{ "best_metric": null, "best_model_checkpoint": null, "epoch": 0.13333333333333333, "eval_steps": 10, "global_step": 100, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "accuracy_reward": 0.21250000596046448, "accuracy_reward/correct": 1.0, "accuracy_reward/correct/max": 1.0, "accuracy_reward/correct/median": 1.0, "accuracy_reward/correct/min": 1.0, "accuracy_reward/correct/p25": 1.0, "accuracy_reward/correct/p75": 1.0, "accuracy_reward/correct/var": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 1.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.16946202516555786, "adam_stats/lm_head/lr_effective_max": 4.743288627651054e-06, "adam_stats/lm_head/lr_effective_mean": -1.830192941643105e-11, "adam_stats/lm_head/lr_effective_min": -4.743269073514966e-06, "adam_stats/lm_head/lr_effective_std": 3.580759653232235e-07, "adam_stats/lr_effective_max": 4.743401859741425e-06, "adam_stats/lr_effective_mean": 1.0161699953714631e-10, "adam_stats/lr_effective_min": -4.743400950246723e-06, "adam_stats/m_t_max": 0.0032175034284591675, "adam_stats/m_t_mean": -5.127846064434394e-11, "adam_stats/m_t_min": -0.0031728155445307493, "adam_stats/v_t_max": 1.0352190429330221e-06, "adam_stats/v_t_mean": 2.1092132760410637e-13, "adam_stats/v_t_min": 0.0, "advantages": 9.685755308908028e-09, "advantages/max": 1.6198352575302124, "advantages/median": 0.0, "advantages/min": -1.6198352575302124, "advantages/p25": -0.5399450659751892, "advantages/p75": 0.0, "advantages/var": 0.4428557753562927, "all_logprobs": -0.5299894213676453, "all_logprobs/max": 0.0, "all_logprobs/median": -0.043212890625, "all_logprobs/min": -10.8125, "all_logprobs/p1": -4.6875, "all_logprobs/p10": -1.734375, "all_logprobs/p25": -0.62109375, "all_logprobs/p5": -2.625, "all_logprobs/p75": -0.00070953369140625, "all_logprobs/var": 0.9874793887138367, "clip_ratio": 0.0, "completion_length": 519.0875244140625, "completion_length/correct": 310.5882263183594, "completion_length/correct/max": 500.0, "completion_length/correct/median": 284.0, "completion_length/correct/min": 169.0, "completion_length/correct/p25": 241.0, "completion_length/correct/p75": 351.0, "completion_length/correct/var": 9566.6328125, "completion_length/incorrect": 575.3492431640625, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 494.0, "completion_length/incorrect/min": 161.0, "completion_length/incorrect/p25": 299.0, "completion_length/incorrect/p75": 801.5, "completion_length/incorrect/var": 88547.84375, "completion_length/max": 1024.0, "completion_length/median": 448.0, "completion_length/min": 161.0, "completion_length/p25": 267.25, "completion_length/p75": 760.0, "completion_length/var": 83309.796875, "epoch": 0.0013333333333333333, "feature_vector_variance/max_squared_error": 26903.267578125, "feature_vector_variance/metric": 15306.8408203125, "generated_tokens/total": 41527.0, "global_fisher_curvature": 2208.0, "global_fisher_curvature/max": 2208.0, "global_fisher_curvature/median": 2208.0, "global_fisher_curvature/min": 2208.0, "global_fisher_curvature/p25": 2208.0, "global_fisher_curvature/p75": 2208.0, "global_fisher_curvature/p85": 2208.0, "global_fisher_curvature/p90": 2208.0, "global_fisher_curvature/p95": 2208.0, "global_fisher_curvature/p99": 2208.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 1552.0, "global_hessian_coeff/max": 1552.0, "global_hessian_coeff/median": 1552.0, "global_hessian_coeff/min": 1552.0, "global_hessian_coeff/p25": 1552.0, "global_hessian_coeff/p75": 1552.0, "global_hessian_coeff/p99": 1552.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 1552.0, "global_hessian_coeff_abs/max": 1552.0, "global_hessian_coeff_abs/median": 1552.0, "global_hessian_coeff_abs/min": 1552.0, "global_hessian_coeff_abs/p25": 1552.0, "global_hessian_coeff_abs/p75": 1552.0, "global_hessian_coeff_abs/p99": 1552.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 1.0926555395126343, "grouped_std_rewards": 0.24408018589019775, "learning_rate": 1.5e-06, "loss": 0.0, "mean_logprobs": -0.486328125, "mean_logprobs/var": 0.044189453125, "num_completions/total": 80, "per_sentence_gradient_norm": 55.234375, "per_sentence_gradient_norm/max": 332.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 97.0, "per_sentence_gradient_norm/var": 5091.6865234375, "per_token_feature_norm": 140.52308654785156, "per_token_feature_norm/max": 151.0, "per_token_feature_norm/median": 141.0, "per_token_feature_norm/min": 97.0, "per_token_feature_norm/p25": 138.0, "per_token_feature_norm/p75": 144.0, "per_token_feature_norm/var": 29.137338638305664, "per_token_gradient_norm": 11.204014778137207, "per_token_gradient_norm/max": 334.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0810546875, "per_token_gradient_norm/var": 1000.6014404296875, "per_token_policy_error_norm": 0.23531226813793182, "per_token_policy_error_norm/max": 2.0, "per_token_policy_error_norm/median": 0.0, "per_token_policy_error_norm/min": 0.0, "per_token_policy_error_norm/p25": 0.0, "per_token_policy_error_norm/p75": 0.326171875, "per_token_policy_error_norm/var": 0.15869013965129852, "policy_entropy": 0.5879753232002258, "policy_entropy/max": 3.609375, "policy_entropy/median": 0.2060546875, "policy_entropy/min": 7.683411240577698e-08, "policy_entropy/p25": 0.00653076171875, "policy_entropy/p75": 0.94921875, "policy_entropy/var": 0.5754079818725586, "policy_loss": -4.470348535789981e-09, "policy_loss/max": 1.619835376739502, "policy_loss/median": 0.0, "policy_loss/min": -1.6198352575302124, "policy_loss/p25": 0.0, "policy_loss/p75": 0.5399450659751892, "policy_loss/var": 0.4428557753562927, "policy_sharpness": 3.4709718227386475, "policy_sharpness/max": 10.0, "policy_sharpness/median": 1.679185152053833, "policy_sharpness/min": -0.0, "policy_sharpness/p25": 0.4441272020339966, "policy_sharpness/p75": 6.75, "policy_sharpness/var": 13.782241821289062, "reward": 0.21250000596046448, "reward/max": 1.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.16946202516555786, "rewards/accuracy_reward": 0.21250000596046448, "rewards/accuracy_reward/max": 1.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.16946202516555786, "sentence_fisher_curvature": 407788.8125, "sentence_fisher_curvature/max": 4620288.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 449536.0, "sentence_fisher_curvature/p85": 886374.5, "sentence_fisher_curvature/p90": 1044479.5625, "sentence_fisher_curvature/p95": 1745304.0, "sentence_fisher_curvature/p99": 4464967.5, "sentence_fisher_curvature/var": 702831198208.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 7972.8759765625, "sentence_full_gradient_variance/metric": 7972.8759765625, "sentence_full_gradient_variance/p75": 7972.8759765625, "sentence_full_gradient_variance/p90": 7972.8759765625, "sentence_full_gradient_variance/p95": 7972.8759765625, "sentence_full_gradient_variance/p99": 7972.8759765625, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 151176.0, "sentence_hessian_coeff/max": 6651904.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": -2555904.0, "sentence_hessian_coeff/p25": -28384.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 5098699.0, "sentence_hessian_coeff/var": 1127910473728.0, "sentence_hessian_coeff_abs": 356203.21875, "sentence_hessian_coeff_abs/max": 6651904.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 216576.0, "sentence_hessian_coeff_abs/p99": 5098699.0, "sentence_hessian_coeff_abs/var": 1022567186432.0, "step": 1, "token_fisher_curvature": 20147862.0, "token_fisher_curvature/max": 4764729344.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 7.724761962890625e-05, "token_fisher_curvature/p85": 72704.0, "token_fisher_curvature/p90": 9437184.0, "token_fisher_curvature/p95": 101187584.0, "token_fisher_curvature/p99": 436207616.0, "token_fisher_curvature/var": 1.492465165008896e+16, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 4024080.75, "token_hessian_coeff/max": 7583301632.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": -5737807872.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 142336000.0, "token_hessian_coeff/var": 2.3223549594435584e+16, "token_hessian_coeff_abs": 16085694.0, "token_hessian_coeff_abs/max": 7583301632.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0611572265625, "token_hessian_coeff_abs/p99": 316669952.0, "token_hessian_coeff_abs/var": 2.298098487394304e+16 }, { "accuracy_reward": 0.10000000149011612, "accuracy_reward/correct": 1.0, "accuracy_reward/correct/max": 1.0, "accuracy_reward/correct/median": 1.0, "accuracy_reward/correct/min": 1.0, "accuracy_reward/correct/p25": 1.0, "accuracy_reward/correct/p75": 1.0, "accuracy_reward/correct/var": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 1.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.09113924205303192, "adam_stats/lm_head/lr_effective_max": 1.2765639439749066e-05, "adam_stats/lm_head/lr_effective_mean": -4.8650087430823774e-11, "adam_stats/lm_head/lr_effective_min": -1.276469811273273e-05, "adam_stats/lm_head/lr_effective_std": 9.052323548530694e-07, "adam_stats/lr_effective_max": 1.2765858627972193e-05, "adam_stats/lr_effective_mean": 2.0238552500551066e-10, "adam_stats/lr_effective_min": -1.2765946848958265e-05, "adam_stats/m_t_max": 0.0042352816089987755, "adam_stats/m_t_mean": -1.3516828281656679e-11, "adam_stats/m_t_min": -0.004976320080459118, "adam_stats/v_t_max": 1.4554231029251241e-06, "adam_stats/v_t_mean": 4.3605817199275643e-13, "adam_stats/v_t_min": 0.0, "advantages": 1.4901161415892261e-09, "advantages/max": 2.4741740226745605, "advantages/median": 0.0, "advantages/min": -0.9352393746376038, "advantages/p25": -0.3534534275531769, "advantages/p75": 0.0, "advantages/var": 0.4428044557571411, "all_logprobs": -0.6014763712882996, "all_logprobs/max": 0.0, "all_logprobs/median": -0.07373046875, "all_logprobs/min": -12.5625, "all_logprobs/p1": -4.84375, "all_logprobs/p10": -1.9375, "all_logprobs/p25": -0.76953125, "all_logprobs/p5": -2.875, "all_logprobs/p75": -0.00077056884765625, "all_logprobs/var": 1.1045200824737549, "clip_ratio": 0.0, "completion_length": 538.0499877929688, "completion_length/correct": 485.625, "completion_length/correct/max": 737.0, "completion_length/correct/median": 440.0, "completion_length/correct/min": 273.0, "completion_length/correct/p25": 409.5, "completion_length/correct/p75": 564.5, "completion_length/correct/var": 19693.41015625, "completion_length/incorrect": 543.875, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 479.0, "completion_length/incorrect/min": 188.0, "completion_length/incorrect/p25": 320.0, "completion_length/incorrect/p75": 704.75, "completion_length/incorrect/var": 64050.75390625, "completion_length/max": 1024.0, "completion_length/median": 479.0, "completion_length/min": 188.0, "completion_length/p25": 321.75, "completion_length/p75": 677.75, "completion_length/var": 59618.828125, "epoch": 0.0026666666666666666, "feature_vector_variance/max_squared_error": 26865.744140625, "feature_vector_variance/metric": 15321.1279296875, "generated_tokens/total": 84571.0, "global_fisher_curvature": 1200.0, "global_fisher_curvature/max": 1200.0, "global_fisher_curvature/median": 1200.0, "global_fisher_curvature/min": 1200.0, "global_fisher_curvature/p25": 1200.0, "global_fisher_curvature/p75": 1200.0, "global_fisher_curvature/p85": 1200.0, "global_fisher_curvature/p90": 1200.0, "global_fisher_curvature/p95": 1200.0, "global_fisher_curvature/p99": 1200.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 1.3533281162381172e-09, "global_fisher_kl_divergence/max": 1.3533281162381172e-09, "global_fisher_kl_divergence/median": 1.3533281162381172e-09, "global_fisher_kl_divergence/min": 1.3533281162381172e-09, "global_fisher_kl_divergence/p25": 1.3533281162381172e-09, "global_fisher_kl_divergence/p75": 1.3533281162381172e-09, "global_fisher_kl_divergence/p85": 1.3533281162381172e-09, "global_fisher_kl_divergence/p90": 1.3533281162381172e-09, "global_fisher_kl_divergence/p95": 1.3533281162381172e-09, "global_fisher_kl_divergence/p99": 1.3533281162381172e-09, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.004302978515625, "global_full_update_term/max": 0.004302978515625, "global_full_update_term/median": 0.004302978515625, "global_full_update_term/min": 0.004302978515625, "global_full_update_term/p25": 0.004302978515625, "global_full_update_term/p75": 0.004302978515625, "global_full_update_term/p85": 0.004302978515625, "global_full_update_term/p90": 0.004302978515625, "global_full_update_term/p95": 0.004302978515625, "global_full_update_term/p99": 0.004302978515625, "global_full_update_term/var": NaN, "global_hessian_coeff": 820.0, "global_hessian_coeff/max": 820.0, "global_hessian_coeff/median": 820.0, "global_hessian_coeff/min": 820.0, "global_hessian_coeff/p25": 820.0, "global_hessian_coeff/p75": 820.0, "global_hessian_coeff/p99": 820.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 820.0, "global_hessian_coeff_abs/max": 820.0, "global_hessian_coeff_abs/median": 820.0, "global_hessian_coeff_abs/min": 820.0, "global_hessian_coeff_abs/p25": 820.0, "global_hessian_coeff_abs/p75": 820.0, "global_hessian_coeff_abs/p99": 820.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 1.0648407936096191, "grouped_std_rewards": 0.1948736011981964, "learning_rate": 3e-06, "loss": -0.0, "mean_logprobs": -0.55078125, "mean_logprobs/var": 0.08544921875, "num_completions/total": 160, "per_sentence_gradient_norm": 52.37031173706055, "per_sentence_gradient_norm/max": 560.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 61.5625, "per_sentence_gradient_norm/var": 11077.2333984375, "per_token_feature_norm": 140.73651123046875, "per_token_feature_norm/max": 151.0, "per_token_feature_norm/median": 142.0, "per_token_feature_norm/min": 99.5, "per_token_feature_norm/p25": 139.0, "per_token_feature_norm/p75": 144.0, "per_token_feature_norm/var": 28.16759490966797, "per_token_gradient_norm": 14.501708030700684, "per_token_gradient_norm/max": 494.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 6.84375, "per_token_gradient_norm/var": 1714.1929931640625, "per_token_policy_error_norm": 0.2623652517795563, "per_token_policy_error_norm/max": 2.0, "per_token_policy_error_norm/median": 0.0078125, "per_token_policy_error_norm/min": 0.0, "per_token_policy_error_norm/p25": 0.0, "per_token_policy_error_norm/p75": 0.421875, "per_token_policy_error_norm/var": 0.1692146211862564, "policy_entropy": 0.6707882881164551, "policy_entropy/max": 3.765625, "policy_entropy/median": 0.3046875, "policy_entropy/min": 1.4062970876693726e-07, "policy_entropy/p25": 0.006988525390625, "policy_entropy/p75": 1.1171875, "policy_entropy/var": 0.6675649285316467, "policy_loss": -8.940697071579962e-09, "policy_loss/max": 0.9352393746376038, "policy_loss/median": 0.0, "policy_loss/min": -2.4741742610931396, "policy_loss/p25": 0.0, "policy_loss/p75": 0.3534534275531769, "policy_loss/var": 0.4428044557571411, "policy_sharpness": 3.342435121536255, "policy_sharpness/max": 10.0, "policy_sharpness/median": 1.4138765335083008, "policy_sharpness/min": -0.0, "policy_sharpness/p25": 0.321397989988327, "policy_sharpness/p75": 6.5, "policy_sharpness/var": 14.17398452758789, "reward": 0.10000000149011612, "reward/max": 1.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.09113924205303192, "rewards/accuracy_reward": 0.10000000149011612, "rewards/accuracy_reward/max": 1.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.09113924205303192, "sentence_fisher_curvature": 462619.21875, "sentence_fisher_curvature/max": 14090240.