07,
+      "grpo_mean_group_score": 0.5858271718025208,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.1309343790344428e-06,
+      "learning_rate": 4.842242961384211e-06,
+      "loss": 0.0277,
       "step": 80
     },
     {
-      "epoch": 0.03459915611814346,
-      "grad_norm": 0.35242700576782227,
-      "learning_rate": 1.4235500878734624e-05,
-      "loss": 1.2710685729980469,
-      "step": 82
-    },
-    {
-      "epoch": 0.035443037974683546,
-      "grad_norm": 0.38094228506088257,
-      "learning_rate": 1.4586994727592269e-05,
-      "loss": 1.253411889076233,
-      "step": 84
-    },
-    {
-      "epoch": 0.036286919831223625,
-      "grad_norm": 0.36837366223335266,
-      "learning_rate": 1.4938488576449914e-05,
-      "loss": 1.3064342737197876,
-      "step": 86
-    },
-    {
-      "epoch": 0.03713080168776371,
-      "grad_norm": 0.3443569242954254,
-      "learning_rate": 1.5289982425307557e-05,
-      "loss": 1.293562412261963,
-      "step": 88
-    },
-    {
-      "epoch": 0.0379746835443038,
-      "grad_norm": 0.3799338936805725,
-      "learning_rate": 1.5641476274165202e-05,
-      "loss": 1.3382648229599,
+      "epoch": 0.34034034034034033,
+      "grad_norm": 0.08629189431667328,
+      "grpo_mean_advantage": -1.817941665649414e-06,
+      "grpo_mean_group_score": 0.5871662497520447,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.1141768482048064e-05,
+      "learning_rate": 4.812064558034847e-06,
+      "loss": 0.0246,
+      "step": 85
+    },
+    {
+      "epoch": 0.36036036036036034,
+      "grad_norm": 0.0998779758810997,
+      "grpo_mean_advantage": 1.8179416372277046e-07,
+      "grpo_mean_group_score": 0.5330992937088013,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 6.210335072864837e-07,
+      "learning_rate": 4.779357946036662e-06,
+      "loss": 0.0056,
       "step": 90
     },
     {
-      "epoch": 0.038818565400843885,
-      "grad_norm": 0.40501922369003296,
-      "learning_rate": 1.599297012302285e-05,
-      "loss": 1.3925724029541016,
-      "step": 92
-    },
-    {
-      "epoch": 0.039662447257383965,
-      "grad_norm": 0.4419630467891693,
-      "learning_rate": 1.6344463971880492e-05,
-      "loss": 1.357171893119812,
-      "step": 94
-    },
-    {
-      "epoch": 0.04050632911392405,
-      "grad_norm": 0.3619817793369293,
-      "learning_rate": 1.6695957820738137e-05,
-      "loss": 1.3029985427856445,
-      "step": 96
-    },
-    {
-      "epoch": 0.04135021097046414,
-      "grad_norm": 0.4851357340812683,
-      "learning_rate": 1.7047451669595782e-05,
-      "loss": 1.3498191833496094,
-      "step": 98
-    },
-    {
-      "epoch": 0.04219409282700422,
-      "grad_norm": 0.418658584356308,
-      "learning_rate": 1.7398945518453427e-05,
-      "loss": 1.185287356376648,
-      "step": 100
-    },
-    {
-      "epoch": 0.04219409282700422,
-      "eval_loss": 1.2979938983917236,
-      "eval_runtime": 682.1979,
-      "eval_samples_per_second": 3.089,
-      "eval_steps_per_second": 3.089,
+      "epoch": 0.38038038038038036,
+      "grad_norm": 0.10614689439535141,
+      "grpo_mean_advantage": -2.972781771859445e-07,
+      "grpo_mean_group_score": 0.5265295505523682,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 3.1582342217006953e-06,
+      "learning_rate": 4.74415888958968e-06,
+      "loss": 0.0053,
+      "step": 95
+    },
+    {
+      "epoch": 0.4004004004004004,
+      "grad_norm": 0.10345634073019028,
+      "grpo_mean_advantage": -7.033348197182931e-07,
+      "grpo_mean_group_score": 0.5660771131515503,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 4.245831405569334e-06,
+      "learning_rate": 4.706505878345343e-06,
+      "loss": 0.0134,
       "step": 100
     },
     {
-      "epoch": 0.043037974683544304,
-      "grad_norm": 0.4464418888092041,
-      "learning_rate": 1.7750439367311073e-05,
-      "loss": 1.2217272520065308,
-      "step": 102
-    },
-    {
-      "epoch": 0.04388185654008439,
-      "grad_norm": 0.4706237316131592,
-      "learning_rate": 1.8101933216168718e-05,
-      "loss": 1.2052050828933716,
-      "step": 104
-    },
-    {
-      "epoch": 0.04472573839662447,
-      "grad_norm": 0.46394404768943787,
-      "learning_rate": 1.8453427065026363e-05,
-      "loss": 1.221343994140625,
-      "step": 106
-    },
-    {
-      "epoch": 0.04556962025316456,
-      "grad_norm": 0.4726889431476593,
-      "learning_rate": 1.8804920913884008e-05,
-      "loss": 1.2387475967407227,
-      "step": 108
-    },
-    {
-      "epoch": 0.046413502109704644,
-      "grad_norm": 0.42130985856056213,
-      "learning_rate": 1.9156414762741653e-05,
-      "loss": 1.2851309776306152,
+      "epoch": 0.42042042042042044,
+      "grad_norm": 0.10077933222055435,
+      "grpo_mean_advantage": 1.1920928955078125e-07,
+      "grpo_mean_group_score": 0.57631915807724,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 3.2809634831210133e-07,
+      "learning_rate": 4.666440085318626e-06,
+      "loss": 0.0004,
+      "step": 105
+    },
+    {
+      "epoch": 0.44044044044044045,
+      "grad_norm": 0.09548182785511017,
+      "grpo_mean_advantage": -4.0978193283081055e-07,
+      "grpo_mean_group_score": 0.546563982963562,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 6.0397578636184335e-06,
+      "learning_rate": 4.624005321865968e-06,
+      "loss": 0.0033,
       "step": 110
     },
     {
-      "epoch": 0.04725738396624472,
-      "grad_norm": 0.4504576623439789,
-      "learning_rate": 1.9507908611599298e-05,
-      "loss": 1.2753145694732666,
-      "step": 112
-    },
-    {
-      "epoch": 0.04810126582278481,
-      "grad_norm": 0.396085262298584,
-      "learning_rate": 1.9859402460456943e-05,
-      "loss": 1.2427717447280884,
-      "step": 114
-    },
-    {
-      "epoch": 0.048945147679324896,
-      "grad_norm": 0.5106491446495056,
-      "learning_rate": 2.0210896309314588e-05,
-      "loss": 1.2943825721740723,
-      "step": 116
-    },
-    {
-      "epoch": 0.049789029535864976,
-      "grad_norm": 0.42351317405700684,
-      "learning_rate": 2.0562390158172233e-05,
-      "loss": 1.263301134109497,
-      "step": 118
-    },
-    {
-      "epoch": 0.05063291139240506,
-      "grad_norm": 0.4403539299964905,
-      "learning_rate": 2.0913884007029878e-05,
-      "loss": 1.2647849321365356,
+      "epoch": 0.46046046046046046,
+      "grad_norm": 0.09417816251516342,
+      "grpo_mean_advantage": -1.467764434437413e-07,
+      "grpo_mean_group_score": 0.5519219636917114,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 2.2689375782647403e-06,
+      "learning_rate": 4.57924798977818e-06,
+      "loss": 0.0095,
+      "step": 115
+    },
+    {
+      "epoch": 0.4804804804804805,
+      "grad_norm": 0.10022275149822235,
+      "grpo_mean_advantage": -5.215406329028838e-09,
+      "grpo_mean_group_score": 0.5490407943725586,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 7.929010621410271e-07,
+      "learning_rate": 4.532217030540781e-06,
+      "loss": 0.0006,
       "step": 120
     },
     {
-      "epoch": 0.05147679324894515,
-      "grad_norm": 0.5260752439498901,
-      "learning_rate": 2.1265377855887523e-05,
-      "loss": 1.2351393699645996,
-      "step": 122
-    },
-    {
-      "epoch": 0.05232067510548523,
-      "grad_norm": 0.44978851079940796,
-      "learning_rate": 2.1616871704745168e-05,
-      "loss": 1.0384471416473389,
-      "step": 124
-    },
-    {
-      "epoch": 0.053164556962025315,
-      "grad_norm": 0.47732362151145935,
-      "learning_rate": 2.1968365553602813e-05,
-      "loss": 1.1518068313598633,
-      "step": 126
-    },
-    {
-      "epoch": 0.0540084388185654,
-      "grad_norm": 0.5473551750183105,
-      "learning_rate": 2.231985940246046e-05,
-      "loss": 1.2264912128448486,
-      "step": 128
-    },
-    {
-      "epoch": 0.05485232067510549,
-      "grad_norm": 0.4473855197429657,
-      "learning_rate": 2.2671353251318103e-05,
-      "loss": 1.1615246534347534,
+      "epoch": 0.5005005005005005,
+      "grad_norm": 0.14057794213294983,
+      "grpo_mean_advantage": -5.7369469175228005e-08,
+      "grpo_mean_group_score": 0.5646580457687378,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.2823379620385822e-06,
+      "learning_rate": 4.482963871817195e-06,
+      "loss": -0.0046,
+      "step": 125
+    },
+    {
+      "epoch": 0.5205205205205206,
+      "grad_norm": 0.12420658767223358,
+      "grpo_mean_advantage": 2.9876827056796174e-07,
+      "grpo_mean_group_score": 0.6111599802970886,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.0496698905626545e-06,
+      "learning_rate": 4.4315423712133595e-06,
+      "loss": -0.003,
       "step": 130
     },
     {
-      "epoch": 0.05569620253164557,
-      "grad_norm": 0.5980377197265625,
-      "learning_rate": 2.302284710017575e-05,
-      "loss": 1.1334880590438843,
-      "step": 132
-    },
-    {
-      "epoch": 0.056540084388185655,
-      "grad_norm": 0.5987792015075684,
-      "learning_rate": 2.3374340949033394e-05,
-      "loss": 1.1546804904937744,
-      "step": 134
-    },
-    {
-      "epoch": 0.05738396624472574,
-      "grad_norm": 0.45355498790740967,
-      "learning_rate": 2.372583479789104e-05,
-      "loss": 1.194953441619873,
-      "step": 136
-    },
-    {
-      "epoch": 0.05822784810126582,
-      "grad_norm": 0.5373698472976685,
-      "learning_rate": 2.4077328646748684e-05,
-      "loss": 1.1067466735839844,
-      "step": 138
-    },
-    {
-      "epoch": 0.05907172995780591,
-      "grad_norm": 0.48734328150749207,
-      "learning_rate": 2.442882249560633e-05,
-      "loss": 1.188468098640442,
+      "epoch": 0.5405405405405406,
+      "grad_norm": 0.14342808723449707,
+      "grpo_mean_advantage": 1.5869736103013565e-07,
+      "grpo_mean_group_score": 0.5619662404060364,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.2748531617035042e-06,
+      "learning_rate": 4.378008757385222e-06,
+      "loss": 0.0154,
+      "step": 135
+    },
+    {
+      "epoch": 0.5605605605605606,
+      "grad_norm": 0.14729444682598114,
+      "grpo_mean_advantage": 3.0100346748440643e-07,
+      "grpo_mean_group_score": 0.5795454978942871,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 2.4499684059264837e-06,
+      "learning_rate": 4.322421568553529e-06,
+      "loss": -0.0262,
       "step": 140
     },
     {
-      "epoch": 0.059915611814345994,
-      "grad_norm": 0.4692173898220062,
-      "learning_rate": 2.478031634446397e-05,
-      "loss": 1.1624362468719482,
-      "step": 142
-    },
-    {
-      "epoch": 0.060759493670886074,
-      "grad_norm": 0.532554030418396,
-      "learning_rate": 2.513181019332162e-05,
-      "loss": 1.0978907346725464,
-      "step": 144
-    },
-    {
-      "epoch": 0.06160337552742616,
-      "grad_norm": 0.5853802561759949,
-      "learning_rate": 2.5483304042179264e-05,
-      "loss": 1.2030781507492065,
-      "step": 146
-    },
-    {
-      "epoch": 0.06244725738396625,
-      "grad_norm": 0.5061611533164978,
-      "learning_rate": 2.583479789103691e-05,
-      "loss": 1.082366943359375,
-      "step": 148
-    },
-    {
-      "epoch": 0.06329113924050633,
-      "grad_norm": 0.49426141381263733,
-      "learning_rate": 2.6186291739894554e-05,
-      "loss": 1.10564386844635,
+      "epoch": 0.5805805805805806,
+      "grad_norm": 0.15249410271644592,
+      "grpo_mean_advantage": -3.233552092751779e-07,
+      "grpo_mean_group_score": 0.5804953575134277,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.248456669600273e-06,
+      "learning_rate": 4.2648415884931476e-06,
+      "loss": 0.0018,
+      "step": 145
+    },
+    {
+      "epoch": 0.6006006006006006,
+      "grad_norm": 0.1841023564338684,
+      "grpo_mean_advantage": 3.2261013416245987e-07,
+      "grpo_mean_group_score": 0.5628539323806763,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.4773489738217904e-06,
+      "learning_rate": 4.205331780066892e-06,
+      "loss": -0.017,
       "step": 150
     },
     {
-      "epoch": 0.06413502109704641,
-      "grad_norm": 0.5846618413925171,
-      "learning_rate": 2.6537785588752196e-05,
-      "loss": 1.1992807388305664,
-      "step": 152
-    },
-    {
-      "epoch": 0.06497890295358649,
-      "grad_norm": 0.5517552495002747,
-      "learning_rate": 2.6889279437609844e-05,
-      "loss": 1.1757566928863525,
-      "step": 154
-    },
-    {
-      "epoch": 0.06582278481012659,
-      "grad_norm": 0.5667305588722229,
-      "learning_rate": 2.7240773286467486e-05,
-      "loss": 1.0548783540725708,
-      "step": 156
-    },
-    {
-      "epoch": 0.06666666666666667,
-      "grad_norm": 0.6760414242744446,
-      "learning_rate": 2.7592267135325134e-05,
-      "loss": 1.184364914894104,
-      "step": 158
-    },
-    {
-      "epoch": 0.06751054852320675,
-      "grad_norm": 0.5261430740356445,
-      "learning_rate": 2.7943760984182776e-05,
-      "loss": 1.1945042610168457,
+      "epoch": 0.6206206206206206,
+      "grad_norm": 0.18597163259983063,
+      "grpo_mean_advantage": -2.5331974029541016e-07,
+      "grpo_mean_group_score": 0.5727725625038147,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.5092309695319273e-06,
+      "learning_rate": 4.1439572163765615e-06,
+      "loss": 0.0044,
+      "step": 155
+    },
+    {
+      "epoch": 0.6406406406406406,
+      "grad_norm": 0.18310388922691345,
+      "grpo_mean_advantage": -6.780028627417778e-08,
+      "grpo_mean_group_score": 0.5833909511566162,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 8.550978805033083e-07,
+      "learning_rate": 4.0807850096064605e-06,
+      "loss": -0.005,
       "step": 160
     },
     {
-      "epoch": 0.06835443037974684,
-      "grad_norm": 0.6155015230178833,
-      "learning_rate": 2.8295254833040425e-05,
-      "loss": 1.2021973133087158,
-      "step": 162
-    },
-    {
-      "epoch": 0.06919831223628692,
-      "grad_norm": 0.6131619215011597,
-      "learning_rate": 2.8646748681898066e-05,
-      "loss": 1.144123911857605,
-      "step": 164
-    },
-    {
-      "epoch": 0.070042194092827,
-      "grad_norm": 0.5749185681343079,
-      "learning_rate": 2.8998242530755715e-05,
-      "loss": 1.1329256296157837,
-      "step": 166
-    },
-    {
-      "epoch": 0.07088607594936709,
-      "grad_norm": 0.5243118405342102,
-      "learning_rate": 2.9349736379613356e-05,
-      "loss": 1.0892387628555298,
-      "step": 168
-    },
-    {
-      "epoch": 0.07172995780590717,
-      "grad_norm": 0.7190104722976685,
-      "learning_rate": 2.9701230228471005e-05,
-      "loss": 1.163260817527771,
+      "epoch": 0.6606606606606606,
+      "grad_norm": 0.2192923128604889,
+      "grpo_mean_advantage": -5.587935447692871e-08,
+      "grpo_mean_group_score": 0.5742615461349487,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 3.564579174053506e-07,
+      "learning_rate": 4.015884237637206e-06,
+      "loss": -0.015,
+      "step": 165
+    },
+    {
+      "epoch": 0.6806806806806807,
+      "grad_norm": 0.16708803176879883,
+      "grpo_mean_advantage": -5.327165126800537e-07,
+      "grpo_mean_group_score": 0.5758188962936401,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 2.309018327650847e-06,
+      "learning_rate": 3.949325868510083e-06,
+      "loss": -0.0314,
       "step": 170
     },
     {
-      "epoch": 0.07257383966244725,
-      "grad_norm": 0.5486982464790344,
-      "learning_rate": 3.0052724077328647e-05,
-      "loss": 1.0880777835845947,
-      "step": 172
-    },
-    {
-      "epoch": 0.07341772151898734,
-      "grad_norm": 0.5020889043807983,
-      "learning_rate": 3.0404217926186295e-05,
-      "loss": 1.0433368682861328,
-      "step": 174
-    },
-    {
-      "epoch": 0.07426160337552742,
-      "grad_norm": 0.47329774498939514,
-      "learning_rate": 3.075571177504394e-05,
-      "loss": 1.0528991222381592,
-      "step": 176
-    },
-    {
-      "epoch": 0.0751054852320675,
-      "grad_norm": 0.6635547876358032,
-      "learning_rate": 3.110720562390158e-05,
-      "loss": 1.1627811193466187,
-      "step": 178
-    },
-    {
-      "epoch": 0.0759493670886076,
-      "grad_norm": 0.5624618530273438,
-      "learning_rate": 3.145869947275923e-05,
-      "loss": 1.084869384765625,
+      "epoch": 0.7007007007007007,
+      "grad_norm": 0.3401262164115906,
+      "grpo_mean_advantage": 5.863606702405377e-07,
+      "grpo_mean_group_score": 0.5767683982849121,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 2.4449204829579685e-06,
+      "learning_rate": 3.881182682824534e-06,
+      "loss": -0.0441,
+      "step": 175
+    },
+    {
+      "epoch": 0.7207207207207207,
+      "grad_norm": 0.1931898146867752,
+      "grpo_mean_advantage": 3.2186508747145126e-07,
+      "grpo_mean_group_score": 0.586772084236145,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 2.293551688126172e-06,
+      "learning_rate": 3.811529194153635e-06,
+      "loss": -0.0162,
       "step": 180
     },
     {
-      "epoch": 0.07679324894514768,
-      "grad_norm": 0.6029536724090576,
-      "learning_rate": 3.181019332161687e-05,
-      "loss": 1.2227671146392822,
-      "step": 182
-    },
-    {
-      "epoch": 0.07763713080168777,
-      "grad_norm": 0.930959939956665,
-      "learning_rate": 3.216168717047452e-05,
-      "loss": 1.0955452919006348,
-      "step": 184
-    },
-    {
-      "epoch": 0.07848101265822785,
-      "grad_norm": 0.5326952338218689,
-      "learning_rate": 3.251318101933216e-05,
-      "loss": 1.0640798807144165,
-      "step": 186
-    },
-    {
-      "epoch": 0.07932489451476793,
-      "grad_norm": 0.5484727621078491,
-      "learning_rate": 3.286467486818981e-05,
-      "loss": 1.0700589418411255,
-      "step": 188
-    },
-    {
-      "epoch": 0.08016877637130802,
-      "grad_norm": 0.605273425579071,
-      "learning_rate": 3.3216168717047456e-05,
-      "loss": 1.1593081951141357,
+      "epoch": 0.7407407407407407,
+      "grad_norm": 0.2537969648838043,
+      "grpo_mean_advantage": -4.470348358154297e-08,
+      "grpo_mean_group_score": 0.549396276473999,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 3.7067667335577426e-07,
+      "learning_rate": 3.7404415675646054e-06,
+      "loss": -0.0386,
+      "step": 185
+    },
+    {
+      "epoch": 0.7607607607607607,
+      "grad_norm": 0.20326584577560425,
+      "grpo_mean_advantage": -2.1010637851759384e-07,
+      "grpo_mean_group_score": 0.5798425078392029,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.1695076409523608e-06,
+      "learning_rate": 3.667997536333424e-06,
+      "loss": -0.037,
       "step": 190
     },
     {
-      "epoch": 0.0810126582278481,
-      "grad_norm": 0.5704394578933716,
-      "learning_rate": 3.35676625659051e-05,
-      "loss": 1.1617076396942139,
-      "step": 192
-    },
-    {
-      "epoch": 0.08185654008438818,
-      "grad_norm": 0.5929452180862427,
-      "learning_rate": 3.3919156414762746e-05,
-      "loss": 1.1346839666366577,
-      "step": 194
-    },
-    {
-      "epoch": 0.08270042194092828,
-      "grad_norm": 0.5624077916145325,
-      "learning_rate": 3.427065026362039e-05,
-      "loss": 1.0934710502624512,
-      "step": 196
-    },
-    {
-      "epoch": 0.08354430379746836,
-      "grad_norm": 0.6717425584793091,
-      "learning_rate": 3.4622144112478036e-05,
-      "loss": 1.1810534000396729,
-      "step": 198
-    },
-    {
-      "epoch": 0.08438818565400844,
-      "grad_norm": 0.5120199918746948,
-      "learning_rate": 3.4973637961335674e-05,
-      "loss": 1.1525514125823975,
+      "epoch": 0.7807807807807807,
+      "grad_norm": 0.25048357248306274,
+      "grpo_mean_advantage": 1.765787658314366e-07,
+      "grpo_mean_group_score": 0.5584167838096619,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 2.429934738756856e-06,
+      "learning_rate": 3.59427631694463e-06,
+      "loss": -0.0292,
+      "step": 195
+    },
+    {
+      "epoch": 0.8008008008008008,
+      "grad_norm": 0.2687569260597229,
+      "grpo_mean_advantage": 1.6540289493605087e-07,
+      "grpo_mean_group_score": 0.5676193237304688,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 2.6342788714828203e-06,
+      "learning_rate": 3.5193585224692595e-06,
+      "loss": -0.0454,
       "step": 200
     },
     {
-      "epoch": 0.08438818565400844,
-      "eval_loss": 1.142486810684204,
-      "eval_runtime": 668.2356,
-      "eval_samples_per_second": 3.153,
-      "eval_steps_per_second": 3.153,
-      "step": 200
-    },
-    {
-      "epoch": 0.08523206751054853,
-      "grad_norm": 0.5144487023353577,
-      "learning_rate": 3.5325131810193326e-05,
-      "loss": 1.0243735313415527,
-      "step": 202
-    },
-    {
-      "epoch": 0.08607594936708861,
-      "grad_norm": 0.6325069069862366,
-      "learning_rate": 3.5676625659050964e-05,
-      "loss": 1.118743896484375,
-      "step": 204
-    },
-    {
-      "epoch": 0.08691983122362869,
-      "grad_norm": 0.5501633882522583,
-      "learning_rate": 3.6028119507908616e-05,
-      "loss": 1.0380504131317139,
-      "step": 206
-    },
-    {
-      "epoch": 0.08776371308016878,
-      "grad_norm": 0.6133899688720703,
-      "learning_rate": 3.6379613356766254e-05,
-      "loss": 0.9837555885314941,
-      "step": 208
-    },
-    {
-      "epoch": 0.08860759493670886,
-      "grad_norm": 0.5799810886383057,
-      "learning_rate": 3.6731107205623906e-05,
-      "loss": 1.090720295906067,
+      "epoch": 0.8208208208208209,
+      "grad_norm": 0.22301620244979858,
+      "grpo_mean_advantage": -1.0944902442133753e-06,
+      "grpo_mean_group_score": 0.5669739842414856,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 5.346942998585291e-06,
+      "learning_rate": 3.44332607441564e-06,
+      "loss": -0.0423,
+      "step": 205
+    },
+    {
+      "epoch": 0.8408408408408409,
+      "grad_norm": 0.3040211498737335,
+      "grpo_mean_advantage": 2.4065374759629776e-07,
+      "grpo_mean_group_score": 0.5922158360481262,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.6327536513927043e-06,
+      "learning_rate": 3.3662621131494204e-06,
+      "loss": -0.0857,
       "step": 210
     },
     {
-      "epoch": 0.08945147679324894,
-      "grad_norm": 0.6039511561393738,
-      "learning_rate": 3.7082601054481544e-05,
-      "loss": 1.120232343673706,
-      "step": 212
-    },
-    {
-      "epoch": 0.09029535864978903,
-      "grad_norm": 0.5983024835586548,
-      "learning_rate": 3.7434094903339196e-05,
-      "loss": 1.096949815750122,
-      "step": 214
-    },
-    {
-      "epoch": 0.09113924050632911,
-      "grad_norm": 0.5641079545021057,
-      "learning_rate": 3.7785588752196835e-05,
-      "loss": 1.1226298809051514,
-      "step": 216
-    },
-    {
-      "epoch": 0.0919831223628692,
-      "grad_norm": 0.655717134475708,
-      "learning_rate": 3.8137082601054486e-05,
-      "loss": 1.1260643005371094,
-      "step": 218
-    },
-    {
-      "epoch": 0.09282700421940929,
-      "grad_norm": 0.6111898422241211,
-      "learning_rate": 3.848857644991213e-05,
-      "loss": 1.0777709484100342,
+      "epoch": 0.8608608608608609,
+      "grad_norm": 0.27231141924858093,
+      "grpo_mean_advantage": -5.21540641784668e-08,
+      "grpo_mean_group_score": 0.5473950505256653,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 5.847922466273303e-07,
+      "learning_rate": 3.2882509069808044e-06,
+      "loss": -0.0278,
+      "step": 215
+    },
+    {
+      "epoch": 0.8808808808808809,
+      "grad_norm": 0.3571636378765106,
+      "grpo_mean_advantage": 6.541609991472797e-07,
+      "grpo_mean_group_score": 0.5880032777786255,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 4.072162937518442e-06,
+      "learning_rate": 3.2093777600183873e-06,
+      "loss": -0.0727,
       "step": 220
     },
     {
-      "epoch": 0.09367088607594937,
-      "grad_norm": 0.6821302771568298,
-      "learning_rate": 3.884007029876977e-05,
-      "loss": 1.10588800907135,
-      "step": 222
-    },
-    {
-      "epoch": 0.09451476793248945,
-      "grad_norm": 0.693175733089447,
-      "learning_rate": 3.919156414762742e-05,
-      "loss": 1.1498671770095825,
-      "step": 224
-    },
-    {
-      "epoch": 0.09535864978902954,
-      "grad_norm": 0.5288166403770447,
-      "learning_rate": 3.954305799648506e-05,
-      "loss": 1.0587562322616577,
-      "step": 226
-    },
-    {
-      "epoch": 0.09620253164556962,
-      "grad_norm": 0.6882867813110352,
-      "learning_rate": 3.989455184534271e-05,
-      "loss": 1.1107512712478638,
-      "step": 228
-    },
-    {
-      "epoch": 0.0970464135021097,
-      "grad_norm": 0.5834154486656189,
-      "learning_rate": 4.024604569420035e-05,
-      "loss": 1.020510196685791,
+      "epoch": 0.9009009009009009,
+      "grad_norm": 0.306273490190506,
+      "grpo_mean_advantage": -1.2218951894737984e-07,
+      "grpo_mean_group_score": 0.5835092663764954,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 4.386006935419573e-07,
+      "learning_rate": 3.1297289188903705e-06,
+      "loss": -0.0464,
+      "step": 225
+    },
+    {
+      "epoch": 0.9209209209209209,
+      "grad_norm": 0.2700377106666565,
+      "grpo_mean_advantage": 1.7605722177904681e-06,
+      "grpo_mean_group_score": 0.5394966006278992,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 8.007580618141219e-06,
+      "learning_rate": 3.049391478435133e-06,
+      "loss": -0.0295,
       "step": 230
     },
     {
-      "epoch": 0.09789029535864979,
-      "grad_norm": 0.7157064080238342,
-      "learning_rate": 4.0597539543058e-05,
-      "loss": 1.0642449855804443,
-      "step": 232
-    },
-    {
-      "epoch": 0.09873417721518987,
-      "grad_norm": 0.6530708074569702,
-      "learning_rate": 4.094903339191564e-05,
-      "loss": 1.0359872579574585,
-      "step": 234
-    },
-    {
-      "epoch": 0.09957805907172995,
-      "grad_norm": 0.6329686045646667,
-      "learning_rate": 4.130052724077329e-05,
-      "loss": 1.050504446029663,
-      "step": 236
-    },
-    {
-      "epoch": 0.10042194092827005,
-      "grad_norm": 0.6597026586532593,
-      "learning_rate": 4.165202108963093e-05,
-      "loss": 1.2621175050735474,
-      "step": 238
-    },
-    {
-      "epoch": 0.10126582278481013,
-      "grad_norm": 0.6195225119590759,
-      "learning_rate": 4.200351493848858e-05,
-      "loss": 1.1218310594558716,
+      "epoch": 0.9409409409409409,
+      "grad_norm": 0.39531761407852173,
+      "grpo_mean_advantage": -3.3080578987210174e-07,
+      "grpo_mean_group_score": 0.5687432289123535,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.551636614749441e-06,
+      "learning_rate": 2.9684532864643123e-06,
+      "loss": -0.031,
+      "step": 235
+    },
+    {
+      "epoch": 0.960960960960961,
+      "grad_norm": 0.5987040996551514,
+      "grpo_mean_advantage": 2.712011450967111e-07,
+      "grpo_mean_group_score": 0.5550583600997925,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.4400844747797237e-06,
+      "learning_rate": 2.887002847702504e-06,
+      "loss": -0.0789,
       "step": 240
     },
     {
-      "epoch": 0.1021097046413502,
-      "grad_norm": 0.6764137744903564,
-      "learning_rate": 4.235500878734622e-05,
-      "loss": 1.1250728368759155,
-      "step": 242
-    },
-    {
-      "epoch": 0.1029535864978903,
-      "grad_norm": 0.552363395690918,
-      "learning_rate": 4.270650263620387e-05,
-      "loss": 1.028212308883667,
-      "step": 244
-    },
-    {
-      "epoch": 0.10379746835443038,
-      "grad_norm": 0.5620495676994324,
-      "learning_rate": 4.305799648506151e-05,
-      "loss": 1.0425450801849365,
-      "step": 246
-    },
-    {
-      "epoch": 0.10464135021097046,
-      "grad_norm": 0.6860032081604004,
-      "learning_rate": 4.3409490333919156e-05,
-      "loss": 1.144278883934021,
-      "step": 248
-    },
-    {
-      "epoch": 0.10548523206751055,
-      "grad_norm": 0.6033259034156799,
-      "learning_rate": 4.37609841827768e-05,
-      "loss": 1.1223982572555542,
+      "epoch": 0.980980980980981,
+      "grad_norm": 0.5680716037750244,
+      "grpo_mean_advantage": -3.2857059295565705e-07,
+      "grpo_mean_group_score": 0.558111310005188,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 2.105091425619321e-06,
+      "learning_rate": 2.8051292270086506e-06,
+      "loss": -0.1131,
+      "step": 245
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 0.6204046010971069,
+      "grpo_mean_advantage": 4.470348358154297e-08,
+      "grpo_mean_group_score": 0.6196198463439941,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 5.315724820320611e-07,
+      "learning_rate": 2.722921951984927e-06,
+      "loss": -0.2232,
       "step": 250
     },
     {
-      "epoch": 0.10632911392405063,
-      "grad_norm": 0.6292146444320679,
-      "learning_rate": 4.4112478031634446e-05,
-      "loss": 1.1609960794448853,
-      "step": 252
-    },
-    {
-      "epoch": 0.10717299578059072,
-      "grad_norm": 0.7982883453369141,
-      "learning_rate": 4.44639718804921e-05,
-      "loss": 1.063547968864441,
-      "step": 254
-    },
-    {
-      "epoch": 0.1080168776371308,
-      "grad_norm": 0.7719110250473022,
-      "learning_rate": 4.4815465729349736e-05,
-      "loss": 1.0719804763793945,
-      "step": 256
-    },
-    {
-      "epoch": 0.10886075949367088,
-      "grad_norm": 0.6101011633872986,
-      "learning_rate": 4.516695957820739e-05,
-      "loss": 1.0778400897979736,
-      "step": 258
-    },
-    {
-      "epoch": 0.10970464135021098,
-      "grad_norm": 0.7300994396209717,
-      "learning_rate": 4.5518453427065026e-05,
-      "loss": 1.2129558324813843,
+      "epoch": 1.02002002002002,
+      "grad_norm": 0.8389026522636414,
+      "grpo_mean_advantage": 9.290873776990338e-07,
+      "grpo_mean_group_score": 0.582168459892273,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 4.219644324621186e-06,
+      "learning_rate": 2.640470915079614e-06,
+      "loss": -0.1363,
+      "step": 255
+    },
+    {
+      "epoch": 1.04004004004004,
+      "grad_norm": 0.9067686796188354,
+      "grpo_mean_advantage": 2.533197474008375e-08,
+      "grpo_mean_group_score": 0.5551307797431946,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.6600588992332632e-07,
+      "learning_rate": 2.557866275291035e-06,
+      "loss": -0.1868,
       "step": 260
     },
     {
-      "epoch": 0.11054852320675106,
-      "grad_norm": 0.8348747491836548,
-      "learning_rate": 4.586994727592268e-05,
-      "loss": 1.221714735031128,
-      "step": 262
-    },
-    {
-      "epoch": 0.11139240506329114,
-      "grad_norm": 0.5445612072944641,
-      "learning_rate": 4.6221441124780316e-05,
-      "loss": 1.0187978744506836,
-      "step": 264
-    },
-    {
-      "epoch": 0.11223628691983123,
-      "grad_norm": 0.6230319738388062,
-      "learning_rate": 4.657293497363797e-05,
-      "loss": 1.096561312675476,
-      "step": 266
-    },
-    {
-      "epoch": 0.11308016877637131,
-      "grad_norm": 0.6231237649917603,
-      "learning_rate": 4.6924428822495606e-05,
-      "loss": 1.089842438697815,
-      "step": 268
-    },
-    {
-      "epoch": 0.11392405063291139,
-      "grad_norm": 0.7178627252578735,
-      "learning_rate": 4.727592267135325e-05,
-      "loss": 1.0696645975112915,
+      "epoch": 1.06006006006006,
+      "grad_norm": 0.9277902841567993,
+      "grpo_mean_advantage": -5.662441182607836e-08,
+      "grpo_mean_group_score": 0.535040020942688,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.0909400316450046e-06,
+      "learning_rate": 2.4751983595800093e-06,
+      "loss": -0.1792,
+      "step": 265
+    },
+    {
+      "epoch": 1.08008008008008,
+      "grad_norm": 1.0715463161468506,
+      "grpo_mean_advantage": -9.536743306171047e-08,
+      "grpo_mean_group_score": 0.5673571825027466,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 5.838213610331877e-07,
+      "learning_rate": 2.392557564098649e-06,
+      "loss": -0.1691,
       "step": 270
     },
     {
-      "epoch": 0.11476793248945148,
-      "grad_norm": 0.6895854473114014,
-      "learning_rate": 4.7627416520210896e-05,
-      "loss": 1.0511361360549927,
-      "step": 272
-    },
-    {
-      "epoch": 0.11561181434599156,
-      "grad_norm": 0.6046878695487976,
-      "learning_rate": 4.797891036906854e-05,
-      "loss": 1.1373958587646484,
-      "step": 274
-    },
-    {
-      "epoch": 0.11645569620253164,
-      "grad_norm": 0.6524552702903748,
-      "learning_rate": 4.833040421792619e-05,
-      "loss": 1.0734186172485352,
-      "step": 276
-    },
-    {
-      "epoch": 0.11729957805907174,
-      "grad_norm": 0.6331019997596741,
-      "learning_rate": 4.868189806678383e-05,
-      "loss": 1.123913049697876,
-      "step": 278
-    },
-    {
-      "epoch": 0.11814345991561181,
-      "grad_norm": 0.5919018983840942,
-      "learning_rate": 4.903339191564148e-05,
-      "loss": 1.0635710954666138,
+      "epoch": 1.1001001001001,
+      "grad_norm": 0.7759184837341309,
+      "grpo_mean_advantage": 3.278255533700758e-08,
+      "grpo_mean_group_score": 0.5874732732772827,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 9.317170679423725e-07,
+      "learning_rate": 2.3100342553434924e-06,
+      "loss": -0.1655,
+      "step": 275
+    },
+    {
+      "epoch": 1.12012012012012,
+      "grad_norm": 0.9387398958206177,
+      "grpo_mean_advantage": -1.206994113545079e-07,
+      "grpo_mean_group_score": 0.5569106340408325,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 6.201085511747806e-07,
+      "learning_rate": 2.2277186713410688e-06,
+      "loss": -0.1821,
       "step": 280
     },
     {
-      "epoch": 0.1189873417721519,
-      "grad_norm": 0.6067633032798767,
-      "learning_rate": 4.938488576449912e-05,
-      "loss": 1.0429247617721558,
-      "step": 282
-    },
-    {
-      "epoch": 0.11983122362869199,
-      "grad_norm": 0.6583750247955322,
-      "learning_rate": 4.9736379613356774e-05,
-      "loss": 1.1397464275360107,
-      "step": 284
-    },
-    {
-      "epoch": 0.12067510548523207,
-      "grad_norm": 0.6200069785118103,
-      "learning_rate": 5.008787346221442e-05,
-      "loss": 1.0590803623199463,
-      "step": 286
-    },
-    {
-      "epoch": 0.12151898734177215,
-      "grad_norm": 0.6798665523529053,
-      "learning_rate": 5.0439367311072064e-05,
-      "loss": 1.1318789720535278,
-      "step": 288
-    },
-    {
-      "epoch": 0.12236286919831224,
-      "grad_norm": 0.7508794069290161,
-      "learning_rate": 5.07908611599297e-05,
-      "loss": 1.0934956073760986,
+      "epoch": 1.14014014014014,
+      "grad_norm": 1.6132302284240723,
+      "grpo_mean_advantage": 4.470348358154297e-08,
+      "grpo_mean_group_score": 0.5578873157501221,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 6.115651558502577e-07,
+      "learning_rate": 2.1457008229739395e-06,
+      "loss": -0.2102,
+      "step": 285
+    },
+    {
+      "epoch": 1.16016016016016,
+      "grad_norm": 0.8679026961326599,
+      "grpo_mean_advantage": -3.3453108017056365e-07,
+      "grpo_mean_group_score": 0.5735999345779419,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 3.5326345368957845e-06,
+      "learning_rate": 2.0640703955551214e-06,
+      "loss": -0.2937,
       "step": 290
     },
     {
-      "epoch": 0.12320675105485232,
-      "grad_norm": 0.6901452541351318,
-      "learning_rate": 5.114235500878735e-05,
-      "loss": 1.163407802581787,
-      "step": 292
-    },
-    {
-      "epoch": 0.1240506329113924,
-      "grad_norm": 0.6423285603523254,
-      "learning_rate": 5.1493848857644985e-05,
-      "loss": 1.09059476852417,
-      "step": 294
-    },
-    {
