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/vgprtp13) 
+[<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/tzt1v6gw) 
 
 
 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:fd417f8052f768e2bb60d6f379953d6c0e49f32e7a32a179c7feeb00525c423a
+oid sha256:8bfde5566b21a196d8ea312ebd610a085b3b5113e4b769acf1ffdac157255d9e
 size 3554214752

trainer_state.json

@@ -14,25 +14,25 @@
       "clip_ratio/low_mean": 0.0,
       "clip_ratio/low_min": 0.0,
       "clip_ratio/region_mean": 0.0,
-      "completions/clipped_ratio": 0.016796875,
+      "completions/clipped_ratio": 0.045703125,
       "completions/max_length": 1536.0,
-      "completions/max_terminated_length": 1448.0,
-      "completions/mean_length": 268.53125,
-      "completions/mean_terminated_length": 246.8666229248047,
-      "completions/min_length": 19.0,
-      "completions/min_terminated_length": 19.0,
+      "completions/max_terminated_length": 1497.8,
+      "completions/mean_length": 521.414453125,
+      "completions/mean_terminated_length": 472.765625,
+      "completions/min_length": 74.0,
+      "completions/min_terminated_length": 74.0,
       "epoch": 0.2576489533011272,
-      "grad_norm": 0.24433724582195282,
+      "grad_norm": 0.1361742913722992,
       "learning_rate": 1e-06,
-      "loss": 0.054,
-      "num_tokens": 2605240.0,
-      "reward": 0.4578125,
-      "reward_std": 0.25091842710971834,
-      "rewards/accuracy_reward": 0.105859375,
-      "rewards/brier_reward": 0.0,
-      "rewards/confidence_one_or_zero": 0.0,
-      "rewards/format_reward": 0.809765625,
-      "rewards/mean_confidence_reward": 0.0,
+      "loss": 0.0562,
+      "num_tokens": 4038541.0,
+      "reward": 0.4758795499801636,
+      "reward_std": 0.4125692009925842,
+      "rewards/accuracy_reward": 0.064453125,
+      "rewards/brier_reward": 0.2697267413139343,
+      "rewards/confidence_one_or_zero": 0.06171875,
+      "rewards/format_reward": 0.617578125,
+      "rewards/mean_confidence_reward": 0.500839501619339,
       "step": 5
     },
     {
@@ -41,25 +41,25 @@
       "clip_ratio/low_mean": 0.0,
       "clip_ratio/low_min": 0.0,
       "clip_ratio/region_mean": 0.0,
-      "completions/clipped_ratio": 0.01796875,
+      "completions/clipped_ratio": 0.04609375,
       "completions/max_length": 1536.0,
-      "completions/max_terminated_length": 1433.0,
-      "completions/mean_length": 250.86875,
-      "completions/mean_terminated_length": 227.32728271484376,
-      "completions/min_length": 17.6,
-      "completions/min_terminated_length": 17.6,
+      "completions/max_terminated_length": 1505.2,
+      "completions/mean_length": 523.770703125,
+      "completions/mean_terminated_length": 474.846337890625,
+      "completions/min_length": 71.0,
+      "completions/min_terminated_length": 71.0,
       "epoch": 0.5152979066022544,
-      "grad_norm": 0.003483051899820566,
+      "grad_norm": 0.0030647902749478817,
       "learning_rate": 1e-06,
-      "loss": 0.0464,
-      "num_tokens": 5188600.0,
-      "reward": 0.47890625,
-      "reward_std": 0.22365978062152864,
-      "rewards/accuracy_reward": 0.104296875,
-      "rewards/brier_reward": 0.0,
-      "rewards/confidence_one_or_zero": 0.0,
-      "rewards/format_reward": 0.853515625,
-      "rewards/mean_confidence_reward": 0.0,
+      "loss": 0.0583,
+      "num_tokens": 8106450.0,
