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/ryanzhang123-the-university-of-auckland/RLCR/runs/7t6rinro) 
+[<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/ryanzhang123-the-university-of-auckland/RLCR/runs/no0v40a9) 
 
 
 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).

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d56de5b2370feb4dbb509c230ac580149cec26e9aad73a718a4595f40806a488
+oid sha256:d66cb65a1041f8fda1a53da903ef1c492178657071c945a0890410b1df4818ca
 size 3554214752

trainer_state.json

@@ -1055,24 +1055,24 @@
       "clip_ratio/low_mean": 0.0,
       "clip_ratio/low_min": 0.0,
       "clip_ratio/region_mean": 0.0,
-      "completions/clipped_ratio": 0.0125,
-      "completions/max_length": 1480.2,
-      "completions/max_terminated_length": 1096.6,
-      "completions/mean_length": 138.0765625,
-      "completions/mean_terminated_length": 120.40022125244141,
-      "completions/min_length": 21.4,
-      "completions/min_terminated_length": 21.4,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 1218.0,
+      "completions/max_terminated_length": 1218.0,
+      "completions/mean_length": 109.996875,
+      "completions/mean_terminated_length": 109.996875,
+      "completions/min_length": 17.2,
+      "completions/min_terminated_length": 17.2,
       "epoch": 2.399355877616747,
-      "grad_norm": 0.002646031556650996,
+      "grad_norm": 0.0019124990794807673,
       "learning_rate": 1e-06,
-      "loss": 0.0047,
-      "num_tokens": 21079397.0,
-      "reward": 0.4921875,
-      "reward_std": 0.022097086161375047,
+      "loss": 0.0006,
+      "num_tokens": 21060786.0,
+      "reward": 0.49921875,
+      "reward_std": 0.0022097086533904076,
       "rewards/accuracy_reward": 0.0,
       "rewards/brier_reward": 0.0,
       "rewards/confidence_one_or_zero": 0.0,
-      "rewards/format_reward": 0.984375,
+      "rewards/format_reward": 0.9984375,
       "rewards/mean_confidence_reward": 0.0,
       "step": 185
     },
@@ -1083,23 +1083,23 @@
       "clip_ratio/low_min": 0.0,
       "clip_ratio/region_mean": 0.0,
       "completions/clipped_ratio": 0.0015625,
-      "completions/max_length": 1132.6,
-      "completions/max_terminated_length": 983.0,
-      "completions/mean_length": 115.8234375,
-      "completions/mean_terminated_length": 113.60880889892579,
-      "completions/min_length": 17.4,
-      "completions/min_terminated_length": 17.4,
+      "completions/max_length": 949.8,
+      "completions/max_terminated_length": 824.4,
+      "completions/mean_length": 105.5375,
+      "completions/mean_terminated_length": 103.31301727294922,
+      "completions/min_length": 17.0,
+      "completions/min_terminated_length": 17.0,
       "epoch": 2.463768115942029,
-      "grad_norm": 0.005643462762236595,
+      "grad_norm": 0.0021424347069114447,
       "learning_rate": 1e-06,
-      "loss": 0.0008,
-      "num_tokens": 21628532.0,
-      "reward": 0.49765625,
-      "reward_std": 0.0066291259601712225,
+      "loss": 0.0002,
+      "num_tokens": 21602698.0,
+      "reward": 0.4984375,
+      "reward_std": 0.004419417306780815,
       "rewards/accuracy_reward": 0.0,
       "rewards/brier_reward": 0.0,
       "rewards/confidence_one_or_zero": 0.0,
-      "rewards/format_reward": 0.9953125,
+      "rewards/format_reward": 0.996875,
       "rewards/mean_confidence_reward": 0.0,
       "step": 190
     },
@@ -1109,24 +1109,24 @@
       "clip_ratio/low_mean": 0.0,
       "clip_ratio/low_min": 0.0,
       "clip_ratio/region_mean": 0.0,
-      "completions/clipped_ratio": 0.009375,
