| { |
| "best_global_step": null, |
| "best_metric": null, |
| "best_model_checkpoint": null, |
| "epoch": 0.35486160397445, |
| "eval_steps": 500, |
| "global_step": 500, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.93359375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4080.0, |
| "completions/mean_length": 4003.748046875, |
| "completions/mean_terminated_length": 2706.794189453125, |
| "completions/min_length": 1057.0, |
| "completions/min_terminated_length": 1057.0, |
| "epoch": 0.0007097232079489, |
| "frac_reward_zero_std": 0.8125, |
| "grad_norm": 0.06193803250789642, |
| "learning_rate": 1e-06, |
| "loss": 0.0106, |
| "num_tokens": 2095727.0, |
| "reward": 0.068359375, |
| "reward_std": 0.0723423957824707, |
| "rewards/simpleverify_reward/mean": 0.068359375, |
| "rewards/simpleverify_reward/std": 0.25260838866233826, |
| "step": 1 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.78515625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4072.0, |
| "completions/mean_length": 3713.083984375, |
| "completions/mean_terminated_length": 2313.699951171875, |
| "completions/min_length": 773.0, |
| "completions/min_terminated_length": 773.0, |
| "epoch": 0.0014194464158978, |
| "frac_reward_zero_std": 0.625, |
| "grad_norm": 0.09702980518341064, |
| "learning_rate": 9.98e-07, |
| "loss": 0.014, |
| "num_tokens": 4043866.0, |
| "reward": 0.146484375, |
| "reward_std": 0.1376640498638153, |
| "rewards/simpleverify_reward/mean": 0.146484375, |
| "rewards/simpleverify_reward/std": 0.35393697023391724, |
| "step": 2 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.8515625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4077.0, |
| "completions/mean_length": 3900.49609375, |
| "completions/mean_terminated_length": 2778.921142578125, |
| "completions/min_length": 1308.0, |
| "completions/min_terminated_length": 1308.0, |
| "epoch": 0.0021291696238466998, |
| "frac_reward_zero_std": 0.75, |
| "grad_norm": 0.0868891030550003, |
| "learning_rate": 9.959999999999999e-07, |
| "loss": 0.006, |
| "num_tokens": 6087544.0, |
| "reward": 0.146484375, |
| "reward_std": 0.09110259264707565, |
| "rewards/simpleverify_reward/mean": 0.146484375, |
| "rewards/simpleverify_reward/std": 0.35393697023391724, |
| "step": 3 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.849609375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4066.0, |
| "completions/mean_length": 3933.599609375, |
| "completions/mean_terminated_length": 3016.142822265625, |
| "completions/min_length": 1442.0, |
| "completions/min_terminated_length": 1442.0, |
| "epoch": 0.0028388928317956, |
| "frac_reward_zero_std": 0.625, |
| "grad_norm": 0.09985340386629105, |
| "learning_rate": 9.94e-07, |
| "loss": 0.0125, |
| "num_tokens": 8142747.0, |
| "reward": 0.12890625, |
| "reward_std": 0.14248856902122498, |
| "rewards/simpleverify_reward/mean": 0.12890625, |
| "rewards/simpleverify_reward/std": 0.33542385697364807, |
| "step": 4 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.779296875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4050.0, |
| "completions/mean_length": 3744.625, |
| "completions/mean_terminated_length": 2503.92919921875, |
| "completions/min_length": 1020.0, |
| "completions/min_terminated_length": 1020.0, |
| "epoch": 0.0035486160397444995, |
| "frac_reward_zero_std": 0.71875, |
| "grad_norm": 0.08821574598550797, |
| "learning_rate": 9.92e-07, |
| "loss": 0.0084, |
| "num_tokens": 10111595.0, |
| "reward": 0.13671875, |
| "reward_std": 0.10212816298007965, |
| "rewards/simpleverify_reward/mean": 0.13671875, |
| "rewards/simpleverify_reward/std": 0.3438861668109894, |
| "step": 5 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.87109375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 3972.0, |
| "completions/mean_length": 3919.439453125, |
| "completions/mean_terminated_length": 2726.318359375, |
| "completions/min_length": 1073.0, |
| "completions/min_terminated_length": 1073.0, |
| "epoch": 0.0042583392476933995, |
| "frac_reward_zero_std": 0.65625, |
| "grad_norm": 0.10246389359235764, |
| "learning_rate": 9.9e-07, |
| "loss": 0.0043, |
| "num_tokens": 12160764.0, |
| "reward": 0.12109375, |
| "reward_std": 0.12669725716114044, |
| "rewards/simpleverify_reward/mean": 0.12109375, |
| "rewards/simpleverify_reward/std": 0.3265552520751953, |
| "step": 6 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.765625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4086.0, |
| "completions/mean_length": 3722.5625, |
| "completions/mean_terminated_length": 2502.666748046875, |
| "completions/min_length": 1037.0, |
| "completions/min_terminated_length": 1037.0, |
| "epoch": 0.0049680624556423, |
| "frac_reward_zero_std": 0.5625, |
| "grad_norm": 0.11841621994972229, |
| "learning_rate": 9.88e-07, |
| "loss": 0.0058, |
| "num_tokens": 14115068.0, |
| "reward": 0.162109375, |
| "reward_std": 0.1561369150876999, |
| "rewards/simpleverify_reward/mean": 0.162109375, |
| "rewards/simpleverify_reward/std": 0.3689115643501282, |
| "step": 7 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.861328125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4088.0, |
| "completions/mean_length": 3913.859375, |
| "completions/mean_terminated_length": 2782.53515625, |
| "completions/min_length": 1221.0, |
| "completions/min_terminated_length": 1221.0, |
| "epoch": 0.0056777856635912, |
| "frac_reward_zero_std": 0.625, |
| "grad_norm": 0.10141696780920029, |
| "learning_rate": 9.86e-07, |
| "loss": 0.0229, |
| "num_tokens": 16162324.0, |
| "reward": 0.111328125, |
| "reward_std": 0.150614395737648, |
| "rewards/simpleverify_reward/mean": 0.111328125, |
| "rewards/simpleverify_reward/std": 0.31484565138816833, |
| "step": 8 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.79296875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4060.0, |
| "completions/mean_length": 3759.650390625, |
| "completions/mean_terminated_length": 2471.367919921875, |
| "completions/min_length": 1101.0, |
| "completions/min_terminated_length": 1101.0, |
| "epoch": 0.0063875088715401, |
| "frac_reward_zero_std": 0.75, |
| "grad_norm": 0.0836983472108841, |
| "learning_rate": 9.84e-07, |
| "loss": 0.0101, |
| "num_tokens": 18127505.0, |
| "reward": 0.1796875, |
| "reward_std": 0.09651977568864822, |
| "rewards/simpleverify_reward/mean": 0.1796875, |
| "rewards/simpleverify_reward/std": 0.38430243730545044, |
| "step": 9 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.767578125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4091.0, |
| "completions/mean_length": 3743.142578125, |
| "completions/mean_terminated_length": 2577.82373046875, |
| "completions/min_length": 962.0, |
| "completions/min_terminated_length": 962.0, |
| "epoch": 0.007097232079488999, |
| "frac_reward_zero_std": 0.53125, |
| "grad_norm": 0.1163906678557396, |
| "learning_rate": 9.819999999999999e-07, |
| "loss": 0.0073, |
| "num_tokens": 20097274.0, |
| "reward": 0.169921875, |
| "reward_std": 0.1669841706752777, |
| "rewards/simpleverify_reward/mean": 0.169921875, |
| "rewards/simpleverify_reward/std": 0.3759314715862274, |
| "step": 10 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.791015625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4090.0, |
| "completions/mean_length": 3868.671875, |
| "completions/mean_terminated_length": 3008.22412109375, |
| "completions/min_length": 1433.0, |
| "completions/min_terminated_length": 1433.0, |
| "epoch": 0.007806955287437899, |
| "frac_reward_zero_std": 0.59375, |
| "grad_norm": 0.09666936099529266, |
| "learning_rate": 9.8e-07, |
| "loss": 0.0192, |
| "num_tokens": 22129650.0, |
| "reward": 0.146484375, |
| "reward_std": 0.17313450574874878, |
| "rewards/simpleverify_reward/mean": 0.146484375, |
| "rewards/simpleverify_reward/std": 0.35393697023391724, |
| "step": 11 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.88671875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4083.0, |
| "completions/mean_length": 3967.998046875, |
| "completions/mean_terminated_length": 2966.0517578125, |
| "completions/min_length": 771.0, |
| "completions/min_terminated_length": 771.0, |
| "epoch": 0.008516678495386799, |
| "frac_reward_zero_std": 0.59375, |
| "grad_norm": 0.10196839272975922, |
| "learning_rate": 9.78e-07, |
| "loss": 0.0202, |
| "num_tokens": 24202113.0, |
| "reward": 0.107421875, |
| "reward_std": 0.16589424014091492, |
| "rewards/simpleverify_reward/mean": 0.107421875, |
| "rewards/simpleverify_reward/std": 0.30995169281959534, |
| "step": 12 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.751953125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4095.0, |
| "completions/mean_length": 3760.6796875, |
| "completions/mean_terminated_length": 2744.157470703125, |
| "completions/min_length": 1060.0, |
| "completions/min_terminated_length": 1060.0, |
| "epoch": 0.009226401703335699, |
| "frac_reward_zero_std": 0.5625, |
| "grad_norm": 0.10873758792877197, |
| "learning_rate": 9.759999999999998e-07, |
| "loss": 0.0179, |
| "num_tokens": 26175549.0, |
| "reward": 0.216796875, |
| "reward_std": 0.16234853863716125, |
| "rewards/simpleverify_reward/mean": 0.216796875, |
| "rewards/simpleverify_reward/std": 0.4124660789966583, |
| "step": 13 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.662109375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4068.0, |
| "completions/mean_length": 3702.04296875, |
| "completions/mean_terminated_length": 2930.0693359375, |
| "completions/min_length": 1573.0, |
| "completions/min_terminated_length": 1573.0, |
| "epoch": 0.0099361249112846, |
| "frac_reward_zero_std": 0.46875, |
| "grad_norm": 0.1272289901971817, |
| "learning_rate": 9.74e-07, |
| "loss": 0.0182, |
| "num_tokens": 28110531.0, |
| "reward": 0.265625, |
| "reward_std": 0.22406208515167236, |
| "rewards/simpleverify_reward/mean": 0.265625, |
| "rewards/simpleverify_reward/std": 0.44209739565849304, |
| "step": 14 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.76171875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4092.0, |
| "completions/mean_length": 3747.599609375, |
| "completions/mean_terminated_length": 2633.860595703125, |
| "completions/min_length": 1285.0, |
| "completions/min_terminated_length": 1285.0, |
| "epoch": 0.0106458481192335, |
| "frac_reward_zero_std": 0.5625, |
| "grad_norm": 0.1079520657658577, |
| "learning_rate": 9.72e-07, |
| "loss": 0.0102, |
| "num_tokens": 30073526.0, |
| "reward": 0.189453125, |
| "reward_std": 0.16478893160820007, |
| "rewards/simpleverify_reward/mean": 0.189453125, |
| "rewards/simpleverify_reward/std": 0.3922513723373413, |
| "step": 15 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.697265625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 3973.0, |
| "completions/mean_length": 3606.822265625, |
| "completions/mean_terminated_length": 2480.135498046875, |
| "completions/min_length": 1252.0, |
| "completions/min_terminated_length": 1252.0, |
| "epoch": 0.0113555713271824, |
| "frac_reward_zero_std": 0.53125, |
| "grad_norm": 0.11524096131324768, |
| "learning_rate": 9.7e-07, |
| "loss": 0.0159, |
| "num_tokens": 31961883.0, |
| "reward": 0.255859375, |
| "reward_std": 0.1792701929807663, |
| "rewards/simpleverify_reward/mean": 0.255859375, |
| "rewards/simpleverify_reward/std": 0.43676990270614624, |
| "step": 16 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.912109375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4087.0, |
| "completions/mean_length": 4031.818359375, |
| "completions/mean_terminated_length": 3365.755615234375, |
| "completions/min_length": 1804.0, |
| "completions/min_terminated_length": 1804.0, |
| "epoch": 0.0120652945351313, |
| "frac_reward_zero_std": 0.625, |
| "grad_norm": 0.0938996747136116, |
| "learning_rate": 9.679999999999999e-07, |
| "loss": 0.0085, |
| "num_tokens": 34076542.0, |
| "reward": 0.091796875, |
| "reward_std": 0.1439877450466156, |
| "rewards/simpleverify_reward/mean": 0.091796875, |
| "rewards/simpleverify_reward/std": 0.289021372795105, |
| "step": 17 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.765625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4089.0, |
| "completions/mean_length": 3758.666015625, |
| "completions/mean_terminated_length": 2656.70849609375, |
| "completions/min_length": 1080.0, |
| "completions/min_terminated_length": 1080.0, |
| "epoch": 0.0127750177430802, |
| "frac_reward_zero_std": 0.5625, |
| "grad_norm": 0.10947775095701218, |
| "learning_rate": 9.66e-07, |
| "loss": 0.0145, |
| "num_tokens": 36048915.0, |
| "reward": 0.244140625, |
| "reward_std": 0.17310172319412231, |
| "rewards/simpleverify_reward/mean": 0.244140625, |
| "rewards/simpleverify_reward/std": 0.42999663949012756, |
| "step": 18 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.828125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4096.0, |
| "completions/mean_length": 3837.333984375, |
| "completions/mean_terminated_length": 2591.0341796875, |
| "completions/min_length": 1190.0, |
| "completions/min_terminated_length": 1190.0, |
| "epoch": 0.0134847409510291, |
| "frac_reward_zero_std": 0.71875, |
| "grad_norm": 0.08318022638559341, |
| "learning_rate": 9.64e-07, |
| "loss": 0.0036, |
| "num_tokens": 38062398.0, |
| "reward": 0.126953125, |
| "reward_std": 0.10612908005714417, |
| "rewards/simpleverify_reward/mean": 0.126953125, |
| "rewards/simpleverify_reward/std": 0.33324605226516724, |
| "step": 19 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.8828125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 3932.6796875, |
| "completions/mean_terminated_length": 2702.33349609375, |
| "completions/min_length": 1172.0, |
| "completions/min_terminated_length": 1172.0, |
| "epoch": 0.014194464158977998, |
| "frac_reward_zero_std": 0.78125, |
| "grad_norm": 0.07120879739522934, |
| "learning_rate": 9.619999999999999e-07, |
| "loss": 0.0061, |
| "num_tokens": 40124362.0, |
| "reward": 0.0859375, |
| "reward_std": 0.08619783818721771, |
| "rewards/simpleverify_reward/mean": 0.0859375, |
| "rewards/simpleverify_reward/std": 0.28054583072662354, |
| "step": 20 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.73828125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4093.0, |
| "completions/mean_length": 3657.986328125, |
| "completions/mean_terminated_length": 2422.3955078125, |
| "completions/min_length": 1145.0, |
| "completions/min_terminated_length": 1145.0, |
| "epoch": 0.014904187366926898, |
| "frac_reward_zero_std": 0.71875, |
| "grad_norm": 0.08722103387117386, |
| "learning_rate": 9.6e-07, |
| "loss": 0.0131, |
| "num_tokens": 42040435.0, |
| "reward": 0.24609375, |
| "reward_std": 0.11635598540306091, |
| "rewards/simpleverify_reward/mean": 0.24609375, |
| "rewards/simpleverify_reward/std": 0.4311550557613373, |
| "step": 21 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.689453125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4096.0, |
| "completions/mean_length": 3573.80078125, |
| "completions/mean_terminated_length": 2414.45263671875, |
| "completions/min_length": 709.0, |
| "completions/min_terminated_length": 709.0, |
| "epoch": 0.015613910574875798, |
| "frac_reward_zero_std": 0.5625, |
| "grad_norm": 0.11641096323728561, |
| "learning_rate": 9.58e-07, |
| "loss": 0.0248, |
| "num_tokens": 43920765.0, |
| "reward": 0.197265625, |
| "reward_std": 0.1678110957145691, |
| "rewards/simpleverify_reward/mean": 0.197265625, |
| "rewards/simpleverify_reward/std": 0.3983237147331238, |
| "step": 22 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.720703125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4045.0, |
| "completions/mean_length": 3699.919921875, |
| "completions/mean_terminated_length": 2677.8671875, |
| "completions/min_length": 1078.0, |
| "completions/min_terminated_length": 1078.0, |
| "epoch": 0.016323633782824698, |
| "frac_reward_zero_std": 0.53125, |
| "grad_norm": 0.11638127267360687, |
| "learning_rate": 9.559999999999998e-07, |
| "loss": 0.019, |
| "num_tokens": 45855636.0, |
| "reward": 0.169921875, |
| "reward_std": 0.18488094210624695, |
| "rewards/simpleverify_reward/mean": 0.169921875, |
| "rewards/simpleverify_reward/std": 0.3759314715862274, |
| "step": 23 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.75390625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4079.0, |
| "completions/mean_length": 3795.697265625, |
| "completions/mean_terminated_length": 2875.722412109375, |
| "completions/min_length": 1220.0, |
| "completions/min_terminated_length": 1220.0, |
| "epoch": 0.017033356990773598, |
| "frac_reward_zero_std": 0.53125, |
| "grad_norm": 0.11083409935235977, |
| "learning_rate": 9.539999999999999e-07, |
| "loss": 0.0095, |
| "num_tokens": 47841929.0, |
| "reward": 0.185546875, |
| "reward_std": 0.1795351207256317, |
| "rewards/simpleverify_reward/mean": 0.185546875, |
| "rewards/simpleverify_reward/std": 0.38912075757980347, |
| "step": 24 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.796875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 3814.24609375, |
| "completions/mean_terminated_length": 2708.904052734375, |
| "completions/min_length": 846.0, |
| "completions/min_terminated_length": 846.0, |
| "epoch": 0.017743080198722498, |
| "frac_reward_zero_std": 0.59375, |
| "grad_norm": 0.11103121191263199, |
| "learning_rate": 9.52e-07, |
| "loss": 0.0088, |
| "num_tokens": 49840247.0, |
| "reward": 0.1484375, |
| "reward_std": 0.14493131637573242, |
| "rewards/simpleverify_reward/mean": 0.1484375, |
| "rewards/simpleverify_reward/std": 0.35588082671165466, |
| "step": 25 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.767578125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4092.0, |
| "completions/mean_length": 3834.658203125, |
| "completions/mean_terminated_length": 2971.571533203125, |
| "completions/min_length": 1049.0, |
| "completions/min_terminated_length": 1049.0, |
| "epoch": 0.018452803406671398, |
| "frac_reward_zero_std": 0.53125, |
| "grad_norm": 0.11120044440031052, |
| "learning_rate": 9.499999999999999e-07, |
| "loss": 0.0124, |
| "num_tokens": 51847624.0, |
| "reward": 0.185546875, |
| "reward_std": 0.17720650136470795, |
| "rewards/simpleverify_reward/mean": 0.185546875, |
| "rewards/simpleverify_reward/std": 0.38912075757980347, |
| "step": 26 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.814453125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4048.0, |
| "completions/mean_length": 3873.873046875, |
| "completions/mean_terminated_length": 2898.852783203125, |
| "completions/min_length": 1598.0, |
| "completions/min_terminated_length": 1598.0, |
| "epoch": 0.0191625266146203, |
| "frac_reward_zero_std": 0.53125, |
| "grad_norm": 0.1120850145816803, |
| "learning_rate": 9.479999999999999e-07, |
| "loss": 0.0146, |
| "num_tokens": 53872295.0, |
| "reward": 0.15625, |
| "reward_std": 0.16736537218093872, |
| "rewards/simpleverify_reward/mean": 0.15625, |
| "rewards/simpleverify_reward/std": 0.36344730854034424, |
| "step": 27 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.7890625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4080.0, |
| "completions/mean_length": 3711.517578125, |
| "completions/mean_terminated_length": 2273.2685546875, |
| "completions/min_length": 859.0, |
| "completions/min_terminated_length": 859.0, |
| "epoch": 0.0198722498225692, |
| "frac_reward_zero_std": 0.53125, |
| "grad_norm": 0.11432869732379913, |
| "learning_rate": 9.459999999999999e-07, |
| "loss": 0.0044, |
| "num_tokens": 55823728.0, |
| "reward": 0.146484375, |
| "reward_std": 0.16484221816062927, |
| "rewards/simpleverify_reward/mean": 0.146484375, |
| "rewards/simpleverify_reward/std": 0.35393697023391724, |
| "step": 28 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.689453125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4086.0, |
| "completions/mean_length": 3760.837890625, |
| "completions/mean_terminated_length": 3016.73583984375, |
| "completions/min_length": 1176.0, |
| "completions/min_terminated_length": 1176.0, |
| "epoch": 0.0205819730305181, |
| "frac_reward_zero_std": 0.5, |
| "grad_norm": 0.11176938563585281, |
| "learning_rate": 9.439999999999999e-07, |
| "loss": 0.0178, |
| "num_tokens": 57795965.0, |
| "reward": 0.30078125, |
| "reward_std": 0.21012087166309357, |
| "rewards/simpleverify_reward/mean": 0.30078125, |
| "rewards/simpleverify_reward/std": 0.45904624462127686, |
| "step": 29 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.822265625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4073.0, |
| "completions/mean_length": 3856.87109375, |
| "completions/mean_terminated_length": 2750.571533203125, |
| "completions/min_length": 1278.0, |
| "completions/min_terminated_length": 1278.0, |
| "epoch": 0.021291696238467, |
| "frac_reward_zero_std": 0.65625, |
| "grad_norm": 0.09087643027305603, |
| "learning_rate": 9.419999999999999e-07, |
| "loss": 0.0107, |
| "num_tokens": 59818587.0, |
| "reward": 0.1640625, |
| "reward_std": 0.12603998184204102, |
| "rewards/simpleverify_reward/mean": 0.1640625, |
| "rewards/simpleverify_reward/std": 0.37069445848464966, |
| "step": 30 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.765625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4084.0, |
| "completions/mean_length": 3760.57421875, |
| "completions/mean_terminated_length": 2664.85009765625, |
| "completions/min_length": 1003.0, |
| "completions/min_terminated_length": 1003.0, |
| "epoch": 0.0220014194464159, |
| "frac_reward_zero_std": 0.5625, |
| "grad_norm": 0.11131440103054047, |
| "learning_rate": 9.399999999999999e-07, |
| "loss": 0.0155, |
| "num_tokens": 61782113.0, |
| "reward": 0.2421875, |
| "reward_std": 0.1788366734981537, |
| "rewards/simpleverify_reward/mean": 0.2421875, |
| "rewards/simpleverify_reward/std": 0.42882615327835083, |
| "step": 31 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.84375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4086.0, |
| "completions/mean_length": 3907.611328125, |
| "completions/mean_terminated_length": 2890.3125, |
| "completions/min_length": 1531.0, |
| "completions/min_terminated_length": 1531.0, |
| "epoch": 0.0227111426543648, |
| "frac_reward_zero_std": 0.625, |
| "grad_norm": 0.09195547550916672, |
| "learning_rate": 9.379999999999998e-07, |
| "loss": 0.0109, |
| "num_tokens": 63831162.0, |
| "reward": 0.138671875, |
| "reward_std": 0.14747919142246246, |
| "rewards/simpleverify_reward/mean": 0.138671875, |
| "rewards/simpleverify_reward/std": 0.34594178199768066, |
| "step": 32 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.662109375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4076.0, |
| "completions/mean_length": 3647.27734375, |
| "completions/mean_terminated_length": 2767.98828125, |
| "completions/min_length": 1122.0, |
| "completions/min_terminated_length": 1122.0, |
| "epoch": 0.0234208658623137, |
| "frac_reward_zero_std": 0.5, |
| "grad_norm": 0.11667177081108093, |
| "learning_rate": 9.36e-07, |
| "loss": 0.023, |
| "num_tokens": 65746728.0, |
| "reward": 0.322265625, |
| "reward_std": 0.18821430206298828, |
| "rewards/simpleverify_reward/mean": 0.322265625, |
| "rewards/simpleverify_reward/std": 0.46780112385749817, |
| "step": 33 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.669921875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4081.0, |
| "completions/mean_length": 3667.45703125, |
| "completions/mean_terminated_length": 2797.6923828125, |
| "completions/min_length": 973.0, |
| "completions/min_terminated_length": 973.0, |
| "epoch": 0.0241305890702626, |
| "frac_reward_zero_std": 0.40625, |
| "grad_norm": 0.13087135553359985, |
| "learning_rate": 9.34e-07, |
| "loss": 0.0295, |
| "num_tokens": 67686290.0, |
| "reward": 0.29296875, |
| "reward_std": 0.23394519090652466, |
| "rewards/simpleverify_reward/mean": 0.29296875, |
| "rewards/simpleverify_reward/std": 0.455569326877594, |
| "step": 34 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.658203125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4090.0, |
| "completions/mean_length": 3774.703125, |
| "completions/mean_terminated_length": 3155.97705078125, |
| "completions/min_length": 1493.0, |
| "completions/min_terminated_length": 1493.0, |
| "epoch": 0.0248403122782115, |
| "frac_reward_zero_std": 0.40625, |
| "grad_norm": 0.12667018175125122, |
| "learning_rate": 9.32e-07, |
| "loss": 0.0281, |
| "num_tokens": 69662058.0, |
| "reward": 0.3125, |
| "reward_std": 0.2654282748699188, |
| "rewards/simpleverify_reward/mean": 0.3125, |
| "rewards/simpleverify_reward/std": 0.4639657139778137, |
| "step": 35 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.734375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4095.0, |
| "completions/mean_length": 3775.576171875, |
| "completions/mean_terminated_length": 2889.698486328125, |
| "completions/min_length": 1423.0, |
| "completions/min_terminated_length": 1423.0, |
| "epoch": 0.0255500354861604, |
| "frac_reward_zero_std": 0.40625, |
| "grad_norm": 0.1257808655500412, |
| "learning_rate": 9.3e-07, |
| "loss": 0.0114, |
| "num_tokens": 71644081.0, |
| "reward": 0.208984375, |
| "reward_std": 0.2165631353855133, |
| "rewards/simpleverify_reward/mean": 0.208984375, |
| "rewards/simpleverify_reward/std": 0.40698084235191345, |
| "step": 36 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.759765625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4075.0, |
| "completions/mean_length": 3745.580078125, |
| "completions/mean_terminated_length": 2637.34130859375, |
| "completions/min_length": 1419.0, |
| "completions/min_terminated_length": 1419.0, |
| "epoch": 0.0262597586941093, |
| "frac_reward_zero_std": 0.625, |
| "grad_norm": 0.0989406630396843, |
| "learning_rate": 9.28e-07, |
| "loss": 0.014, |
| "num_tokens": 73606858.0, |
| "reward": 0.236328125, |
| "reward_std": 0.143095001578331, |
| "rewards/simpleverify_reward/mean": 0.236328125, |
| "rewards/simpleverify_reward/std": 0.42524150013923645, |
| "step": 37 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.62890625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4010.0, |
| "completions/mean_length": 3375.416015625, |
| "completions/mean_terminated_length": 2154.2158203125, |
| "completions/min_length": 857.0, |
| "completions/min_terminated_length": 857.0, |
| "epoch": 0.0269694819020582, |
| "frac_reward_zero_std": 0.40625, |
| "grad_norm": 0.135038822889328, |
| "learning_rate": 9.26e-07, |
| "loss": 0.0205, |
| "num_tokens": 75383439.0, |
| "reward": 0.3828125, |
| "reward_std": 0.20536984503269196, |
| "rewards/simpleverify_reward/mean": 0.3828125, |
| "rewards/simpleverify_reward/std": 0.486548513174057, |
| "step": 38 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.806640625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4045.0, |
| "completions/mean_length": 3823.478515625, |
| "completions/mean_terminated_length": 2686.595947265625, |
| "completions/min_length": 614.0, |
| "completions/min_terminated_length": 614.0, |
| "epoch": 0.027679205110007096, |
| "frac_reward_zero_std": 0.46875, |
| "grad_norm": 0.12824682891368866, |
| "learning_rate": 9.24e-07, |
| "loss": 0.0126, |
| "num_tokens": 77391412.0, |
| "reward": 0.18359375, |
| "reward_std": 0.19765016436576843, |
| "rewards/simpleverify_reward/mean": 0.18359375, |
| "rewards/simpleverify_reward/std": 0.3875311613082886, |
| "step": 39 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.794921875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4075.0, |
| "completions/mean_length": 3873.90234375, |
| "completions/mean_terminated_length": 3013.009521484375, |
| "completions/min_length": 1147.0, |
| "completions/min_terminated_length": 1147.0, |
| "epoch": 0.028388928317955996, |
| "frac_reward_zero_std": 0.53125, |
| "grad_norm": 0.11737268418073654, |
| "learning_rate": 9.22e-07, |
| "loss": 0.0182, |
| "num_tokens": 79420210.0, |
| "reward": 0.193359375, |
| "reward_std": 0.1833476573228836, |
| "rewards/simpleverify_reward/mean": 0.193359375, |
| "rewards/simpleverify_reward/std": 0.39531853795051575, |
| "step": 40 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.767578125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4089.0, |
| "completions/mean_length": 3707.45703125, |
| "completions/mean_terminated_length": 2424.285888671875, |
| "completions/min_length": 814.0, |
| "completions/min_terminated_length": 814.0, |
| "epoch": 0.029098651525904896, |
| "frac_reward_zero_std": 0.59375, |
| "grad_norm": 0.09561893343925476, |
| "learning_rate": 9.2e-07, |
| "loss": 0.01, |
| "num_tokens": 81359212.0, |
| "reward": 0.240234375, |
| "reward_std": 0.15835317969322205, |
| "rewards/simpleverify_reward/mean": 0.240234375, |
| "rewards/simpleverify_reward/std": 0.4276435375213623, |
| "step": 41 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.69921875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 3780.330078125, |
| "completions/mean_terminated_length": 3046.5, |
| "completions/min_length": 1411.0, |
| "completions/min_terminated_length": 1411.0, |
| "epoch": 0.029808374733853796, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.12864162027835846, |
| "learning_rate": 9.18e-07, |
| "loss": 0.0211, |
| "num_tokens": 83344885.0, |
| "reward": 0.26953125, |
| "reward_std": 0.2715563476085663, |
| "rewards/simpleverify_reward/mean": 0.26953125, |
| "rewards/simpleverify_reward/std": 0.44415023922920227, |
| "step": 42 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.7421875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4093.0, |
| "completions/mean_length": 3769.748046875, |
| "completions/mean_terminated_length": 2830.537841796875, |
| "completions/min_length": 1023.0, |
| "completions/min_terminated_length": 1023.0, |
| "epoch": 0.030518097941802696, |
| "frac_reward_zero_std": 0.46875, |
| "grad_norm": 0.11613088846206665, |
| "learning_rate": 9.16e-07, |
| "loss": 0.0204, |
| "num_tokens": 85318772.0, |
| "reward": 0.224609375, |
| "reward_std": 0.2001463770866394, |
| "rewards/simpleverify_reward/mean": 0.224609375, |
| "rewards/simpleverify_reward/std": 0.41773295402526855, |
| "step": 43 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.75, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4091.0, |
| "completions/mean_length": 3768.869140625, |
| "completions/mean_terminated_length": 2787.4765625, |
| "completions/min_length": 820.0, |
| "completions/min_terminated_length": 820.0, |
| "epoch": 0.031227821149751596, |
| "frac_reward_zero_std": 0.5625, |
| "grad_norm": 0.11177936941385269, |
| "learning_rate": 9.14e-07, |
| "loss": 0.0251, |
| "num_tokens": 87290897.0, |
| "reward": 0.216796875, |
| "reward_std": 0.16860191524028778, |
| "rewards/simpleverify_reward/mean": 0.216796875, |
| "rewards/simpleverify_reward/std": 0.4124660789966583, |
| "step": 44 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.578125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4087.0, |
| "completions/mean_length": 3504.71875, |
| "completions/mean_terminated_length": 2694.444580078125, |
| "completions/min_length": 920.0, |
| "completions/min_terminated_length": 920.0, |
| "epoch": 0.0319375443577005, |
| "frac_reward_zero_std": 0.375, |
| "grad_norm": 0.13649967312812805, |
| "learning_rate": 9.12e-07, |
| "loss": 0.0184, |
| "num_tokens": 89124305.0, |
| "reward": 0.337890625, |
| "reward_std": 0.25768181681632996, |
| "rewards/simpleverify_reward/mean": 0.337890625, |
| "rewards/simpleverify_reward/std": 0.4734536409378052, |
| "step": 45 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.712890625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4035.0, |
| "completions/mean_length": 3758.248046875, |
| "completions/mean_terminated_length": 2919.6123046875, |
| "completions/min_length": 1507.0, |
| "completions/min_terminated_length": 1507.0, |
| "epoch": 0.032647267565649396, |
| "frac_reward_zero_std": 0.4375, |
| "grad_norm": 0.12174121290445328, |
| "learning_rate": 9.1e-07, |
| "loss": 0.0259, |
| "num_tokens": 91093072.0, |
| "reward": 0.267578125, |
| "reward_std": 0.2229965329170227, |
| "rewards/simpleverify_reward/mean": 0.267578125, |
| "rewards/simpleverify_reward/std": 0.4431293308734894, |
| "step": 46 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.626953125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4091.0, |
| "completions/mean_length": 3610.69140625, |
| "completions/mean_terminated_length": 2795.068115234375, |
| "completions/min_length": 737.0, |
| "completions/min_terminated_length": 737.0, |
| "epoch": 0.0333569907735983, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.1326490193605423, |
| "learning_rate": 9.08e-07, |
| "loss": 0.0285, |
| "num_tokens": 92984930.0, |
| "reward": 0.27734375, |
| "reward_std": 0.2685825824737549, |
| "rewards/simpleverify_reward/mean": 0.27734375, |
| "rewards/simpleverify_reward/std": 0.4481254518032074, |
| "step": 47 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.669921875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4091.0, |
| "completions/mean_length": 3668.0859375, |
| "completions/mean_terminated_length": 2799.59765625, |
| "completions/min_length": 710.0, |
| "completions/min_terminated_length": 710.0, |
| "epoch": 0.034066713981547196, |
| "frac_reward_zero_std": 0.46875, |
| "grad_norm": 0.12671765685081482, |
| "learning_rate": 9.06e-07, |
| "loss": 0.0158, |
| "num_tokens": 94913950.0, |
| "reward": 0.25, |
| "reward_std": 0.20195434987545013, |
| "rewards/simpleverify_reward/mean": 0.25, |
| "rewards/simpleverify_reward/std": 0.43343618512153625, |
| "step": 48 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.703125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 3541.064453125, |
| "completions/mean_terminated_length": 2226.743408203125, |
| "completions/min_length": 823.0, |
| "completions/min_terminated_length": 823.0, |
| "epoch": 0.0347764371894961, |
| "frac_reward_zero_std": 0.46875, |
| "grad_norm": 0.1286790817975998, |
| "learning_rate": 9.039999999999999e-07, |
| "loss": 0.0178, |
| "num_tokens": 96774703.0, |
| "reward": 0.232421875, |
| "reward_std": 0.1943105161190033, |
| "rewards/simpleverify_reward/mean": 0.232421875, |
| "rewards/simpleverify_reward/std": 0.42278963327407837, |
| "step": 49 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.642578125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4060.0, |
| "completions/mean_length": 3555.24609375, |
| "completions/mean_terminated_length": 2583.071044921875, |
| "completions/min_length": 778.0, |
| "completions/min_terminated_length": 778.0, |
| "epoch": 0.035486160397444996, |
| "frac_reward_zero_std": 0.40625, |
| "grad_norm": 0.1379772573709488, |
| "learning_rate": 9.02e-07, |
| "loss": 0.0186, |
| "num_tokens": 98637453.0, |
| "reward": 0.34375, |
| "reward_std": 0.24374546110630035, |
| "rewards/simpleverify_reward/mean": 0.34375, |
| "rewards/simpleverify_reward/std": 0.4754233956336975, |
| "step": 50 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.61328125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4096.0, |
| "completions/mean_length": 3425.3359375, |
| "completions/mean_terminated_length": 2361.757568359375, |
| "completions/min_length": 565.0, |
| "completions/min_terminated_length": 565.0, |
| "epoch": 0.0361958836053939, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.169500932097435, |
| "learning_rate": 9e-07, |
| "loss": 0.002, |
| "num_tokens": 100437353.0, |
| "reward": 0.310546875, |
| "reward_std": 0.25840920209884644, |
| "rewards/simpleverify_reward/mean": 0.310546875, |
| "rewards/simpleverify_reward/std": 0.46317005157470703, |
| "step": 51 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.642578125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4092.0, |
| "completions/mean_length": 3535.466796875, |
| "completions/mean_terminated_length": 2527.732177734375, |
| "completions/min_length": 713.0, |
| "completions/min_terminated_length": 713.0, |
| "epoch": 0.036905606813342796, |
| "frac_reward_zero_std": 0.5, |
| "grad_norm": 0.12625886499881744, |
| "learning_rate": 8.98e-07, |
| "loss": 0.0244, |
| "num_tokens": 102292088.0, |
| "reward": 0.291015625, |
| "reward_std": 0.19838237762451172, |
| "rewards/simpleverify_reward/mean": 0.291015625, |
| "rewards/simpleverify_reward/std": 0.45467492938041687, |
| "step": 52 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.658203125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4082.0, |
| "completions/mean_length": 3512.44921875, |
| "completions/mean_terminated_length": 2388.697021484375, |
| "completions/min_length": 604.0, |
| "completions/min_terminated_length": 604.0, |
| "epoch": 0.03761533002129169, |
| "frac_reward_zero_std": 0.40625, |
| "grad_norm": 0.1417085975408554, |
| "learning_rate": 8.96e-07, |
| "loss": 0.0266, |
| "num_tokens": 104134846.0, |
| "reward": 0.29296875, |
| "reward_std": 0.23736441135406494, |
| "rewards/simpleverify_reward/mean": 0.29296875, |
| "rewards/simpleverify_reward/std": 0.455569326877594, |
| "step": 53 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.61328125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4089.0, |
| "completions/mean_length": 3548.041015625, |
| "completions/mean_terminated_length": 2679.0556640625, |
| "completions/min_length": 730.0, |
| "completions/min_terminated_length": 730.0, |
| "epoch": 0.0383250532292406, |
| "frac_reward_zero_std": 0.5, |
| "grad_norm": 0.13296587765216827, |
| "learning_rate": 8.939999999999999e-07, |
| "loss": 0.028, |
| "num_tokens": 105991827.0, |
| "reward": 0.341796875, |
| "reward_std": 0.19836962223052979, |
| "rewards/simpleverify_reward/mean": 0.341796875, |
| "rewards/simpleverify_reward/std": 0.4747757613658905, |
| "step": 54 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.7265625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4092.0, |
| "completions/mean_length": 3763.9375, |
| "completions/mean_terminated_length": 2881.60009765625, |
| "completions/min_length": 1166.0, |
| "completions/min_terminated_length": 1166.0, |
| "epoch": 0.03903477643718949, |
| "frac_reward_zero_std": 0.46875, |
| "grad_norm": 0.11660730093717575, |
| "learning_rate": 8.92e-07, |
| "loss": 0.0242, |
| "num_tokens": 107974307.0, |
| "reward": 0.263671875, |
| "reward_std": 0.20324639976024628, |
| "rewards/simpleverify_reward/mean": 0.263671875, |
| "rewards/simpleverify_reward/std": 0.4410543739795685, |
| "step": 55 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.697265625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4061.0, |
| "completions/mean_length": 3559.408203125, |
| "completions/mean_terminated_length": 2323.51611328125, |
| "completions/min_length": 557.0, |
| "completions/min_terminated_length": 557.0, |
| "epoch": 0.0397444996451384, |
| "frac_reward_zero_std": 0.46875, |
| "grad_norm": 0.13143253326416016, |
| "learning_rate": 8.9e-07, |
| "loss": 0.0208, |
| "num_tokens": 109838772.0, |
| "reward": 0.255859375, |
| "reward_std": 0.20659980177879333, |
| "rewards/simpleverify_reward/mean": 0.255859375, |
| "rewards/simpleverify_reward/std": 0.43676990270614624, |
| "step": 56 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.8046875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4067.0, |
| "completions/mean_length": 3907.0625, |
| "completions/mean_terminated_length": 3128.639892578125, |
| "completions/min_length": 1984.0, |
| "completions/min_terminated_length": 1984.0, |
| "epoch": 0.04045422285308729, |
| "frac_reward_zero_std": 0.53125, |
| "grad_norm": 0.10804349929094315, |
| "learning_rate": 8.88e-07, |
| "loss": 0.0224, |
| "num_tokens": 111887220.0, |
| "reward": 0.185546875, |
| "reward_std": 0.18597042560577393, |
| "rewards/simpleverify_reward/mean": 0.185546875, |
| "rewards/simpleverify_reward/std": 0.38912075757980347, |
| "step": 57 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.67578125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4080.0, |
| "completions/mean_length": 3622.884765625, |
| "completions/mean_terminated_length": 2636.7529296875, |
| "completions/min_length": 991.0, |
| "completions/min_terminated_length": 991.0, |
| "epoch": 0.0411639460610362, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.1402440220117569, |
| "learning_rate": 8.86e-07, |
| "loss": 0.0252, |
| "num_tokens": 113790841.0, |
| "reward": 0.255859375, |
| "reward_std": 0.2719079852104187, |
| "rewards/simpleverify_reward/mean": 0.255859375, |
| "rewards/simpleverify_reward/std": 0.43676990270614624, |
| "step": 58 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.763671875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4095.0, |
| "completions/mean_length": 3691.001953125, |
| "completions/mean_terminated_length": 2382.2890625, |
| "completions/min_length": 35.0, |
| "completions/min_terminated_length": 35.0, |
| "epoch": 0.04187366926898509, |
| "frac_reward_zero_std": 0.40625, |
| "grad_norm": 0.1322338730096817, |
| "learning_rate": 8.839999999999999e-07, |
| "loss": 0.0215, |
| "num_tokens": 115733242.0, |
| "reward": 0.18359375, |
| "reward_std": 0.2083035707473755, |
| "rewards/simpleverify_reward/mean": 0.18359375, |
| "rewards/simpleverify_reward/std": 0.3875311613082886, |
| "step": 59 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.580078125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4082.0, |
| "completions/mean_length": 3518.625, |
| "completions/mean_terminated_length": 2721.041748046875, |
| "completions/min_length": 1149.0, |
| "completions/min_terminated_length": 1149.0, |
| "epoch": 0.042583392476934, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.14263875782489777, |
| "learning_rate": 8.82e-07, |
| "loss": 0.0199, |
| "num_tokens": 117593786.0, |
| "reward": 0.345703125, |
| "reward_std": 0.30645856261253357, |
| "rewards/simpleverify_reward/mean": 0.345703125, |
| "rewards/simpleverify_reward/std": 0.4760620892047882, |
| "step": 60 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.587890625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 3540.1796875, |
| "completions/mean_terminated_length": 2747.27978515625, |
| "completions/min_length": 911.0, |
| "completions/min_terminated_length": 911.0, |
| "epoch": 0.043293115684882894, |
| "frac_reward_zero_std": 0.4375, |
| "grad_norm": 0.1268397569656372, |
| "learning_rate": 8.799999999999999e-07, |
| "loss": 0.0317, |
| "num_tokens": 119451654.0, |
| "reward": 0.3125, |
| "reward_std": 0.2413923293352127, |
| "rewards/simpleverify_reward/mean": 0.3125, |
| "rewards/simpleverify_reward/std": 0.4639657139778137, |
| "step": 61 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.6796875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4088.0, |
| "completions/mean_length": 3692.986328125, |
| "completions/mean_terminated_length": 2837.810791015625, |
| "completions/min_length": 1315.0, |
| "completions/min_terminated_length": 1315.0, |
| "epoch": 0.0440028388928318, |
| "frac_reward_zero_std": 0.46875, |
| "grad_norm": 0.12313452363014221, |
| "learning_rate": 8.78e-07, |
| "loss": 0.0255, |
| "num_tokens": 121391551.0, |
| "reward": 0.267578125, |
| "reward_std": 0.19513413310050964, |
| "rewards/simpleverify_reward/mean": 0.267578125, |
| "rewards/simpleverify_reward/std": 0.4431293308734894, |
| "step": 62 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.50390625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4095.0, |
| "completions/mean_length": 3233.09375, |
| "completions/mean_terminated_length": 2356.598388671875, |
| "completions/min_length": 868.0, |
