{ "best_global_step": null, "best_metric": null, "best_model_checkpoint": null, "epoch": 0.04, "eval_steps": 50, "global_step": 100, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0004, "frac_reward_zero_std": 0.75, "grad_norm": 31.350610733032227, "kl": 0.0, "learning_rate": 0.0, "loss": -0.0, "num_tokens": 8880.0, "reward": 0.8125, "reward_std": 0.125, "rewards/compute_reward/mean": 0.8125, "rewards/compute_reward/std": 0.40311288833618164, "step": 1 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0008, "frac_reward_zero_std": 0.75, "grad_norm": 40.88920593261719, "kl": 0.0, "learning_rate": 4e-08, "loss": 0.0, "num_tokens": 17760.0, "reward": 0.375, "reward_std": 0.14433756470680237, "rewards/compute_reward/mean": 0.375, "rewards/compute_reward/std": 0.5, "step": 2 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0012, "frac_reward_zero_std": 0.75, "grad_norm": 41.24170684814453, "kl": 0.0013443923089653254, "learning_rate": 8e-08, "loss": 0.0, "num_tokens": 26640.0, "reward": 0.3125, "reward_std": 0.125, "rewards/compute_reward/mean": 0.3125, "rewards/compute_reward/std": 0.4787135720252991, "step": 3 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0016, "frac_reward_zero_std": 1.0, "grad_norm": 0.025073383003473282, "kl": 0.0012363345449557528, "learning_rate": 1.2e-07, "loss": 0.0, "num_tokens": 35520.0, "reward": 0.25, "reward_std": 0.0, "rewards/compute_reward/mean": 0.25, "rewards/compute_reward/std": 0.44721361994743347, "step": 4 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.002, "frac_reward_zero_std": 0.5, "grad_norm": 45.65277862548828, "kl": 0.0009983214986277744, "learning_rate": 1.6e-07, "loss": 0.0, "num_tokens": 44400.0, "reward": 0.625, "reward_std": 0.25, "rewards/compute_reward/mean": 0.625, "rewards/compute_reward/std": 0.5, "step": 5 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0024, "frac_reward_zero_std": 0.5, "grad_norm": 40.60847854614258, "kl": 0.0010500029457034543, "learning_rate": 2e-07, "loss": 0.0, "num_tokens": 53280.0, "reward": 0.5, "reward_std": 0.25, "rewards/compute_reward/mean": 0.5, "rewards/compute_reward/std": 0.5163977742195129, "step": 6 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0028, "frac_reward_zero_std": 0.5, "grad_norm": 37.455440521240234, "kl": 0.0013627598236780614, "learning_rate": 2.4e-07, "loss": 0.0, "num_tokens": 62160.0, "reward": 0.5625, "reward_std": 0.26933756470680237, "rewards/compute_reward/mean": 0.5625, "rewards/compute_reward/std": 0.5123475790023804, "step": 7 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0032, "frac_reward_zero_std": 0.5, "grad_norm": 49.363807678222656, "kl": 0.0019289999618194997, "learning_rate": 2.8e-07, "loss": 0.0, "num_tokens": 71040.0, "reward": 0.4375, "reward_std": 0.26933756470680237, "rewards/compute_reward/mean": 0.4375, "rewards/compute_reward/std": 0.5123475790023804, "step": 8 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0036, "frac_reward_zero_std": 1.0, "grad_norm": 0.055168308317661285, "kl": 0.0023864044342190027, "learning_rate": 3.2e-07, "loss": 0.0, "num_tokens": 79920.0, "reward": 0.5, "reward_std": 0.0, "rewards/compute_reward/mean": 0.5, "rewards/compute_reward/std": 0.5163977742195129, "step": 9 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.004, "frac_reward_zero_std": 0.75, "grad_norm": 27.4397029876709, "kl": 0.0029534806380979717, "learning_rate": 3.6e-07, "loss": 0.0, "num_tokens": 88800.0, "reward": 0.3125, "reward_std": 0.125, "rewards/compute_reward/mean": 0.3125, "rewards/compute_reward/std": 0.4787135720252991, "step": 10 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0044, "frac_reward_zero_std": 0.5, "grad_norm": 34.00825881958008, "kl": 0.004601971711963415, "learning_rate": 4e-07, "loss": 0.0, "num_tokens": 97680.0, "reward": 0.1875, "reward_std": 0.26933756470680237, "rewards/compute_reward/mean": 0.1875, "rewards/compute_reward/std": 0.40311288833618164, "step": 11 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0048, "frac_reward_zero_std": 0.5, "grad_norm": 41.115177154541016, "kl": 0.005590100074186921, "learning_rate": 4.3999999999999997e-07, "loss": 0.0, "num_tokens": 106560.0, "reward": 0.5, "reward_std": 0.25, "rewards/compute_reward/mean": 0.5, "rewards/compute_reward/std": 0.5163977742195129, "step": 12 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0052, "frac_reward_zero_std": 0.5, "grad_norm": 37.690364837646484, "kl": 0.008083905442617834, "learning_rate": 4.8e-07, "loss": 0.0, "num_tokens": 115440.0, "reward": 0.25, "reward_std": 0.28867512941360474, "rewards/compute_reward/mean": 0.25, "rewards/compute_reward/std": 0.44721361994743347, "step": 13 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0056, "frac_reward_zero_std": 