-      "epoch": 0.1248945147679325,
-      "grad_norm": 0.6839275360107422,
-      "learning_rate": 5.1845342706502644e-05,
-      "loss": 1.0690211057662964,
-      "step": 296
-    },
-    {
-      "epoch": 0.1257383966244726,
-      "grad_norm": 0.6350128054618835,
-      "learning_rate": 5.219683655536028e-05,
-      "loss": 0.982322096824646,
-      "step": 298
-    },
-    {
-      "epoch": 0.12658227848101267,
-      "grad_norm": 0.7136530876159668,
-      "learning_rate": 5.254833040421793e-05,
-      "loss": 1.1132930517196655,
-      "step": 300
-    },
-    {
-      "epoch": 0.12658227848101267,
-      "eval_loss": 1.0952109098434448,
-      "eval_runtime": 677.0652,
-      "eval_samples_per_second": 3.112,
-      "eval_steps_per_second": 3.112,
+      "epoch": 1.1801801801801801,
+      "grad_norm": 1.0550166368484497,
+      "grpo_mean_advantage": -1.110136480519941e-07,
+      "grpo_mean_group_score": 0.5626259446144104,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 4.731904823529476e-07,
+      "learning_rate": 1.9829166507585084e-06,
+      "loss": -0.2598,
+      "step": 295
+    },
+    {
+      "epoch": 1.2002002002002001,
+      "grad_norm": 1.2819372415542603,
+      "grpo_mean_advantage": -5.08874677507265e-07,
+      "grpo_mean_group_score": 0.5463050603866577,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.840126174101897e-06,
+      "learning_rate": 1.90232832901255e-06,
+      "loss": -0.2546,
       "step": 300
     },
     {
-      "epoch": 0.12742616033755275,
-      "grad_norm": 0.7339721322059631,
-      "learning_rate": 5.289982425307557e-05,
-      "loss": 0.973595917224884,
-      "step": 302
-    },
-    {
-      "epoch": 0.12827004219409283,
-      "grad_norm": 0.5941481590270996,
-      "learning_rate": 5.3251318101933224e-05,
-      "loss": 0.9819849729537964,
-      "step": 304
-    },
-    {
-      "epoch": 0.1291139240506329,
-      "grad_norm": 0.7153938412666321,
-      "learning_rate": 5.360281195079086e-05,
-      "loss": 1.0315470695495605,
-      "step": 306
-    },
-    {
-      "epoch": 0.12995780590717299,
-      "grad_norm": 0.5167180299758911,
-      "learning_rate": 5.395430579964851e-05,
-      "loss": 0.9492001533508301,
-      "step": 308
-    },
-    {
-      "epoch": 0.1308016877637131,
-      "grad_norm": 0.6055944561958313,
-      "learning_rate": 5.430579964850615e-05,
-      "loss": 1.0156209468841553,
+      "epoch": 1.2202202202202201,
+      "grad_norm": 1.0188143253326416,
+      "grpo_mean_advantage": 1.01327898960335e-07,
+      "grpo_mean_group_score": 0.5352144241333008,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 7.798533943059738e-07,
+      "learning_rate": 1.82239355246389e-06,
+      "loss": -0.1809,
+      "step": 305
+    },
+    {
+      "epoch": 1.2402402402402402,
+      "grad_norm": 2.0709052085876465,
+      "grpo_mean_advantage": 1.341104507446289e-07,
+      "grpo_mean_group_score": 0.5547868013381958,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 7.821902840987605e-07,
+      "learning_rate": 1.7431997286170923e-06,
+      "loss": -0.3559,
       "step": 310
     },
     {
-      "epoch": 0.13164556962025317,
-      "grad_norm": 0.7662386298179626,
-      "learning_rate": 5.4657293497363805e-05,
-      "loss": 1.1791651248931885,
-      "step": 312
-    },
-    {
-      "epoch": 0.13248945147679325,
-      "grad_norm": 0.6065546274185181,
-      "learning_rate": 5.500878734622145e-05,
-      "loss": 1.0009297132492065,
-      "step": 314
-    },
-    {
-      "epoch": 0.13333333333333333,
-      "grad_norm": 0.604225754737854,
-      "learning_rate": 5.536028119507909e-05,
-      "loss": 1.0208244323730469,
-      "step": 316
-    },
-    {
-      "epoch": 0.1341772151898734,
-      "grad_norm": 0.6186763048171997,
-      "learning_rate": 5.571177504393673e-05,
-      "loss": 0.9968416690826416,
-      "step": 318
-    },
-    {
-      "epoch": 0.1350210970464135,
-      "grad_norm": 0.7100363969802856,
-      "learning_rate": 5.606326889279437e-05,
-      "loss": 0.9540256857872009,
+      "epoch": 1.2602602602602602,
+      "grad_norm": 1.8516215085983276,
+      "grpo_mean_advantage": 9.015202806494926e-08,
+      "grpo_mean_group_score": 0.5859472751617432,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.0693488547985908e-06,
+      "learning_rate": 1.6648334547558227e-06,
+      "loss": -0.3874,
+      "step": 315
+    },
+    {
+      "epoch": 1.2802802802802802,
+      "grad_norm": 1.283104419708252,
+      "grpo_mean_advantage": -2.443790378947597e-07,
+      "grpo_mean_group_score": 0.5751550793647766,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.183122208203713e-06,
+      "learning_rate": 1.5873804232499862e-06,
+      "loss": -0.3467,
       "step": 320
     },
     {
-      "epoch": 0.1358649789029536,
-      "grad_norm": 0.6979711055755615,
-      "learning_rate": 5.641476274165203e-05,
-      "loss": 1.0631953477859497,
-      "step": 322
-    },
-    {
-      "epoch": 0.13670886075949368,
-      "grad_norm": 0.6237109303474426,
-      "learning_rate": 5.676625659050967e-05,
-      "loss": 1.0170501470565796,
-      "step": 324
-    },
-    {
-      "epoch": 0.13755274261603376,
-      "grad_norm": 0.6525548696517944,
-      "learning_rate": 5.711775043936731e-05,
-      "loss": 1.0715603828430176,
-      "step": 326
-    },
-    {
-      "epoch": 0.13839662447257384,
-      "grad_norm": 0.6869221329689026,
-      "learning_rate": 5.746924428822495e-05,
-      "loss": 1.0111541748046875,
-      "step": 328
-    },
-    {
-      "epoch": 0.13924050632911392,
-      "grad_norm": 0.553188145160675,
-      "learning_rate": 5.782073813708261e-05,
-      "loss": 1.0311682224273682,
+      "epoch": 1.3003003003003002,
+      "grad_norm": 1.4108576774597168,
+      "grpo_mean_advantage": -6.705522537231445e-08,
+      "grpo_mean_group_score": 0.5497723817825317,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 6.109748937888071e-07,
+      "learning_rate": 1.51092532785238e-06,
+      "loss": -0.1703,
+      "step": 325
+    },
+    {
+      "epoch": 1.3203203203203202,
+      "grad_norm": 1.0421361923217773,
+      "grpo_mean_advantage": -1.639127766850379e-08,
+      "grpo_mean_group_score": 0.55989670753479,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 5.529495297196263e-07,
+      "learning_rate": 1.4355517710873184e-06,
+      "loss": -0.2918,
       "step": 330
     },
     {
-      "epoch": 0.140084388185654,
-      "grad_norm": 0.6760852932929993,
-      "learning_rate": 5.817223198594025e-05,
-      "loss": 1.0213634967803955,
-      "step": 332
-    },
-    {
-      "epoch": 0.1409282700421941,
-      "grad_norm": 0.5907419919967651,
-      "learning_rate": 5.8523725834797894e-05,
-      "loss": 0.9748594164848328,
-      "step": 334
-    },
-    {
-      "epoch": 0.14177215189873418,
-      "grad_norm": 0.7044920921325684,
-      "learning_rate": 5.887521968365554e-05,
-      "loss": 1.05863356590271,
-      "step": 336
-    },
-    {
-      "epoch": 0.14261603375527426,
-      "grad_norm": 0.679073691368103,
-      "learning_rate": 5.922671353251318e-05,
-      "loss": 1.1341127157211304,
-      "step": 338
-    },
-    {
-      "epoch": 0.14345991561181434,
-      "grad_norm": 0.7676237225532532,
-      "learning_rate": 5.957820738137083e-05,
-      "loss": 0.9540836215019226,
+      "epoch": 1.3403403403403402,
+      "grad_norm": 1.3465828895568848,
+      "grpo_mean_advantage": 4.418194237132411e-07,
+      "grpo_mean_group_score": 0.5809233784675598,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 2.9275292945385445e-06,
+      "learning_rate": 1.361342172832502e-06,
+      "loss": -0.3069,
+      "step": 335
+    },
+    {
+      "epoch": 1.3603603603603602,
+      "grad_norm": 1.1959459781646729,
+      "grpo_mean_advantage": 9.685754776000977e-08,
+      "grpo_mean_group_score": 0.5568087100982666,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 3.754235251562932e-07,
+      "learning_rate": 1.2883776801940884e-06,
+      "loss": -0.5594,
       "step": 340
     },
     {
-      "epoch": 0.14430379746835442,
-      "grad_norm": 0.6313899755477905,
-      "learning_rate": 5.9929701230228474e-05,
-      "loss": 1.0585911273956299,
-      "step": 342
-    },
-    {
-      "epoch": 0.1451476793248945,
-      "grad_norm": 0.7123099565505981,
-      "learning_rate": 6.028119507908612e-05,
-      "loss": 1.0760118961334229,
-      "step": 344
-    },
-    {
-      "epoch": 0.1459915611814346,
-      "grad_norm": 0.585935652256012,
-      "learning_rate": 6.063268892794376e-05,
-      "loss": 1.036866307258606,
-      "step": 346
-    },
-    {
-      "epoch": 0.1468354430379747,
-      "grad_norm": 0.5643263459205627,
-      "learning_rate": 6.0984182776801416e-05,
-      "loss": 1.0242938995361328,
-      "step": 348
-    },
-    {
-      "epoch": 0.14767932489451477,
-      "grad_norm": 0.626761794090271,
-      "learning_rate": 6.133567662565906e-05,
-      "loss": 1.0497376918792725,
+      "epoch": 1.3803803803803802,
+      "grad_norm": 1.8967422246932983,
+      "grpo_mean_advantage": -2.384185791015625e-07,
+      "grpo_mean_group_score": 0.5655568838119507,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 6.821086913078034e-07,
+      "learning_rate": 1.216738078773522e-06,
+      "loss": -0.4102,
+      "step": 345
+    },
+    {
+      "epoch": 1.4004004004004005,
+      "grad_norm": 2.221132755279541,
+      "grpo_mean_advantage": -8.717179156292332e-08,
+      "grpo_mean_group_score": 0.6089578866958618,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 2.500940354366321e-06,
+      "learning_rate": 1.146501705423155e-06,
+      "loss": -0.338,
       "step": 350
     },
     {
-      "epoch": 0.14852320675105485,
-      "grad_norm": 0.5106956958770752,
-      "learning_rate": 6.16871704745167e-05,
-      "loss": 0.9811885356903076,
-      "step": 352
-    },
-    {
-      "epoch": 0.14936708860759493,
-      "grad_norm": 0.6948089003562927,
-      "learning_rate": 6.203866432337434e-05,
-      "loss": 1.0715330839157104,
-      "step": 354
-    },
-    {
-      "epoch": 0.150210970464135,
-      "grad_norm": 0.699713945388794,
-      "learning_rate": 6.239015817223199e-05,
-      "loss": 1.0405226945877075,
-      "step": 356
-    },
-    {
-      "epoch": 0.15105485232067511,
-      "grad_norm": 0.6437667012214661,
-      "learning_rate": 6.274165202108964e-05,
-      "loss": 1.0490930080413818,
-      "step": 358
-    },
-    {
-      "epoch": 0.1518987341772152,
-      "grad_norm": 0.6952699422836304,
-      "learning_rate": 6.309314586994728e-05,
-      "loss": 0.9267548322677612,
+      "epoch": 1.4204204204204205,
+      "grad_norm": 2.3640377521514893,
+      "grpo_mean_advantage": 2.1606683731079102e-07,
+      "grpo_mean_group_score": 0.6129671335220337,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.4568390724889468e-06,
+      "learning_rate": 1.0777453625860474e-06,
+      "loss": -0.4985,
+      "step": 355
+    },
+    {
+      "epoch": 1.4404404404404405,
+      "grad_norm": 1.9084734916687012,
+      "grpo_mean_advantage": -3.725290298461914e-09,
+      "grpo_mean_group_score": 0.5562310814857483,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 2.965894054796081e-06,
+      "learning_rate": 1.0105442343136184e-06,
+      "loss": -0.4347,
       "step": 360
     },
     {
-      "epoch": 0.15274261603375527,
-      "grad_norm": 0.6713186502456665,
-      "learning_rate": 6.344463971880492e-05,
-      "loss": 1.0427420139312744,
-      "step": 362
-    },
-    {
-      "epoch": 0.15358649789029535,
-      "grad_norm": 0.6750379800796509,
-      "learning_rate": 6.379613356766257e-05,
-      "loss": 1.048950433731079,
-      "step": 364
-    },
-    {
-      "epoch": 0.15443037974683543,
-      "grad_norm": 0.6053379774093628,
-      "learning_rate": 6.414762741652022e-05,
-      "loss": 1.0156004428863525,
-      "step": 366
-    },
-    {
-      "epoch": 0.15527426160337554,
-      "grad_norm": 0.8063633441925049,
-      "learning_rate": 6.449912126537786e-05,
-      "loss": 1.0020819902420044,
-      "step": 368
-    },
-    {
-      "epoch": 0.15611814345991562,
-      "grad_norm": 0.8027494549751282,
-      "learning_rate": 6.48506151142355e-05,
-      "loss": 1.055633783340454,
+      "epoch": 1.4604604604604605,
+      "grad_norm": 1.6063904762268066,
+      "grpo_mean_advantage": 4.313886279305734e-07,
+      "grpo_mean_group_score": 0.5884170532226562,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.9621948013082147e-06,
+      "learning_rate": 9.449718040529987e-07,
+      "loss": -0.6217,
+      "step": 365
+    },
+    {
+      "epoch": 1.4804804804804805,
+      "grad_norm": 2.114664077758789,
+      "grpo_mean_advantage": 2.0489096641540527e-07,
+      "grpo_mean_group_score": 0.5795440673828125,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.0235522722723545e-06,
+      "learning_rate": 8.810997742939531e-07,
+      "loss": -0.5364,
       "step": 370
     },
     {
-      "epoch": 0.1569620253164557,
-      "grad_norm": 0.6580121517181396,
-      "learning_rate": 6.520210896309315e-05,
-      "loss": 1.0149940252304077,
-      "step": 372
-    },
-    {
-      "epoch": 0.15780590717299578,
-      "grad_norm": 0.6561233997344971,
-      "learning_rate": 6.55536028119508e-05,
-      "loss": 0.9769611954689026,
-      "step": 374
-    },
-    {
-      "epoch": 0.15864978902953586,
-      "grad_norm": 0.6444346308708191,
-      "learning_rate": 6.590509666080844e-05,
-      "loss": 0.9099349975585938,
-      "step": 376
-    },
-    {
-      "epoch": 0.15949367088607594,
-      "grad_norm": 0.5879359245300293,
-      "learning_rate": 6.625659050966608e-05,
-      "loss": 1.0797548294067383,
-      "step": 378
-    },
-    {
-      "epoch": 0.16033755274261605,
-      "grad_norm": 0.6994144916534424,
-      "learning_rate": 6.660808435852373e-05,
-      "loss": 1.0336791276931763,
+      "epoch": 1.5005005005005005,
+      "grad_norm": 1.8450465202331543,
+      "grpo_mean_advantage": -1.4185905001795618e-06,
+      "grpo_mean_group_score": 0.5607603788375854,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.0947338523692451e-05,
+      "learning_rate": 8.189979881632634e-07,
+      "loss": -0.4798,
+      "step": 375
+    },
+    {
+      "epoch": 1.5205205205205206,
+      "grad_norm": 2.673438787460327,
+      "grpo_mean_advantage": -1.758337049295733e-07,
+      "grpo_mean_group_score": 0.5381432771682739,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 9.663675655247062e-07,
+      "learning_rate": 7.587343530522945e-07,
+      "loss": -0.4805,
       "step": 380
     },
     {
-      "epoch": 0.16118143459915613,
-      "grad_norm": 0.6128669381141663,
-      "learning_rate": 6.695957820738138e-05,
-      "loss": 1.018118143081665,
-      "step": 382
-    },
-    {
-      "epoch": 0.1620253164556962,
-      "grad_norm": 1.0237540006637573,
-      "learning_rate": 6.731107205623902e-05,
-      "loss": 1.1405497789382935,
-      "step": 384
-    },
-    {
-      "epoch": 0.16286919831223629,
-      "grad_norm": 0.6091578006744385,
-      "learning_rate": 6.766256590509666e-05,
-      "loss": 1.0314189195632935,
-      "step": 386
-    },
-    {
-      "epoch": 0.16371308016877636,
-      "grad_norm": 0.5916037559509277,
-      "learning_rate": 6.801405975395431e-05,
-      "loss": 0.9564052820205688,
-      "step": 388
-    },
-    {
-      "epoch": 0.16455696202531644,
-      "grad_norm": 0.771653950214386,
-      "learning_rate": 6.836555360281195e-05,
-      "loss": 1.0023859739303589,
+      "epoch": 1.5405405405405406,
+      "grad_norm": 2.2263550758361816,
+      "grpo_mean_advantage": -6.973743325033865e-07,
+      "grpo_mean_group_score": 0.5528443455696106,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 4.341973180999048e-06,
+      "learning_rate": 7.003747663612581e-07,
+      "loss": -0.433,
+      "step": 385
+    },
+    {
+      "epoch": 1.5605605605605606,
+      "grad_norm": 2.3657093048095703,
+      "grpo_mean_advantage": 1.7881394143159923e-08,
+      "grpo_mean_group_score": 0.6091476678848267,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.3004198251564958e-07,
+      "learning_rate": 6.439830434413754e-07,
+      "loss": -0.6021,
       "step": 390
     },
     {
-      "epoch": 0.16540084388185655,
-      "grad_norm": 0.654658317565918,
-      "learning_rate": 6.87170474516696e-05,
-      "loss": 1.07024085521698,
-      "step": 392
-    },
-    {
-      "epoch": 0.16624472573839663,
-      "grad_norm": 0.6611968874931335,
-      "learning_rate": 6.906854130052724e-05,
-      "loss": 1.0552500486373901,
-      "step": 394
-    },
-    {
-      "epoch": 0.1670886075949367,
-      "grad_norm": 0.6955893039703369,
-      "learning_rate": 6.942003514938489e-05,
-      "loss": 1.0562875270843506,
-      "step": 396
-    },
-    {
-      "epoch": 0.1679324894514768,
-      "grad_norm": 0.6666058301925659,
-      "learning_rate": 6.977152899824253e-05,
-      "loss": 0.9850592017173767,
-      "step": 398
-    },
-    {
-      "epoch": 0.16877637130801687,
-      "grad_norm": 0.6131711006164551,
-      "learning_rate": 7.012302284710018e-05,
-      "loss": 1.0077755451202393,
+      "epoch": 1.5805805805805806,
+      "grad_norm": 1.9847129583358765,
+      "grpo_mean_advantage": 3.4868716625169327e-07,
+      "grpo_mean_group_score": 0.5397372245788574,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 2.059372718576924e-06,
+      "learning_rate": 5.896208478137222e-07,
+      "loss": -0.5595,
+      "step": 395
+    },
+    {
+      "epoch": 1.6006006006006006,
+      "grad_norm": 2.922114133834839,
+      "grpo_mean_advantage": -2.1636485598719446e-06,
+      "grpo_mean_group_score": 0.5873125195503235,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 9.725940799398813e-06,
+      "learning_rate": 5.373476237410808e-07,
+      "loss": -0.5592,
       "step": 400
     },
     {
-      "epoch": 0.16877637130801687,
-      "eval_loss": 1.0625108480453491,
-      "eval_runtime": 691.0068,
-      "eval_samples_per_second": 3.049,
-      "eval_steps_per_second": 3.049,
-      "step": 400
-    },
-    {
-      "epoch": 0.16962025316455695,
-      "grad_norm": 0.6286499500274658,
-      "learning_rate": 7.047451669595783e-05,
-      "loss": 1.1012427806854248,
-      "step": 402
-    },
-    {
-      "epoch": 0.17046413502109706,
-      "grad_norm": 0.6639351844787598,
-      "learning_rate": 7.082601054481547e-05,
-      "loss": 1.0379719734191895,
-      "step": 404
-    },
-    {
-      "epoch": 0.17130801687763714,
-      "grad_norm": 0.750401496887207,
-      "learning_rate": 7.117750439367311e-05,
-      "loss": 1.031856656074524,
-      "step": 406
-    },
-    {
-      "epoch": 0.17215189873417722,
-      "grad_norm": 0.8084847331047058,
-      "learning_rate": 7.152899824253075e-05,
-      "loss": 1.0493193864822388,
-      "step": 408
-    },
-    {
-      "epoch": 0.1729957805907173,
-      "grad_norm": 0.7448462247848511,
-      "learning_rate": 7.188049209138841e-05,
-      "loss": 1.1012418270111084,
+      "epoch": 1.6206206206206206,
+      "grad_norm": 1.8524045944213867,
+      "grpo_mean_advantage": -5.960464477539063e-08,
+      "grpo_mean_group_score": 0.5601426362991333,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 3.460792754594877e-07,
+      "learning_rate": 4.872205312265074e-07,
+      "loss": -0.5623,
+      "step": 405
+    },
+    {
+      "epoch": 1.6406406406406406,
+      "grad_norm": 1.7269790172576904,
+      "grpo_mean_advantage": 2.6226044269606064e-07,
+      "grpo_mean_group_score": 0.578656792640686,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 7.928817922220333e-07,
+      "learning_rate": 4.3929438350970687e-07,
+      "loss": -0.5943,
       "step": 410
     },
     {
-      "epoch": 0.17383966244725738,
-      "grad_norm": 0.5841867923736572,
-      "learning_rate": 7.223198594024605e-05,
-      "loss": 0.9926692247390747,
-      "step": 412
-    },
-    {
-      "epoch": 0.17468354430379746,
-      "grad_norm": 0.7125606536865234,
-      "learning_rate": 7.258347978910369e-05,
-      "loss": 1.0588877201080322,
-      "step": 414
-    },
-    {
-      "epoch": 0.17552742616033756,
-      "grad_norm": 0.5750942230224609,
-      "learning_rate": 7.293497363796134e-05,
-      "loss": 1.038270354270935,
-      "step": 416
-    },
-    {
-      "epoch": 0.17637130801687764,
-      "grad_norm": 0.565444827079773,
-      "learning_rate": 7.328646748681899e-05,
-      "loss": 0.9843021035194397,
-      "step": 418
-    },
-    {
-      "epoch": 0.17721518987341772,
-      "grad_norm": 0.5825693011283875,
-      "learning_rate": 7.363796133567663e-05,
-      "loss": 1.0731632709503174,
+      "epoch": 1.6606606606606606,
+      "grad_norm": 2.26530122756958,
+      "grpo_mean_advantage": 2.3558736756967846e-06,
+      "grpo_mean_group_score": 0.5885810852050781,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.4469559573626611e-05,
+      "learning_rate": 3.936215871295634e-07,
+      "loss": -0.6193,
+      "step": 415
+    },
+    {
+      "epoch": 1.6806806806806807,
+      "grad_norm": 2.6794464588165283,
+      "grpo_mean_advantage": 1.639127766850379e-08,
+      "grpo_mean_group_score": 0.5805023312568665,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 9.352411325380672e-07,
+      "learning_rate": 3.502520846183577e-07,
+      "loss": -0.6934,
       "step": 420
     },
     {
-      "epoch": 0.1780590717299578,
-      "grad_norm": 0.6267391443252563,
-      "learning_rate": 7.398945518453427e-05,
-      "loss": 1.0061273574829102,
-      "step": 422
-    },
-    {
-      "epoch": 0.17890295358649788,
-      "grad_norm": 0.6621372103691101,
-      "learning_rate": 7.434094903339192e-05,
-      "loss": 1.0461612939834595,
-      "step": 424
-    },
-    {
-      "epoch": 0.17974683544303796,
-      "grad_norm": 0.6635435223579407,
-      "learning_rate": 7.469244288224957e-05,
-      "loss": 0.9789207577705383,
-      "step": 426
-    },
-    {
-      "epoch": 0.18059071729957807,
-      "grad_norm": 0.6342346668243408,
-      "learning_rate": 7.504393673110721e-05,
-      "loss": 1.0527069568634033,
-      "step": 428
-    },
-    {
-      "epoch": 0.18143459915611815,
-      "grad_norm": 0.6762149930000305,
-      "learning_rate": 7.539543057996485e-05,
-      "loss": 0.9708702564239502,
+      "epoch": 1.7007007007007007,
+      "grad_norm": 2.100447654724121,
+      "grpo_mean_advantage": 3.2387674764322583e-06,
+      "grpo_mean_group_score": 0.5655918121337891,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.999079904635437e-05,
+      "learning_rate": 3.092332998903416e-07,
+      "loss": -0.5126,
+      "step": 425
+    },
+    {
+      "epoch": 1.7207207207207207,
+      "grad_norm": 2.1027915477752686,
+      "grpo_mean_advantage": 3.5464762504489045e-07,
+      "grpo_mean_group_score": 0.5474504232406616,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.7663603557593888e-06,
+      "learning_rate": 2.706100863843822e-07,
+      "loss": -0.5446,
       "step": 430
     },
     {
-      "epoch": 0.18227848101265823,
-      "grad_norm": 0.7073282599449158,
-      "learning_rate": 7.57469244288225e-05,
-      "loss": 1.0509834289550781,
-      "step": 432
-    },
-    {
-      "epoch": 0.1831223628691983,
-      "grad_norm": 0.6917856931686401,
-      "learning_rate": 7.609841827768014e-05,
-      "loss": 1.0128819942474365,
-      "step": 434
-    },
-    {
-      "epoch": 0.1839662447257384,
-      "grad_norm": 0.5574942231178284,
-      "learning_rate": 7.644991212653779e-05,
-      "loss": 0.989395797252655,
-      "step": 436
-    },
-    {
-      "epoch": 0.1848101265822785,
-      "grad_norm": 0.640765905380249,
-      "learning_rate": 7.680140597539543e-05,
-      "loss": 0.9846042990684509,
-      "step": 438
-    },
-    {
-      "epoch": 0.18565400843881857,
-      "grad_norm": 0.6699127554893494,
-      "learning_rate": 7.715289982425308e-05,
-      "loss": 1.0344442129135132,
+      "epoch": 1.7407407407407407,
+      "grad_norm": 2.289045572280884,
+      "grpo_mean_advantage": 3.6135315895080566e-07,
+      "grpo_mean_group_score": 0.5874254703521729,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 2.356920958845876e-06,
+      "learning_rate": 2.3442467801738867e-07,
+      "loss": -0.5125,
+      "step": 435
+    },
+    {
+      "epoch": 1.7607607607607607,
+      "grad_norm": 2.278038501739502,
+      "grpo_mean_advantage": 2.7567148563889532e-08,
+      "grpo_mean_group_score": 0.5815118551254272,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 9.97340521280421e-07,
+      "learning_rate": 2.007166430021415e-07,
+      "loss": -0.595,
       "step": 440
     },
     {
-      "epoch": 0.18649789029535865,
-      "grad_norm": 0.6164930462837219,
-      "learning_rate": 7.750439367311072e-05,
-      "loss": 1.0179373025894165,
-      "step": 442
-    },
-    {
-      "epoch": 0.18734177215189873,
-      "grad_norm": 0.6880720853805542,
-      "learning_rate": 7.785588752196837e-05,
-      "loss": 1.0518895387649536,
-      "step": 444
-    },
-    {
-      "epoch": 0.1881856540084388,
-      "grad_norm": 0.6501413583755493,
-      "learning_rate": 7.820738137082601e-05,
-      "loss": 1.0442606210708618,
-      "step": 446
-    },
-    {
-      "epoch": 0.1890295358649789,
-      "grad_norm": 0.6076085567474365,
-      "learning_rate": 7.855887521968366e-05,
-      "loss": 0.9828442335128784,
-      "step": 448
-    },
-    {
-      "epoch": 0.189873417721519,
-      "grad_norm": 0.6418202519416809,
-      "learning_rate": 7.89103690685413e-05,
-      "loss": 1.0573710203170776,
+      "epoch": 1.7807807807807807,
+      "grad_norm": 2.340942621231079,
+      "grpo_mean_advantage": -8.34465012644614e-08,
+      "grpo_mean_group_score": 0.5611211061477661,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 5.558832185670326e-07,
+      "learning_rate": 1.6952284058003366e-07,
+      "loss": -0.8055,
+      "step": 445
+    },
+    {
+      "epoch": 1.800800800800801,
+      "grad_norm": 2.4256298542022705,
+      "grpo_mean_advantage": -1.9818544672034477e-07,
+      "grpo_mean_group_score": 0.590424656867981,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 6.800727305744658e-07,
+      "learning_rate": 1.4087738071603075e-07,
+      "loss": -0.8561,
       "step": 450
     },
     {
-      "epoch": 0.19071729957805908,
-      "grad_norm": 0.7055076360702515,
-      "learning_rate": 7.926186291739895e-05,
-      "loss": 1.0216103792190552,
-      "step": 452
-    },
-    {
-      "epoch": 0.19156118143459916,
-      "grad_norm": 0.5668330192565918,
-      "learning_rate": 7.961335676625659e-05,
-      "loss": 0.9837722778320312,
-      "step": 454
-    },
-    {
-      "epoch": 0.19240506329113924,
-      "grad_norm": 0.6419380307197571,
-      "learning_rate": 7.996485061511424e-05,
-      "loss": 1.0003894567489624,
-      "step": 456
-    },
-    {
-      "epoch": 0.19324894514767932,
-      "grad_norm": 0.5949198007583618,
-      "learning_rate": 8.031634446397188e-05,
-      "loss": 1.0609031915664673,
-      "step": 458
-    },
-    {
-      "epoch": 0.1940928270042194,
-      "grad_norm": 0.7032039761543274,
-      "learning_rate": 8.066783831282952e-05,
-      "loss": 1.0543403625488281,
+      "epoch": 1.820820820820821,
+      "grad_norm": 1.6453255414962769,
+      "grpo_mean_advantage": -1.9371508841459217e-08,
+      "grpo_mean_group_score": 0.5670351982116699,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 3.142378943721269e-07,
+      "learning_rate": 1.1481158679992554e-07,
+      "loss": -0.429,
+      "step": 455
+    },
+    {
+      "epoch": 1.840840840840841,
+      "grad_norm": 2.3458049297332764,
+      "grpo_mean_advantage": 2.3692845729783585e-07,
+      "grpo_mean_group_score": 0.5640432834625244,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.682946731307311e-06,
+      "learning_rate": 9.135396139467151e-08,
+      "loss": -0.642,
       "step": 460
     },
     {
-      "epoch": 0.1949367088607595,
-      "grad_norm": 0.5775868892669678,
-      "learning_rate": 8.101933216168718e-05,
-      "loss": 0.9819303154945374,
-      "step": 462
-    },
-    {
-      "epoch": 0.19578059071729959,
-      "grad_norm": 0.9301062226295471,
-      "learning_rate": 8.137082601054482e-05,
-      "loss": 1.0542067289352417,
-      "step": 464
-    },
-    {
-      "epoch": 0.19662447257383966,
-      "grad_norm": 0.6193217039108276,
-      "learning_rate": 8.172231985940246e-05,
-      "loss": 0.9966341257095337,
-      "step": 466
-    },
-    {
-      "epoch": 0.19746835443037974,
-      "grad_norm": 0.6286146640777588,
-      "learning_rate": 8.20738137082601e-05,
-      "loss": 1.0474121570587158,
-      "step": 468
-    },
-    {
-      "epoch": 0.19831223628691982,
-      "grad_norm": 0.7418972253799438,
-      "learning_rate": 8.242530755711776e-05,
-      "loss": 0.9549239277839661,
+      "epoch": 1.860860860860861,
+      "grad_norm": 2.730945110321045,
+      "grpo_mean_advantage": 1.110136480519941e-07,
+      "grpo_mean_group_score": 0.5584251284599304,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 8.930008448260196e-07,
+      "learning_rate": 7.053015506924749e-08,
+      "loss": -0.583,
+      "step": 465
+    },
+    {
+      "epoch": 1.880880880880881,
+      "grad_norm": 2.1463465690612793,
+      "grpo_mean_advantage": 2.5406478698641877e-07,
+      "grpo_mean_group_score": 0.5392154455184937,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 9.93092498902115e-07,
+      "learning_rate": 5.236293835013839e-08,
+      "loss": -0.5197,
       "step": 470
     },
     {
-      "epoch": 0.1991561181434599,
-      "grad_norm": 0.6122808456420898,
-      "learning_rate": 8.27768014059754e-05,
-      "loss": 1.0191338062286377,
-      "step": 472
-    },
-    {
-      "epoch": 0.2,
-      "grad_norm": 0.6375362277030945,
-      "learning_rate": 8.312829525483304e-05,
-      "loss": 1.0987539291381836,
-      "step": 474
-    },
-    {
-      "epoch": 0.2008438818565401,
-      "grad_norm": 0.6459513306617737,
-      "learning_rate": 8.347978910369068e-05,
-      "loss": 1.0369136333465576,
-      "step": 476
-    },
-    {
-      "epoch": 0.20168776371308017,
-      "grad_norm": 0.7029640674591064,
-      "learning_rate": 8.383128295254833e-05,
-      "loss": 1.0582096576690674,
-      "step": 478
-    },
-    {
-      "epoch": 0.20253164556962025,
-      "grad_norm": 0.6345387697219849,
-      "learning_rate": 8.418277680140598e-05,
-      "loss": 1.022916316986084,
+      "epoch": 1.900900900900901,
+      "grad_norm": 2.427900791168213,
+      "grpo_mean_advantage": -8.940697071579962e-09,
+      "grpo_mean_group_score": 0.5686308741569519,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 5.835169645251881e-07,
+      "learning_rate": 3.687217682209837e-08,
+      "loss": -0.5864,
+      "step": 475
+    },
+    {
+      "epoch": 1.920920920920921,
+      "grad_norm": 2.042795419692993,
+      "grpo_mean_advantage": 4.0605664253234863e-07,
+      "grpo_mean_group_score": 0.5842767357826233,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 2.3210795916384086e-06,
+      "learning_rate": 2.4074809405425227e-08,
+      "loss": -0.5721,
       "step": 480
     },
     {
-      "epoch": 0.20337552742616033,
-      "grad_norm": 0.5764590501785278,
-      "learning_rate": 8.453427065026362e-05,
-      "loss": 0.973024308681488,
-      "step": 482
-    },
-    {
-      "epoch": 0.2042194092827004,
-      "grad_norm": 0.5884482860565186,
-      "learning_rate": 8.488576449912127e-05,
-      "loss": 1.0292812585830688,
-      "step": 484
-    },
-    {
-      "epoch": 0.20506329113924052,
-      "grad_norm": 0.616357147693634,
-      "learning_rate": 8.523725834797891e-05,
-      "loss": 1.0083447694778442,
-      "step": 486
-    },
-    {
-      "epoch": 0.2059071729957806,
-      "grad_norm": 0.7671196460723877,
-      "learning_rate": 8.558875219683656e-05,
-      "loss": 0.9936985373497009,
-      "step": 488
-    },
-    {
-      "epoch": 0.20675105485232068,
-      "grad_norm": 0.6197299957275391,
-      "learning_rate": 8.59402460456942e-05,
-      "loss": 1.051513910293579,
+      "epoch": 1.940940940940941,
+      "grad_norm": 2.800136089324951,
+      "grpo_mean_advantage": 1.341104507446289e-07,
+      "grpo_mean_group_score": 0.5495311617851257,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 1.507950400991831e-06,
+      "learning_rate": 1.3984829833499636e-08,
+      "loss": -0.5944,
+      "step": 485
+    },
+    {
+      "epoch": 1.960960960960961,
+      "grad_norm": 2.8475866317749023,
+      "grpo_mean_advantage": 2.689659481802664e-07,
+      "grpo_mean_group_score": 0.549436628818512,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 8.491958851664094e-07,
+      "learning_rate": 6.6132713508446075e-09,
+      "loss": -0.7015,
       "step": 490
     },
     {
-      "epoch": 0.20759493670886076,
-      "grad_norm": 0.6912890672683716,
-      "learning_rate": 8.629173989455185e-05,
-      "loss": 0.9474978446960449,
-      "step": 492
-    },
-    {
-      "epoch": 0.20843881856540084,
-      "grad_norm": 0.6941592693328857,
-      "learning_rate": 8.664323374340949e-05,
-      "loss": 1.0671660900115967,
-      "step": 494
-    },
-    {
-      "epoch": 0.20928270042194091,
-      "grad_norm": 0.5889528393745422,
-      "learning_rate": 8.699472759226714e-05,
-      "loss": 1.0020159482955933,
-      "step": 496
-    },
-    {
-      "epoch": 0.21012658227848102,
-      "grad_norm": 0.6478549838066101,
-      "learning_rate": 8.734622144112478e-05,
-      "loss": 1.0165860652923584,
-      "step": 498
-    },
-    {
-      "epoch": 0.2109704641350211,
-      "grad_norm": 0.6018255949020386,
-      "learning_rate": 8.769771528998243e-05,
-      "loss": 0.8798263072967529,
-      "step": 500
-    },
-    {
-      "epoch": 0.2109704641350211,
-      "eval_loss": 1.042096495628357,
-      "eval_runtime": 692.4361,
-      "eval_samples_per_second": 3.043,
-      "eval_steps_per_second": 3.043,
+      "epoch": 1.980980980980981,
+      "grad_norm": 2.9422402381896973,
+      "grpo_mean_advantage": 8.195638656616211e-08,
+      "grpo_mean_group_score": 0.544632077217102,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 3.802849732892355e-06,
+      "learning_rate": 1.9681946484320645e-09,
+      "loss": -0.4033,
+      "step": 495
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 2.66204833984375,
+      "grpo_mean_advantage": 7.552536089860951e-07,
+      "grpo_mean_group_score": 0.5968535542488098,
+      "grpo_mean_kl_div": 0.0,
+      "grpo_std_advantage": 4.143997102801222e-06,
+      "learning_rate": 5.467904943851077e-11,
+      "loss": -0.6773,
       "step": 500
     }
   ],
-  "logging_steps": 2,
-  "max_steps": 14220,
+  "logging_steps": 5,
+  "max_steps": 500,
   "num_input_tokens_seen": 0,
-  "num_train_epochs": 6,
-  "save_steps": 500,
+  "num_train_epochs": 2,
+  "save_steps": 100,
   "stateful_callbacks": {
-    "EarlyStoppingCallback": {
-      "args": {
-        "early_stopping_patience": 5,
-        "early_stopping_threshold": 0.001
-      },
-      "attributes": {
-        "early_stopping_patience_counter": 0
-      }
-    },
     "TrainerControl": {
       "args": {
         "should_epoch_stop": false,
         "should_evaluate": false,
         "should_log": false,
         "should_save": true,
-        "should_training_stop": false
+        "should_training_stop": true
       },
       "attributes": {}
     }
   },
-  "total_flos": 2.9886635097296486e+17,
+  "total_flos": 0.0,
   "train_batch_size": 1,
   "trial_name": null,
   "trial_params": null