| "completions/min_terminated_length": 868.0, |
| "epoch": 0.044712562100780694, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.15140196681022644, |
| "learning_rate": 8.76e-07, |
| "loss": 0.0295, |
| "num_tokens": 123082447.0, |
| "reward": 0.45703125, |
| "reward_std": 0.2732217311859131, |
| "rewards/simpleverify_reward/mean": 0.45703125, |
| "rewards/simpleverify_reward/std": 0.49863746762275696, |
| "step": 63 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.728515625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4084.0, |
| "completions/mean_length": 3691.208984375, |
| "completions/mean_terminated_length": 2604.97119140625, |
| "completions/min_length": 698.0, |
| "completions/min_terminated_length": 698.0, |
| "epoch": 0.0454222853087296, |
| "frac_reward_zero_std": 0.59375, |
| "grad_norm": 0.11032789200544357, |
| "learning_rate": 8.739999999999999e-07, |
| "loss": 0.009, |
| "num_tokens": 125020330.0, |
| "reward": 0.240234375, |
| "reward_std": 0.15764161944389343, |
| "rewards/simpleverify_reward/mean": 0.240234375, |
| "rewards/simpleverify_reward/std": 0.4276435375213623, |
| "step": 64 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.572265625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4072.0, |
| "completions/mean_length": 3524.568359375, |
| "completions/mean_terminated_length": 2760.050048828125, |
| "completions/min_length": 855.0, |
| "completions/min_terminated_length": 855.0, |
| "epoch": 0.046132008516678494, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.14532308280467987, |
| "learning_rate": 8.72e-07, |
| "loss": 0.0191, |
| "num_tokens": 126861613.0, |
| "reward": 0.349609375, |
| "reward_std": 0.2725938856601715, |
| "rewards/simpleverify_reward/mean": 0.349609375, |
| "rewards/simpleverify_reward/std": 0.47731292247772217, |
| "step": 65 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.615234375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 3441.744140625, |
| "completions/mean_terminated_length": 2395.598876953125, |
| "completions/min_length": 93.0, |
| "completions/min_terminated_length": 93.0, |
| "epoch": 0.0468417317246274, |
| "frac_reward_zero_std": 0.375, |
| "grad_norm": 0.14534695446491241, |
| "learning_rate": 8.699999999999999e-07, |
| "loss": 0.0123, |
| "num_tokens": 128671098.0, |
| "reward": 0.328125, |
| "reward_std": 0.2424219250679016, |
| "rewards/simpleverify_reward/mean": 0.328125, |
| "rewards/simpleverify_reward/std": 0.4699897766113281, |
| "step": 66 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.685546875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4026.0, |
| "completions/mean_length": 3645.5390625, |
| "completions/mean_terminated_length": 2663.478271484375, |
| "completions/min_length": 984.0, |
| "completions/min_terminated_length": 984.0, |
| "epoch": 0.047551454932576294, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.1429520696401596, |
| "learning_rate": 8.68e-07, |
| "loss": 0.036, |
| "num_tokens": 130587550.0, |
| "reward": 0.259765625, |
| "reward_std": 0.2727014124393463, |
| "rewards/simpleverify_reward/mean": 0.259765625, |
| "rewards/simpleverify_reward/std": 0.4389347732067108, |
| "step": 67 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.5625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4091.0, |
| "completions/mean_length": 3632.630859375, |
| "completions/mean_terminated_length": 3036.87060546875, |
| "completions/min_length": 1250.0, |
| "completions/min_terminated_length": 1250.0, |
| "epoch": 0.0482611781405252, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.12898331880569458, |
| "learning_rate": 8.659999999999999e-07, |
| "loss": 0.0228, |
| "num_tokens": 132491441.0, |
| "reward": 0.333984375, |
| "reward_std": 0.283235639333725, |
| "rewards/simpleverify_reward/mean": 0.333984375, |
| "rewards/simpleverify_reward/std": 0.47209542989730835, |
| "step": 68 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.685546875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 3672.296875, |
| "completions/mean_terminated_length": 2748.571533203125, |
| "completions/min_length": 711.0, |
| "completions/min_terminated_length": 711.0, |
| "epoch": 0.048970901348474094, |
| "frac_reward_zero_std": 0.40625, |
| "grad_norm": 0.12706109881401062, |
| "learning_rate": 8.639999999999999e-07, |
| "loss": 0.0348, |
| "num_tokens": 134409401.0, |
| "reward": 0.23828125, |
| "reward_std": 0.24831342697143555, |
| "rewards/simpleverify_reward/mean": 0.23828125, |
| "rewards/simpleverify_reward/std": 0.42644867300987244, |
| "step": 69 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.578125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4086.0, |
| "completions/mean_length": 3485.134765625, |
| "completions/mean_terminated_length": 2648.023193359375, |
| "completions/min_length": 989.0, |
| "completions/min_terminated_length": 989.0, |
| "epoch": 0.049680624556423, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.14765338599681854, |
| "learning_rate": 8.62e-07, |
| "loss": 0.0323, |
| "num_tokens": 136248318.0, |
| "reward": 0.373046875, |
| "reward_std": 0.24161046743392944, |
| "rewards/simpleverify_reward/mean": 0.373046875, |
| "rewards/simpleverify_reward/std": 0.48408737778663635, |
| "step": 70 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.6015625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4092.0, |
| "completions/mean_length": 3417.923828125, |
| "completions/mean_terminated_length": 2394.161865234375, |
| "completions/min_length": 861.0, |
| "completions/min_terminated_length": 861.0, |
| "epoch": 0.050390347764371894, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.14964719116687775, |
| "learning_rate": 8.599999999999999e-07, |
| "loss": 0.0401, |
| "num_tokens": 138046279.0, |
| "reward": 0.3359375, |
| "reward_std": 0.28595834970474243, |
| "rewards/simpleverify_reward/mean": 0.3359375, |
| "rewards/simpleverify_reward/std": 0.4727790653705597, |
| "step": 71 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.634765625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4087.0, |
| "completions/mean_length": 3597.6484375, |
| "completions/mean_terminated_length": 2731.529541015625, |
| "completions/min_length": 684.0, |
| "completions/min_terminated_length": 684.0, |
| "epoch": 0.0511000709723208, |
| "frac_reward_zero_std": 0.375, |
| "grad_norm": 0.14101062715053558, |
| "learning_rate": 8.58e-07, |
| "loss": 0.0221, |
| "num_tokens": 139934867.0, |
| "reward": 0.310546875, |
| "reward_std": 0.23631229996681213, |
| "rewards/simpleverify_reward/mean": 0.310546875, |
| "rewards/simpleverify_reward/std": 0.46317005157470703, |
| "step": 72 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.501953125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 3245.390625, |
| "completions/mean_terminated_length": 2388.10986328125, |
| "completions/min_length": 588.0, |
| "completions/min_terminated_length": 588.0, |
| "epoch": 0.051809794180269694, |
| "frac_reward_zero_std": 0.46875, |
| "grad_norm": 0.1243625283241272, |
| "learning_rate": 8.559999999999999e-07, |
| "loss": 0.0239, |
| "num_tokens": 141639019.0, |
| "reward": 0.41015625, |
| "reward_std": 0.19890032708644867, |
| "rewards/simpleverify_reward/mean": 0.41015625, |
| "rewards/simpleverify_reward/std": 0.49234291911125183, |
| "step": 73 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.58984375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4081.0, |
| "completions/mean_length": 3475.8359375, |
| "completions/mean_terminated_length": 2583.98095703125, |
| "completions/min_length": 749.0, |
| "completions/min_terminated_length": 749.0, |
| "epoch": 0.0525195173882186, |
| "frac_reward_zero_std": 0.375, |
| "grad_norm": 0.13435564935207367, |
| "learning_rate": 8.539999999999999e-07, |
| "loss": 0.0306, |
| "num_tokens": 143465911.0, |
| "reward": 0.314453125, |
| "reward_std": 0.27067843079566956, |
| "rewards/simpleverify_reward/mean": 0.314453125, |
| "rewards/simpleverify_reward/std": 0.4647517800331116, |
| "step": 74 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.513671875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4068.0, |
| "completions/mean_length": 3314.13671875, |
| "completions/mean_terminated_length": 2488.313232421875, |
| "completions/min_length": 860.0, |
| "completions/min_terminated_length": 860.0, |
| "epoch": 0.053229240596167494, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.15054352581501007, |
| "learning_rate": 8.52e-07, |
| "loss": 0.0495, |
| "num_tokens": 145203613.0, |
| "reward": 0.412109375, |
| "reward_std": 0.28322935104370117, |
| "rewards/simpleverify_reward/mean": 0.412109375, |
| "rewards/simpleverify_reward/std": 0.49269601702690125, |
| "step": 75 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.59765625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4082.0, |
| "completions/mean_length": 3421.45703125, |
| "completions/mean_terminated_length": 2419.466064453125, |
| "completions/min_length": 658.0, |
| "completions/min_terminated_length": 658.0, |
| "epoch": 0.0539389638041164, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.1501568853855133, |
| "learning_rate": 8.499999999999999e-07, |
| "loss": 0.0295, |
| "num_tokens": 147004119.0, |
| "reward": 0.333984375, |
| "reward_std": 0.2716749310493469, |
| "rewards/simpleverify_reward/mean": 0.333984375, |
| "rewards/simpleverify_reward/std": 0.47209542989730835, |
| "step": 76 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.55078125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4081.0, |
| "completions/mean_length": 3292.57421875, |
| "completions/mean_terminated_length": 2307.504150390625, |
| "completions/min_length": 923.0, |
| "completions/min_terminated_length": 923.0, |
| "epoch": 0.054648687012065295, |
| "frac_reward_zero_std": 0.40625, |
| "grad_norm": 0.14861498773097992, |
| "learning_rate": 8.48e-07, |
| "loss": 0.0273, |
| "num_tokens": 148739581.0, |
| "reward": 0.33984375, |
| "reward_std": 0.23522385954856873, |
| "rewards/simpleverify_reward/mean": 0.33984375, |
| "rewards/simpleverify_reward/std": 0.4741191864013672, |
| "step": 77 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.525390625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4062.0, |
| "completions/mean_length": 3357.7734375, |
| "completions/mean_terminated_length": 2540.5595703125, |
| "completions/min_length": 605.0, |
| "completions/min_terminated_length": 605.0, |
| "epoch": 0.05535841022001419, |
| "frac_reward_zero_std": 0.375, |
| "grad_norm": 0.14416228234767914, |
| "learning_rate": 8.459999999999999e-07, |
| "loss": 0.0225, |
| "num_tokens": 150502569.0, |
| "reward": 0.412109375, |
| "reward_std": 0.23270782828330994, |
| "rewards/simpleverify_reward/mean": 0.412109375, |
| "rewards/simpleverify_reward/std": 0.49269601702690125, |
| "step": 78 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.525390625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4090.0, |
| "completions/mean_length": 3199.4765625, |
| "completions/mean_terminated_length": 2207.02880859375, |
| "completions/min_length": 875.0, |
| "completions/min_terminated_length": 875.0, |
| "epoch": 0.056068133427963095, |
| "frac_reward_zero_std": 0.5, |
| "grad_norm": 0.12976644933223724, |
| "learning_rate": 8.439999999999999e-07, |
| "loss": 0.0258, |
| "num_tokens": 152187485.0, |
| "reward": 0.404296875, |
| "reward_std": 0.17362096905708313, |
| "rewards/simpleverify_reward/mean": 0.404296875, |
| "rewards/simpleverify_reward/std": 0.4912354052066803, |
| "step": 79 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.640625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4096.0, |
| "completions/mean_length": 3498.765625, |
| "completions/mean_terminated_length": 2434.13037109375, |
| "completions/min_length": 952.0, |
| "completions/min_terminated_length": 952.0, |
| "epoch": 0.05677785663591199, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.14961488544940948, |
| "learning_rate": 8.419999999999999e-07, |
| "loss": 0.0325, |
| "num_tokens": 154027957.0, |
| "reward": 0.279296875, |
| "reward_std": 0.2631939649581909, |
| "rewards/simpleverify_reward/mean": 0.279296875, |
| "rewards/simpleverify_reward/std": 0.44909247756004333, |
| "step": 80 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.59375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 3399.11328125, |
| "completions/mean_terminated_length": 2380.586669921875, |
| "completions/min_length": 756.0, |
| "completions/min_terminated_length": 756.0, |
| "epoch": 0.057487579843860895, |
| "frac_reward_zero_std": 0.375, |
| "grad_norm": 0.13716691732406616, |
| "learning_rate": 8.399999999999999e-07, |
| "loss": 0.0436, |
| "num_tokens": 155814991.0, |
| "reward": 0.33984375, |
| "reward_std": 0.24288929998874664, |
| "rewards/simpleverify_reward/mean": 0.33984375, |
| "rewards/simpleverify_reward/std": 0.4741191864013672, |
| "step": 81 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.6875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4078.0, |
| "completions/mean_length": 3703.5546875, |
| "completions/mean_terminated_length": 2840.175048828125, |
| "completions/min_length": 1280.0, |
| "completions/min_terminated_length": 1280.0, |
| "epoch": 0.05819730305180979, |
| "frac_reward_zero_std": 0.375, |
| "grad_norm": 0.1524653434753418, |
| "learning_rate": 8.38e-07, |
| "loss": 0.0146, |
| "num_tokens": 157765035.0, |
| "reward": 0.23828125, |
| "reward_std": 0.24724091589450836, |
| "rewards/simpleverify_reward/mean": 0.23828125, |
| "rewards/simpleverify_reward/std": 0.42644867300987244, |
| "step": 82 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.60546875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 3453.5234375, |
| "completions/mean_terminated_length": 2467.54443359375, |
| "completions/min_length": 707.0, |
| "completions/min_terminated_length": 707.0, |
| "epoch": 0.058907026259758695, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.14475785195827484, |
| "learning_rate": 8.359999999999999e-07, |
| "loss": 0.0404, |
| "num_tokens": 159574327.0, |
| "reward": 0.3515625, |
| "reward_std": 0.2888769507408142, |
| "rewards/simpleverify_reward/mean": 0.3515625, |
| "rewards/simpleverify_reward/std": 0.4779251217842102, |
| "step": 83 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.474609375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4087.0, |
| "completions/mean_length": 3084.685546875, |
| "completions/mean_terminated_length": 2171.118896484375, |
| "completions/min_length": 598.0, |
| "completions/min_terminated_length": 598.0, |
| "epoch": 0.05961674946770759, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.15658195316791534, |
| "learning_rate": 8.34e-07, |
| "loss": 0.0475, |
| "num_tokens": 161199382.0, |
| "reward": 0.365234375, |
| "reward_std": 0.26315879821777344, |
| "rewards/simpleverify_reward/mean": 0.365234375, |
| "rewards/simpleverify_reward/std": 0.4819667339324951, |
| "step": 84 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.669921875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4092.0, |
| "completions/mean_length": 3572.185546875, |
| "completions/mean_terminated_length": 2509.059326171875, |
| "completions/min_length": 970.0, |
| "completions/min_terminated_length": 970.0, |
| "epoch": 0.060326472675656495, |
| "frac_reward_zero_std": 0.40625, |
| "grad_norm": 0.13710473477840424, |
| "learning_rate": 8.319999999999999e-07, |
| "loss": 0.0233, |
| "num_tokens": 163075989.0, |
| "reward": 0.26953125, |
| "reward_std": 0.24437159299850464, |
| "rewards/simpleverify_reward/mean": 0.26953125, |
| "rewards/simpleverify_reward/std": 0.44415023922920227, |
| "step": 85 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.55078125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4093.0, |
| "completions/mean_length": 3351.669921875, |
| "completions/mean_terminated_length": 2439.056396484375, |
| "completions/min_length": 426.0, |
| "completions/min_terminated_length": 426.0, |
| "epoch": 0.06103619588360539, |
| "frac_reward_zero_std": 0.40625, |
| "grad_norm": 0.13624952733516693, |
| "learning_rate": 8.299999999999999e-07, |
| "loss": 0.0139, |
| "num_tokens": 164833180.0, |
| "reward": 0.357421875, |
| "reward_std": 0.25511330366134644, |
| "rewards/simpleverify_reward/mean": 0.357421875, |
| "rewards/simpleverify_reward/std": 0.4797092080116272, |
| "step": 86 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.5234375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4085.0, |
| "completions/mean_length": 3346.98828125, |
| "completions/mean_terminated_length": 2524.30322265625, |
| "completions/min_length": 710.0, |
| "completions/min_terminated_length": 710.0, |
| "epoch": 0.061745919091554295, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.14369390904903412, |
| "learning_rate": 8.28e-07, |
| "loss": 0.0281, |
| "num_tokens": 166603478.0, |
| "reward": 0.3671875, |
| "reward_std": 0.2677637040615082, |
| "rewards/simpleverify_reward/mean": 0.3671875, |
| "rewards/simpleverify_reward/std": 0.48250964283943176, |
| "step": 87 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.376953125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4064.0, |
| "completions/mean_length": 2929.115234375, |
| "completions/mean_terminated_length": 2223.131591796875, |
| "completions/min_length": 496.0, |
| "completions/min_terminated_length": 496.0, |
| "epoch": 0.06245564229950319, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.1705641895532608, |
| "learning_rate": 8.259999999999999e-07, |
| "loss": 0.0199, |
| "num_tokens": 168157169.0, |
| "reward": 0.484375, |
| "reward_std": 0.24600842595100403, |
| "rewards/simpleverify_reward/mean": 0.484375, |
| "rewards/simpleverify_reward/std": 0.5002445578575134, |
| "step": 88 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.525390625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4082.0, |
| "completions/mean_length": 3290.2578125, |
| "completions/mean_terminated_length": 2398.304443359375, |
| "completions/min_length": 44.0, |
| "completions/min_terminated_length": 44.0, |
| "epoch": 0.0631653655074521, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.16099271178245544, |
| "learning_rate": 8.24e-07, |
| "loss": 0.0079, |
| "num_tokens": 169888709.0, |
| "reward": 0.369140625, |
| "reward_std": 0.2765613794326782, |
| "rewards/simpleverify_reward/mean": 0.369140625, |
| "rewards/simpleverify_reward/std": 0.4830440282821655, |
| "step": 89 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.517578125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4093.0, |
| "completions/mean_length": 3118.796875, |
| "completions/mean_terminated_length": 2070.380615234375, |
| "completions/min_length": 876.0, |
| "completions/min_terminated_length": 876.0, |
| "epoch": 0.063875088715401, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.1820123791694641, |
| "learning_rate": 8.219999999999999e-07, |
| "loss": 0.0402, |
| "num_tokens": 171532765.0, |
| "reward": 0.396484375, |
| "reward_std": 0.251827597618103, |
| "rewards/simpleverify_reward/mean": 0.396484375, |
| "rewards/simpleverify_reward/std": 0.4896455705165863, |
| "step": 90 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.46484375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4082.0, |
| "completions/mean_length": 3211.8203125, |
| "completions/mean_terminated_length": 2443.810302734375, |
| "completions/min_length": 486.0, |
| "completions/min_terminated_length": 486.0, |
| "epoch": 0.06458481192334989, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.14464332163333893, |
| "learning_rate": 8.199999999999999e-07, |
| "loss": 0.0224, |
| "num_tokens": 173233073.0, |
| "reward": 0.40625, |
| "reward_std": 0.2866305112838745, |
| "rewards/simpleverify_reward/mean": 0.40625, |
| "rewards/simpleverify_reward/std": 0.49161264300346375, |
| "step": 91 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.556640625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4074.0, |
| "completions/mean_length": 3479.8984375, |
| "completions/mean_terminated_length": 2706.378662109375, |
| "completions/min_length": 738.0, |
| "completions/min_terminated_length": 738.0, |
| "epoch": 0.06529453513129879, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.15506874024868011, |
| "learning_rate": 8.179999999999999e-07, |
| "loss": 0.0234, |
| "num_tokens": 175064925.0, |
| "reward": 0.328125, |
| "reward_std": 0.2913217544555664, |
| "rewards/simpleverify_reward/mean": 0.328125, |
| "rewards/simpleverify_reward/std": 0.4699897766113281, |
| "step": 92 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.3671875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4078.0, |
| "completions/mean_length": 2936.869140625, |
| "completions/mean_terminated_length": 2264.287109375, |
| "completions/min_length": 528.0, |
| "completions/min_terminated_length": 528.0, |
| "epoch": 0.0660042583392477, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.16830310225486755, |
| "learning_rate": 8.159999999999999e-07, |
| "loss": 0.0248, |
| "num_tokens": 176609306.0, |
| "reward": 0.541015625, |
| "reward_std": 0.26065725088119507, |
| "rewards/simpleverify_reward/mean": 0.541015625, |
| "rewards/simpleverify_reward/std": 0.49880221486091614, |
| "step": 93 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.486328125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4095.0, |
| "completions/mean_length": 3244.033203125, |
| "completions/mean_terminated_length": 2437.418212890625, |
| "completions/min_length": 968.0, |
| "completions/min_terminated_length": 968.0, |
| "epoch": 0.0667139815471966, |
| "frac_reward_zero_std": 0.375, |
| "grad_norm": 0.16703960299491882, |
| "learning_rate": 8.14e-07, |
| "loss": 0.0109, |
| "num_tokens": 178307611.0, |
| "reward": 0.396484375, |
| "reward_std": 0.24155621230602264, |
| "rewards/simpleverify_reward/mean": 0.396484375, |
| "rewards/simpleverify_reward/std": 0.4896455705165863, |
| "step": 94 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.482421875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4087.0, |
| "completions/mean_length": 3195.640625, |
| "completions/mean_terminated_length": 2356.437744140625, |
| "completions/min_length": 702.0, |
| "completions/min_terminated_length": 702.0, |
| "epoch": 0.06742370475514549, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.1675298810005188, |
| "learning_rate": 8.12e-07, |
| "loss": 0.0124, |
| "num_tokens": 179988291.0, |
| "reward": 0.390625, |
| "reward_std": 0.2485167682170868, |
| "rewards/simpleverify_reward/mean": 0.390625, |
| "rewards/simpleverify_reward/std": 0.48836761713027954, |
| "step": 95 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.466796875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4076.0, |
| "completions/mean_length": 3258.248046875, |
| "completions/mean_terminated_length": 2524.83154296875, |
| "completions/min_length": 693.0, |
| "completions/min_terminated_length": 693.0, |
| "epoch": 0.06813342796309439, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.15315528213977814, |
| "learning_rate": 8.1e-07, |
| "loss": 0.0298, |
| "num_tokens": 181711650.0, |
| "reward": 0.421875, |
| "reward_std": 0.2681019902229309, |
| "rewards/simpleverify_reward/mean": 0.421875, |
| "rewards/simpleverify_reward/std": 0.49434176087379456, |
| "step": 96 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.447265625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4096.0, |
| "completions/mean_length": 3138.087890625, |
| "completions/mean_terminated_length": 2362.95751953125, |
| "completions/min_length": 520.0, |
| "completions/min_terminated_length": 520.0, |
| "epoch": 0.0688431511710433, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.16633346676826477, |
| "learning_rate": 8.08e-07, |
| "loss": 0.0512, |
| "num_tokens": 183360303.0, |
| "reward": 0.392578125, |
| "reward_std": 0.28456395864486694, |
| "rewards/simpleverify_reward/mean": 0.392578125, |
| "rewards/simpleverify_reward/std": 0.4888018071651459, |
| "step": 97 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.41796875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4058.0, |
| "completions/mean_length": 3016.6875, |
| "completions/mean_terminated_length": 2241.61083984375, |
| "completions/min_length": 639.0, |
| "completions/min_terminated_length": 639.0, |
| "epoch": 0.0695528743789922, |
| "frac_reward_zero_std": 0.375, |
| "grad_norm": 0.15453054010868073, |
| "learning_rate": 8.06e-07, |
| "loss": 0.004, |
| "num_tokens": 184946223.0, |
| "reward": 0.37890625, |
| "reward_std": 0.26614123582839966, |
| "rewards/simpleverify_reward/mean": 0.37890625, |
| "rewards/simpleverify_reward/std": 0.4855891764163971, |
| "step": 98 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.419921875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4084.0, |
| "completions/mean_length": 3211.828125, |
| "completions/mean_terminated_length": 2571.77099609375, |
| "completions/min_length": 670.0, |
| "completions/min_terminated_length": 670.0, |
| "epoch": 0.07026259758694109, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.16378699243068695, |
| "learning_rate": 8.04e-07, |
| "loss": 0.0297, |
| "num_tokens": 186627719.0, |
| "reward": 0.4453125, |
| "reward_std": 0.3155153691768646, |
| "rewards/simpleverify_reward/mean": 0.4453125, |
| "rewards/simpleverify_reward/std": 0.49748632311820984, |
| "step": 99 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4021.0, |
| "completions/mean_length": 2843.70703125, |
| "completions/mean_terminated_length": 2092.331298828125, |
| "completions/min_length": 556.0, |
| "completions/min_terminated_length": 556.0, |
| "epoch": 0.07097232079488999, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.15830565989017487, |
| "learning_rate": 8.02e-07, |
| "loss": 0.0046, |
| "num_tokens": 188125825.0, |
| "reward": 0.419921875, |
| "reward_std": 0.27873051166534424, |
| "rewards/simpleverify_reward/mean": 0.419921875, |
| "rewards/simpleverify_reward/std": 0.4940285086631775, |
| "step": 100 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.50390625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4090.0, |
| "completions/mean_length": 3203.998046875, |
| "completions/mean_terminated_length": 2297.94873046875, |
| "completions/min_length": 223.0, |
| "completions/min_terminated_length": 223.0, |
| "epoch": 0.0716820440028389, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.14174576103687286, |
| "learning_rate": 8e-07, |
| "loss": 0.0423, |
| "num_tokens": 189818896.0, |
| "reward": 0.3671875, |
| "reward_std": 0.2969106137752533, |
| "rewards/simpleverify_reward/mean": 0.3671875, |
| "rewards/simpleverify_reward/std": 0.48250964283943176, |
| "step": 101 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.404296875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4084.0, |
| "completions/mean_length": 2994.482421875, |
| "completions/mean_terminated_length": 2246.89501953125, |
| "completions/min_length": 45.0, |
| "completions/min_terminated_length": 45.0, |
| "epoch": 0.0723917672107878, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.17433840036392212, |
| "learning_rate": 7.98e-07, |
| "loss": 0.044, |
| "num_tokens": 191395479.0, |
| "reward": 0.517578125, |
| "reward_std": 0.27554452419281006, |
| "rewards/simpleverify_reward/mean": 0.517578125, |
| "rewards/simpleverify_reward/std": 0.5001795887947083, |
| "step": 102 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.369140625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4088.0, |
| "completions/mean_length": 2808.177734375, |
| "completions/mean_terminated_length": 2054.622314453125, |
| "completions/min_length": 766.0, |
| "completions/min_terminated_length": 766.0, |
| "epoch": 0.07310149041873669, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.17473237216472626, |
| "learning_rate": 7.96e-07, |
| "loss": 0.0562, |
| "num_tokens": 192869970.0, |
| "reward": 0.521484375, |
| "reward_std": 0.2670045793056488, |
| "rewards/simpleverify_reward/mean": 0.521484375, |
| "rewards/simpleverify_reward/std": 0.5000267624855042, |
| "step": 103 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.40625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4090.0, |
| "completions/mean_length": 3121.095703125, |
| "completions/mean_terminated_length": 2454.055908203125, |
| "completions/min_length": 517.0, |
| "completions/min_terminated_length": 517.0, |
| "epoch": 0.07381121362668559, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.1645442545413971, |
| "learning_rate": 7.94e-07, |
| "loss": 0.0394, |
| "num_tokens": 194511155.0, |
| "reward": 0.451171875, |
| "reward_std": 0.31630560755729675, |
| "rewards/simpleverify_reward/mean": 0.451171875, |
| "rewards/simpleverify_reward/std": 0.498096764087677, |
| "step": 104 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.388671875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4087.0, |
| "completions/mean_length": 2946.859375, |
| "completions/mean_terminated_length": 2216.255615234375, |
| "completions/min_length": 645.0, |
| "completions/min_terminated_length": 645.0, |
| "epoch": 0.0745209368346345, |
| "frac_reward_zero_std": 0.375, |
| "grad_norm": 0.16051265597343445, |
| "learning_rate": 7.92e-07, |
| "loss": 0.0444, |
| "num_tokens": 196069499.0, |
| "reward": 0.533203125, |
| "reward_std": 0.24437262117862701, |
| "rewards/simpleverify_reward/mean": 0.533203125, |
| "rewards/simpleverify_reward/std": 0.4993842542171478, |
| "step": 105 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.466796875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4080.0, |
| "completions/mean_length": 3163.041015625, |
| "completions/mean_terminated_length": 2346.274658203125, |
| "completions/min_length": 780.0, |
| "completions/min_terminated_length": 780.0, |
| "epoch": 0.07523066004258339, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.16436053812503815, |
| "learning_rate": 7.9e-07, |
| "loss": 0.0453, |
| "num_tokens": 197730528.0, |
| "reward": 0.404296875, |
| "reward_std": 0.2917400002479553, |
| "rewards/simpleverify_reward/mean": 0.404296875, |
| "rewards/simpleverify_reward/std": 0.4912354052066803, |
| "step": 106 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.296875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4083.0, |
| "completions/mean_length": 2787.265625, |
| "completions/mean_terminated_length": 2234.68896484375, |
| "completions/min_length": 834.0, |
| "completions/min_terminated_length": 834.0, |
| "epoch": 0.07594038325053229, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.19523406028747559, |
| "learning_rate": 7.88e-07, |
| "loss": 0.0305, |
| "num_tokens": 199204536.0, |
| "reward": 0.5078125, |
| "reward_std": 0.3203083872795105, |
| "rewards/simpleverify_reward/mean": 0.5078125, |
| "rewards/simpleverify_reward/std": 0.5004279017448425, |
| "step": 107 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.458984375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4090.0, |
| "completions/mean_length": 3188.09375, |
| "completions/mean_terminated_length": 2417.848388671875, |
| "completions/min_length": 580.0, |
| "completions/min_terminated_length": 580.0, |
| "epoch": 0.0766501064584812, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.15973350405693054, |
| "learning_rate": 7.86e-07, |
| "loss": 0.0298, |
| "num_tokens": 200880856.0, |
| "reward": 0.396484375, |
| "reward_std": 0.32139918208122253, |
| "rewards/simpleverify_reward/mean": 0.396484375, |
| "rewards/simpleverify_reward/std": 0.4896455705165863, |
| "step": 108 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.54296875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4096.0, |
| "completions/mean_length": 3346.439453125, |
| "completions/mean_terminated_length": 2455.93603515625, |
| "completions/min_length": 648.0, |
| "completions/min_terminated_length": 648.0, |
| "epoch": 0.0773598296664301, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.15497760474681854, |
| "learning_rate": 7.84e-07, |
| "loss": 0.0571, |
| "num_tokens": 202639657.0, |
| "reward": 0.419921875, |
| "reward_std": 0.30249959230422974, |
| "rewards/simpleverify_reward/mean": 0.419921875, |
| "rewards/simpleverify_reward/std": 0.4940285086631775, |
| "step": 109 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.404296875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4096.0, |
| "completions/mean_length": 3191.220703125, |
| "completions/mean_terminated_length": 2577.157470703125, |
| "completions/min_length": 755.0, |
| "completions/min_terminated_length": 755.0, |
| "epoch": 0.07806955287437899, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.15472838282585144, |
| "learning_rate": 7.82e-07, |
| "loss": 0.0303, |
| "num_tokens": 204320554.0, |
| "reward": 0.494140625, |
| "reward_std": 0.3083037734031677, |
| "rewards/simpleverify_reward/mean": 0.494140625, |
| "rewards/simpleverify_reward/std": 0.5004546642303467, |
| "step": 110 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4084.0, |
| "completions/mean_length": 3123.916015625, |
| "completions/mean_terminated_length": 2540.665771484375, |
| "completions/min_length": 789.0, |
| "completions/min_terminated_length": 789.0, |
| "epoch": 0.07877927608232789, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.17517930269241333, |
| "learning_rate": 7.799999999999999e-07, |
| "loss": 0.0438, |
| "num_tokens": 205964303.0, |
| "reward": 0.51171875, |
| "reward_std": 0.3467816114425659, |
| "rewards/simpleverify_reward/mean": 0.51171875, |
| "rewards/simpleverify_reward/std": 0.5003514885902405, |
| "step": 111 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.333984375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4095.0, |
| "completions/mean_length": 2957.005859375, |
| "completions/mean_terminated_length": 2385.838623046875, |
| "completions/min_length": 723.0, |
| "completions/min_terminated_length": 723.0, |
| "epoch": 0.0794889992902768, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.16625703871250153, |
| "learning_rate": 7.78e-07, |
| "loss": 0.0445, |
| "num_tokens": 207520338.0, |
| "reward": 0.53515625, |
| "reward_std": 0.34160950779914856, |
| "rewards/simpleverify_reward/mean": 0.53515625, |
| "rewards/simpleverify_reward/std": 0.49925029277801514, |
| "step": 112 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.361328125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 3972.0, |
| "completions/mean_length": 2845.6328125, |
| "completions/mean_terminated_length": 2138.238525390625, |
| "completions/min_length": 708.0, |
| "completions/min_terminated_length": 708.0, |
| "epoch": 0.0801987224982257, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.15262147784233093, |
| "learning_rate": 7.76e-07, |
| "loss": 0.0429, |
| "num_tokens": 209031462.0, |
| "reward": 0.501953125, |
| "reward_std": 0.2475665807723999, |
| "rewards/simpleverify_reward/mean": 0.501953125, |
| "rewards/simpleverify_reward/std": 0.5004851818084717, |
| "step": 113 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.349609375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4078.0, |
| "completions/mean_length": 3088.0546875, |
| "completions/mean_terminated_length": 2546.246337890625, |
| "completions/min_length": 828.0, |
| "completions/min_terminated_length": 828.0, |
| "epoch": 0.08090844570617459, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.16556212306022644, |
| "learning_rate": 7.74e-07, |
| "loss": 0.0437, |
| "num_tokens": 210657362.0, |
| "reward": 0.572265625, |
| "reward_std": 0.30386435985565186, |
| "rewards/simpleverify_reward/mean": 0.572265625, |
| "rewards/simpleverify_reward/std": 0.4952339828014374, |
| "step": 114 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.392578125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4093.0, |
| "completions/mean_length": 2965.228515625, |
| "completions/mean_terminated_length": 2234.408203125, |
| "completions/min_length": 864.0, |
| "completions/min_terminated_length": 864.0, |
| "epoch": 0.08161816891412349, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.16230656206607819, |
| "learning_rate": 7.72e-07, |
| "loss": 0.0373, |
| "num_tokens": 212220167.0, |
| "reward": 0.470703125, |
| "reward_std": 0.2588341236114502, |
| "rewards/simpleverify_reward/mean": 0.470703125, |
| "rewards/simpleverify_reward/std": 0.49962911009788513, |
| "step": 115 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.2421875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4072.0, |
| "completions/mean_length": 2611.724609375, |
| "completions/mean_terminated_length": 2137.368408203125, |
| "completions/min_length": 579.0, |
| "completions/min_terminated_length": 579.0, |
| "epoch": 0.0823278921220724, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.19717451930046082, |
| "learning_rate": 7.699999999999999e-07, |
| "loss": 0.0425, |
| "num_tokens": 213599290.0, |
| "reward": 0.62109375, |
| "reward_std": 0.29221269488334656, |
| "rewards/simpleverify_reward/mean": 0.62109375, |
| "rewards/simpleverify_reward/std": 0.4855891764163971, |
| "step": 116 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.3359375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4081.0, |
| "completions/mean_length": 3072.814453125, |
| "completions/mean_terminated_length": 2555.202880859375, |
| "completions/min_length": 760.0, |
| "completions/min_terminated_length": 760.0, |
| "epoch": 0.0830376153300213, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.1594112515449524, |
| "learning_rate": 7.68e-07, |
| "loss": 0.0242, |
| "num_tokens": 215214091.0, |
| "reward": 0.541015625, |
| "reward_std": 0.3178325295448303, |
| "rewards/simpleverify_reward/mean": 0.541015625, |
| "rewards/simpleverify_reward/std": 0.49880221486091614, |
| "step": 117 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.408203125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4095.0, |
| "completions/mean_length": 3111.041015625, |
| "completions/mean_terminated_length": 2431.64697265625, |
| "completions/min_length": 690.0, |
| "completions/min_terminated_length": 690.0, |
| "epoch": 0.08374733853797019, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.16258352994918823, |
| "learning_rate": 7.66e-07, |
| "loss": 0.0425, |
| "num_tokens": 216846896.0, |
| "reward": 0.427734375, |
| "reward_std": 0.326577752828598, |
| "rewards/simpleverify_reward/mean": 0.427734375, |
| "rewards/simpleverify_reward/std": 0.4952339828014374, |
| "step": 118 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.275390625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 2811.35546875, |
| "completions/mean_terminated_length": 2323.121337890625, |
| "completions/min_length": 564.0, |
| "completions/min_terminated_length": 564.0, |
| "epoch": 0.08445706174591909, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.16988536715507507, |
| "learning_rate": 7.64e-07, |
| "loss": 0.044, |
| "num_tokens": 218331798.0, |
| "reward": 0.53515625, |
| "reward_std": 0.3496755063533783, |
| "rewards/simpleverify_reward/mean": 0.53515625, |
| "rewards/simpleverify_reward/std": 0.49925029277801514, |
| "step": 119 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.326171875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4096.0, |
| "completions/mean_length": 3152.51171875, |
| "completions/mean_terminated_length": 2695.808837890625, |
| "completions/min_length": 875.0, |
| "completions/min_terminated_length": 875.0, |
| "epoch": 0.085166784953868, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.1705336719751358, |
| "learning_rate": 7.62e-07, |
| "loss": 0.0309, |
| "num_tokens": 219990252.0, |
| "reward": 0.513671875, |
| "reward_std": 0.34677326679229736, |
| "rewards/simpleverify_reward/mean": 0.513671875, |
| "rewards/simpleverify_reward/std": 0.5003018379211426, |
| "step": 120 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.345703125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 2998.013671875, |
| "completions/mean_terminated_length": 2417.883544921875, |
| "completions/min_length": 463.0, |
| "completions/min_terminated_length": 463.0, |
| "epoch": 0.0858765081618169, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.18733693659305573, |
| "learning_rate": 7.599999999999999e-07, |
| "loss": 0.0505, |
| "num_tokens": 221566339.0, |
| "reward": 0.501953125, |
| "reward_std": 0.3544197976589203, |
| "rewards/simpleverify_reward/mean": 0.501953125, |
| "rewards/simpleverify_reward/std": 0.5004851818084717, |
| "step": 121 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.291015625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4074.0, |
| "completions/mean_length": 2790.142578125, |
| "completions/mean_terminated_length": 2254.12939453125, |
| "completions/min_length": 596.0, |
| "completions/min_terminated_length": 596.0, |
| "epoch": 0.08658623136976579, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.19825592637062073, |
| "learning_rate": 7.58e-07, |
| "loss": 0.0289, |
| "num_tokens": 223037452.0, |
| "reward": 0.5625, |
| "reward_std": 0.31943148374557495, |
| "rewards/simpleverify_reward/mean": 0.5625, |
| "rewards/simpleverify_reward/std": 0.49656352400779724, |
| "step": 122 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.265625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4083.0, |
| "completions/mean_length": 2775.15234375, |
| "completions/mean_terminated_length": 2297.39892578125, |
| "completions/min_length": 715.0, |
| "completions/min_terminated_length": 715.0, |
| "epoch": 0.08729595457771469, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.18365372717380524, |
| "learning_rate": 7.559999999999999e-07, |
| "loss": 0.0167, |
| "num_tokens": 224514554.0, |
| "reward": 0.552734375, |
| "reward_std": 0.3342917263507843, |
| "rewards/simpleverify_reward/mean": 0.552734375, |
| "rewards/simpleverify_reward/std": 0.4976975917816162, |
| "step": 123 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.240234375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4096.0, |
| "completions/mean_length": 2583.75, |
| "completions/mean_terminated_length": 2105.58349609375, |
| "completions/min_length": 553.0, |
| "completions/min_terminated_length": 553.0, |
| "epoch": 0.0880056777856636, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.20780374109745026, |
| "learning_rate": 7.54e-07, |
| "loss": 0.0447, |
| "num_tokens": 225879834.0, |
| "reward": 0.513671875, |
| "reward_std": 0.3267039656639099, |
| "rewards/simpleverify_reward/mean": 0.513671875, |
| "rewards/simpleverify_reward/std": 0.5003018379211426, |
| "step": 124 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.357421875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4089.0, |
| "completions/mean_length": 3052.068359375, |
| "completions/mean_terminated_length": 2471.4013671875, |
| "completions/min_length": 803.0, |
| "completions/min_terminated_length": 803.0, |
| "epoch": 0.0887154009936125, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.1777561753988266, |
| "learning_rate": 7.52e-07, |
| "loss": 0.0652, |
| "num_tokens": 227486765.0, |
| "reward": 0.51171875, |
| "reward_std": 0.36513233184814453, |
| "rewards/simpleverify_reward/mean": 0.51171875, |
| "rewards/simpleverify_reward/std": 0.5003514885902405, |
| "step": 125 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.2109375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4059.0, |
| "completions/mean_length": 2696.140625, |
| "completions/mean_terminated_length": 2321.920654296875, |