0.5, "grad_norm": 44.36626052856445, "kl": 0.01064743846654892, "learning_rate": 5.2e-07, "loss": 0.0, "num_tokens": 124320.0, "reward": 0.25, "reward_std": 0.25, "rewards/compute_reward/mean": 0.25, "rewards/compute_reward/std": 0.44721361994743347, "step": 14 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.006, "frac_reward_zero_std": 0.25, "grad_norm": 35.651023864746094, "kl": 0.01316027413122356, "learning_rate": 5.6e-07, "loss": 0.0001, "num_tokens": 133200.0, "reward": 0.75, "reward_std": 0.39433756470680237, "rewards/compute_reward/mean": 0.75, "rewards/compute_reward/std": 0.44721361994743347, "step": 15 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0064, "frac_reward_zero_std": 0.25, "grad_norm": 52.94963836669922, "kl": 0.020378376357257366, "learning_rate": 6e-07, "loss": 0.0001, "num_tokens": 142080.0, "reward": 0.4375, "reward_std": 0.41367512941360474, "rewards/compute_reward/mean": 0.4375, "rewards/compute_reward/std": 0.5123475790023804, "step": 16 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0068, "frac_reward_zero_std": 0.75, "grad_norm": 36.93665313720703, "kl": 0.025172261521220207, "learning_rate": 6.4e-07, "loss": 0.0001, "num_tokens": 150960.0, "reward": 0.3125, "reward_std": 0.125, "rewards/compute_reward/mean": 0.3125, "rewards/compute_reward/std": 0.4787135720252991, "step": 17 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0072, "frac_reward_zero_std": 1.0, "grad_norm": 0.5544130206108093, "kl": 0.032024398911744356, "learning_rate": 6.800000000000001e-07, "loss": 0.0001, "num_tokens": 159840.0, "reward": 0.25, "reward_std": 0.0, "rewards/compute_reward/mean": 0.25, "rewards/compute_reward/std": 0.44721361994743347, "step": 18 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0076, "frac_reward_zero_std": 0.25, "grad_norm": 44.66011047363281, "kl": 0.039961141999810934, "learning_rate": 7.2e-07, "loss": 0.0002, "num_tokens": 168720.0, "reward": 0.25, "reward_std": 0.39433756470680237, "rewards/compute_reward/mean": 0.25, "rewards/compute_reward/std": 0.44721361994743347, "step": 19 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.008, "frac_reward_zero_std": 0.75, "grad_norm": 28.504152297973633, "kl": 0.036296515725553036, "learning_rate": 7.599999999999999e-07, "loss": 0.0001, "num_tokens": 177600.0, "reward": 0.9375, "reward_std": 0.125, "rewards/compute_reward/mean": 0.9375, "rewards/compute_reward/std": 0.25, "step": 20 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0084, "frac_reward_zero_std": 0.25, "grad_norm": 50.410888671875, "kl": 0.04341549798846245, "learning_rate": 8e-07, "loss": 0.0002, "num_tokens": 186480.0, "reward": 0.375, "reward_std": 0.39433756470680237, "rewards/compute_reward/mean": 0.375, "rewards/compute_reward/std": 0.5, "step": 21 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0088, "frac_reward_zero_std": 0.5, "grad_norm": 51.3606071472168, "kl": 0.050860715098679066, "learning_rate": 8.399999999999999e-07, "loss": 0.0002, "num_tokens": 195360.0, "reward": 0.5625, "reward_std": 0.26933756470680237, "rewards/compute_reward/mean": 0.5625, "rewards/compute_reward/std": 0.5123475790023804, "step": 22 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0092, "frac_reward_zero_std": 0.75, "grad_norm": 24.91278076171875, "kl": 0.06213559303432703, "learning_rate": 8.799999999999999e-07, "loss": 0.0002, "num_tokens": 204240.0, "reward": 0.3125, "reward_std": 0.125, "rewards/compute_reward/mean": 0.3125, "rewards/compute_reward/std": 0.4787135720252991, "step": 23 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0096, "frac_reward_zero_std": 0.5, "grad_norm": 39.62419891357422, "kl": 0.0647846395149827, "learning_rate": 9.2e-07, "loss": 0.0003, "num_tokens": 213120.0, "reward": 0.625, "reward_std": 0.25, "rewards/compute_reward/mean": 0.625, "rewards/compute_reward/std": 0.5, "step": 24 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.01, "frac_reward_zero_std": 0.25, "grad_norm": 63.561767578125, "kl": 0.0892756637185812, "learning_rate": 9.6e-07, "loss": 0.0004, "num_tokens": 222000.0, "reward": 0.5625, "reward_std": 0.41367512941360474, "rewards/compute_reward/mean": 0.5625, "rewards/compute_reward/std": 0.5123475790023804, "step": 25 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0104, "frac_reward_zero_std": 0.5, "grad_norm": 42.179954528808594, "kl": 0.10400468856096268, "learning_rate": 1e-06, "loss": 0.0004, "num_tokens": 230880.0, "reward": 0.5, "reward_std": 0.25, "rewards/compute_reward/mean": 0.5, "rewards/compute_reward/std": 0.5163977742195129, "step": 26 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0108, "frac_reward_zero_std": 0.75, "grad_norm": 25.707307815551758, "kl": 