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/sjhxbug218-nvidia/huggingface/runs/jika9i2p) 
+[<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/sjhxbug218-nvidia/huggingface/runs/csdr8cxx) 
 
 
 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": 0.008086015470325947,
-    "train_runtime": 126.4484,
+    "train_loss": 0.007936205714941025,
+    "train_runtime": 131.0047,
     "train_samples": 7500,
-    "train_samples_per_second": 1.012,
+    "train_samples_per_second": 0.977,
     "train_steps_per_second": 0.008
 }

train_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 0.008086015470325947,
-    "train_runtime": 126.4484,
+    "train_loss": 0.007936205714941025,
+    "train_runtime": 131.0047,
     "train_samples": 7500,
-    "train_samples_per_second": 1.012,
+    "train_samples_per_second": 0.977,
     "train_steps_per_second": 0.008
 }

trainer_state.json

@@ -10,15 +10,15 @@
   "log_history": [
     {
       "clip_ratio": 0.0,
-      "completion_length": 516.796875,
+      "completion_length": 521.4765625,
       "epoch": 1.0,
-      "grad_norm": 0.5813263058662415,
+      "grad_norm": 0.5955476760864258,
       "kl": 0.0,
       "learning_rate": 3e-06,
-      "loss": 0.0081,
-      "reward": 0.7098214626312256,
-      "reward_std": 0.3545062728226185,
-      "rewards/accuracy_reward": 0.7098214626312256,
+      "loss": 0.0079,
+      "reward": 0.7031250298023224,
+      "reward_std": 0.3489644229412079,
+      "rewards/accuracy_reward": 0.7031250298023224,
       "rewards/format_reward": 0.0,
       "step": 1
     },
@@ -26,9 +26,9 @@
       "epoch": 1.0,
       "step": 1,
       "total_flos": 0.0,
-      "train_loss": 0.008086015470325947,
-      "train_runtime": 126.4484,
-      "train_samples_per_second": 1.012,
+      "train_loss": 0.007936205714941025,
+      "train_runtime": 131.0047,
+      "train_samples_per_second": 0.977,
       "train_steps_per_second": 0.008
     }
   ],

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cf04e394615126d67569c09d6f52f0442c2fbadf27759f7ca1f142ab47ec0a69
+oid sha256:67ef1e0d5558a29538503adf7369f40c57aa17a1139078ac8cd2fd0a8f64b90a
 size 7992