+      "reward": 0.5272839426994324,
+      "reward_std": 0.40863736867904665,
+      "rewards/accuracy_reward": 0.07109375,
+      "rewards/brier_reward": 0.3018323004245758,
+      "rewards/confidence_one_or_zero": 0.078125,
+      "rewards/format_reward": 0.681640625,
+      "rewards/mean_confidence_reward": 0.5452994704246521,
       "step": 10
     },
     {
@@ -68,25 +68,25 @@
       "clip_ratio/low_mean": 0.0,
       "clip_ratio/low_min": 0.0,
       "clip_ratio/region_mean": 0.0,
-      "completions/clipped_ratio": 0.01796875,
+      "completions/clipped_ratio": 0.039453125,
       "completions/max_length": 1536.0,
-      "completions/max_terminated_length": 1412.0,
-      "completions/mean_length": 231.741015625,
-      "completions/mean_terminated_length": 207.91524353027344,
-      "completions/min_length": 19.2,
-      "completions/min_terminated_length": 19.2,
+      "completions/max_terminated_length": 1519.0,
+      "completions/mean_length": 499.593359375,
+      "completions/mean_terminated_length": 457.06886596679686,
+      "completions/min_length": 67.2,
+      "completions/min_terminated_length": 67.2,
       "epoch": 0.7729468599033816,
-      "grad_norm": 0.002466069534420967,
+      "grad_norm": 0.002637381898239255,
       "learning_rate": 1e-06,
-      "loss": 0.0538,
-      "num_tokens": 7692025.0,
-      "reward": 0.521875,
-      "reward_std": 0.21285174489021302,
-      "rewards/accuracy_reward": 0.134375,
-      "rewards/brier_reward": 0.0,
-      "rewards/confidence_one_or_zero": 0.0,
-      "rewards/format_reward": 0.909375,
-      "rewards/mean_confidence_reward": 0.0,
+      "loss": 0.047,
+      "num_tokens": 12081497.0,
+      "reward": 0.6300971508026123,
+      "reward_std": 0.3623956501483917,
+      "rewards/accuracy_reward": 0.086328125,
+      "rewards/brier_reward": 0.35863040685653685,
+      "rewards/confidence_one_or_zero": 0.071484375,
+      "rewards/format_reward": 0.815234375,
+      "rewards/mean_confidence_reward": 0.6512776494026185,
       "step": 15
     },
     {
@@ -95,25 +95,25 @@
       "clip_ratio/low_mean": 0.0,
       "clip_ratio/low_min": 0.0,
       "clip_ratio/region_mean": 0.0,
-      "completions/clipped_ratio": 0.01015625,
-      "completions/max_length": 1534.4,
-      "completions/max_terminated_length": 1478.6,
-      "completions/mean_length": 208.05107421875,
-      "completions/mean_terminated_length": 194.41161499023437,
-      "completions/min_length": 23.4,
-      "completions/min_terminated_length": 23.4,
+      "completions/clipped_ratio": 0.0331206995412844,
+      "completions/max_length": 1536.0,
+      "completions/max_terminated_length": 1507.6,
+      "completions/mean_length": 500.6675964355469,
+      "completions/mean_terminated_length": 465.1861938476562,
+      "completions/min_length": 76.0,
+      "completions/min_terminated_length": 76.0,
       "epoch": 1.0515297906602254,
-      "grad_norm": 0.0027691556606441736,
+      "grad_norm": 0.001991912489756942,
       "learning_rate": 1e-06,
-      "loss": 0.049,
-      "num_tokens": 10130483.0,
-      "reward": 0.5451171875,
-      "reward_std": 0.19625933170318605,
-      "rewards/accuracy_reward": 0.146484375,
-      "rewards/brier_reward": 0.0,
-      "rewards/confidence_one_or_zero": 0.0,
-      "rewards/format_reward": 0.94375,