0.1136582437902689, "learning_rate": 1.04e-06, "loss": 0.0005, "num_tokens": 239760.0, "reward": 0.125, "reward_std": 0.14433756470680237, "rewards/compute_reward/mean": 0.125, "rewards/compute_reward/std": 0.3415650427341461, "step": 27 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0112, "frac_reward_zero_std": 0.75, "grad_norm": 26.41197967529297, "kl": 0.13468273170292377, "learning_rate": 1.08e-06, "loss": 0.0005, "num_tokens": 248640.0, "reward": 0.5625, "reward_std": 0.125, "rewards/compute_reward/mean": 0.5625, "rewards/compute_reward/std": 0.5123475790023804, "step": 28 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0116, "frac_reward_zero_std": 0.75, "grad_norm": 35.851348876953125, "kl": 0.1477714516222477, "learning_rate": 1.12e-06, "loss": 0.0006, "num_tokens": 257520.0, "reward": 0.625, "reward_std": 0.14433756470680237, "rewards/compute_reward/mean": 0.625, "rewards/compute_reward/std": 0.5, "step": 29 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.012, "frac_reward_zero_std": 0.75, "grad_norm": 29.354488372802734, "kl": 0.14891785010695457, "learning_rate": 1.16e-06, "loss": 0.0006, "num_tokens": 266400.0, "reward": 0.6875, "reward_std": 0.125, "rewards/compute_reward/mean": 0.6875, "rewards/compute_reward/std": 0.4787135720252991, "step": 30 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0124, "frac_reward_zero_std": 0.5, "grad_norm": 39.174476623535156, "kl": 0.15899883769452572, "learning_rate": 1.2e-06, "loss": 0.0006, "num_tokens": 275280.0, "reward": 0.25, "reward_std": 0.28867512941360474, "rewards/compute_reward/mean": 0.25, "rewards/compute_reward/std": 0.44721361994743347, "step": 31 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0128, "frac_reward_zero_std": 0.75, "grad_norm": 44.24154281616211, "kl": 0.22540221363306046, "learning_rate": 1.24e-06, "loss": 0.0009, "num_tokens": 284160.0, "reward": 0.5625, "reward_std": 0.125, "rewards/compute_reward/mean": 0.5625, "rewards/compute_reward/std": 0.5123475790023804, "step": 32 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0132, "frac_reward_zero_std": 0.5, "grad_norm": 32.92380142211914, "kl": 0.19463030248880386, "learning_rate": 1.28e-06, "loss": 0.0008, "num_tokens": 293040.0, "reward": 0.5, "reward_std": 0.25, "rewards/compute_reward/mean": 0.5, "rewards/compute_reward/std": 0.5163977742195129, "step": 33 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0136, "frac_reward_zero_std": 0.75, "grad_norm": 32.65846633911133, "kl": 0.2380228154361248, "learning_rate": 1.32e-06, "loss": 0.001, "num_tokens": 301920.0, "reward": 0.5625, "reward_std": 0.125, "rewards/compute_reward/mean": 0.5625, "rewards/compute_reward/std": 0.5123475790023804, "step": 34 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.014, "frac_reward_zero_std": 0.5, "grad_norm": 77.5787582397461, "kl": 0.21238164231181145, "learning_rate": 1.3600000000000001e-06, "loss": 0.0008, "num_tokens": 310800.0, "reward": 0.125, "reward_std": 0.25, "rewards/compute_reward/mean": 0.125, "rewards/compute_reward/std": 0.3415650427341461, "step": 35 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0144, "frac_reward_zero_std": 0.75, "grad_norm": 29.662382125854492, "kl": 0.29322075098752975, "learning_rate": 1.4e-06, "loss": 0.0012, "num_tokens": 319680.0, "reward": 0.0625, "reward_std": 0.125, "rewards/compute_reward/mean": 0.0625, "rewards/compute_reward/std": 0.25, "step": 36 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0148, "frac_reward_zero_std": 0.5, "grad_norm": 48.11992263793945, "kl": 0.25715991109609604, "learning_rate": 1.44e-06, "loss": 0.001, "num_tokens": 328560.0, "reward": 0.125, "reward_std": 0.25, "rewards/compute_reward/mean": 0.125, "rewards/compute_reward/std": 0.3415650427341461, "step": 37 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0152, "frac_reward_zero_std": 0.75, "grad_norm": 38.42544937133789, "kl": 0.2538900189101696, "learning_rate": 1.48e-06, "loss": 0.001, "num_tokens": 337440.0, "reward": 0.125, "reward_std": 0.14433756470680237, "rewards/compute_reward/mean": 0.125, "rewards/compute_reward/std": 0.3415650427341461, "step": 38 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0156, "frac_reward_zero_std": 0.0, "grad_norm": 76.46257019042969, "kl": 0.29158805310726166, "learning_rate": 1.5199999999999998e-06, "loss": 0.0012, "num_tokens": 346320.0, "reward": 0.4375, "reward_std": 0.5580127239227295, "rewards/compute_reward/mean": 0.4375, "rewards/compute_reward/std": 0.5123475790023804, "step": 39 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.016, "frac_reward_zero_std": 0.5, "grad_norm": 50.54710388183594, "kl": 0.42541511356830597, "learning_rate": 1.5599999999999999e-06, "loss": 0.0017, "num_tokens": 