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 162816.0, "sentence_fisher_curvature/p85": 244787.25, "sentence_fisher_curvature/p90": 349798.34375, "sentence_fisher_curvature/p95": 1038943.375, "sentence_fisher_curvature/p99": 8291608.0, "sentence_fisher_curvature/var": 3545436782592.0, "sentence_fisher_kl_divergence": 5.204754529586353e-07, "sentence_fisher_kl_divergence/max": 1.5854835510253906e-05, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 1.8277205526828766e-07, "sentence_fisher_kl_divergence/p85": 2.74833325875079e-07, "sentence_fisher_kl_divergence/p90": 3.933905929898174e-07, "sentence_fisher_kl_divergence/p95": 1.16638057079399e-06, "sentence_fisher_kl_divergence/p99": 9.333185516879894e-06, "sentence_fisher_kl_divergence/var": 4.48893699989128e-12, "sentence_full_gradient_variance/max_squared_error": 13513.4169921875, "sentence_full_gradient_variance/metric": 13513.4169921875, "sentence_full_gradient_variance/p75": 13513.4169921875, "sentence_full_gradient_variance/p90": 13513.4169921875, "sentence_full_gradient_variance/p95": 13513.4169921875, "sentence_full_gradient_variance/p99": 13513.4169921875, "sentence_full_update_term": 0.019916439428925514, "sentence_full_update_term/max": 0.455078125, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.00556182861328125, "sentence_full_update_term/p85": 0.008746338076889515, "sentence_full_update_term/p90": 0.011273188516497612, "sentence_full_update_term/p95": 0.034215882420539856, "sentence_full_update_term/p99": 0.4041600823402405, "sentence_full_update_term/var": 0.00623113289475441, "sentence_hessian_coeff": 614733.1875, "sentence_hessian_coeff/max": 22282240.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": -190464.0, "sentence_hessian_coeff/p25": -24512.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 15189270.0, "sentence_hessian_coeff/var": 9849724207104.0, "sentence_hessian_coeff_abs": 662569.1875, "sentence_hessian_coeff_abs/max": 22282240.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 48896.0, "sentence_hessian_coeff_abs/p99": 15189270.0, "sentence_hessian_coeff_abs/var": 9787848785920.0, "step": 2, "token_fisher_curvature": 34819368.0, "token_fisher_curvature/max": 9797894144.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 9536.0, "token_fisher_curvature/p85": 6979584.0, "token_fisher_curvature/p90": 29884416.0, "token_fisher_curvature/p95": 72351744.0, "token_fisher_curvature/p99": 687865856.0, "token_fisher_curvature/var": 8.114187936569754e+16, "token_fisher_kl_divergence": 3.918255242751911e-05, "token_fisher_kl_divergence/max": 0.01104736328125, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 1.0710209608078003e-08, "token_fisher_kl_divergence/p85": 7.867813110351562e-06, "token_fisher_kl_divergence/p90": 3.361701965332031e-05, "token_fisher_kl_divergence/p95": 8.153915405273438e-05, "token_fisher_kl_divergence/p99": 0.000774383544921875, "token_fisher_kl_divergence/var": 1.0275400796899703e-07, "token_full_update_term": 0.002929953159764409, "token_full_update_term/max": 0.39453125, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 7.009506225585938e-05, "token_full_update_term/p85": 0.00167083740234375, "token_full_update_term/p90": 0.0032958984375, "token_full_update_term/p95": 0.00555419921875, "token_full_update_term/p99": 0.07666015625, "token_full_update_term/var": 0.00034996491740457714, "token_hessian_coeff": 38752756.0, "token_hessian_coeff/max": 24024973312.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": -1308622848.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 8.463859558105469e-06, "token_hessian_coeff/p99": 334938112.0, "token_hessian_coeff/var": 3.6823530895468134e+17, "token_hessian_coeff_abs": 50787924.0, "token_hessian_coeff_abs/max": 24024973312.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 41728.0, "token_hessian_coeff_abs/p99": 568623104.0, "token_hessian_coeff_abs/var": 3.67157615760769e+17 }, { "accuracy_reward": 0.0625, "accuracy_reward/correct": 1.0, "accuracy_reward/correct/max": 1.0, "accuracy_reward/correct/median": 1.0, "accuracy_reward/correct/min": 1.0, "accuracy_reward/correct/p25": 1.0, "accuracy_reward/correct/p75": 1.0, "accuracy_reward/correct/var": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 1.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.05933544784784317, "adam_stats/lm_head/lr_effective_max": 2.2338797862175852e-05, "adam_stats/lm_head/lr_effective_mean": -9.675316103852083e-11, "adam_stats/lm_head/lr_effective_min": -2.2339163479045965e-05, "adam_stats/lm_head/lr_effective_std": 1.3744402167503722e-06, "adam_stats/lr_effective_max": 2.2352178348228335e-05, "adam_stats/lr_effective_mean": 2.565324896952603e-10, "adam_stats/lr_effective_min": -2.2352933228830807e-05, "adam_stats/m_t_max": 0.007924361154437065, "adam_stats/m_t_mean": -6.526736096024166e-11, "adam_stats/m_t_min": -0.005473561119288206, "adam_stats/v_t_max": 3.5727923659578664e-06, "adam_stats/v_t_mean": 6.327581687781691e-13, "adam_stats/v_t_min": 0.0, "advantages": 6.705522803684971e-09, "advantages/max": 2.4741740226745605, "advantages/median": 0.0, "advantages/min": -0.7244288325309753, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.26568838953971863, "all_logprobs": -0.5914792418479919, "all_logprobs/max": 0.0, "all_logprobs/median": -0.052490234375, "all_logprobs/min": -11.5625, "all_logprobs/p1": -4.84375, "all_logprobs/p10": -1.9765625, "all_logprobs/p25": -0.73046875, "all_logprobs/p5": -2.921875, "all_logprobs/p75": -0.0004673004150390625, "all_logprobs/var": 1.1390597820281982, "clip_ratio": 0.0, "completion_length": 554.8624877929688, "completion_length/correct": 368.3999938964844, "completion_length/correct/max": 594.0, "completion_length/correct/median": 348.0, "completion_length/correct/min": 229.0, "completion_length/correct/p25": 275.0, "completion_length/correct/p75": 396.0, "completion_length/correct/var": 20057.30078125, "completion_length/incorrect": 567.2933349609375, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 487.0, "completion_length/incorrect/min": 141.0, "completion_length/incorrect/p25": 334.0, "completion_length/incorrect/p75": 818.5, "completion_length/incorrect/var": 83558.640625, "completion_length/max": 1024.0, "completion_length/median": 431.0, "completion_length/min": 141.0, "completion_length/p25": 330.75, "completion_length/p75": 787.0, "completion_length/var": 81632.90625, "epoch": 0.004, "feature_vector_variance/max_squared_error": 26536.228515625, "feature_vector_variance/metric": 15726.2099609375, "generated_tokens/total": 128960.0, "global_fisher_curvature": 1432.0, "global_fisher_curvature/max": 1432.0, "global_fisher_curvature/median": 1432.0, "global_fisher_curvature/min": 1432.0, "global_fisher_curvature/p25": 1432.0, "global_fisher_curvature/p75": 1432.0, "global_fisher_curvature/p85": 1432.0, "global_fisher_curvature/p90": 1432.0, "global_fisher_curvature/p95": 1432.0, "global_fisher_curvature/p99": 1432.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 6.4319465309381485e-09, "global_fisher_kl_divergence/max": 6.4319465309381485e-09, "global_fisher_kl_divergence/median": 6.4319465309381485e-09, "global_fisher_kl_divergence/min": 6.4319465309381485e-09, "global_fisher_kl_divergence/p25": 6.4319465309381485e-09, "global_fisher_kl_divergence/p75": 6.4319465309381485e-09, "global_fisher_kl_divergence/p85": 6.4319465309381485e-09, "global_fisher_kl_divergence/p90": 6.4319465309381485e-09, "global_fisher_kl_divergence/p95": 6.4319465309381485e-09, "global_fisher_kl_divergence/p99": 6.4319465309381485e-09, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0159912109375, "global_full_update_term/max": 0.0159912109375, "global_full_update_term/median": 0.0159912109375, "global_full_update_term/min": 0.0159912109375, "global_full_update_term/p25": 0.0159912109375, "global_full_update_term/p75": 0.0159912109375, "global_full_update_term/p85": 0.0159912109375, "global_full_update_term/p90": 0.0159912109375, "global_full_update_term/p95": 0.0159912109375, "global_full_update_term/p99": 0.0159912109375, "global_full_update_term/var": NaN, "global_hessian_coeff": -672.0, "global_hessian_coeff/max": -672.0, "global_hessian_coeff/median": -672.0, "global_hessian_coeff/min": -672.0, "global_hessian_coeff/p25": -672.0, "global_hessian_coeff/p75": -672.0, "global_hessian_coeff/p99": -672.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 672.0, "global_hessian_coeff_abs/max": 672.0, "global_hessian_coeff_abs/median": 672.0, "global_hessian_coeff_abs/min": 672.0, "global_hessian_coeff_abs/p25": 672.0, "global_hessian_coeff_abs/p75": 672.0, "global_hessian_coeff_abs/p99": 672.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.9356981515884399, "grouped_std_rewards": 0.1224656030535698, "learning_rate": 4.5e-06, "loss": -0.0, "mean_logprobs": -0.5546875, "mean_logprobs/var": 0.099609375, "num_completions/total": 240, "per_sentence_gradient_norm": 29.53125, "per_sentence_gradient_norm/max": 346.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 43.4375, "per_sentence_gradient_norm/var": 3996.315185546875, "per_token_feature_norm": 140.91566467285156, "per_token_feature_norm/max": 151.0, "per_token_feature_norm/median": 142.0, "per_token_feature_norm/min": 100.5, "per_token_feature_norm/p25": 139.0, "per_token_feature_norm/p75": 144.0, "per_token_feature_norm/var": 26.337982177734375, "per_token_gradient_norm": 6.724372863769531, "per_token_gradient_norm/max": 516.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.000667572021484375, "per_token_gradient_norm/var": 749.578857421875, "per_token_policy_error_norm": 0.2525905966758728, "per_token_policy_error_norm/max": 2.0, "per_token_policy_error_norm/median": 0.00390625, "per_token_policy_error_norm/min": 0.0, "per_token_policy_error_norm/p25": 0.0, "per_token_policy_error_norm/p75": 0.38671875, "per_token_policy_error_norm/var": 0.16720865666866302, "policy_entropy": 0.6596906185150146, "policy_entropy/max": 3.703125, "policy_entropy/median": 0.2421875, "policy_entropy/min": 1.7601996660232544e-07, "policy_entropy/p25": 0.0045166015625, "policy_entropy/p75": 1.0859375, "policy_entropy/var": 0.7049999833106995, "policy_loss": -5.215406329028838e-09, "policy_loss/max": 0.7244288921356201, "policy_loss/median": 0.0, "policy_loss/min": -2.4741740226745605, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.26568838953971863, "policy_sharpness": 3.5206596851348877, "policy_sharpness/max": 10.0, "policy_sharpness/median": 1.4730956554412842, "policy_sharpness/min": -0.0, "policy_sharpness/p25": 0.3140608072280884, "policy_sharpness/p75": 7.875, "policy_sharpness/var": 15.176739692687988, "reward": 0.0625, "reward/max": 1.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.05933544784784317, "rewards/accuracy_reward": 0.0625, "rewards/accuracy_reward/max": 1.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.05933544784784317, "sentence_fisher_curvature": 219628.796875, "sentence_fisher_curvature/max": 4620288.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 34304.0, "sentence_fisher_curvature/p85": 247910.546875, "sentence_fisher_curvature/p90": 629964.75, "sentence_fisher_curvature/p95": 1542963.0, "sentence_fisher_curvature/p99": 3183573.5, "sentence_fisher_curvature/var": 473213468672.0, "sentence_fisher_kl_divergence": 9.88558213066426e-07, "sentence_fisher_kl_divergence/max": 2.0742416381835938e-05, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 1.5459954738616943e-07, "sentence_fisher_kl_divergence/p85": 1.1153525747431559e-06, "sentence_fisher_kl_divergence/p90": 2.8312203994573792e-06, "sentence_fisher_kl_divergence/p95": 6.955860953894444e-06, "sentence_fisher_kl_divergence/p99": 1.4338485925691202e-05, "sentence_fisher_kl_divergence/var": 9.573825239528322e-12, "sentence_full_gradient_variance/max_squared_error": 4759.9560546875, "sentence_full_gradient_variance/metric": 4759.9560546875, "sentence_full_gradient_variance/p75": 4759.9560546875, "sentence_full_gradient_variance/p90": 4759.9560546875, "sentence_full_gradient_variance/p95": 4759.9560546875, "sentence_full_gradient_variance/p99": 4759.9560546875, "sentence_full_update_term": 0.014024734497070312, "sentence_full_update_term/max": 0.3515625, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.005462646484375, "sentence_full_update_term/p85": 0.014260871335864067, "sentence_full_update_term/p90": 0.022534174844622612, "sentence_full_update_term/p95": 0.051745541393756866, "sentence_full_update_term/p99": 0.2944725751876831, "sentence_full_update_term/var": 0.0026700447779148817, "sentence_hessian_coeff": 139109.203125, "sentence_hessian_coeff/max": 7372800.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": -1064960.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 3515674.25, "sentence_hessian_coeff/var": 853760278528.0, "sentence_hessian_coeff_abs": 208436.40625, "sentence_hessian_coeff_abs/max": 7372800.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 8168.0, "sentence_hessian_coeff_abs/p99": 3515674.25, "sentence_hessian_coeff_abs/var": 829360898048.0, "step": 3, "token_fisher_curvature": 15088260.0, "token_fisher_curvature/max": 11475615744.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 8.952838470577262e-13, "token_fisher_curvature/p85": 49.900390625, "token_fisher_curvature/p90": 485376.0, "token_fisher_curvature/p95": 28180480.0, "token_fisher_curvature/p99": 203931648.0, "token_fisher_curvature/var": 3.808359527271629e+16, "token_fisher_kl_divergence": 6.792024942114949e-05, "token_fisher_kl_divergence/max": 0.0517578125, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 4.03250548619601e-24, "token_fisher_kl_divergence/p85": 2.2428281454267562e-10, "token_fisher_kl_divergence/p90": 2.1904706954956055e-06, "token_fisher_kl_divergence/p95": 0.00012683868408203125, "token_fisher_kl_divergence/p99": 0.0009178370237350464, "token_fisher_kl_divergence/var": 7.722024406575656e-07, "token_full_update_term": 0.002458591479808092, "token_full_update_term/max": 0.921875, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 1.3287149158713873e-12, "token_full_update_term/p85": 9.119510650634766e-06, "token_full_update_term/p90": 0.0009010285139083862, "token_full_update_term/p95": 0.0063658952713012695, "token_full_update_term/p99": 0.044463157653808594, "token_full_update_term/var": 0.0005861047538928688, "token_hessian_coeff": 16774113.0, "token_hessian_coeff/max": 28185722880.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": -612368384.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 4169472.0, "token_hessian_coeff/var": 1.70708422981845e+17, "token_hessian_coeff_abs": 20560402.0, "token_hessian_coeff_abs/max": 28185722880.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 3.3527612686157227e-08, "token_hessian_coeff_abs/p99": 145752064.0, "token_hessian_coeff_abs/var": 1.7056706701819904e+17 }, { "accuracy_reward": 0.36250001192092896, "accuracy_reward/correct": 1.0, "accuracy_reward/correct/max": 1.0, "accuracy_reward/correct/median": 1.0, "accuracy_reward/correct/min": 1.0, "accuracy_reward/correct/p25": 1.0, "accuracy_reward/correct/p75": 1.0, "accuracy_reward/correct/var": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 1.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 1.0, "accuracy_reward/var": 0.23401899635791779, "adam_stats/lm_head/lr_effective_max": 3.283662226749584e-05, "adam_stats/lm_head/lr_effective_mean": -8.548686758480528e-11, "adam_stats/lm_head/lr_effective_min": -3.284077683929354e-05, "adam_stats/lm_head/lr_effective_std": 1.9233712009736337e-06, "adam_stats/lr_effective_max": 3.284417471149936e-05, "adam_stats/lr_effective_mean": 2.951713595766847e-10, "adam_stats/lr_effective_min": -3.2854142773430794e-05, "adam_stats/m_t_max": 0.0069174496456980705, "adam_stats/m_t_mean": -4.4843566360253106e-11, "adam_stats/m_t_min": -0.005659515503793955, "adam_stats/v_t_max": 3.573819640223519e-06, "adam_stats/v_t_mean": 8.466760278991181e-13, "adam_stats/v_t_min": 0.0, "advantages": 1.1920929132713809e-08, "advantages/max": 1.6198352575302124, "advantages/median": 0.0, "advantages/min": -2.4741740226745605, "advantages/p25": -0.5399450659751892, "advantages/p75": 0.5399450659751892, "advantages/var": 0.7085530757904053, "all_logprobs": -0.5386762022972107, "all_logprobs/max": 0.0, "all_logprobs/median": -0.038818359375, "all_logprobs/min": -11.6875, "all_logprobs/p1": -4.78125, "all_logprobs/p10": -1.796875, "all_logprobs/p25": -0.6328125, "all_logprobs/p5": -2.703125, "all_logprobs/p75": -0.0004367828369140625, "all_logprobs/var": 1.0271810293197632, "clip_ratio": 0.0, "completion_length": 519.6500244140625, "completion_length/correct": 425.6896667480469, "completion_length/correct/max": 851.0, "completion_length/correct/median": 388.0, "completion_length/correct/min": 192.0, "completion_length/correct/p25": 296.0, "completion_length/correct/p75": 499.0, "completion_length/correct/var": 30170.150390625, "completion_length/incorrect": 573.0784301757812, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 431.0, "completion_length/incorrect/min": 200.0, "completion_length/incorrect/p25": 317.5, "completion_length/incorrect/p75": 908.5, "completion_length/incorrect/var": 92243.078125, "completion_length/max": 1024.0, "completion_length/median": 422.0, "completion_length/min": 192.0, "completion_length/p25": 309.0, "completion_length/p75": 743.75, "completion_length/var": 74158.609375, "epoch": 0.005333333333333333, "feature_vector_variance/max_squared_error": 27025.412109375, "feature_vector_variance/metric": 15336.0625, "generated_tokens/total": 170532.0, "global_fisher_curvature": 980.0, "global_fisher_curvature/max": 980.0, "global_fisher_curvature/median": 980.0, "global_fisher_curvature/min": 980.0, "global_fisher_curvature/p25": 980.0, "global_fisher_curvature/p75": 980.0, "global_fisher_curvature/p85": 980.0, "global_fisher_curvature/p90": 980.0, "global_fisher_curvature/p95": 980.0, "global_fisher_curvature/p99": 980.