checkpoints/checkpoint-500/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:afae4a8ce1391f149eb00b14eed8f891c715e892ea546bf754d22db2c2bc5969
-size 4792
+oid sha256:a4a4e48ed61b7c96f3bd2836ac828013a311834ab8a9542ea461fe1ff953396b
+size 5496

config_resolved.yaml

@@ -1,36 +1,34 @@
 run:
-  run_dir: ./task2file/sft_qwen_14B_v2
-  seed: 42
+  run_dir: runs/grpo_14b_run1
+model:
+  repo_id: /workspace/Models/Qwen2.5-Coder-14B-CPT-SFT_v2
+  tokenizer_name: Qwen/Qwen2.5-Coder-14B
+  load_in_8bit: false
+  load_in_4bit: false
+  torch_dtype: bfloat16
+  device_map: auto
+  trust_remote_code: true
 wandb:
   enabled: true
-  project: sft-training
+  project: rl-training
   entity: null
   name: null
   tags:
-  - sft-lora
-  - instruction-tuning
+  - grpo-lora
+  - 14B-QWEN
   notes: null
-model:
-  repo_id: ./Models/Qwen2.5-Coder-14B-CPT
-  revision: null
-  base_local_dir: base_model
-  trust_remote_code: true
-  tokenizer_use_fast: true
-  device_map: auto
-  torch_dtype: bfloat16
-  use_4bit: false
-  bnb_4bit_quant_type: nf4
-  bnb_4bit_use_double_quant: false
-  bnb_4bit_compute_dtype: bfloat16
-  attn_implementation: null
 data:
-  train_jsonl: sft_dataset.jsonl
+  train_jsonl: grpo_dataset.jsonl
   eval_jsonl: null
-  eval_split_ratio: 0.1
-  instruction_field: instruction
-  input_field: input
-  output_field: output
-  format_type: custom
+  eval_split_ratio: 0.0
+  shuffle: true
+  num_proc: 1
+  prompt_field: prompt
+  completions_field: completions
+  scores_field: scores
+  format_type: raw
+  max_length: 2048
+  min_completions: 2
   system_prompt: "You are a Hyperswitch Rust code analyzer. Identify functions/structs\
     \ that need modification for a given task.\n\n## Output Format\n\n##OUTPUT\nExplain\
     \ the data flow and why each component must change:\n- Flow: [Input \u2192 Processing\
@@ -59,44 +57,56 @@ data:
     {input}<|im_end|>
 