355200.0, "reward": 0.3125, "reward_std": 0.26933756470680237, "rewards/compute_reward/mean": 0.3125, "rewards/compute_reward/std": 0.4787135720252991, "step": 40 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0164, "frac_reward_zero_std": 0.5, "grad_norm": 47.023494720458984, "kl": 0.3670000210404396, "learning_rate": 1.6e-06, "loss": 0.0015, "num_tokens": 364080.0, "reward": 0.3125, "reward_std": 0.26933756470680237, "rewards/compute_reward/mean": 0.3125, "rewards/compute_reward/std": 0.4787135720252991, "step": 41 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0168, "frac_reward_zero_std": 0.75, "grad_norm": 32.30452346801758, "kl": 0.38483476638793945, "learning_rate": 1.6399999999999998e-06, "loss": 0.0015, "num_tokens": 372960.0, "reward": 0.0625, "reward_std": 0.125, "rewards/compute_reward/mean": 0.0625, "rewards/compute_reward/std": 0.25, "step": 42 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0172, "frac_reward_zero_std": 0.5, "grad_norm": 42.685611724853516, "kl": 0.3910580351948738, "learning_rate": 1.6799999999999998e-06, "loss": 0.0016, "num_tokens": 381840.0, "reward": 0.1875, "reward_std": 0.26933756470680237, "rewards/compute_reward/mean": 0.1875, "rewards/compute_reward/std": 0.40311288833618164, "step": 43 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0176, "frac_reward_zero_std": 0.75, "grad_norm": 39.6999626159668, "kl": 0.4346514791250229, "learning_rate": 1.7199999999999998e-06, "loss": 0.0017, "num_tokens": 390720.0, "reward": 0.4375, "reward_std": 0.125, "rewards/compute_reward/mean": 0.4375, "rewards/compute_reward/std": 0.5123475790023804, "step": 44 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.018, "frac_reward_zero_std": 0.75, "grad_norm": 32.37580490112305, "kl": 0.38541124016046524, "learning_rate": 1.7599999999999999e-06, "loss": 0.0015, "num_tokens": 399600.0, "reward": 0.0625, "reward_std": 0.125, "rewards/compute_reward/mean": 0.0625, "rewards/compute_reward/std": 0.25, "step": 45 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0184, "frac_reward_zero_std": 0.5, "grad_norm": 37.88615798950195, "kl": 0.3760378360748291, "learning_rate": 1.8e-06, "loss": 0.0015, "num_tokens": 408480.0, "reward": 0.375, "reward_std": 0.25, "rewards/compute_reward/mean": 0.375, "rewards/compute_reward/std": 0.5, "step": 46 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0188, "frac_reward_zero_std": 0.5, "grad_norm": 50.25699996948242, "kl": 0.4850776717066765, "learning_rate": 1.84e-06, "loss": 0.0019, "num_tokens": 417360.0, "reward": 0.1875, "reward_std": 0.26933756470680237, "rewards/compute_reward/mean": 0.1875, "rewards/compute_reward/std": 0.40311288833618164, "step": 47 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0192, "frac_reward_zero_std": 1.0, "grad_norm": 0.5161222815513611, "kl": 0.3658723160624504, "learning_rate": 1.8799999999999998e-06, "loss": 0.0015, "num_tokens": 426240.0, "reward": 0.25, "reward_std": 0.0, "rewards/compute_reward/mean": 0.25, "rewards/compute_reward/std": 0.44721361994743347, "step": 48 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0196, "frac_reward_zero_std": 0.5, "grad_norm": 37.257171630859375, "kl": 0.3248850479722023, "learning_rate": 1.92e-06, "loss": 0.0013, "num_tokens": 435120.0, "reward": 0.875, "reward_std": 0.25, "rewards/compute_reward/mean": 0.875, "rewards/compute_reward/std": 0.3415650427341461, "step": 49 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.02, "frac_reward_zero_std": 0.5, "grad_norm": 40.35070037841797, "kl": 0.3114013075828552, "learning_rate": 1.96e-06, "loss": 0.0012, "num_tokens": 444000.0, "reward": 0.375, "reward_std": 0.25, "rewards/compute_reward/mean": 0.375, "rewards/compute_reward/std": 0.5, "step": 50 }, { "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.02, "eval_loss": 0.0003695653285831213, "eval_runtime": 12968.8828, "eval_samples_per_second": 0.015, "eval_steps_per_second": 0.004, "frac_reward_zero_std": 0.54, "num_tokens": 555000.0, "reward": 0.445, "reward_std": 0.23928203105926513, "rewards/compute_reward/mean": 0.445, "rewards/compute_reward/std": 0.23928203105926513, "step": 50 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0204, "frac_reward_zero_std": 0.75, "grad_norm": 35.21977233886719, "kl": 0.40479009598493576, "learning_rate": 2e-06, "loss": -0.1234, "num_tokens": 563880.0, "reward": 0.5625, "reward_std": 0.125, "rewards/compute_reward/mean": 0.5625, "rewards/compute_reward/std": 0.5123475790023804, "step": 51 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0208, "frac_reward_zero_std": 0.75, "grad_norm": 17.721899032592773, "kl": 0.3238118216395378, "learning_rate": 1.9999756307053944e-06, "loss": 0.1263, "num_tokens": 572760.0, "reward": 