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/dongwei_jiang/huggingface/runs/a455kyhz) 
+[<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/dongwei_jiang/huggingface/runs/r6vxw2n1) 
 
 
 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

@@ -5,9 +5,9 @@
     "eval_samples_per_second": 0.372,
     "eval_steps_per_second": 0.053,
     "total_flos": 0.0,
-    "train_loss": 0.0010332457457126966,
-    "train_runtime": 12336.334,
+    "train_loss": 0.001102253468012187,
+    "train_runtime": 12358.4402,
     "train_samples": 7500,
-    "train_samples_per_second": 0.608,
+    "train_samples_per_second": 0.607,
     "train_steps_per_second": 0.005
 }

model-00001-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:43f4a4415894eb44a772f2a952f60dfd74014ce17da021a268db1c61ff0e8431
+oid sha256:1d16cf0a5855474c715f5400c54f272528e5f7582f8490aabaa3b195585646d1
 size 4877660776

model-00002-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:850e15aa199993ff28b9a5ac3fd5895348dc4b6d174fe973c4d6beb106667e16
+oid sha256:77c341d2b8e43a12cdb2b96d79f7e71ae597157db282a048d831599bca5b0081
 size 4932751008

model-00003-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b91fadec7880886ede2a0d8efec62f95e3342b36161972a121bc0ca58563026b
+oid sha256:287ad58b8bafccdef834389401693a93f4c6e9b26009e2275ecb9dd777a4cd7c
 size 4330865200

model-00004-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:46189c667de83c221c22d722a17469650d924b0a82aaf2bba1387775d3ccd4f1
+oid sha256:f06ae450d81394bd9715da3799234c1b1d93d4d2600f13859bc40677b6007c73
 size 1089994880

train_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 0.0010332457457126966,
-    "train_runtime": 12336.334,
+    "train_loss": 0.001102253468012187,
+    "train_runtime": 12358.4402,
     "train_samples": 7500,
-    "train_samples_per_second": 0.608,
+    "train_samples_per_second": 0.607,
     "train_steps_per_second": 0.005
 }