-      "completions/max_length": 1536.0,
-      "completions/max_terminated_length": 988.6,
-      "completions/mean_length": 128.975,
-      "completions/mean_terminated_length": 115.63809661865234,
-      "completions/min_length": 18.4,
-      "completions/min_terminated_length": 18.4,
+      "completions/clipped_ratio": 0.00625,
+      "completions/max_length": 1301.0,
+      "completions/max_terminated_length": 1011.4,
+      "completions/mean_length": 117.728125,
+      "completions/mean_terminated_length": 108.78486022949218,
+      "completions/min_length": 16.8,
+      "completions/min_terminated_length": 16.8,
       "epoch": 2.528180354267311,
-      "grad_norm": 0.0020768363028764725,
+      "grad_norm": 0.0,
       "learning_rate": 1e-06,
-      "loss": 0.0033,
-      "num_tokens": 22231892.0,
-      "reward": 0.49453125,
-      "reward_std": 0.015467960573732852,
+      "loss": 0.0028,
+      "num_tokens": 22198220.0,
+      "reward": 0.4953125,
+      "reward_std": 0.01020579319447279,
       "rewards/accuracy_reward": 0.0,
       "rewards/brier_reward": 0.0,
       "rewards/confidence_one_or_zero": 0.0,
-      "rewards/format_reward": 0.9890625,
+      "rewards/format_reward": 0.990625,
       "rewards/mean_confidence_reward": 0.0,
       "step": 195
     },
@@ -1136,48 +1136,48 @@
       "clip_ratio/low_mean": 0.0,
       "clip_ratio/low_min": 0.0,
       "clip_ratio/region_mean": 0.0,
-      "completions/clipped_ratio": 0.0078125,
-      "completions/max_length": 1536.0,
-      "completions/max_terminated_length": 1052.8,
-      "completions/mean_length": 134.6171875,
-      "completions/mean_terminated_length": 123.58267517089844,
-      "completions/min_length": 23.4,
-      "completions/min_terminated_length": 23.4,
+      "completions/clipped_ratio": 0.003125,
+      "completions/max_length": 1346.4,
+      "completions/max_terminated_length": 1101.8,
+      "completions/mean_length": 119.5875,
+      "completions/mean_terminated_length": 115.14135131835937,
+      "completions/min_length": 16.6,
+      "completions/min_terminated_length": 16.6,
       "epoch": 2.5925925925925926,
-      "grad_norm": 0.0007894107839092612,
+      "grad_norm": 0.0009198391926474869,
       "learning_rate": 1e-06,
-      "loss": 0.0029,
-      "num_tokens": 22775407.0,
-      "reward": 0.49609375,
-      "reward_std": 0.011048543266952038,
+      "loss": 0.0009,
+      "num_tokens": 22731476.0,
+      "reward": 0.49765625,
+      "reward_std": 0.0066291259601712225,
       "rewards/accuracy_reward": 0.0,
       "rewards/brier_reward": 0.0,
       "rewards/confidence_one_or_zero": 0.0,
-      "rewards/format_reward": 0.9921875,
+      "rewards/format_reward": 0.9953125,
       "rewards/mean_confidence_reward": 0.0,
       "step": 200
     },
     {
       "epoch": 2.5925925925925926,
-      "eval_completions/clipped_ratio": 0.005434782608695652,
-      "eval_completions/max_length": 440.4855072463768,
-      "eval_completions/max_terminated_length": 395.9927536231884,
-      "eval_completions/mean_length": 130.15851449275362,
-      "eval_completions/mean_terminated_length": 122.56198280445044,
-      "eval_completions/min_length": 38.231884057971016,
-      "eval_completions/min_terminated_length": 38.231884057971016,
+      "eval_completions/clipped_ratio": 0.006340579710144928,
+      "eval_completions/max_length": 389.9130434782609,
+      "eval_completions/max_terminated_length": 328.77536231884056,
+      "eval_completions/mean_length": 117.83695652173913,