| "completions/min_length": 700.0, |
| "completions/min_terminated_length": 700.0, |
| "epoch": 0.08942512420156139, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.1802533119916916, |
| "learning_rate": 7.5e-07, |
| "loss": 0.0381, |
| "num_tokens": 228908741.0, |
| "reward": 0.603515625, |
| "reward_std": 0.33434486389160156, |
| "rewards/simpleverify_reward/mean": 0.603515625, |
| "rewards/simpleverify_reward/std": 0.4896455705165863, |
| "step": 126 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.259765625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4093.0, |
| "completions/mean_length": 2583.85546875, |
| "completions/mean_terminated_length": 2053.20849609375, |
| "completions/min_length": 687.0, |
| "completions/min_terminated_length": 687.0, |
| "epoch": 0.09013484740951029, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.17590190470218658, |
| "learning_rate": 7.48e-07, |
| "loss": -0.0011, |
| "num_tokens": 230285851.0, |
| "reward": 0.478515625, |
| "reward_std": 0.2783436179161072, |
| "rewards/simpleverify_reward/mean": 0.478515625, |
| "rewards/simpleverify_reward/std": 0.5000267624855042, |
| "step": 127 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.283203125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4086.0, |
| "completions/mean_length": 2705.388671875, |
| "completions/mean_terminated_length": 2155.964599609375, |
| "completions/min_length": 445.0, |
| "completions/min_terminated_length": 445.0, |
| "epoch": 0.0908445706174592, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.18793454766273499, |
| "learning_rate": 7.459999999999999e-07, |
| "loss": 0.051, |
| "num_tokens": 231714082.0, |
| "reward": 0.59375, |
| "reward_std": 0.338156521320343, |
| "rewards/simpleverify_reward/mean": 0.59375, |
| "rewards/simpleverify_reward/std": 0.49161264300346375, |
| "step": 128 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.3828125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4080.0, |
| "completions/mean_length": 3024.951171875, |
| "completions/mean_terminated_length": 2360.6298828125, |
| "completions/min_length": 578.0, |
| "completions/min_terminated_length": 578.0, |
| "epoch": 0.09155429382540808, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.1606791615486145, |
| "learning_rate": 7.44e-07, |
| "loss": 0.0385, |
| "num_tokens": 233313961.0, |
| "reward": 0.4609375, |
| "reward_std": 0.34139925241470337, |
| "rewards/simpleverify_reward/mean": 0.4609375, |
| "rewards/simpleverify_reward/std": 0.4989593029022217, |
| "step": 129 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.23828125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4084.0, |
| "completions/mean_length": 2674.212890625, |
| "completions/mean_terminated_length": 2229.44873046875, |
| "completions/min_length": 645.0, |
| "completions/min_terminated_length": 645.0, |
| "epoch": 0.09226401703335699, |
| "frac_reward_zero_std": 0.125, |
| "grad_norm": 0.19755835831165314, |
| "learning_rate": 7.42e-07, |
| "loss": 0.033, |
| "num_tokens": 234727318.0, |
| "reward": 0.53125, |
| "reward_std": 0.3984617292881012, |
| "rewards/simpleverify_reward/mean": 0.53125, |
| "rewards/simpleverify_reward/std": 0.4995105266571045, |
| "step": 130 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.212890625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4089.0, |
| "completions/mean_length": 2622.28125, |
| "completions/mean_terminated_length": 2223.682373046875, |
| "completions/min_length": 479.0, |
| "completions/min_terminated_length": 479.0, |
| "epoch": 0.09297374024130589, |
| "frac_reward_zero_std": 0.09375, |
| "grad_norm": 0.20060239732265472, |
| "learning_rate": 7.4e-07, |
| "loss": 0.0216, |
| "num_tokens": 236109766.0, |
| "reward": 0.51171875, |
| "reward_std": 0.39317673444747925, |
| "rewards/simpleverify_reward/mean": 0.51171875, |
| "rewards/simpleverify_reward/std": 0.5003514885902405, |
| "step": 131 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.279296875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4092.0, |
| "completions/mean_length": 2800.283203125, |
| "completions/mean_terminated_length": 2298.149169921875, |
| "completions/min_length": 650.0, |
| "completions/min_terminated_length": 650.0, |
| "epoch": 0.0936834634492548, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.19882658123970032, |
| "learning_rate": 7.38e-07, |
| "loss": 0.0465, |
| "num_tokens": 237590871.0, |
| "reward": 0.59765625, |
| "reward_std": 0.3287273049354553, |
| "rewards/simpleverify_reward/mean": 0.59765625, |
| "rewards/simpleverify_reward/std": 0.4908501207828522, |
| "step": 132 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.345703125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4085.0, |
| "completions/mean_length": 2933.1484375, |
| "completions/mean_terminated_length": 2318.746337890625, |
| "completions/min_length": 351.0, |
| "completions/min_terminated_length": 351.0, |
| "epoch": 0.09439318665720368, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.17065875232219696, |
| "learning_rate": 7.359999999999999e-07, |
| "loss": 0.0355, |
| "num_tokens": 239143315.0, |
| "reward": 0.43359375, |
| "reward_std": 0.2977556884288788, |
| "rewards/simpleverify_reward/mean": 0.43359375, |
| "rewards/simpleverify_reward/std": 0.4960552453994751, |
| "step": 133 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.294921875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4082.0, |
| "completions/mean_length": 2796.146484375, |
| "completions/mean_terminated_length": 2252.4404296875, |
| "completions/min_length": 530.0, |
| "completions/min_terminated_length": 530.0, |
| "epoch": 0.09510290986515259, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.18555350601673126, |
| "learning_rate": 7.34e-07, |
| "loss": 0.0384, |
| "num_tokens": 240628222.0, |
| "reward": 0.490234375, |
| "reward_std": 0.3166831135749817, |
| "rewards/simpleverify_reward/mean": 0.490234375, |
| "rewards/simpleverify_reward/std": 0.5003935098648071, |
| "step": 134 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.30859375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4093.0, |
| "completions/mean_length": 2909.833984375, |
| "completions/mean_terminated_length": 2380.415283203125, |
| "completions/min_length": 654.0, |
| "completions/min_terminated_length": 654.0, |
| "epoch": 0.09581263307310149, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.20278164744377136, |
| "learning_rate": 7.319999999999999e-07, |
| "loss": 0.0411, |
| "num_tokens": 242170985.0, |
| "reward": 0.462890625, |
| "reward_std": 0.34776583313941956, |
| "rewards/simpleverify_reward/mean": 0.462890625, |
| "rewards/simpleverify_reward/std": 0.4991086423397064, |
| "step": 135 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.52734375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4083.0, |
| "completions/mean_length": 3346.70703125, |
| "completions/mean_terminated_length": 2510.718994140625, |
| "completions/min_length": 810.0, |
| "completions/min_terminated_length": 810.0, |
| "epoch": 0.0965223562810504, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.15448293089866638, |
| "learning_rate": 7.3e-07, |
| "loss": 0.0411, |
| "num_tokens": 243930195.0, |
| "reward": 0.33203125, |
| "reward_std": 0.3065604269504547, |
| "rewards/simpleverify_reward/mean": 0.33203125, |
| "rewards/simpleverify_reward/std": 0.47140273451805115, |
| "step": 136 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.37109375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4092.0, |
| "completions/mean_length": 3009.916015625, |
| "completions/mean_terminated_length": 2369.05908203125, |
| "completions/min_length": 614.0, |
| "completions/min_terminated_length": 614.0, |
| "epoch": 0.09723207948899928, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.1710393875837326, |
| "learning_rate": 7.28e-07, |
| "loss": 0.0521, |
| "num_tokens": 245523608.0, |
| "reward": 0.431640625, |
| "reward_std": 0.33728110790252686, |
| "rewards/simpleverify_reward/mean": 0.431640625, |
| "rewards/simpleverify_reward/std": 0.4957893490791321, |
| "step": 137 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.35546875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4093.0, |
| "completions/mean_length": 3135.19140625, |
| "completions/mean_terminated_length": 2605.290771484375, |
| "completions/min_length": 621.0, |
| "completions/min_terminated_length": 621.0, |
| "epoch": 0.09794180269694819, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.17173364758491516, |
| "learning_rate": 7.259999999999999e-07, |
| "loss": 0.046, |
| "num_tokens": 247177930.0, |
| "reward": 0.4921875, |
| "reward_std": 0.3344770669937134, |
| "rewards/simpleverify_reward/mean": 0.4921875, |
| "rewards/simpleverify_reward/std": 0.5004279017448425, |
| "step": 138 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4051.0, |
| "completions/mean_length": 2993.515625, |
| "completions/mean_terminated_length": 2332.025146484375, |
| "completions/min_length": 473.0, |
| "completions/min_terminated_length": 473.0, |
| "epoch": 0.09865152590489709, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.17654848098754883, |
| "learning_rate": 7.24e-07, |
| "loss": 0.0151, |
| "num_tokens": 248758306.0, |
| "reward": 0.390625, |
| "reward_std": 0.3386094272136688, |
| "rewards/simpleverify_reward/mean": 0.390625, |
| "rewards/simpleverify_reward/std": 0.48836761713027954, |
| "step": 139 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.25390625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 2618.998046875, |
| "completions/mean_terminated_length": 2116.353515625, |
| "completions/min_length": 34.0, |
| "completions/min_terminated_length": 34.0, |
| "epoch": 0.099361249112846, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.20156492292881012, |
| "learning_rate": 7.219999999999999e-07, |
| "loss": 0.0655, |
| "num_tokens": 250149953.0, |
| "reward": 0.54296875, |
| "reward_std": 0.32388830184936523, |
| "rewards/simpleverify_reward/mean": 0.54296875, |
| "rewards/simpleverify_reward/std": 0.49863746762275696, |
| "step": 140 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.31640625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4075.0, |
| "completions/mean_length": 3052.642578125, |
| "completions/mean_terminated_length": 2569.717041015625, |
| "completions/min_length": 657.0, |
| "completions/min_terminated_length": 657.0, |
| "epoch": 0.10007097232079488, |
| "frac_reward_zero_std": 0.09375, |
| "grad_norm": 0.18135222792625427, |
| "learning_rate": 7.2e-07, |
| "loss": 0.0575, |
| "num_tokens": 251760826.0, |
| "reward": 0.474609375, |
| "reward_std": 0.3774464726448059, |
| "rewards/simpleverify_reward/mean": 0.474609375, |
| "rewards/simpleverify_reward/std": 0.4998432695865631, |
| "step": 141 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.294921875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4078.0, |
| "completions/mean_length": 2642.79296875, |
| "completions/mean_terminated_length": 2034.9417724609375, |
| "completions/min_length": 448.0, |
| "completions/min_terminated_length": 448.0, |
| "epoch": 0.10078069552874379, |
| "frac_reward_zero_std": 0.375, |
| "grad_norm": 0.16045616567134857, |
| "learning_rate": 7.179999999999999e-07, |
| "loss": 0.0112, |
| "num_tokens": 253159136.0, |
| "reward": 0.48828125, |
| "reward_std": 0.2591387927532196, |
| "rewards/simpleverify_reward/mean": 0.48828125, |
| "rewards/simpleverify_reward/std": 0.5003514885902405, |
| "step": 142 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.3203125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4063.0, |
| "completions/mean_length": 2769.447265625, |
| "completions/mean_terminated_length": 2144.290283203125, |
| "completions/min_length": 525.0, |
| "completions/min_terminated_length": 525.0, |
| "epoch": 0.10149041873669269, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.18575704097747803, |
| "learning_rate": 7.159999999999999e-07, |
| "loss": 0.0615, |
| "num_tokens": 254623765.0, |
| "reward": 0.53125, |
| "reward_std": 0.31973132491111755, |
| "rewards/simpleverify_reward/mean": 0.53125, |
| "rewards/simpleverify_reward/std": 0.4995105266571045, |
| "step": 143 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.232421875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4052.0, |
| "completions/mean_length": 2472.5703125, |
| "completions/mean_terminated_length": 1980.9974365234375, |
| "completions/min_length": 656.0, |
| "completions/min_terminated_length": 656.0, |
| "epoch": 0.1022001419446416, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.19655199348926544, |
| "learning_rate": 7.14e-07, |
| "loss": 0.0139, |
| "num_tokens": 255932073.0, |
| "reward": 0.537109375, |
| "reward_std": 0.2763058841228485, |
| "rewards/simpleverify_reward/mean": 0.537109375, |
| "rewards/simpleverify_reward/std": 0.4991086423397064, |
| "step": 144 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.1953125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4056.0, |
| "completions/mean_length": 2539.6796875, |
| "completions/mean_terminated_length": 2161.93212890625, |
| "completions/min_length": 629.0, |
| "completions/min_terminated_length": 629.0, |
| "epoch": 0.10290986515259049, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.18420714139938354, |
| "learning_rate": 7.119999999999999e-07, |
| "loss": 0.0195, |
| "num_tokens": 257274789.0, |
| "reward": 0.560546875, |
| "reward_std": 0.32961738109588623, |
| "rewards/simpleverify_reward/mean": 0.560546875, |
| "rewards/simpleverify_reward/std": 0.49680593609809875, |
| "step": 145 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.203125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4095.0, |
| "completions/mean_length": 2650.40625, |
| "completions/mean_terminated_length": 2281.921630859375, |
| "completions/min_length": 530.0, |
| "completions/min_terminated_length": 530.0, |
| "epoch": 0.10361958836053939, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.1894926279783249, |
| "learning_rate": 7.1e-07, |
| "loss": 0.0411, |
| "num_tokens": 258681253.0, |
| "reward": 0.564453125, |
| "reward_std": 0.3288867473602295, |
| "rewards/simpleverify_reward/mean": 0.564453125, |
| "rewards/simpleverify_reward/std": 0.49631330370903015, |
| "step": 146 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.224609375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4088.0, |
| "completions/mean_length": 2723.36328125, |
| "completions/mean_terminated_length": 2325.748046875, |
| "completions/min_length": 442.0, |
| "completions/min_terminated_length": 442.0, |
| "epoch": 0.10432931156848829, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.18728479743003845, |
| "learning_rate": 7.079999999999999e-07, |
| "loss": 0.0305, |
| "num_tokens": 260134335.0, |
| "reward": 0.548828125, |
| "reward_std": 0.3487515449523926, |
| "rewards/simpleverify_reward/mean": 0.548828125, |
| "rewards/simpleverify_reward/std": 0.498096764087677, |
| "step": 147 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.2890625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 2839.04296875, |
| "completions/mean_terminated_length": 2327.97265625, |
| "completions/min_length": 963.0, |
| "completions/min_terminated_length": 963.0, |
| "epoch": 0.1050390347764372, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.17495302855968475, |
| "learning_rate": 7.059999999999999e-07, |
| "loss": 0.0344, |
| "num_tokens": 261630805.0, |
| "reward": 0.419921875, |
| "reward_std": 0.31567054986953735, |
| "rewards/simpleverify_reward/mean": 0.419921875, |
| "rewards/simpleverify_reward/std": 0.4940285086631775, |
| "step": 148 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.298828125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4069.0, |
| "completions/mean_length": 2685.63671875, |
| "completions/mean_terminated_length": 2084.562744140625, |
| "completions/min_length": 694.0, |
| "completions/min_terminated_length": 694.0, |
| "epoch": 0.10574875798438609, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.18346737325191498, |
| "learning_rate": 7.04e-07, |
| "loss": 0.0254, |
| "num_tokens": 263048907.0, |
| "reward": 0.5, |
| "reward_std": 0.28010767698287964, |
| "rewards/simpleverify_reward/mean": 0.5, |
| "rewards/simpleverify_reward/std": 0.5004889965057373, |
| "step": 149 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.392578125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4073.0, |
| "completions/mean_length": 2839.798828125, |
| "completions/mean_terminated_length": 2027.9132080078125, |
| "completions/min_length": 676.0, |
| "completions/min_terminated_length": 676.0, |
| "epoch": 0.10645848119233499, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.14945359528064728, |
| "learning_rate": 7.019999999999999e-07, |
| "loss": 0.039, |
| "num_tokens": 264550196.0, |
| "reward": 0.44140625, |
| "reward_std": 0.26879581809043884, |
| "rewards/simpleverify_reward/mean": 0.44140625, |
| "rewards/simpleverify_reward/std": 0.4970405399799347, |
| "step": 150 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.27734375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4067.0, |
| "completions/mean_length": 2831.484375, |
| "completions/mean_terminated_length": 2346.183837890625, |
| "completions/min_length": 55.0, |
| "completions/min_terminated_length": 55.0, |
| "epoch": 0.10716820440028389, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.19210544228553772, |
| "learning_rate": 7e-07, |
| "loss": 0.0064, |
| "num_tokens": 266051740.0, |
| "reward": 0.505859375, |
| "reward_std": 0.33529749512672424, |
| "rewards/simpleverify_reward/mean": 0.505859375, |
| "rewards/simpleverify_reward/std": 0.5004546642303467, |
| "step": 151 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.421875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4076.0, |
| "completions/mean_length": 3012.7109375, |
| "completions/mean_terminated_length": 2222.20263671875, |
| "completions/min_length": 697.0, |
| "completions/min_terminated_length": 697.0, |
| "epoch": 0.1078779276082328, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.17350085079669952, |
| "learning_rate": 6.979999999999999e-07, |
| "loss": 0.043, |
| "num_tokens": 267634664.0, |
| "reward": 0.42578125, |
| "reward_std": 0.32796716690063477, |
| "rewards/simpleverify_reward/mean": 0.42578125, |
| "rewards/simpleverify_reward/std": 0.4949444830417633, |
| "step": 152 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.3125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4093.0, |
| "completions/mean_length": 2946.61328125, |
| "completions/mean_terminated_length": 2424.164794921875, |
| "completions/min_length": 718.0, |
| "completions/min_terminated_length": 718.0, |
| "epoch": 0.10858765081618169, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.15166601538658142, |
| "learning_rate": 6.959999999999999e-07, |
| "loss": 0.0274, |
| "num_tokens": 269195186.0, |
| "reward": 0.552734375, |
| "reward_std": 0.29393789172172546, |
| "rewards/simpleverify_reward/mean": 0.552734375, |
| "rewards/simpleverify_reward/std": 0.4976975917816162, |
| "step": 153 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.36328125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4068.0, |
| "completions/mean_length": 2910.419921875, |
| "completions/mean_terminated_length": 2233.984619140625, |
| "completions/min_length": 202.0, |
| "completions/min_terminated_length": 202.0, |
| "epoch": 0.10929737402413059, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.1705927699804306, |
| "learning_rate": 6.939999999999999e-07, |
| "loss": 0.0447, |
| "num_tokens": 270732345.0, |
| "reward": 0.4375, |
| "reward_std": 0.3233838379383087, |
| "rewards/simpleverify_reward/mean": 0.4375, |
| "rewards/simpleverify_reward/std": 0.49656352400779724, |
| "step": 154 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.404296875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4082.0, |
| "completions/mean_length": 2965.12890625, |
| "completions/mean_terminated_length": 2197.61962890625, |
| "completions/min_length": 747.0, |
| "completions/min_terminated_length": 747.0, |
| "epoch": 0.11000709723207949, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.17731772363185883, |
| "learning_rate": 6.919999999999999e-07, |
| "loss": 0.053, |
| "num_tokens": 272299323.0, |
| "reward": 0.45703125, |
| "reward_std": 0.293290913105011, |
| "rewards/simpleverify_reward/mean": 0.45703125, |
| "rewards/simpleverify_reward/std": 0.49863746762275696, |
| "step": 155 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.326171875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4091.0, |
| "completions/mean_length": 2888.28515625, |
| "completions/mean_terminated_length": 2303.68115234375, |
| "completions/min_length": 895.0, |
| "completions/min_terminated_length": 895.0, |
| "epoch": 0.11071682044002838, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.18218308687210083, |
| "learning_rate": 6.9e-07, |
| "loss": 0.0558, |
| "num_tokens": 273843773.0, |
| "reward": 0.53125, |
| "reward_std": 0.3439802825450897, |
| "rewards/simpleverify_reward/mean": 0.53125, |
| "rewards/simpleverify_reward/std": 0.4995105266571045, |
| "step": 156 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.341796875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4093.0, |
| "completions/mean_length": 2923.806640625, |
| "completions/mean_terminated_length": 2315.100830078125, |
| "completions/min_length": 349.0, |
| "completions/min_terminated_length": 349.0, |
| "epoch": 0.11142654364797729, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.15375730395317078, |
| "learning_rate": 6.879999999999999e-07, |
| "loss": 0.0306, |
| "num_tokens": 275392106.0, |
| "reward": 0.447265625, |
| "reward_std": 0.34314247965812683, |
| "rewards/simpleverify_reward/mean": 0.447265625, |
| "rewards/simpleverify_reward/std": 0.4976975917816162, |
| "step": 157 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.28125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4090.0, |
| "completions/mean_length": 2626.232421875, |
| "completions/mean_terminated_length": 2051.10595703125, |
| "completions/min_length": 525.0, |
| "completions/min_terminated_length": 525.0, |
| "epoch": 0.11213626685592619, |
| "frac_reward_zero_std": 0.375, |
| "grad_norm": 0.18197540938854218, |
| "learning_rate": 6.86e-07, |
| "loss": 0.0468, |
| "num_tokens": 276781489.0, |
| "reward": 0.466796875, |
| "reward_std": 0.27893567085266113, |
| "rewards/simpleverify_reward/mean": 0.466796875, |
| "rewards/simpleverify_reward/std": 0.4993842542171478, |
| "step": 158 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.240234375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4088.0, |
| "completions/mean_length": 2603.20703125, |
| "completions/mean_terminated_length": 2131.19287109375, |
| "completions/min_length": 484.0, |
| "completions/min_terminated_length": 484.0, |
| "epoch": 0.11284599006387509, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.1840077042579651, |
| "learning_rate": 6.84e-07, |
| "loss": 0.0506, |
| "num_tokens": 278157323.0, |
| "reward": 0.53515625, |
| "reward_std": 0.3079981505870819, |
| "rewards/simpleverify_reward/mean": 0.53515625, |
| "rewards/simpleverify_reward/std": 0.49925029277801514, |
| "step": 159 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.263671875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4084.0, |
| "completions/mean_length": 2638.595703125, |
| "completions/mean_terminated_length": 2116.71337890625, |
| "completions/min_length": 589.0, |
| "completions/min_terminated_length": 589.0, |
| "epoch": 0.11355571327182398, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.2060907781124115, |
| "learning_rate": 6.82e-07, |
| "loss": 0.0563, |
| "num_tokens": 279553468.0, |
| "reward": 0.552734375, |
| "reward_std": 0.3513737916946411, |
| "rewards/simpleverify_reward/mean": 0.552734375, |
| "rewards/simpleverify_reward/std": 0.4976975917816162, |
| "step": 160 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.30859375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4047.0, |
| "completions/mean_length": 2683.3671875, |
| "completions/mean_terminated_length": 2052.8701171875, |
| "completions/min_length": 655.0, |
| "completions/min_terminated_length": 655.0, |
| "epoch": 0.11426543647977289, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.18641158938407898, |
| "learning_rate": 6.800000000000001e-07, |
| "loss": 0.0334, |
| "num_tokens": 280967768.0, |
| "reward": 0.529296875, |
| "reward_std": 0.3056606948375702, |
| "rewards/simpleverify_reward/mean": 0.529296875, |
| "rewards/simpleverify_reward/std": 0.49962911009788513, |
| "step": 161 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.3828125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4088.0, |
| "completions/mean_length": 2984.599609375, |
| "completions/mean_terminated_length": 2295.25, |
| "completions/min_length": 732.0, |
| "completions/min_terminated_length": 732.0, |
| "epoch": 0.11497515968772179, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.16161923110485077, |
| "learning_rate": 6.78e-07, |
| "loss": 0.0416, |
| "num_tokens": 282543259.0, |
| "reward": 0.49609375, |
| "reward_std": 0.3076329529285431, |
| "rewards/simpleverify_reward/mean": 0.49609375, |
| "rewards/simpleverify_reward/std": 0.5004737377166748, |
| "step": 162 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.2265625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4086.0, |
| "completions/mean_length": 2680.837890625, |
| "completions/mean_terminated_length": 2266.29541015625, |
| "completions/min_length": 616.0, |
| "completions/min_terminated_length": 616.0, |
| "epoch": 0.1156848828956707, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.19069817662239075, |
| "learning_rate": 6.76e-07, |
| "loss": 0.0389, |
| "num_tokens": 283964024.0, |
| "reward": 0.55859375, |
| "reward_std": 0.3719806969165802, |
| "rewards/simpleverify_reward/mean": 0.55859375, |
| "rewards/simpleverify_reward/std": 0.4970405399799347, |
| "step": 163 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.376953125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4081.0, |
| "completions/mean_length": 2977.556640625, |
| "completions/mean_terminated_length": 2300.880859375, |
| "completions/min_length": 708.0, |
| "completions/min_terminated_length": 708.0, |
| "epoch": 0.11639460610361958, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.17289063334465027, |
| "learning_rate": 6.74e-07, |
| "loss": 0.0278, |
| "num_tokens": 285533717.0, |
| "reward": 0.40625, |
| "reward_std": 0.3166877031326294, |
| "rewards/simpleverify_reward/mean": 0.40625, |
| "rewards/simpleverify_reward/std": 0.49161264300346375, |
| "step": 164 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.33203125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4086.0, |
| "completions/mean_length": 2876.431640625, |
| "completions/mean_terminated_length": 2270.21337890625, |
| "completions/min_length": 552.0, |
| "completions/min_terminated_length": 552.0, |
| "epoch": 0.11710432931156849, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.18286117911338806, |
| "learning_rate": 6.72e-07, |
| "loss": 0.0251, |
| "num_tokens": 287060530.0, |
| "reward": 0.474609375, |
| "reward_std": 0.28491973876953125, |
| "rewards/simpleverify_reward/mean": 0.474609375, |
| "rewards/simpleverify_reward/std": 0.4998432695865631, |
| "step": 165 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.33203125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4087.0, |
| "completions/mean_length": 2762.412109375, |
| "completions/mean_terminated_length": 2099.517578125, |
| "completions/min_length": 790.0, |
| "completions/min_terminated_length": 790.0, |
| "epoch": 0.11781405251951739, |
| "frac_reward_zero_std": 0.40625, |
| "grad_norm": 0.1441221833229065, |
| "learning_rate": 6.7e-07, |
| "loss": 0.0177, |
| "num_tokens": 288521429.0, |
| "reward": 0.466796875, |
| "reward_std": 0.2386757880449295, |
| "rewards/simpleverify_reward/mean": 0.466796875, |
| "rewards/simpleverify_reward/std": 0.4993842542171478, |
| "step": 166 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.189453125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4080.0, |
| "completions/mean_length": 2449.177734375, |
| "completions/mean_terminated_length": 2064.2578125, |
| "completions/min_length": 42.0, |
| "completions/min_terminated_length": 42.0, |
| "epoch": 0.1185237757274663, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.19036909937858582, |
| "learning_rate": 6.68e-07, |
| "loss": 0.0198, |
| "num_tokens": 289815392.0, |
| "reward": 0.578125, |
| "reward_std": 0.3070945143699646, |
| "rewards/simpleverify_reward/mean": 0.578125, |
| "rewards/simpleverify_reward/std": 0.49434176087379456, |
| "step": 167 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.248046875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4065.0, |
| "completions/mean_length": 2553.43359375, |
| "completions/mean_terminated_length": 2044.5869140625, |
| "completions/min_length": 536.0, |
| "completions/min_terminated_length": 536.0, |
| "epoch": 0.11923349893541518, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.21847881376743317, |
| "learning_rate": 6.66e-07, |
| "loss": 0.0428, |
| "num_tokens": 291175262.0, |
| "reward": 0.564453125, |
| "reward_std": 0.35230618715286255, |
| "rewards/simpleverify_reward/mean": 0.564453125, |
| "rewards/simpleverify_reward/std": 0.49631330370903015, |
| "step": 168 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.271484375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4082.0, |
| "completions/mean_length": 2552.0234375, |
| "completions/mean_terminated_length": 1976.6541748046875, |
| "completions/min_length": 418.0, |
| "completions/min_terminated_length": 418.0, |
| "epoch": 0.11994322214336409, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.1822100728750229, |
| "learning_rate": 6.64e-07, |
| "loss": 0.0099, |
| "num_tokens": 292524826.0, |
| "reward": 0.5, |
| "reward_std": 0.2985767722129822, |
| "rewards/simpleverify_reward/mean": 0.5, |
| "rewards/simpleverify_reward/std": 0.5004889965057373, |
| "step": 169 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.228515625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4055.0, |
| "completions/mean_length": 2774.5546875, |
| "completions/mean_terminated_length": 2383.13916015625, |
| "completions/min_length": 904.0, |
| "completions/min_terminated_length": 904.0, |
| "epoch": 0.12065294535131299, |
| "frac_reward_zero_std": 0.125, |
| "grad_norm": 0.19010956585407257, |
| "learning_rate": 6.62e-07, |
| "loss": 0.036, |
| "num_tokens": 293991862.0, |
| "reward": 0.572265625, |
| "reward_std": 0.352969765663147, |
| "rewards/simpleverify_reward/mean": 0.572265625, |
| "rewards/simpleverify_reward/std": 0.4952339828014374, |
| "step": 170 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.41015625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4056.0, |
| "completions/mean_length": 3022.46484375, |
| "completions/mean_terminated_length": 2275.966796875, |
| "completions/min_length": 533.0, |
| "completions/min_terminated_length": 533.0, |
| "epoch": 0.1213626685592619, |
| "frac_reward_zero_std": 0.375, |
| "grad_norm": 0.16382890939712524, |
| "learning_rate": 6.6e-07, |
| "loss": 0.0332, |
| "num_tokens": 295587076.0, |
| "reward": 0.37109375, |
| "reward_std": 0.2818239629268646, |
| "rewards/simpleverify_reward/mean": 0.37109375, |
| "rewards/simpleverify_reward/std": 0.4835699498653412, |
| "step": 171 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.23828125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4090.0, |
| "completions/mean_length": 2740.162109375, |
| "completions/mean_terminated_length": 2316.0283203125, |
| "completions/min_length": 420.0, |
| "completions/min_terminated_length": 420.0, |
| "epoch": 0.12207239176721078, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.19507798552513123, |
| "learning_rate": 6.58e-07, |
| "loss": 0.0656, |
| "num_tokens": 297047063.0, |
| "reward": 0.578125, |
| "reward_std": 0.36732351779937744, |
| "rewards/simpleverify_reward/mean": 0.578125, |
| "rewards/simpleverify_reward/std": 0.49434176087379456, |
| "step": 172 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.298828125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4063.0, |
| "completions/mean_length": 2865.62890625, |
| "completions/mean_terminated_length": 2341.2646484375, |
| "completions/min_length": 755.0, |
| "completions/min_terminated_length": 755.0, |
| "epoch": 0.12278211497515969, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.1689491719007492, |
| "learning_rate": 6.56e-07, |
| "loss": 0.0276, |
| "num_tokens": 298561657.0, |
| "reward": 0.408203125, |
| "reward_std": 0.32400238513946533, |
| "rewards/simpleverify_reward/mean": 0.408203125, |
| "rewards/simpleverify_reward/std": 0.49198177456855774, |
| "step": 173 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.216796875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4090.0, |
| "completions/mean_length": 2610.828125, |
| "completions/mean_terminated_length": 2199.720703125, |
| "completions/min_length": 827.0, |
| "completions/min_terminated_length": 827.0, |
| "epoch": 0.12349183818310859, |
| "frac_reward_zero_std": 0.09375, |
| "grad_norm": 0.20500050485134125, |
| "learning_rate": 6.54e-07, |
| "loss": 0.0604, |
| "num_tokens": 299949009.0, |
| "reward": 0.537109375, |
| "reward_std": 0.3801402151584625, |
| "rewards/simpleverify_reward/mean": 0.537109375, |
| "rewards/simpleverify_reward/std": 0.4991086423397064, |
| "step": 174 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.232421875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4089.0, |
| "completions/mean_length": 2786.884765625, |
| "completions/mean_terminated_length": 2390.486083984375, |
| "completions/min_length": 414.0, |
| "completions/min_terminated_length": 414.0, |
| "epoch": 0.1242015613910575, |
| "frac_reward_zero_std": 0.0625, |
| "grad_norm": 0.1940719485282898, |
| "learning_rate": 6.52e-07, |
| "loss": 0.0703, |
| "num_tokens": 301421254.0, |
| "reward": 0.509765625, |
| "reward_std": 0.4262314438819885, |
| "rewards/simpleverify_reward/mean": 0.509765625, |
| "rewards/simpleverify_reward/std": 0.5003935098648071, |
| "step": 175 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.232421875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4080.0, |
| "completions/mean_length": 2701.06640625, |
| "completions/mean_terminated_length": 2278.681884765625, |
| "completions/min_length": 763.0, |
| "completions/min_terminated_length": 763.0, |
| "epoch": 0.12491128459900638, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.17947563529014587, |
| "learning_rate": 6.5e-07, |
| "loss": 0.0165, |
| "num_tokens": 302848504.0, |
| "reward": 0.5625, |
| "reward_std": 0.30335918068885803, |
| "rewards/simpleverify_reward/mean": 0.5625, |
| "rewards/simpleverify_reward/std": 0.49656352400779724, |
| "step": 176 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.244140625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4073.0, |
| "completions/mean_length": 2641.98828125, |
| "completions/mean_terminated_length": 2172.34619140625, |
| "completions/min_length": 743.0, |
| "completions/min_terminated_length": 743.0, |
| "epoch": 0.1256210078069553, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.18924672901630402, |
| "learning_rate": 6.48e-07, |
| "loss": 0.0434, |
| "num_tokens": 304241634.0, |
| "reward": 0.634765625, |
| "reward_std": 0.31121736764907837, |
| "rewards/simpleverify_reward/mean": 0.634765625, |
| "rewards/simpleverify_reward/std": 0.4819667339324951, |
| "step": 177 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.240234375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4090.0, |
| "completions/mean_length": 2556.810546875, |
| "completions/mean_terminated_length": 2070.1259765625, |
| "completions/min_length": 237.0, |
| "completions/min_terminated_length": 237.0, |
| "epoch": 0.1263307310149042, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.19043995440006256, |
| "learning_rate": 6.46e-07, |
| "loss": 0.0327, |
| "num_tokens": 305610785.0, |
| "reward": 0.544921875, |
| "reward_std": 0.3164547085762024, |
| "rewards/simpleverify_reward/mean": 0.544921875, |
| "rewards/simpleverify_reward/std": 0.4984649419784546, |
| "step": 178 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.314453125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4081.0, |
| "completions/mean_length": 2679.068359375, |
| "completions/mean_terminated_length": 2029.13671875, |
| "completions/min_length": 572.0, |
| "completions/min_terminated_length": 572.0, |
| "epoch": 0.1270404542228531, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.18073788285255432, |
| "learning_rate": 6.44e-07, |
| "loss": 0.0512, |
| "num_tokens": 307027540.0, |
| "reward": 0.45703125, |
| "reward_std": 0.2876802980899811, |
| "rewards/simpleverify_reward/mean": 0.45703125, |
| "rewards/simpleverify_reward/std": 0.49863746762275696, |
| "step": 179 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.40234375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4090.0, |
| "completions/mean_length": 3031.56640625, |
| "completions/mean_terminated_length": 2314.987060546875, |
| "completions/min_length": 473.0, |
| "completions/min_terminated_length": 473.0, |
| "epoch": 0.127750177430802, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.17582952976226807, |
| "learning_rate": 6.42e-07, |
| "loss": 0.0437, |
| "num_tokens": 308626294.0, |
| "reward": 0.42578125, |
| "reward_std": 0.33148908615112305, |
| "rewards/simpleverify_reward/mean": 0.42578125, |
| "rewards/simpleverify_reward/std": 0.4949444830417633, |
| "step": 180 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.517578125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4090.0, |
| "completions/mean_length": 3323.435546875, |
| "completions/mean_terminated_length": 2494.57080078125, |
| "completions/min_length": 537.0, |
| "completions/min_terminated_length": 537.0, |
| "epoch": 0.12845990063875087, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.15160709619522095, |
| "learning_rate": 6.4e-07, |
| "loss": 0.0271, |
| "num_tokens": 310374885.0, |
| "reward": 0.357421875, |
| "reward_std": 0.2855038642883301, |
| "rewards/simpleverify_reward/mean": 0.357421875, |
| "rewards/simpleverify_reward/std": 0.4797092080116272, |
| "step": 181 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.322265625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4076.0, |
| "completions/mean_length": 2796.357421875, |
| "completions/mean_terminated_length": 2178.37158203125, |
| "completions/min_length": 640.0, |
| "completions/min_terminated_length": 640.0, |
| "epoch": 0.12916962384669978, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.18909138441085815, |
| "learning_rate": 6.38e-07, |
| "loss": 0.0629, |
| "num_tokens": 311851692.0, |
| "reward": 0.47265625, |
| "reward_std": 0.33643782138824463, |
| "rewards/simpleverify_reward/mean": 0.47265625, |
| "rewards/simpleverify_reward/std": 0.49974003434181213, |
| "step": 182 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.26171875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4035.0, |
| "completions/mean_length": 2663.328125, |
| "completions/mean_terminated_length": 2155.44970703125, |
| "completions/min_length": 454.0, |
| "completions/min_terminated_length": 454.0, |
| "epoch": 0.12987934705464868, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.19046445190906525, |
| "learning_rate": 6.36e-07, |
| "loss": 0.0425, |
| "num_tokens": 313256116.0, |
| "reward": 0.431640625, |
| "reward_std": 0.3443180322647095, |
| "rewards/simpleverify_reward/mean": 0.431640625, |
| "rewards/simpleverify_reward/std": 0.4957893490791321, |
| "step": 183 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.193359375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4090.0, |
| "completions/mean_length": 2522.755859375, |
| "completions/mean_terminated_length": 2145.634521484375, |
| "completions/min_length": 650.0, |
| "completions/min_terminated_length": 650.0, |
| "epoch": 0.13058907026259758, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.22442075610160828, |
| "learning_rate": 6.34e-07, |
| "loss": 0.0279, |
| "num_tokens": 314591655.0, |
| "reward": 0.64453125, |
| "reward_std": 0.34537842869758606, |
| "rewards/simpleverify_reward/mean": 0.64453125, |
| "rewards/simpleverify_reward/std": 0.47912323474884033, |
| "step": 184 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.251953125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4064.0, |
| "completions/mean_length": 2535.09765625, |
| "completions/mean_terminated_length": 2009.3629150390625, |
| "completions/min_length": 465.0, |
| "completions/min_terminated_length": 465.0, |
| "epoch": 0.1312987934705465, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.19926586747169495, |
| "learning_rate": 6.319999999999999e-07, |
| "loss": 0.0228, |
| "num_tokens": 315936025.0, |
| "reward": 0.533203125, |
| "reward_std": 0.321033239364624, |
| "rewards/simpleverify_reward/mean": 0.533203125, |
| "rewards/simpleverify_reward/std": 0.4993842542171478, |
| "step": 185 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.287109375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4082.0, |
| "completions/mean_length": 2769.3515625, |
| "completions/mean_terminated_length": 2235.0576171875, |
| "completions/min_length": 486.0, |
| "completions/min_terminated_length": 486.0, |
| "epoch": 0.1320085166784954, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.19813241064548492, |
| "learning_rate": 6.3e-07, |
| "loss": 0.0279, |
| "num_tokens": 317402029.0, |
| "reward": 0.470703125, |
| "reward_std": 0.34010064601898193, |
| "rewards/simpleverify_reward/mean": 0.470703125, |
| "rewards/simpleverify_reward/std": 0.49962911009788513, |
| "step": 186 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.357421875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4083.0, |
| "completions/mean_length": 3075.984375, |
| "completions/mean_terminated_length": 2508.6201171875, |
| "completions/min_length": 492.0, |
| "completions/min_terminated_length": 492.0, |
| "epoch": 0.1327182398864443, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.18306022882461548, |
| "learning_rate": 6.28e-07, |
| "loss": 0.0362, |
| "num_tokens": 319033877.0, |
| "reward": 0.421875, |
| "reward_std": 0.32141512632369995, |
| "rewards/simpleverify_reward/mean": 0.421875, |
| "rewards/simpleverify_reward/std": 0.49434176087379456, |
| "step": 187 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.330078125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4076.0, |
| "completions/mean_length": 2953.32421875, |
| "completions/mean_terminated_length": 2390.31494140625, |
| "completions/min_length": 603.0, |
| "completions/min_terminated_length": 603.0, |
| "epoch": 0.1334279630943932, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.1696513444185257, |
| "learning_rate": 6.26e-07, |
| "loss": 0.0499, |
| "num_tokens": 320588683.0, |
| "reward": 0.5234375, |
| "reward_std": 0.29684534668922424, |
| "rewards/simpleverify_reward/mean": 0.5234375, |
| "rewards/simpleverify_reward/std": 0.49993884563446045, |