0.875, "reward_std": 0.14433756470680237, "rewards/compute_reward/mean": 0.875, "rewards/compute_reward/std": 0.3415650427341461, "step": 52 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0212, "frac_reward_zero_std": 0.5, "grad_norm": 48.17813491821289, "kl": 0.4712698385119438, "learning_rate": 1.999902524009304e-06, "loss": -0.1231, "num_tokens": 581640.0, "reward": 0.25, "reward_std": 0.25, "rewards/compute_reward/mean": 0.25, "rewards/compute_reward/std": 0.44721361994743347, "step": 53 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0216, "frac_reward_zero_std": 1.0, "grad_norm": 21.3551025390625, "kl": 0.4696747958660126, "learning_rate": 1.999780683474845e-06, "loss": 0.1269, "num_tokens": 590520.0, "reward": 0.25, "reward_std": 0.0, "rewards/compute_reward/mean": 0.25, "rewards/compute_reward/std": 0.44721361994743347, "step": 54 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.022, "frac_reward_zero_std": 0.75, "grad_norm": 42.38340377807617, "kl": 0.5665004104375839, "learning_rate": 1.999610115040354e-06, "loss": 0.1272, "num_tokens": 599400.0, "reward": 0.6875, "reward_std": 0.125, "rewards/compute_reward/mean": 0.6875, "rewards/compute_reward/std": 0.4787135720252991, "step": 55 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0224, "frac_reward_zero_std": 0.5, "grad_norm": 40.7006721496582, "kl": 0.4810464084148407, "learning_rate": 1.9993908270190957e-06, "loss": -0.248, "num_tokens": 608280.0, "reward": 0.1875, "reward_std": 0.26933756470680237, "rewards/compute_reward/mean": 0.1875, "rewards/compute_reward/std": 0.40311288833618164, "step": 56 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0228, "frac_reward_zero_std": 0.25, "grad_norm": 31.320539474487305, "kl": 0.44268205761909485, "learning_rate": 1.999122830098858e-06, "loss": 0.7516, "num_tokens": 617160.0, "reward": 0.3125, "reward_std": 0.375, "rewards/compute_reward/mean": 0.3125, "rewards/compute_reward/std": 0.4787135720252991, "step": 57 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0232, "frac_reward_zero_std": 0.5, "grad_norm": 46.34135818481445, "kl": 0.4858321249485016, "learning_rate": 1.998806137341434e-06, "loss": -0.8729, "num_tokens": 626040.0, "reward": 0.1875, "reward_std": 0.26933756470680237, "rewards/compute_reward/mean": 0.1875, "rewards/compute_reward/std": 0.40311288833618164, "step": 58 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0236, "frac_reward_zero_std": 0.5, "grad_norm": 50.77784729003906, "kl": 0.4124316945672035, "learning_rate": 1.998440764181981e-06, "loss": 0.4681, "num_tokens": 634920.0, "reward": 0.1875, "reward_std": 0.26933756470680237, "rewards/compute_reward/mean": 0.1875, "rewards/compute_reward/std": 0.40311288833618164, "step": 59 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.024, "frac_reward_zero_std": 0.25, "grad_norm": 62.64232635498047, "kl": 0.4631865471601486, "learning_rate": 1.9980267284282714e-06, "loss": -0.0896, "num_tokens": 643800.0, "reward": 0.1875, "reward_std": 0.375, "rewards/compute_reward/mean": 0.1875, "rewards/compute_reward/std": 0.40311288833618164, "step": 60 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0244, "frac_reward_zero_std": 0.5, "grad_norm": 55.10979080200195, "kl": 0.3911132216453552, "learning_rate": 1.997564050259824e-06, "loss": 0.1265, "num_tokens": 652680.0, "reward": 0.375, "reward_std": 0.25, "rewards/compute_reward/mean": 0.375, "rewards/compute_reward/std": 0.5, "step": 61 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0248, "frac_reward_zero_std": 0.25, "grad_norm": 75.01067352294922, "kl": 0.4750595986843109, "learning_rate": 1.99705275222692e-06, "loss": -0.8729, "num_tokens": 661560.0, "reward": 0.5625, "reward_std": 0.375, "rewards/compute_reward/mean": 0.5625, "rewards/compute_reward/std": 0.5123475790023804, "step": 62 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0252, "frac_reward_zero_std": 0.0, "grad_norm": 69.7686996459961, "kl": 0.4622643366456032, "learning_rate": 1.9964928592495045e-06, "loss": -0.3061, "num_tokens": 670440.0, "reward": 0.4375, "reward_std": 0.51933753490448, "rewards/compute_reward/mean": 0.4375, "rewards/compute_reward/std": 0.5123475790023804, "step": 63 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0256, "frac_reward_zero_std": 0.0, "grad_norm": 78.1102294921875, "kl": 0.4316819906234741, "learning_rate": 1.99588439861597e-06, "loss": 0.8431, "num_tokens": 679320.0, "reward": 0.5625, "reward_std": 0.51933753490448, "rewards/compute_reward/mean": 0.5625, "rewards/compute_reward/std": 0.5123475790023804, "step": 64 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.026, "frac_reward_zero_std": 0.25, "grad_norm": 70.63719940185547, "kl": 