trainer_state.json

@@ -9,96 +9,96 @@
   "is_world_process_zero": true,
   "log_history": [
     {
-      "completion_length": 707.4456775665283,
+      "completion_length": 708.9171075820923,
       "epoch": 0.14925373134328357,
-      "grad_norm": 0.06321819126605988,
-      "kl": 0.006711304187774658,
+      "grad_norm": 0.06562932580709457,
+      "kl": 0.005872213840484619,
       "learning_rate": 1.9876883405951378e-05,
-      "loss": 0.0003,
-      "reward": 0.5986607405939139,
-      "reward_std": 0.16735767447389663,
-      "rewards/accuracy_reward": 0.5986607405939139,
+      "loss": 0.0002,
+      "reward": 0.603348241513595,
+      "reward_std": 0.16946082445792854,
+      "rewards/accuracy_reward": 0.603348241513595,
       "rewards/format_reward": 0.0,
       "step": 10
     },
     {
-      "completion_length": 675.7138690948486,
+      "completion_length": 682.0505897521973,
       "epoch": 0.29850746268656714,
-      "grad_norm": 0.07178156822919846,
-      "kl": 0.025733184814453126,
+      "grad_norm": 0.05642814189195633,
+      "kl": 0.024619293212890626,
       "learning_rate": 1.777145961456971e-05,
       "loss": 0.001,
-      "reward": 0.6954241393133997,
-      "reward_std": 0.16923737577162684,
-      "rewards/accuracy_reward": 0.6954241393133997,
+      "reward": 0.7039062824100256,
+      "reward_std": 0.162486822437495,
+      "rewards/accuracy_reward": 0.7039062824100256,
       "rewards/format_reward": 0.0,
       "step": 20
     },
     {
-      "completion_length": 685.2908779144287,
+      "completion_length": 680.5775978088379,
       "epoch": 0.44776119402985076,
-      "grad_norm": 0.07352711260318756,
-      "kl": 0.02706298828125,
+      "grad_norm": 0.07271964848041534,
+      "kl": 0.029143524169921876,
       "learning_rate": 1.3583679495453e-05,
-      "loss": 0.0011,
-      "reward": 0.6889509245753288,
-      "reward_std": 0.15610567890107632,
-      "rewards/accuracy_reward": 0.6889509245753288,
+      "loss": 0.0012,
+      "reward": 0.6966518176719546,
+      "reward_std": 0.15723272264003754,
+      "rewards/accuracy_reward": 0.6966518176719546,
       "rewards/format_reward": 0.0,
       "step": 30
     },
     {
-      "completion_length": 691.1547193527222,
+      "completion_length": 678.4286026000976,
       "epoch": 0.5970149253731343,
-      "grad_norm": 0.059463296085596085,
-      "kl": 0.03000030517578125,
+      "grad_norm": 0.06629566103219986,
+      "kl": 0.029315948486328125,
       "learning_rate": 8.43565534959769e-06,
       "loss": 0.0012,
-      "reward": 0.6849330635741353,
-      "reward_std": 0.16573377684690058,
-      "rewards/accuracy_reward": 0.6849330635741353,
+      "reward": 0.6925223540514708,
+      "reward_std": 0.1698813715018332,
+      "rewards/accuracy_reward": 0.6925223540514708,
       "rewards/format_reward": 0.0,
       "step": 40
     },
     {
-      "completion_length": 690.842552947998,
+      "completion_length": 678.2587354660034,
       "epoch": 0.746268656716418,
-      "grad_norm": 0.06777255237102509,
-      "kl": 0.03056793212890625,
+      "grad_norm": 0.07359491288661957,
+      "kl": 0.03377456665039062,
       "learning_rate": 3.7067960895016277e-06,
-      "loss": 0.0012,
-      "reward": 0.7001116393133998,
-      "reward_std": 0.16762742078863085,
-      "rewards/accuracy_reward": 0.7001116393133998,
+      "loss": 0.0014,
+      "reward": 0.6977678883820773,
+      "reward_std": 0.1756029822397977,
+      "rewards/accuracy_reward": 0.6977678883820773,
       "rewards/format_reward": 0.0,
       "step": 50
     },
     {
-      "completion_length": 668.4015930175781,
+      "completion_length": 676.1143146514893,
       "epoch": 0.8955223880597015,
-      "grad_norm": 0.07001896947622299,
-      "kl": 0.031087493896484374,
+      "grad_norm": 0.07282646745443344,
+      "kl": 0.03675537109375,
       "learning_rate": 6.641957350279838e-07,
-      "loss": 0.0012,
-      "reward": 0.7029018193483353,
-      "reward_std": 0.16094307857565582,
-      "rewards/accuracy_reward": 0.7029018193483353,
+      "loss": 0.0015,
+      "reward": 0.6957589590921998,
+      "reward_std": 0.167550537455827,
+      "rewards/accuracy_reward": 0.6957589590921998,
       "rewards/format_reward": 0.0,
       "step": 60
     },
     {
-      "completion_length": 663.3861323765346,
+      "completion_length": 673.9289711543491,
       "epoch": 1.0,
-      "kl": 0.03179168701171875,
-      "reward": 0.7240115058209214,
-      "reward_std": 0.1582813827054841,
-      "rewards/accuracy_reward": 0.7240115058209214,
+      "kl": 0.03508213588169643,
+      "reward": 0.7025404186653239,
+      "reward_std": 0.1512090131374342,
+      "rewards/accuracy_reward": 0.7025404186653239,
       "rewards/format_reward": 0.0,
       "step": 67,
       "total_flos": 0.0,
-      "train_loss": 0.0010332457457126966,
-      "train_runtime": 12336.334,
-      "train_samples_per_second": 0.608,
+      "train_loss": 0.001102253468012187,
+      "train_runtime": 12358.4402,
+      "train_samples_per_second": 0.607,
       "train_steps_per_second": 0.005
     }
   ],

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:13cbabc08ca0e5abe40c178d16e74839353ac52a59e2740d6b323fe07f596c30
+oid sha256:010ba5f0a144da22011142faca812261dbf0129a623c2c90d9c7561bd143a44b
 size 7032