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/shuheikurita/huggingface/runs/bbn8kalc) 
+[<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/shuheikurita/huggingface/runs/hxoo91ej) 
 
 
 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": -0.0055115707218647,
-    "train_runtime": 312.6496,
+    "train_loss": 4.3852842645719646e-05,
+    "train_runtime": 520.3847,
     "train_samples": 7500,
-    "train_samples_per_second": 1.199,
+    "train_samples_per_second": 1.441,
     "train_steps_per_second": 0.01
 }

model-00001-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:23c14f6bfb72c89d4f34b8d04ce14789497a4fd112ecc774bceae0ed22a83832
+oid sha256:c30a72af7c3cdbaedb6c9d031f0114c654c7df6b62f01468216decd2a4ead5d5
 size 4877660776

model-00002-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e2b9052683a6e1362240df4cd2ab173b87ddfd63e1ba804f56f325ce72f84180
+oid sha256:02cb1280725bb2bb71bcd5d014f1b948a4a1266c1023a853106059780f6bc768
 size 4932751008

model-00003-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:62b0d06d01a2609dc266ec63ff2a4fd112823fa1b141863baf401ab5ceac342a
+oid sha256:58a4caeae83e117b32ed5f6f3bf6f738190ebd45e4d7d7b7cc514aba774b79a4
 size 4330865200

model-00004-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:03c823986e7579a785a55e63983b70e1f55a163c921904d909a4b65c1150f4d5
+oid sha256:613cdbf58c9f65787a7ff4bb299a37aec726ad1ea250eee91b11c5c0bc291847
 size 1089994880

train_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": -0.0055115707218647,
-    "train_runtime": 312.6496,
+    "train_loss": 4.3852842645719646e-05,
+    "train_runtime": 520.3847,
     "train_samples": 7500,
-    "train_samples_per_second": 1.199,
+    "train_samples_per_second": 1.441,
     "train_steps_per_second": 0.01
 }

trainer_state.json

@@ -1,31 +1,38 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 1.0,
+  "epoch": 0.851063829787234,
   "eval_steps": 100,
-  "global_step": 3,
+  "global_step": 5,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "completion_length": 523.3298619588217,
-      "epoch": 1.0,
-      "kl": 0.0003174146016438802,
-      "reward": 0.7875744352738062,
-      "reward_std": 0.3135776606698831,
-      "rewards/accuracy_reward": 0.7864583656191826,
+      "completion_length": 517.0663162231446,
+      "epoch": 0.851063829787234,
+      "grad_norm": 2.043825149536133,
+      "kl": 0.001093578338623047,
+      "learning_rate": 0.0,
+      "loss": 0.0,
+      "reward": 0.794419676065445,
+      "reward_std": 0.3082874551415443,
+      "rewards/accuracy_reward": 0.7933036014437675,
       "rewards/format_reward": 0.0011160714784637094,
-      "step": 3,
+      "step": 5
+    },
+    {
+      "epoch": 0.851063829787234,
+      "step": 5,
       "total_flos": 0.0,
-      "train_loss": -0.0055115707218647,
-      "train_runtime": 312.6496,
-      "train_samples_per_second": 1.199,
+      "train_loss": 4.3852842645719646e-05,
+      "train_runtime": 520.3847,
+      "train_samples_per_second": 1.441,
       "train_steps_per_second": 0.01
     }
   ],
   "logging_steps": 5,
-  "max_steps": 3,
+  "max_steps": 5,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 1,
   "save_steps": 500,

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:95d2d4b4e7c94c87d21458c375e3df7027c724d1a42c486545b75e1d27684772
+oid sha256:ba36b46b141106ba02b4a8a5c1f80fc2e42f38c1a88c6733a611813a04183d37
 size 7544