     {output}<|im_end|>'
-  max_length: 2048
-  shuffle: true
-  num_proc: 4
+grpo:
+  group_size: 4
+  kl_coef: 0.05
+  normalize_advantages: true
+  reward_scaling: 1.0
+  reward_bias: 0.0
+  reward_clip: 5.0
+  advantage_temperature: 1.0
+  use_reference_model: false
+  seed: 42
 peft:
   enabled: true
-  r: 32
-  lora_alpha: 64
+  r: 16
+  lora_alpha: 32
   lora_dropout: 0.05
+  target_modules:
+  - q_proj
+  - k_proj
+  - v_proj
+  - o_proj
+  - gate_proj
+  - up_proj
+  - down_proj
   bias: none
-  target_modules: auto
+  task_type: CAUSAL_LM
 train:
-  num_train_epochs: 6
+  output_dir: runs/grpo_14b_run1
+  num_train_epochs: 2
   per_device_train_batch_size: 1
-  per_device_eval_batch_size: 1
   gradient_accumulation_steps: 8
-  learning_rate: 2e-4
-  weight_decay: 0.0
-  warmup_ratio: 0.08
+  per_device_eval_batch_size: 1
+  learning_rate: 5.0e-06
+  weight_decay: 0.01
+  warmup_ratio: 0.05
   lr_scheduler_type: cosine
-  optim: adamw_torch
+  fp16: false
+  bf16: true
   max_grad_norm: 1.0
   gradient_checkpointing: true
-  logging_steps: 2
-  save_strategy: steps
-  save_steps: 500
-  save_total_limit: 20
-  evaluation_strategy: steps
-  eval_steps: 100
-  load_best_model_at_end: true
-  early_stopping:
-    enabled: true
-    patience: 5
-    min_delta: 0.001
-    metric: eval_loss
-    mode: min
-  resume_from_checkpoint: auto
+  logging_steps: 5
+  save_steps: 100
+  save_total_limit: 2
+  evaluation_strategy: 'no'
+  dataloader_num_workers: 4
+  dataloader_pin_memory: true
+  remove_unused_columns: false
+  report_to: []
+  seed: 42
+  ddp_find_unused_parameters: false
 merge:
-  enabled: true
-  merged_dtype: float16
-  max_shard_size: 2GB
-  output_dir: ./Models/Qwen2.5-Coder-14B-CPT-SFT_v2
+  enabled: false
+upload:
+  enabled: false

logs/grpo_metrics.jsonl

@@ -0,0 +1,101 @@
+{"ts": "2025-12-27T19:40:44", "step": 5, "epoch": 0.02, "grpo_mean_advantage": -1.3560057254835556e-07, "grpo_std_advantage": 3.0318567496578908e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5922331809997559}
+{"ts": "2025-12-27T19:46:55", "step": 5, "epoch": 0.02, "grpo_mean_advantage": -1.3560057254835556e-07, "grpo_std_advantage": 3.0318567496578908e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5922331809997559}
+{"ts": "2025-12-27T19:48:18", "step": 10, "epoch": 0.04, "grpo_mean_advantage": 3.6619603633880615e-06, "grpo_std_advantage": 1.6246918676188216e-05, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5561589002609253}
+{"ts": "2025-12-27T19:49:40", "step": 15, "epoch": 0.0601, "grpo_mean_advantage": -1.0654330395709621e-07, "grpo_std_advantage": 5.399440965447866e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5759152173995972}
+{"ts": "2025-12-27T19:51:01", "step": 20, "epoch": 0.0801, "grpo_mean_advantage": -5.871057737749652e-07, "grpo_std_advantage": 2.6951597646984737e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5127314329147339}
+{"ts": "2025-12-27T19:52:23", "step": 25, "epoch": 0.1001, "grpo_mean_advantage": 6.370246410369873e-07, "grpo_std_advantage": 2.8908377771585947e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.539706826210022}
+{"ts": "2025-12-27T19:53:48", "step": 30, "epoch": 0.1201, "grpo_mean_advantage": 6.705522359595761e-09, "grpo_std_advantage": 6.189450800775376e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5812538862228394}
+{"ts": "2025-12-27T19:55:13", "step": 35, "epoch": 0.1401, "grpo_mean_advantage": 3.859400692363124e-07, "grpo_std_advantage": 1.6833292875162442e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5909844636917114}
+{"ts": "2025-12-27T19:56:38", "step": 40, "epoch": 0.1602, "grpo_mean_advantage": 2.600252742013254e-07, "grpo_std_advantage": 1.4095899132371414e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5630953907966614}
+{"ts": "2025-12-27T19:58:00", "step": 45, "epoch": 0.1802, "grpo_mean_advantage": -1.2591480924584175e-07, "grpo_std_advantage": 1.0309080380466185e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5604403614997864}
+{"ts": "2025-12-27T19:59:27", "step": 50, "epoch": 0.2002, "grpo_mean_advantage": -2.808868941883702e-07, "grpo_std_advantage": 1.5696078889959608e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5971035957336426}
+{"ts": "2025-12-27T20:00:53", "step": 55, "epoch": 0.2202, "grpo_mean_advantage": 2.6822089438383045e-08, "grpo_std_advantage": 3.7878271541558206e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5892971754074097}
+{"ts": "2025-12-27T20:02:17", "step": 60, "epoch": 0.2402, "grpo_mean_advantage": -5.662441182607836e-08, "grpo_std_advantage": 6.128998393251095e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.564322292804718}
+{"ts": "2025-12-27T20:03:45", "step": 65, "epoch": 0.2603, "grpo_mean_advantage": -1.5944242193199898e-07, "grpo_std_advantage": 1.6374274309782777e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.562497615814209}
+{"ts": "2025-12-27T20:05:10", "step": 70, "epoch": 0.2803, "grpo_mean_advantage": 1.6838312433264946e-07, "grpo_std_advantage": 8.536571272088622e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5904761552810669}
+{"ts": "2025-12-27T20:06:36", "step": 75, "epoch": 0.3003, "grpo_mean_advantage": 1.1175870895385742e-07, "grpo_std_advantage": 6.451961667153228e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5765624046325684}
+{"ts": "2025-12-27T20:08:03", "step": 80, "epoch": 0.3203, "grpo_mean_advantage": -1.4603138254187797e-07, "grpo_std_advantage": 1.1309343790344428e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5858271718025208}
+{"ts": "2025-12-27T20:09:31", "step": 85, "epoch": 0.3403, "grpo_mean_advantage": -1.817941665649414e-06, "grpo_std_advantage": 1.1141768482048064e-05, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5871662497520447}
+{"ts": "2025-12-27T20:10:55", "step": 90, "epoch": 0.3604, "grpo_mean_advantage": 1.8179416372277046e-07, "grpo_std_advantage": 6.210335072864837e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5330992937088013}
+{"ts": "2025-12-27T20:12:19", "step": 95, "epoch": 0.3804, "grpo_mean_advantage": -2.972781771859445e-07, "grpo_std_advantage": 3.1582342217006953e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5265295505523682}
+{"ts": "2025-12-27T20:13:46", "step": 100, "epoch": 0.4004, "grpo_mean_advantage": -7.033348197182931e-07, "grpo_std_advantage": 4.245831405569334e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5660771131515503}
+{"ts": "2025-12-27T20:15:15", "step": 105, "epoch": 0.4204, "grpo_mean_advantage": 1.1920928955078125e-07, "grpo_std_advantage": 3.2809634831210133e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.57631915807724}
+{"ts": "2025-12-27T20:16:39", "step": 110, "epoch": 0.4404, "grpo_mean_advantage": -4.0978193283081055e-07, "grpo_std_advantage": 6.0397578636184335e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.546563982963562}
+{"ts": "2025-12-27T20:18:01", "step": 115, "epoch": 0.4605, "grpo_mean_advantage": -1.467764434437413e-07, "grpo_std_advantage": 2.2689375782647403e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5519219636917114}
+{"ts": "2025-12-27T20:19:25", "step": 120, "epoch": 0.4805, "grpo_mean_advantage": -5.215406329028838e-09, "grpo_std_advantage": 7.929010621410271e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5490407943725586}
+{"ts": "2025-12-27T20:20:48", "step": 125, "epoch": 0.5005, "grpo_mean_advantage": -5.7369469175228005e-08, "grpo_std_advantage": 1.2823379620385822e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5646580457687378}
+{"ts": "2025-12-27T20:22:09", "step": 130, "epoch": 0.5205, "grpo_mean_advantage": 2.9876827056796174e-07, "grpo_std_advantage": 1.0496698905626545e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.6111599802970886}
+{"ts": "2025-12-27T20:23:32", "step": 135, "epoch": 0.5405, "grpo_mean_advantage": 1.5869736103013565e-07, "grpo_std_advantage": 1.2748531617035042e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5619662404060364}
+{"ts": "2025-12-27T20:24:57", "step": 140, "epoch": 0.5606, "grpo_mean_advantage": 3.0100346748440643e-07, "grpo_std_advantage": 2.4499684059264837e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5795454978942871}
+{"ts": "2025-12-27T20:26:19", "step": 145, "epoch": 0.5806, "grpo_mean_advantage": -3.233552092751779e-07, "grpo_std_advantage": 1.248456669600273e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5804953575134277}
+{"ts": "2025-12-27T20:27:43", "step": 150, "epoch": 0.6006, "grpo_mean_advantage": 3.2261013416245987e-07, "grpo_std_advantage": 1.4773489738217904e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5628539323806763}
+{"ts": "2025-12-27T20:29:07", "step": 155, "epoch": 0.6206, "grpo_mean_advantage": -2.5331974029541016e-07, "grpo_std_advantage": 1.5092309695319273e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5727725625038147}
+{"ts": "2025-12-27T20:30:30", "step": 160, "epoch": 0.6406, "grpo_mean_advantage": -6.780028627417778e-08, "grpo_std_advantage": 8.550978805033083e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5833909511566162}
+{"ts": "2025-12-27T20:31:51", "step": 165, "epoch": 0.6607, "grpo_mean_advantage": -5.587935447692871e-08, "grpo_std_advantage": 3.564579174053506e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5742615461349487}
+{"ts": "2025-12-27T20:33:17", "step": 170, "epoch": 0.6807, "grpo_mean_advantage": -5.327165126800537e-07, "grpo_std_advantage": 2.309018327650847e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5758188962936401}
+{"ts": "2025-12-27T20:34:40", "step": 175, "epoch": 0.7007, "grpo_mean_advantage": 5.863606702405377e-07, "grpo_std_advantage": 2.4449204829579685e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5767683982849121}
+{"ts": "2025-12-27T20:36:09", "step": 180, "epoch": 0.7207, "grpo_mean_advantage": 3.2186508747145126e-07, "grpo_std_advantage": 2.293551688126172e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.586772084236145}
+{"ts": "2025-12-27T20:37:33", "step": 185, "epoch": 0.7407, "grpo_mean_advantage": -4.470348358154297e-08, "grpo_std_advantage": 3.7067667335577426e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.549396276473999}
+{"ts": "2025-12-27T20:38:56", "step": 190, "epoch": 0.7608, "grpo_mean_advantage": -2.1010637851759384e-07, "grpo_std_advantage": 1.1695076409523608e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5798425078392029}
+{"ts": "2025-12-27T20:40:21", "step": 195, "epoch": 0.7808, "grpo_mean_advantage": 1.765787658314366e-07, "grpo_std_advantage": 2.429934738756856e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5584167838096619}
+{"ts": "2025-12-27T20:41:46", "step": 200, "epoch": 0.8008, "grpo_mean_advantage": 1.6540289493605087e-07, "grpo_std_advantage": 2.6342788714828203e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5676193237304688}
+{"ts": "2025-12-27T20:43:13", "step": 205, "epoch": 0.8208, "grpo_mean_advantage": -1.0944902442133753e-06, "grpo_std_advantage": 5.346942998585291e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5669739842414856}
+{"ts": "2025-12-27T20:44:38", "step": 210, "epoch": 0.8408, "grpo_mean_advantage": 2.4065374759629776e-07, "grpo_std_advantage": 1.6327536513927043e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5922158360481262}
+{"ts": "2025-12-27T20:46:02", "step": 215, "epoch": 0.8609, "grpo_mean_advantage": -5.21540641784668e-08, "grpo_std_advantage": 5.847922466273303e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5473950505256653}
+{"ts": "2025-12-27T20:47:28", "step": 220, "epoch": 0.8809, "grpo_mean_advantage": 6.541609991472797e-07, "grpo_std_advantage": 4.072162937518442e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5880032777786255}
+{"ts": "2025-12-27T20:48:53", "step": 225, "epoch": 0.9009, "grpo_mean_advantage": -1.2218951894737984e-07, "grpo_std_advantage": 4.386006935419573e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5835092663764954}
+{"ts": "2025-12-27T20:50:20", "step": 230, "epoch": 0.9209, "grpo_mean_advantage": 1.7605722177904681e-06, "grpo_std_advantage": 8.007580618141219e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5394966006278992}
+{"ts": "2025-12-27T20:51:41", "step": 235, "epoch": 0.9409, "grpo_mean_advantage": -3.3080578987210174e-07, "grpo_std_advantage": 1.551636614749441e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5687432289123535}
+{"ts": "2025-12-27T20:53:06", "step": 240, "epoch": 0.961, "grpo_mean_advantage": 2.712011450967111e-07, "grpo_std_advantage": 1.4400844747797237e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5550583600997925}
+{"ts": "2025-12-27T20:54:27", "step": 245, "epoch": 0.981, "grpo_mean_advantage": -3.2857059295565705e-07, "grpo_std_advantage": 2.105091425619321e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.558111310005188}
+{"ts": "2025-12-27T20:55:45", "step": 250, "epoch": 1.0, "grpo_mean_advantage": 4.470348358154297e-08, "grpo_std_advantage": 5.315724820320611e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.6196198463439941}
+{"ts": "2025-12-27T20:57:06", "step": 255, "epoch": 1.02, "grpo_mean_advantage": 9.290873776990338e-07, "grpo_std_advantage": 4.219644324621186e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.582168459892273}
+{"ts": "2025-12-27T20:58:26", "step": 260, "epoch": 1.04, "grpo_mean_advantage": 2.533197474008375e-08, "grpo_std_advantage": 1.6600588992332632e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5551307797431946}
+{"ts": "2025-12-27T20:59:48", "step": 265, "epoch": 1.0601, "grpo_mean_advantage": -5.662441182607836e-08, "grpo_std_advantage": 1.0909400316450046e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.535040020942688}
+{"ts": "2025-12-27T21:01:15", "step": 270, "epoch": 1.0801, "grpo_mean_advantage": -9.536743306171047e-08, "grpo_std_advantage": 5.838213610331877e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5673571825027466}
+{"ts": "2025-12-27T21:02:37", "step": 275, "epoch": 1.1001, "grpo_mean_advantage": 3.278255533700758e-08, "grpo_std_advantage": 9.317170679423725e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5874732732772827}
+{"ts": "2025-12-27T21:03:59", "step": 280, "epoch": 1.1201, "grpo_mean_advantage": -1.206994113545079e-07, "grpo_std_advantage": 6.201085511747806e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5569106340408325}
+{"ts": "2025-12-27T21:05:19", "step": 285, "epoch": 1.1401, "grpo_mean_advantage": 4.470348358154297e-08, "grpo_std_advantage": 6.115651558502577e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5578873157501221}
+{"ts": "2025-12-27T21:06:43", "step": 290, "epoch": 1.1602, "grpo_mean_advantage": -3.3453108017056365e-07, "grpo_std_advantage": 3.5326345368957845e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5735999345779419}
+{"ts": "2025-12-27T21:08:10", "step": 295, "epoch": 1.1802, "grpo_mean_advantage": -1.110136480519941e-07, "grpo_std_advantage": 4.731904823529476e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5626259446144104}
+{"ts": "2025-12-27T21:09:34", "step": 300, "epoch": 1.2002, "grpo_mean_advantage": -5.08874677507265e-07, "grpo_std_advantage": 1.840126174101897e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5463050603866577}
+{"ts": "2025-12-27T21:10:59", "step": 305, "epoch": 1.2202, "grpo_mean_advantage": 1.01327898960335e-07, "grpo_std_advantage": 7.798533943059738e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5352144241333008}
+{"ts": "2025-12-27T21:12:25", "step": 310, "epoch": 1.2402, "grpo_mean_advantage": 1.341104507446289e-07, "grpo_std_advantage": 7.821902840987605e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5547868013381958}
+{"ts": "2025-12-27T21:13:52", "step": 315, "epoch": 1.2603, "grpo_mean_advantage": 9.015202806494926e-08, "grpo_std_advantage": 1.0693488547985908e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5859472751617432}
+{"ts": "2025-12-27T21:15:21", "step": 320, "epoch": 1.2803, "grpo_mean_advantage": -2.443790378947597e-07, "grpo_std_advantage": 1.183122208203713e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5751550793647766}
+{"ts": "2025-12-27T21:16:47", "step": 325, "epoch": 1.3003, "grpo_mean_advantage": -6.705522537231445e-08, "grpo_std_advantage": 6.109748937888071e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5497723817825317}
+{"ts": "2025-12-27T21:18:13", "step": 330, "epoch": 1.3203, "grpo_mean_advantage": -1.639127766850379e-08, "grpo_std_advantage": 5.529495297196263e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.55989670753479}
+{"ts": "2025-12-27T21:19:37", "step": 335, "epoch": 1.3403, "grpo_mean_advantage": 4.418194237132411e-07, "grpo_std_advantage": 2.9275292945385445e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5809233784675598}
+{"ts": "2025-12-27T21:20:59", "step": 340, "epoch": 1.3604, "grpo_mean_advantage": 9.685754776000977e-08, "grpo_std_advantage": 3.754235251562932e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5568087100982666}
+{"ts": "2025-12-27T21:22:27", "step": 345, "epoch": 1.3804, "grpo_mean_advantage": -2.384185791015625e-07, "grpo_std_advantage": 6.821086913078034e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5655568838119507}
+{"ts": "2025-12-27T21:23:49", "step": 350, "epoch": 1.4004, "grpo_mean_advantage": -8.717179156292332e-08, "grpo_std_advantage": 2.500940354366321e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.6089578866958618}
+{"ts": "2025-12-27T21:25:13", "step": 355, "epoch": 1.4204, "grpo_mean_advantage": 2.1606683731079102e-07, "grpo_std_advantage": 1.4568390724889468e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.6129671335220337}
+{"ts": "2025-12-27T21:26:36", "step": 360, "epoch": 1.4404, "grpo_mean_advantage": -3.725290298461914e-09, "grpo_std_advantage": 2.965894054796081e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5562310814857483}
+{"ts": "2025-12-27T21:27:58", "step": 365, "epoch": 1.4605, "grpo_mean_advantage": 4.313886279305734e-07, "grpo_std_advantage": 1.9621948013082147e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5884170532226562}
+{"ts": "2025-12-27T21:29:21", "step": 370, "epoch": 1.4805, "grpo_mean_advantage": 2.0489096641540527e-07, "grpo_std_advantage": 1.0235522722723545e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5795440673828125}
+{"ts": "2025-12-27T21:30:45", "step": 375, "epoch": 1.5005, "grpo_mean_advantage": -1.4185905001795618e-06, "grpo_std_advantage": 1.0947338523692451e-05, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5607603788375854}
+{"ts": "2025-12-27T21:32:13", "step": 380, "epoch": 1.5205, "grpo_mean_advantage": -1.758337049295733e-07, "grpo_std_advantage": 9.663675655247062e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5381432771682739}
+{"ts": "2025-12-27T21:33:36", "step": 385, "epoch": 1.5405, "grpo_mean_advantage": -6.973743325033865e-07, "grpo_std_advantage": 4.341973180999048e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5528443455696106}
+{"ts": "2025-12-27T21:34:59", "step": 390, "epoch": 1.5606, "grpo_mean_advantage": 1.7881394143159923e-08, "grpo_std_advantage": 1.3004198251564958e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.6091476678848267}
+{"ts": "2025-12-27T21:36:21", "step": 395, "epoch": 1.5806, "grpo_mean_advantage": 3.4868716625169327e-07, "grpo_std_advantage": 2.059372718576924e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5397372245788574}
+{"ts": "2025-12-27T21:37:48", "step": 400, "epoch": 1.6006, "grpo_mean_advantage": -2.1636485598719446e-06, "grpo_std_advantage": 9.725940799398813e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5873125195503235}
+{"ts": "2025-12-27T21:39:11", "step": 405, "epoch": 1.6206, "grpo_mean_advantage": -5.960464477539063e-08, "grpo_std_advantage": 3.460792754594877e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5601426362991333}
+{"ts": "2025-12-27T21:40:38", "step": 410, "epoch": 1.6406, "grpo_mean_advantage": 2.6226044269606064e-07, "grpo_std_advantage": 7.928817922220333e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.578656792640686}
+{"ts": "2025-12-27T21:42:02", "step": 415, "epoch": 1.6607, "grpo_mean_advantage": 2.3558736756967846e-06, "grpo_std_advantage": 1.4469559573626611e-05, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5885810852050781}
+{"ts": "2025-12-27T21:43:27", "step": 420, "epoch": 1.6807, "grpo_mean_advantage": 1.639127766850379e-08, "grpo_std_advantage": 9.352411325380672e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5805023312568665}
+{"ts": "2025-12-27T21:44:52", "step": 425, "epoch": 1.7007, "grpo_mean_advantage": 3.2387674764322583e-06, "grpo_std_advantage": 1.999079904635437e-05, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5655918121337891}
+{"ts": "2025-12-27T21:46:17", "step": 430, "epoch": 1.7207, "grpo_mean_advantage": 3.5464762504489045e-07, "grpo_std_advantage": 1.7663603557593888e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5474504232406616}
+{"ts": "2025-12-27T21:47:43", "step": 435, "epoch": 1.7407, "grpo_mean_advantage": 3.6135315895080566e-07, "grpo_std_advantage": 2.356920958845876e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5874254703521729}
+{"ts": "2025-12-27T21:49:08", "step": 440, "epoch": 1.7608, "grpo_mean_advantage": 2.7567148563889532e-08, "grpo_std_advantage": 9.97340521280421e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5815118551254272}
+{"ts": "2025-12-27T21:50:35", "step": 445, "epoch": 1.7808, "grpo_mean_advantage": -8.34465012644614e-08, "grpo_std_advantage": 5.558832185670326e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5611211061477661}
+{"ts": "2025-12-27T21:52:06", "step": 450, "epoch": 1.8008, "grpo_mean_advantage": -1.9818544672034477e-07, "grpo_std_advantage": 6.800727305744658e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.590424656867981}
+{"ts": "2025-12-27T21:53:32", "step": 455, "epoch": 1.8208, "grpo_mean_advantage": -1.9371508841459217e-08, "grpo_std_advantage": 3.142378943721269e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5670351982116699}
+{"ts": "2025-12-27T21:55:00", "step": 460, "epoch": 1.8408, "grpo_mean_advantage": 2.3692845729783585e-07, "grpo_std_advantage": 1.682946731307311e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5640432834625244}
+{"ts": "2025-12-27T21:56:27", "step": 465, "epoch": 1.8609, "grpo_mean_advantage": 1.110136480519941e-07, "grpo_std_advantage": 8.930008448260196e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5584251284599304}
+{"ts": "2025-12-27T21:57:50", "step": 470, "epoch": 1.8809, "grpo_mean_advantage": 2.5406478698641877e-07, "grpo_std_advantage": 9.93092498902115e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5392154455184937}
+{"ts": "2025-12-27T21:59:15", "step": 475, "epoch": 1.9009, "grpo_mean_advantage": -8.940697071579962e-09, "grpo_std_advantage": 5.835169645251881e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5686308741569519}
+{"ts": "2025-12-27T22:00:36", "step": 480, "epoch": 1.9209, "grpo_mean_advantage": 4.0605664253234863e-07, "grpo_std_advantage": 2.3210795916384086e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5842767357826233}
+{"ts": "2025-12-27T22:01:58", "step": 485, "epoch": 1.9409, "grpo_mean_advantage": 1.341104507446289e-07, "grpo_std_advantage": 1.507950400991831e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5495311617851257}
+{"ts": "2025-12-27T22:03:20", "step": 490, "epoch": 1.961, "grpo_mean_advantage": 2.689659481802664e-07, "grpo_std_advantage": 8.491958851664094e-07, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.549436628818512}
+{"ts": "2025-12-27T22:04:42", "step": 495, "epoch": 1.981, "grpo_mean_advantage": 8.195638656616211e-08, "grpo_std_advantage": 3.802849732892355e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.544632077217102}
+{"ts": "2025-12-27T22:06:01", "step": 500, "epoch": 2.0, "grpo_mean_advantage": 7.552536089860951e-07, "grpo_std_advantage": 4.143997102801222e-06, "grpo_mean_kl_div": 0.0, "grpo_mean_group_score": 0.5968535542488098}