0.4158880114555359, "learning_rate": 1.995227399981831e-06, "loss": 0.0597, "num_tokens": 688200.0, "reward": 0.5625, "reward_std": 0.41367512941360474, "rewards/compute_reward/mean": 0.5625, "rewards/compute_reward/std": 0.5123475790023804, "step": 65 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0264, "frac_reward_zero_std": 0.25, "grad_norm": 56.348411560058594, "kl": 0.30301499366760254, "learning_rate": 1.994521895368273e-06, "loss": -0.2153, "num_tokens": 697080.0, "reward": 0.5625, "reward_std": 0.41367512941360474, "rewards/compute_reward/mean": 0.5625, "rewards/compute_reward/std": 0.5123475790023804, "step": 66 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0268, "frac_reward_zero_std": 0.5, "grad_norm": 25.096176147460938, "kl": 0.27180512621998787, "learning_rate": 1.9937679191605962e-06, "loss": 0.7175, "num_tokens": 705960.0, "reward": 0.1875, "reward_std": 0.26933756470680237, "rewards/compute_reward/mean": 0.1875, "rewards/compute_reward/std": 0.40311288833618164, "step": 67 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0272, "frac_reward_zero_std": 0.25, "grad_norm": 43.11947250366211, "kl": 0.27942386642098427, "learning_rate": 1.992965508106537e-06, "loss": -0.3403, "num_tokens": 714840.0, "reward": 0.6875, "reward_std": 0.41367512941360474, "rewards/compute_reward/mean": 0.6875, "rewards/compute_reward/std": 0.4787135720252991, "step": 68 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0276, "frac_reward_zero_std": 0.5, "grad_norm": 38.26469802856445, "kl": 0.24121180921792984, "learning_rate": 1.9921147013144777e-06, "loss": -0.124, "num_tokens": 723720.0, "reward": 0.1875, "reward_std": 0.26933756470680237, "rewards/compute_reward/mean": 0.1875, "rewards/compute_reward/std": 0.40311288833618164, "step": 69 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.028, "frac_reward_zero_std": 1.0, "grad_norm": 16.814804077148438, "kl": 0.2414722703397274, "learning_rate": 1.9912155402515414e-06, "loss": 0.001, "num_tokens": 732600.0, "reward": 0.0, "reward_std": 0.0, "rewards/compute_reward/mean": 0.0, "rewards/compute_reward/std": 0.0, "step": 70 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0284, "frac_reward_zero_std": 0.25, "grad_norm": 58.598567962646484, "kl": 0.4162477105855942, "learning_rate": 1.99026806874157e-06, "loss": -0.2483, "num_tokens": 741480.0, "reward": 0.4375, "reward_std": 0.41367512941360474, "rewards/compute_reward/mean": 0.4375, "rewards/compute_reward/std": 0.5123475790023804, "step": 71 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0288, "frac_reward_zero_std": 0.5, "grad_norm": 56.5173454284668, "kl": 0.30467043071985245, "learning_rate": 1.9892723329629885e-06, "loss": 0.2512, "num_tokens": 750360.0, "reward": 0.25, "reward_std": 0.28867512941360474, "rewards/compute_reward/mean": 0.25, "rewards/compute_reward/std": 0.44721361994743347, "step": 72 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0292, "frac_reward_zero_std": 0.5, "grad_norm": 88.95624542236328, "kl": 0.35381778329610825, "learning_rate": 1.9882283814465526e-06, "loss": 0.5593, "num_tokens": 759240.0, "reward": 0.5625, "reward_std": 0.26933756470680237, "rewards/compute_reward/mean": 0.5625, "rewards/compute_reward/std": 0.5123475790023804, "step": 73 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0296, "frac_reward_zero_std": 0.5, "grad_norm": 90.53428649902344, "kl": 0.3509898856282234, "learning_rate": 1.987136265072988e-06, "loss": -0.6815, "num_tokens": 768120.0, "reward": 0.625, "reward_std": 0.25, "rewards/compute_reward/mean": 0.625, "rewards/compute_reward/std": 0.5, "step": 74 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.03, "frac_reward_zero_std": 0.5, "grad_norm": 40.467559814453125, "kl": 0.27240653708577156, "learning_rate": 1.985996037070505e-06, "loss": 0.501, "num_tokens": 777000.0, "reward": 0.375, "reward_std": 0.25, "rewards/compute_reward/mean": 0.375, "rewards/compute_reward/std": 0.5, "step": 75 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0304, "frac_reward_zero_std": 0.75, "grad_norm": 42.016510009765625, "kl": 0.3176127299666405, "learning_rate": 1.984807753012208e-06, "loss": -0.3737, "num_tokens": 785880.0, "reward": 0.3125, "reward_std": 0.125, "rewards/compute_reward/mean": 0.3125, "rewards/compute_reward/std": 0.4787135720252991, "step": 76 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0308, "frac_reward_zero_std": 0.5, "grad_norm": 42.70841598510742, "kl": 0.40913257747888565, "learning_rate": 1.983571470813386e-06, "loss": 0.2181, "num_tokens": 794760.0, "reward": 0.4375, "reward_std": 0.26933756470680237, "rewards/compute_reward/mean": 0.4375, "rewards/compute_reward/std": 0.5123475790023804, "step": 77 