Model save

README.md

@@ -27,7 +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/ziyaow-university-of-maryland/Qwen2.5-FL-GRPO/runs/8qw6jaei) 
+[<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/ziyaow-university-of-maryland/Qwen2.5-FL-GRPO/runs/yw682lur) 
 
 
 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

@@ -1,9 +1,9 @@
 {
     "client_id": 0,
     "total_flos": 0.0,
-    "train_loss": 0.011762881651520729,
-    "train_runtime": 227.9358,
+    "train_loss": 0.02052961972852548,
+    "train_runtime": 228.5975,
     "train_samples": 100,
-    "train_samples_per_second": 0.439,
-    "train_steps_per_second": 0.018
+    "train_samples_per_second": 0.437,
+    "train_steps_per_second": 0.017
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2a02b5d213d10ea7b3216278fb690d8cdd9e10acac8aa479c58376f4cc3f2523
+oid sha256:43e3e6994b555e9237fac62dfb2d3dfc2bac5caa5edb8e497406934675189386
 size 3087467144

train_client_0_results.json

@@ -1,9 +1,9 @@
 {
     "client_id": 0,
     "total_flos": 0.0,
-    "train_loss": 0.011762881651520729,
-    "train_runtime": 227.9358,
+    "train_loss": 0.02052961972852548,
+    "train_runtime": 228.5975,
     "train_samples": 100,
-    "train_samples_per_second": 0.439,
-    "train_steps_per_second": 0.018
+    "train_samples_per_second": 0.437,
+    "train_steps_per_second": 0.017
 }

trainer_state.json

@@ -24,7 +24,7 @@
       "completions/min_terminated_length": 44.0,
       "epoch": 0.3076923076923077,
       "frac_reward_zero_std": 0.28125,
-      "grad_norm": 0.2418988953762188,
+      "grad_norm": 0.24194501054197542,
       "kl": 0.00023603439331054688,
       "learning_rate": 0.0,
       "loss": 0.0155,
@@ -45,36 +45,36 @@
       "clip_ratio/low_mean": 0.0,
       "clip_ratio/low_min": 0.0,
       "clip_ratio/region_mean": 0.0,
-      "completions/clipped_ratio": -6.390625,
+      "completions/clipped_ratio": -6.34375,
       "completions/max_length": 1024.0,
-      "completions/max_terminated_length": 1022.0,
-      "completions/mean_length": 648.22265625,
-      "completions/mean_terminated_length": 617.1923522949219,
+      "completions/max_terminated_length": 1020.0,
+      "completions/mean_length": 647.62109375,
+      "completions/mean_terminated_length": 613.8444519042969,
       "completions/min_length": 71.5,
       "completions/min_terminated_length": 71.5,
       "epoch": 0.9230769230769231,
       "frac_reward_zero_std": 0.171875,
-      "kl": 0.006052374839782715,
-      "num_tokens": 1237385.0,
-      "reward": 0.3359375,
-      "reward_std": 0.2339080274105072,
-      "rewards/accuracy_reward/mean": 0.0283203125,
-      "rewards/accuracy_reward/std": 0.16538910567760468,
-      "rewards/format_reward/mean": 0.0078125,
-      "rewards/format_reward/std": 0.08741512894630432,
-      "rewards/tag_count_reward/mean": 0.2998046875,
-      "rewards/tag_count_reward/std": 0.20496860146522522,
+      "kl": 0.006288886070251465,
+      "num_tokens": 1236769.0,
+      "reward": 0.32763671875,
+      "reward_std": 0.2266850769519806,
+      "rewards/accuracy_reward/mean": 0.021484375,
+      "rewards/accuracy_reward/std": 0.13888774067163467,
+      "rewards/format_reward/mean": 0.0068359375,
+      "rewards/format_reward/std": 0.0822625607252121,
+      "rewards/tag_count_reward/mean": 0.29931640625,
+      "rewards/tag_count_reward/std": 0.20676623284816742,
       "step": 3,
       "total_flos": 0.0,
-      "train_loss": 0.011762881651520729,
-      "train_runtime": 227.9358,
-      "train_samples_per_second": 0.439,
-      "train_steps_per_second": 0.018
+      "train_loss": 0.02052961972852548,
+      "train_runtime": 228.5975,
+      "train_samples_per_second": 0.437,
+      "train_steps_per_second": 0.017
     }
   ],
   "logging_steps": 10,
   "max_steps": 4,
-  "num_input_tokens_seen": 1237385,
+  "num_input_tokens_seen": 1236769,
   "num_train_epochs": 1,
   "save_steps": 500,
   "stateful_callbacks": {

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8a848fc19a889b3819b4eb4627e9e75c9cca4d623fab69f3214d3f027e2d3a56
+oid sha256:c1cbef9eff8ac8cc301cfb6c7a8508d362641a4f3d5ccd2118ea73ec5653460d
 size 8440