-      "rewards/mean_confidence_reward": 0.0,
+      "loss": 0.0578,
+      "num_tokens": 16048198.0,
+      "reward": 0.7004695773124695,
+      "reward_std": 0.32459375262260437,
+      "rewards/accuracy_reward": 0.096875,
+      "rewards/brier_reward": 0.40406264662742614,
+      "rewards/confidence_one_or_zero": 0.078125,
+      "rewards/format_reward": 0.9,
+      "rewards/mean_confidence_reward": 0.6946446537971497,
       "step": 20
     },
     {
@@ -122,25 +122,25 @@
       "clip_ratio/low_mean": 0.0,
       "clip_ratio/low_min": 0.0,
       "clip_ratio/region_mean": 0.0,
-      "completions/clipped_ratio": 0.012109375,
+      "completions/clipped_ratio": 0.02890625,
       "completions/max_length": 1536.0,
-      "completions/max_terminated_length": 1288.4,
-      "completions/mean_length": 198.357421875,
-      "completions/mean_terminated_length": 181.95052490234374,
-      "completions/min_length": 19.6,
-      "completions/min_terminated_length": 19.6,
+      "completions/max_terminated_length": 1522.6,
+      "completions/mean_length": 469.98359375,
+      "completions/mean_terminated_length": 438.2403137207031,
+      "completions/min_length": 66.6,
+      "completions/min_terminated_length": 66.6,
       "epoch": 1.3091787439613527,
-      "grad_norm": 0.0031809473875910044,
+      "grad_norm": 0.32453644275665283,
       "learning_rate": 1e-06,
-      "loss": 0.0439,
-      "num_tokens": 12549638.0,
-      "reward": 0.556640625,
-      "reward_std": 0.19344157576560975,
-      "rewards/accuracy_reward": 0.1640625,
-      "rewards/brier_reward": 0.0,
-      "rewards/confidence_one_or_zero": 0.0,
-      "rewards/format_reward": 0.94921875,
-      "rewards/mean_confidence_reward": 0.0,
+      "loss": 0.0493,
+      "num_tokens": 19948636.0,
+      "reward": 0.7361393928527832,
+      "reward_std": 0.315061616897583,
+      "rewards/accuracy_reward": 0.107421875,
+      "rewards/brier_reward": 0.4398553490638733,
+      "rewards/confidence_one_or_zero": 0.07734375,
+      "rewards/format_reward": 0.925,
+      "rewards/mean_confidence_reward": 0.6970306396484375,
       "step": 25
     },
     {
@@ -149,25 +149,25 @@
       "clip_ratio/low_mean": 0.0,
       "clip_ratio/low_min": 0.0,
       "clip_ratio/region_mean": 0.0,
-      "completions/clipped_ratio": 0.006640625,
+      "completions/clipped_ratio": 0.025,
       "completions/max_length": 1536.0,
-      "completions/max_terminated_length": 1428.2,
-      "completions/mean_length": 164.6171875,
-      "completions/mean_terminated_length": 155.45379943847655,
-      "completions/min_length": 22.0,
-      "completions/min_terminated_length": 22.0,
+      "completions/max_terminated_length": 1524.8,
+      "completions/mean_length": 444.559765625,
+      "completions/mean_terminated_length": 416.5305541992187,
+      "completions/min_length": 77.6,
+      "completions/min_terminated_length": 77.6,
       "epoch": 1.5668276972624797,
-      "grad_norm": 0.003916527610272169,
+      "grad_norm": 0.00270429952070117,
       "learning_rate": 1e-06,
-      "loss": 0.0237,
-      "num_tokens": 14889882.0,
-      "reward": 0.5884765625,
-      "reward_std": 0.19649214446544647,
-      "rewards/accuracy_reward": 0.203515625,
-      "rewards/brier_reward": 0.0,
-      "rewards/confidence_one_or_zero": 0.0,
-      "rewards/format_reward": 0.9734375,