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0312, "frac_reward_zero_std": 0.5, "grad_norm": 48.8704833984375, "kl": 0.30612240731716156, "learning_rate": 1.9822872507286887e-06, "loss": -0.4652, "num_tokens": 803640.0, "reward": 0.4375, "reward_std": 0.26933756470680237, "rewards/compute_reward/mean": 0.4375, "rewards/compute_reward/std": 0.5123475790023804, "step": 78 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0316, "frac_reward_zero_std": 1.0, "grad_norm": 10.744608879089355, "kl": 0.3573462963104248, "learning_rate": 1.9809551553491913e-06, "loss": 0.2514, "num_tokens": 812520.0, "reward": 0.25, "reward_std": 0.0, "rewards/compute_reward/mean": 0.25, "rewards/compute_reward/std": 0.44721361994743347, "step": 79 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.032, "frac_reward_zero_std": 0.25, "grad_norm": 44.66023254394531, "kl": 0.2872606888413429, "learning_rate": 1.979575249599344e-06, "loss": 0.1596, "num_tokens": 821400.0, "reward": 0.625, "reward_std": 0.39433756470680237, "rewards/compute_reward/mean": 0.625, "rewards/compute_reward/std": 0.5, "step": 80 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0324, "frac_reward_zero_std": 0.5, "grad_norm": 48.97246170043945, "kl": 0.41179851442575455, "learning_rate": 1.9781476007338054e-06, "loss": -0.2818, "num_tokens": 830280.0, "reward": 0.375, "reward_std": 0.25, "rewards/compute_reward/mean": 0.375, "rewards/compute_reward/std": 0.5, "step": 81 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0328, "frac_reward_zero_std": 0.75, "grad_norm": 47.648284912109375, "kl": 0.361647367477417, "learning_rate": 1.9766722783341677e-06, "loss": -0.09, "num_tokens": 839160.0, "reward": 0.375, "reward_std": 0.14433756470680237, "rewards/compute_reward/mean": 0.375, "rewards/compute_reward/std": 0.5, "step": 82 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0332, "frac_reward_zero_std": 0.5, "grad_norm": 68.1531982421875, "kl": 0.35677963495254517, "learning_rate": 1.975149354305563e-06, "loss": 0.0929, "num_tokens": 848040.0, "reward": 0.4375, "reward_std": 0.26933756470680237, "rewards/compute_reward/mean": 0.4375, "rewards/compute_reward/std": 0.5123475790023804, "step": 83 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0336, "frac_reward_zero_std": 1.0, "grad_norm": 25.728612899780273, "kl": 0.299656730145216, "learning_rate": 1.97357890287316e-06, "loss": 0.1262, "num_tokens": 856920.0, "reward": 0.25, "reward_std": 0.0, "rewards/compute_reward/mean": 0.25, "rewards/compute_reward/std": 0.44721361994743347, "step": 84 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.034, "frac_reward_zero_std": 0.25, "grad_norm": 43.74058151245117, "kl": 0.3780274763703346, "learning_rate": 1.9719610005785463e-06, "loss": -0.215, "num_tokens": 865800.0, "reward": 0.375, "reward_std": 0.39433756470680237, "rewards/compute_reward/mean": 0.375, "rewards/compute_reward/std": 0.5, "step": 85 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0344, "frac_reward_zero_std": 1.0, "grad_norm": 22.789318084716797, "kl": 0.361059308052063, "learning_rate": 1.9702957262759963e-06, "loss": 0.2179, "num_tokens": 874680.0, "reward": 0.0, "reward_std": 0.0, "rewards/compute_reward/mean": 0.0, "rewards/compute_reward/std": 0.0, "step": 86 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0348, "frac_reward_zero_std": 0.75, "grad_norm": 35.32533264160156, "kl": 0.44058068096637726, "learning_rate": 1.968583161128631e-06, "loss": 0.0018, "num_tokens": 883560.0, "reward": 0.3125, "reward_std": 0.125, "rewards/compute_reward/mean": 0.3125, "rewards/compute_reward/std": 0.4787135720252991, "step": 87 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0352, "frac_reward_zero_std": 0.5, "grad_norm": 48.396873474121094, "kl": 0.40034161880612373, "learning_rate": 1.9668233886044593e-06, "loss": 0.3765, "num_tokens": 892440.0, "reward": 0.4375, "reward_std": 0.26933756470680237, "rewards/compute_reward/mean": 0.4375, "rewards/compute_reward/std": 0.5123475790023804, "step": 88 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0356, "frac_reward_zero_std": 0.5, "grad_norm": 51.92245101928711, "kl": 0.582250103354454, "learning_rate": 1.9650164944723112e-06, "loss": -0.2811, "num_tokens": 901320.0, "reward": 0.4375, "reward_std": 0.26933756470680237, "rewards/compute_reward/mean": 0.4375, "rewards/compute_reward/std": 0.5123475790023804, "step": 89 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.036, "frac_reward_zero_std": 0.75, "grad_norm": 39.04520797729492, "kl": 0.46102364361286163, "learning_rate": 1.963162566797658e-06, "loss": 0.1603, "num_tokens": 910200.0, "reward": 0.6875, "reward_std": 0.125, "rewards/compute_reward/mean": 0.6875, "rewards/compute_reward/std": 0.4787135720252991, "step": 