wandb/debug-internal.log

@@ -1,12 +1,12 @@
-{"time":"2025-12-26T18:15:45.044149374Z","level":"INFO","msg":"stream: starting","core version":"0.23.1"}
-{"time":"2025-12-26T18:15:45.212148231Z","level":"INFO","msg":"stream: created new stream","id":"upub1jan"}
-{"time":"2025-12-26T18:15:45.212312297Z","level":"INFO","msg":"handler: started","stream_id":"upub1jan"}
-{"time":"2025-12-26T18:15:45.212463318Z","level":"INFO","msg":"stream: started","id":"upub1jan"}
-{"time":"2025-12-26T18:15:45.212498387Z","level":"INFO","msg":"writer: started","stream_id":"upub1jan"}
-{"time":"2025-12-26T18:15:45.212503642Z","level":"INFO","msg":"sender: started","stream_id":"upub1jan"}
-{"time":"2025-12-27T15:02:32.177112089Z","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"}
-{"time":"2025-12-27T15:02:32.275585886Z","level":"INFO","msg":"handler: operation stats","stats":{}}
-{"time":"2025-12-27T15:02:32.280819216Z","level":"INFO","msg":"stream: closing","id":"upub1jan"}
-{"time":"2025-12-27T15:02:32.280845154Z","level":"INFO","msg":"handler: closed","stream_id":"upub1jan"}
-{"time":"2025-12-27T15:02:32.280896164Z","level":"INFO","msg":"sender: closed","stream_id":"upub1jan"}
-{"time":"2025-12-27T15:02:32.280915121Z","level":"INFO","msg":"stream: closed","id":"upub1jan"}
+{"time":"2025-12-27T19:44:23.778699792Z","level":"INFO","msg":"stream: starting","core version":"0.23.1"}
+{"time":"2025-12-27T19:44:23.931692267Z","level":"INFO","msg":"stream: created new stream","id":"jz7bptqa"}
+{"time":"2025-12-27T19:44:23.931816408Z","level":"INFO","msg":"handler: started","stream_id":"jz7bptqa"}
+{"time":"2025-12-27T19:44:23.931950499Z","level":"INFO","msg":"stream: started","id":"jz7bptqa"}
+{"time":"2025-12-27T19:44:23.931981018Z","level":"INFO","msg":"writer: started","stream_id":"jz7bptqa"}
+{"time":"2025-12-27T19:44:23.931980519Z","level":"INFO","msg":"sender: started","stream_id":"jz7bptqa"}
+{"time":"2025-12-27T22:06:04.101056437Z","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"}
+{"time":"2025-12-27T22:06:04.298272513Z","level":"INFO","msg":"handler: operation stats","stats":{}}
+{"time":"2025-12-27T22:06:04.303656494Z","level":"INFO","msg":"stream: closing","id":"jz7bptqa"}
+{"time":"2025-12-27T22:06:04.303677603Z","level":"INFO","msg":"handler: closed","stream_id":"jz7bptqa"}
+{"time":"2025-12-27T22:06:04.303750712Z","level":"INFO","msg":"sender: closed","stream_id":"jz7bptqa"}
+{"time":"2025-12-27T22:06:04.303767265Z","level":"INFO","msg":"stream: closed","id":"jz7bptqa"}

wandb/debug.log

@@ -1,29 +1,29 @@
-2025-12-26 18:15:44,766 INFO    MainThread:194421 [wandb_setup.py:_flush():80] Current SDK version is 0.23.1
-2025-12-26 18:15:44,767 INFO    MainThread:194421 [wandb_setup.py:_flush():80] Configure stats pid to 194421
-2025-12-26 18:15:44,767 INFO    MainThread:194421 [wandb_setup.py:_flush():80] Loading settings from /root/.config/wandb/settings
-2025-12-26 18:15:44,767 INFO    MainThread:194421 [wandb_setup.py:_flush():80] Loading settings from /workspace/wandb/settings
-2025-12-26 18:15:44,767 INFO    MainThread:194421 [wandb_setup.py:_flush():80] Loading settings from environment variables
-2025-12-26 18:15:44,767 INFO    MainThread:194421 [wandb_init.py:setup_run_log_directory():714] Logging user logs to task2file/sft_qwen_14B_v2/wandb/run-20251226_181544-upub1jan/logs/debug.log
-2025-12-26 18:15:44,767 INFO    MainThread:194421 [wandb_init.py:setup_run_log_directory():715] Logging internal logs to task2file/sft_qwen_14B_v2/wandb/run-20251226_181544-upub1jan/logs/debug-internal.log
-2025-12-26 18:15:44,767 INFO    MainThread:194421 [wandb_init.py:init():841] calling init triggers
-2025-12-26 18:15:44,767 INFO    MainThread:194421 [wandb_init.py:init():846] wandb.init called with sweep_config: {}
-config: {'model': {'repo_id': './Models/Qwen2.5-Coder-14B-CPT', 'revision': None, 'base_local_dir': 'base_model', 'trust_remote_code': True, 'tokenizer_use_fast': True, 'device_map': 'auto', 'torch_dtype': 'bfloat16', 'use_4bit': False, 'bnb_4bit_quant_type': 'nf4', 'bnb_4bit_use_double_quant': False, 'bnb_4bit_compute_dtype': 'bfloat16', 'attn_implementation': None}, 'data': {'train_jsonl': 'sft_dataset.jsonl', 'eval_jsonl': None, 'eval_split_ratio': 0.1, 'instruction_field': 'instruction', 'input_field': 'input', 'output_field': 'output', 'format_type': 'custom', 'system_prompt': 'You are a Hyperswitch Rust code analyzer. Identify functions/structs that need modification for a given task.\n\n## Output Format\n\n##OUTPUT\nExplain the data flow and why each component must change:\n- Flow: [Input → Processing → Output with arrows]\n- For each component: "The [ComponentName] ([path]) must [action] because [reason]—without this, [consequence]"\n- Explain coupling between components\n\n##SELECT\nmodify::crates/path/to/file.rs::impl::ComponentName\nadd::crates/another/file.rs::function::AnotherComponent\n<EOS>\n\n## Rules\n\n1. Use full paths: `remove::crates/folder/file.rs::Type::Name`\n2. Use `::` for nested items: `status::StructName::Type::Name`\n3. Always explain "must change because" and "without this"\n3. Types of components: function, struct, enum, impl, trait\n4. If there is extra information (e.g., enum variants), include that too.\n5. Start with ##OUTPUT, end with ##SELECT, terminate with <EOS>\n\n## Example\n\n##TASK\nAdd webhook subscription support\n\n##OUTPUT\nThe webhook system routes events via EventClass enum. Flow: webhook → EventClass → handler → processing. The EventClass enum (crates/common_enums/src/enums.rs::EventClass) must add Subscriptions variant because it defines event routing—without this, subscription events cannot be processed. The SubscriptionStatus impl (crates/common_enums/src/transformers.rs::SubscriptionStatus) must map to EventType because it converts status to events—without this, status changes don\'t trigger webhooks. These are coupled: EventClass routes to handlers that use SubscriptionStatus mappings.\n\n##SELECT\ncrates/common_enums/src/enums.rs::EventClass\ncrates/common_enums/src/transformers.rs::SubscriptionStatus\n<EOS>\n', 'custom_template': '##INSTRUCTION\n{instruction}<|im_end|>\n{input}<|im_end|>\n{output}<|im_end|>', 'max_length': 2048, 'shuffle': True, 'num_proc': 4}, 'peft': {'enabled': True, 'r': 32, 'lora_alpha': 64, 'lora_dropout': 0.05, 'bias': 'none', 'target_modules': 'auto'}, 'train': {'num_train_epochs': 6, 'per_device_train_batch_size': 1, 'per_device_eval_batch_size': 1, 'gradient_accumulation_steps': 8, 'learning_rate': '2e-4', 'weight_decay': 0.0, 'warmup_ratio': 0.08, 'lr_scheduler_type': 'cosine', 'optim': 'adamw_torch', 'max_grad_norm': 1.0, 'gradient_checkpointing': True, 'logging_steps': 2, 'save_strategy': 'steps', 'save_steps': 500, 'save_total_limit': 20, 'evaluation_strategy': 'steps', 'eval_steps': 100, 'load_best_model_at_end': True, 'early_stopping': {'enabled': True, 'patience': 5, 'min_delta': 0.001, 'metric': 'eval_loss', 'mode': 'min'}, 'resume_from_checkpoint': 'auto'}, 'run_dir': 'task2file/sft_qwen_14B_v2', '_wandb': {}}
-2025-12-26 18:15:44,767 INFO    MainThread:194421 [wandb_init.py:init():889] starting backend
-2025-12-26 18:15:45,035 INFO    MainThread:194421 [wandb_init.py:init():892] sending inform_init request
-2025-12-26 18:15:45,040 INFO    MainThread:194421 [wandb_init.py:init():900] backend started and connected
-2025-12-26 18:15:45,044 INFO    MainThread:194421 [wandb_init.py:init():970] updated telemetry
-2025-12-26 18:15:45,045 INFO    MainThread:194421 [wandb_init.py:init():994] communicating run to backend with 90.0 second timeout
-2025-12-26 18:15:45,420 INFO    MainThread:194421 [wandb_init.py:init():1041] starting run threads in backend
-2025-12-26 18:15:45,537 INFO    MainThread:194421 [wandb_run.py:_console_start():2521] atexit reg
-2025-12-26 18:15:45,537 INFO    MainThread:194421 [wandb_run.py:_redirect():2369] redirect: wrap_raw
-2025-12-26 18:15:45,537 INFO    MainThread:194421 [wandb_run.py:_redirect():2438] Wrapping output streams.
-2025-12-26 18:15:45,537 INFO    MainThread:194421 [wandb_run.py:_redirect():2461] Redirects installed.
-2025-12-26 18:15:45,542 INFO    MainThread:194421 [wandb_init.py:init():1081] run started, returning control to user process
-2025-12-26 18:17:10,652 INFO    MainThread:194421 [wandb_run.py:_config_callback():1396] config_cb None None {'peft_config': {'default': {'task_type': 'CAUSAL_LM', 'peft_type': 'LORA', 'auto_mapping': None, 'peft_version': '0.18.0', 'base_model_name_or_path': 'Models/Qwen2.5-Coder-14B-CPT', 'revision': None, 'inference_mode': False, 'r': 32, 'target_modules': ['o_proj', 'v_proj', 'k_proj', 'q_proj'], 'exclude_modules': None, 'lora_alpha': 64, 'lora_dropout': 0.05, 'fan_in_fan_out': False, 'bias': 'none', 'use_rslora': False, 'modules_to_save': None, 'init_lora_weights': True, 'layers_to_transform': None, 'layers_pattern': None, 'rank_pattern': {}, 'alpha_pattern': {}, 'megatron_config': None, 'megatron_core': 'megatron.core', 'trainable_token_indices': None, 'loftq_config': {}, 'eva_config': None, 'corda_config': None, 'use_dora': False, 'alora_invocation_tokens': None, 'use_qalora': False, 'qalora_group_size': 16, 'layer_replication': None, 'runtime_config': {'ephemeral_gpu_offload': False}, 'lora_bias': False, 'target_parameters': None, 'arrow_config': None, 'ensure_weight_tying': False}}, 'vocab_size': 152064, 'max_position_embeddings': 32768, 'hidden_size': 5120, 'intermediate_size': 13824, 'num_hidden_layers': 48, 'num_attention_heads': 40, 'use_sliding_window': False, 'sliding_window': None, 'max_window_layers': 48, 'num_key_value_heads': 8, 'hidden_act': 'silu', 'initializer_range': 0.02, 'rms_norm_eps': 1e-06, 'use_cache': False, 'attention_dropout': 0.0, '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', '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'], 'rope_parameters': {'rope_theta': 1000000.0, 'rope_type': 'default'}, 'return_dict': True, 'output_hidden_states': False, 'dtype': 'bfloat16', 'tie_word_embeddings': False, 'chunk_size_feed_forward': 0, 'is_encoder_decoder': False, 'is_decoder': False, 'cross_attention_hidden_size': None, 'add_cross_attention': False, 'architectures': ['Qwen2ForCausalLM'], 'finetuning_task': None, 'id2label': {0: 'LABEL_0', 1: 'LABEL_1'}, 'label2id': {'LABEL_0': 0, 'LABEL_1': 1}, 'task_specific_params': None, 'problem_type': None, 'tokenizer_class': None, 'prefix': None, 'bos_token_id': 151643, 'pad_token_id': None, 'eos_token_id': 151643, 'sep_token_id': None, 'decoder_start_token_id': None, '_name_or_path': 'Models/Qwen2.5-Coder-14B-CPT', 'transformers_version': '5.0.0.dev0', 'model_type': 'qwen2', 'output_attentions': False, 'output_dir': 'task2file/sft_qwen_14B_v2/checkpoints', 'do_train': False, 'do_eval': True, 'do_predict': False, 'eval_strategy': 'steps', 'prediction_loss_only': False, 'per_device_train_batch_size': 1, 'per_device_eval_batch_size': 1, 'gradient_accumulation_steps': 8, 'eval_accumulation_steps': None, 'eval_delay': 0, 'torch_empty_cache_steps': None, 'learning_rate': 0.0002, 'weight_decay': 0.0, 'adam_beta1': 0.9, 'adam_beta2': 0.999, 'adam_epsilon': 1e-08, 'max_grad_norm': 1.0, 'num_train_epochs': 6.0, 'max_steps': -1, 'lr_scheduler_type': 'cosine', 'lr_scheduler_kwargs': None, 'warmup_ratio': 0.08, 'warmup_steps': 0.08, 'log_level': 'passive', 'log_level_replica': 'warning', 'log_on_each_node': True, 'logging_dir': None, 'logging_strategy': 'steps', 'logging_first_step': False, 'logging_steps': 2, 'logging_nan_inf_filter': True, 'save_strategy': 'steps', 'save_steps': 500, 'save_total_limit': 20, 'enable_jit_checkpoint': False, 'save_on_each_node': False, 'save_only_model': False, 'restore_callback_states_from_checkpoint': False, 'use_cpu': False, 'seed': 42, 'data_seed': None, 'bf16': True, 'fp16': False, 'bf16_full_eval': False, 'fp16_full_eval': False, 'tf32': None, 'local_rank': -1, 'ddp_backend': None, 'debug': [], 'dataloader_drop_last': False, 'eval_steps': 100, 'dataloader_num_workers': 0, 'dataloader_prefetch_factor': None, 'run_name': None, 'disable_tqdm': False, 'remove_unused_columns': False, 'label_names': None, 'load_best_model_at_end': True, 'metric_for_best_model': 'eval_loss', 'greater_is_better': False, 'ignore_data_skip': False, 'fsdp': [], 'fsdp_config': {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}, 'accelerator_config': {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}, 'parallelism_config': None, 'deepspeed': None, 'label_smoothing_factor': 0.0, 'optim': 'adamw_torch', 'optim_args': None, 'group_by_length': False, 'length_column_name': 'length', 'report_to': ['wandb'], 'project': 'huggingface', 'trackio_space_id': 'trackio', 'ddp_find_unused_parameters': None, 'ddp_bucket_cap_mb': None, 'ddp_broadcast_buffers': None, 'dataloader_pin_memory': True, 'dataloader_persistent_workers': False, 'skip_memory_metrics': True, 'push_to_hub': False, 'resume_from_checkpoint': None, 'hub_model_id': None, 'hub_strategy': 'every_save', 'hub_token': '<HUB_TOKEN>', 'hub_private_repo': None, 'hub_always_push': False, 'hub_revision': None, 'gradient_checkpointing': False, 'gradient_checkpointing_kwargs': None, 'include_for_metrics': [], 'eval_do_concat_batches': True, 'auto_find_batch_size': False, 'full_determinism': False, 'ddp_timeout': 1800, 'torch_compile': False, 'torch_compile_backend': None, 'torch_compile_mode': None, 'include_num_input_tokens_seen': 'no', 'neftune_noise_alpha': None, 'optim_target_modules': None, 'batch_eval_metrics': False, 'eval_on_start': False, 'use_liger_kernel': False, 'liger_kernel_config': None, 'eval_use_gather_object': False, 'average_tokens_across_devices': True}
-2025-12-26 18:17:10,660 INFO    MainThread:194421 [wandb_config.py:__setitem__():154] [no run ID] config set model/num_parameters = 14820365312 - <bound method Run._config_callback of <wandb.sdk.wandb_run.Run object at 0x7ed24f0556c0>>
-2025-12-26 18:17:10,660 INFO    MainThread:194421 [wandb_run.py:_config_callback():1396] config_cb model/num_parameters 14820365312 None
-2025-12-27 15:02:31,903 INFO    MainThread:194421 [wandb_run.py:_finish():2287] finishing run sirajuddin-shaik-007/sft-training/upub1jan
-2025-12-27 15:02:31,905 INFO    MainThread:194421 [wandb_run.py:_atexit_cleanup():2486] got exitcode: 0
-2025-12-27 15:02:31,906 INFO    MainThread:194421 [wandb_run.py:_restore():2468] restore
-2025-12-27 15:02:31,906 INFO    MainThread:194421 [wandb_run.py:_restore():2474] restore done
-2025-12-27 15:02:32,279 INFO    MainThread:194421 [wandb_run.py:_footer_sync_info():3862] logging synced files
+2025-12-27 19:44:23,485 INFO    MainThread:592889 [wandb_setup.py:_flush():80] Current SDK version is 0.23.1
+2025-12-27 19:44:23,486 INFO    MainThread:592889 [wandb_setup.py:_flush():80] Configure stats pid to 592889
+2025-12-27 19:44:23,486 INFO    MainThread:592889 [wandb_setup.py:_flush():80] Loading settings from /root/.config/wandb/settings
+2025-12-27 19:44:23,486 INFO    MainThread:592889 [wandb_setup.py:_flush():80] Loading settings from /workspace/trainer-kit/GRPO-14b/wandb/settings
+2025-12-27 19:44:23,486 INFO    MainThread:592889 [wandb_setup.py:_flush():80] Loading settings from environment variables
+2025-12-27 19:44:23,486 INFO    MainThread:592889 [wandb_init.py:setup_run_log_directory():714] Logging user logs to runs/grpo_14b_run1/wandb/run-20251227_194423-jz7bptqa/logs/debug.log
+2025-12-27 19:44:23,486 INFO    MainThread:592889 [wandb_init.py:setup_run_log_directory():715] Logging internal logs to runs/grpo_14b_run1/wandb/run-20251227_194423-jz7bptqa/logs/debug-internal.log
+2025-12-27 19:44:23,486 INFO    MainThread:592889 [wandb_init.py:init():841] calling init triggers
+2025-12-27 19:44:23,486 INFO    MainThread:592889 [wandb_init.py:init():846] wandb.init called with sweep_config: {}
+config: {'model': {'repo_id': '/workspace/Models/Qwen2.5-Coder-14B-CPT-SFT_v2', 'tokenizer_name': 'Qwen/Qwen2.5-Coder-14B', 'load_in_8bit': False, 'load_in_4bit': False, 'torch_dtype': 'bfloat16', 'device_map': 'auto', 'trust_remote_code': True}, 'data': {'train_jsonl': 'grpo_dataset.jsonl', 'eval_jsonl': None, 'eval_split_ratio': 0.0, 'shuffle': True, 'num_proc': 1, 'prompt_field': 'prompt', 'completions_field': 'completions', 'scores_field': 'scores', 'format_type': 'raw', 'max_length': 2048, 'min_completions': 2, 'system_prompt': 'You are a Hyperswitch Rust code analyzer. Identify functions/structs that need modification for a given task.\n\n## Output Format\n\n##OUTPUT\nExplain the data flow and why each component must change:\n- Flow: [Input → Processing → Output with arrows]\n- For each component: "The [ComponentName] ([path]) must [action] because [reason]—without this, [consequence]"\n- Explain coupling between components\n\n##SELECT\nmodify::crates/path/to/file.rs::impl::ComponentName\nadd::crates/another/file.rs::function::AnotherComponent\n<EOS>\n\n## Rules\n\n1. Use full paths: `remove::crates/folder/file.rs::Type::Name`\n2. Use `::` for nested items: `status::StructName::Type::Name`\n3. Always explain "must change because" and "without this"\n3. Types of components: function, struct, enum, impl, trait\n4. If there is extra information (e.g., enum variants), include that too.\n5. Start with ##OUTPUT, end with ##SELECT, terminate with <EOS>\n\n## Example\n\n##TASK\nAdd webhook subscription support\n\n##OUTPUT\nThe webhook system routes events via EventClass enum. Flow: webhook → EventClass → handler → processing. The EventClass enum (crates/common_enums/src/enums.rs::EventClass) must add Subscriptions variant because it defines event routing—without this, subscription events cannot be processed. The SubscriptionStatus impl (crates/common_enums/src/transformers.rs::SubscriptionStatus) must map to EventType because it converts status to events—without this, status changes don\'t trigger webhooks. These are coupled: EventClass routes to handlers that use SubscriptionStatus mappings.\n\n##SELECT\ncrates/common_enums/src/enums.rs::EventClass\ncrates/common_enums/src/transformers.rs::SubscriptionStatus\n<EOS>\n', 'custom_template': '##INSTRUCTION\n{instruction}<|im_end|>\n{input}<|im_end|>\n{output}<|im_end|>'}, 'peft': {'enabled': True, 'r': 16, 'lora_alpha': 32, 'lora_dropout': 0.05, 'target_modules': ['q_proj', 'k_proj', 'v_proj', 'o_proj', 'gate_proj', 'up_proj', 'down_proj'], 'bias': 'none', 'task_type': 'CAUSAL_LM'}, 'grpo': {'group_size': 4, 'kl_coef': 0.05, 'normalize_advantages': True, 'reward_scaling': 1.0, 'reward_bias': 0.0, 'reward_clip': 5.0, 'advantage_temperature': 1.0, 'use_reference_model': False, 'seed': 42}, 'train': {'output_dir': 'runs/grpo_14b_run1', 'num_train_epochs': 2, 'per_device_train_batch_size': 1, 'gradient_accumulation_steps': 8, 'per_device_eval_batch_size': 1, 'learning_rate': 5e-06, 'weight_decay': 0.01, 'warmup_ratio': 0.05, 'lr_scheduler_type': 'cosine', 'fp16': False, 'bf16': True, 'max_grad_norm': 1.0, 'gradient_checkpointing': True, 'logging_steps': 5, 'save_steps': 100, 'save_total_limit': 2, 'evaluation_strategy': 'no', 'dataloader_num_workers': 4, 'dataloader_pin_memory': True, 'remove_unused_columns': False, 'report_to': [], 'seed': 42, 'ddp_find_unused_parameters': False}, 'run_dir': 'runs/grpo_14b_run1', '_wandb': {}}
+2025-12-27 19:44:23,486 INFO    MainThread:592889 [wandb_init.py:init():889] starting backend
+2025-12-27 19:44:23,772 INFO    MainThread:592889 [wandb_init.py:init():892] sending inform_init request
+2025-12-27 19:44:23,776 INFO    MainThread:592889 [wandb_init.py:init():900] backend started and connected
+2025-12-27 19:44:23,778 INFO    MainThread:592889 [wandb_init.py:init():970] updated telemetry
+2025-12-27 19:44:23,779 INFO    MainThread:592889 [wandb_init.py:init():994] communicating run to backend with 90.0 second timeout
+2025-12-27 19:44:24,241 INFO    MainThread:592889 [wandb_init.py:init():1041] starting run threads in backend
+2025-12-27 19:44:24,354 INFO    MainThread:592889 [wandb_run.py:_console_start():2521] atexit reg
+2025-12-27 19:44:24,354 INFO    MainThread:592889 [wandb_run.py:_redirect():2369] redirect: wrap_raw
+2025-12-27 19:44:24,354 INFO    MainThread:592889 [wandb_run.py:_redirect():2438] Wrapping output streams.
+2025-12-27 19:44:24,354 INFO    MainThread:592889 [wandb_run.py:_redirect():2461] Redirects installed.
+2025-12-27 19:44:24,358 INFO    MainThread:592889 [wandb_init.py:init():1081] run started, returning control to user process
+2025-12-27 19:45:32,787 INFO    MainThread:592889 [wandb_run.py:_config_callback():1396] config_cb None None {'peft_config': {'default': {'task_type': 'CAUSAL_LM', 'peft_type': 'LORA', 'auto_mapping': None, 'peft_version': '0.18.0', 'base_model_name_or_path': '/workspace/Models/Qwen2.5-Coder-14B-CPT-SFT_v2', 'revision': None, 'inference_mode': False, 'r': 16, 'target_modules': ['q_proj', 'o_proj', 'v_proj', 'up_proj', 'gate_proj', 'down_proj', 'k_proj'], 'exclude_modules': None, 'lora_alpha': 32, 'lora_dropout': 0.05, 'fan_in_fan_out': False, 'bias': 'none', 'use_rslora': False, 'modules_to_save': None, 'init_lora_weights': True, 'layers_to_transform': None, 'layers_pattern': None, 'rank_pattern': {}, 'alpha_pattern': {}, 'megatron_config': None, 'megatron_core': 'megatron.core', 'trainable_token_indices': None, 'loftq_config': {}, 'eva_config': None, 'corda_config': None, 'use_dora': False, 'alora_invocation_tokens': None, 'use_qalora': False, 'qalora_group_size': 16, 'layer_replication': None, 'runtime_config': {'ephemeral_gpu_offload': False}, 'lora_bias': False, 'target_parameters': None, 'arrow_config': None, 'ensure_weight_tying': False}}, 'vocab_size': 152064, 'max_position_embeddings': 32768, 'hidden_size': 5120, 'intermediate_size': 13824, 'num_hidden_layers': 48, 'num_attention_heads': 40, 'use_sliding_window': False, 'sliding_window': None, 'max_window_layers': 48, 'num_key_value_heads': 8, 'hidden_act': 'silu', 'initializer_range': 0.02, 'rms_norm_eps': 1e-06, 'use_cache': False, 'rope_theta': 10000.0, 'rope_scaling': None, 'attention_dropout': 0.0, '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', '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'], 'return_dict': True, 'output_hidden_states': False, 'torchscript': False, 'dtype': 'bfloat16', 'pruned_heads': {}, 'tie_word_embeddings': False, 'chunk_size_feed_forward': 0, 'is_encoder_decoder': False, 'is_decoder': False, 'cross_attention_hidden_size': None, 'add_cross_attention': False, 'tie_encoder_decoder': False, 'architectures': ['Qwen2ForCausalLM'], 'finetuning_task': None, 'id2label': {0: 'LABEL_0', 1: 'LABEL_1'}, 'label2id': {'LABEL_0': 0, 'LABEL_1': 1}, 'task_specific_params': None, 'problem_type': None, 'tokenizer_class': None, 'prefix': None, 'bos_token_id': 151643, 'pad_token_id': None, 'eos_token_id': 151643, 'sep_token_id': None, 'decoder_start_token_id': None, 'max_length': 20, 'min_length': 0, 'do_sample': False, 'early_stopping': False, 'num_beams': 1, 'temperature': 1.0, 'top_k': 50, 'top_p': 1.0, 'typical_p': 1.0, 'repetition_penalty': 1.0, 'length_penalty': 1.0, 'no_repeat_ngram_size': 0, 'encoder_no_repeat_ngram_size': 0, 'bad_words_ids': None, 'num_return_sequences': 1, 'output_scores': False, 'return_dict_in_generate': False, 'forced_bos_token_id': None, 'forced_eos_token_id': None, 'remove_invalid_values': False, 'exponential_decay_length_penalty': None, 'suppress_tokens': None, 'begin_suppress_tokens': None, 'num_beam_groups': 1, 'diversity_penalty': 0.0, '_name_or_path': '/workspace/Models/Qwen2.5-Coder-14B-CPT-SFT_v2', 'transformers_version': '4.57.3', 'model_type': 'qwen2', 'rope_parameters': {'rope_theta': 1000000.0, 'rope_type': 'default'}, 'tf_legacy_loss': False, 'use_bfloat16': False, 'output_attentions': False, 'output_dir': 'runs/grpo_14b_run1/checkpoints', 'overwrite_output_dir': False, 'do_train': False, 'do_eval': False, 'do_predict': False, 'eval_strategy': 'no', 'prediction_loss_only': False, 'per_device_train_batch_size': 1, 'per_device_eval_batch_size': 1, 'per_gpu_train_batch_size': None, 'per_gpu_eval_batch_size': None, 'gradient_accumulation_steps': 8, 'eval_accumulation_steps': None, 'eval_delay': 0, 'torch_empty_cache_steps': None, 'learning_rate': 5e-06, 'weight_decay': 0.01, 'adam_beta1': 0.9, 'adam_beta2': 0.999, 'adam_epsilon': 1e-08, 'max_grad_norm': 1.0, 'num_train_epochs': 2.0, 'max_steps': -1, 'lr_scheduler_type': 'cosine', 'lr_scheduler_kwargs': {}, 'warmup_ratio': 0.05, 'warmup_steps': 0, 'log_level': 'passive', 'log_level_replica': 'warning', 'log_on_each_node': True, 'logging_dir': 'runs/grpo_14b_run1/checkpoints/runs/Dec27_19-45-32_a100-2gpu-shell-session-757d587799-mfdvv', 'logging_strategy': 'steps', 'logging_first_step': False, 'logging_steps': 5, 'logging_nan_inf_filter': True, 'save_strategy': 'steps', 'save_steps': 100, 'save_total_limit': 2, 'save_safetensors': True, 'save_on_each_node': False, 'save_only_model': False, 'restore_callback_states_from_checkpoint': False, 'no_cuda': False, 'use_cpu': False, 'use_mps_device': False, 'seed': 42, 'data_seed': None, 'jit_mode_eval': False, 'bf16': True, 'fp16': False, 'fp16_opt_level': 'O1', 'half_precision_backend': 'auto', 'bf16_full_eval': False, 'fp16_full_eval': False, 'tf32': None, 'local_rank': 0, 'ddp_backend': None, 'tpu_num_cores': None, 'tpu_metrics_debug': False, 'debug': [], 'dataloader_drop_last': False, 'eval_steps': 50, 'dataloader_num_workers': 0, 'dataloader_prefetch_factor': None, 'past_index': -1, 'run_name': None, 'disable_tqdm': False, 'remove_unused_columns': False, 'label_names': None, 'load_best_model_at_end': False, 'metric_for_best_model': 'eval_loss', 'greater_is_better': False, 'ignore_data_skip': False, 'fsdp': [], 'fsdp_min_num_params': 0, 'fsdp_config': {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}, 'fsdp_transformer_layer_cls_to_wrap': None, 'accelerator_config': {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}, 'parallelism_config': None, 'deepspeed': None, 'label_smoothing_factor': 0.0, 'optim': 'adamw_torch', 'optim_args': None, 'adafactor': False, 'group_by_length': False, 'length_column_name': 'length', 'report_to': ['wandb'], 'project': 'huggingface', 'trackio_space_id': 'trackio', 'ddp_find_unused_parameters': None, 'ddp_bucket_cap_mb': None, 'ddp_broadcast_buffers': None, 'dataloader_pin_memory': True, 'dataloader_persistent_workers': False, 'skip_memory_metrics': True, 'use_legacy_prediction_loop': False, 'push_to_hub': False, 'resume_from_checkpoint': None, 'hub_model_id': None, 'hub_strategy': 'every_save', 'hub_token': '<HUB_TOKEN>', 'hub_private_repo': None, 'hub_always_push': False, 'hub_revision': None, 'gradient_checkpointing': False, 'gradient_checkpointing_kwargs': None, 'include_inputs_for_metrics': False, 'include_for_metrics': [], 'eval_do_concat_batches': True, 'fp16_backend': 'auto', 'push_to_hub_model_id': None, 'push_to_hub_organization': None, 'push_to_hub_token': '<PUSH_TO_HUB_TOKEN>', 'mp_parameters': '', 'auto_find_batch_size': False, 'full_determinism': False, 'torchdynamo': None, 'ray_scope': 'last', 'ddp_timeout': 1800, 'torch_compile': False, 'torch_compile_backend': None, 'torch_compile_mode': None, 'include_tokens_per_second': False, 'include_num_input_tokens_seen': 'no', 'neftune_noise_alpha': None, 'optim_target_modules': None, 'batch_eval_metrics': False, 'eval_on_start': False, 'use_liger_kernel': False, 'liger_kernel_config': None, 'eval_use_gather_object': False, 'average_tokens_across_devices': True}
+2025-12-27 19:45:32,798 INFO    MainThread:592889 [wandb_config.py:__setitem__():154] [no run ID] config set model/num_parameters = 14838846464 - <bound method Run._config_callback of <wandb.sdk.wandb_run.Run object at 0x7ada527aff70>>
+2025-12-27 19:45:32,798 INFO    MainThread:592889 [wandb_run.py:_config_callback():1396] config_cb model/num_parameters 14838846464 None
+2025-12-27 22:06:03,800 INFO    MainThread:592889 [wandb_run.py:_finish():2287] finishing run sirajuddin-shaik-007/rl-training/jz7bptqa
+2025-12-27 22:06:03,800 INFO    MainThread:592889 [wandb_run.py:_atexit_cleanup():2486] got exitcode: 0
+2025-12-27 22:06:03,801 INFO    MainThread:592889 [wandb_run.py:_restore():2468] restore
+2025-12-27 22:06:03,801 INFO    MainThread:592889 [wandb_run.py:_restore():2474] restore done
+2025-12-27 22:06:04,302 INFO    MainThread:592889 [wandb_run.py:_footer_sync_info():3862] logging synced files