| "step": 188 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.22265625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4069.0, |
| "completions/mean_length": 2450.611328125, |
| "completions/mean_terminated_length": 1979.319091796875, |
| "completions/min_length": 464.0, |
| "completions/min_terminated_length": 464.0, |
| "epoch": 0.13413768630234207, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.21474160254001617, |
| "learning_rate": 6.24e-07, |
| "loss": 0.057, |
| "num_tokens": 321880196.0, |
| "reward": 0.603515625, |
| "reward_std": 0.32950931787490845, |
| "rewards/simpleverify_reward/mean": 0.603515625, |
| "rewards/simpleverify_reward/std": 0.4896455705165863, |
| "step": 189 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.1640625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4051.0, |
| "completions/mean_length": 2352.560546875, |
| "completions/mean_terminated_length": 2010.39013671875, |
| "completions/min_length": 347.0, |
| "completions/min_terminated_length": 347.0, |
| "epoch": 0.13484740951029098, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.2193438559770584, |
| "learning_rate": 6.219999999999999e-07, |
| "loss": 0.0362, |
| "num_tokens": 323134531.0, |
| "reward": 0.576171875, |
| "reward_std": 0.36345213651657104, |
| "rewards/simpleverify_reward/mean": 0.576171875, |
| "rewards/simpleverify_reward/std": 0.4946470856666565, |
| "step": 190 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.2734375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4088.0, |
| "completions/mean_length": 2678.7578125, |
| "completions/mean_terminated_length": 2145.38720703125, |
| "completions/min_length": 466.0, |
| "completions/min_terminated_length": 466.0, |
| "epoch": 0.13555713271823988, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.1765129268169403, |
| "learning_rate": 6.2e-07, |
| "loss": 0.0188, |
| "num_tokens": 324541799.0, |
| "reward": 0.5625, |
| "reward_std": 0.2784305214881897, |
| "rewards/simpleverify_reward/mean": 0.5625, |
| "rewards/simpleverify_reward/std": 0.49656352400779724, |
| "step": 191 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.3828125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4078.0, |
| "completions/mean_length": 2875.86328125, |
| "completions/mean_terminated_length": 2119.069580078125, |
| "completions/min_length": 474.0, |
| "completions/min_terminated_length": 474.0, |
| "epoch": 0.13626685592618878, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.18329308927059174, |
| "learning_rate": 6.18e-07, |
| "loss": 0.0495, |
| "num_tokens": 326068145.0, |
| "reward": 0.408203125, |
| "reward_std": 0.27528315782546997, |
| "rewards/simpleverify_reward/mean": 0.408203125, |
| "rewards/simpleverify_reward/std": 0.49198177456855774, |
| "step": 192 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.22265625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4086.0, |
| "completions/mean_length": 2729.765625, |
| "completions/mean_terminated_length": 2338.43212890625, |
| "completions/min_length": 745.0, |
| "completions/min_terminated_length": 745.0, |
| "epoch": 0.1369765791341377, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.18101434409618378, |
| "learning_rate": 6.16e-07, |
| "loss": 0.0326, |
| "num_tokens": 327521129.0, |
| "reward": 0.56640625, |
| "reward_std": 0.37592363357543945, |
| "rewards/simpleverify_reward/mean": 0.56640625, |
| "rewards/simpleverify_reward/std": 0.4960552453994751, |
| "step": 193 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.2265625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4091.0, |
| "completions/mean_length": 2599.751953125, |
| "completions/mean_terminated_length": 2161.45703125, |
| "completions/min_length": 580.0, |
| "completions/min_terminated_length": 580.0, |
| "epoch": 0.1376863023420866, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.1901218593120575, |
| "learning_rate": 6.14e-07, |
| "loss": 0.0383, |
| "num_tokens": 328899866.0, |
| "reward": 0.509765625, |
| "reward_std": 0.3250877857208252, |
| "rewards/simpleverify_reward/mean": 0.509765625, |
| "rewards/simpleverify_reward/std": 0.5003935098648071, |
| "step": 194 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.19140625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4080.0, |
| "completions/mean_length": 2396.232421875, |
| "completions/mean_terminated_length": 1993.8719482421875, |
| "completions/min_length": 720.0, |
| "completions/min_terminated_length": 720.0, |
| "epoch": 0.1383960255500355, |
| "frac_reward_zero_std": 0.125, |
| "grad_norm": 0.21104131639003754, |
| "learning_rate": 6.119999999999999e-07, |
| "loss": 0.0615, |
| "num_tokens": 330177057.0, |
| "reward": 0.5546875, |
| "reward_std": 0.3645366430282593, |
| "rewards/simpleverify_reward/mean": 0.5546875, |
| "rewards/simpleverify_reward/std": 0.49748632311820984, |
| "step": 195 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.30859375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4087.0, |
| "completions/mean_length": 2995.041015625, |
| "completions/mean_terminated_length": 2503.652587890625, |
| "completions/min_length": 678.0, |
| "completions/min_terminated_length": 678.0, |
| "epoch": 0.1391057487579844, |
| "frac_reward_zero_std": 0.125, |
| "grad_norm": 0.19640375673770905, |
| "learning_rate": 6.1e-07, |
| "loss": 0.0348, |
| "num_tokens": 331757494.0, |
| "reward": 0.494140625, |
| "reward_std": 0.36113378405570984, |
| "rewards/simpleverify_reward/mean": 0.494140625, |
| "rewards/simpleverify_reward/std": 0.5004546642303467, |
| "step": 196 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.267578125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4073.0, |
| "completions/mean_length": 2608.486328125, |
| "completions/mean_terminated_length": 2065.048095703125, |
| "completions/min_length": 594.0, |
| "completions/min_terminated_length": 594.0, |
| "epoch": 0.13981547196593327, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.19372299313545227, |
| "learning_rate": 6.079999999999999e-07, |
| "loss": 0.0381, |
| "num_tokens": 333139919.0, |
| "reward": 0.537109375, |
| "reward_std": 0.3582150936126709, |
| "rewards/simpleverify_reward/mean": 0.537109375, |
| "rewards/simpleverify_reward/std": 0.4991086423397064, |
| "step": 197 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.376953125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4083.0, |
| "completions/mean_length": 2865.107421875, |
| "completions/mean_terminated_length": 2120.39794921875, |
| "completions/min_length": 542.0, |
| "completions/min_terminated_length": 542.0, |
| "epoch": 0.14052519517388218, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.16689543426036835, |
| "learning_rate": 6.06e-07, |
| "loss": 0.0148, |
| "num_tokens": 334655494.0, |
| "reward": 0.439453125, |
| "reward_std": 0.29074186086654663, |
| "rewards/simpleverify_reward/mean": 0.439453125, |
| "rewards/simpleverify_reward/std": 0.49680593609809875, |
| "step": 198 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.18359375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4016.0, |
| "completions/mean_length": 2516.400390625, |
| "completions/mean_terminated_length": 2161.179443359375, |
| "completions/min_length": 675.0, |
| "completions/min_terminated_length": 675.0, |
| "epoch": 0.14123491838183108, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.18413615226745605, |
| "learning_rate": 6.04e-07, |
| "loss": 0.0398, |
| "num_tokens": 335992515.0, |
| "reward": 0.630859375, |
| "reward_std": 0.3117239475250244, |
| "rewards/simpleverify_reward/mean": 0.630859375, |
| "rewards/simpleverify_reward/std": 0.4830440282821655, |
| "step": 199 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.31640625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4056.0, |
| "completions/mean_length": 2764.994140625, |
| "completions/mean_terminated_length": 2148.928466796875, |
| "completions/min_length": 539.0, |
| "completions/min_terminated_length": 539.0, |
| "epoch": 0.14194464158977999, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.16617894172668457, |
| "learning_rate": 6.019999999999999e-07, |
| "loss": 0.0255, |
| "num_tokens": 337450128.0, |
| "reward": 0.482421875, |
| "reward_std": 0.2826576232910156, |
| "rewards/simpleverify_reward/mean": 0.482421875, |
| "rewards/simpleverify_reward/std": 0.5001795887947083, |
| "step": 200 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.212890625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4055.0, |
| "completions/mean_length": 2483.22265625, |
| "completions/mean_terminated_length": 2047.012451171875, |
| "completions/min_length": 604.0, |
| "completions/min_terminated_length": 604.0, |
| "epoch": 0.1426543647977289, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.19393746554851532, |
| "learning_rate": 6e-07, |
| "loss": 0.0348, |
| "num_tokens": 338773682.0, |
| "reward": 0.57421875, |
| "reward_std": 0.3195977807044983, |
| "rewards/simpleverify_reward/mean": 0.57421875, |
| "rewards/simpleverify_reward/std": 0.4949444830417633, |
| "step": 201 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.32421875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4085.0, |
| "completions/mean_length": 2834.890625, |
| "completions/mean_terminated_length": 2229.849609375, |
| "completions/min_length": 706.0, |
| "completions/min_terminated_length": 706.0, |
| "epoch": 0.1433640880056778, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.17194059491157532, |
| "learning_rate": 5.979999999999999e-07, |
| "loss": 0.0391, |
| "num_tokens": 340269210.0, |
| "reward": 0.482421875, |
| "reward_std": 0.3398098051548004, |
| "rewards/simpleverify_reward/mean": 0.482421875, |
| "rewards/simpleverify_reward/std": 0.5001795887947083, |
| "step": 202 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.224609375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4051.0, |
| "completions/mean_length": 2578.828125, |
| "completions/mean_terminated_length": 2139.344970703125, |
| "completions/min_length": 639.0, |
| "completions/min_terminated_length": 639.0, |
| "epoch": 0.1440738112136267, |
| "frac_reward_zero_std": 0.09375, |
| "grad_norm": 0.20428285002708435, |
| "learning_rate": 5.96e-07, |
| "loss": 0.0628, |
| "num_tokens": 341641298.0, |
| "reward": 0.55078125, |
| "reward_std": 0.3916857838630676, |
| "rewards/simpleverify_reward/mean": 0.55078125, |
| "rewards/simpleverify_reward/std": 0.497901052236557, |
| "step": 203 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.2421875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4088.0, |
| "completions/mean_length": 2691.525390625, |
| "completions/mean_terminated_length": 2242.672607421875, |
| "completions/min_length": 458.0, |
| "completions/min_terminated_length": 458.0, |
| "epoch": 0.1447835344215756, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.24238519370555878, |
| "learning_rate": 5.939999999999999e-07, |
| "loss": 0.0603, |
| "num_tokens": 343065679.0, |
| "reward": 0.630859375, |
| "reward_std": 0.32010918855667114, |
| "rewards/simpleverify_reward/mean": 0.630859375, |
| "rewards/simpleverify_reward/std": 0.4830440282821655, |
| "step": 204 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.197265625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4087.0, |
| "completions/mean_length": 2407.421875, |
| "completions/mean_terminated_length": 1992.4671630859375, |
| "completions/min_length": 587.0, |
| "completions/min_terminated_length": 587.0, |
| "epoch": 0.14549325762952448, |
| "frac_reward_zero_std": 0.125, |
| "grad_norm": 0.21709959208965302, |
| "learning_rate": 5.919999999999999e-07, |
| "loss": 0.0481, |
| "num_tokens": 344350759.0, |
| "reward": 0.564453125, |
| "reward_std": 0.3764348030090332, |
| "rewards/simpleverify_reward/mean": 0.564453125, |
| "rewards/simpleverify_reward/std": 0.49631330370903015, |
| "step": 205 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.23046875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4083.0, |
| "completions/mean_length": 2750.185546875, |
| "completions/mean_terminated_length": 2347.124267578125, |
| "completions/min_length": 467.0, |
| "completions/min_terminated_length": 467.0, |
| "epoch": 0.14620298083747338, |
| "frac_reward_zero_std": 0.09375, |
| "grad_norm": 0.20958703756332397, |
| "learning_rate": 5.9e-07, |
| "loss": 0.0424, |
| "num_tokens": 345801766.0, |
| "reward": 0.505859375, |
| "reward_std": 0.39000988006591797, |
| "rewards/simpleverify_reward/mean": 0.505859375, |
| "rewards/simpleverify_reward/std": 0.5004546642303467, |
| "step": 206 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.279296875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4088.0, |
| "completions/mean_length": 2623.966796875, |
| "completions/mean_terminated_length": 2053.504150390625, |
| "completions/min_length": 580.0, |
| "completions/min_terminated_length": 580.0, |
| "epoch": 0.14691270404542228, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.18957479298114777, |
| "learning_rate": 5.879999999999999e-07, |
| "loss": 0.0498, |
| "num_tokens": 347190533.0, |
| "reward": 0.53125, |
| "reward_std": 0.30021119117736816, |
| "rewards/simpleverify_reward/mean": 0.53125, |
| "rewards/simpleverify_reward/std": 0.4995105266571045, |
| "step": 207 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.36328125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4084.0, |
| "completions/mean_length": 3076.05078125, |
| "completions/mean_terminated_length": 2494.116455078125, |
| "completions/min_length": 531.0, |
| "completions/min_terminated_length": 531.0, |
| "epoch": 0.14762242725337119, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.17310208082199097, |
| "learning_rate": 5.86e-07, |
| "loss": 0.0305, |
| "num_tokens": 348803855.0, |
| "reward": 0.451171875, |
| "reward_std": 0.3601369857788086, |
| "rewards/simpleverify_reward/mean": 0.451171875, |
| "rewards/simpleverify_reward/std": 0.498096764087677, |
| "step": 208 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.27734375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4076.0, |
| "completions/mean_length": 2640.6328125, |
| "completions/mean_terminated_length": 2082.08642578125, |
| "completions/min_length": 496.0, |
| "completions/min_terminated_length": 496.0, |
| "epoch": 0.1483321504613201, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.17670436203479767, |
| "learning_rate": 5.839999999999999e-07, |
| "loss": 0.0542, |
| "num_tokens": 350205411.0, |
| "reward": 0.5625, |
| "reward_std": 0.3062669038772583, |
| "rewards/simpleverify_reward/mean": 0.5625, |
| "rewards/simpleverify_reward/std": 0.49656352400779724, |
| "step": 209 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.25390625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 2867.150390625, |
| "completions/mean_terminated_length": 2448.95556640625, |
| "completions/min_length": 457.0, |
| "completions/min_terminated_length": 457.0, |
| "epoch": 0.149041873669269, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.18433161079883575, |
| "learning_rate": 5.819999999999999e-07, |
| "loss": 0.029, |
| "num_tokens": 351721952.0, |
| "reward": 0.490234375, |
| "reward_std": 0.3143155574798584, |
| "rewards/simpleverify_reward/mean": 0.490234375, |
| "rewards/simpleverify_reward/std": 0.5003935098648071, |
| "step": 210 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.314453125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4074.0, |
| "completions/mean_length": 2713.1484375, |
| "completions/mean_terminated_length": 2078.848876953125, |
| "completions/min_length": 618.0, |
| "completions/min_terminated_length": 618.0, |
| "epoch": 0.1497515968772179, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.17659804224967957, |
| "learning_rate": 5.8e-07, |
| "loss": 0.0421, |
| "num_tokens": 353160940.0, |
| "reward": 0.515625, |
| "reward_std": 0.31040072441101074, |
| "rewards/simpleverify_reward/mean": 0.515625, |
| "rewards/simpleverify_reward/std": 0.5002445578575134, |
| "step": 211 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.2421875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4057.0, |
| "completions/mean_length": 2580.474609375, |
| "completions/mean_terminated_length": 2096.13134765625, |
| "completions/min_length": 568.0, |
| "completions/min_terminated_length": 568.0, |
| "epoch": 0.15046132008516677, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.1837129145860672, |
| "learning_rate": 5.779999999999999e-07, |
| "loss": 0.0311, |
| "num_tokens": 354530447.0, |
| "reward": 0.4765625, |
| "reward_std": 0.31170618534088135, |
| "rewards/simpleverify_reward/mean": 0.4765625, |
| "rewards/simpleverify_reward/std": 0.49993884563446045, |
| "step": 212 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.296875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4070.0, |
| "completions/mean_length": 2861.677734375, |
| "completions/mean_terminated_length": 2340.51953125, |
| "completions/min_length": 530.0, |
| "completions/min_terminated_length": 530.0, |
| "epoch": 0.15117104329311568, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.1911577582359314, |
| "learning_rate": 5.76e-07, |
| "loss": 0.0487, |
| "num_tokens": 356045946.0, |
| "reward": 0.546875, |
| "reward_std": 0.3171061873435974, |
| "rewards/simpleverify_reward/mean": 0.546875, |
| "rewards/simpleverify_reward/std": 0.4982847273349762, |
| "step": 213 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.1875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4043.0, |
| "completions/mean_length": 2595.1640625, |
| "completions/mean_terminated_length": 2248.8173828125, |
| "completions/min_length": 374.0, |
| "completions/min_terminated_length": 374.0, |
| "epoch": 0.15188076650106458, |
| "frac_reward_zero_std": 0.375, |
| "grad_norm": 0.16014865040779114, |
| "learning_rate": 5.739999999999999e-07, |
| "loss": 0.0223, |
| "num_tokens": 357420270.0, |
| "reward": 0.546875, |
| "reward_std": 0.27465832233428955, |
| "rewards/simpleverify_reward/mean": 0.546875, |
| "rewards/simpleverify_reward/std": 0.4982847273349762, |
| "step": 214 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.1875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4091.0, |
| "completions/mean_length": 2533.119140625, |
| "completions/mean_terminated_length": 2172.454345703125, |
| "completions/min_length": 564.0, |
| "completions/min_terminated_length": 564.0, |
| "epoch": 0.15259048970901348, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.18444742262363434, |
| "learning_rate": 5.719999999999999e-07, |
| "loss": 0.0338, |
| "num_tokens": 358756667.0, |
| "reward": 0.6171875, |
| "reward_std": 0.3308834731578827, |
| "rewards/simpleverify_reward/mean": 0.6171875, |
| "rewards/simpleverify_reward/std": 0.486548513174057, |
| "step": 215 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.2578125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4086.0, |
| "completions/mean_length": 2675.3515625, |
| "completions/mean_terminated_length": 2181.86328125, |
| "completions/min_length": 476.0, |
| "completions/min_terminated_length": 476.0, |
| "epoch": 0.1533002129169624, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.19254998862743378, |
| "learning_rate": 5.699999999999999e-07, |
| "loss": 0.0408, |
| "num_tokens": 360167551.0, |
| "reward": 0.478515625, |
| "reward_std": 0.3241611123085022, |
| "rewards/simpleverify_reward/mean": 0.478515625, |
| "rewards/simpleverify_reward/std": 0.5000267624855042, |
| "step": 216 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.2421875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4065.0, |
| "completions/mean_length": 2675.68359375, |
| "completions/mean_terminated_length": 2221.76806640625, |
| "completions/min_length": 460.0, |
| "completions/min_terminated_length": 460.0, |
| "epoch": 0.1540099361249113, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.18435202538967133, |
| "learning_rate": 5.679999999999999e-07, |
| "loss": 0.0404, |
| "num_tokens": 361587421.0, |
| "reward": 0.546875, |
| "reward_std": 0.3397667109966278, |
| "rewards/simpleverify_reward/mean": 0.546875, |
| "rewards/simpleverify_reward/std": 0.4982847273349762, |
| "step": 217 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.1328125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4077.0, |
| "completions/mean_length": 2309.75, |
| "completions/mean_terminated_length": 2036.18017578125, |
| "completions/min_length": 565.0, |
| "completions/min_terminated_length": 565.0, |
| "epoch": 0.1547196593328602, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.20714353024959564, |
| "learning_rate": 5.66e-07, |
| "loss": 0.0478, |
| "num_tokens": 362811277.0, |
| "reward": 0.67578125, |
| "reward_std": 0.33898013830184937, |
| "rewards/simpleverify_reward/mean": 0.67578125, |
| "rewards/simpleverify_reward/std": 0.4685399830341339, |
| "step": 218 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.185546875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4078.0, |
| "completions/mean_length": 2444.544921875, |
| "completions/mean_terminated_length": 2068.314208984375, |
| "completions/min_length": 494.0, |
| "completions/min_terminated_length": 494.0, |
| "epoch": 0.1554293825408091, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.2170497328042984, |
| "learning_rate": 5.639999999999999e-07, |
| "loss": 0.0399, |
| "num_tokens": 364104868.0, |
| "reward": 0.5859375, |
| "reward_std": 0.3199089765548706, |
| "rewards/simpleverify_reward/mean": 0.5859375, |
| "rewards/simpleverify_reward/std": 0.49304109811782837, |
| "step": 219 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.220703125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4030.0, |
| "completions/mean_length": 2530.1796875, |
| "completions/mean_terminated_length": 2086.726806640625, |
| "completions/min_length": 657.0, |
| "completions/min_terminated_length": 657.0, |
| "epoch": 0.15613910574875797, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.21132685244083405, |
| "learning_rate": 5.620000000000001e-07, |
| "loss": 0.0485, |
| "num_tokens": 365450256.0, |
| "reward": 0.58203125, |
| "reward_std": 0.31587207317352295, |
| "rewards/simpleverify_reward/mean": 0.58203125, |
| "rewards/simpleverify_reward/std": 0.4937073290348053, |
| "step": 220 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.21484375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4088.0, |
| "completions/mean_length": 2662.302734375, |
| "completions/mean_terminated_length": 2269.99755859375, |
| "completions/min_length": 459.0, |
| "completions/min_terminated_length": 459.0, |
| "epoch": 0.15684882895670688, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.1874718815088272, |
| "learning_rate": 5.6e-07, |
| "loss": 0.0382, |
| "num_tokens": 366863227.0, |
| "reward": 0.552734375, |
| "reward_std": 0.3736315965652466, |
| "rewards/simpleverify_reward/mean": 0.552734375, |
| "rewards/simpleverify_reward/std": 0.4976975917816162, |
| "step": 221 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.236328125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 2749.794921875, |
| "completions/mean_terminated_length": 2333.1943359375, |
| "completions/min_length": 644.0, |
| "completions/min_terminated_length": 644.0, |
| "epoch": 0.15755855216465578, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.19688737392425537, |
| "learning_rate": 5.58e-07, |
| "loss": 0.0281, |
| "num_tokens": 368319986.0, |
| "reward": 0.478515625, |
| "reward_std": 0.35983237624168396, |
| "rewards/simpleverify_reward/mean": 0.478515625, |
| "rewards/simpleverify_reward/std": 0.5000267624855042, |
| "step": 222 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.212890625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4050.0, |
| "completions/mean_length": 2609.787109375, |
| "completions/mean_terminated_length": 2207.808837890625, |
| "completions/min_length": 670.0, |
| "completions/min_terminated_length": 670.0, |
| "epoch": 0.15826827537260468, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.18813881278038025, |
| "learning_rate": 5.560000000000001e-07, |
| "loss": 0.0409, |
| "num_tokens": 369703941.0, |
| "reward": 0.564453125, |
| "reward_std": 0.3500629961490631, |
| "rewards/simpleverify_reward/mean": 0.564453125, |
| "rewards/simpleverify_reward/std": 0.49631330370903015, |
| "step": 223 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.29296875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4091.0, |
| "completions/mean_length": 2800.796875, |
| "completions/mean_terminated_length": 2264.110595703125, |
| "completions/min_length": 562.0, |
| "completions/min_terminated_length": 562.0, |
| "epoch": 0.1589779985805536, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.1872749626636505, |
| "learning_rate": 5.54e-07, |
| "loss": 0.0394, |
| "num_tokens": 371185069.0, |
| "reward": 0.5078125, |
| "reward_std": 0.36320289969444275, |
| "rewards/simpleverify_reward/mean": 0.5078125, |
| "rewards/simpleverify_reward/std": 0.5004279017448425, |
| "step": 224 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.18359375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4092.0, |
| "completions/mean_length": 2445.109375, |
| "completions/mean_terminated_length": 2073.8564453125, |
| "completions/min_length": 439.0, |
| "completions/min_terminated_length": 439.0, |
| "epoch": 0.1596877217885025, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.21936750411987305, |
| "learning_rate": 5.520000000000001e-07, |
| "loss": 0.0571, |
| "num_tokens": 372479157.0, |
| "reward": 0.62890625, |
| "reward_std": 0.35524022579193115, |
| "rewards/simpleverify_reward/mean": 0.62890625, |
| "rewards/simpleverify_reward/std": 0.4835699498653412, |
| "step": 225 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.34765625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4075.0, |
| "completions/mean_length": 2912.369140625, |
| "completions/mean_terminated_length": 2281.572021484375, |
| "completions/min_length": 919.0, |
| "completions/min_terminated_length": 919.0, |
| "epoch": 0.1603974449964514, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.16307418048381805, |
| "learning_rate": 5.5e-07, |
| "loss": 0.0427, |
| "num_tokens": 374014882.0, |
| "reward": 0.5, |
| "reward_std": 0.32015693187713623, |
| "rewards/simpleverify_reward/mean": 0.5, |
| "rewards/simpleverify_reward/std": 0.5004889965057373, |
| "step": 226 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.150390625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4078.0, |
| "completions/mean_length": 2285.427734375, |
| "completions/mean_terminated_length": 1964.9356689453125, |
| "completions/min_length": 320.0, |
| "completions/min_terminated_length": 320.0, |
| "epoch": 0.1611071682044003, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.22463363409042358, |
| "learning_rate": 5.48e-07, |
| "loss": 0.0235, |
| "num_tokens": 375229837.0, |
| "reward": 0.50390625, |
| "reward_std": 0.35051393508911133, |
| "rewards/simpleverify_reward/mean": 0.50390625, |
| "rewards/simpleverify_reward/std": 0.5004737377166748, |
| "step": 227 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.25, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4081.0, |
| "completions/mean_length": 2629.087890625, |
| "completions/mean_terminated_length": 2140.1171875, |
| "completions/min_length": 677.0, |
| "completions/min_terminated_length": 677.0, |
| "epoch": 0.16181689141234917, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.19044502079486847, |
| "learning_rate": 5.46e-07, |
| "loss": 0.0589, |
| "num_tokens": 376621914.0, |
| "reward": 0.546875, |
| "reward_std": 0.36336594820022583, |
| "rewards/simpleverify_reward/mean": 0.546875, |
| "rewards/simpleverify_reward/std": 0.4982847273349762, |
| "step": 228 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.259765625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4056.0, |
| "completions/mean_length": 2721.607421875, |
| "completions/mean_terminated_length": 2239.30078125, |
| "completions/min_length": 488.0, |
| "completions/min_terminated_length": 488.0, |
| "epoch": 0.16252661462029808, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.19006232917308807, |
| "learning_rate": 5.44e-07, |
| "loss": 0.0359, |
| "num_tokens": 378057921.0, |
| "reward": 0.5703125, |
| "reward_std": 0.3338164687156677, |
| "rewards/simpleverify_reward/mean": 0.5703125, |
| "rewards/simpleverify_reward/std": 0.4955156147480011, |
| "step": 229 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.23046875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4069.0, |
| "completions/mean_length": 2452.115234375, |
| "completions/mean_terminated_length": 1959.7841796875, |
| "completions/min_length": 541.0, |
| "completions/min_terminated_length": 541.0, |
| "epoch": 0.16323633782824698, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.1864621937274933, |
| "learning_rate": 5.420000000000001e-07, |
| "loss": 0.0332, |
| "num_tokens": 379358348.0, |
| "reward": 0.55078125, |
| "reward_std": 0.29035601019859314, |
| "rewards/simpleverify_reward/mean": 0.55078125, |
| "rewards/simpleverify_reward/std": 0.497901052236557, |
| "step": 230 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.3203125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4095.0, |
| "completions/mean_length": 2882.6015625, |
| "completions/mean_terminated_length": 2310.77001953125, |
| "completions/min_length": 879.0, |
| "completions/min_terminated_length": 879.0, |
| "epoch": 0.16394606103619588, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.1875680387020111, |
| "learning_rate": 5.4e-07, |
| "loss": 0.046, |
| "num_tokens": 380876960.0, |
| "reward": 0.51171875, |
| "reward_std": 0.3487931191921234, |
| "rewards/simpleverify_reward/mean": 0.51171875, |
| "rewards/simpleverify_reward/std": 0.5003514885902405, |
| "step": 231 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.216796875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4083.0, |
| "completions/mean_length": 2533.564453125, |
| "completions/mean_terminated_length": 2101.06982421875, |
| "completions/min_length": 508.0, |
| "completions/min_terminated_length": 508.0, |
| "epoch": 0.1646557842441448, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.20639567077159882, |
| "learning_rate": 5.38e-07, |
| "loss": 0.0449, |
| "num_tokens": 382232481.0, |
| "reward": 0.544921875, |
| "reward_std": 0.33554893732070923, |
| "rewards/simpleverify_reward/mean": 0.544921875, |
| "rewards/simpleverify_reward/std": 0.4984649419784546, |
| "step": 232 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.2421875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4090.0, |
| "completions/mean_length": 2716.578125, |
| "completions/mean_terminated_length": 2275.73193359375, |
| "completions/min_length": 706.0, |
| "completions/min_terminated_length": 706.0, |
| "epoch": 0.1653655074520937, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.19332976639270782, |
| "learning_rate": 5.36e-07, |
| "loss": 0.0519, |
| "num_tokens": 383671465.0, |
| "reward": 0.54296875, |
| "reward_std": 0.3540034890174866, |
| "rewards/simpleverify_reward/mean": 0.54296875, |
| "rewards/simpleverify_reward/std": 0.49863746762275696, |
| "step": 233 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.169921875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 2399.111328125, |
| "completions/mean_terminated_length": 2051.748046875, |
| "completions/min_length": 495.0, |
| "completions/min_terminated_length": 495.0, |
| "epoch": 0.1660752306600426, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.2186063975095749, |
| "learning_rate": 5.34e-07, |
| "loss": 0.045, |
| "num_tokens": 384942114.0, |
| "reward": 0.658203125, |
| "reward_std": 0.3039512038230896, |
| "rewards/simpleverify_reward/mean": 0.658203125, |
| "rewards/simpleverify_reward/std": 0.4747757613658905, |
| "step": 234 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.23046875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4093.0, |
| "completions/mean_length": 2722.666015625, |
| "completions/mean_terminated_length": 2311.36279296875, |
| "completions/min_length": 871.0, |
| "completions/min_terminated_length": 871.0, |
| "epoch": 0.16678495386799147, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.18104292452335358, |
| "learning_rate": 5.32e-07, |
| "loss": 0.0612, |
| "num_tokens": 386388743.0, |
| "reward": 0.548828125, |
| "reward_std": 0.36007893085479736, |
| "rewards/simpleverify_reward/mean": 0.548828125, |
| "rewards/simpleverify_reward/std": 0.498096764087677, |
| "step": 235 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.158203125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4084.0, |
| "completions/mean_length": 2583.5859375, |
| "completions/mean_terminated_length": 2299.350341796875, |
| "completions/min_length": 375.0, |
| "completions/min_terminated_length": 375.0, |
| "epoch": 0.16749467707594037, |
| "frac_reward_zero_std": 0.125, |
| "grad_norm": 0.19752179086208344, |
| "learning_rate": 5.3e-07, |
| "loss": 0.0283, |
| "num_tokens": 387754851.0, |
| "reward": 0.62109375, |
| "reward_std": 0.3685027062892914, |
| "rewards/simpleverify_reward/mean": 0.62109375, |
| "rewards/simpleverify_reward/std": 0.4855891764163971, |
| "step": 236 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.171875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4079.0, |
| "completions/mean_length": 2637.44140625, |
| "completions/mean_terminated_length": 2334.7216796875, |
| "completions/min_length": 874.0, |
| "completions/min_terminated_length": 874.0, |
| "epoch": 0.16820440028388928, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.21025946736335754, |
| "learning_rate": 5.28e-07, |
| "loss": 0.0404, |
| "num_tokens": 389152821.0, |
| "reward": 0.630859375, |
| "reward_std": 0.32404178380966187, |
| "rewards/simpleverify_reward/mean": 0.630859375, |
| "rewards/simpleverify_reward/std": 0.4830440282821655, |
| "step": 237 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.251953125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4091.0, |
| "completions/mean_length": 2708.689453125, |
| "completions/mean_terminated_length": 2241.423095703125, |
| "completions/min_length": 597.0, |
| "completions/min_terminated_length": 597.0, |
| "epoch": 0.16891412349183818, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.1909317970275879, |
| "learning_rate": 5.26e-07, |
| "loss": 0.0396, |
| "num_tokens": 390581446.0, |
| "reward": 0.486328125, |
| "reward_std": 0.33489710092544556, |
| "rewards/simpleverify_reward/mean": 0.486328125, |
| "rewards/simpleverify_reward/std": 0.5003018379211426, |
| "step": 238 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.224609375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4092.0, |
| "completions/mean_length": 2440.298828125, |
| "completions/mean_terminated_length": 1960.6876220703125, |
| "completions/min_length": 475.0, |
| "completions/min_terminated_length": 475.0, |
| "epoch": 0.16962384669978708, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.201446533203125, |
| "learning_rate": 5.24e-07, |
| "loss": 0.0511, |
| "num_tokens": 391875279.0, |
| "reward": 0.572265625, |
| "reward_std": 0.3391408920288086, |
| "rewards/simpleverify_reward/mean": 0.572265625, |
| "rewards/simpleverify_reward/std": 0.4952339828014374, |
| "step": 239 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.205078125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4081.0, |
| "completions/mean_length": 2616.126953125, |
| "completions/mean_terminated_length": 2234.341552734375, |
| "completions/min_length": 673.0, |
| "completions/min_terminated_length": 673.0, |
| "epoch": 0.170333569907736, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.19156606495380402, |
| "learning_rate": 5.22e-07, |
| "loss": 0.0296, |
| "num_tokens": 393262880.0, |
| "reward": 0.640625, |
| "reward_std": 0.34353888034820557, |
| "rewards/simpleverify_reward/mean": 0.640625, |
| "rewards/simpleverify_reward/std": 0.48028653860092163, |
| "step": 240 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.330078125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 2805.962890625, |
| "completions/mean_terminated_length": 2170.346923828125, |
| "completions/min_length": 526.0, |
| "completions/min_terminated_length": 526.0, |
| "epoch": 0.1710432931156849, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.20077036321163177, |
| "learning_rate": 5.2e-07, |
| "loss": 0.0556, |
| "num_tokens": 394749037.0, |
| "reward": 0.494140625, |
| "reward_std": 0.3329257071018219, |
| "rewards/simpleverify_reward/mean": 0.494140625, |
| "rewards/simpleverify_reward/std": 0.5004546642303467, |
| "step": 241 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.19921875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4074.0, |
| "completions/mean_length": 2371.22265625, |
| "completions/mean_terminated_length": 1942.1317138671875, |
| "completions/min_length": 477.0, |
| "completions/min_terminated_length": 477.0, |
| "epoch": 0.1717530163236338, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.20660369098186493, |
| "learning_rate": 5.18e-07, |
| "loss": 0.0326, |
| "num_tokens": 396003455.0, |
| "reward": 0.537109375, |
| "reward_std": 0.3553847670555115, |
| "rewards/simpleverify_reward/mean": 0.537109375, |
| "rewards/simpleverify_reward/std": 0.4991086423397064, |
| "step": 242 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.275390625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4067.0, |
| "completions/mean_length": 2811.521484375, |
| "completions/mean_terminated_length": 2323.350341796875, |
| "completions/min_length": 756.0, |
| "completions/min_terminated_length": 756.0, |
| "epoch": 0.17246273953158267, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.18414585292339325, |
| "learning_rate": 5.16e-07, |
| "loss": 0.019, |
| "num_tokens": 397487482.0, |
| "reward": 0.5078125, |
| "reward_std": 0.31871193647384644, |
| "rewards/simpleverify_reward/mean": 0.5078125, |
| "rewards/simpleverify_reward/std": 0.5004279017448425, |
| "step": 243 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.357421875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4018.0, |
| "completions/mean_length": 2778.607421875, |
| "completions/mean_terminated_length": 2045.8328857421875, |
| "completions/min_length": 621.0, |
| "completions/min_terminated_length": 621.0, |
| "epoch": 0.17317246273953157, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.19176249206066132, |
| "learning_rate": 5.14e-07, |
| "loss": 0.0443, |
| "num_tokens": 398957921.0, |
| "reward": 0.4921875, |
| "reward_std": 0.35048142075538635, |
| "rewards/simpleverify_reward/mean": 0.4921875, |
| "rewards/simpleverify_reward/std": 0.5004279017448425, |
| "step": 244 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.314453125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 2844.796875, |
| "completions/mean_terminated_length": 2270.88330078125, |
| "completions/min_length": 645.0, |
| "completions/min_terminated_length": 645.0, |
| "epoch": 0.17388218594748048, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.18036429584026337, |
| "learning_rate": 5.12e-07, |
| "loss": 0.0376, |
| "num_tokens": 400462329.0, |
| "reward": 0.4453125, |
| "reward_std": 0.35625144839286804, |
| "rewards/simpleverify_reward/mean": 0.4453125, |
| "rewards/simpleverify_reward/std": 0.49748632311820984, |
| "step": 245 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.3359375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4095.0, |
| "completions/mean_length": 2742.84765625, |
| "completions/mean_terminated_length": 2058.311767578125, |
| "completions/min_length": 679.0, |
| "completions/min_terminated_length": 679.0, |
| "epoch": 0.17459190915542938, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.17582790553569794, |
| "learning_rate": 5.1e-07, |
| "loss": 0.0394, |
| "num_tokens": 401911083.0, |
| "reward": 0.4609375, |
| "reward_std": 0.2817447781562805, |
| "rewards/simpleverify_reward/mean": 0.4609375, |
| "rewards/simpleverify_reward/std": 0.4989593029022217, |
| "step": 246 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.111328125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4065.0, |
| "completions/mean_length": 2113.640625, |
| "completions/mean_terminated_length": 1865.3011474609375, |
| "completions/min_length": 660.0, |
| "completions/min_terminated_length": 660.0, |
| "epoch": 0.17530163236337828, |
| "frac_reward_zero_std": 0.375, |
| "grad_norm": 0.1930147409439087, |
| "learning_rate": 5.079999999999999e-07, |
| "loss": 0.0271, |
| "num_tokens": 403037043.0, |
| "reward": 0.673828125, |
| "reward_std": 0.25939908623695374, |
| "rewards/simpleverify_reward/mean": 0.673828125, |
| "rewards/simpleverify_reward/std": 0.4692695140838623, |
| "step": 247 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.15625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4061.0, |
| "completions/mean_length": 2267.5234375, |
| "completions/mean_terminated_length": 1928.9166259765625, |
| "completions/min_length": 563.0, |
| "completions/min_terminated_length": 563.0, |
| "epoch": 0.1760113555713272, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.20930366218090057, |
| "learning_rate": 5.06e-07, |
| "loss": 0.0153, |
| "num_tokens": 404233951.0, |
| "reward": 0.6640625, |
| "reward_std": 0.3308311104774475, |
| "rewards/simpleverify_reward/mean": 0.6640625, |
| "rewards/simpleverify_reward/std": 0.4727790653705597, |
| "step": 248 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.2578125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4088.0, |
| "completions/mean_length": 2505.330078125, |
| "completions/mean_terminated_length": 1952.7816162109375, |
| "completions/min_length": 552.0, |
| "completions/min_terminated_length": 552.0, |
| "epoch": 0.1767210787792761, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.17469164729118347, |
| "learning_rate": 5.04e-07, |
| "loss": 0.0357, |
| "num_tokens": 405559688.0, |
| "reward": 0.533203125, |
| "reward_std": 0.2962723672389984, |
| "rewards/simpleverify_reward/mean": 0.533203125, |
| "rewards/simpleverify_reward/std": 0.4993842542171478, |
| "step": 249 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.166015625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4092.0, |
| "completions/mean_length": 2464.44921875, |
| "completions/mean_terminated_length": 2139.667236328125, |
| "completions/min_length": 538.0, |
| "completions/min_terminated_length": 538.0, |
| "epoch": 0.177430801987225, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.19963285326957703, |
| "learning_rate": 5.02e-07, |
| "loss": 0.0558, |
| "num_tokens": 406864158.0, |
| "reward": 0.59375, |
| "reward_std": 0.3067856431007385, |
| "rewards/simpleverify_reward/mean": 0.59375, |
| "rewards/simpleverify_reward/std": 0.49161264300346375, |
| "step": 250 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.220703125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4084.0, |
| "completions/mean_length": 2355.625, |
| "completions/mean_terminated_length": 1862.73681640625, |
| "completions/min_length": 533.0, |
| "completions/min_terminated_length": 533.0, |
| "epoch": 0.17814052519517387, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.21490077674388885, |
| "learning_rate": 5e-07, |
| "loss": 0.0476, |
| "num_tokens": 408115758.0, |
| "reward": 0.607421875, |
| "reward_std": 0.2897767126560211, |