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 9.89530235528946e-09, "global_fisher_kl_divergence/max": 9.89530235528946e-09, "global_fisher_kl_divergence/median": 9.89530235528946e-09, "global_fisher_kl_divergence/min": 9.89530235528946e-09, "global_fisher_kl_divergence/p25": 9.89530235528946e-09, "global_fisher_kl_divergence/p75": 9.89530235528946e-09, "global_fisher_kl_divergence/p85": 9.89530235528946e-09, "global_fisher_kl_divergence/p90": 9.89530235528946e-09, "global_fisher_kl_divergence/p95": 9.89530235528946e-09, "global_fisher_kl_divergence/p99": 9.89530235528946e-09, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.014892578125, "global_full_update_term/max": 0.014892578125, "global_full_update_term/median": 0.014892578125, "global_full_update_term/min": 0.014892578125, "global_full_update_term/p25": 0.014892578125, "global_full_update_term/p75": 0.014892578125, "global_full_update_term/p85": 0.014892578125, "global_full_update_term/p90": 0.014892578125, "global_full_update_term/p95": 0.014892578125, "global_full_update_term/p99": 0.014892578125, "global_full_update_term/var": NaN, "global_hessian_coeff": -44.75, "global_hessian_coeff/max": -44.75, "global_hessian_coeff/median": -44.75, "global_hessian_coeff/min": -44.75, "global_hessian_coeff/p25": -44.75, "global_hessian_coeff/p75": -44.75, "global_hessian_coeff/p99": -44.75, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 44.75, "global_hessian_coeff_abs/max": 44.75, "global_hessian_coeff_abs/median": 44.75, "global_hessian_coeff_abs/min": 44.75, "global_hessian_coeff_abs/p25": 44.75, "global_hessian_coeff_abs/p75": 44.75, "global_hessian_coeff_abs/p99": 44.75, "global_hessian_coeff_abs/var": NaN, "grad_norm": 1.3659777641296387, "grouped_std_rewards": 0.3737148642539978, "learning_rate": 6e-06, "loss": -0.0, "mean_logprobs": -0.46484375, "mean_logprobs/var": 0.0771484375, "num_completions/total": 320, "per_sentence_gradient_norm": 74.91719055175781, "per_sentence_gradient_norm/max": 360.0, "per_sentence_gradient_norm/median": 60.25, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 28.375, "per_sentence_gradient_norm/p75": 100.0, "per_sentence_gradient_norm/var": 5419.10107421875, "per_token_feature_norm": 140.44403076171875, "per_token_feature_norm/max": 151.0, "per_token_feature_norm/median": 141.0, "per_token_feature_norm/min": 86.0, "per_token_feature_norm/p25": 139.0, "per_token_feature_norm/p75": 144.0, "per_token_feature_norm/var": 34.75197219848633, "per_token_gradient_norm": 20.066776275634766, "per_token_gradient_norm/max": 508.0, "per_token_gradient_norm/median": 0.009033203125, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 12.6875, "per_token_gradient_norm/var": 2083.342041015625, "per_token_policy_error_norm": 0.23549284040927887, "per_token_policy_error_norm/max": 1.9921875, "per_token_policy_error_norm/median": 0.0, "per_token_policy_error_norm/min": 0.0, "per_token_policy_error_norm/p25": 0.0, "per_token_policy_error_norm/p75": 0.33203125, "per_token_policy_error_norm/var": 0.15874773263931274, "policy_entropy": 0.5972877740859985, "policy_entropy/max": 3.859375, "policy_entropy/median": 0.1923828125, "policy_entropy/min": 9.89530235528946e-09, "policy_entropy/p25": 0.004241943359375, "policy_entropy/p75": 0.98046875, "policy_entropy/var": 0.616187572479248, "policy_loss": -1.4901161193847656e-08, "policy_loss/max": 2.4741740226745605, "policy_loss/median": 0.0, "policy_loss/min": -1.6198352575302124, "policy_loss/p25": -0.5399450659751892, "policy_loss/p75": 0.5399450659751892, "policy_loss/var": 0.7085530161857605, "policy_sharpness": 3.653594970703125, "policy_sharpness/max": 10.0, "policy_sharpness/median": 1.7012414932250977, "policy_sharpness/min": -0.0, "policy_sharpness/p25": 0.418542742729187, "policy_sharpness/p75": 8.0, "policy_sharpness/var": 14.966253280639648, "reward": 0.36250001192092896, "reward/max": 1.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 1.0, "reward/var": 0.23401899635791779, "rewards/accuracy_reward": 0.36250001192092896, "rewards/accuracy_reward/max": 1.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 1.0, "rewards/accuracy_reward/var": 0.23401899635791779, "sentence_fisher_curvature": 430241.59375, "sentence_fisher_curvature/max": 6356992.0, "sentence_fisher_curvature/median": 141312.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 26336.0, "sentence_fisher_curvature/p75": 403968.0, "sentence_fisher_curvature/p85": 663347.375, "sentence_fisher_curvature/p90": 885555.125, "sentence_fisher_curvature/p95": 2020965.875, "sentence_fisher_curvature/p99": 3690656.75, "sentence_fisher_curvature/var": 789387149312.0, "sentence_fisher_kl_divergence": 4.355970304459333e-06, "sentence_fisher_kl_divergence/max": 6.437301635742188e-05, "sentence_fisher_kl_divergence/median": 1.430511474609375e-06, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 2.6635825634002686e-07, "sentence_fisher_kl_divergence/p75": 4.090368747711182e-06, "sentence_fisher_kl_divergence/p85": 6.70254439683049e-06, "sentence_fisher_kl_divergence/p90": 8.988379704533145e-06, "sentence_fisher_kl_divergence/p95": 2.0444387700990774e-05, "sentence_fisher_kl_divergence/p99": 3.7344663724070415e-05, "sentence_fisher_kl_divergence/var": 8.088873321154821e-11, "sentence_full_gradient_variance/max_squared_error": 10825.9482421875, "sentence_full_gradient_variance/metric": 10825.9482421875, "sentence_full_gradient_variance/p75": 10825.9482421875, "sentence_full_gradient_variance/p90": 10825.9482421875, "sentence_full_gradient_variance/p95": 10825.9482421875, "sentence_full_gradient_variance/p99": 10825.9482421875, "sentence_full_update_term": 0.047915268689394, "sentence_full_update_term/max": 0.5703125, "sentence_full_update_term/median": 0.0159912109375, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.00356292724609375, "sentence_full_update_term/p75": 0.0423583984375, "sentence_full_update_term/p85": 0.06376953423023224, "sentence_full_update_term/p90": 0.1005859300494194, "sentence_full_update_term/p95": 0.15502843260765076, "sentence_full_update_term/p99": 0.5070506930351257, "sentence_full_update_term/var": 0.010126445442438126, "sentence_hessian_coeff": -111274.6015625, "sentence_hessian_coeff/max": 4259840.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": -12779520.0, "sentence_hessian_coeff/p25": -84480.0, "sentence_hessian_coeff/p75": 48384.0, "sentence_hessian_coeff/p99": 2978446.0, "sentence_hessian_coeff/var": 2791961264128.0, "sentence_hessian_coeff_abs": 475384.59375, "sentence_hessian_coeff_abs/max": 12779520.0, "sentence_hessian_coeff_abs/median": 59648.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 5560.0, "sentence_hessian_coeff_abs/p75": 179712.0, "sentence_hessian_coeff_abs/p99": 6048965.0, "sentence_hessian_coeff_abs/var": 2575649472512.0, "step": 4, "token_fisher_curvature": 45799612.0, "token_fisher_curvature/max": 10938744832.0, "token_fisher_curvature/median": 1.4551915228366852e-08, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 51968.0, "token_fisher_curvature/p85": 13631488.0, "token_fisher_curvature/p90": 66584576.0, "token_fisher_curvature/p95": 212860928.0, "token_fisher_curvature/p99": 1090519040.0, "token_fisher_curvature/var": 5.464370926583808e+16, "token_fisher_kl_divergence": 0.0004637640668079257, "token_fisher_kl_divergence/max": 0.11083984375, "token_fisher_kl_divergence/median": 1.4738373282224826e-19, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 5.252659320831299e-07, "token_fisher_kl_divergence/p85": 0.00013828277587890625, "token_fisher_kl_divergence/p90": 0.000675201416015625, "token_fisher_kl_divergence/p95": 0.0021514892578125, "token_fisher_kl_divergence/p99": 0.01104736328125, "token_fisher_kl_divergence/var": 5.603381850960432e-06, "token_full_update_term": 0.010963143780827522, "token_full_update_term/max": 0.88671875, "token_full_update_term/median": 3.6561687011271715e-10, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0007219314575195312, "token_full_update_term/p85": 0.01123046875, "token_full_update_term/p90": 0.0244140625, "token_full_update_term/p95": 0.05992412567138672, "token_full_update_term/p99": 0.2265625, "token_full_update_term/var": 0.0017237757565453649, "token_hessian_coeff": -22144330.0, "token_hessian_coeff/max": 5872025600.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": -26575110144.0, "token_hessian_coeff/p25": -2.16066837310791e-05, "token_hessian_coeff/p75": 0.0003509521484375, "token_hessian_coeff/p99": 241172480.0, "token_hessian_coeff/var": 1.5476127813127373e+17, "token_hessian_coeff_abs": 49792252.0, "token_hessian_coeff_abs/max": 26575110144.0, "token_hessian_coeff_abs/median": 7.867813110351562e-05, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 243712.0, "token_hessian_coeff_abs/p99": 1409286144.0, "token_hessian_coeff_abs/var": 1.5277234749597286e+17 }, { "accuracy_reward": 0.6500000357627869, "accuracy_reward/correct": 1.0, "accuracy_reward/correct/max": 1.0, "accuracy_reward/correct/median": 1.0, "accuracy_reward/correct/min": 1.0, "accuracy_reward/correct/p25": 1.0, "accuracy_reward/correct/p75": 1.0, "accuracy_reward/correct/var": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 1.0, "accuracy_reward/median": 1.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 1.0, "accuracy_reward/var": 0.23037976026535034, "adam_stats/lm_head/lr_effective_max": 4.3810076022055e-05, "adam_stats/lm_head/lr_effective_mean": -6.883183606420928e-11, "adam_stats/lm_head/lr_effective_min": -4.377226287033409e-05, "adam_stats/lm_head/lr_effective_std": 2.3811558094166685e-06, "adam_stats/lr_effective_max": 4.384678686619736e-05, "adam_stats/lr_effective_mean": 5.833396543586389e-10, "adam_stats/lr_effective_min": -4.3856467527803034e-05, "adam_stats/m_t_max": 0.00827067345380783, "adam_stats/m_t_mean": -3.50516178582172e-11, "adam_stats/m_t_min": -0.006699834484606981, "adam_stats/v_t_max": 3.9884298530523665e-06, "adam_stats/v_t_mean": 1.0584728623100337e-12, "adam_stats/v_t_min": 0.0, "advantages": -9.685755308908028e-09, "advantages/max": 0.7244288325309753, "advantages/median": 0.0, "advantages/min": -2.4741740226745605, "advantages/p25": 0.0, "advantages/p75": 0.5399450659751892, "advantages/var": 0.5314004421234131, "all_logprobs": -0.32906797528266907, "all_logprobs/max": 0.0, "all_logprobs/median": -0.001739501953125, "all_logprobs/min": -10.75, "all_logprobs/p1": -3.96875, "all_logprobs/p10": -1.1015625, "all_logprobs/p25": -0.2021484375, "all_logprobs/p5": -1.9765625, "all_logprobs/p75": -1.7762184143066406e-05, "all_logprobs/var": 0.6404090523719788, "clip_ratio": 0.0, "completion_length": 476.8999938964844, "completion_length/correct": 360.3077087402344, "completion_length/correct/max": 770.0, "completion_length/correct/median": 298.0, "completion_length/correct/min": 192.0, "completion_length/correct/p25": 243.25, "completion_length/correct/p75": 475.75, "completion_length/correct/var": 22867.943359375, "completion_length/incorrect": 693.4285888671875, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 602.0, "completion_length/incorrect/min": 268.0, "completion_length/incorrect/p25": 438.25, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 80445.890625, "completion_length/max": 1024.0, "completion_length/median": 387.0, "completion_length/min": 192.0, "completion_length/p25": 272.5, "completion_length/p75": 595.0, "completion_length/var": 67822.140625, "epoch": 0.006666666666666667, "feature_vector_variance/max_squared_error": 26059.130859375, "feature_vector_variance/metric": 16668.056640625, "generated_tokens/total": 208684.0, "global_fisher_curvature": 1840.0, "global_fisher_curvature/max": 1840.0, "global_fisher_curvature/median": 1840.0, "global_fisher_curvature/min": 1840.0, "global_fisher_curvature/p25": 1840.0, "global_fisher_curvature/p75": 1840.0, "global_fisher_curvature/p85": 1840.0, "global_fisher_curvature/p90": 1840.0, "global_fisher_curvature/p95": 1840.0, "global_fisher_curvature/p99": 1840.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 3.306195139884949e-08, "global_fisher_kl_divergence/max": 3.306195139884949e-08, "global_fisher_kl_divergence/median": 3.306195139884949e-08, "global_fisher_kl_divergence/min": 3.306195139884949e-08, "global_fisher_kl_divergence/p25": 3.306195139884949e-08, "global_fisher_kl_divergence/p75": 3.306195139884949e-08, "global_fisher_kl_divergence/p85": 3.306195139884949e-08, "global_fisher_kl_divergence/p90": 3.306195139884949e-08, "global_fisher_kl_divergence/p95": 3.306195139884949e-08, "global_fisher_kl_divergence/p99": 3.306195139884949e-08, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0223388671875, "global_full_update_term/max": 0.0223388671875, "global_full_update_term/median": 0.0223388671875, "global_full_update_term/min": 0.0223388671875, "global_full_update_term/p25": 0.0223388671875, "global_full_update_term/p75": 0.0223388671875, "global_full_update_term/p85": 0.0223388671875, "global_full_update_term/p90": 0.0223388671875, "global_full_update_term/p95": 0.0223388671875, "global_full_update_term/p99": 0.0223388671875, "global_full_update_term/var": NaN, "global_hessian_coeff": -1608.0, "global_hessian_coeff/max": -1608.0, "global_hessian_coeff/median": -1608.0, "global_hessian_coeff/min": -1608.0, "global_hessian_coeff/p25": -1608.0, "global_hessian_coeff/p75": -1608.0, "global_hessian_coeff/p99": -1608.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 1608.0, "global_hessian_coeff_abs/max": 1608.0, "global_hessian_coeff_abs/median": 1608.0, "global_hessian_coeff_abs/min": 1608.0, "global_hessian_coeff_abs/p25": 1608.0, "global_hessian_coeff_abs/p75": 1608.0, "global_hessian_coeff_abs/p99": 1608.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 1.0505958795547485, "grouped_std_rewards": 0.26680251955986023, "learning_rate": 7.5e-06, "loss": 0.0, "mean_logprobs": -0.287109375, "mean_logprobs/var": 0.0311279296875, "num_completions/total": 400, "per_sentence_gradient_norm": 53.759376525878906, "per_sentence_gradient_norm/max": 402.0, "per_sentence_gradient_norm/median": 35.5, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 72.375, "per_sentence_gradient_norm/var": 4883.451171875, "per_token_feature_norm": 141.93946838378906, "per_token_feature_norm/max": 151.0, "per_token_feature_norm/median": 143.0, "per_token_feature_norm/min": 90.5, "per_token_feature_norm/p25": 140.0, "per_token_feature_norm/p75": 146.0, "per_token_feature_norm/var": 34.97080993652344, "per_token_gradient_norm": 12.396733283996582, "per_token_gradient_norm/max": 508.0, "per_token_gradient_norm/median": 2.2172927856445312e-05, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.330078125, "per_token_gradient_norm/var": 1464.120361328125, "per_token_policy_error_norm": 0.15153422951698303, "per_token_policy_error_norm/max": 1.984375, "per_token_policy_error_norm/median": 0.0, "per_token_policy_error_norm/min": 0.0, "per_token_policy_error_norm/p25": 0.0, "per_token_policy_error_norm/p75": 0.0546875, "per_token_policy_error_norm/var": 0.11528592556715012, "policy_entropy": 0.36456283926963806, "policy_entropy/max": 3.796875, "policy_entropy/median": 0.01416015625, "policy_entropy/min": 1.4260876923799515e-08, "policy_entropy/p25": 0.0002295970916748047, "policy_entropy/p75": 0.55078125, "policy_entropy/var": 0.39265722036361694, "policy_loss": 9.685755308908028e-09, "policy_loss/max": 2.4741740226745605, "policy_loss/median": 0.0, "policy_loss/min": -0.7244288921356201, "policy_loss/p25": -0.5399450063705444, "policy_loss/p75": 0.0, "policy_loss/var": 0.5314004421234131, "policy_sharpness": 5.396049499511719, "policy_sharpness/max": 10.0, "policy_sharpness/median": 4.435546875, "policy_sharpness/min": -0.0, "policy_sharpness/p25": 1.1290457248687744, "policy_sharpness/p75": 10.0, "policy_sharpness/var": 17.143077850341797, "reward": 0.6500000357627869, "reward/max": 1.0, "reward/median": 1.