wandb/run-20251227_194423-jz7bptqa/files/config.yaml

@@ -0,0 +1,725 @@
+_name_or_path:
+    value: /workspace/Models/Qwen2.5-Coder-14B-CPT-SFT_v2
+_wandb:
+    value:
+        cli_version: 0.23.1
+        e:
+            flq59s27gswy5t5lhj0bnmni5u7y1dm1:
+                args:
+                    - --config
+                    - config_grpo_14b.yaml
+                codePath: run_grpo_fixed.py
+                codePathLocal: run_grpo_fixed.py
+                cpu_count: 12
+                cpu_count_logical: 24
+                cudaVersion: "13.0"
+                disk:
+                    /:
+                        total: "791251738624"
+                        used: "537236914176"
+                email: shaiksirajuddin9949@gmail.com
+                executable: /root/llm_finetuning_env/bin/python
+                gpu: NVIDIA A100-SXM4-80GB
+                gpu_count: 2
+                gpu_nvidia:
+                    - architecture: Ampere
+                      cudaCores: 6912
+                      memoryTotal: "85899345920"
+                      name: NVIDIA A100-SXM4-80GB
+                      uuid: GPU-989794b0-ec3b-13bf-db9f-3fbe341497ba
+                    - architecture: Ampere
+                      cudaCores: 6912
+                      memoryTotal: "85899345920"
+                      name: NVIDIA A100-SXM4-80GB
+                      uuid: GPU-3790aa64-60ef-9eac-b0b1-b278ee8c0d40
+                host: a100-2gpu-shell-session-757d587799-mfdvv
+                memory:
+                    total: "359047892992"
+                os: Linux-6.12.46+-x86_64-with-glibc2.35
+                program: /workspace/trainer-kit/GRPO-14b/run_grpo_fixed.py
+                python: CPython 3.10.12
+                root: runs/grpo_14b_run1
+                startedAt: "2025-12-27T19:44:23.484378Z"
+                writerId: flq59s27gswy5t5lhj0bnmni5u7y1dm1
+        m:
+            - "1": train/global_step
+              "6":
+                - 3
+              "7": []
+            - "2": '*'
+              "5": 1
+              "6":
+                - 1
+              "7": []
+        python_version: 3.10.12
+        t:
+            "1":
+                - 1
+                - 11
+                - 41
+                - 49
+                - 51
+                - 71
+                - 98
+            "2":
+                - 1
+                - 11
+                - 41
+                - 49
+                - 51
+                - 71
+                - 98
+            "3":
+                - 2
+                - 7
+                - 15
+                - 16
+                - 19
+                - 62
+                - 66
+            "4": 3.10.12
+            "5": 0.23.1
+            "6": 4.57.3
+            "9":
+                "1": transformers_trainer
+            "12": 0.23.1
+            "13": linux-x86_64
+accelerator_config:
+    value:
+        dispatch_batches: null
+        even_batches: true
+        gradient_accumulation_kwargs: null
+        non_blocking: false
+        split_batches: false
+        use_seedable_sampler: true
+adafactor:
+    value: false
+adam_beta1:
+    value: 0.9
+adam_beta2:
+    value: 0.999
+adam_epsilon:
+    value: 1e-08
+add_cross_attention:
+    value: false
+architectures:
+    value:
+        - Qwen2ForCausalLM
+attention_dropout:
+    value: 0
+auto_find_batch_size:
+    value: false
+average_tokens_across_devices:
+    value: true
+bad_words_ids:
+    value: null
+batch_eval_metrics:
+    value: false
+begin_suppress_tokens:
+    value: null
+bf16:
+    value: true
+bf16_full_eval:
+    value: false
+bos_token_id:
+    value: 151643
+chunk_size_feed_forward:
+    value: 0
+cross_attention_hidden_size:
+    value: null
+data:
+    value:
+        completions_field: completions
+        custom_template: |-
+            ##INSTRUCTION
+            {instruction}<|im_end|>
+            {input}<|im_end|>
+            {output}<|im_end|>
+        eval_jsonl: null
+        eval_split_ratio: 0
+        format_type: raw
+        max_length: 2048
+        min_completions: 2
+        num_proc: 1
+        prompt_field: prompt
+        scores_field: scores
+        shuffle: true
+        system_prompt: |
+            You are a Hyperswitch Rust code analyzer. Identify functions/structs that need modification for a given task.
+
+            ## Output Format
+
+            ##OUTPUT
+            Explain the data flow and why each component must change:
+            - Flow: [Input → Processing → Output with arrows]
+            - For each component: "The [ComponentName] ([path]) must [action] because [reason]—without this, [consequence]"
+            - Explain coupling between components
+
+            ##SELECT
+            modify::crates/path/to/file.rs::impl::ComponentName
+            add::crates/another/file.rs::function::AnotherComponent
+            <EOS>
+
+            ## Rules
+
+            1. Use full paths: `remove::crates/folder/file.rs::Type::Name`
+            2. Use `::` for nested items: `status::StructName::Type::Name`
+            3. Always explain "must change because" and "without this"
+            3. Types of components: function, struct, enum, impl, trait
+            4. If there is extra information (e.g., enum variants), include that too.
+            5. Start with ##OUTPUT, end with ##SELECT, terminate with <EOS>
+
+            ## Example
+
+            ##TASK
+            Add webhook subscription support
+
+            ##OUTPUT
+            The webhook system routes events via EventClass enum. Flow: webhook → EventClass → handler → processing. The EventClass enum (crates/common_enums/src/enums.rs::EventClass) must add Subscriptions variant because it defines event routing—without this, subscription events cannot be processed. The SubscriptionStatus impl (crates/common_enums/src/transformers.rs::SubscriptionStatus) must map to EventType because it converts status to events—without this, status changes don't trigger webhooks. These are coupled: EventClass routes to handlers that use SubscriptionStatus mappings.
+
+            ##SELECT
+            crates/common_enums/src/enums.rs::EventClass
+            crates/common_enums/src/transformers.rs::SubscriptionStatus
+            <EOS>
+        train_jsonl: grpo_dataset.jsonl
+data_seed:
+    value: null
+dataloader_drop_last:
+    value: false
+dataloader_num_workers:
+    value: 0
+dataloader_persistent_workers:
+    value: false
+dataloader_pin_memory:
+    value: true
+dataloader_prefetch_factor:
+    value: null
+ddp_backend:
+    value: null
+ddp_broadcast_buffers:
+    value: null
+ddp_bucket_cap_mb:
+    value: null
+ddp_find_unused_parameters:
+    value: null
+ddp_timeout:
+    value: 1800
+debug:
+    value: []
+decoder_start_token_id:
+    value: null
+deepspeed:
+    value: null
+disable_tqdm:
+    value: false
+diversity_penalty:
+    value: 0
+do_eval:
+    value: false
+do_predict:
+    value: false
+do_sample:
+    value: false
+do_train:
+    value: false
+dtype:
+    value: bfloat16
+early_stopping:
+    value: false
+encoder_no_repeat_ngram_size:
+    value: 0
+eos_token_id:
+    value: 151643
+eval_accumulation_steps:
+    value: null
+eval_delay:
+    value: 0
+eval_do_concat_batches:
+    value: true
+eval_on_start:
+    value: false
+eval_steps:
+    value: 50
+eval_strategy:
+    value: "no"
+eval_use_gather_object:
+    value: false
+exponential_decay_length_penalty:
+    value: null
+finetuning_task:
+    value: null
+forced_bos_token_id:
+    value: null
+forced_eos_token_id:
+    value: null
+fp16:
+    value: false
+fp16_backend:
+    value: auto
+fp16_full_eval:
+    value: false
+fp16_opt_level:
+    value: O1
+fsdp:
+    value: []
+fsdp_config:
+    value:
+        min_num_params: 0
+        xla: false
+        xla_fsdp_grad_ckpt: false
+        xla_fsdp_v2: false
+fsdp_min_num_params:
+    value: 0
+fsdp_transformer_layer_cls_to_wrap:
+    value: null
+full_determinism:
+    value: false
+gradient_accumulation_steps:
+    value: 8
+gradient_checkpointing:
+    value: false
+gradient_checkpointing_kwargs:
+    value: null
+greater_is_better:
+    value: false
+group_by_length:
+    value: false
+grpo:
+    value:
+        advantage_temperature: 1
+        group_size: 4
+        kl_coef: 0.05
+        normalize_advantages: true
+        reward_bias: 0
+        reward_clip: 5
+        reward_scaling: 1
+        seed: 42
+        use_reference_model: false
+half_precision_backend:
+    value: auto
+hidden_act:
+    value: silu
+hidden_size:
+    value: 5120
+hub_always_push:
+    value: false
+hub_model_id:
+    value: null
+hub_private_repo:
+    value: null
+hub_revision:
+    value: null
+hub_strategy:
+    value: every_save
+hub_token:
+    value: <HUB_TOKEN>
+id2label:
+    value:
+        "0": LABEL_0
+        "1": LABEL_1
+ignore_data_skip:
+    value: false
+include_for_metrics:
+    value: []
+include_inputs_for_metrics:
+    value: false
+include_num_input_tokens_seen:
+    value: "no"
+include_tokens_per_second:
+    value: false
+initializer_range:
+    value: 0.02
+intermediate_size:
+    value: 13824
+is_decoder:
+    value: false
+is_encoder_decoder:
+    value: false
+jit_mode_eval:
+    value: false
+label_names:
+    value: null
+label_smoothing_factor:
+    value: 0
+label2id:
+    value:
+        LABEL_0: 0
+        LABEL_1: 1
+layer_types:
+    value:
+        - 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
+        - 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
+learning_rate:
+    value: 5e-06
+length_column_name:
+    value: length
+length_penalty:
+    value: 1
+liger_kernel_config:
+    value: null
+load_best_model_at_end:
+    value: false
+local_rank:
+    value: 0
+log_level:
+    value: passive
+log_level_replica:
+    value: warning
+log_on_each_node:
+    value: true
+logging_dir:
+    value: runs/grpo_14b_run1/checkpoints/runs/Dec27_19-45-32_a100-2gpu-shell-session-757d587799-mfdvv
+logging_first_step:
+    value: false
+logging_nan_inf_filter:
+    value: true
+logging_steps:
+    value: 5
+logging_strategy:
+    value: steps
+lr_scheduler_type:
+    value: cosine
+max_grad_norm:
+    value: 1
+max_length:
+    value: 20
+max_position_embeddings:
+    value: 32768
+max_steps:
+    value: -1
+max_window_layers:
+    value: 48
+metric_for_best_model:
+    value: eval_loss
+min_length:
+    value: 0
+model:
+    value:
+        device_map: auto
+        load_in_4bit: false
+        load_in_8bit: false
+        repo_id: /workspace/Models/Qwen2.5-Coder-14B-CPT-SFT_v2
+        tokenizer_name: Qwen/Qwen2.5-Coder-14B
+        torch_dtype: bfloat16
+        trust_remote_code: true
+model/num_parameters:
+    value: 14838846464
+model_type:
+    value: qwen2
+mp_parameters:
+    value: ""
+neftune_noise_alpha:
+    value: null
+no_cuda:
+    value: false
+no_repeat_ngram_size:
+    value: 0
+num_attention_heads:
+    value: 40
+num_beam_groups:
+    value: 1
+num_beams:
+    value: 1
+num_hidden_layers:
+    value: 48
+num_key_value_heads:
+    value: 8
+num_return_sequences:
+    value: 1
+num_train_epochs:
+    value: 2
+optim:
+    value: adamw_torch
+optim_args:
+    value: null
+optim_target_modules:
+    value: null
+output_attentions:
+    value: false
+output_dir:
+    value: runs/grpo_14b_run1/checkpoints
+output_hidden_states:
+    value: false
+output_scores:
+    value: false
+overwrite_output_dir:
+    value: false
+pad_token_id:
+    value: null
+parallelism_config:
+    value: null
+past_index:
+    value: -1
+peft:
+    value:
+        bias: none
+        enabled: true
+        lora_alpha: 32
+        lora_dropout: 0.05
+        r: 16
+        target_modules:
+            - q_proj
+            - k_proj
+            - v_proj
+            - o_proj
+            - gate_proj
+            - up_proj
+            - down_proj
+        task_type: CAUSAL_LM
+peft_config:
+    value:
+        default:
+            alora_invocation_tokens: null
+            arrow_config: null
+            auto_mapping: null
+            base_model_name_or_path: /workspace/Models/Qwen2.5-Coder-14B-CPT-SFT_v2
+            bias: none
+            corda_config: null
+            ensure_weight_tying: false
+            eva_config: null
+            exclude_modules: null
+            fan_in_fan_out: false
+            inference_mode: false
+            init_lora_weights: true
+            layer_replication: null
+            layers_pattern: null
+            layers_to_transform: null
+            lora_alpha: 32
+            lora_bias: false
+            lora_dropout: 0.05
+            megatron_config: null
+            megatron_core: megatron.core
+            modules_to_save: null
+            peft_type: LORA
+            peft_version: 0.18.0
+            qalora_group_size: 16
+            r: 16
+            revision: null
+            runtime_config:
+                ephemeral_gpu_offload: false
+            target_modules:
+                - q_proj
+                - o_proj
+                - v_proj
+                - up_proj
+                - gate_proj
+                - down_proj
+                - k_proj
+            target_parameters: null
+            task_type: CAUSAL_LM
+            trainable_token_indices: null
+            use_dora: false
+            use_qalora: false
+            use_rslora: false
+per_device_eval_batch_size:
+    value: 1
+per_device_train_batch_size:
+    value: 1
+per_gpu_eval_batch_size:
+    value: null
+per_gpu_train_batch_size:
+    value: null
+prediction_loss_only:
+    value: false
+prefix:
+    value: null
+problem_type:
+    value: null
+project:
+    value: huggingface
+push_to_hub:
+    value: false
+push_to_hub_model_id:
+    value: null
+push_to_hub_organization:
+    value: null
+push_to_hub_token:
+    value: <PUSH_TO_HUB_TOKEN>
+ray_scope:
+    value: last
+remove_invalid_values:
+    value: false
+remove_unused_columns:
+    value: false
+repetition_penalty:
+    value: 1
+report_to:
+    value:
+        - wandb
+restore_callback_states_from_checkpoint:
+    value: false
+resume_from_checkpoint:
+    value: null
+return_dict:
+    value: true
+return_dict_in_generate:
+    value: false
+rms_norm_eps:
+    value: 1e-06
+rope_parameters:
+    value:
+        rope_theta: 1e+06
+        rope_type: default
+rope_scaling:
+    value: null
+rope_theta:
+    value: 10000
+run_dir:
+    value: runs/grpo_14b_run1
+run_name:
+    value: null
+save_on_each_node:
+    value: false
+save_only_model:
+    value: false
+save_safetensors:
+    value: true
+save_steps:
+    value: 100
+save_strategy:
+    value: steps
+save_total_limit:
+    value: 2
+seed:
+    value: 42
+sep_token_id:
+    value: null
+skip_memory_metrics:
+    value: true
+sliding_window:
+    value: null
+suppress_tokens:
+    value: null
+task_specific_params:
+    value: null
+temperature:
+    value: 1
+tf_legacy_loss:
+    value: false
+tf32:
+    value: null
+tie_encoder_decoder:
+    value: false
+tie_word_embeddings:
+    value: false
+tokenizer_class:
+    value: null
+top_k:
+    value: 50
+top_p:
+    value: 1
+torch_compile:
+    value: false
+torch_compile_backend:
+    value: null
+torch_compile_mode:
+    value: null
+torch_empty_cache_steps:
+    value: null
+torchdynamo:
+    value: null
+torchscript:
+    value: false
+tpu_metrics_debug:
+    value: false
+tpu_num_cores:
+    value: null
+trackio_space_id:
+    value: trackio
+train:
+    value:
+        bf16: true
+        dataloader_num_workers: 4
+        dataloader_pin_memory: true
+        ddp_find_unused_parameters: false
+        evaluation_strategy: "no"
+        fp16: false
+        gradient_accumulation_steps: 8
+        gradient_checkpointing: true
+        learning_rate: 5e-06
+        logging_steps: 5
+        lr_scheduler_type: cosine
+        max_grad_norm: 1
+        num_train_epochs: 2
+        output_dir: runs/grpo_14b_run1
+        per_device_eval_batch_size: 1
+        per_device_train_batch_size: 1
+        remove_unused_columns: false
+        report_to: []
+        save_steps: 100
+        save_total_limit: 2
+        seed: 42
+        warmup_ratio: 0.05
+        weight_decay: 0.01
+transformers_version:
+    value: 4.57.3
+typical_p:
+    value: 1
+use_bfloat16:
+    value: false
+use_cache:
+    value: false
+use_cpu:
+    value: false
+use_legacy_prediction_loop:
+    value: false
+use_liger_kernel:
+    value: false
+use_mps_device:
+    value: false
+use_sliding_window:
+    value: false
+vocab_size:
+    value: 152064
+warmup_ratio:
+    value: 0.05
+warmup_steps:
+    value: 0
+weight_decay:
+    value: 0.01