| "rewards/simpleverify_reward/mean": 0.607421875, |
| "rewards/simpleverify_reward/std": 0.4888018071651459, |
| "step": 251 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.20703125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4087.0, |
| "completions/mean_length": 2619.30078125, |
| "completions/mean_terminated_length": 2233.758544921875, |
| "completions/min_length": 685.0, |
| "completions/min_terminated_length": 685.0, |
| "epoch": 0.17885024840312277, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.20267431437969208, |
| "learning_rate": 4.979999999999999e-07, |
| "loss": 0.0338, |
| "num_tokens": 409499848.0, |
| "reward": 0.509765625, |
| "reward_std": 0.3494358956813812, |
| "rewards/simpleverify_reward/mean": 0.509765625, |
| "rewards/simpleverify_reward/std": 0.5003935098648071, |
| "step": 252 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.1640625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4095.0, |
| "completions/mean_length": 2509.431640625, |
| "completions/mean_terminated_length": 2198.049072265625, |
| "completions/min_length": 342.0, |
| "completions/min_terminated_length": 342.0, |
| "epoch": 0.17955997161107168, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.18749865889549255, |
| "learning_rate": 4.96e-07, |
| "loss": 0.0281, |
| "num_tokens": 410826661.0, |
| "reward": 0.55078125, |
| "reward_std": 0.3471548855304718, |
| "rewards/simpleverify_reward/mean": 0.55078125, |
| "rewards/simpleverify_reward/std": 0.497901052236557, |
| "step": 253 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.205078125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4092.0, |
| "completions/mean_length": 2580.173828125, |
| "completions/mean_terminated_length": 2189.113037109375, |
| "completions/min_length": 579.0, |
| "completions/min_terminated_length": 579.0, |
| "epoch": 0.18026969481902058, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.1829977035522461, |
| "learning_rate": 4.94e-07, |
| "loss": 0.0416, |
| "num_tokens": 412192718.0, |
| "reward": 0.5390625, |
| "reward_std": 0.3347896337509155, |
| "rewards/simpleverify_reward/mean": 0.5390625, |
| "rewards/simpleverify_reward/std": 0.4989593029022217, |
| "step": 254 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.26171875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4091.0, |
| "completions/mean_length": 2735.30859375, |
| "completions/mean_terminated_length": 2252.947021484375, |
| "completions/min_length": 629.0, |
| "completions/min_terminated_length": 629.0, |
| "epoch": 0.18097941802696949, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.18470896780490875, |
| "learning_rate": 4.92e-07, |
| "loss": 0.0299, |
| "num_tokens": 413644060.0, |
| "reward": 0.451171875, |
| "reward_std": 0.3024138808250427, |
| "rewards/simpleverify_reward/mean": 0.451171875, |
| "rewards/simpleverify_reward/std": 0.498096764087677, |
| "step": 255 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.208984375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4067.0, |
| "completions/mean_length": 2707.5703125, |
| "completions/mean_terminated_length": 2340.750732421875, |
| "completions/min_length": 497.0, |
| "completions/min_terminated_length": 497.0, |
| "epoch": 0.1816891412349184, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.19147031009197235, |
| "learning_rate": 4.9e-07, |
| "loss": 0.0266, |
| "num_tokens": 415080144.0, |
| "reward": 0.541015625, |
| "reward_std": 0.3559653162956238, |
| "rewards/simpleverify_reward/mean": 0.541015625, |
| "rewards/simpleverify_reward/std": 0.49880221486091614, |
| "step": 256 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.30859375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4092.0, |
| "completions/mean_length": 2902.23046875, |
| "completions/mean_terminated_length": 2369.41796875, |
| "completions/min_length": 526.0, |
| "completions/min_terminated_length": 526.0, |
| "epoch": 0.1823988644428673, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.18606849014759064, |
| "learning_rate": 4.879999999999999e-07, |
| "loss": 0.0215, |
| "num_tokens": 416628278.0, |
| "reward": 0.494140625, |
| "reward_std": 0.3223326802253723, |
| "rewards/simpleverify_reward/mean": 0.494140625, |
| "rewards/simpleverify_reward/std": 0.5004546642303467, |
| "step": 257 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.193359375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4069.0, |
| "completions/mean_length": 2577.400390625, |
| "completions/mean_terminated_length": 2213.3779296875, |
| "completions/min_length": 467.0, |
| "completions/min_terminated_length": 467.0, |
| "epoch": 0.18310858765081617, |
| "frac_reward_zero_std": 0.09375, |
| "grad_norm": 0.24080772697925568, |
| "learning_rate": 4.86e-07, |
| "loss": 0.0216, |
| "num_tokens": 417997539.0, |
| "reward": 0.529296875, |
| "reward_std": 0.3667445182800293, |
| "rewards/simpleverify_reward/mean": 0.529296875, |
| "rewards/simpleverify_reward/std": 0.49962911009788513, |
| "step": 258 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.296875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 2800.078125, |
| "completions/mean_terminated_length": 2252.9111328125, |
| "completions/min_length": 348.0, |
| "completions/min_terminated_length": 348.0, |
| "epoch": 0.18381831085876507, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.1979677528142929, |
| "learning_rate": 4.839999999999999e-07, |
| "loss": 0.0517, |
| "num_tokens": 419488315.0, |
| "reward": 0.46484375, |
| "reward_std": 0.3152044415473938, |
| "rewards/simpleverify_reward/mean": 0.46484375, |
| "rewards/simpleverify_reward/std": 0.49925029277801514, |
| "step": 259 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.20703125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4085.0, |
| "completions/mean_length": 2540.7734375, |
| "completions/mean_terminated_length": 2134.72900390625, |
| "completions/min_length": 618.0, |
| "completions/min_terminated_length": 618.0, |
| "epoch": 0.18452803406671398, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.20918329060077667, |
| "learning_rate": 4.82e-07, |
| "loss": 0.0339, |
| "num_tokens": 420834887.0, |
| "reward": 0.625, |
| "reward_std": 0.34839698672294617, |
| "rewards/simpleverify_reward/mean": 0.625, |
| "rewards/simpleverify_reward/std": 0.4845963716506958, |
| "step": 260 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.248046875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4079.0, |
| "completions/mean_length": 2834.91015625, |
| "completions/mean_terminated_length": 2418.914306640625, |
| "completions/min_length": 478.0, |
| "completions/min_terminated_length": 478.0, |
| "epoch": 0.18523775727466288, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.18778075277805328, |
| "learning_rate": 4.8e-07, |
| "loss": 0.0484, |
| "num_tokens": 422327337.0, |
| "reward": 0.591796875, |
| "reward_std": 0.3483273386955261, |
| "rewards/simpleverify_reward/mean": 0.591796875, |
| "rewards/simpleverify_reward/std": 0.49198177456855774, |
| "step": 261 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.158203125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4015.0, |
| "completions/mean_length": 2259.638671875, |
| "completions/mean_terminated_length": 1914.52197265625, |
| "completions/min_length": 420.0, |
| "completions/min_terminated_length": 420.0, |
| "epoch": 0.18594748048261178, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.20406940579414368, |
| "learning_rate": 4.779999999999999e-07, |
| "loss": 0.0499, |
| "num_tokens": 423523584.0, |
| "reward": 0.58984375, |
| "reward_std": 0.283415824174881, |
| "rewards/simpleverify_reward/mean": 0.58984375, |
| "rewards/simpleverify_reward/std": 0.49234291911125183, |
| "step": 262 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.25390625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4091.0, |
| "completions/mean_length": 2559.3671875, |
| "completions/mean_terminated_length": 2036.4293212890625, |
| "completions/min_length": 387.0, |
| "completions/min_terminated_length": 387.0, |
| "epoch": 0.18665720369056069, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.19075004756450653, |
| "learning_rate": 4.76e-07, |
| "loss": 0.0474, |
| "num_tokens": 424878780.0, |
| "reward": 0.56640625, |
| "reward_std": 0.31149914860725403, |
| "rewards/simpleverify_reward/mean": 0.56640625, |
| "rewards/simpleverify_reward/std": 0.4960552453994751, |
| "step": 263 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.23046875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4085.0, |
| "completions/mean_length": 2559.291015625, |
| "completions/mean_terminated_length": 2099.058349609375, |
| "completions/min_length": 703.0, |
| "completions/min_terminated_length": 703.0, |
| "epoch": 0.1873669268985096, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.19555342197418213, |
| "learning_rate": 4.7399999999999993e-07, |
| "loss": 0.0173, |
| "num_tokens": 426232801.0, |
| "reward": 0.51953125, |
| "reward_std": 0.3292984962463379, |
| "rewards/simpleverify_reward/mean": 0.51953125, |
| "rewards/simpleverify_reward/std": 0.5001069903373718, |
| "step": 264 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.248046875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4042.0, |
| "completions/mean_length": 2456.095703125, |
| "completions/mean_terminated_length": 1915.1402587890625, |
| "completions/min_length": 460.0, |
| "completions/min_terminated_length": 460.0, |
| "epoch": 0.1880766501064585, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.19858834147453308, |
| "learning_rate": 4.7199999999999994e-07, |
| "loss": 0.0287, |
| "num_tokens": 427532130.0, |
| "reward": 0.4375, |
| "reward_std": 0.318639874458313, |
| "rewards/simpleverify_reward/mean": 0.4375, |
| "rewards/simpleverify_reward/std": 0.49656352400779724, |
| "step": 265 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.294921875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4057.0, |
| "completions/mean_length": 2659.427734375, |
| "completions/mean_terminated_length": 2058.53466796875, |
| "completions/min_length": 490.0, |
| "completions/min_terminated_length": 490.0, |
| "epoch": 0.18878637331440737, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.19035471975803375, |
| "learning_rate": 4.6999999999999995e-07, |
| "loss": 0.041, |
| "num_tokens": 428939869.0, |
| "reward": 0.423828125, |
| "reward_std": 0.32699698209762573, |
| "rewards/simpleverify_reward/mean": 0.423828125, |
| "rewards/simpleverify_reward/std": 0.4946470856666565, |
| "step": 266 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.283203125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4081.0, |
| "completions/mean_length": 2719.8828125, |
| "completions/mean_terminated_length": 2176.185302734375, |
| "completions/min_length": 679.0, |
| "completions/min_terminated_length": 679.0, |
| "epoch": 0.18949609652235627, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.20378349721431732, |
| "learning_rate": 4.68e-07, |
| "loss": 0.0496, |
| "num_tokens": 430378977.0, |
| "reward": 0.5625, |
| "reward_std": 0.35818740725517273, |
| "rewards/simpleverify_reward/mean": 0.5625, |
| "rewards/simpleverify_reward/std": 0.49656352400779724, |
| "step": 267 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.220703125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 2642.1328125, |
| "completions/mean_terminated_length": 2230.385986328125, |
| "completions/min_length": 469.0, |
| "completions/min_terminated_length": 469.0, |
| "epoch": 0.19020581973030518, |
| "frac_reward_zero_std": 0.125, |
| "grad_norm": 0.2166747897863388, |
| "learning_rate": 4.66e-07, |
| "loss": 0.0359, |
| "num_tokens": 431778613.0, |
| "reward": 0.59765625, |
| "reward_std": 0.3443273603916168, |
| "rewards/simpleverify_reward/mean": 0.59765625, |
| "rewards/simpleverify_reward/std": 0.4908501207828522, |
| "step": 268 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.220703125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4047.0, |
| "completions/mean_length": 2557.037109375, |
| "completions/mean_terminated_length": 2121.1904296875, |
| "completions/min_length": 619.0, |
| "completions/min_terminated_length": 619.0, |
| "epoch": 0.19091554293825408, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.20906975865364075, |
| "learning_rate": 4.64e-07, |
| "loss": 0.0529, |
| "num_tokens": 433135032.0, |
| "reward": 0.5859375, |
| "reward_std": 0.31893813610076904, |
| "rewards/simpleverify_reward/mean": 0.5859375, |
| "rewards/simpleverify_reward/std": 0.49304109811782837, |
| "step": 269 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.14453125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4092.0, |
| "completions/mean_length": 2290.63671875, |
| "completions/mean_terminated_length": 1985.6209716796875, |
| "completions/min_length": 64.0, |
| "completions/min_terminated_length": 64.0, |
| "epoch": 0.19162526614620298, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.21218395233154297, |
| "learning_rate": 4.62e-07, |
| "loss": 0.0597, |
| "num_tokens": 434353726.0, |
| "reward": 0.66796875, |
| "reward_std": 0.3565444350242615, |
| "rewards/simpleverify_reward/mean": 0.66796875, |
| "rewards/simpleverify_reward/std": 0.47140273451805115, |
| "step": 270 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.197265625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 2484.921875, |
| "completions/mean_terminated_length": 2089.01220703125, |
| "completions/min_length": 824.0, |
| "completions/min_terminated_length": 824.0, |
| "epoch": 0.1923349893541519, |
| "frac_reward_zero_std": 0.0625, |
| "grad_norm": 0.21874597668647766, |
| "learning_rate": 4.6e-07, |
| "loss": 0.023, |
| "num_tokens": 435670838.0, |
| "reward": 0.5703125, |
| "reward_std": 0.3861766755580902, |
| "rewards/simpleverify_reward/mean": 0.5703125, |
| "rewards/simpleverify_reward/std": 0.4955156147480011, |
| "step": 271 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.3203125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4091.0, |
| "completions/mean_length": 2785.115234375, |
| "completions/mean_terminated_length": 2167.342041015625, |
| "completions/min_length": 573.0, |
| "completions/min_terminated_length": 573.0, |
| "epoch": 0.1930447125621008, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.18055453896522522, |
| "learning_rate": 4.58e-07, |
| "loss": 0.0288, |
| "num_tokens": 437143137.0, |
| "reward": 0.521484375, |
| "reward_std": 0.26880234479904175, |
| "rewards/simpleverify_reward/mean": 0.521484375, |
| "rewards/simpleverify_reward/std": 0.5000267624855042, |
| "step": 272 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.173828125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 2388.83203125, |
| "completions/mean_terminated_length": 2029.640625, |
| "completions/min_length": 351.0, |
| "completions/min_terminated_length": 351.0, |
| "epoch": 0.1937544357700497, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.2116241157054901, |
| "learning_rate": 4.56e-07, |
| "loss": 0.0307, |
| "num_tokens": 438414155.0, |
| "reward": 0.53515625, |
| "reward_std": 0.3500208854675293, |
| "rewards/simpleverify_reward/mean": 0.53515625, |
| "rewards/simpleverify_reward/std": 0.49925029277801514, |
| "step": 273 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.162109375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4038.0, |
| "completions/mean_length": 2432.2265625, |
| "completions/mean_terminated_length": 2110.3310546875, |
| "completions/min_length": 424.0, |
| "completions/min_terminated_length": 424.0, |
| "epoch": 0.19446415897799857, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.244610995054245, |
| "learning_rate": 4.54e-07, |
| "loss": 0.0417, |
| "num_tokens": 439709199.0, |
| "reward": 0.623046875, |
| "reward_std": 0.3437129259109497, |
| "rewards/simpleverify_reward/mean": 0.623046875, |
| "rewards/simpleverify_reward/std": 0.4850969910621643, |
| "step": 274 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.115234375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4058.0, |
| "completions/mean_length": 2329.455078125, |
| "completions/mean_terminated_length": 2099.375244140625, |
| "completions/min_length": 550.0, |
| "completions/min_terminated_length": 550.0, |
| "epoch": 0.19517388218594747, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.20062685012817383, |
| "learning_rate": 4.5199999999999997e-07, |
| "loss": 0.0115, |
| "num_tokens": 440944792.0, |
| "reward": 0.623046875, |
| "reward_std": 0.2904953062534332, |
| "rewards/simpleverify_reward/mean": 0.623046875, |
| "rewards/simpleverify_reward/std": 0.4850969910621643, |
| "step": 275 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.251953125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4095.0, |
| "completions/mean_length": 2779.072265625, |
| "completions/mean_terminated_length": 2335.51171875, |
| "completions/min_length": 654.0, |
| "completions/min_terminated_length": 654.0, |
| "epoch": 0.19588360539389638, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.17759059369564056, |
| "learning_rate": 4.5e-07, |
| "loss": 0.0378, |
| "num_tokens": 442421981.0, |
| "reward": 0.521484375, |
| "reward_std": 0.3284127712249756, |
| "rewards/simpleverify_reward/mean": 0.521484375, |
| "rewards/simpleverify_reward/std": 0.5000267624855042, |
| "step": 276 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.21484375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4058.0, |
| "completions/mean_length": 2452.150390625, |
| "completions/mean_terminated_length": 2002.3406982421875, |
| "completions/min_length": 768.0, |
| "completions/min_terminated_length": 768.0, |
| "epoch": 0.19659332860184528, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.19238083064556122, |
| "learning_rate": 4.48e-07, |
| "loss": 0.0288, |
| "num_tokens": 443724026.0, |
| "reward": 0.546875, |
| "reward_std": 0.2994121313095093, |
| "rewards/simpleverify_reward/mean": 0.546875, |
| "rewards/simpleverify_reward/std": 0.4982847273349762, |
| "step": 277 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.2890625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4085.0, |
| "completions/mean_length": 2617.166015625, |
| "completions/mean_terminated_length": 2015.8819580078125, |
| "completions/min_length": 452.0, |
| "completions/min_terminated_length": 452.0, |
| "epoch": 0.19730305180979418, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.19540122151374817, |
| "learning_rate": 4.46e-07, |
| "loss": 0.037, |
| "num_tokens": 445111311.0, |
| "reward": 0.47265625, |
| "reward_std": 0.30521726608276367, |
| "rewards/simpleverify_reward/mean": 0.47265625, |
| "rewards/simpleverify_reward/std": 0.49974003434181213, |
| "step": 278 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.25390625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4087.0, |
| "completions/mean_length": 2559.7734375, |
| "completions/mean_terminated_length": 2036.973876953125, |
| "completions/min_length": 625.0, |
| "completions/min_terminated_length": 625.0, |
| "epoch": 0.1980127750177431, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.18888156116008759, |
| "learning_rate": 4.44e-07, |
| "loss": 0.0225, |
| "num_tokens": 446472715.0, |
| "reward": 0.541015625, |
| "reward_std": 0.29477423429489136, |
| "rewards/simpleverify_reward/mean": 0.541015625, |
| "rewards/simpleverify_reward/std": 0.49880221486091614, |
| "step": 279 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.24609375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4091.0, |
| "completions/mean_length": 2631.82421875, |
| "completions/mean_terminated_length": 2153.880859375, |
| "completions/min_length": 814.0, |
| "completions/min_terminated_length": 814.0, |
| "epoch": 0.198722498225692, |
| "frac_reward_zero_std": 0.09375, |
| "grad_norm": 0.21171967685222626, |
| "learning_rate": 4.4199999999999996e-07, |
| "loss": 0.057, |
| "num_tokens": 447862305.0, |
| "reward": 0.556640625, |
| "reward_std": 0.3798057436943054, |
| "rewards/simpleverify_reward/mean": 0.556640625, |
| "rewards/simpleverify_reward/std": 0.49726733565330505, |
| "step": 280 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.25390625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4096.0, |
| "completions/mean_length": 2800.28515625, |
| "completions/mean_terminated_length": 2359.335205078125, |
| "completions/min_length": 771.0, |
| "completions/min_terminated_length": 771.0, |
| "epoch": 0.1994322214336409, |
| "frac_reward_zero_std": 0.0625, |
| "grad_norm": 0.2009730488061905, |
| "learning_rate": 4.3999999999999997e-07, |
| "loss": 0.0253, |
| "num_tokens": 449343923.0, |
| "reward": 0.59375, |
| "reward_std": 0.3875921964645386, |
| "rewards/simpleverify_reward/mean": 0.59375, |
| "rewards/simpleverify_reward/std": 0.49161264300346375, |
| "step": 281 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.240234375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4095.0, |
| "completions/mean_length": 2718.685546875, |
| "completions/mean_terminated_length": 2283.18505859375, |
| "completions/min_length": 568.0, |
| "completions/min_terminated_length": 568.0, |
| "epoch": 0.20014194464158977, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.1929699331521988, |
| "learning_rate": 4.38e-07, |
| "loss": 0.0522, |
| "num_tokens": 450780962.0, |
| "reward": 0.548828125, |
| "reward_std": 0.2961171269416809, |
| "rewards/simpleverify_reward/mean": 0.548828125, |
| "rewards/simpleverify_reward/std": 0.498096764087677, |
| "step": 282 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.134765625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4056.0, |
| "completions/mean_length": 2249.1796875, |
| "completions/mean_terminated_length": 1961.5260009765625, |
| "completions/min_length": 449.0, |
| "completions/min_terminated_length": 449.0, |
| "epoch": 0.20085166784953867, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.2232760637998581, |
| "learning_rate": 4.36e-07, |
| "loss": 0.0587, |
| "num_tokens": 451965086.0, |
| "reward": 0.64453125, |
| "reward_std": 0.35176533460617065, |
| "rewards/simpleverify_reward/mean": 0.64453125, |
| "rewards/simpleverify_reward/std": 0.47912323474884033, |
| "step": 283 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.224609375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4032.0, |
| "completions/mean_length": 2568.193359375, |
| "completions/mean_terminated_length": 2125.629638671875, |
| "completions/min_length": 600.0, |
| "completions/min_terminated_length": 600.0, |
| "epoch": 0.20156139105748758, |
| "frac_reward_zero_std": 0.125, |
| "grad_norm": 0.21754631400108337, |
| "learning_rate": 4.34e-07, |
| "loss": 0.0412, |
| "num_tokens": 453326561.0, |
| "reward": 0.572265625, |
| "reward_std": 0.33707985281944275, |
| "rewards/simpleverify_reward/mean": 0.572265625, |
| "rewards/simpleverify_reward/std": 0.4952339828014374, |
| "step": 284 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.154296875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4087.0, |
| "completions/mean_length": 2504.978515625, |
| "completions/mean_terminated_length": 2214.69970703125, |
| "completions/min_length": 575.0, |
| "completions/min_terminated_length": 575.0, |
| "epoch": 0.20227111426543648, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.1915304958820343, |
| "learning_rate": 4.3199999999999995e-07, |
| "loss": 0.0234, |
| "num_tokens": 454657158.0, |
| "reward": 0.55078125, |
| "reward_std": 0.3348847031593323, |
| "rewards/simpleverify_reward/mean": 0.55078125, |
| "rewards/simpleverify_reward/std": 0.497901052236557, |
| "step": 285 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.314453125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 2776.314453125, |
| "completions/mean_terminated_length": 2170.988525390625, |
| "completions/min_length": 553.0, |
| "completions/min_terminated_length": 553.0, |
| "epoch": 0.20298083747338538, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.19131878018379211, |
| "learning_rate": 4.2999999999999996e-07, |
| "loss": 0.0269, |
| "num_tokens": 456120151.0, |
| "reward": 0.44140625, |
| "reward_std": 0.3147798180580139, |
| "rewards/simpleverify_reward/mean": 0.44140625, |
| "rewards/simpleverify_reward/std": 0.4970405399799347, |
| "step": 286 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.251953125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4083.0, |
| "completions/mean_length": 2559.33984375, |
| "completions/mean_terminated_length": 2041.770263671875, |
| "completions/min_length": 687.0, |
| "completions/min_terminated_length": 687.0, |
| "epoch": 0.2036905606813343, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.2084338217973709, |
| "learning_rate": 4.2799999999999997e-07, |
| "loss": 0.06, |
| "num_tokens": 457473317.0, |
| "reward": 0.513671875, |
| "reward_std": 0.33566099405288696, |
| "rewards/simpleverify_reward/mean": 0.513671875, |
| "rewards/simpleverify_reward/std": 0.5003018379211426, |
| "step": 287 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.23828125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4088.0, |
| "completions/mean_length": 2596.2421875, |
| "completions/mean_terminated_length": 2127.087158203125, |
| "completions/min_length": 648.0, |
| "completions/min_terminated_length": 648.0, |
| "epoch": 0.2044002838892832, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.1931682676076889, |
| "learning_rate": 4.26e-07, |
| "loss": 0.0294, |
| "num_tokens": 458844049.0, |
| "reward": 0.544921875, |
| "reward_std": 0.31266218423843384, |
| "rewards/simpleverify_reward/mean": 0.544921875, |
| "rewards/simpleverify_reward/std": 0.4984649419784546, |
| "step": 288 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.20703125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4093.0, |
| "completions/mean_length": 2625.65234375, |
| "completions/mean_terminated_length": 2241.7685546875, |
| "completions/min_length": 667.0, |
| "completions/min_terminated_length": 667.0, |
| "epoch": 0.20511000709723207, |
| "frac_reward_zero_std": 0.09375, |
| "grad_norm": 0.21760718524456024, |
| "learning_rate": 4.24e-07, |
| "loss": 0.051, |
| "num_tokens": 460239087.0, |
| "reward": 0.6015625, |
| "reward_std": 0.344471275806427, |
| "rewards/simpleverify_reward/mean": 0.6015625, |
| "rewards/simpleverify_reward/std": 0.4900552034378052, |
| "step": 289 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.3125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4085.0, |
| "completions/mean_length": 2729.314453125, |
| "completions/mean_terminated_length": 2108.09375, |
| "completions/min_length": 498.0, |
| "completions/min_terminated_length": 498.0, |
| "epoch": 0.20581973030518097, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.20011965930461884, |
| "learning_rate": 4.2199999999999994e-07, |
| "loss": 0.0346, |
| "num_tokens": 461686592.0, |
| "reward": 0.40625, |
| "reward_std": 0.3546728491783142, |
| "rewards/simpleverify_reward/mean": 0.40625, |
| "rewards/simpleverify_reward/std": 0.49161264300346375, |
| "step": 290 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.267578125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4090.0, |
| "completions/mean_length": 2735.642578125, |
| "completions/mean_terminated_length": 2238.65869140625, |
| "completions/min_length": 707.0, |
| "completions/min_terminated_length": 707.0, |
| "epoch": 0.20652945351312987, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.18933384120464325, |
| "learning_rate": 4.1999999999999995e-07, |
| "loss": 0.0469, |
| "num_tokens": 463133753.0, |
| "reward": 0.505859375, |
| "reward_std": 0.3232646882534027, |
| "rewards/simpleverify_reward/mean": 0.505859375, |
| "rewards/simpleverify_reward/std": 0.5004546642303467, |
| "step": 291 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.23828125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4053.0, |
| "completions/mean_length": 2769.330078125, |
| "completions/mean_terminated_length": 2354.320556640625, |
| "completions/min_length": 834.0, |
| "completions/min_terminated_length": 834.0, |
| "epoch": 0.20723917672107878, |
| "frac_reward_zero_std": 0.09375, |
| "grad_norm": 0.19974802434444427, |
| "learning_rate": 4.1799999999999996e-07, |
| "loss": 0.0552, |
| "num_tokens": 464600146.0, |
| "reward": 0.501953125, |
| "reward_std": 0.3730611205101013, |
| "rewards/simpleverify_reward/mean": 0.501953125, |
| "rewards/simpleverify_reward/std": 0.5004851818084717, |
| "step": 292 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.201171875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4073.0, |
| "completions/mean_length": 2459.93359375, |
| "completions/mean_terminated_length": 2047.9169921875, |
| "completions/min_length": 669.0, |
| "completions/min_terminated_length": 669.0, |
| "epoch": 0.20794889992902768, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.19120380282402039, |
| "learning_rate": 4.1599999999999997e-07, |
| "loss": 0.0544, |
| "num_tokens": 465907024.0, |
| "reward": 0.580078125, |
| "reward_std": 0.36129409074783325, |
| "rewards/simpleverify_reward/mean": 0.580078125, |
| "rewards/simpleverify_reward/std": 0.4940285086631775, |
| "step": 293 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.193359375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 2386.83984375, |
| "completions/mean_terminated_length": 1977.1380615234375, |
| "completions/min_length": 534.0, |
| "completions/min_terminated_length": 534.0, |
| "epoch": 0.20865862313697658, |
| "frac_reward_zero_std": 0.09375, |
| "grad_norm": 0.23504528403282166, |
| "learning_rate": 4.14e-07, |
| "loss": 0.05, |
| "num_tokens": 467172910.0, |
| "reward": 0.58984375, |
| "reward_std": 0.3808078169822693, |
| "rewards/simpleverify_reward/mean": 0.58984375, |
| "rewards/simpleverify_reward/std": 0.49234291911125183, |
| "step": 294 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.2109375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4091.0, |
| "completions/mean_length": 2502.728515625, |
| "completions/mean_terminated_length": 2076.804443359375, |
| "completions/min_length": 491.0, |
| "completions/min_terminated_length": 491.0, |
| "epoch": 0.2093683463449255, |
| "frac_reward_zero_std": 0.09375, |
| "grad_norm": 0.22503319382667542, |
| "learning_rate": 4.12e-07, |
| "loss": 0.0362, |
| "num_tokens": 468501395.0, |
| "reward": 0.560546875, |
| "reward_std": 0.38168463110923767, |
| "rewards/simpleverify_reward/mean": 0.560546875, |
| "rewards/simpleverify_reward/std": 0.49680593609809875, |
| "step": 295 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.22265625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4096.0, |
| "completions/mean_length": 2501.8984375, |
| "completions/mean_terminated_length": 2045.2965087890625, |
| "completions/min_length": 288.0, |
| "completions/min_terminated_length": 288.0, |
| "epoch": 0.2100780695528744, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.19616208970546722, |
| "learning_rate": 4.0999999999999994e-07, |
| "loss": 0.0294, |
| "num_tokens": 469824751.0, |
| "reward": 0.505859375, |
| "reward_std": 0.34041157364845276, |
| "rewards/simpleverify_reward/mean": 0.505859375, |
| "rewards/simpleverify_reward/std": 0.5004546642303467, |
| "step": 296 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.298828125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4091.0, |
| "completions/mean_length": 2777.767578125, |
| "completions/mean_terminated_length": 2215.958251953125, |
| "completions/min_length": 322.0, |
| "completions/min_terminated_length": 322.0, |
| "epoch": 0.21078779276082327, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.21250969171524048, |
| "learning_rate": 4.0799999999999995e-07, |
| "loss": 0.0485, |
| "num_tokens": 471290408.0, |
| "reward": 0.54296875, |
| "reward_std": 0.3311998248100281, |
| "rewards/simpleverify_reward/mean": 0.54296875, |
| "rewards/simpleverify_reward/std": 0.49863746762275696, |
| "step": 297 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.2109375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4091.0, |
| "completions/mean_length": 2423.396484375, |
| "completions/mean_terminated_length": 1976.2647705078125, |
| "completions/min_length": 394.0, |
| "completions/min_terminated_length": 394.0, |
| "epoch": 0.21149751596877217, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.20817621052265167, |
| "learning_rate": 4.06e-07, |
| "loss": 0.046, |
| "num_tokens": 472573651.0, |
| "reward": 0.580078125, |
| "reward_std": 0.3214206397533417, |
| "rewards/simpleverify_reward/mean": 0.580078125, |
| "rewards/simpleverify_reward/std": 0.4940285086631775, |
| "step": 298 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.244140625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4087.0, |
| "completions/mean_length": 2429.607421875, |
| "completions/mean_terminated_length": 1891.366943359375, |
| "completions/min_length": 134.0, |
| "completions/min_terminated_length": 134.0, |
| "epoch": 0.21220723917672107, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.20557406544685364, |
| "learning_rate": 4.04e-07, |
| "loss": 0.0273, |
| "num_tokens": 473865114.0, |
| "reward": 0.515625, |
| "reward_std": 0.28485965728759766, |
| "rewards/simpleverify_reward/mean": 0.515625, |
| "rewards/simpleverify_reward/std": 0.5002445578575134, |
| "step": 299 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.197265625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4086.0, |
| "completions/mean_length": 2523.693359375, |
| "completions/mean_terminated_length": 2137.3115234375, |
| "completions/min_length": 511.0, |
| "completions/min_terminated_length": 511.0, |
| "epoch": 0.21291696238466998, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.2058168649673462, |
| "learning_rate": 4.02e-07, |
| "loss": 0.0221, |
| "num_tokens": 475198365.0, |
| "reward": 0.470703125, |
| "reward_std": 0.3237093687057495, |
| "rewards/simpleverify_reward/mean": 0.470703125, |
| "rewards/simpleverify_reward/std": 0.49962911009788513, |
| "step": 300 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.34765625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 2936.412109375, |
| "completions/mean_terminated_length": 2318.42822265625, |
| "completions/min_length": 865.0, |
| "completions/min_terminated_length": 865.0, |
| "epoch": 0.21362668559261888, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.17465318739414215, |
| "learning_rate": 4e-07, |
| "loss": 0.0441, |
| "num_tokens": 476748688.0, |
| "reward": 0.4765625, |
| "reward_std": 0.30614206194877625, |
| "rewards/simpleverify_reward/mean": 0.4765625, |
| "rewards/simpleverify_reward/std": 0.49993884563446045, |
| "step": 301 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.1875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4055.0, |
| "completions/mean_length": 2463.22265625, |
| "completions/mean_terminated_length": 2086.427978515625, |
| "completions/min_length": 536.0, |
| "completions/min_terminated_length": 536.0, |
| "epoch": 0.21433640880056778, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.18531689047813416, |
| "learning_rate": 3.98e-07, |
| "loss": 0.0256, |
| "num_tokens": 478062850.0, |
| "reward": 0.541015625, |
| "reward_std": 0.3139249086380005, |
| "rewards/simpleverify_reward/mean": 0.541015625, |
| "rewards/simpleverify_reward/std": 0.49880221486091614, |
| "step": 302 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.212890625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4074.0, |
| "completions/mean_length": 2592.806640625, |
| "completions/mean_terminated_length": 2186.235595703125, |
| "completions/min_length": 674.0, |
| "completions/min_terminated_length": 674.0, |
| "epoch": 0.2150461320085167, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.20510566234588623, |
| "learning_rate": 3.96e-07, |
| "loss": 0.0695, |
| "num_tokens": 479429599.0, |
| "reward": 0.6015625, |
| "reward_std": 0.35381215810775757, |
| "rewards/simpleverify_reward/mean": 0.6015625, |
| "rewards/simpleverify_reward/std": 0.4900552034378052, |
| "step": 303 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.212890625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4083.0, |
| "completions/mean_length": 2612.115234375, |
| "completions/mean_terminated_length": 2210.766845703125, |
| "completions/min_length": 283.0, |
| "completions/min_terminated_length": 283.0, |
| "epoch": 0.2157558552164656, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.18846213817596436, |
| "learning_rate": 3.94e-07, |
| "loss": 0.0364, |
| "num_tokens": 480811850.0, |
| "reward": 0.5546875, |
| "reward_std": 0.3335626423358917, |
| "rewards/simpleverify_reward/mean": 0.5546875, |
| "rewards/simpleverify_reward/std": 0.49748632311820984, |
| "step": 304 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.3125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4082.0, |
| "completions/mean_length": 2702.5, |
| "completions/mean_terminated_length": 2069.091064453125, |
| "completions/min_length": 476.0, |
| "completions/min_terminated_length": 476.0, |
| "epoch": 0.21646557842441447, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.19156986474990845, |
| "learning_rate": 3.92e-07, |
| "loss": 0.034, |
| "num_tokens": 482236858.0, |
| "reward": 0.451171875, |
| "reward_std": 0.2684345543384552, |
| "rewards/simpleverify_reward/mean": 0.451171875, |
| "rewards/simpleverify_reward/std": 0.498096764087677, |
| "step": 305 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.19140625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4088.0, |
| "completions/mean_length": 2485.8046875, |
| "completions/mean_terminated_length": 2104.647216796875, |
| "completions/min_length": 590.0, |
| "completions/min_terminated_length": 590.0, |
| "epoch": 0.21717530163236337, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.18935483694076538, |
| "learning_rate": 3.8999999999999997e-07, |
| "loss": 0.0488, |
| "num_tokens": 483560038.0, |
| "reward": 0.6484375, |
| "reward_std": 0.29434189200401306, |
| "rewards/simpleverify_reward/mean": 0.6484375, |
| "rewards/simpleverify_reward/std": 0.4779251217842102, |
| "step": 306 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.287109375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4093.0, |
| "completions/mean_length": 2552.220703125, |
| "completions/mean_terminated_length": 1930.4794921875, |
| "completions/min_length": 341.0, |
| "completions/min_terminated_length": 341.0, |
| "epoch": 0.21788502484031227, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.19606468081474304, |
| "learning_rate": 3.88e-07, |
| "loss": 0.0389, |
| "num_tokens": 484923991.0, |
| "reward": 0.40625, |
| "reward_std": 0.30811700224876404, |
| "rewards/simpleverify_reward/mean": 0.40625, |
| "rewards/simpleverify_reward/std": 0.49161264300346375, |
| "step": 307 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.138671875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4093.0, |
| "completions/mean_length": 2208.017578125, |
| "completions/mean_terminated_length": 1904.0567626953125, |
| "completions/min_length": 433.0, |
| "completions/min_terminated_length": 433.0, |
| "epoch": 0.21859474804826118, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.21783199906349182, |
| "learning_rate": 3.86e-07, |
| "loss": 0.0384, |
| "num_tokens": 486098096.0, |
| "reward": 0.5, |
| "reward_std": 0.30791789293289185, |
| "rewards/simpleverify_reward/mean": 0.5, |
| "rewards/simpleverify_reward/std": 0.5004889965057373, |
| "step": 308 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.177734375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4043.0, |
| "completions/mean_length": 2350.04296875, |
| "completions/mean_terminated_length": 1972.65087890625, |
| "completions/min_length": 610.0, |
| "completions/min_terminated_length": 610.0, |
| "epoch": 0.21930447125621008, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.21395067870616913, |
| "learning_rate": 3.84e-07, |
| "loss": 0.0659, |
| "num_tokens": 487350358.0, |
| "reward": 0.6171875, |
| "reward_std": 0.31130343675613403, |
| "rewards/simpleverify_reward/mean": 0.6171875, |
| "rewards/simpleverify_reward/std": 0.486548513174057, |
| "step": 309 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.224609375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4064.0, |
| "completions/mean_length": 2562.32421875, |
| "completions/mean_terminated_length": 2118.060302734375, |
| "completions/min_length": 486.0, |
| "completions/min_terminated_length": 486.0, |
| "epoch": 0.22001419446415899, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.20008878409862518, |
| "learning_rate": 3.82e-07, |
| "loss": 0.0598, |
| "num_tokens": 488713660.0, |
| "reward": 0.568359375, |
| "reward_std": 0.32896625995635986, |
| "rewards/simpleverify_reward/mean": 0.568359375, |
| "rewards/simpleverify_reward/std": 0.4957893490791321, |
| "step": 310 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.244140625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4046.0, |
| "completions/mean_length": 2519.048828125, |
| "completions/mean_terminated_length": 2009.69775390625, |
| "completions/min_length": 519.0, |
| "completions/min_terminated_length": 519.0, |
| "epoch": 0.2207239176721079, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.20936205983161926, |
| "learning_rate": 3.7999999999999996e-07, |
| "loss": 0.0477, |
| "num_tokens": 490053189.0, |
| "reward": 0.5546875, |
| "reward_std": 0.33175498247146606, |
| "rewards/simpleverify_reward/mean": 0.5546875, |
| "rewards/simpleverify_reward/std": 0.49748632311820984, |
| "step": 311 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.33203125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4083.0, |
| "completions/mean_length": 2858.96484375, |
| "completions/mean_terminated_length": 2244.064453125, |
| "completions/min_length": 544.0, |
| "completions/min_terminated_length": 544.0, |
| "epoch": 0.22143364088005676, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.1876993179321289, |
| "learning_rate": 3.7799999999999997e-07, |
| "loss": 0.0518, |
| "num_tokens": 491568707.0, |
| "reward": 0.490234375, |
| "reward_std": 0.30799734592437744, |
| "rewards/simpleverify_reward/mean": 0.490234375, |
| "rewards/simpleverify_reward/std": 0.5003935098648071, |
| "step": 312 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.189453125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4081.0, |
| "completions/mean_length": 2328.810546875, |
| "completions/mean_terminated_length": 1915.7567138671875, |
| "completions/min_length": 591.0, |
| "completions/min_terminated_length": 591.0, |
| "epoch": 0.22214336408800567, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.21870994567871094, |
| "learning_rate": 3.76e-07, |
| "loss": 0.0349, |
| "num_tokens": 492807954.0, |
| "reward": 0.576171875, |
| "reward_std": 0.34467804431915283, |
| "rewards/simpleverify_reward/mean": 0.576171875, |
| "rewards/simpleverify_reward/std": 0.4946470856666565, |
| "step": 313 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.28125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4091.0, |
| "completions/mean_length": 2511.748046875, |
| "completions/mean_terminated_length": 1891.8233642578125, |
| "completions/min_length": 390.0, |
| "completions/min_terminated_length": 390.0, |
| "epoch": 0.22285308729595457, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.18180464208126068, |