-      "rewards/mean_confidence_reward": 0.0,
+      "loss": 0.0415,
+      "num_tokens": 23791453.0,
+      "reward": 0.7462241291999817,
+      "reward_std": 0.30905682444572447,
+      "rewards/accuracy_reward": 0.110546875,
+      "rewards/brier_reward": 0.4447904586791992,
+      "rewards/confidence_one_or_zero": 0.06484375,
+      "rewards/format_reward": 0.937109375,
+      "rewards/mean_confidence_reward": 0.7072680354118347,
       "step": 30
     },
     {
@@ -176,25 +176,25 @@
       "clip_ratio/low_mean": 0.0,
       "clip_ratio/low_min": 0.0,
       "clip_ratio/region_mean": 0.0,
-      "completions/clipped_ratio": 0.005078125,
-      "completions/max_length": 1519.6,
-      "completions/max_terminated_length": 1345.6,
-      "completions/mean_length": 146.447265625,
-      "completions/mean_terminated_length": 139.37295532226562,
-      "completions/min_length": 20.0,
-      "completions/min_terminated_length": 20.0,
+      "completions/clipped_ratio": 0.025390625,
+      "completions/max_length": 1536.0,
+      "completions/max_terminated_length": 1511.2,
+      "completions/mean_length": 425.152734375,
+      "completions/mean_terminated_length": 396.2151611328125,
+      "completions/min_length": 50.8,
+      "completions/min_terminated_length": 50.8,
       "epoch": 1.8244766505636072,
-      "grad_norm": 0.003678182838484645,
+      "grad_norm": 0.002345304237678647,
       "learning_rate": 1e-06,
-      "loss": 0.0283,
-      "num_tokens": 17175619.0,
-      "reward": 0.60078125,
-      "reward_std": 0.1955224335193634,
-      "rewards/accuracy_reward": 0.216796875,
-      "rewards/brier_reward": 0.0,
-      "rewards/confidence_one_or_zero": 0.0,
-      "rewards/format_reward": 0.984765625,
-      "rewards/mean_confidence_reward": 0.0,
+      "loss": 0.0458,
+      "num_tokens": 27576596.0,
+      "reward": 0.7759710311889648,
+      "reward_std": 0.30494046211242676,
+      "rewards/accuracy_reward": 0.1234375,
+      "rewards/brier_reward": 0.4824093222618103,
+      "rewards/confidence_one_or_zero": 0.06796875,
+      "rewards/format_reward": 0.94609375,
+      "rewards/mean_confidence_reward": 0.681143045425415,
       "step": 35
     },
     {
@@ -203,25 +203,25 @@
       "clip_ratio/low_mean": 0.0,
       "clip_ratio/low_min": 0.0,
       "clip_ratio/region_mean": 0.0,
-      "completions/clipped_ratio": 0.00574111238532109,
-      "completions/max_length": 1488.6,
-      "completions/max_terminated_length": 1305.6,
-      "completions/mean_length": 141.31993713378907,
-      "completions/mean_terminated_length": 133.36893768310546,
-      "completions/min_length": 20.8,
-      "completions/min_terminated_length": 20.8,
+      "completions/clipped_ratio": 0.01511611238532109,
+      "completions/max_length": 1536.0,
+      "completions/max_terminated_length": 1485.4,
+      "completions/mean_length": 392.34654541015624,
+      "completions/mean_terminated_length": 374.88779907226564,
+      "completions/min_length": 82.6,
+      "completions/min_terminated_length": 82.6,
       "epoch": 2.103059581320451,
-      "grad_norm": 0.31614598631858826,
+      "grad_norm": 0.33194369077682495,
       "learning_rate": 1e-06,
-      "loss": 0.028,
-      "num_tokens": 19445973.0,
-      "reward": 0.61953125,
-      "reward_std": 0.20828622579574585,
-      "rewards/accuracy_reward": 0.257421875,