+      "eval_completions/mean_terminated_length": 108.77691534291144,
+      "eval_completions/min_length": 33.07971014492754,
+      "eval_completions/min_terminated_length": 33.07971014492754,
       "eval_loss": 0.0,
-      "eval_num_tokens": 22775407.0,
-      "eval_reward": 0.49682971014492755,
-      "eval_reward_std": 0.008966933665932089,
+      "eval_num_tokens": 22731476.0,
+      "eval_reward": 0.4963768115942029,
+      "eval_reward_std": 0.010247924189636673,
       "eval_rewards/accuracy_reward": 0.0,
       "eval_rewards/brier_reward": 0.0,
       "eval_rewards/confidence_one_or_zero": 0.0,
-      "eval_rewards/format_reward": 0.9936594202898551,
+      "eval_rewards/format_reward": 0.9927536231884058,
       "eval_rewards/mean_confidence_reward": 0.0,
-      "eval_runtime": 370.7818,
-      "eval_samples_per_second": 0.372,
-      "eval_steps_per_second": 0.049,
+      "eval_runtime": 342.0823,
+      "eval_samples_per_second": 0.403,
+      "eval_steps_per_second": 0.053,
       "step": 200
     },
     {
@@ -1186,24 +1186,24 @@
       "clip_ratio/low_mean": 0.0,
       "clip_ratio/low_min": 0.0,
       "clip_ratio/region_mean": 0.0,
-      "completions/clipped_ratio": 0.00625,
-      "completions/max_length": 1263.6,
-      "completions/max_terminated_length": 911.0,
-      "completions/mean_length": 116.421875,
-      "completions/mean_terminated_length": 107.53674011230468,
-      "completions/min_length": 16.8,
-      "completions/min_terminated_length": 16.8,
+      "completions/clipped_ratio": 0.0015625,
+      "completions/max_length": 1091.6,
+      "completions/max_terminated_length": 1001.0,
+      "completions/mean_length": 112.4625,
+      "completions/mean_terminated_length": 110.25907897949219,
+      "completions/min_length": 17.0,
+      "completions/min_terminated_length": 17.0,
       "epoch": 2.6570048309178746,
-      "grad_norm": 0.00095662398962304,
+      "grad_norm": 0.0014365437673404813,
       "learning_rate": 1e-06,
-      "loss": 0.0029,
-      "num_tokens": 23355853.0,
-      "reward": 0.49609375,
-      "reward_std": 0.011048543266952038,
+      "loss": 0.0008,
+      "num_tokens": 23308748.0,
+      "reward": 0.49921875,
+      "reward_std": 0.0022097086533904076,
       "rewards/accuracy_reward": 0.0,
       "rewards/brier_reward": 0.0,
       "rewards/confidence_one_or_zero": 0.0,
-      "rewards/format_reward": 0.9921875,
+      "rewards/format_reward": 0.9984375,
       "rewards/mean_confidence_reward": 0.0,
       "step": 205
     },
@@ -1213,24 +1213,24 @@
       "clip_ratio/low_mean": 0.0,
       "clip_ratio/low_min": 0.0,
       "clip_ratio/region_mean": 0.0,
-      "completions/clipped_ratio": 0.0078125,
-      "completions/max_length": 1519.0,
-      "completions/max_terminated_length": 1180.4,
-      "completions/mean_length": 142.6796875,
-      "completions/mean_terminated_length": 131.7208999633789,
-      "completions/min_length": 17.4,
-      "completions/min_terminated_length": 17.4,
+      "completions/clipped_ratio": 0.00625,
+      "completions/max_length": 1477.8,
+      "completions/max_terminated_length": 1071.2,
+      "completions/mean_length": 117.846875,
+      "completions/mean_terminated_length": 108.93180236816406,
+      "completions/min_length": 16.0,
+      "completions/min_terminated_length": 16.0,
       "epoch": 2.7214170692431563,
-      "grad_norm": 0.00048346660332754254,
+      "grad_norm": 0.0007683393196202815,
       "learning_rate": 1e-06,
-      "loss": 0.0025,
-      "num_tokens": 23956464.0,
-      "reward": 0.4953125,
-      "reward_std": 0.013258251920342445,