| "learning_rate": 3.74e-07, |
| "loss": 0.0385, |
| "num_tokens": 494139697.0, |
| "reward": 0.482421875, |
| "reward_std": 0.3053285479545593, |
| "rewards/simpleverify_reward/mean": 0.482421875, |
| "rewards/simpleverify_reward/std": 0.5001795887947083, |
| "step": 314 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.212890625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 2587.30078125, |
| "completions/mean_terminated_length": 2179.24072265625, |
| "completions/min_length": 605.0, |
| "completions/min_terminated_length": 605.0, |
| "epoch": 0.22356281050390348, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.21861045062541962, |
| "learning_rate": 3.72e-07, |
| "loss": 0.0432, |
| "num_tokens": 495509771.0, |
| "reward": 0.58203125, |
| "reward_std": 0.320464551448822, |
| "rewards/simpleverify_reward/mean": 0.58203125, |
| "rewards/simpleverify_reward/std": 0.4937073290348053, |
| "step": 315 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.2109375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4082.0, |
| "completions/mean_length": 2631.041015625, |
| "completions/mean_terminated_length": 2239.418212890625, |
| "completions/min_length": 694.0, |
| "completions/min_terminated_length": 694.0, |
| "epoch": 0.22427253371185238, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.20442113280296326, |
| "learning_rate": 3.7e-07, |
| "loss": 0.0477, |
| "num_tokens": 496904976.0, |
| "reward": 0.640625, |
| "reward_std": 0.31070080399513245, |
| "rewards/simpleverify_reward/mean": 0.640625, |
| "rewards/simpleverify_reward/std": 0.48028653860092163, |
| "step": 316 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.1953125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4060.0, |
| "completions/mean_length": 2392.9453125, |
| "completions/mean_terminated_length": 1979.58251953125, |
| "completions/min_length": 556.0, |
| "completions/min_terminated_length": 556.0, |
| "epoch": 0.22498225691980128, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.2131560742855072, |
| "learning_rate": 3.6799999999999996e-07, |
| "loss": 0.0352, |
| "num_tokens": 498179060.0, |
| "reward": 0.59765625, |
| "reward_std": 0.2884575128555298, |
| "rewards/simpleverify_reward/mean": 0.59765625, |
| "rewards/simpleverify_reward/std": 0.4908501207828522, |
| "step": 317 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.255859375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4090.0, |
| "completions/mean_length": 2674.9609375, |
| "completions/mean_terminated_length": 2186.3623046875, |
| "completions/min_length": 555.0, |
| "completions/min_terminated_length": 555.0, |
| "epoch": 0.22569198012775019, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.1953941434621811, |
| "learning_rate": 3.6599999999999997e-07, |
| "loss": 0.0638, |
| "num_tokens": 499595536.0, |
| "reward": 0.51171875, |
| "reward_std": 0.3371179699897766, |
| "rewards/simpleverify_reward/mean": 0.51171875, |
| "rewards/simpleverify_reward/std": 0.5003514885902405, |
| "step": 318 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.18359375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4044.0, |
| "completions/mean_length": 2494.587890625, |
| "completions/mean_terminated_length": 2134.461669921875, |
| "completions/min_length": 511.0, |
| "completions/min_terminated_length": 511.0, |
| "epoch": 0.2264017033356991, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.2018555849790573, |
| "learning_rate": 3.64e-07, |
| "loss": 0.0438, |
| "num_tokens": 500920381.0, |
| "reward": 0.642578125, |
| "reward_std": 0.345365047454834, |
| "rewards/simpleverify_reward/mean": 0.642578125, |
| "rewards/simpleverify_reward/std": 0.4797092080116272, |
| "step": 319 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.208984375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4074.0, |
| "completions/mean_length": 2400.3828125, |
| "completions/mean_terminated_length": 1952.405029296875, |
| "completions/min_length": 641.0, |
| "completions/min_terminated_length": 641.0, |
| "epoch": 0.22711142654364797, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.1993652582168579, |
| "learning_rate": 3.62e-07, |
| "loss": 0.0194, |
| "num_tokens": 502192385.0, |
| "reward": 0.556640625, |
| "reward_std": 0.29038089513778687, |
| "rewards/simpleverify_reward/mean": 0.556640625, |
| "rewards/simpleverify_reward/std": 0.49726733565330505, |
| "step": 320 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.201171875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4095.0, |
| "completions/mean_length": 2490.8828125, |
| "completions/mean_terminated_length": 2086.66015625, |
| "completions/min_length": 612.0, |
| "completions/min_terminated_length": 612.0, |
| "epoch": 0.22782114975159687, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.19828158617019653, |
| "learning_rate": 3.6e-07, |
| "loss": 0.0414, |
| "num_tokens": 503504629.0, |
| "reward": 0.5546875, |
| "reward_std": 0.3170880675315857, |
| "rewards/simpleverify_reward/mean": 0.5546875, |
| "rewards/simpleverify_reward/std": 0.49748632311820984, |
| "step": 321 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.291015625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4074.0, |
| "completions/mean_length": 2914.22265625, |
| "completions/mean_terminated_length": 2429.140625, |
| "completions/min_length": 788.0, |
| "completions/min_terminated_length": 788.0, |
| "epoch": 0.22853087295954577, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.17794601619243622, |
| "learning_rate": 3.5799999999999995e-07, |
| "loss": 0.0287, |
| "num_tokens": 505046071.0, |
| "reward": 0.501953125, |
| "reward_std": 0.3345019817352295, |
| "rewards/simpleverify_reward/mean": 0.501953125, |
| "rewards/simpleverify_reward/std": 0.5004851818084717, |
| "step": 322 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.224609375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4096.0, |
| "completions/mean_length": 2575.568359375, |
| "completions/mean_terminated_length": 2135.140869140625, |
| "completions/min_length": 389.0, |
| "completions/min_terminated_length": 389.0, |
| "epoch": 0.22924059616749468, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.19339078664779663, |
| "learning_rate": 3.5599999999999996e-07, |
| "loss": 0.0234, |
| "num_tokens": 506414410.0, |
| "reward": 0.55859375, |
| "reward_std": 0.3279615044593811, |
| "rewards/simpleverify_reward/mean": 0.55859375, |
| "rewards/simpleverify_reward/std": 0.4970405399799347, |
| "step": 323 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.2421875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4059.0, |
| "completions/mean_length": 2568.919921875, |
| "completions/mean_terminated_length": 2080.884033203125, |
| "completions/min_length": 746.0, |
| "completions/min_terminated_length": 746.0, |
| "epoch": 0.22995031937544358, |
| "frac_reward_zero_std": 0.09375, |
| "grad_norm": 0.2050626575946808, |
| "learning_rate": 3.5399999999999997e-07, |
| "loss": 0.0576, |
| "num_tokens": 507779953.0, |
| "reward": 0.572265625, |
| "reward_std": 0.3650937080383301, |
| "rewards/simpleverify_reward/mean": 0.572265625, |
| "rewards/simpleverify_reward/std": 0.4952339828014374, |
| "step": 324 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.232421875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4076.0, |
| "completions/mean_length": 2719.3203125, |
| "completions/mean_terminated_length": 2302.463134765625, |
| "completions/min_length": 569.0, |
| "completions/min_terminated_length": 569.0, |
| "epoch": 0.23066004258339248, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.18556873500347137, |
| "learning_rate": 3.52e-07, |
| "loss": 0.0349, |
| "num_tokens": 509215045.0, |
| "reward": 0.513671875, |
| "reward_std": 0.32115447521209717, |
| "rewards/simpleverify_reward/mean": 0.513671875, |
| "rewards/simpleverify_reward/std": 0.5003018379211426, |
| "step": 325 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.228515625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4079.0, |
| "completions/mean_length": 2713.556640625, |
| "completions/mean_terminated_length": 2304.073486328125, |
| "completions/min_length": 606.0, |
| "completions/min_terminated_length": 606.0, |
| "epoch": 0.2313697657913414, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.19560633599758148, |
| "learning_rate": 3.5e-07, |
| "loss": 0.0578, |
| "num_tokens": 510652018.0, |
| "reward": 0.5625, |
| "reward_std": 0.34959104657173157, |
| "rewards/simpleverify_reward/mean": 0.5625, |
| "rewards/simpleverify_reward/std": 0.49656352400779724, |
| "step": 326 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.154296875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4092.0, |
| "completions/mean_length": 2246.021484375, |
| "completions/mean_terminated_length": 1908.49658203125, |
| "completions/min_length": 541.0, |
| "completions/min_terminated_length": 541.0, |
| "epoch": 0.2320794889992903, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.2100205421447754, |
| "learning_rate": 3.4799999999999994e-07, |
| "loss": 0.0403, |
| "num_tokens": 511848077.0, |
| "reward": 0.650390625, |
| "reward_std": 0.3186723589897156, |
| "rewards/simpleverify_reward/mean": 0.650390625, |
| "rewards/simpleverify_reward/std": 0.47731292247772217, |
| "step": 327 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.154296875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4096.0, |
| "completions/mean_length": 2353.546875, |
| "completions/mean_terminated_length": 2035.6397705078125, |
| "completions/min_length": 485.0, |
| "completions/min_terminated_length": 485.0, |
| "epoch": 0.23278921220723917, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.21640923619270325, |
| "learning_rate": 3.4599999999999995e-07, |
| "loss": 0.0245, |
| "num_tokens": 513102949.0, |
| "reward": 0.5625, |
| "reward_std": 0.3023741543292999, |
| "rewards/simpleverify_reward/mean": 0.5625, |
| "rewards/simpleverify_reward/std": 0.49656352400779724, |
| "step": 328 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.208984375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 2591.345703125, |
| "completions/mean_terminated_length": 2193.81982421875, |
| "completions/min_length": 389.0, |
| "completions/min_terminated_length": 389.0, |
| "epoch": 0.23349893541518807, |
| "frac_reward_zero_std": 0.09375, |
| "grad_norm": 0.21333864331245422, |
| "learning_rate": 3.4399999999999996e-07, |
| "loss": 0.0395, |
| "num_tokens": 514474534.0, |
| "reward": 0.5546875, |
| "reward_std": 0.37220126390457153, |
| "rewards/simpleverify_reward/mean": 0.5546875, |
| "rewards/simpleverify_reward/std": 0.49748632311820984, |
| "step": 329 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.24609375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4048.0, |
| "completions/mean_length": 2776.080078125, |
| "completions/mean_terminated_length": 2345.225341796875, |
| "completions/min_length": 629.0, |
| "completions/min_terminated_length": 629.0, |
| "epoch": 0.23420865862313697, |
| "frac_reward_zero_std": 0.0625, |
| "grad_norm": 0.21197707951068878, |
| "learning_rate": 3.42e-07, |
| "loss": 0.047, |
| "num_tokens": 515940047.0, |
| "reward": 0.46875, |
| "reward_std": 0.400264173746109, |
| "rewards/simpleverify_reward/mean": 0.46875, |
| "rewards/simpleverify_reward/std": 0.4995105266571045, |
| "step": 330 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.38671875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4073.0, |
| "completions/mean_length": 2941.484375, |
| "completions/mean_terminated_length": 2213.477783203125, |
| "completions/min_length": 794.0, |
| "completions/min_terminated_length": 794.0, |
| "epoch": 0.23491838183108588, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.1675199419260025, |
| "learning_rate": 3.4000000000000003e-07, |
| "loss": 0.0382, |
| "num_tokens": 517506775.0, |
| "reward": 0.421875, |
| "reward_std": 0.2874833941459656, |
| "rewards/simpleverify_reward/mean": 0.421875, |
| "rewards/simpleverify_reward/std": 0.49434176087379456, |
| "step": 331 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.203125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4089.0, |
| "completions/mean_length": 2468.76171875, |
| "completions/mean_terminated_length": 2053.9755859375, |
| "completions/min_length": 594.0, |
| "completions/min_terminated_length": 594.0, |
| "epoch": 0.23562810503903478, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.18241873383522034, |
| "learning_rate": 3.38e-07, |
| "loss": 0.0326, |
| "num_tokens": 518821965.0, |
| "reward": 0.583984375, |
| "reward_std": 0.3025062084197998, |
| "rewards/simpleverify_reward/mean": 0.583984375, |
| "rewards/simpleverify_reward/std": 0.493378221988678, |
| "step": 332 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.232421875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4072.0, |
| "completions/mean_length": 2589.4609375, |
| "completions/mean_terminated_length": 2133.282470703125, |
| "completions/min_length": 683.0, |
| "completions/min_terminated_length": 683.0, |
| "epoch": 0.23633782824698368, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.19192776083946228, |
| "learning_rate": 3.36e-07, |
| "loss": 0.0396, |
| "num_tokens": 520204777.0, |
| "reward": 0.634765625, |
| "reward_std": 0.3036050796508789, |
| "rewards/simpleverify_reward/mean": 0.634765625, |
| "rewards/simpleverify_reward/std": 0.4819667339324951, |
| "step": 333 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.19140625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4090.0, |
| "completions/mean_length": 2475.267578125, |
| "completions/mean_terminated_length": 2091.615966796875, |
| "completions/min_length": 76.0, |
| "completions/min_terminated_length": 76.0, |
| "epoch": 0.2370475514549326, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.17822405695915222, |
| "learning_rate": 3.34e-07, |
| "loss": 0.0389, |
| "num_tokens": 521518802.0, |
| "reward": 0.58203125, |
| "reward_std": 0.33402031660079956, |
| "rewards/simpleverify_reward/mean": 0.58203125, |
| "rewards/simpleverify_reward/std": 0.4937073290348053, |
| "step": 334 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.201171875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4076.0, |
| "completions/mean_length": 2533.154296875, |
| "completions/mean_terminated_length": 2139.5771484375, |
| "completions/min_length": 900.0, |
| "completions/min_terminated_length": 900.0, |
| "epoch": 0.23775727466288146, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.19185730814933777, |
| "learning_rate": 3.32e-07, |
| "loss": 0.0523, |
| "num_tokens": 522862945.0, |
| "reward": 0.580078125, |
| "reward_std": 0.3231615722179413, |
| "rewards/simpleverify_reward/mean": 0.580078125, |
| "rewards/simpleverify_reward/std": 0.4940285086631775, |
| "step": 335 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.240234375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4071.0, |
| "completions/mean_length": 2710.27734375, |
| "completions/mean_terminated_length": 2272.1181640625, |
| "completions/min_length": 895.0, |
| "completions/min_terminated_length": 895.0, |
| "epoch": 0.23846699787083037, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.17365916073322296, |
| "learning_rate": 3.3e-07, |
| "loss": 0.027, |
| "num_tokens": 524301903.0, |
| "reward": 0.583984375, |
| "reward_std": 0.2996189594268799, |
| "rewards/simpleverify_reward/mean": 0.583984375, |
| "rewards/simpleverify_reward/std": 0.493378221988678, |
| "step": 336 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.23046875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4087.0, |
| "completions/mean_length": 2743.625, |
| "completions/mean_terminated_length": 2338.598876953125, |
| "completions/min_length": 673.0, |
| "completions/min_terminated_length": 673.0, |
| "epoch": 0.23917672107877927, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.19701598584651947, |
| "learning_rate": 3.28e-07, |
| "loss": 0.0345, |
| "num_tokens": 525749423.0, |
| "reward": 0.501953125, |
| "reward_std": 0.35752877593040466, |
| "rewards/simpleverify_reward/mean": 0.501953125, |
| "rewards/simpleverify_reward/std": 0.5004851818084717, |
| "step": 337 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.28515625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4077.0, |
| "completions/mean_length": 2835.85546875, |
| "completions/mean_terminated_length": 2333.1748046875, |
| "completions/min_length": 435.0, |
| "completions/min_terminated_length": 435.0, |
| "epoch": 0.23988644428672817, |
| "frac_reward_zero_std": 0.125, |
| "grad_norm": 0.20235873758792877, |
| "learning_rate": 3.26e-07, |
| "loss": 0.0516, |
| "num_tokens": 527246293.0, |
| "reward": 0.46484375, |
| "reward_std": 0.33568209409713745, |
| "rewards/simpleverify_reward/mean": 0.46484375, |
| "rewards/simpleverify_reward/std": 0.49925029277801514, |
| "step": 338 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.23046875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4070.0, |
| "completions/mean_length": 2632.314453125, |
| "completions/mean_terminated_length": 2193.95166015625, |
| "completions/min_length": 567.0, |
| "completions/min_terminated_length": 567.0, |
| "epoch": 0.24059616749467708, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.20360615849494934, |
| "learning_rate": 3.24e-07, |
| "loss": 0.0382, |
| "num_tokens": 528635062.0, |
| "reward": 0.513671875, |
| "reward_std": 0.3104659914970398, |
| "rewards/simpleverify_reward/mean": 0.513671875, |
| "rewards/simpleverify_reward/std": 0.5003018379211426, |
| "step": 339 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.2421875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4082.0, |
| "completions/mean_length": 2619.37890625, |
| "completions/mean_terminated_length": 2147.468994140625, |
| "completions/min_length": 689.0, |
| "completions/min_terminated_length": 689.0, |
| "epoch": 0.24130589070262598, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.23346677422523499, |
| "learning_rate": 3.22e-07, |
| "loss": 0.0246, |
| "num_tokens": 530024536.0, |
| "reward": 0.53125, |
| "reward_std": 0.34536439180374146, |
| "rewards/simpleverify_reward/mean": 0.53125, |
| "rewards/simpleverify_reward/std": 0.4995105266571045, |
| "step": 340 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.208984375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4078.0, |
| "completions/mean_length": 2551.814453125, |
| "completions/mean_terminated_length": 2143.844482421875, |
| "completions/min_length": 550.0, |
| "completions/min_terminated_length": 550.0, |
| "epoch": 0.24201561391057488, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.21286055445671082, |
| "learning_rate": 3.2e-07, |
| "loss": 0.0412, |
| "num_tokens": 531373593.0, |
| "reward": 0.541015625, |
| "reward_std": 0.3267485797405243, |
| "rewards/simpleverify_reward/mean": 0.541015625, |
| "rewards/simpleverify_reward/std": 0.49880221486091614, |
| "step": 341 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.154296875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4049.0, |
| "completions/mean_length": 2384.0703125, |
| "completions/mean_terminated_length": 2071.732177734375, |
| "completions/min_length": 431.0, |
| "completions/min_terminated_length": 431.0, |
| "epoch": 0.2427253371185238, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.2067318558692932, |
| "learning_rate": 3.18e-07, |
| "loss": 0.0598, |
| "num_tokens": 532638861.0, |
| "reward": 0.666015625, |
| "reward_std": 0.33365142345428467, |
| "rewards/simpleverify_reward/mean": 0.666015625, |
| "rewards/simpleverify_reward/std": 0.47209542989730835, |
| "step": 342 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.283203125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4086.0, |
| "completions/mean_length": 2653.115234375, |
| "completions/mean_terminated_length": 2083.0380859375, |
| "completions/min_length": 659.0, |
| "completions/min_terminated_length": 659.0, |
| "epoch": 0.24343506032647266, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.18490897119045258, |
| "learning_rate": 3.1599999999999997e-07, |
| "loss": 0.038, |
| "num_tokens": 534045032.0, |
| "reward": 0.58984375, |
| "reward_std": 0.26106908917427063, |
| "rewards/simpleverify_reward/mean": 0.58984375, |
| "rewards/simpleverify_reward/std": 0.49234291911125183, |
| "step": 343 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.20703125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 2561.11328125, |
| "completions/mean_terminated_length": 2160.37939453125, |
| "completions/min_length": 434.0, |
| "completions/min_terminated_length": 434.0, |
| "epoch": 0.24414478353442157, |
| "frac_reward_zero_std": 0.09375, |
| "grad_norm": 0.21138200163841248, |
| "learning_rate": 3.14e-07, |
| "loss": 0.0301, |
| "num_tokens": 535410498.0, |
| "reward": 0.580078125, |
| "reward_std": 0.3660527467727661, |
| "rewards/simpleverify_reward/mean": 0.580078125, |
| "rewards/simpleverify_reward/std": 0.4940285086631775, |
| "step": 344 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.185546875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4084.0, |
| "completions/mean_length": 2658.9140625, |
| "completions/mean_terminated_length": 2331.5205078125, |
| "completions/min_length": 698.0, |
| "completions/min_terminated_length": 698.0, |
| "epoch": 0.24485450674237047, |
| "frac_reward_zero_std": 0.125, |
| "grad_norm": 0.21464751660823822, |
| "learning_rate": 3.12e-07, |
| "loss": 0.022, |
| "num_tokens": 536816310.0, |
| "reward": 0.595703125, |
| "reward_std": 0.37484991550445557, |
| "rewards/simpleverify_reward/mean": 0.595703125, |
| "rewards/simpleverify_reward/std": 0.4912354052066803, |
| "step": 345 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.130859375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4096.0, |
| "completions/mean_length": 2440.67578125, |
| "completions/mean_terminated_length": 2191.447265625, |
| "completions/min_length": 548.0, |
| "completions/min_terminated_length": 548.0, |
| "epoch": 0.24556422995031937, |
| "frac_reward_zero_std": 0.0625, |
| "grad_norm": 0.22174955904483795, |
| "learning_rate": 3.1e-07, |
| "loss": 0.0472, |
| "num_tokens": 538116048.0, |
| "reward": 0.5703125, |
| "reward_std": 0.3892859220504761, |
| "rewards/simpleverify_reward/mean": 0.5703125, |
| "rewards/simpleverify_reward/std": 0.4955156147480011, |
| "step": 346 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.138671875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4077.0, |
| "completions/mean_length": 2346.408203125, |
| "completions/mean_terminated_length": 2064.72802734375, |
| "completions/min_length": 372.0, |
| "completions/min_terminated_length": 372.0, |
| "epoch": 0.24627395315826828, |
| "frac_reward_zero_std": 0.09375, |
| "grad_norm": 0.23982973396778107, |
| "learning_rate": 3.08e-07, |
| "loss": 0.0311, |
| "num_tokens": 539359937.0, |
| "reward": 0.607421875, |
| "reward_std": 0.3656797707080841, |
| "rewards/simpleverify_reward/mean": 0.607421875, |
| "rewards/simpleverify_reward/std": 0.4888018071651459, |
| "step": 347 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.208984375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4083.0, |
| "completions/mean_length": 2690.763671875, |
| "completions/mean_terminated_length": 2319.50390625, |
| "completions/min_length": 565.0, |
| "completions/min_terminated_length": 565.0, |
| "epoch": 0.24698367636621718, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.19287671148777008, |
| "learning_rate": 3.0599999999999996e-07, |
| "loss": 0.0515, |
| "num_tokens": 540788824.0, |
| "reward": 0.52734375, |
| "reward_std": 0.3201083242893219, |
| "rewards/simpleverify_reward/mean": 0.52734375, |
| "rewards/simpleverify_reward/std": 0.49974003434181213, |
| "step": 348 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.28125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4079.0, |
| "completions/mean_length": 2702.34765625, |
| "completions/mean_terminated_length": 2157.00537109375, |
| "completions/min_length": 598.0, |
| "completions/min_terminated_length": 598.0, |
| "epoch": 0.24769339957416608, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.20683422684669495, |
| "learning_rate": 3.0399999999999997e-07, |
| "loss": 0.0432, |
| "num_tokens": 542218090.0, |
| "reward": 0.4765625, |
| "reward_std": 0.3258872926235199, |
| "rewards/simpleverify_reward/mean": 0.4765625, |
| "rewards/simpleverify_reward/std": 0.49993884563446045, |
| "step": 349 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.25, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4091.0, |
| "completions/mean_length": 2714.150390625, |
| "completions/mean_terminated_length": 2253.533935546875, |
| "completions/min_length": 759.0, |
| "completions/min_terminated_length": 759.0, |
| "epoch": 0.248403122782115, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.1979198008775711, |
| "learning_rate": 3.02e-07, |
| "loss": 0.0434, |
| "num_tokens": 543652087.0, |
| "reward": 0.52734375, |
| "reward_std": 0.35835325717926025, |
| "rewards/simpleverify_reward/mean": 0.52734375, |
| "rewards/simpleverify_reward/std": 0.49974003434181213, |
| "step": 350 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.185546875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4004.0, |
| "completions/mean_length": 2260.5, |
| "completions/mean_terminated_length": 1842.340576171875, |
| "completions/min_length": 628.0, |
| "completions/min_terminated_length": 628.0, |
| "epoch": 0.24911284599006386, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.2081659883260727, |
| "learning_rate": 3e-07, |
| "loss": 0.058, |
| "num_tokens": 544852983.0, |
| "reward": 0.673828125, |
| "reward_std": 0.2824767231941223, |
| "rewards/simpleverify_reward/mean": 0.673828125, |
| "rewards/simpleverify_reward/std": 0.4692695140838623, |
| "step": 351 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.25, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4052.0, |
| "completions/mean_length": 2431.7421875, |
| "completions/mean_terminated_length": 1876.9896240234375, |
| "completions/min_length": 516.0, |
| "completions/min_terminated_length": 516.0, |
| "epoch": 0.24982256919801277, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.2134397029876709, |
| "learning_rate": 2.98e-07, |
| "loss": 0.0398, |
| "num_tokens": 546142099.0, |
| "reward": 0.517578125, |
| "reward_std": 0.29715022444725037, |
| "rewards/simpleverify_reward/mean": 0.517578125, |
| "rewards/simpleverify_reward/std": 0.5001795887947083, |
| "step": 352 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.294921875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4082.0, |
| "completions/mean_length": 2839.96875, |
| "completions/mean_terminated_length": 2314.5927734375, |
| "completions/min_length": 645.0, |
| "completions/min_terminated_length": 645.0, |
| "epoch": 0.2505322924059617, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.19456526637077332, |
| "learning_rate": 2.9599999999999995e-07, |
| "loss": 0.0402, |
| "num_tokens": 547636451.0, |
| "reward": 0.521484375, |
| "reward_std": 0.354428768157959, |
| "rewards/simpleverify_reward/mean": 0.521484375, |
| "rewards/simpleverify_reward/std": 0.5000267624855042, |
| "step": 353 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.162109375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4090.0, |
| "completions/mean_length": 2327.4765625, |
| "completions/mean_terminated_length": 1985.314697265625, |
| "completions/min_length": 692.0, |
| "completions/min_terminated_length": 692.0, |
| "epoch": 0.2512420156139106, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.21156735718250275, |
| "learning_rate": 2.9399999999999996e-07, |
| "loss": 0.0653, |
| "num_tokens": 548870759.0, |
| "reward": 0.638671875, |
| "reward_std": 0.3378239572048187, |
| "rewards/simpleverify_reward/mean": 0.638671875, |
| "rewards/simpleverify_reward/std": 0.48085519671440125, |
| "step": 354 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.26953125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4096.0, |
| "completions/mean_length": 2719.87109375, |
| "completions/mean_terminated_length": 2212.1015625, |
| "completions/min_length": 633.0, |
| "completions/min_terminated_length": 633.0, |
| "epoch": 0.25195173882185945, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.17233306169509888, |
| "learning_rate": 2.9199999999999997e-07, |
| "loss": 0.0376, |
| "num_tokens": 550303605.0, |
| "reward": 0.58203125, |
| "reward_std": 0.3008376359939575, |
| "rewards/simpleverify_reward/mean": 0.58203125, |
| "rewards/simpleverify_reward/std": 0.4937073290348053, |
| "step": 355 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.21875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4069.0, |
| "completions/mean_length": 2491.25, |
| "completions/mean_terminated_length": 2041.919921875, |
| "completions/min_length": 444.0, |
| "completions/min_terminated_length": 444.0, |
| "epoch": 0.2526614620298084, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.18799129128456116, |
| "learning_rate": 2.9e-07, |
| "loss": 0.0423, |
| "num_tokens": 551622229.0, |
| "reward": 0.5703125, |
| "reward_std": 0.2613755464553833, |
| "rewards/simpleverify_reward/mean": 0.5703125, |
| "rewards/simpleverify_reward/std": 0.4955156147480011, |
| "step": 356 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.244140625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4093.0, |
| "completions/mean_length": 2679.66015625, |
| "completions/mean_terminated_length": 2222.18603515625, |
| "completions/min_length": 518.0, |
| "completions/min_terminated_length": 518.0, |
| "epoch": 0.25337118523775726, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.19577065110206604, |
| "learning_rate": 2.88e-07, |
| "loss": 0.0622, |
| "num_tokens": 553035735.0, |
| "reward": 0.54296875, |
| "reward_std": 0.34782466292381287, |
| "rewards/simpleverify_reward/mean": 0.54296875, |
| "rewards/simpleverify_reward/std": 0.49863746762275696, |
| "step": 357 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.259765625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4092.0, |
| "completions/mean_length": 2751.294921875, |
| "completions/mean_terminated_length": 2279.406494140625, |
| "completions/min_length": 655.0, |
| "completions/min_terminated_length": 655.0, |
| "epoch": 0.2540809084457062, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.19107285141944885, |
| "learning_rate": 2.8599999999999994e-07, |
| "loss": 0.061, |
| "num_tokens": 554490846.0, |
| "reward": 0.537109375, |
| "reward_std": 0.34905558824539185, |
| "rewards/simpleverify_reward/mean": 0.537109375, |
| "rewards/simpleverify_reward/std": 0.4991086423397064, |
| "step": 358 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.1953125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4081.0, |
| "completions/mean_length": 2450.25390625, |
| "completions/mean_terminated_length": 2050.801025390625, |
| "completions/min_length": 630.0, |
| "completions/min_terminated_length": 630.0, |
| "epoch": 0.25479063165365506, |
| "frac_reward_zero_std": 0.09375, |
| "grad_norm": 0.2313884198665619, |
| "learning_rate": 2.8399999999999995e-07, |
| "loss": 0.0357, |
| "num_tokens": 555795424.0, |
| "reward": 0.57421875, |
| "reward_std": 0.37918421626091003, |
| "rewards/simpleverify_reward/mean": 0.57421875, |
| "rewards/simpleverify_reward/std": 0.4949444830417633, |
| "step": 359 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.255859375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4092.0, |
| "completions/mean_length": 2476.509765625, |
| "completions/mean_terminated_length": 1919.6771240234375, |
| "completions/min_length": 565.0, |
| "completions/min_terminated_length": 565.0, |
| "epoch": 0.255500354861604, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.20385627448558807, |
| "learning_rate": 2.8199999999999996e-07, |
| "loss": 0.0284, |
| "num_tokens": 557101781.0, |
| "reward": 0.533203125, |
| "reward_std": 0.3176800608634949, |
| "rewards/simpleverify_reward/mean": 0.533203125, |
| "rewards/simpleverify_reward/std": 0.4993842542171478, |
| "step": 360 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.20703125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4055.0, |
| "completions/mean_length": 2334.49609375, |
| "completions/mean_terminated_length": 1874.5960693359375, |
| "completions/min_length": 481.0, |
| "completions/min_terminated_length": 481.0, |
| "epoch": 0.25621007806955287, |
| "frac_reward_zero_std": 0.125, |
| "grad_norm": 0.2169909030199051, |
| "learning_rate": 2.8e-07, |
| "loss": 0.0314, |
| "num_tokens": 558345987.0, |
| "reward": 0.517578125, |
| "reward_std": 0.34519070386886597, |
| "rewards/simpleverify_reward/mean": 0.517578125, |
| "rewards/simpleverify_reward/std": 0.5001795887947083, |
| "step": 361 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.197265625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4075.0, |
| "completions/mean_length": 2457.859375, |
| "completions/mean_terminated_length": 2055.29931640625, |
| "completions/min_length": 245.0, |
| "completions/min_terminated_length": 245.0, |
| "epoch": 0.25691980127750175, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.2180250883102417, |
| "learning_rate": 2.7800000000000003e-07, |
| "loss": 0.0633, |
| "num_tokens": 559654603.0, |
| "reward": 0.5546875, |
| "reward_std": 0.3741927146911621, |
| "rewards/simpleverify_reward/mean": 0.5546875, |
| "rewards/simpleverify_reward/std": 0.49748632311820984, |
| "step": 362 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.146484375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4090.0, |
| "completions/mean_length": 2425.732421875, |
| "completions/mean_terminated_length": 2139.0732421875, |
| "completions/min_length": 346.0, |
| "completions/min_terminated_length": 346.0, |
| "epoch": 0.2576295244854507, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.21454890072345734, |
| "learning_rate": 2.7600000000000004e-07, |
| "loss": 0.0405, |
| "num_tokens": 560937394.0, |
| "reward": 0.54296875, |
| "reward_std": 0.3476009964942932, |
| "rewards/simpleverify_reward/mean": 0.54296875, |
| "rewards/simpleverify_reward/std": 0.49863746762275696, |
| "step": 363 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.142578125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4080.0, |
| "completions/mean_length": 2378.091796875, |
| "completions/mean_terminated_length": 2092.426025390625, |
| "completions/min_length": 566.0, |
| "completions/min_terminated_length": 566.0, |
| "epoch": 0.25833924769339955, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.19333621859550476, |
| "learning_rate": 2.74e-07, |
| "loss": 0.0192, |
| "num_tokens": 562200753.0, |
| "reward": 0.5703125, |
| "reward_std": 0.34931451082229614, |
| "rewards/simpleverify_reward/mean": 0.5703125, |
| "rewards/simpleverify_reward/std": 0.4955156147480011, |
| "step": 364 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.158203125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4071.0, |
| "completions/mean_length": 2346.185546875, |
| "completions/mean_terminated_length": 2017.333984375, |
| "completions/min_length": 618.0, |
| "completions/min_terminated_length": 618.0, |
| "epoch": 0.2590489709013485, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.24230532348155975, |
| "learning_rate": 2.72e-07, |
| "loss": 0.0189, |
| "num_tokens": 563445520.0, |
| "reward": 0.552734375, |
| "reward_std": 0.3605913519859314, |
| "rewards/simpleverify_reward/mean": 0.552734375, |
| "rewards/simpleverify_reward/std": 0.4976975917816162, |
| "step": 365 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.1328125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4045.0, |
| "completions/mean_length": 2336.62109375, |
| "completions/mean_terminated_length": 2067.166748046875, |
| "completions/min_length": 828.0, |
| "completions/min_terminated_length": 828.0, |
| "epoch": 0.25975869410929736, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.17702725529670715, |
| "learning_rate": 2.7e-07, |
| "loss": 0.0398, |
| "num_tokens": 564686654.0, |
| "reward": 0.6640625, |
| "reward_std": 0.28119152784347534, |
| "rewards/simpleverify_reward/mean": 0.6640625, |
| "rewards/simpleverify_reward/std": 0.4727790653705597, |
| "step": 366 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.16015625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4084.0, |
| "completions/mean_length": 2564.375, |
| "completions/mean_terminated_length": 2272.297607421875, |
| "completions/min_length": 304.0, |
| "completions/min_terminated_length": 304.0, |
| "epoch": 0.2604684173172463, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.19529995322227478, |
| "learning_rate": 2.68e-07, |
| "loss": 0.0363, |
| "num_tokens": 566048846.0, |
| "reward": 0.634765625, |
| "reward_std": 0.3344035744667053, |
| "rewards/simpleverify_reward/mean": 0.634765625, |
| "rewards/simpleverify_reward/std": 0.4819667339324951, |
| "step": 367 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.28125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4087.0, |
| "completions/mean_length": 2759.208984375, |
| "completions/mean_terminated_length": 2236.116943359375, |
| "completions/min_length": 483.0, |
| "completions/min_terminated_length": 483.0, |
| "epoch": 0.26117814052519517, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.19797979295253754, |
| "learning_rate": 2.66e-07, |
| "loss": 0.0419, |
| "num_tokens": 567505657.0, |
| "reward": 0.45703125, |
| "reward_std": 0.3104194402694702, |
| "rewards/simpleverify_reward/mean": 0.45703125, |
| "rewards/simpleverify_reward/std": 0.49863746762275696, |
| "step": 368 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.177734375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4087.0, |
| "completions/mean_length": 2745.6015625, |
| "completions/mean_terminated_length": 2453.710205078125, |
| "completions/min_length": 45.0, |
| "completions/min_terminated_length": 45.0, |
| "epoch": 0.2618878637331441, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.18262818455696106, |
| "learning_rate": 2.64e-07, |
| "loss": 0.0353, |
| "num_tokens": 568963229.0, |
| "reward": 0.630859375, |
| "reward_std": 0.3537280857563019, |
| "rewards/simpleverify_reward/mean": 0.630859375, |
| "rewards/simpleverify_reward/std": 0.4830440282821655, |
| "step": 369 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.3359375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4087.0, |
| "completions/mean_length": 2936.83203125, |
| "completions/mean_terminated_length": 2350.429443359375, |
| "completions/min_length": 318.0, |
| "completions/min_terminated_length": 318.0, |
| "epoch": 0.262597586941093, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.18226830661296844, |
| "learning_rate": 2.62e-07, |
| "loss": 0.064, |
| "num_tokens": 570518999.0, |
| "reward": 0.4375, |
| "reward_std": 0.3118121027946472, |
| "rewards/simpleverify_reward/mean": 0.4375, |
| "rewards/simpleverify_reward/std": 0.49656352400779724, |
| "step": 370 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.1328125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4083.0, |
| "completions/mean_length": 2030.396484375, |
| "completions/mean_terminated_length": 1714.0428466796875, |
| "completions/min_length": 570.0, |
| "completions/min_terminated_length": 570.0, |
| "epoch": 0.26330731014904185, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.22261498868465424, |
| "learning_rate": 2.6e-07, |
| "loss": 0.0387, |
| "num_tokens": 571601474.0, |
| "reward": 0.693359375, |
| "reward_std": 0.29824966192245483, |
| "rewards/simpleverify_reward/mean": 0.693359375, |
| "rewards/simpleverify_reward/std": 0.4615498185157776, |
| "step": 371 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.224609375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4084.0, |
| "completions/mean_length": 2534.048828125, |
| "completions/mean_terminated_length": 2081.594482421875, |
| "completions/min_length": 473.0, |
| "completions/min_terminated_length": 473.0, |
| "epoch": 0.2640170333569908, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.18592584133148193, |
| "learning_rate": 2.58e-07, |
| "loss": 0.0276, |
| "num_tokens": 572947595.0, |
| "reward": 0.46875, |
| "reward_std": 0.309254914522171, |
| "rewards/simpleverify_reward/mean": 0.46875, |
| "rewards/simpleverify_reward/std": 0.4995105266571045, |
| "step": 372 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.2578125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4068.0, |
| "completions/mean_length": 2558.4296875, |
| "completions/mean_terminated_length": 2024.326416015625, |
| "completions/min_length": 307.0, |
| "completions/min_terminated_length": 307.0, |
| "epoch": 0.26472675656493966, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.20139005780220032, |
| "learning_rate": 2.56e-07, |
| "loss": 0.0683, |
| "num_tokens": 574317399.0, |
| "reward": 0.55078125, |
| "reward_std": 0.3291076421737671, |
| "rewards/simpleverify_reward/mean": 0.55078125, |
| "rewards/simpleverify_reward/std": 0.497901052236557, |
| "step": 373 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.2109375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4037.0, |
| "completions/mean_length": 2402.427734375, |
| "completions/mean_terminated_length": 1949.6905517578125, |
| "completions/min_length": 697.0, |
| "completions/min_terminated_length": 697.0, |
| "epoch": 0.2654364797728886, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.21713021397590637, |
| "learning_rate": 2.5399999999999997e-07, |
| "loss": 0.0266, |
| "num_tokens": 575591634.0, |
| "reward": 0.58984375, |
| "reward_std": 0.3220381736755371, |
| "rewards/simpleverify_reward/mean": 0.58984375, |
| "rewards/simpleverify_reward/std": 0.49234291911125183, |
| "step": 374 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.232421875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4079.0, |
| "completions/mean_length": 2629.169921875, |
| "completions/mean_terminated_length": 2185.015380859375, |
| "completions/min_length": 816.0, |
| "completions/min_terminated_length": 816.0, |
| "epoch": 0.26614620298083747, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.19079548120498657, |
| "learning_rate": 2.52e-07, |
| "loss": 0.0399, |
| "num_tokens": 576981385.0, |
| "reward": 0.619140625, |
| "reward_std": 0.2839146852493286, |
| "rewards/simpleverify_reward/mean": 0.619140625, |
| "rewards/simpleverify_reward/std": 0.48607301712036133, |
| "step": 375 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.19140625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4077.0, |
| "completions/mean_length": 2501.681640625, |
| "completions/mean_terminated_length": 2124.282470703125, |
| "completions/min_length": 745.0, |
| "completions/min_terminated_length": 745.0, |
| "epoch": 0.2668559261887864, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.1964111179113388, |
| "learning_rate": 2.5e-07, |
| "loss": 0.0483, |
| "num_tokens": 578312070.0, |
| "reward": 0.548828125, |
| "reward_std": 0.32082968950271606, |
| "rewards/simpleverify_reward/mean": 0.548828125, |
| "rewards/simpleverify_reward/std": 0.498096764087677, |
| "step": 376 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.2421875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4093.0, |