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 1.0, "reward/var": 0.23037976026535034, "rewards/accuracy_reward": 0.6500000357627869, "rewards/accuracy_reward/max": 1.0, "rewards/accuracy_reward/median": 1.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 1.0, "rewards/accuracy_reward/var": 0.23037976026535034, "sentence_fisher_curvature": 301088.0, "sentence_fisher_curvature/max": 7241728.0, "sentence_fisher_curvature/median": 46848.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 214016.0, "sentence_fisher_curvature/p85": 418816.15625, "sentence_fisher_curvature/p90": 729497.375, "sentence_fisher_curvature/p95": 1174323.0, "sentence_fisher_curvature/p99": 3164565.0, "sentence_fisher_curvature/var": 784971923456.0, "sentence_fisher_kl_divergence": 5.4243605518422555e-06, "sentence_fisher_kl_divergence/max": 0.00013065338134765625, "sentence_fisher_kl_divergence/median": 8.419156074523926e-07, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 3.855675458908081e-06, "sentence_fisher_kl_divergence/p85": 7.548931080236798e-06, "sentence_fisher_kl_divergence/p90": 1.310705738433171e-05, "sentence_fisher_kl_divergence/p95": 2.115368261002004e-05, "sentence_fisher_kl_divergence/p99": 5.70081829209812e-05, "sentence_fisher_kl_divergence/var": 2.553269262683955e-10, "sentence_full_gradient_variance/max_squared_error": 7616.97802734375, "sentence_full_gradient_variance/metric": 7616.97802734375, "sentence_full_gradient_variance/p75": 7616.97802734375, "sentence_full_gradient_variance/p90": 7616.97802734375, "sentence_full_gradient_variance/p95": 7616.97802734375, "sentence_full_gradient_variance/p99": 7616.97802734375, "sentence_full_update_term": 0.04531135782599449, "sentence_full_update_term/max": 0.95703125, "sentence_full_update_term/median": 0.007537841796875, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.03045654296875, "sentence_full_update_term/p85": 0.08366699516773224, "sentence_full_update_term/p90": 0.13134759664535522, "sentence_full_update_term/p95": 0.17607419192790985, "sentence_full_update_term/p99": 0.4571087956428528, "sentence_full_update_term/var": 0.0143965482711792, "sentence_hessian_coeff": -294216.40625, "sentence_hessian_coeff/max": 407552.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": -13959168.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 27584.0, "sentence_hessian_coeff/p99": 404316.15625, "sentence_hessian_coeff/var": 2730654957568.0, "sentence_hessian_coeff_abs": 363946.8125, "sentence_hessian_coeff_abs/max": 13959168.0, "sentence_hessian_coeff_abs/median": 11840.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 71168.0, "sentence_hessian_coeff_abs/p99": 5986049.0, "sentence_hessian_coeff_abs/var": 2684180234240.0, "step": 5, "token_fisher_curvature": 30471578.0, "token_fisher_curvature/max": 11005853696.0, "token_fisher_curvature/median": 8.199278929421627e-19, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0450439453125, "token_fisher_curvature/p85": 68608.0, "token_fisher_curvature/p90": 6717440.0, "token_fisher_curvature/p95": 88604672.0, "token_fisher_curvature/p99": 719290368.0, "token_fisher_curvature/var": 4.852767583633408e+16, "token_fisher_kl_divergence": 0.0005486081936396658, "token_fisher_kl_divergence/max": 0.1982421875, "token_fisher_kl_divergence/median": 1.479114197289397e-29, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 8.117950756059145e-13, "token_fisher_kl_divergence/p85": 1.2367963790893555e-06, "token_fisher_kl_divergence/p90": 0.00012111663818359375, "token_fisher_kl_divergence/p95": 0.00159454345703125, "token_fisher_kl_divergence/p99": 0.012969493865966797, "token_fisher_kl_divergence/var": 1.5733146938146092e-05, "token_full_update_term": 0.009199532680213451, "token_full_update_term/max": 1.0625, "token_full_update_term/median": 2.9698465908722937e-15, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 6.528571248054504e-07, "token_full_update_term/p85": 0.001007080078125, "token_full_update_term/p90": 0.00970458984375, "token_full_update_term/p95": 0.046630859375, "token_full_update_term/p99": 0.197265625, "token_full_update_term/var": 0.002087129047140479, "token_hessian_coeff": -28138112.0, "token_hessian_coeff/max": 616562688.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": -26709327872.0, "token_hessian_coeff/p25": -1.1094380170106888e-07, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 95420416.0, "token_hessian_coeff/var": 1.605435238824018e+17, "token_hessian_coeff_abs": 34791180.0, "token_hessian_coeff_abs/max": 26709327872.0, "token_hessian_coeff_abs/median": 1.1226575225009583e-12, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 5.4453125, "token_hessian_coeff_abs/p99": 771751936.0, "token_hessian_coeff_abs/var": 1.601248504703877e+17 }, { "accuracy_reward": 0.6625000238418579, "accuracy_reward/correct": 1.0, "accuracy_reward/correct/max": 1.0, "accuracy_reward/correct/median": 1.0, "accuracy_reward/correct/min": 1.0, "accuracy_reward/correct/p25": 1.0, "accuracy_reward/correct/p75": 1.0, "accuracy_reward/correct/var": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 1.0, "accuracy_reward/median": 1.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 1.0, "accuracy_reward/var": 0.22642405331134796, "adam_stats/lm_head/lr_effective_max": 5.49818214494735e-05, "adam_stats/lm_head/lr_effective_mean": -2.3507324559135512e-11, "adam_stats/lm_head/lr_effective_min": -5.5011772928992286e-05, "adam_stats/lm_head/lr_effective_std": 2.799985395540716e-06, "adam_stats/lr_effective_max": 5.524799053091556e-05, "adam_stats/lr_effective_mean": 8.140540463941193e-10, "adam_stats/lr_effective_min": -5.514492659131065e-05, "adam_stats/m_t_max": 0.007571779657155275, "adam_stats/m_t_mean": -3.911718171600853e-11, "adam_stats/m_t_min": -0.0069343955256044865, "adam_stats/v_t_max": 3.986084266216494e-06, "adam_stats/v_t_mean": 1.1453413087739173e-12, "adam_stats/v_t_min": 0.0, "advantages": -4.470348535789981e-09, "advantages/max": 2.4741740226745605, "advantages/median": 0.0, "advantages/min": -2.4741740226745605, "advantages/p25": 0.0, "advantages/p75": 0.3534534275531769, "advantages/var": 0.35424181818962097, "all_logprobs": -0.3733700215816498, "all_logprobs/max": 0.0, "all_logprobs/median": -0.000926971435546875, "all_logprobs/min": -13.75, "all_logprobs/p1": -4.59375, "all_logprobs/p10": -1.28125, "all_logprobs/p25": -0.201171875, "all_logprobs/p5": -2.3453121185302734, "all_logprobs/p75": -3.6954879760742188e-06, "all_logprobs/var": 0.8291133642196655, "clip_ratio": 0.0, "completion_length": 502.7375183105469, "completion_length/correct": 404.2075500488281, "completion_length/correct/max": 916.0, "completion_length/correct/median": 324.0, "completion_length/correct/min": 177.0, "completion_length/correct/p25": 255.0, "completion_length/correct/p75": 425.0, "completion_length/correct/var": 47365.3984375, "completion_length/incorrect": 696.1481323242188, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 622.0, "completion_length/incorrect/min": 292.0, "completion_length/incorrect/p25": 510.5, "completion_length/incorrect/p75": 923.0, "completion_length/incorrect/var": 56579.27734375, "completion_length/max": 1024.0, "completion_length/median": 409.0, "completion_length/min": 177.0, "completion_length/p25": 296.25, "completion_length/p75": 732.5, "completion_length/var": 69096.21875, "epoch": 0.008, "feature_vector_variance/max_squared_error": 25653.79296875, "feature_vector_variance/metric": 17233.087890625, "generated_tokens/total": 248903.0, "global_fisher_curvature": 624.0, "global_fisher_curvature/max": 624.0, "global_fisher_curvature/median": 624.0, "global_fisher_curvature/min": 624.0, "global_fisher_curvature/p25": 624.0, "global_fisher_curvature/p75": 624.0, "global_fisher_curvature/p85": 624.0, "global_fisher_curvature/p90": 624.0, "global_fisher_curvature/p95": 624.0, "global_fisher_curvature/p99": 624.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 1.7578713595867157e-08, "global_fisher_kl_divergence/max": 1.7578713595867157e-08, "global_fisher_kl_divergence/median": 1.7578713595867157e-08, "global_fisher_kl_divergence/min": 1.7578713595867157e-08, "global_fisher_kl_divergence/p25": 1.7578713595867157e-08, "global_fisher_kl_divergence/p75": 1.7578713595867157e-08, "global_fisher_kl_divergence/p85": 1.7578713595867157e-08, "global_fisher_kl_divergence/p90": 1.7578713595867157e-08, "global_fisher_kl_divergence/p95": 1.7578713595867157e-08, "global_fisher_kl_divergence/p99": 1.7578713595867157e-08, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.041015625, "global_full_update_term/max": 0.041015625, "global_full_update_term/median": 0.041015625, "global_full_update_term/min": 0.041015625, "global_full_update_term/p25": 0.041015625, "global_full_update_term/p75": 0.041015625, "global_full_update_term/p85": 0.041015625, "global_full_update_term/p90": 0.041015625, "global_full_update_term/p95": 0.041015625, "global_full_update_term/p99": 0.041015625, "global_full_update_term/var": NaN, "global_hessian_coeff": -720.0, "global_hessian_coeff/max": -720.0, "global_hessian_coeff/median": -720.0, "global_hessian_coeff/min": -720.0, "global_hessian_coeff/p25": -720.0, "global_hessian_coeff/p75": -720.0, "global_hessian_coeff/p99": -720.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 720.0, "global_hessian_coeff_abs/max": 720.0, "global_hessian_coeff_abs/median": 720.0, "global_hessian_coeff_abs/min": 720.0, "global_hessian_coeff_abs/p25": 720.0, "global_hessian_coeff_abs/p75": 720.0, "global_hessian_coeff_abs/p99": 720.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.6593793034553528, "grouped_std_rewards": 0.15235702693462372, "learning_rate": 9e-06, "loss": 0.0, "mean_logprobs": -0.31640625, "mean_logprobs/var": 0.09521484375, "num_completions/total": 480, "per_sentence_gradient_norm": 27.803125381469727, "per_sentence_gradient_norm/max": 400.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 39.375, "per_sentence_gradient_norm/var": 3485.4755859375, "per_token_feature_norm": 142.32034301757812, "per_token_feature_norm/max": 151.0, "per_token_feature_norm/median": 143.0, "per_token_feature_norm/min": 97.0, "per_token_feature_norm/p25": 140.0, "per_token_feature_norm/p75": 146.0, "per_token_feature_norm/var": 36.47355651855469, "per_token_gradient_norm": 11.320646286010742, "per_token_gradient_norm/max": 512.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.01446533203125, "per_token_gradient_norm/var": 2205.93310546875, "per_token_policy_error_norm": 0.15885263681411743, "per_token_policy_error_norm/max": 2.0, "per_token_policy_error_norm/median": 0.0, "per_token_policy_error_norm/min": 0.0, "per_token_policy_error_norm/p25": 0.0, "per_token_policy_error_norm/p75": 0.0546875, "per_token_policy_error_norm/var": 0.12212774157524109, "policy_entropy": 0.4046971797943115, "policy_entropy/max": 3.71875, "policy_entropy/median": 0.0081787109375, "policy_entropy/min": 1.064108801074326e-10, "policy_entropy/p25": 5.364418029785156e-05, "policy_entropy/p75": 0.5546875, "policy_entropy/var": 0.5451201796531677, "policy_loss": 4.470348535789981e-09, "policy_loss/max": 2.4741742610931396, "policy_loss/median": 0.0, "policy_loss/min": -2.4741740226745605, "policy_loss/p25": -0.3534534275531769, "policy_loss/p75": 0.0, "policy_loss/var": 0.35424187779426575, "policy_sharpness": 5.643321514129639, "policy_sharpness/max": 10.0, "policy_sharpness/median": 6.24609375, "policy_sharpness/min": -0.0, "policy_sharpness/p25": 1.082991123199463, "policy_sharpness/p75": 10.0, "policy_sharpness/var": 17.95266342163086, "reward": 0.6625000238418579, "reward/max": 1.0, "reward/median": 1.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 1.0, "reward/var": 0.22642405331134796, "rewards/accuracy_reward": 0.6625000238418579, "rewards/accuracy_reward/max": 1.0, "rewards/accuracy_reward/median": 1.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 1.0, "rewards/accuracy_reward/var": 0.22642405331134796, "sentence_fisher_curvature": 95138.3984375, "sentence_fisher_curvature/max": 2981888.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 55232.0, "sentence_fisher_curvature/p85": 80204.8125, "sentence_fisher_curvature/p90": 130047.9375, "sentence_fisher_curvature/p95": 353279.90625, "sentence_fisher_curvature/p99": 1480456.5, "sentence_fisher_curvature/var": 134836895744.0, "sentence_fisher_kl_divergence": 2.6756199531519087e-06, "sentence_fisher_kl_divergence/max": 8.392333984375e-05, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 1.5534460544586182e-06, "sentence_fisher_kl_divergence/p85": 2.2575261482415954e-06, "sentence_fisher_kl_divergence/p90": 3.6627036479330854e-06, "sentence_fisher_kl_divergence/p95": 9.924170626618434e-06, "sentence_fisher_kl_divergence/p99": 4.163856283412315e-05, "sentence_fisher_kl_divergence/var": 1.0673278233452876e-10, "sentence_full_gradient_variance/max_squared_error": 4160.9208984375, "sentence_full_gradient_variance/metric": 4160.9208984375, "sentence_full_gradient_variance/p75": 4160.9208984375, "sentence_full_gradient_variance/p90": 4160.9208984375, "sentence_full_gradient_variance/p95": 4160.9208984375, "sentence_full_gradient_variance/p99": 4160.9208984375, "sentence_full_update_term": 0.030424881726503372, "sentence_full_update_term/max": 1.15625, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0114593505859375, "sentence_full_update_term/p85": 0.01613159105181694, "sentence_full_update_term/p90": 0.02509765513241291, "sentence_full_update_term/p95": 0.07097131013870239, "sentence_full_update_term/p99": 0.5575774312019348, "sentence_full_update_term/var": 0.019335588440299034, "sentence_hessian_coeff": -63459.30078125, "sentence_hessian_coeff/max": 2211840.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": -5799936.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 6320.0, "sentence_hessian_coeff/p99": 597154.0, "sentence_hessian_coeff/var": 511287689216.0, "sentence_hessian_coeff_abs": 133497.703125, "sentence_hessian_coeff_abs/max": 5799936.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 11424.0, "sentence_hessian_coeff_abs/p99": 2965337.0, "sentence_hessian_coeff_abs/var": 497318461440.0, "step": 6, "token_fisher_curvature": 43069864.0, "token_fisher_curvature/max": 11207180288.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 1.51805579662323e-07, "token_fisher_curvature/p85": 265.203125, "token_fisher_curvature/p90": 450560.0, "token_fisher_curvature/p95": 36700160.0, "token_fisher_curvature/p99": 1417674752.0, "token_fisher_curvature/var": 1.191949030612009e+17, "token_fisher_kl_divergence": 0.0012110108509659767, "token_fisher_kl_divergence/max": 0.314453125, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 4.2825985813177425e-18, "token_fisher_kl_divergence/p85": 7.447738425980788e-09, "token_fisher_kl_divergence/p90": 1.2695789337158203e-05, "token_fisher_kl_divergence/p95": 0.00102996826171875, "token_fisher_kl_divergence/p99": 0.039794921875, "token_fisher_kl_divergence/var": 9.42251062951982e-05, "token_full_update_term": 0.01581099070608616, "token_full_update_term/max": 2.734375, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 1.57160684466362e-09, "token_full_update_term/p85": 7.62939453125e-05, "token_full_update_term/p90": 0.002581775188446045, "token_full_update_term/p95": 0.021240234375, "token_full_update_term/p99": 0.59765625, "token_full_update_term/var": 0.009477803483605385, "token_hessian_coeff": -60318760.0, "token_hessian_coeff/max": 27648851968.