+      "loss": 0.004,
+      "num_tokens": 23892826.0,
+      "reward": 0.49375,
+      "reward_std": 0.017677669040858747,
       "rewards/accuracy_reward": 0.0,
       "rewards/brier_reward": 0.0,
       "rewards/confidence_one_or_zero": 0.0,
-      "rewards/format_reward": 0.990625,
+      "rewards/format_reward": 0.9875,
       "rewards/mean_confidence_reward": 0.0,
       "step": 210
     },
@@ -1240,24 +1240,24 @@
       "clip_ratio/low_mean": 0.0,
       "clip_ratio/low_min": 0.0,
       "clip_ratio/region_mean": 0.0,
-      "completions/clipped_ratio": 0.00625,
-      "completions/max_length": 1356.6,
-      "completions/max_terminated_length": 878.0,
-      "completions/mean_length": 116.9703125,
-      "completions/mean_terminated_length": 108.06337432861328,
-      "completions/min_length": 21.4,
-      "completions/min_terminated_length": 21.4,
+      "completions/clipped_ratio": 0.0046875,
+      "completions/max_length": 1254.2,
+      "completions/max_terminated_length": 974.2,
+      "completions/mean_length": 111.115625,
+      "completions/mean_terminated_length": 104.36334838867188,
+      "completions/min_length": 16.8,
+      "completions/min_terminated_length": 16.8,
       "epoch": 2.785829307568438,
-      "grad_norm": 0.0006744141574017704,
+      "grad_norm": 0.0009448982309550047,
       "learning_rate": 1e-06,
-      "loss": 0.0026,
-      "num_tokens": 24487189.0,
-      "reward": 0.496875,
-      "reward_std": 0.00883883461356163,
+      "loss": 0.0029,
+      "num_tokens": 24419164.0,
+      "reward": 0.49609375,
+      "reward_std": 0.011048543266952038,
       "rewards/accuracy_reward": 0.0,
       "rewards/brier_reward": 0.0,
       "rewards/confidence_one_or_zero": 0.0,
-      "rewards/format_reward": 0.99375,
+      "rewards/format_reward": 0.9921875,
       "rewards/mean_confidence_reward": 0.0,
       "step": 215
     },
@@ -1267,24 +1267,24 @@
       "clip_ratio/low_mean": 0.0,
       "clip_ratio/low_min": 0.0,
       "clip_ratio/region_mean": 0.0,
-      "completions/clipped_ratio": 0.0046875,
-      "completions/max_length": 1141.2,
-      "completions/max_terminated_length": 915.0,
-      "completions/mean_length": 123.528125,
-      "completions/mean_terminated_length": 116.93916931152344,
-      "completions/min_length": 22.4,
-      "completions/min_terminated_length": 22.4,
+      "completions/clipped_ratio": 0.00625,
+      "completions/max_length": 1252.0,
+      "completions/max_terminated_length": 985.6,
+      "completions/mean_length": 103.759375,
+      "completions/mean_terminated_length": 94.76344604492188,
+      "completions/min_length": 16.4,
+      "completions/min_terminated_length": 16.4,
       "epoch": 2.85024154589372,
-      "grad_norm": 0.0016728757182136178,
+      "grad_norm": 0.0,
       "learning_rate": 1e-06,
-      "loss": 0.002,
-      "num_tokens": 25022543.0,
-      "reward": 0.4953125,
-      "reward_std": 0.01325825173407793,
+      "loss": 0.0022,
+      "num_tokens": 24941226.0,
+      "reward": 0.496875,
+      "reward_std": 0.00883883461356163,
       "rewards/accuracy_reward": 0.0,
       "rewards/brier_reward": 0.0,
       "rewards/confidence_one_or_zero": 0.0,
-      "rewards/format_reward": 0.990625,
+      "rewards/format_reward": 0.99375,
       "rewards/mean_confidence_reward": 0.0,
       "step": 220
     },
@@ -1294,24 +1294,24 @@
       "clip_ratio/low_mean": 0.0,
       "clip_ratio/low_min": 0.0,
       "clip_ratio/region_mean": 0.0,
-      "completions/clipped_ratio": 0.009375,