| "completions/mean_length": 2648.36328125, |
| "completions/mean_terminated_length": 2185.71630859375, |
| "completions/min_length": 594.0, |
| "completions/min_terminated_length": 594.0, |
| "epoch": 0.2675656493967353, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.19703461229801178, |
| "learning_rate": 2.48e-07, |
| "loss": 0.0185, |
| "num_tokens": 579716208.0, |
| "reward": 0.51171875, |
| "reward_std": 0.35496723651885986, |
| "rewards/simpleverify_reward/mean": 0.51171875, |
| "rewards/simpleverify_reward/std": 0.5003514885902405, |
| "step": 377 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.2890625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4080.0, |
| "completions/mean_length": 2895.693359375, |
| "completions/mean_terminated_length": 2407.65673828125, |
| "completions/min_length": 879.0, |
| "completions/min_terminated_length": 879.0, |
| "epoch": 0.26827537260468415, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.19431690871715546, |
| "learning_rate": 2.46e-07, |
| "loss": 0.0574, |
| "num_tokens": 581246387.0, |
| "reward": 0.513671875, |
| "reward_std": 0.33509624004364014, |
| "rewards/simpleverify_reward/mean": 0.513671875, |
| "rewards/simpleverify_reward/std": 0.5003018379211426, |
| "step": 378 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.16015625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4090.0, |
| "completions/mean_length": 2416.4375, |
| "completions/mean_terminated_length": 2096.14892578125, |
| "completions/min_length": 501.0, |
| "completions/min_terminated_length": 501.0, |
| "epoch": 0.2689850958126331, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.1963028609752655, |
| "learning_rate": 2.4399999999999996e-07, |
| "loss": 0.0278, |
| "num_tokens": 582525635.0, |
| "reward": 0.6328125, |
| "reward_std": 0.3088630437850952, |
| "rewards/simpleverify_reward/mean": 0.6328125, |
| "rewards/simpleverify_reward/std": 0.48250964283943176, |
| "step": 379 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.19921875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4073.0, |
| "completions/mean_length": 2446.3125, |
| "completions/mean_terminated_length": 2035.9024658203125, |
| "completions/min_length": 486.0, |
| "completions/min_terminated_length": 486.0, |
| "epoch": 0.26969481902058196, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.19548539817333221, |
| "learning_rate": 2.4199999999999997e-07, |
| "loss": 0.0175, |
| "num_tokens": 583821923.0, |
| "reward": 0.61328125, |
| "reward_std": 0.3079766631126404, |
| "rewards/simpleverify_reward/mean": 0.61328125, |
| "rewards/simpleverify_reward/std": 0.48747459053993225, |
| "step": 380 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.263671875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4092.0, |
| "completions/mean_length": 2705.947265625, |
| "completions/mean_terminated_length": 2208.18310546875, |
| "completions/min_length": 585.0, |
| "completions/min_terminated_length": 585.0, |
| "epoch": 0.2704045422285309, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.1892731636762619, |
| "learning_rate": 2.4e-07, |
| "loss": 0.0087, |
| "num_tokens": 585254856.0, |
| "reward": 0.505859375, |
| "reward_std": 0.28356197476387024, |
| "rewards/simpleverify_reward/mean": 0.505859375, |
| "rewards/simpleverify_reward/std": 0.5004546642303467, |
| "step": 381 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.1953125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4090.0, |
| "completions/mean_length": 2471.662109375, |
| "completions/mean_terminated_length": 2077.4052734375, |
| "completions/min_length": 723.0, |
| "completions/min_terminated_length": 723.0, |
| "epoch": 0.27111426543647976, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.19215938448905945, |
| "learning_rate": 2.38e-07, |
| "loss": 0.0187, |
| "num_tokens": 586565355.0, |
| "reward": 0.556640625, |
| "reward_std": 0.30890369415283203, |
| "rewards/simpleverify_reward/mean": 0.556640625, |
| "rewards/simpleverify_reward/std": 0.49726733565330505, |
| "step": 382 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.224609375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4096.0, |
| "completions/mean_length": 2829.3984375, |
| "completions/mean_terminated_length": 2462.49853515625, |
| "completions/min_length": 895.0, |
| "completions/min_terminated_length": 895.0, |
| "epoch": 0.2718239886444287, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.18462873995304108, |
| "learning_rate": 2.3599999999999997e-07, |
| "loss": 0.0195, |
| "num_tokens": 588063479.0, |
| "reward": 0.5078125, |
| "reward_std": 0.33669960498809814, |
| "rewards/simpleverify_reward/mean": 0.5078125, |
| "rewards/simpleverify_reward/std": 0.5004279017448425, |
| "step": 383 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.224609375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4051.0, |
| "completions/mean_length": 2729.845703125, |
| "completions/mean_terminated_length": 2334.108154296875, |
| "completions/min_length": 791.0, |
| "completions/min_terminated_length": 791.0, |
| "epoch": 0.27253371185237757, |
| "frac_reward_zero_std": 0.125, |
| "grad_norm": 0.19513589143753052, |
| "learning_rate": 2.34e-07, |
| "loss": 0.0468, |
| "num_tokens": 589503656.0, |
| "reward": 0.580078125, |
| "reward_std": 0.3703019320964813, |
| "rewards/simpleverify_reward/mean": 0.580078125, |
| "rewards/simpleverify_reward/std": 0.4940285086631775, |
| "step": 384 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.2265625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4096.0, |
| "completions/mean_length": 2450.4140625, |
| "completions/mean_terminated_length": 1968.373779296875, |
| "completions/min_length": 524.0, |
| "completions/min_terminated_length": 524.0, |
| "epoch": 0.27324343506032645, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.19112029671669006, |
| "learning_rate": 2.32e-07, |
| "loss": 0.0218, |
| "num_tokens": 590806060.0, |
| "reward": 0.546875, |
| "reward_std": 0.3013434410095215, |
| "rewards/simpleverify_reward/mean": 0.546875, |
| "rewards/simpleverify_reward/std": 0.4982847273349762, |
| "step": 385 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.25, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4085.0, |
| "completions/mean_length": 2695.744140625, |
| "completions/mean_terminated_length": 2228.9921875, |
| "completions/min_length": 570.0, |
| "completions/min_terminated_length": 570.0, |
| "epoch": 0.2739531582682754, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.20222526788711548, |
| "learning_rate": 2.3e-07, |
| "loss": 0.0254, |
| "num_tokens": 592245705.0, |
| "reward": 0.484375, |
| "reward_std": 0.3360751271247864, |
| "rewards/simpleverify_reward/mean": 0.484375, |
| "rewards/simpleverify_reward/std": 0.5002445578575134, |
| "step": 386 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.267578125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4066.0, |
| "completions/mean_length": 2729.251953125, |
| "completions/mean_terminated_length": 2229.933349609375, |
| "completions/min_length": 711.0, |
| "completions/min_terminated_length": 711.0, |
| "epoch": 0.27466288147622425, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.20924566686153412, |
| "learning_rate": 2.28e-07, |
| "loss": 0.0564, |
| "num_tokens": 593685770.0, |
| "reward": 0.572265625, |
| "reward_std": 0.33446913957595825, |
| "rewards/simpleverify_reward/mean": 0.572265625, |
| "rewards/simpleverify_reward/std": 0.4952339828014374, |
| "step": 387 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.1640625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4088.0, |
| "completions/mean_length": 2437.62890625, |
| "completions/mean_terminated_length": 2112.154052734375, |
| "completions/min_length": 733.0, |
| "completions/min_terminated_length": 733.0, |
| "epoch": 0.2753726046841732, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.19301064312458038, |
| "learning_rate": 2.2599999999999999e-07, |
| "loss": 0.0451, |
| "num_tokens": 594977452.0, |
| "reward": 0.630859375, |
| "reward_std": 0.3132881224155426, |
| "rewards/simpleverify_reward/mean": 0.630859375, |
| "rewards/simpleverify_reward/std": 0.4830440282821655, |
| "step": 388 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.29296875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4081.0, |
| "completions/mean_length": 2792.505859375, |
| "completions/mean_terminated_length": 2252.384033203125, |
| "completions/min_length": 635.0, |
| "completions/min_terminated_length": 635.0, |
| "epoch": 0.27608232789212206, |
| "frac_reward_zero_std": 0.125, |
| "grad_norm": 0.21344466507434845, |
| "learning_rate": 2.24e-07, |
| "loss": 0.0408, |
| "num_tokens": 596448591.0, |
| "reward": 0.513671875, |
| "reward_std": 0.343060702085495, |
| "rewards/simpleverify_reward/mean": 0.513671875, |
| "rewards/simpleverify_reward/std": 0.5003018379211426, |
| "step": 389 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.24609375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4095.0, |
| "completions/mean_length": 2589.4921875, |
| "completions/mean_terminated_length": 2097.73046875, |
| "completions/min_length": 656.0, |
| "completions/min_terminated_length": 656.0, |
| "epoch": 0.276792051100071, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.20714245736598969, |
| "learning_rate": 2.22e-07, |
| "loss": 0.036, |
| "num_tokens": 597820411.0, |
| "reward": 0.505859375, |
| "reward_std": 0.29776811599731445, |
| "rewards/simpleverify_reward/mean": 0.505859375, |
| "rewards/simpleverify_reward/std": 0.5004546642303467, |
| "step": 390 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.259765625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4092.0, |
| "completions/mean_length": 2666.634765625, |
| "completions/mean_terminated_length": 2165.037109375, |
| "completions/min_length": 66.0, |
| "completions/min_terminated_length": 66.0, |
| "epoch": 0.27750177430801987, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.20907177031040192, |
| "learning_rate": 2.1999999999999998e-07, |
| "loss": 0.0609, |
| "num_tokens": 599237344.0, |
| "reward": 0.576171875, |
| "reward_std": 0.33809125423431396, |
| "rewards/simpleverify_reward/mean": 0.576171875, |
| "rewards/simpleverify_reward/std": 0.4946470856666565, |
| "step": 391 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.1875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4077.0, |
| "completions/mean_length": 2470.0625, |
| "completions/mean_terminated_length": 2094.84619140625, |
| "completions/min_length": 482.0, |
| "completions/min_terminated_length": 482.0, |
| "epoch": 0.2782114975159688, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.2034214287996292, |
| "learning_rate": 2.18e-07, |
| "loss": 0.0374, |
| "num_tokens": 600550320.0, |
| "reward": 0.609375, |
| "reward_std": 0.33818933367729187, |
| "rewards/simpleverify_reward/mean": 0.609375, |
| "rewards/simpleverify_reward/std": 0.48836761713027954, |
| "step": 392 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.23046875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4086.0, |
| "completions/mean_length": 2583.693359375, |
| "completions/mean_terminated_length": 2130.76904296875, |
| "completions/min_length": 470.0, |
| "completions/min_terminated_length": 470.0, |
| "epoch": 0.2789212207239177, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.21120710670948029, |
| "learning_rate": 2.1599999999999998e-07, |
| "loss": 0.0687, |
| "num_tokens": 601914083.0, |
| "reward": 0.568359375, |
| "reward_std": 0.3515923023223877, |
| "rewards/simpleverify_reward/mean": 0.568359375, |
| "rewards/simpleverify_reward/std": 0.4957893490791321, |
| "step": 393 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.248046875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4051.0, |
| "completions/mean_length": 2677.060546875, |
| "completions/mean_terminated_length": 2208.994873046875, |
| "completions/min_length": 681.0, |
| "completions/min_terminated_length": 681.0, |
| "epoch": 0.27963094393186655, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.22164218127727509, |
| "learning_rate": 2.1399999999999998e-07, |
| "loss": 0.0606, |
| "num_tokens": 603327586.0, |
| "reward": 0.564453125, |
| "reward_std": 0.3427889347076416, |
| "rewards/simpleverify_reward/mean": 0.564453125, |
| "rewards/simpleverify_reward/std": 0.49631330370903015, |
| "step": 394 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.255859375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4088.0, |
| "completions/mean_length": 2599.65625, |
| "completions/mean_terminated_length": 2085.165283203125, |
| "completions/min_length": 513.0, |
| "completions/min_terminated_length": 513.0, |
| "epoch": 0.2803406671398155, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.2094106376171112, |
| "learning_rate": 2.12e-07, |
| "loss": 0.0308, |
| "num_tokens": 604706114.0, |
| "reward": 0.548828125, |
| "reward_std": 0.31317460536956787, |
| "rewards/simpleverify_reward/mean": 0.548828125, |
| "rewards/simpleverify_reward/std": 0.498096764087677, |
| "step": 395 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.126953125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4091.0, |
| "completions/mean_length": 2222.607421875, |
| "completions/mean_terminated_length": 1950.190185546875, |
| "completions/min_length": 370.0, |
| "completions/min_terminated_length": 370.0, |
| "epoch": 0.28105039034776436, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.2299289107322693, |
| "learning_rate": 2.0999999999999997e-07, |
| "loss": 0.0397, |
| "num_tokens": 605887913.0, |
| "reward": 0.56640625, |
| "reward_std": 0.30883562564849854, |
| "rewards/simpleverify_reward/mean": 0.56640625, |
| "rewards/simpleverify_reward/std": 0.4960552453994751, |
| "step": 396 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.1953125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4087.0, |
| "completions/mean_length": 2592.517578125, |
| "completions/mean_terminated_length": 2227.5947265625, |
| "completions/min_length": 810.0, |
| "completions/min_terminated_length": 810.0, |
| "epoch": 0.2817601135557133, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.18124863505363464, |
| "learning_rate": 2.0799999999999998e-07, |
| "loss": 0.0271, |
| "num_tokens": 607270386.0, |
| "reward": 0.607421875, |
| "reward_std": 0.30711591243743896, |
| "rewards/simpleverify_reward/mean": 0.607421875, |
| "rewards/simpleverify_reward/std": 0.4888018071651459, |
| "step": 397 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.177734375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4056.0, |
| "completions/mean_length": 2422.72265625, |
| "completions/mean_terminated_length": 2061.04052734375, |
| "completions/min_length": 513.0, |
| "completions/min_terminated_length": 513.0, |
| "epoch": 0.28246983676366216, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.20141863822937012, |
| "learning_rate": 2.06e-07, |
| "loss": 0.0364, |
| "num_tokens": 608550772.0, |
| "reward": 0.623046875, |
| "reward_std": 0.3151366412639618, |
| "rewards/simpleverify_reward/mean": 0.623046875, |
| "rewards/simpleverify_reward/std": 0.4850969910621643, |
| "step": 398 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.12890625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4093.0, |
| "completions/mean_length": 2611.263671875, |
| "completions/mean_terminated_length": 2391.549560546875, |
| "completions/min_length": 679.0, |
| "completions/min_terminated_length": 679.0, |
| "epoch": 0.2831795599716111, |
| "frac_reward_zero_std": 0.09375, |
| "grad_norm": 0.22008755803108215, |
| "learning_rate": 2.0399999999999997e-07, |
| "loss": 0.0448, |
| "num_tokens": 609930811.0, |
| "reward": 0.5546875, |
| "reward_std": 0.36058270931243896, |
| "rewards/simpleverify_reward/mean": 0.5546875, |
| "rewards/simpleverify_reward/std": 0.49748632311820984, |
| "step": 399 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.298828125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 3991.0, |
| "completions/mean_length": 2707.166015625, |
| "completions/mean_terminated_length": 2115.267333984375, |
| "completions/min_length": 455.0, |
| "completions/min_terminated_length": 455.0, |
| "epoch": 0.28388928317955997, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.1868460774421692, |
| "learning_rate": 2.02e-07, |
| "loss": 0.0314, |
| "num_tokens": 611358048.0, |
| "reward": 0.384765625, |
| "reward_std": 0.276626318693161, |
| "rewards/simpleverify_reward/mean": 0.384765625, |
| "rewards/simpleverify_reward/std": 0.4870156943798065, |
| "step": 400 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.220703125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4096.0, |
| "completions/mean_length": 2678.25390625, |
| "completions/mean_terminated_length": 2276.73681640625, |
| "completions/min_length": 499.0, |
| "completions/min_terminated_length": 499.0, |
| "epoch": 0.28459900638750885, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.29515475034713745, |
| "learning_rate": 2e-07, |
| "loss": 0.0573, |
| "num_tokens": 612774306.0, |
| "reward": 0.49609375, |
| "reward_std": 0.34244608879089355, |
| "rewards/simpleverify_reward/mean": 0.49609375, |
| "rewards/simpleverify_reward/std": 0.5004737377166748, |
| "step": 401 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.18359375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4082.0, |
| "completions/mean_length": 2651.271484375, |
| "completions/mean_terminated_length": 2326.38037109375, |
| "completions/min_length": 531.0, |
| "completions/min_terminated_length": 531.0, |
| "epoch": 0.2853087295954578, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.19947567582130432, |
| "learning_rate": 1.98e-07, |
| "loss": 0.0485, |
| "num_tokens": 614184701.0, |
| "reward": 0.55078125, |
| "reward_std": 0.36476659774780273, |
| "rewards/simpleverify_reward/mean": 0.55078125, |
| "rewards/simpleverify_reward/std": 0.497901052236557, |
| "step": 402 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.18359375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4083.0, |
| "completions/mean_length": 2462.048828125, |
| "completions/mean_terminated_length": 2094.605224609375, |
| "completions/min_length": 566.0, |
| "completions/min_terminated_length": 566.0, |
| "epoch": 0.28601845280340665, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.19088207185268402, |
| "learning_rate": 1.96e-07, |
| "loss": 0.0082, |
| "num_tokens": 615493862.0, |
| "reward": 0.484375, |
| "reward_std": 0.339627742767334, |
| "rewards/simpleverify_reward/mean": 0.484375, |
| "rewards/simpleverify_reward/std": 0.5002445578575134, |
| "step": 403 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.171875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4078.0, |
| "completions/mean_length": 2311.41015625, |
| "completions/mean_terminated_length": 1941.0235595703125, |
| "completions/min_length": 556.0, |
| "completions/min_terminated_length": 556.0, |
| "epoch": 0.2867281760113556, |
| "frac_reward_zero_std": 0.125, |
| "grad_norm": 0.23991873860359192, |
| "learning_rate": 1.94e-07, |
| "loss": 0.04, |
| "num_tokens": 616726760.0, |
| "reward": 0.626953125, |
| "reward_std": 0.34829896688461304, |
| "rewards/simpleverify_reward/mean": 0.626953125, |
| "rewards/simpleverify_reward/std": 0.48408737778663635, |
| "step": 404 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.189453125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4060.0, |
| "completions/mean_length": 2491.587890625, |
| "completions/mean_terminated_length": 2116.580810546875, |
| "completions/min_length": 373.0, |
| "completions/min_terminated_length": 373.0, |
| "epoch": 0.28743789921930446, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.21169212460517883, |
| "learning_rate": 1.92e-07, |
| "loss": 0.063, |
| "num_tokens": 618057557.0, |
| "reward": 0.607421875, |
| "reward_std": 0.3342360556125641, |
| "rewards/simpleverify_reward/mean": 0.607421875, |
| "rewards/simpleverify_reward/std": 0.4888018071651459, |
| "step": 405 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.271484375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4087.0, |
| "completions/mean_length": 2806.130859375, |
| "completions/mean_terminated_length": 2325.455810546875, |
| "completions/min_length": 983.0, |
| "completions/min_terminated_length": 983.0, |
| "epoch": 0.2881476224272534, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.22285404801368713, |
| "learning_rate": 1.8999999999999998e-07, |
| "loss": 0.036, |
| "num_tokens": 619537336.0, |
| "reward": 0.490234375, |
| "reward_std": 0.3338390290737152, |
| "rewards/simpleverify_reward/mean": 0.490234375, |
| "rewards/simpleverify_reward/std": 0.5003935098648071, |
| "step": 406 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.232421875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4083.0, |
| "completions/mean_length": 2791.228515625, |
| "completions/mean_terminated_length": 2396.14501953125, |
| "completions/min_length": 821.0, |
| "completions/min_terminated_length": 821.0, |
| "epoch": 0.28885734563520227, |
| "frac_reward_zero_std": 0.09375, |
| "grad_norm": 0.21185067296028137, |
| "learning_rate": 1.88e-07, |
| "loss": 0.0209, |
| "num_tokens": 621010925.0, |
| "reward": 0.552734375, |
| "reward_std": 0.37926238775253296, |
| "rewards/simpleverify_reward/mean": 0.552734375, |
| "rewards/simpleverify_reward/std": 0.4976975917816162, |
| "step": 407 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.162109375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4096.0, |
| "completions/mean_length": 2362.775390625, |
| "completions/mean_terminated_length": 2027.4429931640625, |
| "completions/min_length": 599.0, |
| "completions/min_terminated_length": 599.0, |
| "epoch": 0.2895670688431512, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.20628294348716736, |
| "learning_rate": 1.86e-07, |
| "loss": 0.0195, |
| "num_tokens": 622272954.0, |
| "reward": 0.5703125, |
| "reward_std": 0.3565436899662018, |
| "rewards/simpleverify_reward/mean": 0.5703125, |
| "rewards/simpleverify_reward/std": 0.4955156147480011, |
| "step": 408 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.25, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4064.0, |
| "completions/mean_length": 2726.677734375, |
| "completions/mean_terminated_length": 2270.237060546875, |
| "completions/min_length": 562.0, |
| "completions/min_terminated_length": 562.0, |
| "epoch": 0.2902767920511001, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.20102724432945251, |
| "learning_rate": 1.8399999999999998e-07, |
| "loss": 0.0487, |
| "num_tokens": 623718837.0, |
| "reward": 0.46875, |
| "reward_std": 0.35779038071632385, |
| "rewards/simpleverify_reward/mean": 0.46875, |
| "rewards/simpleverify_reward/std": 0.4995105266571045, |
| "step": 409 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.197265625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4091.0, |
| "completions/mean_length": 2450.330078125, |
| "completions/mean_terminated_length": 2045.919677734375, |
| "completions/min_length": 390.0, |
| "completions/min_terminated_length": 390.0, |
| "epoch": 0.29098651525904895, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.20330436527729034, |
| "learning_rate": 1.82e-07, |
| "loss": 0.0489, |
| "num_tokens": 625023422.0, |
| "reward": 0.59765625, |
| "reward_std": 0.28087252378463745, |
| "rewards/simpleverify_reward/mean": 0.59765625, |
| "rewards/simpleverify_reward/std": 0.4908501207828522, |
| "step": 410 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.20703125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4096.0, |
| "completions/mean_length": 2602.982421875, |
| "completions/mean_terminated_length": 2213.1796875, |
| "completions/min_length": 576.0, |
| "completions/min_terminated_length": 576.0, |
| "epoch": 0.2916962384669979, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.20151516795158386, |
| "learning_rate": 1.8e-07, |
| "loss": 0.0239, |
| "num_tokens": 626404069.0, |
| "reward": 0.51953125, |
| "reward_std": 0.333137184381485, |
| "rewards/simpleverify_reward/mean": 0.51953125, |
| "rewards/simpleverify_reward/std": 0.5001069903373718, |
| "step": 411 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.140625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4089.0, |
| "completions/mean_length": 2367.087890625, |
| "completions/mean_terminated_length": 2084.175048828125, |
| "completions/min_length": 727.0, |
| "completions/min_terminated_length": 727.0, |
| "epoch": 0.29240596167494676, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.1748477816581726, |
| "learning_rate": 1.7799999999999998e-07, |
| "loss": 0.0403, |
| "num_tokens": 627659346.0, |
| "reward": 0.63671875, |
| "reward_std": 0.263751745223999, |
| "rewards/simpleverify_reward/mean": 0.63671875, |
| "rewards/simpleverify_reward/std": 0.4814152419567108, |
| "step": 412 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.15625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4074.0, |
| "completions/mean_length": 2492.1640625, |
| "completions/mean_terminated_length": 2195.157470703125, |
| "completions/min_length": 666.0, |
| "completions/min_terminated_length": 666.0, |
| "epoch": 0.2931156848828957, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.20372775197029114, |
| "learning_rate": 1.76e-07, |
| "loss": 0.0278, |
| "num_tokens": 628980902.0, |
| "reward": 0.537109375, |
| "reward_std": 0.3521461486816406, |
| "rewards/simpleverify_reward/mean": 0.537109375, |
| "rewards/simpleverify_reward/std": 0.4991086423397064, |
| "step": 413 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.220703125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4082.0, |
| "completions/mean_length": 2392.298828125, |
| "completions/mean_terminated_length": 1909.7969970703125, |
| "completions/min_length": 339.0, |
| "completions/min_terminated_length": 339.0, |
| "epoch": 0.29382540809084456, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.2087661325931549, |
| "learning_rate": 1.7399999999999997e-07, |
| "loss": 0.0229, |
| "num_tokens": 630257263.0, |
| "reward": 0.525390625, |
| "reward_std": 0.3238239586353302, |
| "rewards/simpleverify_reward/mean": 0.525390625, |
| "rewards/simpleverify_reward/std": 0.4998432695865631, |
| "step": 414 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.19921875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4031.0, |
| "completions/mean_length": 2455.6953125, |
| "completions/mean_terminated_length": 2047.6195068359375, |
| "completions/min_length": 585.0, |
| "completions/min_terminated_length": 585.0, |
| "epoch": 0.2945351312987935, |
| "frac_reward_zero_std": 0.125, |
| "grad_norm": 0.22420771420001984, |
| "learning_rate": 1.7199999999999998e-07, |
| "loss": 0.0371, |
| "num_tokens": 631563843.0, |
| "reward": 0.58984375, |
| "reward_std": 0.3680419325828552, |
| "rewards/simpleverify_reward/mean": 0.58984375, |
| "rewards/simpleverify_reward/std": 0.49234291911125183, |
| "step": 415 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.240234375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4093.0, |
| "completions/mean_length": 2694.7109375, |
| "completions/mean_terminated_length": 2251.6298828125, |
| "completions/min_length": 586.0, |
| "completions/min_terminated_length": 586.0, |
| "epoch": 0.29524485450674237, |
| "frac_reward_zero_std": 0.375, |
| "grad_norm": 0.1733558028936386, |
| "learning_rate": 1.7000000000000001e-07, |
| "loss": 0.039, |
| "num_tokens": 632998239.0, |
| "reward": 0.529296875, |
| "reward_std": 0.2837464511394501, |
| "rewards/simpleverify_reward/mean": 0.529296875, |
| "rewards/simpleverify_reward/std": 0.49962911009788513, |
| "step": 416 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.15234375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4079.0, |
| "completions/mean_length": 2237.58203125, |
| "completions/mean_terminated_length": 1903.5806884765625, |
| "completions/min_length": 580.0, |
| "completions/min_terminated_length": 580.0, |
| "epoch": 0.29595457771469125, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.1862211376428604, |
| "learning_rate": 1.68e-07, |
| "loss": 0.0084, |
| "num_tokens": 634185225.0, |
| "reward": 0.6484375, |
| "reward_std": 0.2632393538951874, |
| "rewards/simpleverify_reward/mean": 0.6484375, |
| "rewards/simpleverify_reward/std": 0.4779251217842102, |
| "step": 417 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.173828125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4078.0, |
| "completions/mean_length": 2280.76171875, |
| "completions/mean_terminated_length": 1898.8321533203125, |
| "completions/min_length": 603.0, |
| "completions/min_terminated_length": 603.0, |
| "epoch": 0.2966643009226402, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.207467719912529, |
| "learning_rate": 1.66e-07, |
| "loss": 0.0282, |
| "num_tokens": 635400959.0, |
| "reward": 0.546875, |
| "reward_std": 0.2981082797050476, |
| "rewards/simpleverify_reward/mean": 0.546875, |
| "rewards/simpleverify_reward/std": 0.4982847273349762, |
| "step": 418 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.28125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4078.0, |
| "completions/mean_length": 2729.48828125, |
| "completions/mean_terminated_length": 2194.766357421875, |
| "completions/min_length": 747.0, |
| "completions/min_terminated_length": 747.0, |
| "epoch": 0.29737402413058905, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.1952201873064041, |
| "learning_rate": 1.64e-07, |
| "loss": 0.0594, |
| "num_tokens": 636850201.0, |
| "reward": 0.48828125, |
| "reward_std": 0.3350939452648163, |
| "rewards/simpleverify_reward/mean": 0.48828125, |
| "rewards/simpleverify_reward/std": 0.5003514885902405, |
| "step": 419 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.197265625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4043.0, |
| "completions/mean_length": 2527.142578125, |
| "completions/mean_terminated_length": 2141.608154296875, |
| "completions/min_length": 381.0, |
| "completions/min_terminated_length": 381.0, |
| "epoch": 0.298083747338538, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.21695901453495026, |
| "learning_rate": 1.62e-07, |
| "loss": 0.0279, |
| "num_tokens": 638186002.0, |
| "reward": 0.576171875, |
| "reward_std": 0.3417827785015106, |
| "rewards/simpleverify_reward/mean": 0.576171875, |
| "rewards/simpleverify_reward/std": 0.4946470856666565, |
| "step": 420 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.17578125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4023.0, |
| "completions/mean_length": 2442.638671875, |
| "completions/mean_terminated_length": 2090.026123046875, |
| "completions/min_length": 514.0, |
| "completions/min_terminated_length": 514.0, |
| "epoch": 0.29879347054648686, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.20775683224201202, |
| "learning_rate": 1.6e-07, |
| "loss": 0.0372, |
| "num_tokens": 639480137.0, |
| "reward": 0.546875, |
| "reward_std": 0.3300327658653259, |
| "rewards/simpleverify_reward/mean": 0.546875, |
| "rewards/simpleverify_reward/std": 0.4982847273349762, |
| "step": 421 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.177734375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4086.0, |
| "completions/mean_length": 2371.712890625, |
| "completions/mean_terminated_length": 1999.0048828125, |
| "completions/min_length": 537.0, |
| "completions/min_terminated_length": 537.0, |
| "epoch": 0.2995031937544358, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.20475967228412628, |
| "learning_rate": 1.5799999999999999e-07, |
| "loss": 0.0315, |
| "num_tokens": 640740534.0, |
| "reward": 0.646484375, |
| "reward_std": 0.29911237955093384, |
| "rewards/simpleverify_reward/mean": 0.646484375, |
| "rewards/simpleverify_reward/std": 0.47852855920791626, |
| "step": 422 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.234375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4085.0, |
| "completions/mean_length": 2564.6953125, |
| "completions/mean_terminated_length": 2095.928466796875, |
| "completions/min_length": 614.0, |
| "completions/min_terminated_length": 614.0, |
| "epoch": 0.30021291696238467, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.19138917326927185, |
| "learning_rate": 1.56e-07, |
| "loss": 0.0265, |
| "num_tokens": 642101130.0, |
| "reward": 0.541015625, |
| "reward_std": 0.31755903363227844, |
| "rewards/simpleverify_reward/mean": 0.541015625, |
| "rewards/simpleverify_reward/std": 0.49880221486091614, |
| "step": 423 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.18359375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4074.0, |
| "completions/mean_length": 2560.54296875, |
| "completions/mean_terminated_length": 2215.248779296875, |
| "completions/min_length": 601.0, |
| "completions/min_terminated_length": 601.0, |
| "epoch": 0.30092264017033354, |
| "frac_reward_zero_std": 0.125, |
| "grad_norm": 0.2158213108778, |
| "learning_rate": 1.54e-07, |
| "loss": 0.0277, |
| "num_tokens": 643463632.0, |
| "reward": 0.609375, |
| "reward_std": 0.37185513973236084, |
| "rewards/simpleverify_reward/mean": 0.609375, |
| "rewards/simpleverify_reward/std": 0.48836761713027954, |
| "step": 424 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.236328125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4091.0, |
| "completions/mean_length": 2806.25390625, |
| "completions/mean_terminated_length": 2407.125244140625, |
| "completions/min_length": 683.0, |
| "completions/min_terminated_length": 683.0, |
| "epoch": 0.3016323633782825, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.17949548363685608, |
| "learning_rate": 1.5199999999999998e-07, |
| "loss": 0.0372, |
| "num_tokens": 644951570.0, |
| "reward": 0.5703125, |
| "reward_std": 0.32787638902664185, |
| "rewards/simpleverify_reward/mean": 0.5703125, |
| "rewards/simpleverify_reward/std": 0.4955156147480011, |
| "step": 425 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.228515625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4088.0, |
| "completions/mean_length": 2545.083984375, |
| "completions/mean_terminated_length": 2085.69873046875, |
| "completions/min_length": 734.0, |
| "completions/min_terminated_length": 734.0, |
| "epoch": 0.30234208658623135, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.19720643758773804, |
| "learning_rate": 1.5e-07, |
| "loss": 0.0428, |
| "num_tokens": 646300925.0, |
| "reward": 0.61328125, |
| "reward_std": 0.2884518802165985, |
| "rewards/simpleverify_reward/mean": 0.61328125, |
| "rewards/simpleverify_reward/std": 0.48747459053993225, |
| "step": 426 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.232421875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4079.0, |
| "completions/mean_length": 2626.689453125, |
| "completions/mean_terminated_length": 2181.78369140625, |
| "completions/min_length": 479.0, |
| "completions/min_terminated_length": 479.0, |
| "epoch": 0.3030518097941803, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.20522646605968475, |
| "learning_rate": 1.4799999999999998e-07, |
| "loss": 0.0268, |
| "num_tokens": 647692654.0, |
| "reward": 0.50390625, |
| "reward_std": 0.3583560883998871, |
| "rewards/simpleverify_reward/mean": 0.50390625, |
| "rewards/simpleverify_reward/std": 0.5004737377166748, |
| "step": 427 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.1953125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4068.0, |
| "completions/mean_length": 2491.482421875, |
| "completions/mean_terminated_length": 2102.036376953125, |
| "completions/min_length": 573.0, |
| "completions/min_terminated_length": 573.0, |
| "epoch": 0.30376153300212916, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.20029541850090027, |
| "learning_rate": 1.4599999999999998e-07, |
| "loss": 0.0654, |
| "num_tokens": 649010293.0, |
| "reward": 0.689453125, |
| "reward_std": 0.288011759519577, |
| "rewards/simpleverify_reward/mean": 0.689453125, |
| "rewards/simpleverify_reward/std": 0.46317005157470703, |
| "step": 428 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.177734375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4080.0, |
| "completions/mean_length": 2522.53515625, |
| "completions/mean_terminated_length": 2182.427734375, |
| "completions/min_length": 685.0, |
| "completions/min_terminated_length": 685.0, |
| "epoch": 0.3044712562100781, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.20148970186710358, |
| "learning_rate": 1.44e-07, |
| "loss": 0.0411, |
| "num_tokens": 650349863.0, |
| "reward": 0.55859375, |
| "reward_std": 0.34723415970802307, |
| "rewards/simpleverify_reward/mean": 0.55859375, |
| "rewards/simpleverify_reward/std": 0.4970405399799347, |
| "step": 429 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.349609375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 2968.71875, |
| "completions/mean_terminated_length": 2362.762939453125, |
| "completions/min_length": 765.0, |
| "completions/min_terminated_length": 765.0, |
| "epoch": 0.30518097941802697, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.1832769811153412, |
| "learning_rate": 1.4199999999999997e-07, |
| "loss": 0.0574, |
| "num_tokens": 651928759.0, |
| "reward": 0.46875, |
| "reward_std": 0.31103262305259705, |
| "rewards/simpleverify_reward/mean": 0.46875, |
| "rewards/simpleverify_reward/std": 0.4995105266571045, |
| "step": 430 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.21875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4047.0, |
| "completions/mean_length": 2518.2265625, |
| "completions/mean_terminated_length": 2076.449951171875, |
| "completions/min_length": 735.0, |
| "completions/min_terminated_length": 735.0, |
| "epoch": 0.3058907026259759, |
| "frac_reward_zero_std": 0.125, |
| "grad_norm": 0.20330217480659485, |
| "learning_rate": 1.4e-07, |
| "loss": 0.0263, |
| "num_tokens": 653262075.0, |
| "reward": 0.5234375, |
| "reward_std": 0.3495761752128601, |
| "rewards/simpleverify_reward/mean": 0.5234375, |
| "rewards/simpleverify_reward/std": 0.49993884563446045, |
| "step": 431 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.19921875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4082.0, |
| "completions/mean_length": 2619.62109375, |
| "completions/mean_terminated_length": 2252.326904296875, |
| "completions/min_length": 529.0, |
| "completions/min_terminated_length": 529.0, |
| "epoch": 0.3066004258339248, |
| "frac_reward_zero_std": 0.0625, |
| "grad_norm": 0.21751154959201813, |
| "learning_rate": 1.3800000000000002e-07, |
| "loss": 0.0391, |
| "num_tokens": 654657177.0, |
| "reward": 0.552734375, |
| "reward_std": 0.38228219747543335, |
| "rewards/simpleverify_reward/mean": 0.552734375, |
| "rewards/simpleverify_reward/std": 0.4976975917816162, |
| "step": 432 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.1484375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4091.0, |
| "completions/mean_length": 2274.912109375, |
| "completions/mean_terminated_length": 1957.4747314453125, |
| "completions/min_length": 587.0, |
| "completions/min_terminated_length": 587.0, |
| "epoch": 0.30731014904187365, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.20458753407001495, |
| "learning_rate": 1.36e-07, |
| "loss": 0.0205, |
| "num_tokens": 655864572.0, |
| "reward": 0.65625, |
| "reward_std": 0.3174334168434143, |
| "rewards/simpleverify_reward/mean": 0.65625, |
| "rewards/simpleverify_reward/std": 0.4754233956336975, |
| "step": 433 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.216796875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4067.0, |
| "completions/mean_length": 2541.388671875, |
| "completions/mean_terminated_length": 2111.059814453125, |
| "completions/min_length": 493.0, |
| "completions/min_terminated_length": 493.0, |
| "epoch": 0.3080198722498226, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.2247285395860672, |
| "learning_rate": 1.34e-07, |
| "loss": 0.0508, |
| "num_tokens": 657208259.0, |
| "reward": 0.54296875, |
| "reward_std": 0.33892136812210083, |
| "rewards/simpleverify_reward/mean": 0.54296875, |
| "rewards/simpleverify_reward/std": 0.49863746762275696, |
| "step": 434 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.234375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4096.0, |
| "completions/mean_length": 2654.669921875, |
| "completions/mean_terminated_length": 2213.4462890625, |
| "completions/min_length": 899.0, |
| "completions/min_terminated_length": 899.0, |
| "epoch": 0.30872959545777146, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.18462000787258148, |
| "learning_rate": 1.32e-07, |
| "loss": 0.0514, |
| "num_tokens": 658617706.0, |
| "reward": 0.5625, |
| "reward_std": 0.31916558742523193, |
| "rewards/simpleverify_reward/mean": 0.5625, |
| "rewards/simpleverify_reward/std": 0.49656352400779724, |
| "step": 435 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.228515625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4093.0, |
| "completions/mean_length": 2601.193359375, |
| "completions/mean_terminated_length": 2158.427978515625, |
| "completions/min_length": 648.0, |
| "completions/min_terminated_length": 648.0, |
| "epoch": 0.3094393186657204, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.1939697414636612, |
| "learning_rate": 1.3e-07, |
| "loss": 0.0283, |
| "num_tokens": 659993933.0, |
| "reward": 0.55859375, |
| "reward_std": 0.3170611262321472, |
| "rewards/simpleverify_reward/mean": 0.55859375, |
| "rewards/simpleverify_reward/std": 0.4970405399799347, |
| "step": 436 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.107421875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4059.0, |
| "completions/mean_length": 2269.04296875, |
| "completions/mean_terminated_length": 2049.16845703125, |
| "completions/min_length": 460.0, |
| "completions/min_terminated_length": 460.0, |
| "epoch": 0.31014904187366926, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.19287249445915222, |
| "learning_rate": 1.28e-07, |
| "loss": 0.0426, |
| "num_tokens": 661204131.0, |
| "reward": 0.556640625, |
| "reward_std": 0.28369688987731934, |
| "rewards/simpleverify_reward/mean": 0.556640625, |
| "rewards/simpleverify_reward/std": 0.49726733565330505, |
| "step": 437 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.22265625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4088.0, |
| "completions/mean_length": 2625.787109375, |
| "completions/mean_terminated_length": 2204.6708984375, |
| "completions/min_length": 700.0, |
| "completions/min_terminated_length": 700.0, |
| "epoch": 0.3108587650816182, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.18917374312877655, |
| "learning_rate": 1.26e-07, |
| "loss": 0.0453, |
| "num_tokens": 662589094.0, |
| "reward": 0.587890625, |
| "reward_std": 0.3366950452327728, |
| "rewards/simpleverify_reward/mean": 0.587890625, |
| "rewards/simpleverify_reward/std": 0.49269601702690125, |
| "step": 438 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.20703125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4020.0, |
| "completions/mean_length": 2585.107421875, |
| "completions/mean_terminated_length": 2190.637939453125, |
| "completions/min_length": 828.0, |
| "completions/min_terminated_length": 828.0, |
| "epoch": 0.31156848828956707, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.19559752941131592, |
| "learning_rate": 1.24e-07, |
| "loss": 0.0525, |