wandb/run-20251227_194423-jz7bptqa/files/output.log

@@ -0,0 +1,134 @@
+Wandb initialized: project='rl-training', name='auto-generated'
+`torch_dtype` is deprecated! Use `dtype` instead!
+Loading checkpoint shards: 100%|████████████████████████████████████████████| 16/16 [00:17<00:00,  1.10s/it]
+Formatting and tokenizing train GRPO data (num_proc=1): 100%|████| 2000/2000 [00:37<00:00, 53.65 examples/s]
+Filter: 100%|███████████████████████████████████████████████████| 1998/1998 [00:08<00:00, 244.33 examples/s]
+Training samples: 1998
+Warning: group_size (4) > min_completions (2)
+Setting group_size to 2
+
+=== GRPO Training Configuration ===
+Group size: 2
+KL coefficient: 0.05
+Normalize advantages: True
+Reward scaling: 1.0
+Advantage temperature: 1.0
+Use reference model: False
+===================================
+
+Starting GRPO training...
+  0%|                                                                               | 0/500 [00:00<?, ?it/s]/workspace/trainer-kit/GRPO-14b/run_grpo_fixed.py:700: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead.
+  with torch.cuda.amp.autocast(enabled=self.args.fp16 or self.args.bf16):
+Could not estimate the number of tokens of the input, floating-point operations will not be computed
+ 20%|█████████████▍                                                     | 100/500 [28:13<1:54:10, 17.13s/it]/workspace/trainer-kit/GRPO-14b/run_grpo_fixed.py:700: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead.
+{'loss': 0.007, 'grad_norm': 0.05460292845964432, 'learning_rate': 8.000000000000001e-07, 'grpo_mean_advantage': -1.3560057254835556e-07, 'grpo_std_advantage': 3.0318567496578908e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5922331809997559, 'epoch': 0.02}
+{'loss': 0.0107, 'grad_norm': 0.0679207444190979, 'learning_rate': 1.8000000000000001e-06, 'grpo_mean_advantage': 3.6619603633880615e-06, 'grpo_std_advantage': 1.6246918676188216e-05, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5561589002609253, 'epoch': 0.04}
+{'loss': 0.007, 'grad_norm': 0.05788416787981987, 'learning_rate': 2.8000000000000003e-06, 'grpo_mean_advantage': -1.0654330395709621e-07, 'grpo_std_advantage': 5.399440965447866e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5759152173995972, 'epoch': 0.06}
+{'loss': 0.0246, 'grad_norm': 0.0746568813920021, 'learning_rate': 3.8000000000000005e-06, 'grpo_mean_advantage': -5.871057737749652e-07, 'grpo_std_advantage': 2.6951597646984737e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5127314329147339, 'epoch': 0.08}
+{'loss': 0.0337, 'grad_norm': 0.11442846059799194, 'learning_rate': 4.800000000000001e-06, 'grpo_mean_advantage': 6.370246410369873e-07, 'grpo_std_advantage': 2.8908377771585947e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.539706826210022, 'epoch': 0.1}
+{'loss': 0.0171, 'grad_norm': 0.05778791010379791, 'learning_rate': 4.999125183044924e-06, 'grpo_mean_advantage': 6.705522359595761e-09, 'grpo_std_advantage': 6.189450800775376e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5812538862228394, 'epoch': 0.12}
+{'loss': 0.0145, 'grad_norm': 0.05819695070385933, 'learning_rate': 4.995572288443412e-06, 'grpo_mean_advantage': 3.859400692363124e-07, 'grpo_std_advantage': 1.6833292875162442e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5909844636917114, 'epoch': 0.14}
+{'loss': 0.0196, 'grad_norm': 0.07968433201313019, 'learning_rate': 4.98929052218411e-06, 'grpo_mean_advantage': 2.600252742013254e-07, 'grpo_std_advantage': 1.4095899132371414e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5630953907966614, 'epoch': 0.16}
+{'loss': 0.0186, 'grad_norm': 0.0733402892947197, 'learning_rate': 4.980286753286196e-06, 'grpo_mean_advantage': -1.2591480924584175e-07, 'grpo_std_advantage': 1.0309080380466185e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5604403614997864, 'epoch': 0.18}
+{'loss': 0.0286, 'grad_norm': 0.07136482000350952, 'learning_rate': 4.9685708272387645e-06, 'grpo_mean_advantage': -2.808868941883702e-07, 'grpo_std_advantage': 1.5696078889959608e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5971035957336426, 'epoch': 0.2}
+{'loss': 0.0054, 'grad_norm': 0.08851475268602371, 'learning_rate': 4.9541555552349404e-06, 'grpo_mean_advantage': 2.6822089438383045e-08, 'grpo_std_advantage': 3.7878271541558206e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5892971754074097, 'epoch': 0.22}
+{'loss': -0.0074, 'grad_norm': 0.07778509706258774, 'learning_rate': 4.9370567001630155e-06, 'grpo_mean_advantage': -5.662441182607836e-08, 'grpo_std_advantage': 6.128998393251095e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.564322292804718, 'epoch': 0.24}
+{'loss': 0.0145, 'grad_norm': 0.08740051090717316, 'learning_rate': 4.917292959369968e-06, 'grpo_mean_advantage': -1.5944242193199898e-07, 'grpo_std_advantage': 1.6374274309782777e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.562497615814209, 'epoch': 0.26}
+{'loss': 0.0257, 'grad_norm': 0.19070060551166534, 'learning_rate': 4.8948859442161876e-06, 'grpo_mean_advantage': 1.6838312433264946e-07, 'grpo_std_advantage': 8.536571272088622e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5904761552810669, 'epoch': 0.28}
+{'loss': 0.0024, 'grad_norm': 0.07321271300315857, 'learning_rate': 4.869860156443768e-06, 'grpo_mean_advantage': 1.1175870895385742e-07, 'grpo_std_advantage': 6.451961667153228e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5765624046325684, 'epoch': 0.3}
+{'loss': 0.0277, 'grad_norm': 0.07126748561859131, 'learning_rate': 4.842242961384211e-06, 'grpo_mean_advantage': -1.4603138254187797e-07, 'grpo_std_advantage': 1.1309343790344428e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5858271718025208, 'epoch': 0.32}
+{'loss': 0.0246, 'grad_norm': 0.08629189431667328, 'learning_rate': 4.812064558034847e-06, 'grpo_mean_advantage': -1.817941665649414e-06, 'grpo_std_advantage': 1.1141768482048064e-05, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5871662497520447, 'epoch': 0.34}
+{'loss': 0.0056, 'grad_norm': 0.0998779758810997, 'learning_rate': 4.779357946036662e-06, 'grpo_mean_advantage': 1.8179416372277046e-07, 'grpo_std_advantage': 6.210335072864837e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5330992937088013, 'epoch': 0.36}
+{'loss': 0.0053, 'grad_norm': 0.10614689439535141, 'learning_rate': 4.74415888958968e-06, 'grpo_mean_advantage': -2.972781771859445e-07, 'grpo_std_advantage': 3.1582342217006953e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5265295505523682, 'epoch': 0.38}
+{'loss': 0.0134, 'grad_norm': 0.10345634073019028, 'learning_rate': 4.706505878345343e-06, 'grpo_mean_advantage': -7.033348197182931e-07, 'grpo_std_advantage': 4.245831405569334e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5660771131515503, 'epoch': 0.4}
+  with torch.cuda.amp.autocast(enabled=self.args.fp16 or self.args.bf16):
+ 40%|██████████████████████████▊                                        | 200/500 [56:13<1:25:51, 17.17s/it]/workspace/trainer-kit/GRPO-14b/run_grpo_fixed.py:700: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead.
+{'loss': 0.0004, 'grad_norm': 0.10077933222055435, 'learning_rate': 4.666440085318626e-06, 'grpo_mean_advantage': 1.1920928955078125e-07, 'grpo_std_advantage': 3.2809634831210133e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.57631915807724, 'epoch': 0.42}
+{'loss': 0.0033, 'grad_norm': 0.09548182785511017, 'learning_rate': 4.624005321865968e-06, 'grpo_mean_advantage': -4.0978193283081055e-07, 'grpo_std_advantage': 6.0397578636184335e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.546563982963562, 'epoch': 0.44}
+{'loss': 0.0095, 'grad_norm': 0.09417816251516342, 'learning_rate': 4.57924798977818e-06, 'grpo_mean_advantage': -1.467764434437413e-07, 'grpo_std_advantage': 2.2689375782647403e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5519219636917114, 'epoch': 0.46}
+{'loss': 0.0006, 'grad_norm': 0.10022275149822235, 'learning_rate': 4.532217030540781e-06, 'grpo_mean_advantage': -5.215406329028838e-09, 'grpo_std_advantage': 7.929010621410271e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5490407943725586, 'epoch': 0.48}
+{'loss': -0.0046, 'grad_norm': 0.14057794213294983, 'learning_rate': 4.482963871817195e-06, 'grpo_mean_advantage': -5.7369469175228005e-08, 'grpo_std_advantage': 1.2823379620385822e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5646580457687378, 'epoch': 0.5}
+{'loss': -0.003, 'grad_norm': 0.12420658767223358, 'learning_rate': 4.4315423712133595e-06, 'grpo_mean_advantage': 2.9876827056796174e-07, 'grpo_std_advantage': 1.0496698905626545e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.6111599802970886, 'epoch': 0.52}
+{'loss': 0.0154, 'grad_norm': 0.14342808723449707, 'learning_rate': 4.378008757385222e-06, 'grpo_mean_advantage': 1.5869736103013565e-07, 'grpo_std_advantage': 1.2748531617035042e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5619662404060364, 'epoch': 0.54}
+{'loss': -0.0262, 'grad_norm': 0.14729444682598114, 'learning_rate': 4.322421568553529e-06, 'grpo_mean_advantage': 3.0100346748440643e-07, 'grpo_std_advantage': 2.4499684059264837e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5795454978942871, 'epoch': 0.56}
+{'loss': 0.0018, 'grad_norm': 0.15249410271644592, 'learning_rate': 4.2648415884931476e-06, 'grpo_mean_advantage': -3.233552092751779e-07, 'grpo_std_advantage': 1.248456669600273e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5804953575134277, 'epoch': 0.58}
+{'loss': -0.017, 'grad_norm': 0.1841023564338684, 'learning_rate': 4.205331780066892e-06, 'grpo_mean_advantage': 3.2261013416245987e-07, 'grpo_std_advantage': 1.4773489738217904e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5628539323806763, 'epoch': 0.6}
+{'loss': 0.0044, 'grad_norm': 0.18597163259983063, 'learning_rate': 4.1439572163765615e-06, 'grpo_mean_advantage': -2.5331974029541016e-07, 'grpo_std_advantage': 1.5092309695319273e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5727725625038147, 'epoch': 0.62}
+{'loss': -0.005, 'grad_norm': 0.18310388922691345, 'learning_rate': 4.0807850096064605e-06, 'grpo_mean_advantage': -6.780028627417778e-08, 'grpo_std_advantage': 8.550978805033083e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5833909511566162, 'epoch': 0.64}
+{'loss': -0.015, 'grad_norm': 0.2192923128604889, 'learning_rate': 4.015884237637206e-06, 'grpo_mean_advantage': -5.587935447692871e-08, 'grpo_std_advantage': 3.564579174053506e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5742615461349487, 'epoch': 0.66}
+{'loss': -0.0314, 'grad_norm': 0.16708803176879883, 'learning_rate': 3.949325868510083e-06, 'grpo_mean_advantage': -5.327165126800537e-07, 'grpo_std_advantage': 2.309018327650847e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5758188962936401, 'epoch': 0.68}
+{'loss': -0.0441, 'grad_norm': 0.3401262164115906, 'learning_rate': 3.881182682824534e-06, 'grpo_mean_advantage': 5.863606702405377e-07, 'grpo_std_advantage': 2.4449204829579685e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5767683982849121, 'epoch': 0.7}
+{'loss': -0.0162, 'grad_norm': 0.1931898146867752, 'learning_rate': 3.811529194153635e-06, 'grpo_mean_advantage': 3.2186508747145126e-07, 'grpo_std_advantage': 2.293551688126172e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.586772084236145, 'epoch': 0.72}
+{'loss': -0.0386, 'grad_norm': 0.2537969648838043, 'learning_rate': 3.7404415675646054e-06, 'grpo_mean_advantage': -4.470348358154297e-08, 'grpo_std_advantage': 3.7067667335577426e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.549396276473999, 'epoch': 0.74}
+{'loss': -0.037, 'grad_norm': 0.20326584577560425, 'learning_rate': 3.667997536333424e-06, 'grpo_mean_advantage': -2.1010637851759384e-07, 'grpo_std_advantage': 1.1695076409523608e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5798425078392029, 'epoch': 0.76}
+{'loss': -0.0292, 'grad_norm': 0.25048357248306274, 'learning_rate': 3.59427631694463e-06, 'grpo_mean_advantage': 1.765787658314366e-07, 'grpo_std_advantage': 2.429934738756856e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5584167838096619, 'epoch': 0.78}
+{'loss': -0.0454, 'grad_norm': 0.2687569260597229, 'learning_rate': 3.5193585224692595e-06, 'grpo_mean_advantage': 1.6540289493605087e-07, 'grpo_std_advantage': 2.6342788714828203e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5676193237304688, 'epoch': 0.8}
+  with torch.cuda.amp.autocast(enabled=self.args.fp16 or self.args.bf16):
+ 60%|████████████████████████████████████████▏                          | 300/500 [1:24:01<55:53, 16.77s/it]/workspace/trainer-kit/GRPO-14b/run_grpo_fixed.py:700: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead.
+{'loss': -0.0423, 'grad_norm': 0.22301620244979858, 'learning_rate': 3.44332607441564e-06, 'grpo_mean_advantage': -1.0944902442133753e-06, 'grpo_std_advantage': 5.346942998585291e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5669739842414856, 'epoch': 0.82}
+{'loss': -0.0857, 'grad_norm': 0.3040211498737335, 'learning_rate': 3.3662621131494204e-06, 'grpo_mean_advantage': 2.4065374759629776e-07, 'grpo_std_advantage': 1.6327536513927043e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5922158360481262, 'epoch': 0.84}
+{'loss': -0.0278, 'grad_norm': 0.27231141924858093, 'learning_rate': 3.2882509069808044e-06, 'grpo_mean_advantage': -5.21540641784668e-08, 'grpo_std_advantage': 5.847922466273303e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5473950505256653, 'epoch': 0.86}
+{'loss': -0.0727, 'grad_norm': 0.3571636378765106, 'learning_rate': 3.2093777600183873e-06, 'grpo_mean_advantage': 6.541609991472797e-07, 'grpo_std_advantage': 4.072162937518442e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5880032777786255, 'epoch': 0.88}
+{'loss': -0.0464, 'grad_norm': 0.306273490190506, 'learning_rate': 3.1297289188903705e-06, 'grpo_mean_advantage': -1.2218951894737984e-07, 'grpo_std_advantage': 4.386006935419573e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5835092663764954, 'epoch': 0.9}
+{'loss': -0.0295, 'grad_norm': 0.2700377106666565, 'learning_rate': 3.049391478435133e-06, 'grpo_mean_advantage': 1.7605722177904681e-06, 'grpo_std_advantage': 8.007580618141219e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5394966006278992, 'epoch': 0.92}
+{'loss': -0.031, 'grad_norm': 0.39531761407852173, 'learning_rate': 2.9684532864643123e-06, 'grpo_mean_advantage': -3.3080578987210174e-07, 'grpo_std_advantage': 1.551636614749441e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5687432289123535, 'epoch': 0.94}
+{'loss': -0.0789, 'grad_norm': 0.5987040996551514, 'learning_rate': 2.887002847702504e-06, 'grpo_mean_advantage': 2.712011450967111e-07, 'grpo_std_advantage': 1.4400844747797237e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5550583600997925, 'epoch': 0.96}
+{'loss': -0.1131, 'grad_norm': 0.5680716037750244, 'learning_rate': 2.8051292270086506e-06, 'grpo_mean_advantage': -3.2857059295565705e-07, 'grpo_std_advantage': 2.105091425619321e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.558111310005188, 'epoch': 0.98}
+{'loss': -0.2232, 'grad_norm': 0.6204046010971069, 'learning_rate': 2.722921951984927e-06, 'grpo_mean_advantage': 4.470348358154297e-08, 'grpo_std_advantage': 5.315724820320611e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.6196198463439941, 'epoch': 1.0}
+{'loss': -0.1363, 'grad_norm': 0.8389026522636414, 'learning_rate': 2.640470915079614e-06, 'grpo_mean_advantage': 9.290873776990338e-07, 'grpo_std_advantage': 4.219644324621186e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.582168459892273, 'epoch': 1.02}
+{'loss': -0.1868, 'grad_norm': 0.9067686796188354, 'learning_rate': 2.557866275291035e-06, 'grpo_mean_advantage': 2.533197474008375e-08, 'grpo_std_advantage': 1.6600588992332632e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5551307797431946, 'epoch': 1.04}
+{'loss': -0.1792, 'grad_norm': 0.9277902841567993, 'learning_rate': 2.4751983595800093e-06, 'grpo_mean_advantage': -5.662441182607836e-08, 'grpo_std_advantage': 1.0909400316450046e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.535040020942688, 'epoch': 1.06}
+{'loss': -0.1691, 'grad_norm': 1.0715463161468506, 'learning_rate': 2.392557564098649e-06, 'grpo_mean_advantage': -9.536743306171047e-08, 'grpo_std_advantage': 5.838213610331877e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5673571825027466, 'epoch': 1.08}
+{'loss': -0.1655, 'grad_norm': 0.7759184837341309, 'learning_rate': 2.3100342553434924e-06, 'grpo_mean_advantage': 3.278255533700758e-08, 'grpo_std_advantage': 9.317170679423725e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5874732732772827, 'epoch': 1.1}
+{'loss': -0.1821, 'grad_norm': 0.9387398958206177, 'learning_rate': 2.2277186713410688e-06, 'grpo_mean_advantage': -1.206994113545079e-07, 'grpo_std_advantage': 6.201085511747806e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5569106340408325, 'epoch': 1.12}
+{'loss': -0.2102, 'grad_norm': 1.6132302284240723, 'learning_rate': 2.1457008229739395e-06, 'grpo_mean_advantage': 4.470348358154297e-08, 'grpo_std_advantage': 6.115651558502577e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5578873157501221, 'epoch': 1.14}
+{'loss': -0.2937, 'grad_norm': 0.8679026961326599, 'learning_rate': 2.0640703955551214e-06, 'grpo_mean_advantage': -3.3453108017056365e-07, 'grpo_std_advantage': 3.5326345368957845e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5735999345779419, 'epoch': 1.16}
+{'loss': -0.2598, 'grad_norm': 1.0550166368484497, 'learning_rate': 1.9829166507585084e-06, 'grpo_mean_advantage': -1.110136480519941e-07, 'grpo_std_advantage': 4.731904823529476e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5626259446144104, 'epoch': 1.18}
+{'loss': -0.2546, 'grad_norm': 1.2819372415542603, 'learning_rate': 1.90232832901255e-06, 'grpo_mean_advantage': -5.08874677507265e-07, 'grpo_std_advantage': 1.840126174101897e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5463050603866577, 'epoch': 1.2}
+  with torch.cuda.amp.autocast(enabled=self.args.fp16 or self.args.bf16):
+ 80%|█████████████████████████████████████████████████████▌             | 400/500 [1:52:15<28:44, 17.25s/it]/workspace/trainer-kit/GRPO-14b/run_grpo_fixed.py:700: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead.
+{'loss': -0.1809, 'grad_norm': 1.0188143253326416, 'learning_rate': 1.82239355246389e-06, 'grpo_mean_advantage': 1.01327898960335e-07, 'grpo_std_advantage': 7.798533943059738e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5352144241333008, 'epoch': 1.22}
+{'loss': -0.3559, 'grad_norm': 2.0709052085876465, 'learning_rate': 1.7431997286170923e-06, 'grpo_mean_advantage': 1.341104507446289e-07, 'grpo_std_advantage': 7.821902840987605e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5547868013381958, 'epoch': 1.24}
+{'loss': -0.3874, 'grad_norm': 1.8516215085983276, 'learning_rate': 1.6648334547558227e-06, 'grpo_mean_advantage': 9.015202806494926e-08, 'grpo_std_advantage': 1.0693488547985908e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5859472751617432, 'epoch': 1.26}
+{'loss': -0.3467, 'grad_norm': 1.283104419708252, 'learning_rate': 1.5873804232499862e-06, 'grpo_mean_advantage': -2.443790378947597e-07, 'grpo_std_advantage': 1.183122208203713e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5751550793647766, 'epoch': 1.28}
+{'loss': -0.1703, 'grad_norm': 1.4108576774597168, 'learning_rate': 1.51092532785238e-06, 'grpo_mean_advantage': -6.705522537231445e-08, 'grpo_std_advantage': 6.109748937888071e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5497723817825317, 'epoch': 1.3}
+{'loss': -0.2918, 'grad_norm': 1.0421361923217773, 'learning_rate': 1.4355517710873184e-06, 'grpo_mean_advantage': -1.639127766850379e-08, 'grpo_std_advantage': 5.529495297196263e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.55989670753479, 'epoch': 1.32}
+{'loss': -0.3069, 'grad_norm': 1.3465828895568848, 'learning_rate': 1.361342172832502e-06, 'grpo_mean_advantage': 4.418194237132411e-07, 'grpo_std_advantage': 2.9275292945385445e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5809233784675598, 'epoch': 1.34}
+{'loss': -0.5594, 'grad_norm': 1.1959459781646729, 'learning_rate': 1.2883776801940884e-06, 'grpo_mean_advantage': 9.685754776000977e-08, 'grpo_std_advantage': 3.754235251562932e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5568087100982666, 'epoch': 1.36}
+{'loss': -0.4102, 'grad_norm': 1.8967422246932983, 'learning_rate': 1.216738078773522e-06, 'grpo_mean_advantage': -2.384185791015625e-07, 'grpo_std_advantage': 6.821086913078034e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5655568838119507, 'epoch': 1.38}
+{'loss': -0.338, 'grad_norm': 2.221132755279541, 'learning_rate': 1.146501705423155e-06, 'grpo_mean_advantage': -8.717179156292332e-08, 'grpo_std_advantage': 2.500940354366321e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.6089578866958618, 'epoch': 1.4}
+{'loss': -0.4985, 'grad_norm': 2.3640377521514893, 'learning_rate': 1.0777453625860474e-06, 'grpo_mean_advantage': 2.1606683731079102e-07, 'grpo_std_advantage': 1.4568390724889468e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.6129671335220337, 'epoch': 1.42}
+{'loss': -0.4347, 'grad_norm': 1.9084734916687012, 'learning_rate': 1.0105442343136184e-06, 'grpo_mean_advantage': -3.725290298461914e-09, 'grpo_std_advantage': 2.965894054796081e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5562310814857483, 'epoch': 1.44}
+{'loss': -0.6217, 'grad_norm': 1.6063904762268066, 'learning_rate': 9.449718040529987e-07, 'grpo_mean_advantage': 4.313886279305734e-07, 'grpo_std_advantage': 1.9621948013082147e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5884170532226562, 'epoch': 1.46}
+{'loss': -0.5364, 'grad_norm': 2.114664077758789, 'learning_rate': 8.810997742939531e-07, 'grpo_mean_advantage': 2.0489096641540527e-07, 'grpo_std_advantage': 1.0235522722723545e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5795440673828125, 'epoch': 1.48}
+{'loss': -0.4798, 'grad_norm': 1.8450465202331543, 'learning_rate': 8.189979881632634e-07, 'grpo_mean_advantage': -1.4185905001795618e-06, 'grpo_std_advantage': 1.0947338523692451e-05, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5607603788375854, 'epoch': 1.5}
+{'loss': -0.4805, 'grad_norm': 2.673438787460327, 'learning_rate': 7.587343530522945e-07, 'grpo_mean_advantage': -1.758337049295733e-07, 'grpo_std_advantage': 9.663675655247062e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5381432771682739, 'epoch': 1.52}
+{'loss': -0.433, 'grad_norm': 2.2263550758361816, 'learning_rate': 7.003747663612581e-07, 'grpo_mean_advantage': -6.973743325033865e-07, 'grpo_std_advantage': 4.341973180999048e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5528443455696106, 'epoch': 1.54}
+{'loss': -0.6021, 'grad_norm': 2.3657093048095703, 'learning_rate': 6.439830434413754e-07, 'grpo_mean_advantage': 1.7881394143159923e-08, 'grpo_std_advantage': 1.3004198251564958e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.6091476678848267, 'epoch': 1.56}
+{'loss': -0.5595, 'grad_norm': 1.9847129583358765, 'learning_rate': 5.896208478137222e-07, 'grpo_mean_advantage': 3.4868716625169327e-07, 'grpo_std_advantage': 2.059372718576924e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5397372245788574, 'epoch': 1.58}
+{'loss': -0.5592, 'grad_norm': 2.922114133834839, 'learning_rate': 5.373476237410808e-07, 'grpo_mean_advantage': -2.1636485598719446e-06, 'grpo_std_advantage': 9.725940799398813e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5873125195503235, 'epoch': 1.6}
+  with torch.cuda.amp.autocast(enabled=self.args.fp16 or self.args.bf16):
+100%|███████████████████████████████████████████████████████████████████| 500/500 [2:20:30<00:00, 16.86s/it]
+{'loss': -0.5623, 'grad_norm': 1.8524045944213867, 'learning_rate': 4.872205312265074e-07, 'grpo_mean_advantage': -5.960464477539063e-08, 'grpo_std_advantage': 3.460792754594877e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5601426362991333, 'epoch': 1.62}
+{'loss': -0.5943, 'grad_norm': 1.7269790172576904, 'learning_rate': 4.3929438350970687e-07, 'grpo_mean_advantage': 2.6226044269606064e-07, 'grpo_std_advantage': 7.928817922220333e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.578656792640686, 'epoch': 1.64}
+{'loss': -0.6193, 'grad_norm': 2.26530122756958, 'learning_rate': 3.936215871295634e-07, 'grpo_mean_advantage': 2.3558736756967846e-06, 'grpo_std_advantage': 1.4469559573626611e-05, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5885810852050781, 'epoch': 1.66}
+{'loss': -0.6934, 'grad_norm': 2.6794464588165283, 'learning_rate': 3.502520846183577e-07, 'grpo_mean_advantage': 1.639127766850379e-08, 'grpo_std_advantage': 9.352411325380672e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5805023312568665, 'epoch': 1.68}
+{'loss': -0.5126, 'grad_norm': 2.100447654724121, 'learning_rate': 3.092332998903416e-07, 'grpo_mean_advantage': 3.2387674764322583e-06, 'grpo_std_advantage': 1.999079904635437e-05, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5655918121337891, 'epoch': 1.7}
+{'loss': -0.5446, 'grad_norm': 2.1027915477752686, 'learning_rate': 2.706100863843822e-07, 'grpo_mean_advantage': 3.5464762504489045e-07, 'grpo_std_advantage': 1.7663603557593888e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5474504232406616, 'epoch': 1.72}
+{'loss': -0.5125, 'grad_norm': 2.289045572280884, 'learning_rate': 2.3442467801738867e-07, 'grpo_mean_advantage': 3.6135315895080566e-07, 'grpo_std_advantage': 2.356920958845876e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5874254703521729, 'epoch': 1.74}
+{'loss': -0.595, 'grad_norm': 2.278038501739502, 'learning_rate': 2.007166430021415e-07, 'grpo_mean_advantage': 2.7567148563889532e-08, 'grpo_std_advantage': 9.97340521280421e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5815118551254272, 'epoch': 1.76}
+{'loss': -0.8055, 'grad_norm': 2.340942621231079, 'learning_rate': 1.6952284058003366e-07, 'grpo_mean_advantage': -8.34465012644614e-08, 'grpo_std_advantage': 5.558832185670326e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5611211061477661, 'epoch': 1.78}
+{'loss': -0.8561, 'grad_norm': 2.4256298542022705, 'learning_rate': 1.4087738071603075e-07, 'grpo_mean_advantage': -1.9818544672034477e-07, 'grpo_std_advantage': 6.800727305744658e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.590424656867981, 'epoch': 1.8}
+{'loss': -0.429, 'grad_norm': 1.6453255414962769, 'learning_rate': 1.1481158679992554e-07, 'grpo_mean_advantage': -1.9371508841459217e-08, 'grpo_std_advantage': 3.142378943721269e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5670351982116699, 'epoch': 1.82}
+{'loss': -0.642, 'grad_norm': 2.3458049297332764, 'learning_rate': 9.135396139467151e-08, 'grpo_mean_advantage': 2.3692845729783585e-07, 'grpo_std_advantage': 1.682946731307311e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5640432834625244, 'epoch': 1.84}
+{'loss': -0.583, 'grad_norm': 2.730945110321045, 'learning_rate': 7.053015506924749e-08, 'grpo_mean_advantage': 1.110136480519941e-07, 'grpo_std_advantage': 8.930008448260196e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5584251284599304, 'epoch': 1.86}
+{'loss': -0.5197, 'grad_norm': 2.1463465690612793, 'learning_rate': 5.236293835013839e-08, 'grpo_mean_advantage': 2.5406478698641877e-07, 'grpo_std_advantage': 9.93092498902115e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5392154455184937, 'epoch': 1.88}
+{'loss': -0.5864, 'grad_norm': 2.427900791168213, 'learning_rate': 3.687217682209837e-08, 'grpo_mean_advantage': -8.940697071579962e-09, 'grpo_std_advantage': 5.835169645251881e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5686308741569519, 'epoch': 1.9}
+{'loss': -0.5721, 'grad_norm': 2.042795419692993, 'learning_rate': 2.4074809405425227e-08, 'grpo_mean_advantage': 4.0605664253234863e-07, 'grpo_std_advantage': 2.3210795916384086e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5842767357826233, 'epoch': 1.92}
+{'loss': -0.5944, 'grad_norm': 2.800136089324951, 'learning_rate': 1.3984829833499636e-08, 'grpo_mean_advantage': 1.341104507446289e-07, 'grpo_std_advantage': 1.507950400991831e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5495311617851257, 'epoch': 1.94}
+{'loss': -0.7015, 'grad_norm': 2.8475866317749023, 'learning_rate': 6.6132713508446075e-09, 'grpo_mean_advantage': 2.689659481802664e-07, 'grpo_std_advantage': 8.491958851664094e-07, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.549436628818512, 'epoch': 1.96}
+{'loss': -0.4033, 'grad_norm': 2.9422402381896973, 'learning_rate': 1.9681946484320645e-09, 'grpo_mean_advantage': 8.195638656616211e-08, 'grpo_std_advantage': 3.802849732892355e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.544632077217102, 'epoch': 1.98}
+{'loss': -0.6773, 'grad_norm': 2.66204833984375, 'learning_rate': 5.467904943851077e-11, 'grpo_mean_advantage': 7.552536089860951e-07, 'grpo_std_advantage': 4.143997102801222e-06, 'grpo_mean_kl_div': 0.0, 'grpo_mean_group_score': 0.5968535542488098, 'epoch': 2.0}
+{'train_runtime': 8430.3461, 'train_samples_per_second': 0.474, 'train_steps_per_second': 0.059, 'train_loss': -0.23323759501613678, 'epoch': 2.0}
+Saved best adapter -> runs/grpo_14b_run1/best_adapter
+Merge disabled. Run with --merge-only later if needed.