-      "completions/max_length": 1212.0,
-      "completions/max_terminated_length": 884.8,
-      "completions/mean_length": 133.2875,
-      "completions/mean_terminated_length": 120.0623291015625,
-      "completions/min_length": 19.8,
-      "completions/min_terminated_length": 19.8,
+      "completions/clipped_ratio": 0.0046875,
+      "completions/max_length": 1380.2,
+      "completions/max_terminated_length": 939.6,
+      "completions/mean_length": 97.8796875,
+      "completions/mean_terminated_length": 91.06793060302735,
+      "completions/min_length": 16.4,
+      "completions/min_terminated_length": 16.4,
       "epoch": 2.9146537842190017,
       "grad_norm": 0.0,
       "learning_rate": 1e-06,
-      "loss": 0.0028,
-      "num_tokens": 25553799.0,
-      "reward": 0.4953125,
-      "reward_std": 0.009863808192312718,
+      "loss": 0.0017,
+      "num_tokens": 25449181.0,
+      "reward": 0.49609375,
+      "reward_std": 0.011048543080687524,
       "rewards/accuracy_reward": 0.0,
       "rewards/brier_reward": 0.0,
       "rewards/confidence_one_or_zero": 0.0,
-      "rewards/format_reward": 0.990625,
+      "rewards/format_reward": 0.9921875,
       "rewards/mean_confidence_reward": 0.0,
       "step": 225
     },
@@ -1321,24 +1321,24 @@
       "clip_ratio/low_mean": 0.0,
       "clip_ratio/low_min": 0.0,
       "clip_ratio/region_mean": 0.0,
-      "completions/clipped_ratio": 0.0046875,
-      "completions/max_length": 1162.8,
-      "completions/max_terminated_length": 894.4,
-      "completions/mean_length": 114.178125,
-      "completions/mean_terminated_length": 107.45401611328126,
-      "completions/min_length": 17.8,
-      "completions/min_terminated_length": 17.8,
+      "completions/clipped_ratio": 0.0015625,
+      "completions/max_length": 1033.6,
+      "completions/max_terminated_length": 831.8,
+      "completions/mean_length": 78.1453125,
+      "completions/mean_terminated_length": 75.84559478759766,
+      "completions/min_length": 16.6,
+      "completions/min_terminated_length": 16.6,
       "epoch": 2.9790660225442833,
-      "grad_norm": 0.0010454885195940733,
+      "grad_norm": 0.0008915381622500718,
       "learning_rate": 1e-06,
-      "loss": 0.003,
-      "num_tokens": 26101145.0,
-      "reward": 0.49609375,
-      "reward_std": 0.011048543266952038,
+      "loss": 0.0005,
+      "num_tokens": 25972826.0,
+      "reward": 0.4984375,
+      "reward_std": 0.004419417306780815,
       "rewards/accuracy_reward": 0.0,
       "rewards/brier_reward": 0.0,
       "rewards/confidence_one_or_zero": 0.0,
-      "rewards/format_reward": 0.9921875,
+      "rewards/format_reward": 0.996875,
       "rewards/mean_confidence_reward": 0.0,
       "step": 230
     },
@@ -1346,15 +1346,15 @@
       "epoch": 2.9790660225442833,
       "step": 230,
       "total_flos": 0.0,
-      "train_loss": 0.0005941401033297829,
-      "train_runtime": 2082.8626,
-      "train_samples_per_second": 1.789,
-      "train_steps_per_second": 0.111
+      "train_loss": 0.00036267064835714255,
+      "train_runtime": 2152.1672,
+      "train_samples_per_second": 1.731,
+      "train_steps_per_second": 0.107
     }
   ],
   "logging_steps": 5,
   "max_steps": 231,
-  "num_input_tokens_seen": 26101145,
+  "num_input_tokens_seen": 25972826,
   "num_train_epochs": 3,
   "save_steps": 60,
   "stateful_callbacks": {

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e4552b4f72af82a334750ed040fd1a3e4be960cc8d5a42cdf1d5cfa672c2bc4c
+oid sha256:cf06cc29d6059820188588aba3f689e5b3c53c90771469a03c4e2de494c9e2a4
 size 7992