90 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0364, "frac_reward_zero_std": 0.25, "grad_norm": 64.48860168457031, "kl": 0.7562894448637962, "learning_rate": 1.9612616959383188e-06, "loss": 0.2195, "num_tokens": 919080.0, "reward": 0.625, "reward_std": 0.39433756470680237, "rewards/compute_reward/mean": 0.625, "rewards/compute_reward/std": 0.5, "step": 91 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0368, "frac_reward_zero_std": 0.5, "grad_norm": 71.13645935058594, "kl": 0.584452673792839, "learning_rate": 1.9593139745400573e-06, "loss": -0.2141, "num_tokens": 927960.0, "reward": 0.625, "reward_std": 0.25, "rewards/compute_reward/mean": 0.625, "rewards/compute_reward/std": 0.5, "step": 92 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0372, "frac_reward_zero_std": 0.25, "grad_norm": 44.049346923828125, "kl": 0.5378477424383163, "learning_rate": 1.957319497532067e-06, "loss": -0.2143, "num_tokens": 936840.0, "reward": 0.5, "reward_std": 0.39433756470680237, "rewards/compute_reward/mean": 0.5, "rewards/compute_reward/std": 0.5163977742195129, "step": 93 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0376, "frac_reward_zero_std": 0.25, "grad_norm": 75.5696792602539, "kl": 0.5544816628098488, "learning_rate": 1.9552783621223435e-06, "loss": 0.5601, "num_tokens": 945720.0, "reward": 0.75, "reward_std": 0.39433756470680237, "rewards/compute_reward/mean": 0.75, "rewards/compute_reward/std": 0.44721361994743347, "step": 94 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.038, "frac_reward_zero_std": 1.0, "grad_norm": 17.455116271972656, "kl": 0.3371001221239567, "learning_rate": 1.953190667792947e-06, "loss": -0.59, "num_tokens": 954600.0, "reward": 0.5, "reward_std": 0.0, "rewards/compute_reward/mean": 0.5, "rewards/compute_reward/std": 0.5163977742195129, "step": 95 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0384, "frac_reward_zero_std": 0.25, "grad_norm": 59.06554412841797, "kl": 0.3672589622437954, "learning_rate": 1.9510565162951534e-06, "loss": 0.2179, "num_tokens": 963480.0, "reward": 0.625, "reward_std": 0.39433756470680237, "rewards/compute_reward/mean": 0.625, "rewards/compute_reward/std": 0.5, "step": 96 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0388, "frac_reward_zero_std": 0.5, "grad_norm": 22.509157180786133, "kl": 0.18590721115469933, "learning_rate": 1.9488760116444964e-06, "loss": -0.3407, "num_tokens": 972360.0, "reward": 0.6875, "reward_std": 0.26933756470680237, "rewards/compute_reward/mean": 0.6875, "rewards/compute_reward/std": 0.4787135720252991, "step": 97 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0392, "frac_reward_zero_std": 0.5, "grad_norm": 36.51370620727539, "kl": 0.25807878002524376, "learning_rate": 1.9466492601156963e-06, "loss": 0.126, "num_tokens": 981240.0, "reward": 0.25, "reward_std": 0.25, "rewards/compute_reward/mean": 0.25, "rewards/compute_reward/std": 0.44721361994743347, "step": 98 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.0396, "frac_reward_zero_std": 0.75, "grad_norm": 13.788187026977539, "kl": 0.23087293654680252, "learning_rate": 1.944376370237481e-06, "loss": 0.2509, "num_tokens": 990120.0, "reward": 0.6875, "reward_std": 0.125, "rewards/compute_reward/mean": 0.6875, "rewards/compute_reward/std": 0.4787135720252991, "step": 99 }, { "completion_length": 512.0, "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.04, "frac_reward_zero_std": 0.5, "grad_norm": 8.3355073928833, "kl": 0.1271139644086361, "learning_rate": 1.9420574527872966e-06, "loss": -0.1245, "num_tokens": 999000.0, "reward": 0.375, "reward_std": 0.25, "rewards/compute_reward/mean": 0.375, "rewards/compute_reward/std": 0.5, "step": 100 }, { "completions/clipped_ratio": 1.0, "completions/max_length": 512.0, "completions/max_terminated_length": 0.0, "completions/mean_length": 512.0, "completions/mean_terminated_length": 0.0, "completions/min_length": 512.0, "completions/min_terminated_length": 0.0, "epoch": 0.04, "eval_loss": -0.00047406076919287443, "eval_runtime": 10044.1197, "eval_samples_per_second": 0.02, "eval_steps_per_second": 0.005, "frac_reward_zero_std": 0.66, "num_tokens": 1110000.0, "reward": 0.415, "reward_std": 0.17928203105926513, "rewards/compute_reward/mean": 0.415, "rewards/compute_reward/std": 0.17928203105926513, "step": 100 } ], "logging_steps": 1, "max_steps": 500, "num_input_tokens_seen": 1110000, "num_train_epochs": 1, "save_steps": 50, "stateful_callbacks": { "TrainerControl": { "args": { "should_epoch_stop": false, "should_evaluate": false, "should_log": false, "should_save": true, "should_training_stop": false }, "attributes": {} } }, "total_flos": 0.0, "train_batch_size": 4, "trial_name": null, "trial_params": null }