wandb/run-20251227_194423-jz7bptqa/files/requirements.txt

@@ -0,0 +1,102 @@
+exceptiongroup==1.3.1
+wheel==0.45.1
+python-dateutil==2.9.0.post0
+huggingface-hub==0.36.0
+idna==3.11
+click==8.3.1
+numpy==2.2.6
+httpx==0.28.1
+tokenizers==0.22.1
+sympy==1.13.1
+yarl==1.22.0
+async-timeout==5.0.1
+datasets==4.4.2
+einops==0.8.1
+platformdirs==4.5.1
+nvidia-cuda-cupti-cu12==12.1.105
+nvidia-nvtx-cu12==12.1.105
+smmap==5.0.2
+accelerate==1.12.0
+requests==2.32.5
+aiohttp==3.13.2
+transformers==4.57.3
+bitsandbytes==0.49.0
+evaluate==0.4.6
+nvidia-cublas-cu12==12.1.3.1
+mpmath==1.3.0
+torchaudio==2.5.1+cu121
+nvidia-cuda-runtime-cu12==12.1.105
+typing-inspection==0.4.2
+GitPython==3.1.45
+xxhash==3.6.0
+nvidia-cusolver-cu12==11.4.5.107
+pydantic_core==2.41.5
+six==1.17.0
+torchvision==0.20.1+cu121
+typing_extensions==4.15.0
+triton==3.1.0
+charset-normalizer==3.4.4
+wandb==0.23.1
+regex==2025.11.3
+pip==25.3
+nvidia-cusparse-cu12==12.1.0.106
+pytz==2025.2
+Jinja2==3.1.6
+psutil==7.2.0
+pillow==12.0.0
+packaging==25.0
+safetensors==0.7.0
+sentry-sdk==2.48.0
+gitdb==4.0.12
+httpcore==1.0.9
+setuptools==80.9.0
+nvidia-cufft-cu12==11.0.2.54
+flash_attn==2.8.3
+anyio==4.12.0
+pydantic==2.12.5
+fsspec==2025.10.0
+filelock==3.20.0
+PyYAML==6.0.3
+hf-xet==1.2.0
+nvidia-cudnn-cu12==9.1.0.70
+tqdm==4.67.1
+MarkupSafe==2.1.5
+attrs==25.4.0
+nvidia-cuda-nvrtc-cu12==12.1.105
+peft==0.18.0
+aiohappyeyeballs==2.6.1
+networkx==3.4.2
+nvidia-nvjitlink-cu12==12.9.86
+certifi==2025.11.12
+pyarrow==22.0.0
+dill==0.4.0
+protobuf==6.33.2
+aiosignal==1.4.0
+frozenlist==1.8.0
+urllib3==2.6.2
+propcache==0.4.1
+tzdata==2025.3
+pandas==2.3.3
+annotated-types==0.7.0
+nvidia-nccl-cu12==2.21.5
+multidict==6.7.0
+nvidia-curand-cu12==10.3.2.106
+torch==2.5.1+cu121
+h11==0.16.0
+multiprocess==0.70.18
+wheel==0.45.1
+tomli==2.0.1
+autocommand==2.2.2
+jaraco.context==5.3.0
+zipp==3.19.2
+packaging==24.2
+inflect==7.3.1
+typing_extensions==4.12.2
+platformdirs==4.2.2
+jaraco.functools==4.0.1
+jaraco.collections==5.1.0
+jaraco.text==3.12.1
+backports.tarfile==1.2.0
+more-itertools==10.3.0
+importlib_metadata==8.0.0
+typeguard==4.3.0

wandb/run-20251227_194423-jz7bptqa/files/wandb-metadata.json

@@ -0,0 +1,47 @@
+{
+  "os":  "Linux-6.12.46+-x86_64-with-glibc2.35",
+  "python":  "CPython 3.10.12",
+  "startedAt":  "2025-12-27T19:44:23.484378Z",
+  "args":  [
+    "--config",
+    "config_grpo_14b.yaml"
+  ],
+  "program":  "/workspace/trainer-kit/GRPO-14b/run_grpo_fixed.py",
+  "codePath":  "run_grpo_fixed.py",
+  "codePathLocal":  "run_grpo_fixed.py",
+  "email":  "shaiksirajuddin9949@gmail.com",
+  "root":  "runs/grpo_14b_run1",
+  "host":  "a100-2gpu-shell-session-757d587799-mfdvv",
+  "executable":  "/root/llm_finetuning_env/bin/python",
+  "cpu_count":  12,
+  "cpu_count_logical":  24,
+  "gpu":  "NVIDIA A100-SXM4-80GB",
+  "gpu_count":  2,
+  "disk":  {
+    "/":  {
+      "total":  "791251738624",
+      "used":  "537236914176"
+    }
+  },
+  "memory":  {
+    "total":  "359047892992"
+  },
+  "gpu_nvidia":  [
+    {
+      "name":  "NVIDIA A100-SXM4-80GB",
+      "memoryTotal":  "85899345920",
+      "cudaCores":  6912,
+      "architecture":  "Ampere",
+      "uuid":  "GPU-989794b0-ec3b-13bf-db9f-3fbe341497ba"
+    },
+    {
+      "name":  "NVIDIA A100-SXM4-80GB",
+      "memoryTotal":  "85899345920",
+      "cudaCores":  6912,
+      "architecture":  "Ampere",
+      "uuid":  "GPU-3790aa64-60ef-9eac-b0b1-b278ee8c0d40"
+    }
+  ],
+  "cudaVersion":  "13.0",
+  "writerId":  "flq59s27gswy5t5lhj0bnmni5u7y1dm1"
+}

wandb/run-20251227_194423-jz7bptqa/files/wandb-summary.json

@@ -0,0 +1 @@
+{"_timestamp":1.7668731631222892e+09,"total_flos":0,"train/grpo_mean_group_score":0.5968535542488098,"_wandb":{"runtime":8499},"train/loss":-0.6773,"train/epoch":2,"train/grpo_mean_kl_div":0,"train/learning_rate":5.467904943851077e-11,"train/grpo_mean_advantage":7.552536089860951e-07,"train_steps_per_second":0.059,"train_samples_per_second":0.474,"_step":100,"train/grad_norm":2.66204833984375,"train_runtime":8430.3461,"_runtime":8499,"train_loss":-0.23323759501613678,"train/global_step":500,"train/grpo_std_advantage":4.143997102801222e-06}

wandb/run-20251227_194423-jz7bptqa/logs/debug-core.log

@@ -0,0 +1,16 @@
+{"time":"2025-12-27T19:44:23.589075701Z","level":"INFO","msg":"main: starting server","port-filename":"/tmp/tmpzezpmqfn/port-592889.txt","pid":592889,"log-level":0,"disable-analytics":false,"shutdown-on-parent-exit":false,"enable-dcgm-profiling":false}
+{"time":"2025-12-27T19:44:23.589840679Z","level":"INFO","msg":"server: will exit if parent process dies","ppid":592889}
+{"time":"2025-12-27T19:44:23.589796884Z","level":"INFO","msg":"server: accepting connections","addr":{"Name":"/tmp/wandb-592889-592949-1509800020/socket","Net":"unix"}}
+{"time":"2025-12-27T19:44:23.771854762Z","level":"INFO","msg":"connection: ManageConnectionData: new connection created","id":"1(@)"}
+{"time":"2025-12-27T19:44:23.778560595Z","level":"INFO","msg":"handleInformInit: received","streamId":"jz7bptqa","id":"1(@)"}
+{"time":"2025-12-27T19:44:23.931963354Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"jz7bptqa","id":"1(@)"}
+{"time":"2025-12-27T22:06:04.30361987Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"jz7bptqa","id":"1(@)"}
+{"time":"2025-12-27T22:06:04.304644358Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"jz7bptqa","id":"1(@)"}
+{"time":"2025-12-27T22:06:04.36465465Z","level":"INFO","msg":"handleInformTeardown: server teardown initiated","id":"1(@)"}
+{"time":"2025-12-27T22:06:04.364694525Z","level":"INFO","msg":"handleInformTeardown: server shutdown complete","id":"1(@)"}
+{"time":"2025-12-27T22:06:04.36471081Z","level":"INFO","msg":"connection: closing","id":"1(@)"}
+{"time":"2025-12-27T22:06:04.364733088Z","level":"INFO","msg":"server is shutting down"}
+{"time":"2025-12-27T22:06:04.364754928Z","level":"INFO","msg":"connection: closed successfully","id":"1(@)"}
+{"time":"2025-12-27T22:06:04.364766781Z","level":"INFO","msg":"connection: ManageConnectionData: connection closed","id":"1(@)"}
+{"time":"2025-12-27T22:06:04.364871744Z","level":"INFO","msg":"server: listener closed","addr":{"Name":"/tmp/wandb-592889-592949-1509800020/socket","Net":"unix"}}
+{"time":"2025-12-27T22:06:04.364906416Z","level":"INFO","msg":"server is closed"}

wandb/run-20251227_194423-jz7bptqa/logs/debug-internal.log

@@ -0,0 +1,12 @@
+{"time":"2025-12-27T19:44:23.778699792Z","level":"INFO","msg":"stream: starting","core version":"0.23.1"}
+{"time":"2025-12-27T19:44:23.931692267Z","level":"INFO","msg":"stream: created new stream","id":"jz7bptqa"}
+{"time":"2025-12-27T19:44:23.931816408Z","level":"INFO","msg":"handler: started","stream_id":"jz7bptqa"}
+{"time":"2025-12-27T19:44:23.931950499Z","level":"INFO","msg":"stream: started","id":"jz7bptqa"}
+{"time":"2025-12-27T19:44:23.931981018Z","level":"INFO","msg":"writer: started","stream_id":"jz7bptqa"}
+{"time":"2025-12-27T19:44:23.931980519Z","level":"INFO","msg":"sender: started","stream_id":"jz7bptqa"}
+{"time":"2025-12-27T22:06:04.101056437Z","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"}
+{"time":"2025-12-27T22:06:04.298272513Z","level":"INFO","msg":"handler: operation stats","stats":{}}
+{"time":"2025-12-27T22:06:04.303656494Z","level":"INFO","msg":"stream: closing","id":"jz7bptqa"}
+{"time":"2025-12-27T22:06:04.303677603Z","level":"INFO","msg":"handler: closed","stream_id":"jz7bptqa"}
+{"time":"2025-12-27T22:06:04.303750712Z","level":"INFO","msg":"sender: closed","stream_id":"jz7bptqa"}
+{"time":"2025-12-27T22:06:04.303767265Z","level":"INFO","msg":"stream: closed","id":"jz7bptqa"}

wandb/run-20251227_194423-jz7bptqa/logs/debug.log

@@ -0,0 +1,29 @@
+2025-12-27 19:44:23,485 INFO    MainThread:592889 [wandb_setup.py:_flush():80] Current SDK version is 0.23.1
+2025-12-27 19:44:23,486 INFO    MainThread:592889 [wandb_setup.py:_flush():80] Configure stats pid to 592889
+2025-12-27 19:44:23,486 INFO    MainThread:592889 [wandb_setup.py:_flush():80] Loading settings from /root/.config/wandb/settings
+2025-12-27 19:44:23,486 INFO    MainThread:592889 [wandb_setup.py:_flush():80] Loading settings from /workspace/trainer-kit/GRPO-14b/wandb/settings
+2025-12-27 19:44:23,486 INFO    MainThread:592889 [wandb_setup.py:_flush():80] Loading settings from environment variables
+2025-12-27 19:44:23,486 INFO    MainThread:592889 [wandb_init.py:setup_run_log_directory():714] Logging user logs to runs/grpo_14b_run1/wandb/run-20251227_194423-jz7bptqa/logs/debug.log
+2025-12-27 19:44:23,486 INFO    MainThread:592889 [wandb_init.py:setup_run_log_directory():715] Logging internal logs to runs/grpo_14b_run1/wandb/run-20251227_194423-jz7bptqa/logs/debug-internal.log
+2025-12-27 19:44:23,486 INFO    MainThread:592889 [wandb_init.py:init():841] calling init triggers
+2025-12-27 19:44:23,486 INFO    MainThread:592889 [wandb_init.py:init():846] wandb.init called with sweep_config: {}
+config: {'model': {'repo_id': '/workspace/Models/Qwen2.5-Coder-14B-CPT-SFT_v2', 'tokenizer_name': 'Qwen/Qwen2.5-Coder-14B', 'load_in_8bit': False, 'load_in_4bit': False, 'torch_dtype': 'bfloat16', 'device_map': 'auto', 'trust_remote_code': True}, 'data': {'train_jsonl': 'grpo_dataset.jsonl', 'eval_jsonl': None, 'eval_split_ratio': 0.0, 'shuffle': True, 'num_proc': 1, 'prompt_field': 'prompt', 'completions_field': 'completions', 'scores_field': 'scores', 'format_type': 'raw', 'max_length': 2048, 'min_completions': 2, 'system_prompt': 'You are a Hyperswitch Rust code analyzer. Identify functions/structs that need modification for a given task.\n\n## Output Format\n\n##OUTPUT\nExplain the data flow and why each component must change:\n- Flow: [Input → Processing → Output with arrows]\n- For each component: "The [ComponentName] ([path]) must [action] because [reason]—without this, [consequence]"\n- Explain coupling between components\n\n##SELECT\nmodify::crates/path/to/file.rs::impl::ComponentName\nadd::crates/another/file.rs::function::AnotherComponent\n<EOS>\n\n## Rules\n\n1. Use full paths: `remove::crates/folder/file.rs::Type::Name`\n2. Use `::` for nested items: `status::StructName::Type::Name`\n3. Always explain "must change because" and "without this"\n3. Types of components: function, struct, enum, impl, trait\n4. If there is extra information (e.g., enum variants), include that too.\n5. Start with ##OUTPUT, end with ##SELECT, terminate with <EOS>\n\n## Example\n\n##TASK\nAdd webhook subscription support\n\n##OUTPUT\nThe webhook system routes events via EventClass enum. Flow: webhook → EventClass → handler → processing. The EventClass enum (crates/common_enums/src/enums.rs::EventClass) must add Subscriptions variant because it defines event routing—without this, subscription events cannot be processed. The SubscriptionStatus impl (crates/common_enums/src/transformers.rs::SubscriptionStatus) must map to EventType because it converts status to events—without this, status changes don\'t trigger webhooks. These are coupled: EventClass routes to handlers that use SubscriptionStatus mappings.\n\n##SELECT\ncrates/common_enums/src/enums.rs::EventClass\ncrates/common_enums/src/transformers.rs::SubscriptionStatus\n<EOS>\n', 'custom_template': '##INSTRUCTION\n{instruction}<|im_end|>\n{input}<|im_end|>\n{output}<|im_end|>'}, 'peft': {'enabled': True, 'r': 16, 'lora_alpha': 32, 'lora_dropout': 0.05, 'target_modules': ['q_proj', 'k_proj', 'v_proj', 'o_proj', 'gate_proj', 'up_proj', 'down_proj'], 'bias': 'none', 'task_type': 'CAUSAL_LM'}, 'grpo': {'group_size': 4, 'kl_coef': 0.05, 'normalize_advantages': True, 'reward_scaling': 1.0, 'reward_bias': 0.0, 'reward_clip': 5.0, 'advantage_temperature': 1.0, 'use_reference_model': False, 'seed': 42}, 'train': {'output_dir': 'runs/grpo_14b_run1', 'num_train_epochs': 2, 'per_device_train_batch_size': 1, 'gradient_accumulation_steps': 8, 'per_device_eval_batch_size': 1, 'learning_rate': 5e-06, 'weight_decay': 0.01, 'warmup_ratio': 0.05, 'lr_scheduler_type': 'cosine', 'fp16': False, 'bf16': True, 'max_grad_norm': 1.0, 'gradient_checkpointing': True, 'logging_steps': 5, 'save_steps': 100, 'save_total_limit': 2, 'evaluation_strategy': 'no', 'dataloader_num_workers': 4, 'dataloader_pin_memory': True, 'remove_unused_columns': False, 'report_to': [], 'seed': 42, 'ddp_find_unused_parameters': False}, 'run_dir': 'runs/grpo_14b_run1', '_wandb': {}}
+2025-12-27 19:44:23,486 INFO    MainThread:592889 [wandb_init.py:init():889] starting backend
+2025-12-27 19:44:23,772 INFO    MainThread:592889 [wandb_init.py:init():892] sending inform_init request
+2025-12-27 19:44:23,776 INFO    MainThread:592889 [wandb_init.py:init():900] backend started and connected
+2025-12-27 19:44:23,778 INFO    MainThread:592889 [wandb_init.py:init():970] updated telemetry
+2025-12-27 19:44:23,779 INFO    MainThread:592889 [wandb_init.py:init():994] communicating run to backend with 90.0 second timeout
+2025-12-27 19:44:24,241 INFO    MainThread:592889 [wandb_init.py:init():1041] starting run threads in backend
+2025-12-27 19:44:24,354 INFO    MainThread:592889 [wandb_run.py:_console_start():2521] atexit reg
+2025-12-27 19:44:24,354 INFO    MainThread:592889 [wandb_run.py:_redirect():2369] redirect: wrap_raw
+2025-12-27 19:44:24,354 INFO    MainThread:592889 [wandb_run.py:_redirect():2438] Wrapping output streams.
+2025-12-27 19:44:24,354 INFO    MainThread:592889 [wandb_run.py:_redirect():2461] Redirects installed.
+2025-12-27 19:44:24,358 INFO    MainThread:592889 [wandb_init.py:init():1081] run started, returning control to user process
+2025-12-27 19:45:32,787 INFO    MainThread:592889 [wandb_run.py:_config_callback():1396] config_cb None None {'peft_config': {'default': {'task_type': 'CAUSAL_LM', 'peft_type': 'LORA', 'auto_mapping': None, 'peft_version': '0.18.0', 'base_model_name_or_path': '/workspace/Models/Qwen2.5-Coder-14B-CPT-SFT_v2', 'revision': None, 'inference_mode': False, 'r': 16, 'target_modules': ['q_proj', 'o_proj', 'v_proj', 'up_proj', 'gate_proj', 'down_proj', 'k_proj'], 'exclude_modules': None, 'lora_alpha': 32, 'lora_dropout': 0.05, 'fan_in_fan_out': False, 'bias': 'none', 'use_rslora': False, 'modules_to_save': None, 'init_lora_weights': True, 'layers_to_transform': None, 'layers_pattern': None, 'rank_pattern': {}, 'alpha_pattern': {}, 'megatron_config': None, 'megatron_core': 'megatron.core', 'trainable_token_indices': None, 'loftq_config': {}, 'eva_config': None, 'corda_config': None, 'use_dora': False, 'alora_invocation_tokens': None, 'use_qalora': False, 'qalora_group_size': 16, 'layer_replication': None, 'runtime_config': {'ephemeral_gpu_offload': False}, 'lora_bias': False, 'target_parameters': None, 'arrow_config': None, 'ensure_weight_tying': False}}, 'vocab_size': 152064, 'max_position_embeddings': 32768, 'hidden_size': 5120, 'intermediate_size': 13824, 'num_hidden_layers': 48, 'num_attention_heads': 40, 'use_sliding_window': False, 'sliding_window': None, 'max_window_layers': 48, 'num_key_value_heads': 8, 'hidden_act': 'silu', 'initializer_range': 0.02, 'rms_norm_eps': 1e-06, 'use_cache': False, 'rope_theta': 10000.0, 'rope_scaling': None, 'attention_dropout': 0.0, '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', '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'], 'return_dict': True, 'output_hidden_states': False, 'torchscript': False, 'dtype': 'bfloat16', 'pruned_heads': {}, 'tie_word_embeddings': False, 'chunk_size_feed_forward': 0, 'is_encoder_decoder': False, 'is_decoder': False, 'cross_attention_hidden_size': None, 'add_cross_attention': False, 'tie_encoder_decoder': False, 'architectures': ['Qwen2ForCausalLM'], 'finetuning_task': None, 'id2label': {0: 'LABEL_0', 1: 'LABEL_1'}, 'label2id': {'LABEL_0': 0, 'LABEL_1': 1}, 'task_specific_params': None, 'problem_type': None, 'tokenizer_class': None, 'prefix': None, 'bos_token_id': 151643, 'pad_token_id': None, 'eos_token_id': 151643, 'sep_token_id': None, 'decoder_start_token_id': None, 'max_length': 20, 'min_length': 0, 'do_sample': False, 'early_stopping': False, 'num_beams': 1, 'temperature': 1.0, 'top_k': 50, 'top_p': 1.0, 'typical_p': 1.0, 'repetition_penalty': 1.0, 'length_penalty': 1.0, 'no_repeat_ngram_size': 0, 'encoder_no_repeat_ngram_size': 0, 'bad_words_ids': None, 'num_return_sequences': 1, 'output_scores': False, 'return_dict_in_generate': False, 'forced_bos_token_id': None, 'forced_eos_token_id': None, 'remove_invalid_values': False, 'exponential_decay_length_penalty': None, 'suppress_tokens': None, 'begin_suppress_tokens': None, 'num_beam_groups': 1, 'diversity_penalty': 0.0, '_name_or_path': '/workspace/Models/Qwen2.5-Coder-14B-CPT-SFT_v2', 'transformers_version': '4.57.3', 'model_type': 'qwen2', 'rope_parameters': {'rope_theta': 1000000.0, 'rope_type': 'default'}, 'tf_legacy_loss': False, 'use_bfloat16': False, 'output_attentions': False, 'output_dir': 'runs/grpo_14b_run1/checkpoints', 'overwrite_output_dir': False, 'do_train': False, 'do_eval': False, 'do_predict': False, 'eval_strategy': 'no', 'prediction_loss_only': False, 'per_device_train_batch_size': 1, 'per_device_eval_batch_size': 1, 'per_gpu_train_batch_size': None, 'per_gpu_eval_batch_size': None, 'gradient_accumulation_steps': 8, 'eval_accumulation_steps': None, 'eval_delay': 0, 'torch_empty_cache_steps': None, 'learning_rate': 5e-06, 'weight_decay': 0.01, 'adam_beta1': 0.9, 'adam_beta2': 0.999, 'adam_epsilon': 1e-08, 'max_grad_norm': 1.0, 'num_train_epochs': 2.0, 'max_steps': -1, 'lr_scheduler_type': 'cosine', 'lr_scheduler_kwargs': {}, 'warmup_ratio': 0.05, 'warmup_steps': 0, 'log_level': 'passive', 'log_level_replica': 'warning', 'log_on_each_node': True, 'logging_dir': 'runs/grpo_14b_run1/checkpoints/runs/Dec27_19-45-32_a100-2gpu-shell-session-757d587799-mfdvv', 'logging_strategy': 'steps', 'logging_first_step': False, 'logging_steps': 5, 'logging_nan_inf_filter': True, 'save_strategy': 'steps', 'save_steps': 100, 'save_total_limit': 2, 'save_safetensors': True, 'save_on_each_node': False, 'save_only_model': False, 'restore_callback_states_from_checkpoint': False, 'no_cuda': False, 'use_cpu': False, 'use_mps_device': False, 'seed': 42, 'data_seed': None, 'jit_mode_eval': False, 'bf16': True, 'fp16': False, 'fp16_opt_level': 'O1', 'half_precision_backend': 'auto', 'bf16_full_eval': False, 'fp16_full_eval': False, 'tf32': None, 'local_rank': 0, 'ddp_backend': None, 'tpu_num_cores': None, 'tpu_metrics_debug': False, 'debug': [], 'dataloader_drop_last': False, 'eval_steps': 50, 'dataloader_num_workers': 0, 'dataloader_prefetch_factor': None, 'past_index': -1, 'run_name': None, 'disable_tqdm': False, 'remove_unused_columns': False, 'label_names': None, 'load_best_model_at_end': False, 'metric_for_best_model': 'eval_loss', 'greater_is_better': False, 'ignore_data_skip': False, 'fsdp': [], 'fsdp_min_num_params': 0, 'fsdp_config': {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}, 'fsdp_transformer_layer_cls_to_wrap': None, 'accelerator_config': {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}, 'parallelism_config': None, 'deepspeed': None, 'label_smoothing_factor': 0.0, 'optim': 'adamw_torch', 'optim_args': None, 'adafactor': False, 'group_by_length': False, 'length_column_name': 'length', 'report_to': ['wandb'], 'project': 'huggingface', 'trackio_space_id': 'trackio', 'ddp_find_unused_parameters': None, 'ddp_bucket_cap_mb': None, 'ddp_broadcast_buffers': None, 'dataloader_pin_memory': True, 'dataloader_persistent_workers': False, 'skip_memory_metrics': True, 'use_legacy_prediction_loop': False, 'push_to_hub': False, 'resume_from_checkpoint': None, 'hub_model_id': None, 'hub_strategy': 'every_save', 'hub_token': '<HUB_TOKEN>', 'hub_private_repo': None, 'hub_always_push': False, 'hub_revision': None, 'gradient_checkpointing': False, 'gradient_checkpointing_kwargs': None, 'include_inputs_for_metrics': False, 'include_for_metrics': [], 'eval_do_concat_batches': True, 'fp16_backend': 'auto', 'push_to_hub_model_id': None, 'push_to_hub_organization': None, 'push_to_hub_token': '<PUSH_TO_HUB_TOKEN>', 'mp_parameters': '', 'auto_find_batch_size': False, 'full_determinism': False, 'torchdynamo': None, 'ray_scope': 'last', 'ddp_timeout': 1800, 'torch_compile': False, 'torch_compile_backend': None, 'torch_compile_mode': None, 'include_tokens_per_second': False, 'include_num_input_tokens_seen': 'no', 'neftune_noise_alpha': None, 'optim_target_modules': None, 'batch_eval_metrics': False, 'eval_on_start': False, 'use_liger_kernel': False, 'liger_kernel_config': None, 'eval_use_gather_object': False, 'average_tokens_across_devices': True}
+2025-12-27 19:45:32,798 INFO    MainThread:592889 [wandb_config.py:__setitem__():154] [no run ID] config set model/num_parameters = 14838846464 - <bound method Run._config_callback of <wandb.sdk.wandb_run.Run object at 0x7ada527aff70>>
+2025-12-27 19:45:32,798 INFO    MainThread:592889 [wandb_run.py:_config_callback():1396] config_cb model/num_parameters 14838846464 None
+2025-12-27 22:06:03,800 INFO    MainThread:592889 [wandb_run.py:_finish():2287] finishing run sirajuddin-shaik-007/rl-training/jz7bptqa
+2025-12-27 22:06:03,800 INFO    MainThread:592889 [wandb_run.py:_atexit_cleanup():2486] got exitcode: 0
+2025-12-27 22:06:03,801 INFO    MainThread:592889 [wandb_run.py:_restore():2468] restore
+2025-12-27 22:06:03,801 INFO    MainThread:592889 [wandb_run.py:_restore():2474] restore done
+2025-12-27 22:06:04,302 INFO    MainThread:592889 [wandb_run.py:_footer_sync_info():3862] logging synced files