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": -21877489664.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 61294592.0, "token_hessian_coeff/var": 5.553117492223672e+17, "token_hessian_coeff_abs": 77289464.0, "token_hessian_coeff_abs/max": 27648851968.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0003528594970703125, "token_hessian_coeff_abs/p99": 2278686720.0, "token_hessian_coeff_abs/var": 5.5297635216523264e+17 }, { "accuracy_reward": 0.75, "accuracy_reward/correct": 1.0, "accuracy_reward/correct/max": 1.0, "accuracy_reward/correct/median": 1.0, "accuracy_reward/correct/min": 1.0, "accuracy_reward/correct/p25": 1.0, "accuracy_reward/correct/p75": 1.0, "accuracy_reward/correct/var": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 1.0, "accuracy_reward/median": 1.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.75, "accuracy_reward/p75": 1.0, "accuracy_reward/var": 0.18987341225147247, "adam_stats/lm_head/lr_effective_max": 6.579548062290996e-05, "adam_stats/lm_head/lr_effective_mean": 1.1277245109964085e-10, "adam_stats/lm_head/lr_effective_min": -6.647277041338384e-05, "adam_stats/lm_head/lr_effective_std": 3.6423425626708195e-06, "adam_stats/lr_effective_max": 6.665560067631304e-05, "adam_stats/lr_effective_mean": 6.070495217613825e-10, "adam_stats/lr_effective_min": -6.651657167822123e-05, "adam_stats/m_t_max": 0.010892059653997421, "adam_stats/m_t_mean": -9.465773304073721e-11, "adam_stats/m_t_min": -0.007430214434862137, "adam_stats/v_t_max": 5.644642897095764e-06, "adam_stats/v_t_mean": 1.4150009289282117e-12, "adam_stats/v_t_min": 0.0, "advantages": -5.9604645663569045e-09, "advantages/max": 2.4741740226745605, "advantages/median": 0.0, "advantages/min": -2.4741740226745605, "advantages/p25": -0.08836335688829422, "advantages/p75": 0.3534534275531769, "advantages/var": 0.6199482083320618, "all_logprobs": -0.5707439184188843, "all_logprobs/max": 0.0, "all_logprobs/median": -0.0020904541015625, "all_logprobs/min": -10.3125, "all_logprobs/p1": -5.375, "all_logprobs/p10": -2.40625, "all_logprobs/p25": -0.314453125, "all_logprobs/p5": -3.90625, "all_logprobs/p75": -7.3909759521484375e-06, "all_logprobs/var": 1.5825971364974976, "clip_ratio": 0.0, "completion_length": 516.5750122070312, "completion_length/correct": 468.5000305175781, "completion_length/correct/max": 1024.0, "completion_length/correct/median": 438.0, "completion_length/correct/min": 195.0, "completion_length/correct/p25": 339.25, "completion_length/correct/p75": 524.25, "completion_length/correct/var": 33261.7109375, "completion_length/incorrect": 660.7999877929688, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 633.0, "completion_length/incorrect/min": 327.0, "completion_length/incorrect/p25": 450.25, "completion_length/incorrect/p75": 826.75, "completion_length/incorrect/var": 61564.27734375, "completion_length/max": 1024.0, "completion_length/median": 462.0, "completion_length/min": 195.0, "completion_length/p25": 359.0, "completion_length/p75": 577.25, "completion_length/var": 46669.0078125, "epoch": 0.009333333333333334, "feature_vector_variance/max_squared_error": 25298.990234375, "feature_vector_variance/metric": 17848.375, "generated_tokens/total": 290229.0, "global_fisher_curvature": 19968.0, "global_fisher_curvature/max": 19968.0, "global_fisher_curvature/median": 19968.0, "global_fisher_curvature/min": 19968.0, "global_fisher_curvature/p25": 19968.0, "global_fisher_curvature/p75": 19968.0, "global_fisher_curvature/p85": 19968.0, "global_fisher_curvature/p90": 19968.0, "global_fisher_curvature/p95": 19968.0, "global_fisher_curvature/p99": 19968.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 8.083879947662354e-07, "global_fisher_kl_divergence/max": 8.083879947662354e-07, "global_fisher_kl_divergence/median": 8.083879947662354e-07, "global_fisher_kl_divergence/min": 8.083879947662354e-07, "global_fisher_kl_divergence/p25": 8.083879947662354e-07, "global_fisher_kl_divergence/p75": 8.083879947662354e-07, "global_fisher_kl_divergence/p85": 8.083879947662354e-07, "global_fisher_kl_divergence/p90": 8.083879947662354e-07, "global_fisher_kl_divergence/p95": 8.083879947662354e-07, "global_fisher_kl_divergence/p99": 8.083879947662354e-07, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.46484375, "global_full_update_term/max": 0.46484375, "global_full_update_term/median": 0.46484375, "global_full_update_term/min": 0.46484375, "global_full_update_term/p25": 0.46484375, "global_full_update_term/p75": 0.46484375, "global_full_update_term/p85": 0.46484375, "global_full_update_term/p90": 0.46484375, "global_full_update_term/p95": 0.46484375, "global_full_update_term/p99": 0.46484375, "global_full_update_term/var": NaN, "global_hessian_coeff": 832.0, "global_hessian_coeff/max": 832.0, "global_hessian_coeff/median": 832.0, "global_hessian_coeff/min": 832.0, "global_hessian_coeff/p25": 832.0, "global_hessian_coeff/p75": 832.0, "global_hessian_coeff/p99": 832.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 832.0, "global_hessian_coeff_abs/max": 832.0, "global_hessian_coeff_abs/median": 832.0, "global_hessian_coeff_abs/min": 832.0, "global_hessian_coeff_abs/p25": 832.0, "global_hessian_coeff_abs/p75": 832.0, "global_hessian_coeff_abs/p99": 832.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 2.155524492263794, "grouped_std_rewards": 0.2946168780326843, "learning_rate": 1.05e-05, "loss": 0.0, "mean_logprobs": -0.431640625, "mean_logprobs/var": 0.287109375, "num_completions/total": 560, "per_sentence_gradient_norm": 93.3031234741211, "per_sentence_gradient_norm/max": 948.0, "per_sentence_gradient_norm/median": 58.25, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 115.625, "per_sentence_gradient_norm/var": 20639.185546875, "per_token_feature_norm": 142.6071319580078, "per_token_feature_norm/max": 151.0, "per_token_feature_norm/median": 144.0, "per_token_feature_norm/min": 89.0, "per_token_feature_norm/p25": 139.0, "per_token_feature_norm/p75": 147.0, "per_token_feature_norm/var": 40.37157440185547, "per_token_gradient_norm": 26.222776412963867, "per_token_gradient_norm/max": 508.0, "per_token_gradient_norm/median": 0.005035400390625, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 14.0625, "per_token_gradient_norm/var": 4495.5, "per_token_policy_error_norm": 0.1989469975233078, "per_token_policy_error_norm/max": 2.0, "per_token_policy_error_norm/median": 0.0, "per_token_policy_error_norm/min": 0.0, "per_token_policy_error_norm/p25": 0.0, "per_token_policy_error_norm/p75": 0.12109375, "per_token_policy_error_norm/var": 0.1504020243883133, "policy_entropy": 0.5577514171600342, "policy_entropy/max": 3.90625, "policy_entropy/median": 0.0164794921875, "policy_entropy/min": 3.9472070056945086e-10, "policy_entropy/p25": 0.00010251998901367188, "policy_entropy/p75": 0.65234375, "policy_entropy/var": 1.039861798286438, "policy_loss": 1.639127766850379e-08, "policy_loss/max": 2.4741742610931396, "policy_loss/median": 0.0, "policy_loss/min": -2.4741740226745605, "policy_loss/p25": -0.3534534275531769, "policy_loss/p75": 0.08836335688829422, "policy_loss/var": 0.6199482679367065, "policy_sharpness": 5.364227294921875, "policy_sharpness/max": 10.0, "policy_sharpness/median": 4.437499523162842, "policy_sharpness/min": -0.0, "policy_sharpness/p25": 0.9638770818710327, "policy_sharpness/p75": 10.0, "policy_sharpness/var": 17.88347816467285, "reward": 0.75, "reward/max": 1.0, "reward/median": 1.0, "reward/min": 0.0, "reward/p25": 0.75, "reward/p75": 1.0, "reward/var": 0.18987341225147247, "rewards/accuracy_reward": 0.75, "rewards/accuracy_reward/max": 1.0, "rewards/accuracy_reward/median": 1.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.75, "rewards/accuracy_reward/p75": 1.0, "rewards/accuracy_reward/var": 0.18987341225147247, "sentence_fisher_curvature": 893580.8125, "sentence_fisher_curvature/max": 27787264.0, "sentence_fisher_curvature/median": 97792.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 487424.0, "sentence_fisher_curvature/p85": 829235.25, "sentence_fisher_curvature/p90": 1137049.375, "sentence_fisher_curvature/p95": 3550408.75, "sentence_fisher_curvature/p99": 12151667.0, "sentence_fisher_curvature/var": 11201055555584.0, "sentence_fisher_kl_divergence": 3.6240882764104754e-05, "sentence_fisher_kl_divergence/max": 0.001129150390625, "sentence_fisher_kl_divergence/median": 3.9637088775634766e-06, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 1.9729137420654297e-05, "sentence_fisher_kl_divergence/p85": 3.368854595464654e-05, "sentence_fisher_kl_divergence/p90": 4.613398778019473e-05, "sentence_fisher_kl_divergence/p95": 0.00014386160182766616, "sentence_fisher_kl_divergence/p99": 0.000493277795612812, "sentence_fisher_kl_divergence/var": 1.8478395702459238e-08, "sentence_full_gradient_variance/max_squared_error": 28710.669921875, "sentence_full_gradient_variance/metric": 28710.669921875, "sentence_full_gradient_variance/p75": 28710.669921875, "sentence_full_gradient_variance/p90": 28710.669921875, "sentence_full_gradient_variance/p95": 28710.669921875, "sentence_full_gradient_variance/p99": 28710.669921875, "sentence_full_update_term": 0.2523452937602997, "sentence_full_update_term/max": 7.84375, "sentence_full_update_term/median": 0.0299072265625, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.11669921875, "sentence_full_update_term/p85": 0.2026367336511612, "sentence_full_update_term/p90": 0.3039061427116394, "sentence_full_update_term/p95": 1.0582005977630615, "sentence_full_update_term/p99": 4.128276824951172, "sentence_full_update_term/var": 0.9339436292648315, "sentence_hessian_coeff": 195829.0, "sentence_hessian_coeff/max": 58720256.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": -16515072.0, "sentence_hessian_coeff/p25": -1200.0, "sentence_hessian_coeff/p75": 30464.0, "sentence_hessian_coeff/p99": 13091623.0, "sentence_hessian_coeff/var": 50267588395008.0, "sentence_hessian_coeff_abs": 1432145.875, "sentence_hessian_coeff_abs/max": 58720256.0, "sentence_hessian_coeff_abs/median": 23936.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 234752.0, "sentence_hessian_coeff_abs/p99": 25378122.0, "sentence_hessian_coeff_abs/var": 48229420892160.0, "step": 7, "token_fisher_curvature": 95546240.0, "token_fisher_curvature/max": 10938744832.0, "token_fisher_curvature/median": 2.2992026060819626e-09, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 128000.0, "token_fisher_curvature/p85": 33816576.0, "token_fisher_curvature/p90": 73400320.0, "token_fisher_curvature/p95": 335544320.0, "token_fisher_curvature/p99": 2189426688.0, "token_fisher_curvature/var": 2.0743808994403942e+17, "token_fisher_kl_divergence": 0.0038702560123056173, "token_fisher_kl_divergence/max": 0.443359375, "token_fisher_kl_divergence/median": 9.317362419797304e-20, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 5.185604095458984e-06, "token_fisher_kl_divergence/p85": 0.001373291015625, "token_fisher_kl_divergence/p90": 0.0029754638671875, "token_fisher_kl_divergence/p95": 0.01361083984375, "token_fisher_kl_divergence/p99": 0.088623046875, "token_fisher_kl_divergence/var": 0.0003404419112484902, "token_full_update_term": 0.04182954505085945, "token_full_update_term/max": 2.9375, "token_full_update_term/median": 2.28283170145005e-10, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.00177001953125, "token_full_update_term/p85": 0.02001953125, "token_full_update_term/p90": 0.0491943359375, "token_full_update_term/p95": 0.175537109375, "token_full_update_term/p99": 0.8359375, "token_full_update_term/var": 0.027788909152150154, "token_hessian_coeff": -119165792.0, "token_hessian_coeff/max": 21072183296.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": -26709327872.0, "token_hessian_coeff/p25": -0.033447265625, "token_hessian_coeff/p75": 1.7516867956146598e-09, "token_hessian_coeff/p99": 278921216.0, "token_hessian_coeff/var": 1.0599345005757727e+18, "token_hessian_coeff_abs": 178114688.0, "token_hessian_coeff_abs/max": 26709327872.0, "token_hessian_coeff_abs/median": 1.3649463653564453e-05, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 264192.0, "token_hessian_coeff_abs/p99": 5167382528.0, "token_hessian_coeff_abs/var": 1.0424097970575114e+18 }, { "accuracy_reward": 0.4375, "accuracy_reward/correct": 1.0, "accuracy_reward/correct/max": 1.0, "accuracy_reward/correct/median": 1.0, "accuracy_reward/correct/min": 1.0, "accuracy_reward/correct/p25": 1.0, "accuracy_reward/correct/p75": 1.0, "accuracy_reward/correct/var": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 1.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 1.0, "accuracy_reward/var": 0.24920886754989624, "adam_stats/lm_head/lr_effective_max": 7.758141146041453e-05, "adam_stats/lm_head/lr_effective_mean": 1.8269486617983333e-11, "adam_stats/lm_head/lr_effective_min": -7.662208372494206e-05, "adam_stats/lm_head/lr_effective_std": 4.6528039092663676e-06, "adam_stats/lr_effective_max": 7.789342635078356e-05, "adam_stats/lr_effective_mean": 7.766037257717073e-10, "adam_stats/lr_effective_min": -7.786529022268951e-05, "adam_stats/m_t_max": 0.0057770125567913055, "adam_stats/m_t_mean": -8.033457232770047e-11, "adam_stats/m_t_min": -0.006767276208847761, "adam_stats/v_t_max": 1.5640172932762653e-05, "adam_stats/v_t_mean": 1.68379546416908e-12, "adam_stats/v_t_min": 0.0, "advantages": -5.9604645663569045e-09, "advantages/max": 2.4741740226745605, "advantages/median": 0.0, "advantages/min": -1.6198352575302124, "advantages/p25": -0.7244288325309753, "advantages/p75": 0.5399450659751892, "advantages/var": 0.7085610628128052, "all_logprobs": -0.7340061068534851, "all_logprobs/max": 0.0, "all_logprobs/median": -0.00830078125, "all_logprobs/min": -10.9375, "all_logprobs/p1": -6.09375, "all_logprobs/p10": -3.09375, "all_logprobs/p25": -0.5234375, "all_logprobs/p5": -4.875, "all_logprobs/p75": -2.956390380859375e-05, "all_logprobs/var": 2.2956182956695557, "clip_ratio": 0.0, "completion_length": 736.0875244140625, "completion_length/correct": 700.028564453125, "completion_length/correct/max": 1024.0, "completion_length/correct/median": 656.0, "completion_length/correct/min": 287.0, "completion_length/correct/p25": 594.0, "completion_length/correct/p75": 862.5, "completion_length/correct/var": 44266.96875, "completion_length/incorrect": 764.1333618164062, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 771.0, "completion_length/incorrect/min": 371.0, "completion_length/incorrect/p25": 574.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 49544.02734375, "completion_length/max": 1024.0, "completion_length/median": 688.0, "completion_length/min": 287.0, "completion_length/p25": 584.75, "completion_length/p75": 987.75, "completion_length/var": 47669.8515625, "epoch": 0.010666666666666666, "feature_vector_variance/max_squared_error": 25212.73046875, "feature_vector_variance/metric": 17627.9765625, "generated_tokens/total": 349116.0, "global_fisher_curvature": 20736.0, "global_fisher_curvature/max": 20736.0, "global_fisher_curvature/median": 20736.0, "global_fisher_curvature/min": 20736.0, "global_fisher_curvature/p25": 20736.0, "global_fisher_curvature/p75": 20736.0, "global_fisher_curvature/p85": 20736.0, "global_fisher_curvature/p90": 20736.0, "global_fisher_curvature/p95": 20736.0, "global_fisher_curvature/p99": 20736.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 1.1399388313293457e-06, "global_fisher_kl_divergence/max": 1.1399388313293457e-06, "global_fisher_kl_divergence/median": 1.1399388313293457e-06, "global_fisher_kl_divergence/min": 1.1399388313293457e-06, "global_fisher_kl_divergence/p25": 1.1399388313293457e-06, "global_fisher_kl_divergence/p75": 1.1399388313293457e-06, "global_fisher_kl_divergence/p85": 1.1399388313293457e-06, "global_fisher_kl_divergence/p90": 1.1399388313293457e-06, "global_fisher_kl_divergence/p95": 1.1399388313293457e-06, "global_fisher_kl_divergence/p99": 1.1399388313293457e-06, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.7421875, "global_full_update_term/max": 0.7421875, "global_full_update_term/median": 0.7421875, "global_full_update_term/min": 0.7421875, "global_full_update_term/p25": 0.7421875, "global_full_update_term/p75": 0.7421875, "global_full_update_term/p85": 0.7421875, "global_full_update_term/p90": 0.7421875, "global_full_update_term/p95": 0.7421875, "global_full_update_term/p99": 0.7421875, "global_full_update_term/var": NaN, "global_hessian_coeff": 6688.0, "global_hessian_coeff/max": 6688.0, "global_hessian_coeff/median": 6688.0, "global_hessian_coeff/min": 6688.0, "global_hessian_coeff/p25": 6688.0, "global_hessian_coeff/p75": 6688.0, "global_hessian_coeff/p99": 6688.