| "num_tokens": 663959085.0, |
| "reward": 0.6484375, |
| "reward_std": 0.32663968205451965, |
| "rewards/simpleverify_reward/mean": 0.6484375, |
| "rewards/simpleverify_reward/std": 0.4779251217842102, |
| "step": 439 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.173828125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4083.0, |
| "completions/mean_length": 2340.537109375, |
| "completions/mean_terminated_length": 1971.1844482421875, |
| "completions/min_length": 509.0, |
| "completions/min_terminated_length": 509.0, |
| "epoch": 0.31227821149751595, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.2090507596731186, |
| "learning_rate": 1.2199999999999998e-07, |
| "loss": 0.0191, |
| "num_tokens": 665206016.0, |
| "reward": 0.578125, |
| "reward_std": 0.2642821669578552, |
| "rewards/simpleverify_reward/mean": 0.578125, |
| "rewards/simpleverify_reward/std": 0.49434176087379456, |
| "step": 440 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.2734375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4057.0, |
| "completions/mean_length": 2836.95703125, |
| "completions/mean_terminated_length": 2363.12353515625, |
| "completions/min_length": 617.0, |
| "completions/min_terminated_length": 617.0, |
| "epoch": 0.3129879347054649, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.18568089604377747, |
| "learning_rate": 1.2e-07, |
| "loss": 0.043, |
| "num_tokens": 666707754.0, |
| "reward": 0.51171875, |
| "reward_std": 0.3048306107521057, |
| "rewards/simpleverify_reward/mean": 0.51171875, |
| "rewards/simpleverify_reward/std": 0.5003514885902405, |
| "step": 441 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.2421875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4082.0, |
| "completions/mean_length": 2634.58984375, |
| "completions/mean_terminated_length": 2167.541259765625, |
| "completions/min_length": 664.0, |
| "completions/min_terminated_length": 664.0, |
| "epoch": 0.31369765791341375, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.1874980479478836, |
| "learning_rate": 1.1799999999999998e-07, |
| "loss": 0.0168, |
| "num_tokens": 668105688.0, |
| "reward": 0.46875, |
| "reward_std": 0.3303448557853699, |
| "rewards/simpleverify_reward/mean": 0.46875, |
| "rewards/simpleverify_reward/std": 0.4995105266571045, |
| "step": 442 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.12890625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4064.0, |
| "completions/mean_length": 2362.162109375, |
| "completions/mean_terminated_length": 2105.585205078125, |
| "completions/min_length": 856.0, |
| "completions/min_terminated_length": 856.0, |
| "epoch": 0.3144073811213627, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.20077484846115112, |
| "learning_rate": 1.16e-07, |
| "loss": 0.0393, |
| "num_tokens": 669371419.0, |
| "reward": 0.65234375, |
| "reward_std": 0.3421695828437805, |
| "rewards/simpleverify_reward/mean": 0.65234375, |
| "rewards/simpleverify_reward/std": 0.47669193148612976, |
| "step": 443 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.1484375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4048.0, |
| "completions/mean_length": 2211.998046875, |
| "completions/mean_terminated_length": 1883.593994140625, |
| "completions/min_length": 567.0, |
| "completions/min_terminated_length": 567.0, |
| "epoch": 0.31511710432931156, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.195366770029068, |
| "learning_rate": 1.14e-07, |
| "loss": 0.0307, |
| "num_tokens": 670548042.0, |
| "reward": 0.671875, |
| "reward_std": 0.2847995162010193, |
| "rewards/simpleverify_reward/mean": 0.671875, |
| "rewards/simpleverify_reward/std": 0.4699897766113281, |
| "step": 444 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.21484375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4088.0, |
| "completions/mean_length": 2583.197265625, |
| "completions/mean_terminated_length": 2169.246337890625, |
| "completions/min_length": 63.0, |
| "completions/min_terminated_length": 63.0, |
| "epoch": 0.3158268275372605, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.2089722454547882, |
| "learning_rate": 1.12e-07, |
| "loss": 0.0536, |
| "num_tokens": 671917215.0, |
| "reward": 0.53515625, |
| "reward_std": 0.3474613130092621, |
| "rewards/simpleverify_reward/mean": 0.53515625, |
| "rewards/simpleverify_reward/std": 0.49925029277801514, |
| "step": 445 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.205078125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4077.0, |
| "completions/mean_length": 2448.65234375, |
| "completions/mean_terminated_length": 2023.660888671875, |
| "completions/min_length": 740.0, |
| "completions/min_terminated_length": 740.0, |
| "epoch": 0.31653655074520937, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.18917009234428406, |
| "learning_rate": 1.0999999999999999e-07, |
| "loss": 0.018, |
| "num_tokens": 673221885.0, |
| "reward": 0.544921875, |
| "reward_std": 0.32501986622810364, |
| "rewards/simpleverify_reward/mean": 0.544921875, |
| "rewards/simpleverify_reward/std": 0.4984649419784546, |
| "step": 446 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.216796875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4090.0, |
| "completions/mean_length": 2349.2734375, |
| "completions/mean_terminated_length": 1865.765625, |
| "completions/min_length": 513.0, |
| "completions/min_terminated_length": 513.0, |
| "epoch": 0.31724627395315824, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.19692853093147278, |
| "learning_rate": 1.0799999999999999e-07, |
| "loss": 0.0471, |
| "num_tokens": 674470025.0, |
| "reward": 0.623046875, |
| "reward_std": 0.2650914788246155, |
| "rewards/simpleverify_reward/mean": 0.623046875, |
| "rewards/simpleverify_reward/std": 0.4850969910621643, |
| "step": 447 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.21875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4073.0, |
| "completions/mean_length": 2697.513671875, |
| "completions/mean_terminated_length": 2305.9375, |
| "completions/min_length": 764.0, |
| "completions/min_terminated_length": 764.0, |
| "epoch": 0.3179559971611072, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.18412169814109802, |
| "learning_rate": 1.06e-07, |
| "loss": 0.0449, |
| "num_tokens": 675901440.0, |
| "reward": 0.537109375, |
| "reward_std": 0.31187042593955994, |
| "rewards/simpleverify_reward/mean": 0.537109375, |
| "rewards/simpleverify_reward/std": 0.4991086423397064, |
| "step": 448 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.208984375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4091.0, |
| "completions/mean_length": 2505.6171875, |
| "completions/mean_terminated_length": 2085.442138671875, |
| "completions/min_length": 480.0, |
| "completions/min_terminated_length": 480.0, |
| "epoch": 0.31866572036905605, |
| "frac_reward_zero_std": 0.125, |
| "grad_norm": 0.20870903134346008, |
| "learning_rate": 1.0399999999999999e-07, |
| "loss": 0.0597, |
| "num_tokens": 677232460.0, |
| "reward": 0.55078125, |
| "reward_std": 0.3743041157722473, |
| "rewards/simpleverify_reward/mean": 0.55078125, |
| "rewards/simpleverify_reward/std": 0.497901052236557, |
| "step": 449 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.33203125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4060.0, |
| "completions/mean_length": 2962.447265625, |
| "completions/mean_terminated_length": 2398.9853515625, |
| "completions/min_length": 747.0, |
| "completions/min_terminated_length": 747.0, |
| "epoch": 0.319375443577005, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.18738172948360443, |
| "learning_rate": 1.0199999999999999e-07, |
| "loss": 0.0341, |
| "num_tokens": 678799761.0, |
| "reward": 0.515625, |
| "reward_std": 0.3560102880001068, |
| "rewards/simpleverify_reward/mean": 0.515625, |
| "rewards/simpleverify_reward/std": 0.5002445578575134, |
| "step": 450 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.158203125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4090.0, |
| "completions/mean_length": 2532.12890625, |
| "completions/mean_terminated_length": 2238.22265625, |
| "completions/min_length": 506.0, |
| "completions/min_terminated_length": 506.0, |
| "epoch": 0.32008516678495386, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.1913900524377823, |
| "learning_rate": 1e-07, |
| "loss": 0.0341, |
| "num_tokens": 680138979.0, |
| "reward": 0.560546875, |
| "reward_std": 0.30554908514022827, |
| "rewards/simpleverify_reward/mean": 0.560546875, |
| "rewards/simpleverify_reward/std": 0.49680593609809875, |
| "step": 451 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.123046875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4078.0, |
| "completions/mean_length": 2321.8359375, |
| "completions/mean_terminated_length": 2072.89990234375, |
| "completions/min_length": 752.0, |
| "completions/min_terminated_length": 752.0, |
| "epoch": 0.3207948899929028, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.19899791479110718, |
| "learning_rate": 9.8e-08, |
| "loss": 0.0464, |
| "num_tokens": 681377215.0, |
| "reward": 0.583984375, |
| "reward_std": 0.3319348394870758, |
| "rewards/simpleverify_reward/mean": 0.583984375, |
| "rewards/simpleverify_reward/std": 0.493378221988678, |
| "step": 452 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.1796875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 3999.0, |
| "completions/mean_length": 2482.111328125, |
| "completions/mean_terminated_length": 2128.593017578125, |
| "completions/min_length": 760.0, |
| "completions/min_terminated_length": 760.0, |
| "epoch": 0.32150461320085166, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.20236912369728088, |
| "learning_rate": 9.6e-08, |
| "loss": 0.0328, |
| "num_tokens": 682686856.0, |
| "reward": 0.55078125, |
| "reward_std": 0.3160087466239929, |
| "rewards/simpleverify_reward/mean": 0.55078125, |
| "rewards/simpleverify_reward/std": 0.497901052236557, |
| "step": 453 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.177734375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4065.0, |
| "completions/mean_length": 2398.599609375, |
| "completions/mean_terminated_length": 2031.703125, |
| "completions/min_length": 556.0, |
| "completions/min_terminated_length": 556.0, |
| "epoch": 0.3222143364088006, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.19401243329048157, |
| "learning_rate": 9.4e-08, |
| "loss": 0.0584, |
| "num_tokens": 683963803.0, |
| "reward": 0.62109375, |
| "reward_std": 0.29754698276519775, |
| "rewards/simpleverify_reward/mean": 0.62109375, |
| "rewards/simpleverify_reward/std": 0.4855891764163971, |
| "step": 454 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.130859375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4044.0, |
| "completions/mean_length": 2265.833984375, |
| "completions/mean_terminated_length": 1990.2808837890625, |
| "completions/min_length": 414.0, |
| "completions/min_terminated_length": 414.0, |
| "epoch": 0.32292405961674947, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.22952602803707123, |
| "learning_rate": 9.199999999999999e-08, |
| "loss": 0.0448, |
| "num_tokens": 685169862.0, |
| "reward": 0.658203125, |
| "reward_std": 0.28380024433135986, |
| "rewards/simpleverify_reward/mean": 0.658203125, |
| "rewards/simpleverify_reward/std": 0.4747757613658905, |
| "step": 455 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.31640625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4087.0, |
| "completions/mean_length": 2856.16015625, |
| "completions/mean_terminated_length": 2282.29150390625, |
| "completions/min_length": 935.0, |
| "completions/min_terminated_length": 935.0, |
| "epoch": 0.32363378282469835, |
| "frac_reward_zero_std": 0.125, |
| "grad_norm": 0.20674359798431396, |
| "learning_rate": 9e-08, |
| "loss": 0.0314, |
| "num_tokens": 686689976.0, |
| "reward": 0.47265625, |
| "reward_std": 0.34934449195861816, |
| "rewards/simpleverify_reward/mean": 0.47265625, |
| "rewards/simpleverify_reward/std": 0.49974003434181213, |
| "step": 456 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.224609375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4076.0, |
| "completions/mean_length": 2630.287109375, |
| "completions/mean_terminated_length": 2205.710205078125, |
| "completions/min_length": 568.0, |
| "completions/min_terminated_length": 568.0, |
| "epoch": 0.3243435060326473, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.19664593040943146, |
| "learning_rate": 8.8e-08, |
| "loss": 0.0331, |
| "num_tokens": 688089595.0, |
| "reward": 0.525390625, |
| "reward_std": 0.3401782512664795, |
| "rewards/simpleverify_reward/mean": 0.525390625, |
| "rewards/simpleverify_reward/std": 0.4998432695865631, |
| "step": 457 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.251953125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4084.0, |
| "completions/mean_length": 2526.00390625, |
| "completions/mean_terminated_length": 1997.206298828125, |
| "completions/min_length": 636.0, |
| "completions/min_terminated_length": 636.0, |
| "epoch": 0.32505322924059615, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.2007729709148407, |
| "learning_rate": 8.599999999999999e-08, |
| "loss": 0.0368, |
| "num_tokens": 689424333.0, |
| "reward": 0.546875, |
| "reward_std": 0.32215848565101624, |
| "rewards/simpleverify_reward/mean": 0.546875, |
| "rewards/simpleverify_reward/std": 0.4982847273349762, |
| "step": 458 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.1484375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4091.0, |
| "completions/mean_length": 2411.619140625, |
| "completions/mean_terminated_length": 2118.011474609375, |
| "completions/min_length": 721.0, |
| "completions/min_terminated_length": 721.0, |
| "epoch": 0.3257629524485451, |
| "frac_reward_zero_std": 0.09375, |
| "grad_norm": 0.229851633310318, |
| "learning_rate": 8.4e-08, |
| "loss": 0.0424, |
| "num_tokens": 690712906.0, |
| "reward": 0.59375, |
| "reward_std": 0.37180325388908386, |
| "rewards/simpleverify_reward/mean": 0.59375, |
| "rewards/simpleverify_reward/std": 0.49161264300346375, |
| "step": 459 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.177734375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4091.0, |
| "completions/mean_length": 2571.98046875, |
| "completions/mean_terminated_length": 2242.560546875, |
| "completions/min_length": 604.0, |
| "completions/min_terminated_length": 604.0, |
| "epoch": 0.32647267565649396, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.19815188646316528, |
| "learning_rate": 8.2e-08, |
| "loss": 0.0258, |
| "num_tokens": 692071312.0, |
| "reward": 0.564453125, |
| "reward_std": 0.33011096715927124, |
| "rewards/simpleverify_reward/mean": 0.564453125, |
| "rewards/simpleverify_reward/std": 0.49631330370903015, |
| "step": 460 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.25390625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4086.0, |
| "completions/mean_length": 2651.5, |
| "completions/mean_terminated_length": 2159.916259765625, |
| "completions/min_length": 672.0, |
| "completions/min_terminated_length": 672.0, |
| "epoch": 0.3271823988644429, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.20204110443592072, |
| "learning_rate": 8e-08, |
| "loss": 0.0416, |
| "num_tokens": 693469216.0, |
| "reward": 0.52734375, |
| "reward_std": 0.3107943534851074, |
| "rewards/simpleverify_reward/mean": 0.52734375, |
| "rewards/simpleverify_reward/std": 0.49974003434181213, |
| "step": 461 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.1953125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4075.0, |
| "completions/mean_length": 2442.111328125, |
| "completions/mean_terminated_length": 2040.6820068359375, |
| "completions/min_length": 416.0, |
| "completions/min_terminated_length": 416.0, |
| "epoch": 0.32789212207239177, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.18333791196346283, |
| "learning_rate": 7.8e-08, |
| "loss": 0.0303, |
| "num_tokens": 694759449.0, |
| "reward": 0.671875, |
| "reward_std": 0.2824768126010895, |
| "rewards/simpleverify_reward/mean": 0.671875, |
| "rewards/simpleverify_reward/std": 0.4699897766113281, |
| "step": 462 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.2578125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4069.0, |
| "completions/mean_length": 2659.630859375, |
| "completions/mean_terminated_length": 2160.681640625, |
| "completions/min_length": 546.0, |
| "completions/min_terminated_length": 546.0, |
| "epoch": 0.32860184528034064, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.1875414401292801, |
| "learning_rate": 7.599999999999999e-08, |
| "loss": 0.0337, |
| "num_tokens": 696169692.0, |
| "reward": 0.4765625, |
| "reward_std": 0.3258887231349945, |
| "rewards/simpleverify_reward/mean": 0.4765625, |
| "rewards/simpleverify_reward/std": 0.49993884563446045, |
| "step": 463 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.193359375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4095.0, |
| "completions/mean_length": 2398.31640625, |
| "completions/mean_terminated_length": 1991.36572265625, |
| "completions/min_length": 502.0, |
| "completions/min_terminated_length": 502.0, |
| "epoch": 0.3293115684882896, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.22139491140842438, |
| "learning_rate": 7.399999999999999e-08, |
| "loss": 0.052, |
| "num_tokens": 697452814.0, |
| "reward": 0.63671875, |
| "reward_std": 0.3055083751678467, |
| "rewards/simpleverify_reward/mean": 0.63671875, |
| "rewards/simpleverify_reward/std": 0.4814152419567108, |
| "step": 464 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.224609375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4042.0, |
| "completions/mean_length": 2464.287109375, |
| "completions/mean_terminated_length": 1991.6246337890625, |
| "completions/min_length": 489.0, |
| "completions/min_terminated_length": 489.0, |
| "epoch": 0.33002129169623845, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.2113555371761322, |
| "learning_rate": 7.2e-08, |
| "loss": 0.0193, |
| "num_tokens": 698768001.0, |
| "reward": 0.58984375, |
| "reward_std": 0.30105698108673096, |
| "rewards/simpleverify_reward/mean": 0.58984375, |
| "rewards/simpleverify_reward/std": 0.49234291911125183, |
| "step": 465 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.287109375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4089.0, |
| "completions/mean_length": 2681.994140625, |
| "completions/mean_terminated_length": 2112.517822265625, |
| "completions/min_length": 521.0, |
| "completions/min_terminated_length": 521.0, |
| "epoch": 0.3307310149041874, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.19308342039585114, |
| "learning_rate": 7e-08, |
| "loss": 0.0289, |
| "num_tokens": 700189246.0, |
| "reward": 0.509765625, |
| "reward_std": 0.3248995840549469, |
| "rewards/simpleverify_reward/mean": 0.509765625, |
| "rewards/simpleverify_reward/std": 0.5003935098648071, |
| "step": 466 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.267578125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4082.0, |
| "completions/mean_length": 2677.640625, |
| "completions/mean_terminated_length": 2159.466552734375, |
| "completions/min_length": 621.0, |
| "completions/min_terminated_length": 621.0, |
| "epoch": 0.33144073811213626, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.1962798535823822, |
| "learning_rate": 6.8e-08, |
| "loss": 0.0541, |
| "num_tokens": 701601286.0, |
| "reward": 0.59765625, |
| "reward_std": 0.34637540578842163, |
| "rewards/simpleverify_reward/mean": 0.59765625, |
| "rewards/simpleverify_reward/std": 0.4908501207828522, |
| "step": 467 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.271484375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4067.0, |
| "completions/mean_length": 2686.1015625, |
| "completions/mean_terminated_length": 2160.697021484375, |
| "completions/min_length": 424.0, |
| "completions/min_terminated_length": 424.0, |
| "epoch": 0.3321504613200852, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.19502991437911987, |
| "learning_rate": 6.6e-08, |
| "loss": 0.0427, |
| "num_tokens": 703026506.0, |
| "reward": 0.568359375, |
| "reward_std": 0.36134985089302063, |
| "rewards/simpleverify_reward/mean": 0.568359375, |
| "rewards/simpleverify_reward/std": 0.4957893490791321, |
| "step": 468 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.22265625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4084.0, |
| "completions/mean_length": 2428.751953125, |
| "completions/mean_terminated_length": 1951.198486328125, |
| "completions/min_length": 625.0, |
| "completions/min_terminated_length": 625.0, |
| "epoch": 0.33286018452803406, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.19219937920570374, |
| "learning_rate": 6.4e-08, |
| "loss": 0.027, |
| "num_tokens": 704322507.0, |
| "reward": 0.525390625, |
| "reward_std": 0.26301252841949463, |
| "rewards/simpleverify_reward/mean": 0.525390625, |
| "rewards/simpleverify_reward/std": 0.4998432695865631, |
| "step": 469 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.302734375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4081.0, |
| "completions/mean_length": 2864.3984375, |
| "completions/mean_terminated_length": 2329.66943359375, |
| "completions/min_length": 759.0, |
| "completions/min_terminated_length": 759.0, |
| "epoch": 0.33356990773598294, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.16894395649433136, |
| "learning_rate": 6.2e-08, |
| "loss": 0.0535, |
| "num_tokens": 705844039.0, |
| "reward": 0.431640625, |
| "reward_std": 0.33326664566993713, |
| "rewards/simpleverify_reward/mean": 0.431640625, |
| "rewards/simpleverify_reward/std": 0.4957893490791321, |
| "step": 470 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.2421875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4074.0, |
| "completions/mean_length": 2560.2734375, |
| "completions/mean_terminated_length": 2069.47412109375, |
| "completions/min_length": 506.0, |
| "completions/min_terminated_length": 506.0, |
| "epoch": 0.33427963094393187, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.19876760244369507, |
| "learning_rate": 6e-08, |
| "loss": 0.0326, |
| "num_tokens": 707205891.0, |
| "reward": 0.533203125, |
| "reward_std": 0.32176676392555237, |
| "rewards/simpleverify_reward/mean": 0.533203125, |
| "rewards/simpleverify_reward/std": 0.4993842542171478, |
| "step": 471 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.19921875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4062.0, |
| "completions/mean_length": 2626.654296875, |
| "completions/mean_terminated_length": 2261.10986328125, |
| "completions/min_length": 111.0, |
| "completions/min_terminated_length": 111.0, |
| "epoch": 0.33498935415188075, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.2102649211883545, |
| "learning_rate": 5.8e-08, |
| "loss": 0.0299, |
| "num_tokens": 708596354.0, |
| "reward": 0.56640625, |
| "reward_std": 0.3318033814430237, |
| "rewards/simpleverify_reward/mean": 0.56640625, |
| "rewards/simpleverify_reward/std": 0.4960552453994751, |
| "step": 472 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.224609375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4093.0, |
| "completions/mean_length": 2480.357421875, |
| "completions/mean_terminated_length": 2012.35009765625, |
| "completions/min_length": 613.0, |
| "completions/min_terminated_length": 613.0, |
| "epoch": 0.3356990773598297, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.20455724000930786, |
| "learning_rate": 5.6e-08, |
| "loss": 0.0529, |
| "num_tokens": 709908985.0, |
| "reward": 0.51953125, |
| "reward_std": 0.3447939157485962, |
| "rewards/simpleverify_reward/mean": 0.51953125, |
| "rewards/simpleverify_reward/std": 0.5001069903373718, |
| "step": 473 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.173828125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4095.0, |
| "completions/mean_length": 2382.51953125, |
| "completions/mean_terminated_length": 2022.0, |
| "completions/min_length": 354.0, |
| "completions/min_terminated_length": 354.0, |
| "epoch": 0.33640880056777855, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.20652587711811066, |
| "learning_rate": 5.3999999999999994e-08, |
| "loss": 0.0221, |
| "num_tokens": 711169395.0, |
| "reward": 0.58203125, |
| "reward_std": 0.2907620668411255, |
| "rewards/simpleverify_reward/mean": 0.58203125, |
| "rewards/simpleverify_reward/std": 0.4937073290348053, |
| "step": 474 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.17578125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4089.0, |
| "completions/mean_length": 2526.796875, |
| "completions/mean_terminated_length": 2192.1328125, |
| "completions/min_length": 469.0, |
| "completions/min_terminated_length": 469.0, |
| "epoch": 0.3371185237757275, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.20143774151802063, |
| "learning_rate": 5.1999999999999996e-08, |
| "loss": 0.0462, |
| "num_tokens": 712509291.0, |
| "reward": 0.654296875, |
| "reward_std": 0.3276016414165497, |
| "rewards/simpleverify_reward/mean": 0.654296875, |
| "rewards/simpleverify_reward/std": 0.4760620892047882, |
| "step": 475 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.14453125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4076.0, |
| "completions/mean_length": 2127.32421875, |
| "completions/mean_terminated_length": 1794.716796875, |
| "completions/min_length": 564.0, |
| "completions/min_terminated_length": 564.0, |
| "epoch": 0.33782824698367636, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.1997940093278885, |
| "learning_rate": 5e-08, |
| "loss": 0.0351, |
| "num_tokens": 713642545.0, |
| "reward": 0.54296875, |
| "reward_std": 0.2737141251564026, |
| "rewards/simpleverify_reward/mean": 0.54296875, |
| "rewards/simpleverify_reward/std": 0.49863746762275696, |
| "step": 476 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.28125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4090.0, |
| "completions/mean_length": 2575.216796875, |
| "completions/mean_terminated_length": 1980.1278076171875, |
| "completions/min_length": 393.0, |
| "completions/min_terminated_length": 393.0, |
| "epoch": 0.3385379701916253, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.21228104829788208, |
| "learning_rate": 4.8e-08, |
| "loss": 0.052, |
| "num_tokens": 715016704.0, |
| "reward": 0.51953125, |
| "reward_std": 0.3314701318740845, |
| "rewards/simpleverify_reward/mean": 0.51953125, |
| "rewards/simpleverify_reward/std": 0.5001069903373718, |
| "step": 477 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.162109375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4094.0, |
| "completions/mean_length": 2458.94140625, |
| "completions/mean_terminated_length": 2142.214599609375, |
| "completions/min_length": 514.0, |
| "completions/min_terminated_length": 514.0, |
| "epoch": 0.33924769339957417, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.20244455337524414, |
| "learning_rate": 4.5999999999999995e-08, |
| "loss": 0.0366, |
| "num_tokens": 716324258.0, |
| "reward": 0.568359375, |
| "reward_std": 0.2914852201938629, |
| "rewards/simpleverify_reward/mean": 0.568359375, |
| "rewards/simpleverify_reward/std": 0.4957893490791321, |
| "step": 478 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.248046875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4092.0, |
| "completions/mean_length": 2619.390625, |
| "completions/mean_terminated_length": 2132.30126953125, |
| "completions/min_length": 619.0, |
| "completions/min_terminated_length": 619.0, |
| "epoch": 0.33995741660752304, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.18973155319690704, |
| "learning_rate": 4.4e-08, |
| "loss": 0.0361, |
| "num_tokens": 717711994.0, |
| "reward": 0.4609375, |
| "reward_std": 0.28933775424957275, |
| "rewards/simpleverify_reward/mean": 0.4609375, |
| "rewards/simpleverify_reward/std": 0.4989593029022217, |
| "step": 479 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.203125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4084.0, |
| "completions/mean_length": 2490.474609375, |
| "completions/mean_terminated_length": 2081.22314453125, |
| "completions/min_length": 437.0, |
| "completions/min_terminated_length": 437.0, |
| "epoch": 0.340667139815472, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.19357015192508698, |
| "learning_rate": 4.2e-08, |
| "loss": 0.0399, |
| "num_tokens": 719028509.0, |
| "reward": 0.572265625, |
| "reward_std": 0.29836899042129517, |
| "rewards/simpleverify_reward/mean": 0.572265625, |
| "rewards/simpleverify_reward/std": 0.4952339828014374, |
| "step": 480 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.22265625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4073.0, |
| "completions/mean_length": 2538.724609375, |
| "completions/mean_terminated_length": 2092.6708984375, |
| "completions/min_length": 527.0, |
| "completions/min_terminated_length": 527.0, |
| "epoch": 0.34137686302342085, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.2020120471715927, |
| "learning_rate": 4e-08, |
| "loss": 0.0146, |
| "num_tokens": 720370992.0, |
| "reward": 0.490234375, |
| "reward_std": 0.3300512135028839, |
| "rewards/simpleverify_reward/mean": 0.490234375, |
| "rewards/simpleverify_reward/std": 0.5003935098648071, |
| "step": 481 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.158203125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4089.0, |
| "completions/mean_length": 2545.5234375, |
| "completions/mean_terminated_length": 2254.134521484375, |
| "completions/min_length": 486.0, |
| "completions/min_terminated_length": 486.0, |
| "epoch": 0.3420865862313698, |
| "frac_reward_zero_std": 0.0625, |
| "grad_norm": 0.2268870770931244, |
| "learning_rate": 3.7999999999999996e-08, |
| "loss": 0.0549, |
| "num_tokens": 721717580.0, |
| "reward": 0.564453125, |
| "reward_std": 0.40695860981941223, |
| "rewards/simpleverify_reward/mean": 0.564453125, |
| "rewards/simpleverify_reward/std": 0.49631330370903015, |
| "step": 482 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.185546875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4038.0, |
| "completions/mean_length": 2472.9609375, |
| "completions/mean_terminated_length": 2103.203857421875, |
| "completions/min_length": 558.0, |
| "completions/min_terminated_length": 558.0, |
| "epoch": 0.34279630943931866, |
| "frac_reward_zero_std": 0.125, |
| "grad_norm": 0.21122798323631287, |
| "learning_rate": 3.6e-08, |
| "loss": 0.0319, |
| "num_tokens": 723021848.0, |
| "reward": 0.498046875, |
| "reward_std": 0.3750291168689728, |
| "rewards/simpleverify_reward/mean": 0.498046875, |
| "rewards/simpleverify_reward/std": 0.5004851818084717, |
| "step": 483 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.23828125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4088.0, |
| "completions/mean_length": 2631.126953125, |
| "completions/mean_terminated_length": 2172.884765625, |
| "completions/min_length": 67.0, |
| "completions/min_terminated_length": 67.0, |
| "epoch": 0.3435060326472676, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.19792085886001587, |
| "learning_rate": 3.4e-08, |
| "loss": 0.0357, |
| "num_tokens": 724418057.0, |
| "reward": 0.470703125, |
| "reward_std": 0.3424098491668701, |
| "rewards/simpleverify_reward/mean": 0.470703125, |
| "rewards/simpleverify_reward/std": 0.49962911009788513, |
| "step": 484 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.216796875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4081.0, |
| "completions/mean_length": 2422.490234375, |
| "completions/mean_terminated_length": 1959.2493896484375, |
| "completions/min_length": 423.0, |
| "completions/min_terminated_length": 423.0, |
| "epoch": 0.34421575585521647, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.2298462837934494, |
| "learning_rate": 3.2e-08, |
| "loss": 0.0122, |
| "num_tokens": 725704212.0, |
| "reward": 0.53125, |
| "reward_std": 0.32577192783355713, |
| "rewards/simpleverify_reward/mean": 0.53125, |
| "rewards/simpleverify_reward/std": 0.4995105266571045, |
| "step": 485 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.21875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4089.0, |
| "completions/mean_length": 2523.984375, |
| "completions/mean_terminated_length": 2083.820068359375, |
| "completions/min_length": 613.0, |
| "completions/min_terminated_length": 613.0, |
| "epoch": 0.34492547906316534, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.21126613020896912, |
| "learning_rate": 3e-08, |
| "loss": 0.0347, |
| "num_tokens": 727040556.0, |
| "reward": 0.525390625, |
| "reward_std": 0.34958311915397644, |
| "rewards/simpleverify_reward/mean": 0.525390625, |
| "rewards/simpleverify_reward/std": 0.4998432695865631, |
| "step": 486 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.19921875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4088.0, |
| "completions/mean_length": 2409.19921875, |
| "completions/mean_terminated_length": 1989.55615234375, |
| "completions/min_length": 497.0, |
| "completions/min_terminated_length": 497.0, |
| "epoch": 0.3456352022711143, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.22722694277763367, |
| "learning_rate": 2.8e-08, |
| "loss": 0.0399, |
| "num_tokens": 728324674.0, |
| "reward": 0.61328125, |
| "reward_std": 0.3058747947216034, |
| "rewards/simpleverify_reward/mean": 0.61328125, |
| "rewards/simpleverify_reward/std": 0.48747459053993225, |
| "step": 487 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.2265625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4096.0, |
| "completions/mean_length": 2757.87890625, |
| "completions/mean_terminated_length": 2365.904052734375, |
| "completions/min_length": 517.0, |
| "completions/min_terminated_length": 517.0, |
| "epoch": 0.34634492547906315, |
| "frac_reward_zero_std": 0.125, |
| "grad_norm": 0.1983136683702469, |
| "learning_rate": 2.5999999999999998e-08, |
| "loss": 0.0442, |
| "num_tokens": 729783172.0, |
| "reward": 0.4921875, |
| "reward_std": 0.3653619885444641, |
| "rewards/simpleverify_reward/mean": 0.4921875, |
| "rewards/simpleverify_reward/std": 0.5004279017448425, |
| "step": 488 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.22265625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4084.0, |
| "completions/mean_length": 2693.92578125, |
| "completions/mean_terminated_length": 2292.32666015625, |
| "completions/min_length": 630.0, |
| "completions/min_terminated_length": 630.0, |
| "epoch": 0.3470546486870121, |
| "frac_reward_zero_std": 0.21875, |
| "grad_norm": 0.20351643860340118, |
| "learning_rate": 2.4e-08, |
| "loss": 0.0217, |
| "num_tokens": 731208510.0, |
| "reward": 0.509765625, |
| "reward_std": 0.34122371673583984, |
| "rewards/simpleverify_reward/mean": 0.509765625, |
| "rewards/simpleverify_reward/std": 0.5003935098648071, |
| "step": 489 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.3046875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4083.0, |
| "completions/mean_length": 2812.48046875, |
| "completions/mean_terminated_length": 2250.039306640625, |
| "completions/min_length": 805.0, |
| "completions/min_terminated_length": 805.0, |
| "epoch": 0.34776437189496096, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.1772644817829132, |
| "learning_rate": 2.2e-08, |
| "loss": 0.0311, |
| "num_tokens": 732707748.0, |
| "reward": 0.427734375, |
| "reward_std": 0.31647050380706787, |
| "rewards/simpleverify_reward/mean": 0.427734375, |
| "rewards/simpleverify_reward/std": 0.4952339828014374, |
| "step": 490 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.21875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4037.0, |
| "completions/mean_length": 2507.548828125, |
| "completions/mean_terminated_length": 2062.782470703125, |
| "completions/min_length": 560.0, |
| "completions/min_terminated_length": 560.0, |
| "epoch": 0.3484740951029099, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.188960999250412, |
| "learning_rate": 2e-08, |
| "loss": 0.0395, |
| "num_tokens": 734036941.0, |
| "reward": 0.630859375, |
| "reward_std": 0.2751677632331848, |
| "rewards/simpleverify_reward/mean": 0.630859375, |
| "rewards/simpleverify_reward/std": 0.4830440282821655, |
| "step": 491 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.2421875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4095.0, |
| "completions/mean_length": 2527.130859375, |
| "completions/mean_terminated_length": 2025.7396240234375, |
| "completions/min_length": 410.0, |
| "completions/min_terminated_length": 410.0, |
| "epoch": 0.34918381831085876, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.21934561431407928, |
| "learning_rate": 1.8e-08, |
| "loss": 0.0378, |
| "num_tokens": 735378128.0, |
| "reward": 0.5234375, |
| "reward_std": 0.3594200611114502, |
| "rewards/simpleverify_reward/mean": 0.5234375, |
| "rewards/simpleverify_reward/std": 0.49993884563446045, |
| "step": 492 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.2421875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4090.0, |
| "completions/mean_length": 2541.4921875, |
| "completions/mean_terminated_length": 2044.690673828125, |
| "completions/min_length": 623.0, |
| "completions/min_terminated_length": 623.0, |
| "epoch": 0.34989354151880764, |
| "frac_reward_zero_std": 0.34375, |
| "grad_norm": 0.17691676318645477, |
| "learning_rate": 1.6e-08, |
| "loss": 0.0383, |
| "num_tokens": 736725308.0, |
| "reward": 0.630859375, |
| "reward_std": 0.2815321087837219, |
| "rewards/simpleverify_reward/mean": 0.630859375, |
| "rewards/simpleverify_reward/std": 0.4830440282821655, |
| "step": 493 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.181640625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4072.0, |
| "completions/mean_length": 2610.72265625, |
| "completions/mean_terminated_length": 2281.054931640625, |
| "completions/min_length": 578.0, |
| "completions/min_terminated_length": 578.0, |
| "epoch": 0.35060326472675657, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.1951213926076889, |
| "learning_rate": 1.4e-08, |
| "loss": 0.0514, |
| "num_tokens": 738108334.0, |
| "reward": 0.626953125, |
| "reward_std": 0.31276935338974, |
| "rewards/simpleverify_reward/mean": 0.626953125, |
| "rewards/simpleverify_reward/std": 0.48408737778663635, |
| "step": 494 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.16796875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4095.0, |
| "completions/mean_length": 2517.6796875, |
| "completions/mean_terminated_length": 2199.0517578125, |
| "completions/min_length": 443.0, |
| "completions/min_terminated_length": 443.0, |
| "epoch": 0.35131298793470545, |
| "frac_reward_zero_std": 0.28125, |
| "grad_norm": 0.19650261104106903, |
| "learning_rate": 1.2e-08, |
| "loss": 0.0346, |
| "num_tokens": 739445162.0, |
| "reward": 0.6171875, |
| "reward_std": 0.28051531314849854, |
| "rewards/simpleverify_reward/mean": 0.6171875, |
| "rewards/simpleverify_reward/std": 0.486548513174057, |
| "step": 495 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.14453125, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4088.0, |
| "completions/mean_length": 2296.611328125, |
| "completions/mean_terminated_length": 1992.60498046875, |
| "completions/min_length": 549.0, |
| "completions/min_terminated_length": 549.0, |
| "epoch": 0.3520227111426544, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.19922779500484467, |
| "learning_rate": 1e-08, |
| "loss": 0.0229, |
| "num_tokens": 740668291.0, |
| "reward": 0.65234375, |
| "reward_std": 0.30340611934661865, |
| "rewards/simpleverify_reward/mean": 0.65234375, |
| "rewards/simpleverify_reward/std": 0.47669193148612976, |
| "step": 496 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.216796875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4095.0, |
| "completions/mean_length": 2517.544921875, |
| "completions/mean_terminated_length": 2080.615966796875, |
| "completions/min_length": 537.0, |
| "completions/min_terminated_length": 537.0, |
| "epoch": 0.35273243435060325, |
| "frac_reward_zero_std": 0.1875, |
| "grad_norm": 0.21617993712425232, |
| "learning_rate": 8e-09, |
| "loss": 0.0583, |
| "num_tokens": 741996362.0, |
| "reward": 0.568359375, |
| "reward_std": 0.34969520568847656, |
| "rewards/simpleverify_reward/mean": 0.568359375, |
| "rewards/simpleverify_reward/std": 0.4957893490791321, |
| "step": 497 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.134765625, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4081.0, |
| "completions/mean_length": 2285.59765625, |
| "completions/mean_terminated_length": 2003.6163330078125, |
| "completions/min_length": 603.0, |
| "completions/min_terminated_length": 603.0, |
| "epoch": 0.3534421575585522, |
| "frac_reward_zero_std": 0.25, |
| "grad_norm": 0.21092663705348969, |
| "learning_rate": 6e-09, |
| "loss": 0.0333, |
| "num_tokens": 743210908.0, |
| "reward": 0.662109375, |
| "reward_std": 0.3217788338661194, |
| "rewards/simpleverify_reward/mean": 0.662109375, |
| "rewards/simpleverify_reward/std": 0.4734536409378052, |
| "step": 498 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.171875, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4072.0, |
| "completions/mean_length": 2503.658203125, |
| "completions/mean_terminated_length": 2173.172119140625, |
| "completions/min_length": 1.0, |
| "completions/min_terminated_length": 1.0, |
| "epoch": 0.35415188076650106, |
| "frac_reward_zero_std": 0.15625, |
| "grad_norm": 0.2160365730524063, |
| "learning_rate": 4e-09, |
| "loss": 0.0534, |
| "num_tokens": 744548221.0, |
| "reward": 0.55859375, |
| "reward_std": 0.3613226115703583, |
| "rewards/simpleverify_reward/mean": 0.55859375, |
| "rewards/simpleverify_reward/std": 0.4970405399799347, |
| "step": 499 |
| }, |
| { |
| "clip_ratio/high_max": 0.0, |
| "clip_ratio/high_mean": 0.0, |
| "clip_ratio/low_mean": 0.0, |
| "clip_ratio/low_min": 0.0, |
| "clip_ratio/region_mean": 0.0, |
| "completions/clipped_ratio": 0.224609375, |
| "completions/max_length": 4096.0, |
| "completions/max_terminated_length": 4035.0, |
| "completions/mean_length": 2561.30078125, |
| "completions/mean_terminated_length": 2116.740478515625, |
| "completions/min_length": 720.0, |
| "completions/min_terminated_length": 720.0, |
| "epoch": 0.35486160397445, |
| "frac_reward_zero_std": 0.3125, |
| "grad_norm": 0.19467459619045258, |
| "learning_rate": 2e-09, |
| "loss": 0.0451, |
| "num_tokens": 745906839.0, |
| "reward": 0.55078125, |
| "reward_std": 0.28372111916542053, |
| "rewards/simpleverify_reward/mean": 0.55078125, |
| "rewards/simpleverify_reward/std": 0.497901052236557, |
| "step": 500 |
| }, |
| { |
| "epoch": 0.35486160397445, |
| "step": 500, |
| "total_flos": 0.0, |
| "train_loss": 0.035182354090502486, |
| "train_runtime": 48959.3587, |
| "train_samples_per_second": 5.229, |
| "train_steps_per_second": 0.01 |
| } |
| ], |
| "logging_steps": 1, |
| "max_steps": 500, |
| "num_input_tokens_seen": 745906839, |
| "num_train_epochs": 1, |
| "save_steps": 100, |
| "stateful_callbacks": { |
| "TrainerControl": { |
| "args": { |
| "should_epoch_stop": false, |
| "should_evaluate": false, |
| "should_log": false, |
| "should_save": true, |
| "should_training_stop": true |
| }, |
| "attributes": {} |
| } |
| }, |
| "total_flos": 0.0, |
| "train_batch_size": 4, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
