Model save

README.md

@@ -1,11 +1,9 @@
 ---
 base_model: coolcui/DeepSeek-R1-Distill-Qwen-1.5B-GRPO
-datasets: chansung/verifiable-coding-problems-python
 library_name: transformers
 model_name: Qwen2.5-1.5B-CCRL-1
 tags:
 - generated_from_trainer
-- open-r1
 - trl
 - grpo
 licence: license
@@ -13,7 +11,7 @@ licence: license
 
 # Model Card for Qwen2.5-1.5B-CCRL-1
 
-This model is a fine-tuned version of [coolcui/DeepSeek-R1-Distill-Qwen-1.5B-GRPO](https://huggingface.co/coolcui/DeepSeek-R1-Distill-Qwen-1.5B-GRPO) on the [chansung/verifiable-coding-problems-python](https://huggingface.co/datasets/chansung/verifiable-coding-problems-python) dataset.
+This model is a fine-tuned version of [coolcui/DeepSeek-R1-Distill-Qwen-1.5B-GRPO](https://huggingface.co/coolcui/DeepSeek-R1-Distill-Qwen-1.5B-GRPO).
 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/chansung18/huggingface/runs/opz13onj) 
+[<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/chansung18/huggingface/runs/oyhi6d8w) 
 
 
 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": 0.0,
-    "train_runtime": 13359.4688,
+    "train_runtime": 295.3751,
     "train_samples": 949,
-    "train_samples_per_second": 0.479,
-    "train_steps_per_second": 0.004
+    "train_samples_per_second": 21.667,
+    "train_steps_per_second": 0.169
 }

train_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
     "train_loss": 0.0,
-    "train_runtime": 13359.4688,
+    "train_runtime": 295.3751,
     "train_samples": 949,
-    "train_samples_per_second": 0.479,
-    "train_steps_per_second": 0.004
+    "train_samples_per_second": 21.667,
+    "train_steps_per_second": 0.169
 }

trainer_state.json

@@ -2,9 +2,9 @@
   "best_global_step": null,
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 0.8403361344537815,
+  "epoch": 0.8571428571428571,
   "eval_steps": 500,
-  "global_step": 50,
+  "global_step": 51,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -710,13 +710,27 @@
       "step": 50
     },
     {
-      "epoch": 0.8403361344537815,
-      "step": 50,
+      "clip_ratio": 0.0,
+      "completion_length": 1704.546875,
+      "epoch": 0.8571428571428571,
+      "grad_norm": 0.0,
+      "kl": 0.0,
+      "learning_rate": 0.0,
+      "loss": 0.0,
+      "num_tokens": 281542.0,
+      "reward": 0.0,
+      "reward_std": 0.0,
+      "rewards/curriculum_aware_reward_fn": 0.0,
+      "step": 51
+    },
+    {
+      "epoch": 0.8571428571428571,
+      "step": 51,
       "total_flos": 0.0,
       "train_loss": 0.0,
-      "train_runtime": 13359.4688,
-      "train_samples_per_second": 0.479,
-      "train_steps_per_second": 0.004
+      "train_runtime": 295.3751,
+      "train_samples_per_second": 21.667,
+      "train_steps_per_second": 0.169
     }
   ],
   "logging_steps": 1,