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 6688.0, "global_hessian_coeff_abs/max": 6688.0, "global_hessian_coeff_abs/median": 6688.0, "global_hessian_coeff_abs/min": 6688.0, "global_hessian_coeff_abs/p25": 6688.0, "global_hessian_coeff_abs/p75": 6688.0, "global_hessian_coeff_abs/p99": 6688.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 3.9059946537017822, "grouped_std_rewards": 0.38464269042015076, "learning_rate": 1.2e-05, "loss": -0.0, "mean_logprobs": -0.6640625, "mean_logprobs/var": 0.490234375, "num_completions/total": 640, "per_sentence_gradient_norm": 127.02812957763672, "per_sentence_gradient_norm/max": 368.0, "per_sentence_gradient_norm/median": 99.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 56.125, "per_sentence_gradient_norm/p75": 193.25, "per_sentence_gradient_norm/var": 10304.8544921875, "per_token_feature_norm": 141.91372680664062, "per_token_feature_norm/max": 151.0, "per_token_feature_norm/median": 144.0, "per_token_feature_norm/min": 87.0, "per_token_feature_norm/p25": 139.0, "per_token_feature_norm/p75": 147.0, "per_token_feature_norm/var": 42.56299591064453, "per_token_gradient_norm": 27.60826873779297, "per_token_gradient_norm/max": 512.0, "per_token_gradient_norm/median": 0.0191650390625, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 1.1742115020751953e-05, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 23.75, "per_token_gradient_norm/var": 3130.25634765625, "per_token_policy_error_norm": 0.23339435458183289, "per_token_policy_error_norm/max": 2.0, "per_token_policy_error_norm/median": 0.0, "per_token_policy_error_norm/min": 0.0, "per_token_policy_error_norm/p25": 0.0, "per_token_policy_error_norm/p75": 0.259765625, "per_token_policy_error_norm/var": 0.16958481073379517, "policy_entropy": 0.6688003540039062, "policy_entropy/max": 3.90625, "policy_entropy/median": 0.05419921875, "policy_entropy/min": 4.160938260611147e-11, "policy_entropy/p25": 0.0003681182861328125, "policy_entropy/p75": 0.83984375, "policy_entropy/var": 1.2053395509719849, "policy_loss": -7.450580596923828e-09, "policy_loss/max": 1.6198352575302124, "policy_loss/median": 0.0, "policy_loss/min": -2.4741740226745605, "policy_loss/p25": -0.5399450659751892, "policy_loss/p75": 0.7244287729263306, "policy_loss/var": 0.7085610628128052, "policy_sharpness": 4.692589282989502, "policy_sharpness/max": 10.0, "policy_sharpness/median": 2.8863282203674316, "policy_sharpness/min": -0.0, "policy_sharpness/p25": 0.4966130256652832, "policy_sharpness/p75": 10.0, "policy_sharpness/var": 18.074506759643555, "reward": 0.4375, "reward/max": 1.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 1.0, "reward/var": 0.24920886754989624, "rewards/accuracy_reward": 0.4375, "rewards/accuracy_reward/max": 1.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 1.0, "rewards/accuracy_reward/var": 0.24920886754989624, "sentence_fisher_curvature": 644771.1875, "sentence_fisher_curvature/max": 3850240.0, "sentence_fisher_curvature/median": 225280.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 63872.0, "sentence_fisher_curvature/p75": 1054720.0, "sentence_fisher_curvature/p85": 1409433.75, "sentence_fisher_curvature/p90": 1754726.375, "sentence_fisher_curvature/p95": 2054551.5, "sentence_fisher_curvature/p99": 3461938.75, "sentence_fisher_curvature/var": 690429034496.0, "sentence_fisher_kl_divergence": 3.554969953256659e-05, "sentence_fisher_kl_divergence/max": 0.00021266937255859375, "sentence_fisher_kl_divergence/median": 1.239776611328125e-05, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 3.5241246223449707e-06, "sentence_fisher_kl_divergence/p75": 5.823373794555664e-05, "sentence_fisher_kl_divergence/p85": 7.774830737616867e-05, "sentence_fisher_kl_divergence/p90": 9.689330909168348e-05, "sentence_fisher_kl_divergence/p95": 0.00011332023132126778, "sentence_fisher_kl_divergence/p99": 0.00019082066137343645, "sentence_fisher_kl_divergence/var": 2.0997168448388948e-09, "sentence_full_gradient_variance/max_squared_error": 26004.1875, "sentence_full_gradient_variance/metric": 26004.1875, "sentence_full_gradient_variance/p75": 26004.1875, "sentence_full_gradient_variance/p90": 26004.1875, "sentence_full_gradient_variance/p95": 26004.1875, "sentence_full_gradient_variance/p99": 26004.1875, "sentence_full_update_term": 0.2626838684082031, "sentence_full_update_term/max": 1.375, "sentence_full_update_term/median": 0.10009765625, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.03192138671875, "sentence_full_update_term/p75": 0.37353515625, "sentence_full_update_term/p85": 0.591015636920929, "sentence_full_update_term/p90": 0.7160153388977051, "sentence_full_update_term/p95": 0.998632550239563, "sentence_full_update_term/p99": 1.2021872997283936, "sentence_full_update_term/var": 0.1063842922449112, "sentence_hessian_coeff": -150347.609375, "sentence_hessian_coeff/max": 5734400.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": -4456448.0, "sentence_hessian_coeff/p25": -171264.0, "sentence_hessian_coeff/p75": 89728.0, "sentence_hessian_coeff/p99": 2479141.25, "sentence_hessian_coeff/var": 1359086616576.0, "sentence_hessian_coeff_abs": 591747.625, "sentence_hessian_coeff_abs/max": 5734400.0, "sentence_hessian_coeff_abs/median": 113664.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 19904.0, "sentence_hessian_coeff_abs/p75": 820224.0, "sentence_hessian_coeff_abs/p99": 4724817.0, "sentence_hessian_coeff_abs/var": 1027379494912.0, "step": 8, "token_fisher_curvature": 70213984.0, "token_fisher_curvature/max": 11341398016.0, "token_fisher_curvature/median": 2.5704503059387207e-07, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 3.7692966152816365e-20, "token_fisher_curvature/p75": 679936.0, "token_fisher_curvature/p85": 57409536.0, "token_fisher_curvature/p90": 196083712.0, "token_fisher_curvature/p95": 501219328.0, "token_fisher_curvature/p99": 977272832.0, "token_fisher_curvature/var": 6.345048816667853e+16, "token_fisher_kl_divergence": 0.0038707866333425045, "token_fisher_kl_divergence/max": 0.625, "token_fisher_kl_divergence/median": 1.4203048459560108e-17, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 2.0830858278492343e-30, "token_fisher_kl_divergence/p75": 3.743171691894531e-05, "token_fisher_kl_divergence/p85": 0.0031585693359375, "token_fisher_kl_divergence/p90": 0.01080322265625, "token_fisher_kl_divergence/p95": 0.027587890625, "token_fisher_kl_divergence/p99": 0.053955078125, "token_fisher_kl_divergence/var": 0.00019285046437289566, "token_full_update_term": 0.038133442401885986, "token_full_update_term/max": 4.28125, "token_full_update_term/median": 3.841705620288849e-09, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 1.4502288259166107e-15, "token_full_update_term/p75": 0.005950927734375, "token_full_update_term/p85": 0.054443359375, "token_full_update_term/p90": 0.1337890625, "token_full_update_term/p95": 0.26171875, "token_full_update_term/p99": 0.439453125, "token_full_update_term/var": 0.013049041852355003, "token_hessian_coeff": -49652524.0, "token_hessian_coeff/max": 27783069696.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": -7549747200.0, "token_hessian_coeff/p25": -0.1474609375, "token_hessian_coeff/p75": 6.631016731262207e-06, "token_hessian_coeff/p99": 262144000.0, "token_hessian_coeff/var": 1.6795756514823373e+17, "token_hessian_coeff_abs": 83615976.0, "token_hessian_coeff_abs/max": 27783069696.0, "token_hessian_coeff_abs/median": 0.000797271728515625, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 1.5631940186722204e-13, "token_hessian_coeff_abs/p75": 950272.0, "token_hessian_coeff_abs/p99": 1417674752.0, "token_hessian_coeff_abs/var": 1.634312365539328e+17 }, { "accuracy_reward": 0.4124999940395355, "accuracy_reward/correct": 1.0, "accuracy_reward/correct/max": 1.0, "accuracy_reward/correct/median": 1.0, "accuracy_reward/correct/min": 1.0, "accuracy_reward/correct/p25": 1.0, "accuracy_reward/correct/p75": 1.0, "accuracy_reward/correct/var": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 1.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 1.0, "accuracy_reward/var": 0.24541139602661133, "adam_stats/lm_head/lr_effective_max": 8.770824933890253e-05, "adam_stats/lm_head/lr_effective_mean": -2.836348100831998e-10, "adam_stats/lm_head/lr_effective_min": -8.748142136028036e-05, "adam_stats/lm_head/lr_effective_std": 5.596394657914061e-06, "adam_stats/lr_effective_max": 8.948147296905518e-05, "adam_stats/lr_effective_mean": 6.16877604553423e-10, "adam_stats/lr_effective_min": -8.909030293580145e-05, "adam_stats/m_t_max": 0.006040527485311031, "adam_stats/m_t_mean": -1.0957609819506331e-10, "adam_stats/m_t_min": -0.007101481314748526, "adam_stats/v_t_max": 1.657758912188001e-05, "adam_stats/v_t_mean": 1.9379500174748765e-12, "adam_stats/v_t_min": 0.0, "advantages": 2.9802322831784522e-09, "advantages/max": 2.4741740226745605, "advantages/median": 0.0, "advantages/min": -2.4741740226745605, "advantages/p25": -0.5860660076141357, "advantages/p75": 0.5860660076141357, "advantages/var": 0.7971175312995911, "all_logprobs": -1.1177340745925903, "all_logprobs/max": 0.0, "all_logprobs/median": -0.0067138671875, "all_logprobs/min": -14.375, "all_logprobs/p1": -7.21875, "all_logprobs/p10": -5.5625, "all_logprobs/p25": -0.73046875, "all_logprobs/p5": -6.53125, "all_logprobs/p75": -3.170967102050781e-05, "all_logprobs/var": 4.609470844268799, "clip_ratio": 0.0, "completion_length": 762.6375122070312, "completion_length/correct": 743.45458984375, "completion_length/correct/max": 1024.0, "completion_length/correct/median": 679.0, "completion_length/correct/min": 307.0, "completion_length/correct/p25": 492.0, "completion_length/correct/p75": 1024.0, "completion_length/correct/var": 65021.19921875, "completion_length/incorrect": 776.1063842773438, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 754.0, "completion_length/incorrect/min": 384.0, "completion_length/incorrect/p25": 539.5, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 56961.18359375, "completion_length/max": 1024.0, "completion_length/median": 747.0, "completion_length/min": 307.0, "completion_length/p25": 526.25, "completion_length/p75": 1024.0, "completion_length/var": 59766.61328125, "epoch": 0.012, "feature_vector_variance/max_squared_error": 25328.041015625, "feature_vector_variance/metric": 17638.310546875, "generated_tokens/total": 410127.0, "global_fisher_curvature": 21888.0, "global_fisher_curvature/max": 21888.0, "global_fisher_curvature/median": 21888.0, "global_fisher_curvature/min": 21888.0, "global_fisher_curvature/p25": 21888.0, "global_fisher_curvature/p75": 21888.0, "global_fisher_curvature/p85": 21888.0, "global_fisher_curvature/p90": 21888.0, "global_fisher_curvature/p95": 21888.0, "global_fisher_curvature/p99": 21888.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 1.5795230865478516e-06, "global_fisher_kl_divergence/max": 1.5795230865478516e-06, "global_fisher_kl_divergence/median": 1.5795230865478516e-06, "global_fisher_kl_divergence/min": 1.5795230865478516e-06, "global_fisher_kl_divergence/p25": 1.5795230865478516e-06, "global_fisher_kl_divergence/p75": 1.5795230865478516e-06, "global_fisher_kl_divergence/p85": 1.5795230865478516e-06, "global_fisher_kl_divergence/p90": 1.5795230865478516e-06, "global_fisher_kl_divergence/p95": 1.5795230865478516e-06, "global_fisher_kl_divergence/p99": 1.5795230865478516e-06, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.82421875, "global_full_update_term/max": 0.82421875, "global_full_update_term/median": 0.82421875, "global_full_update_term/min": 0.82421875, "global_full_update_term/p25": 0.82421875, "global_full_update_term/p75": 0.82421875, "global_full_update_term/p85": 0.82421875, "global_full_update_term/p90": 0.82421875, "global_full_update_term/p95": 0.82421875, "global_full_update_term/p99": 0.82421875, "global_full_update_term/var": NaN, "global_hessian_coeff": -3152.0, "global_hessian_coeff/max": -3152.0, "global_hessian_coeff/median": -3152.0, "global_hessian_coeff/min": -3152.0, "global_hessian_coeff/p25": -3152.0, "global_hessian_coeff/p75": -3152.0, "global_hessian_coeff/p99": -3152.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 3152.0, "global_hessian_coeff_abs/max": 3152.0, "global_hessian_coeff_abs/median": 3152.0, "global_hessian_coeff_abs/min": 3152.0, "global_hessian_coeff_abs/p25": 3152.0, "global_hessian_coeff_abs/p75": 3152.0, "global_hessian_coeff_abs/p99": 3152.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 5.978602886199951, "grouped_std_rewards": 0.41830071806907654, "learning_rate": 1.3500000000000001e-05, "loss": 0.0, "mean_logprobs": -1.09375, "mean_logprobs/var": 1.0703125, "num_completions/total": 720, "per_sentence_gradient_norm": 177.4406280517578, "per_sentence_gradient_norm/max": 808.0, "per_sentence_gradient_norm/median": 169.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 94.25, "per_sentence_gradient_norm/p75": 231.0, "per_sentence_gradient_norm/var": 18671.529296875, "per_token_feature_norm": 141.26901245117188, "per_token_feature_norm/max": 151.0, "per_token_feature_norm/median": 144.0, "per_token_feature_norm/min": 81.0, "per_token_feature_norm/p25": 137.0, "per_token_feature_norm/p75": 147.0, "per_token_feature_norm/var": 57.62664794921875, "per_token_gradient_norm": 27.130184173583984, "per_token_gradient_norm/max": 512.0, "per_token_gradient_norm/median": 0.06982421875, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.00010967254638671875, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 35.5, "per_token_gradient_norm/var": 2979.458984375, "per_token_policy_error_norm": 0.2686036229133606, "per_token_policy_error_norm/max": 2.0, "per_token_policy_error_norm/median": 0.0, "per_token_policy_error_norm/min": 0.0, "per_token_policy_error_norm/p25": 0.0, "per_token_policy_error_norm/p75": 0.44140625, "per_token_policy_error_norm/var": 0.19393374025821686, "policy_entropy": 0.8168915510177612, "policy_entropy/max": 3.921875, "policy_entropy/median": 0.044677734375, "policy_entropy/min": 1.5234036254696548e-11, "policy_entropy/p25": 0.000396728515625, "policy_entropy/p75": 0.984375, "policy_entropy/var": 1.7003840208053589, "policy_loss": 5.9604645663569045e-09, "policy_loss/max": 2.4741742610931396, "policy_loss/median": 0.0, "policy_loss/min": -2.4741742610931396, "policy_loss/p25": -0.5860660076141357, "policy_loss/p75": 0.5860660076141357, "policy_loss/var": 0.7971174716949463, "policy_sharpness": 4.55228853225708, "policy_sharpness/max": 10.0, "policy_sharpness/median": 2.820251226425171, "policy_sharpness/min": -0.0, "policy_sharpness/p25": 0.2431628406047821, "policy_sharpness/p75": 10.0, "policy_sharpness/var": 18.192394256591797, "reward": 0.4124999940395355, "reward/max": 1.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 1.0, "reward/var": 0.24541139602661133, "rewards/accuracy_reward": 0.4124999940395355, "rewards/accuracy_reward/max": 1.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 1.0, "rewards/accuracy_reward/var": 0.24541139602661133, "sentence_fisher_curvature": 1164809.625, "sentence_fisher_curvature/max": 10878976.0, "sentence_fisher_curvature/median": 520192.