Model save

README.md

@@ -1,11 +1,9 @@
 ---
 base_model: Qwen/Qwen2.5-3B-Instruct
-datasets: open-r1/verifiable-coding-problems-python-10k_decontaminated
 library_name: transformers
 model_name: Qwen2.5-3B-Open-R1-Code-GRPO-r2
 tags:
 - generated_from_trainer
-- open-r1
 - trl
 - grpo
 licence: license
@@ -13,7 +11,7 @@ licence: license
 
 # Model Card for Qwen2.5-3B-Open-R1-Code-GRPO-r2
 
-This model is a fine-tuned version of [Qwen/Qwen2.5-3B-Instruct](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct) on the [open-r1/verifiable-coding-problems-python-10k_decontaminated](https://huggingface.co/datasets/open-r1/verifiable-coding-problems-python-10k_decontaminated) dataset.
+This model is a fine-tuned version of [Qwen/Qwen2.5-3B-Instruct](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct).
 It has been trained using [TRL](https://github.com/huggingface/trl).
 
 ## Quick start
@@ -29,7 +27,7 @@ print(output["generated_text"])
 
 ## Training procedure
 
-[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/calads/huggingface/runs/ok073653) 
+[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/calads/huggingface/runs/ihnrm7q2) 
 
 
 This model was trained with GRPO, a method introduced in [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://huggingface.co/papers/2402.03300).

all_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 0.39302948499191553,
-    "train_runtime": 15190.2462,
+    "train_loss": 4.563784550228519e-06,
+    "train_runtime": 168.9071,
     "train_samples": 1574,
-    "train_samples_per_second": 0.23,
-    "train_steps_per_second": 0.033
+    "train_samples_per_second": 20.721,
+    "train_steps_per_second": 2.96
 }

train_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 0.39302948499191553,
-    "train_runtime": 15190.2462,
+    "train_loss": 4.563784550228519e-06,
+    "train_runtime": 168.9071,
     "train_samples": 1574,
-    "train_samples_per_second": 0.23,
-    "train_steps_per_second": 0.033
+    "train_samples_per_second": 20.721,
+    "train_steps_per_second": 2.96
 }

trainer_state.json

@@ -1,9 +1,9 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 0.3176620076238882,
+  "epoch": 0.318297331639136,
   "eval_steps": 500,
-  "global_step": 500,
+  "global_step": 501,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -7509,13 +7509,28 @@
       "step": 500
     },
     {
-      "epoch": 0.3176620076238882,
-      "step": 500,
+      "clip_ratio": 0.0,
+      "completion_length": 444.857177734375,
+      "epoch": 0.318297331639136,
+      "grad_norm": 0.047799259424209595,
+      "kl": 0.19921875,
+      "learning_rate": 5.000472027468528e-07,
+      "loss": 0.0023,
+      "reward": 0.40000003576278687,
+      "reward_std": 0.0,
+      "rewards/code_format_reward": 1.0,
+      "rewards/code_reward": 0.0,
+      "rewards/format_reward": 1.0,
+      "step": 501
+    },
+    {
+      "epoch": 0.318297331639136,
+      "step": 501,
       "total_flos": 0.0,
-      "train_loss": 0.39302948499191553,
-      "train_runtime": 15190.2462,
-      "train_samples_per_second": 0.23,
-      "train_steps_per_second": 0.033
+      "train_loss": 4.563784550228519e-06,
+      "train_runtime": 168.9071,
+      "train_samples_per_second": 20.721,
+      "train_steps_per_second": 2.96
     }
   ],
   "logging_steps": 1,