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/o2gdvnvm) 
+[<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/tydougqd) 
 
 
 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": 0.001063353899838542,
-    "train_runtime": 12365.8628,
-    "train_samples": 7500,
-    "train_samples_per_second": 0.607,
-    "train_steps_per_second": 0.005
+    "train_loss": 0.0,
+    "train_runtime": 15.0247,
+    "train_samples": 7,
+    "train_samples_per_second": 0.466,
+    "train_steps_per_second": 0.067
 }

model-00001-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:95e4d9b363b9d9c1be358fed849e21f093cf04c05b95dbbbb7140f0c8c4c47d0
+oid sha256:922999e9590ae1378683dd123275b3600e7febc0371df308a7cdb4e43d8fd726
 size 4877660776

model-00002-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dccb6d918b62ea985c982407610f5a865ee213ae4e6fae7ba13d89f02fda2a1e
+oid sha256:f50d30c51a8a098e235f4df8bcf46f526b005dc6038f3652ecd6387559400e54
 size 4932751008

model-00003-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:63871bd8992cfe32ebdbf2482fecb9884fa410d48b0066429ae59530226eec52
+oid sha256:10542b140f31ad53d834cbf91aa02e6d27af3e8134bab2e65d58a92d39ba93c6
 size 4330865200

model-00004-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e93d867e4461d8ca90a20d548a4d84140d587c83bf681054a4f07eab9f80af98
+oid sha256:7946740ab848b4a02904f05bff81211bf9600fc20bfad39bbcc8d1703a40ce1c
 size 1089994880

train_results.json

@@ -1,8 +1,8 @@
 {
     "total_flos": 0.0,
-    "train_loss": 0.001063353899838542,
-    "train_runtime": 12365.8628,
-    "train_samples": 7500,
-    "train_samples_per_second": 0.607,
-    "train_steps_per_second": 0.005
+    "train_loss": 0.0,
+    "train_runtime": 15.0247,
+    "train_samples": 7,
+    "train_samples_per_second": 0.466,
+    "train_steps_per_second": 0.067
 }

trainer_state.json

@@ -3,107 +3,29 @@
   "best_model_checkpoint": null,
   "epoch": 1.0,
   "eval_steps": 100,
-  "global_step": 67,
+  "global_step": 1,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "completion_length": 708.4377550125122,
-      "epoch": 0.14925373134328357,
-      "grad_norm": 0.059954091906547546,
-      "kl": 0.006372016668319702,
-      "learning_rate": 1.9876883405951378e-05,
-      "loss": 0.0003,
-      "reward": 0.606026814377401,
-      "reward_std": 0.16658300049602986,
-      "rewards/accuracy_reward": 0.606026814377401,
-      "rewards/format_reward": 0.0,
-      "step": 10
-    },
-    {
-      "completion_length": 689.9649854660034,
-      "epoch": 0.29850746268656714,
-      "grad_norm": 0.40041017532348633,
-      "kl": 0.02557830810546875,
-      "learning_rate": 1.777145961456971e-05,
-      "loss": 0.001,
-      "reward": 0.6754464585334062,
-      "reward_std": 0.16620420585386456,
-      "rewards/accuracy_reward": 0.6754464585334062,
-      "rewards/format_reward": 0.0,
-      "step": 20
-    },
-    {
-      "completion_length": 685.777486038208,
-      "epoch": 0.44776119402985076,
-      "grad_norm": 0.06629940867424011,
-      "kl": 0.027927398681640625,
-      "learning_rate": 1.3583679495453e-05,
-      "loss": 0.0011,
-      "reward": 0.681250031106174,
-      "reward_std": 0.16514827804639937,
-      "rewards/accuracy_reward": 0.681250031106174,
-      "rewards/format_reward": 0.0,
-      "step": 30
-    },
-    {
-      "completion_length": 694.0344058990479,
-      "epoch": 0.5970149253731343,
-      "grad_norm": 0.05779099091887474,
-      "kl": 0.0300262451171875,
-      "learning_rate": 8.43565534959769e-06,
-      "loss": 0.0012,
-      "reward": 0.6813616389408708,
-      "reward_std": 0.15597098912112414,
-      "rewards/accuracy_reward": 0.6813616389408708,
-      "rewards/format_reward": 0.0,
-      "step": 40
-    },
-    {
-      "completion_length": 685.4475748062134,
-      "epoch": 0.746268656716418,
-      "grad_norm": 0.07657386362552643,
-      "kl": 0.030538177490234374,
-      "learning_rate": 3.7067960895016277e-06,
-      "loss": 0.0012,
-      "reward": 0.6813616370782256,
-      "reward_std": 0.17094493978656827,
-      "rewards/accuracy_reward": 0.6813616370782256,
-      "rewards/format_reward": 0.0,
-      "step": 50
-    },
-    {
-      "completion_length": 676.2002536773682,
-      "epoch": 0.8955223880597015,
-      "grad_norm": 0.06962116062641144,
-      "kl": 0.032154083251953125,
-      "learning_rate": 6.641957350279838e-07,
-      "loss": 0.0013,
-      "reward": 0.6899553881958127,
-      "reward_std": 0.1660784802865237,
-      "rewards/accuracy_reward": 0.6899553881958127,
-      "rewards/format_reward": 0.0,
-      "step": 60
-    },
-    {
-      "completion_length": 667.0321821485247,
+      "completion_length": 688.6964721679688,
       "epoch": 1.0,
-      "kl": 0.03317805698939732,
-      "reward": 0.6945684791675636,
-      "reward_std": 0.1684813142221953,
-      "rewards/accuracy_reward": 0.6945684791675636,
+      "kl": 0.0,
+      "reward": 0.785714328289032,
+      "reward_std": 0.12444323301315308,
+      "rewards/accuracy_reward": 0.785714328289032,
       "rewards/format_reward": 0.0,
-      "step": 67,
+      "step": 1,
       "total_flos": 0.0,
-      "train_loss": 0.001063353899838542,
-      "train_runtime": 12365.8628,
-      "train_samples_per_second": 0.607,
-      "train_steps_per_second": 0.005
+      "train_loss": 0.0,
+      "train_runtime": 15.0247,
+      "train_samples_per_second": 0.466,
+      "train_steps_per_second": 0.067
     }
   ],
   "logging_steps": 10,
-  "max_steps": 67,
+  "max_steps": 1,
   "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 sha256:0fbcd43da27115f3fe989a9567434e80e77c3b53b4650d0a65b152eff3e13c16
+oid sha256:63e920eb20fd9a47847c1936f4cdfeb92754f51d824c1538067eb254c1bb131b
 size 7032