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 191488.0, "sentence_fisher_curvature/p75": 1572864.0, "sentence_fisher_curvature/p85": 1948467.5, "sentence_fisher_curvature/p90": 2396978.0, "sentence_fisher_curvature/p95": 3558604.5, "sentence_fisher_curvature/p99": 8963356.0, "sentence_fisher_curvature/var": 3052639617024.0, "sentence_fisher_kl_divergence": 8.387929119635373e-05, "sentence_fisher_kl_divergence/max": 0.000782012939453125, "sentence_fisher_kl_divergence/median": 3.743171691894531e-05, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 1.3768672943115234e-05, "sentence_fisher_kl_divergence/p75": 0.00011301040649414062, "sentence_fisher_kl_divergence/p85": 0.00013999940711073577, "sentence_fisher_kl_divergence/p90": 0.00017290108371526003, "sentence_fisher_kl_divergence/p95": 0.0002558708074502647, "sentence_fisher_kl_divergence/p99": 0.0006464002653956413, "sentence_fisher_kl_divergence/var": 1.5828751642743555e-08, "sentence_full_gradient_variance/max_squared_error": 49291.30859375, "sentence_full_gradient_variance/metric": 49291.30859375, "sentence_full_gradient_variance/p75": 49291.30859375, "sentence_full_gradient_variance/p90": 49291.30859375, "sentence_full_gradient_variance/p95": 49291.30859375, "sentence_full_gradient_variance/p99": 49291.30859375, "sentence_full_update_term": 0.5739395022392273, "sentence_full_update_term/max": 7.46875, "sentence_full_update_term/median": 0.322265625, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.103271484375, "sentence_full_update_term/p75": 0.625, "sentence_full_update_term/p85": 0.8140627145767212, "sentence_full_update_term/p90": 1.0796873569488525, "sentence_full_update_term/p95": 1.8933570384979248, "sentence_full_update_term/p99": 5.197497367858887, "sentence_full_update_term/var": 1.074422836303711, "sentence_hessian_coeff": -278940.40625, "sentence_hessian_coeff/max": 6750208.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": -24379392.0, "sentence_hessian_coeff/p25": -355840.0, "sentence_hessian_coeff/p75": 233216.0, "sentence_hessian_coeff/p99": 5714738.0, "sentence_hessian_coeff/var": 11605904457728.0, "sentence_hessian_coeff_abs": 1246021.25, "sentence_hessian_coeff_abs/max": 24379392.0, "sentence_hessian_coeff_abs/median": 288768.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 65216.0, "sentence_hessian_coeff_abs/p75": 1126400.0, "sentence_hessian_coeff_abs/p99": 15111936.0, "sentence_hessian_coeff_abs/var": 10112475332608.0, "step": 9, "token_fisher_curvature": 64524484.0, "token_fisher_curvature/max": 11140071424.0, "token_fisher_curvature/median": 4.38690185546875e-05, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 2.498001805406602e-16, "token_fisher_curvature/p75": 3817472.0, "token_fisher_curvature/p85": 72351744.0, "token_fisher_curvature/p90": 115343360.0, "token_fisher_curvature/p95": 314572800.0, "token_fisher_curvature/p99": 1476395008.0, "token_fisher_curvature/var": 7.399247680477594e+16, "token_fisher_kl_divergence": 0.0046470798552036285, "token_fisher_kl_divergence/max": 0.80078125, "token_fisher_kl_divergence/median": 3.164135620181696e-15, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 1.7973406864555533e-26, "token_fisher_kl_divergence/p75": 0.000274658203125, "token_fisher_kl_divergence/p85": 0.005218505859375, "token_fisher_kl_divergence/p90": 0.00830078125, "token_fisher_kl_divergence/p95": 0.02264404296875, "token_fisher_kl_divergence/p99": 0.1064453125, "token_fisher_kl_divergence/var": 0.0003838377888314426, "token_full_update_term": 0.04257186874747276, "token_full_update_term/max": 4.375, "token_full_update_term/median": 5.820766091346741e-08, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 1.4432899320127035e-13, "token_full_update_term/p75": 0.0152587890625, "token_full_update_term/p85": 0.05419921875, "token_full_update_term/p90": 0.0908203125, "token_full_update_term/p95": 0.22314453125, "token_full_update_term/p99": 0.875, "token_full_update_term/var": 0.02211654931306839, "token_hessian_coeff": -27905614.0, "token_hessian_coeff/max": 21743271936.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": -26977763328.0, "token_hessian_coeff/p25": -15.875, "token_hessian_coeff/p75": 0.00141143798828125, "token_hessian_coeff/p99": 968884224.0, "token_hessian_coeff/var": 2.926089063314227e+17, "token_hessian_coeff_abs": 88677080.0, "token_hessian_coeff_abs/max": 26977763328.0, "token_hessian_coeff_abs/median": 0.040283203125, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 1.4097167877480388e-10, "token_hessian_coeff_abs/p75": 2392064.0, "token_hessian_coeff_abs/p99": 3019898880.0, "token_hessian_coeff_abs/var": 2.8552389392020275e+17 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 8.775201422395185e-05, "adam_stats/lm_head/lr_effective_mean": -2.8375582439288394e-10, "adam_stats/lm_head/lr_effective_min": -8.752512076171115e-05, "adam_stats/lm_head/lr_effective_std": 5.598900770564796e-06, "adam_stats/lr_effective_max": 8.95262201083824e-05, "adam_stats/lr_effective_mean": 6.171871902438397e-10, "adam_stats/lr_effective_min": -8.913453348213807e-05, "adam_stats/m_t_max": 0.005436474457383156, "adam_stats/m_t_mean": -9.861850919223869e-11, "adam_stats/m_t_min": -0.006391332950443029, "adam_stats/v_t_max": 1.6561010852456093e-05, "adam_stats/v_t_mean": 1.9360114639904724e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -7.231379985809326, "all_logprobs/max": 0.0, "all_logprobs/median": -7.34375, "all_logprobs/min": -8.5625, "all_logprobs/p1": -8.125, "all_logprobs/p10": -7.84375, "all_logprobs/p25": -7.625, "all_logprobs/p5": -7.9375, "all_logprobs/p75": -7.0625, "all_logprobs/var": 0.9089216589927673, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.013333333333333334, "feature_vector_variance/max_squared_error": 38910.4140625, "feature_vector_variance/metric": 1158.8470458984375, "generated_tokens/total": 492047.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.5e-05, "loss": 0.0, "mean_logprobs": -7.21875, "mean_logprobs/var": 0.15234375, "num_completions/total": 800, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 125.302490234375, "per_token_feature_norm/max": 151.0, "per_token_feature_norm/median": 125.5, "per_token_feature_norm/min": 111.5, "per_token_feature_norm/p25": 124.5, "per_token_feature_norm/p75": 125.5, "per_token_feature_norm/var": 7.18102502822876, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0090538263320923, "per_token_policy_error_norm/max": 1.8125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 0.0, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 0.013394422829151154, "policy_entropy": 3.8113491535186768, "policy_entropy/max": 3.921875, "policy_entropy/median": 3.859375, "policy_entropy/min": 9.1313268058002e-10, "policy_entropy/p25": 3.859375, "policy_entropy/p75": 3.875, "policy_entropy/var": 0.19471031427383423, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.10548378527164459, "policy_sharpness/max": 10.0, "policy_sharpness/median": 0.018505403771996498, "policy_sharpness/min": -0.0, "policy_sharpness/p25": 0.016265006735920906, "policy_sharpness/p75": 0.02064225636422634, "policy_sharpness/var": 0.7519396543502808, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 10, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 7.899217598605901e-05, "adam_stats/lm_head/lr_effective_mean": -2.554114697517207e-10, "adam_stats/lm_head/lr_effective_min": -7.878795440774411e-05, "adam_stats/lm_head/lr_effective_std": 5.039732059231028e-06, "adam_stats/lr_effective_max": 8.058933599386364e-05, "adam_stats/lr_effective_mean": 5.555789717170967e-10, "adam_stats/lr_effective_min": -8.023647387744859e-05, "adam_stats/m_t_max": 0.00489282701164484, "adam_stats/m_t_mean": -8.875660623131054e-11, "adam_stats/m_t_min": -0.005752199329435825, "adam_stats/v_t_max": 1.654445077292621e-05, "adam_stats/v_t_mean": 1.9340768136338893e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -7.727969646453857, "all_logprobs/max": 0.0, "all_logprobs/median": -7.75, "all_logprobs/min": -8.8125, "all_logprobs/p1": -8.375, "all_logprobs/p10": -8.0625, "all_logprobs/p25": -7.9375, "all_logprobs/p5": -8.1875, "all_logprobs/p75": -7.5625, "all_logprobs/var": 0.15204206109046936, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.014666666666666666, "feature_vector_variance/max_squared_error": 36554.9375, "feature_vector_variance/metric": 228.84056091308594, "generated_tokens/total": 573967.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.4995431202643219e-05, "loss": 0.0, "mean_logprobs": -7.71875, "mean_logprobs/var": 0.0022430419921875, "num_completions/total": 880, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 124.78032684326172, "per_token_feature_norm/max": 149.0, "per_token_feature_norm/median": 125.0, "per_token_feature_norm/min": 122.0, "per_token_feature_norm/p25": 124.5, "per_token_feature_norm/p75": 125.0, "per_token_feature_norm/var": 0.5776880979537964, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.022491216659546, "per_token_policy_error_norm/max": 1.9375, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 0.0, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 0.0009322454570792615, "policy_entropy": 3.8823132514953613, "policy_entropy/max": 3.921875, "policy_entropy/median": 3.890625, "policy_entropy/min": 4.7031790018081665e-08, "policy_entropy/p25": 3.875, "policy_entropy/p75": 3.890625, "policy_entropy/var": 0.014464255422353745, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.01647840067744255, "policy_sharpness/max": 10.0, "policy_sharpness/median": 0.013500149361789227, "policy_sharpness/min": 0.006967287044972181, "policy_sharpness/p25": 0.01256677694618702, "policy_sharpness/p75": 0.014522808603942394, "policy_sharpness/var": 0.021470708772540092, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 11, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 7.10634485585615e-05, "adam_stats/lm_head/lr_effective_mean": -2.2975817082215855e-10, "adam_stats/lm_head/lr_effective_min": -7.087975245667621e-05, "adam_stats/lm_head/lr_effective_std": 4.533644187176833e-06, "adam_stats/lr_effective_max": 7.25003847037442e-05, "adam_stats/lr_effective_mean": 4.998153557700391e-10, "adam_stats/lr_effective_min": -7.218267273856327e-05, "adam_stats/m_t_max": 0.004403544124215841, "adam_stats/m_t_mean": -7.988098377209596e-11, "adam_stats/m_t_min": -0.00517697911709547, "adam_stats/v_t_max": 1.6527907064300962e-05, "adam_stats/v_t_mean": 1.9321423801177406e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -7.842897891998291, "all_logprobs/max": -6.78125, "all_logprobs/median": -7.875, "all_logprobs/min": -8.875, "all_logprobs/p1": -8.4375, "all_logprobs/p10": -8.1875, "all_logprobs/p25": -8.0625, "all_logprobs/p5": -8.25, "all_logprobs/p75": -7.65625, "all_logprobs/var": 0.09006453305482864, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.016, "feature_vector_variance/max_squared_error": 2655.00732421875, "feature_vector_variance/metric": 87.36547088623047, "generated_tokens/total": 655887.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.4981730376948682e-05, "loss": 0.0, "mean_logprobs": -7.84375, "mean_logprobs/var": 0.0003452301025390625, "num_completions/total": 960, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 124.7724838256836, "per_token_feature_norm/max": 126.5, "per_token_feature_norm/median": 125.0, "per_token_feature_norm/min": 122.5, "per_token_feature_norm/p25": 124.5, "per_token_feature_norm/p75": 125.0, "per_token_feature_norm/var": 0.25738289952278137, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0234334468841553, "per_token_policy_error_norm/max": 1.0234375, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 3.2021048923525086e-08, "policy_entropy": 3.8772342205047607, "policy_entropy/max": 3.921875, "policy_entropy/median": 3.875, "policy_entropy/min": 3.84375, "policy_entropy/p25": 3.875, "policy_entropy/p75": 3.890625, "policy_entropy/var": 0.00011521411943249404, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.015646211802959442, "policy_sharpness/max": 0.02088160440325737, "policy_sharpness/median": 0.015670768916606903, "policy_sharpness/min": 0.008689976297318935, "policy_sharpness/p25": 0.014834933914244175, "policy_sharpness/p75": 0.016474340111017227, "policy_sharpness/var": 1.4342933809530223e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 12, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 6.389156624209136e-05, "adam_stats/lm_head/lr_effective_mean": -2.0655546717485151e-10, "adam_stats/lm_head/lr_effective_min": -6.372643838403746e-05, "adam_stats/lm_head/lr_effective_std": 4.075890501553658e-06, "adam_stats/lr_effective_max": 6.518354348372668e-05, "adam_stats/lr_effective_mean": 4.4937475962569806e-10, "adam_stats/lr_effective_min": -6.489767110906541e-05, "adam_stats/m_t_max": 0.003963189665228128, "adam_stats/m_t_mean": -7.189285972097892e-11, "adam_stats/m_t_min": -0.004659281112253666, "adam_stats/v_t_max": 1.6511379726580344e-05, "adam_stats/v_t_mean": 1.9302107655272405e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -7.897039890289307, "all_logprobs/max": -6.96875, "all_logprobs/median": -7.9375, "all_logprobs/min": -8.8125, "all_logprobs/p1": -8.5, "all_logprobs/p10": -8.3125, "all_logprobs/p25": -8.125, "all_logprobs/p5": -8.375, "all_logprobs/p75": -7.6875, "all_logprobs/var": 0.10149098187685013, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.017333333333333333, "feature_vector_variance/max_squared_error": 1118.609130859375, "feature_vector_variance/metric": 56.63972854614258, "generated_tokens/total": 737807.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.495891421526205e-05, "loss": 0.0, "mean_logprobs": -7.90625, "mean_logprobs/var": 0.000316619873046875, "num_completions/total": 1040, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 125.06697082519531, "per_token_feature_norm/max": 126.5, "per_token_feature_norm/median": 125.0, "per_token_feature_norm/min": 124.0, "per_token_feature_norm/p25": 125.0, "per_token_feature_norm/p75": 125.5, "per_token_feature_norm/var": 0.1525765359401703, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0234375, "per_token_policy_error_norm/max": 1.0234375, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.0234375, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 0.0, "policy_entropy": 3.8704583644866943, "policy_entropy/max": 3.90625, "policy_entropy/median": 3.875, "policy_entropy/min": 3.84375, "policy_entropy/p25": 3.859375, "policy_entropy/p75": 3.875, "policy_entropy/var": 9.60149263846688e-05, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.01782369427382946, "policy_sharpness/max": 0.020763056352734566, "policy_sharpness/median": 0.017875149846076965, "policy_sharpness/min": 0.010449429973959923, "policy_sharpness/p25": 0.01727265864610672, "policy_sharpness/p75": 0.018449129536747932, "policy_sharpness/var": 8.230691719290917e-07, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 13, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 5.740841515944339e-05, "adam_stats/lm_head/lr_effective_mean": -1.855826464725041e-10, "adam_stats/lm_head/lr_effective_min": -5.726006565964781e-05, "adam_stats/lm_head/lr_effective_std": 3.662117023850442e-06, "adam_stats/lr_effective_max": 5.8569348766468465e-05, "adam_stats/lr_effective_mean": 4.037777612264648e-10, "adam_stats/lr_effective_min": -5.831228190800175e-05, "adam_stats/m_t_max": 0.0035668706987053156, "adam_stats/m_t_mean": -6.470359664723091e-11, "adam_stats/m_t_min": -0.004193353001028299, "adam_stats/v_t_max": 1.649486875976436e-05, "adam_stats/v_t_mean": 1.9282798014580438e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -7.928572177886963, "all_logprobs/max": -7.03125, "all_logprobs/median": -7.96875, "all_logprobs/min": -8.8125, "all_logprobs/p1": -8.5, "all_logprobs/p10": -8.3125, "all_logprobs/p25": -8.1875, "all_logprobs/p5": -8.375, "all_logprobs/p75": -7.6875, "all_logprobs/var": 0.10576073080301285, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.018666666666666668, "feature_vector_variance/max_squared_error": 357.3382568359375, "feature_vector_variance/metric": 39.297935485839844, "generated_tokens/total": 819727.