-      "rewards/brier_reward": 0.0,
-      "rewards/confidence_one_or_zero": 0.0,
-      "rewards/format_reward": 0.981640625,
-      "rewards/mean_confidence_reward": 0.0,
+      "loss": 0.0346,
+      "num_tokens": 31277252.0,
+      "reward": 0.8171104550361633,
+      "reward_std": 0.3093654751777649,
+      "rewards/accuracy_reward": 0.157421875,
+      "rewards/brier_reward": 0.5166412889957428,
+      "rewards/confidence_one_or_zero": 0.06953125,
+      "rewards/format_reward": 0.96015625,
+      "rewards/mean_confidence_reward": 0.6755032181739807,
       "step": 40
     },
     {
@@ -230,25 +230,25 @@
       "clip_ratio/low_mean": 0.0,
       "clip_ratio/low_min": 0.0,
       "clip_ratio/region_mean": 0.0,
-      "completions/clipped_ratio": 0.006640625,
+      "completions/clipped_ratio": 0.013671875,
       "completions/max_length": 1536.0,
-      "completions/max_terminated_length": 1335.6,
-      "completions/mean_length": 132.665234375,
-      "completions/mean_terminated_length": 123.2823257446289,
-      "completions/min_length": 21.0,
-      "completions/min_terminated_length": 21.0,
+      "completions/max_terminated_length": 1418.8,
+      "completions/mean_length": 357.22265625,
+      "completions/mean_terminated_length": 340.8806396484375,
+      "completions/min_length": 74.4,
+      "completions/min_terminated_length": 74.4,
       "epoch": 2.360708534621578,
-      "grad_norm": 0.0037020090967416763,
+      "grad_norm": 0.00267302175052464,
       "learning_rate": 1e-06,
-      "loss": 0.0276,
-      "num_tokens": 21705156.0,
-      "reward": 0.6427734375,
-      "reward_std": 0.21320482790470124,
-      "rewards/accuracy_reward": 0.298828125,
-      "rewards/brier_reward": 0.0,
-      "rewards/confidence_one_or_zero": 0.0,
-      "rewards/format_reward": 0.98671875,
-      "rewards/mean_confidence_reward": 0.0,
+      "loss": 0.035,
+      "num_tokens": 34897222.0,
+      "reward": 0.8488460063934327,
+      "reward_std": 0.31457024812698364,
+      "rewards/accuracy_reward": 0.181640625,
+      "rewards/brier_reward": 0.5469092965126038,
+      "rewards/confidence_one_or_zero": 0.066015625,
+      "rewards/format_reward": 0.969140625,
+      "rewards/mean_confidence_reward": 0.6639328360557556,
       "step": 45
     },
     {
@@ -257,48 +257,48 @@
       "clip_ratio/low_mean": 0.0,
       "clip_ratio/low_min": 0.0,
       "clip_ratio/region_mean": 0.0,
-      "completions/clipped_ratio": 0.0015625,
-      "completions/max_length": 1374.2,
-      "completions/max_terminated_length": 1175.2,
-      "completions/mean_length": 113.80234375,
-      "completions/mean_terminated_length": 111.57572784423829,
-      "completions/min_length": 18.4,
-      "completions/min_terminated_length": 18.4,
+      "completions/clipped_ratio": 0.013671875,
+      "completions/max_length": 1536.0,
+      "completions/max_terminated_length": 1464.4,
+      "completions/mean_length": 345.26796875,
+      "completions/mean_terminated_length": 328.7436828613281,
+      "completions/min_length": 70.8,
+      "completions/min_terminated_length": 70.8,
       "epoch": 2.6183574879227054,
-      "grad_norm": 0.3358711898326874,
+      "grad_norm": 0.003008283441886306,
       "learning_rate": 1e-06,
-      "loss": 0.0166,
-      "num_tokens": 23885994.0,
-      "reward": 0.665625,
-      "reward_std": 0.2116748332977295,
-      "rewards/accuracy_reward": 0.338671875,
-      "rewards/brier_reward": 0.0,
-      "rewards/confidence_one_or_zero": 0.0,
-      "rewards/format_reward": 0.992578125,
-      "rewards/mean_confidence_reward": 0.0,
+      "loss": 0.0415,
+      "num_tokens": 38456532.0,
+      "reward": 0.8908233880996704,
+      "reward_std": 0.3146331191062927,
+      "rewards/accuracy_reward": 0.21796875,
+      "rewards/brier_reward": 0.589457881450653,
+      "rewards/confidence_one_or_zero": 0.055078125,
+      "rewards/format_reward": 0.97421875,
+      "rewards/mean_confidence_reward": 0.6358984112739563,
       "step": 50
     },
     {
       "epoch": 2.6183574879227054,
-      "eval_completions/clipped_ratio": 0.0036231884057971015,
-      "eval_completions/max_length": 385.3985507246377,
-      "eval_completions/max_terminated_length": 350.32608695652175,
-      "eval_completions/mean_length": 117.5498188405797,
-      "eval_completions/mean_terminated_length": 112.37914088152456,
-      "eval_completions/min_length": 38.92028985507246,
-      "eval_completions/min_terminated_length": 38.92028985507246,
+      "eval_completions/clipped_ratio": 0.009057971014492754,
+      "eval_completions/max_length": 788.3478260869565,
+      "eval_completions/max_terminated_length": 733.0144927536232,
+      "eval_completions/mean_length": 323.0208333333333,
+      "eval_completions/mean_terminated_length": 312.0772527058919,
+      "eval_completions/min_length": 131.19565217391303,
+      "eval_completions/min_terminated_length": 131.19565217391303,
       "eval_loss": 0.0,
-      "eval_num_tokens": 23885994.0,
-      "eval_reward": 0.6829710144927537,
-      "eval_reward_std": 0.21439332471809525,
-      "eval_rewards/accuracy_reward": 0.3740942028985507,
-      "eval_rewards/brier_reward": 0.0,
-      "eval_rewards/confidence_one_or_zero": 0.0,
-      "eval_rewards/format_reward": 0.9918478260869565,
-      "eval_rewards/mean_confidence_reward": 0.0,
-      "eval_runtime": 283.8399,
-      "eval_samples_per_second": 0.486,
-      "eval_steps_per_second": 0.063,
+      "eval_num_tokens": 38456532.0,
+      "eval_reward": 0.9158611794312795,
+      "eval_reward_std": 0.3136252526571785,
+      "eval_rewards/accuracy_reward": 0.24094202898550723,
+      "eval_rewards/brier_reward": 0.6107065263001815,
+      "eval_rewards/confidence_one_or_zero": 0.050724637681159424,
+      "eval_rewards/format_reward": 0.980072463768116,
+      "eval_rewards/mean_confidence_reward": 0.6292391296314157,
+      "eval_runtime": 570.9473,
+      "eval_samples_per_second": 0.242,
+      "eval_steps_per_second": 0.032,
       "step": 50
     },
     {
@@ -307,25 +307,25 @@
       "clip_ratio/low_mean": 0.0,
       "clip_ratio/low_min": 0.0,
       "clip_ratio/region_mean": 0.0,
-      "completions/clipped_ratio": 0.001953125,
-      "completions/max_length": 1499.4,
-      "completions/max_terminated_length": 1196.8,
-      "completions/mean_length": 109.181640625,
-      "completions/mean_terminated_length": 106.39108734130859,
-      "completions/min_length": 19.2,
-      "completions/min_terminated_length": 19.2,
+      "completions/clipped_ratio": 0.01328125,
+      "completions/max_length": 1536.0,
+      "completions/max_terminated_length": 1483.8,
+      "completions/mean_length": 309.042578125,
+      "completions/mean_terminated_length": 292.53196411132814,
+      "completions/min_length": 71.8,
+      "completions/min_terminated_length": 71.8,
       "epoch": 2.8760064412238324,
-      "grad_norm": 0.004169007763266563,
+      "grad_norm": 0.0024981796741485596,
       "learning_rate": 1e-06,
-      "loss": 0.0147,
-      "num_tokens": 26106195.0,
-      "reward": 0.680078125,
-      "reward_std": 0.2129088193178177,
-      "rewards/accuracy_reward": 0.3640625,
-      "rewards/brier_reward": 0.0,
-      "rewards/confidence_one_or_zero": 0.0,
-      "rewards/format_reward": 0.99609375,
-      "rewards/mean_confidence_reward": 0.0,
+      "loss": 0.0362,
+      "num_tokens": 41974297.0,
+      "reward": 0.936638355255127,
+      "reward_std": 0.3146317660808563,
+      "rewards/accuracy_reward": 0.264453125,
+      "rewards/brier_reward": 0.6279629111289978,
+      "rewards/confidence_one_or_zero": 0.05390625,
+      "rewards/format_reward": 0.980859375,
+      "rewards/mean_confidence_reward": 0.6118968725204468,
       "step": 55
     },
     {
@@ -334,33 +334,33 @@
       "clip_ratio/low_mean": 0.0,
       "clip_ratio/low_min": 0.0,
       "clip_ratio/region_mean": 0.0,
-      "completions/clipped_ratio": 0.005563718463302725,
-      "completions/max_length": 1536.0,
-      "completions/max_terminated_length": 1031.0,
-      "completions/mean_length": 108.30762481689453,
-      "completions/mean_terminated_length": 100.29183578491211,
-      "completions/min_length": 22.5,
-      "completions/min_terminated_length": 22.5,
+      "completions/clipped_ratio": 0.0022935779816513624,
+      "completions/max_length": 1418.0,
+      "completions/max_terminated_length": 1323.5,
+      "completions/mean_length": 279.01670837402344,
+      "completions/mean_terminated_length": 276.12876892089844,
+      "completions/min_length": 71.5,
+      "completions/min_terminated_length": 71.5,
       "epoch": 2.9790660225442833,
-      "num_tokens": 26989330.0,
-      "reward": 0.6923828125,
-      "reward_std": 0.2056785449385643,
-      "rewards/accuracy_reward": 0.3896484375,
-      "rewards/brier_reward": 0.0,
-      "rewards/confidence_one_or_zero": 0.0,
-      "rewards/format_reward": 0.9951171875,
-      "rewards/mean_confidence_reward": 0.0,
+      "num_tokens": 43354056.0,
+      "reward": 0.970618486404419,
+      "reward_std": 0.30738210678100586,
+      "rewards/accuracy_reward": 0.3017578125,
+      "rewards/brier_reward": 0.6502200961112976,
+      "rewards/confidence_one_or_zero": 0.041015625,
+      "rewards/format_reward": 0.9892578125,
+      "rewards/mean_confidence_reward": 0.6022265553474426,
       "step": 57,
       "total_flos": 0.0,
-      "train_loss": 0.034499143952863257,
-      "train_runtime": 7332.8053,
-      "train_samples_per_second": 0.508,
-      "train_steps_per_second": 0.008
+      "train_loss": 0.04462047953877533,
+      "train_runtime": 11303.71,
+      "train_samples_per_second": 0.33,
+      "train_steps_per_second": 0.005
     }
   ],
   "logging_steps": 5,
   "max_steps": 57,
-  "num_input_tokens_seen": 26989330,
+  "num_input_tokens_seen": 43354056,
   "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 sha256:8fb871c75d4ce1a66d4015a05ff738fe7ff4eb2e301a47d588ffc368350e7357
+oid sha256:927b7c96664da05b1fa64705c9b22ad61bc228181267f557e71e3e2bd61e9470
 size 7992