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.4927010515561777e-05, "loss": 0.0, "mean_logprobs": -7.9375, "mean_logprobs/var": 0.00032806396484375, "num_completions/total": 1120, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 125.63748931884766, "per_token_feature_norm/max": 126.5, "per_token_feature_norm/median": 125.5, "per_token_feature_norm/min": 124.5, "per_token_feature_norm/p25": 125.5, "per_token_feature_norm/p75": 126.0, "per_token_feature_norm/var": 0.07189968973398209, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0234375, "per_token_policy_error_norm/max": 1.0234375, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.0234375, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 0.0, "policy_entropy": 3.8673133850097656, "policy_entropy/max": 3.890625, "policy_entropy/median": 3.875, "policy_entropy/min": 3.84375, "policy_entropy/p25": 3.859375, "policy_entropy/p75": 3.875, "policy_entropy/var": 8.693650306668133e-05, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.018725410103797913, "policy_sharpness/max": 0.021527612581849098, "policy_sharpness/median": 0.018796516582369804, "policy_sharpness/min": 0.012336434796452522, "policy_sharpness/p25": 0.018298320472240448, "policy_sharpness/p75": 0.019244447350502014, "policy_sharpness/var": 6.887941594868607e-07, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 14, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 5.155153485247865e-05, "adam_stats/lm_head/lr_effective_mean": -1.666372179132125e-10, "adam_stats/lm_head/lr_effective_min": -5.141834117239341e-05, "adam_stats/lm_head/lr_effective_std": 3.2883347103052074e-06, "adam_stats/lr_effective_max": 5.2594084991142154e-05, "adam_stats/lr_effective_mean": 3.6258557267743186e-10, "adam_stats/lr_effective_min": -5.2363055146997795e-05, "adam_stats/m_t_max": 0.0032101834658533335, "adam_stats/m_t_mean": -5.823318216524598e-11, "adam_stats/m_t_min": -0.003774017561227083, "adam_stats/v_t_max": 1.6478374163853005e-05, "adam_stats/v_t_mean": 1.926352306835799e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -7.942009925842285, "all_logprobs/max": -7.03125, "all_logprobs/median": -7.96875, "all_logprobs/min": -8.75, "all_logprobs/p1": -8.5, "all_logprobs/p10": -8.3125, "all_logprobs/p25": -8.1875, "all_logprobs/p5": -8.375, "all_logprobs/p75": -7.71875, "all_logprobs/var": 0.10831429809331894, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.02, "feature_vector_variance/max_squared_error": 287.9678039550781, "feature_vector_variance/metric": 31.875091552734375, "generated_tokens/total": 901647.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.488605814759156e-05, "loss": 0.0, "mean_logprobs": -7.9375, "mean_logprobs/var": 0.00026702880859375, "num_completions/total": 1200, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 126.63665008544922, "per_token_feature_norm/max": 127.5, "per_token_feature_norm/median": 126.5, "per_token_feature_norm/min": 126.0, "per_token_feature_norm/p25": 126.5, "per_token_feature_norm/p75": 127.0, "per_token_feature_norm/var": 0.051403045654296875, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0234375, "per_token_policy_error_norm/max": 1.0234375, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.0234375, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 0.0, "policy_entropy": 3.8654191493988037, "policy_entropy/max": 3.90625, "policy_entropy/median": 3.859375, "policy_entropy/min": 3.828125, "policy_entropy/p25": 3.859375, "policy_entropy/p75": 3.875, "policy_entropy/var": 8.62446177052334e-05, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.018949735909700394, "policy_sharpness/max": 0.022062573581933975, "policy_sharpness/median": 0.019103391095995903, "policy_sharpness/min": 0.011222238652408123, "policy_sharpness/p25": 0.018313325941562653, "policy_sharpness/p75": 0.0196481104940176, "policy_sharpness/var": 1.063846184479189e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 15, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 4.6263776312116534e-05, "adam_stats/lm_head/lr_effective_mean": -1.4953399629646924e-10, "adam_stats/lm_head/lr_effective_min": -4.614425779436715e-05, "adam_stats/lm_head/lr_effective_std": 2.9508912575693103e-06, "adam_stats/lr_effective_max": 4.7199435357470065e-05, "adam_stats/lr_effective_mean": 3.253954050652652e-10, "adam_stats/lr_effective_min": -4.699192868429236e-05, "adam_stats/m_t_max": 0.0028891649562865496, "adam_stats/m_t_mean": -5.240987852039858e-11, "adam_stats/m_t_min": -0.00339661561883986, "adam_stats/v_t_max": 1.6461895938846283e-05, "adam_stats/v_t_mean": 1.9244252458944233e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -7.958939552307129, "all_logprobs/max": -7.09375, "all_logprobs/median": -8.0, "all_logprobs/min": -8.6875, "all_logprobs/p1": -8.4375, "all_logprobs/p10": -8.3125, "all_logprobs/p25": -8.25, "all_logprobs/p5": -8.375, "all_logprobs/p75": -7.71875, "all_logprobs/var": 0.1044360026717186, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.021333333333333333, "feature_vector_variance/max_squared_error": 234.59645080566406, "feature_vector_variance/metric": 28.19806480407715, "generated_tokens/total": 983567.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.4836107005503543e-05, "loss": 0.0, "mean_logprobs": -7.96875, "mean_logprobs/var": 0.000301361083984375, "num_completions/total": 1280, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 127.57856750488281, "per_token_feature_norm/max": 128.0, "per_token_feature_norm/median": 127.5, "per_token_feature_norm/min": 127.0, "per_token_feature_norm/p25": 127.5, "per_token_feature_norm/p75": 127.5, "per_token_feature_norm/var": 0.03387927636504173, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0234375, "per_token_policy_error_norm/max": 1.0234375, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.0234375, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 0.0, "policy_entropy": 3.866119861602783, "policy_entropy/max": 3.90625, "policy_entropy/median": 3.859375, "policy_entropy/min": 3.84375, "policy_entropy/p25": 3.859375, "policy_entropy/p75": 3.875, "policy_entropy/var": 8.45247705001384e-05, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.018326640129089355, "policy_sharpness/max": 0.021538862958550453, "policy_sharpness/median": 0.018458884209394455, "policy_sharpness/min": 0.009582833386957645, "policy_sharpness/p25": 0.017866898328065872, "policy_sharpness/p75": 0.018945079296827316, "policy_sharpness/var": 9.473572504248295e-07, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 16, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 4.149282176513225e-05, "adam_stats/lm_head/lr_effective_mean": -1.341035477109287e-10, "adam_stats/lm_head/lr_effective_min": -4.138565418543294e-05, "adam_stats/lm_head/lr_effective_std": 2.6464449547347613e-06, "adam_stats/lr_effective_max": 4.233203435433097e-05, "adam_stats/lr_effective_mean": 2.9184013539129694e-10, "adam_stats/lr_effective_min": -4.2145784391323105e-05, "adam_stats/m_t_max": 0.0026002484373748302, "adam_stats/m_t_mean": -4.716887644362622e-11, "adam_stats/m_t_min": -0.0030569538939744234, "adam_stats/v_t_max": 1.6445434084744193e-05, "adam_stats/v_t_mean": 1.922500570197827e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -8.012951850891113, "all_logprobs/max": -3.984375, "all_logprobs/median": -8.0625, "all_logprobs/min": -8.875, "all_logprobs/p1": -8.5625, "all_logprobs/p10": -8.375, "all_logprobs/p25": -8.3125, "all_logprobs/p5": -8.4375, "all_logprobs/p75": -7.78125, "all_logprobs/var": 0.10604716092348099, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.02266666666666667, "feature_vector_variance/max_squared_error": 9799.734375, "feature_vector_variance/metric": 61.41018295288086, "generated_tokens/total": 1065487.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.4777217947069972e-05, "loss": 0.0, "mean_logprobs": -8.0, "mean_logprobs/var": 0.004791259765625, "num_completions/total": 1360, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 128.26129150390625, "per_token_feature_norm/max": 130.0, "per_token_feature_norm/median": 128.0, "per_token_feature_norm/min": 121.0, "per_token_feature_norm/p25": 128.0, "per_token_feature_norm/p75": 129.0, "per_token_feature_norm/var": 0.198994442820549, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0234403610229492, "per_token_policy_error_norm/max": 1.1484375, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.0, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 2.1457260857005167e-07, "policy_entropy": 3.8667073249816895, "policy_entropy/max": 3.90625, "policy_entropy/median": 3.859375, "policy_entropy/min": 3.0625, "policy_entropy/p25": 3.859375, "policy_entropy/p75": 3.875, "policy_entropy/var": 0.00010221919364994392, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.017725765705108643, "policy_sharpness/max": 0.043938323855400085, "policy_sharpness/median": 0.017965184524655342, "policy_sharpness/min": 0.011000149883329868, "policy_sharpness/p25": 0.017117347568273544, "policy_sharpness/p75": 0.018482893705368042, "policy_sharpness/var": 1.2301951528570498e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 17, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 3.719086817000061e-05, "adam_stats/lm_head/lr_effective_mean": -1.2019110706695812e-10, "adam_stats/lm_head/lr_effective_min": -3.709482552949339e-05, "adam_stats/lm_head/lr_effective_std": 2.371941263845656e-06, "adam_stats/lr_effective_max": 3.7943111237837e-05, "adam_stats/lr_effective_mean": 2.6158342159021686e-10, "adam_stats/lr_effective_min": -3.777603342314251e-05, "adam_stats/m_t_max": 0.002340223640203476, "adam_stats/m_t_mean": -4.2451979431756826e-11, "adam_stats/m_t_min": -0.002751258434727788, "adam_stats/v_t_max": 1.6428988601546735e-05, "adam_stats/v_t_mean": 1.9205782797460103e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -8.013799667358398, "all_logprobs/max": -6.84375, "all_logprobs/median": -8.0625, "all_logprobs/min": -8.9375, "all_logprobs/p1": -8.5, "all_logprobs/p10": -8.375, "all_logprobs/p25": -8.3125, "all_logprobs/p5": -8.375, "all_logprobs/p75": -7.78125, "all_logprobs/var": 0.09917321801185608, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.024, "feature_vector_variance/max_squared_error": 2791.044677734375, "feature_vector_variance/metric": 52.66582107543945, "generated_tokens/total": 1147407.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.4709462719537392e-05, "loss": 0.0, "mean_logprobs": -8.0, "mean_logprobs/var": 0.0023040771484375, "num_completions/total": 1440, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 128.9973602294922, "per_token_feature_norm/max": 130.0, "per_token_feature_norm/median": 129.0, "per_token_feature_norm/min": 127.0, "per_token_feature_norm/p25": 129.0, "per_token_feature_norm/p75": 129.0, "per_token_feature_norm/var": 0.01921013928949833, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0234378576278687, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.0430940555750112e-08, "policy_entropy": 3.867819309234619, "policy_entropy/max": 3.90625, "policy_entropy/median": 3.875, "policy_entropy/min": 3.78125, "policy_entropy/p25": 3.859375, "policy_entropy/p75": 3.875, "policy_entropy/var": 9.673376916907728e-05, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.017221957445144653, "policy_sharpness/max": 0.020201079547405243, "policy_sharpness/median": 0.017335684970021248, "policy_sharpness/min": 0.009526560083031654, "policy_sharpness/p25": 0.01658613421022892, "policy_sharpness/p75": 0.017937427386641502, "policy_sharpness/var": 1.0325984476367012e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 18, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 3.3314234315184876e-05, "adam_stats/lm_head/lr_effective_mean": -1.0765500990661536e-10, "adam_stats/lm_head/lr_effective_min": -3.322821430629119e-05, "adam_stats/lm_head/lr_effective_std": 2.1245900825306308e-06, "adam_stats/lr_effective_max": 3.3988097129622474e-05, "adam_stats/lr_effective_mean": 2.343177318842038e-10, "adam_stats/lr_effective_min": -3.38383179041557e-05, "adam_stats/m_t_max": 0.002106201136484742, "adam_stats/m_t_mean": -3.8206802305262855e-11, "adam_stats/m_t_min": -0.0024761324748396873, "adam_stats/v_t_max": 1.641255948925391e-05, "adam_stats/v_t_mean": 1.918657940858104e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -7.963238716125488, "all_logprobs/max": -6.90625, "all_logprobs/median": -8.0, "all_logprobs/min": -8.5625, "all_logprobs/p1": -8.375, "all_logprobs/p10": -8.3125, "all_logprobs/p25": -8.25, "all_logprobs/p5": -8.3125, "all_logprobs/p75": -7.75, "all_logprobs/var": 0.09156106412410736, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.025333333333333333, "feature_vector_variance/max_squared_error": 245.1176300048828, "feature_vector_variance/metric": 30.007482528686523, "generated_tokens/total": 1229327.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.4632923872213653e-05, "loss": 0.0, "mean_logprobs": -7.96875, "mean_logprobs/var": 0.0004291534423828125, "num_completions/total": 1520, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 129.4246826171875, "per_token_feature_norm/max": 130.0, "per_token_feature_norm/median": 129.0, "per_token_feature_norm/min": 128.0, "per_token_feature_norm/p25": 129.0, "per_token_feature_norm/p75": 130.0, "per_token_feature_norm/var": 0.24440084397792816, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0234373807907104, "per_token_policy_error_norm/max": 1.0234375, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 7.450570160827397e-10, "policy_entropy": 3.8700530529022217, "policy_entropy/max": 3.90625, "policy_entropy/median": 3.875, "policy_entropy/min": 3.84375, "policy_entropy/p25": 3.859375, "policy_entropy/p75": 3.875, "policy_entropy/var": 0.0001046869729179889, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.016682013869285583, "policy_sharpness/max": 0.01960534043610096, "policy_sharpness/median": 0.01672418974339962, "policy_sharpness/min": 0.011138956062495708, "policy_sharpness/p25": 0.01620948314666748, "policy_sharpness/p75": 0.017195377498865128, "policy_sharpness/var": 5.858717599949159e-07, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 19, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 2.982305159093812e-05, "adam_stats/lm_head/lr_effective_mean": -9.636626219222677e-11, "adam_stats/lm_head/lr_effective_min": -2.9746057407464832e-05, "adam_stats/lm_head/lr_effective_std": 1.9018450529983966e-06, "adam_stats/lr_effective_max": 3.0426330340560526e-05, "adam_stats/lr_effective_mean": 2.0976315129317413e-10, "adam_stats/lr_effective_min": -3.0292138035292737e-05, "adam_stats/m_t_max": 0.001895581022836268, "adam_stats/m_t_mean": -3.438612936057517e-11, "adam_stats/m_t_min": -0.00222851918078959, "adam_stats/v_t_max": 1.6396146747865714e-05, "adam_stats/v_t_mean": 1.9167393366936736e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -7.91851282119751, "all_logprobs/max": -5.6875, "all_logprobs/median": -8.0, "all_logprobs/min": -8.9375, "all_logprobs/p1": -8.4375, "all_logprobs/p10": -8.25, "all_logprobs/p25": -8.1875, "all_logprobs/p5": -8.3125, "all_logprobs/p75": -7.75, "all_logprobs/var": 0.10135087370872498, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.02666666666666667, "feature_vector_variance/max_squared_error": 3483.492919921875, "feature_vector_variance/metric": 56.0156364440918, "generated_tokens/total": 1311247.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.4547694655894313e-05, "loss": 0.0, "mean_logprobs": -7.90625, "mean_logprobs/var": 0.0037689208984375, "num_completions/total": 1600, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 130.0208740234375, "per_token_feature_norm/max": 131.0, "per_token_feature_norm/median": 130.0, "per_token_feature_norm/min": 126.5, "per_token_feature_norm/p25": 130.0, "per_token_feature_norm/p75": 130.0, "per_token_feature_norm/var": 0.0668836385011673, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0234311819076538, "per_token_policy_error_norm/max": 1.0703125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.944232792538969e-07, "policy_entropy": 3.8688111305236816, "policy_entropy/max": 3.90625, "policy_entropy/median": 3.875, "policy_entropy/min": 3.421875, "policy_entropy/p25": 3.859375, "policy_entropy/p75": 3.875, "policy_entropy/var": 0.0001442609791411087, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.016034049913287163, "policy_sharpness/max": 0.023615321144461632, "policy_sharpness/median": 0.016081932932138443, "policy_sharpness/min": 0.011448079720139503, "policy_sharpness/p25": 0.015565726906061172, "policy_sharpness/p75": 0.016546370461583138, "policy_sharpness/var": 6.546519557559805e-07, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 20, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 2.668096567504108e-05, "adam_stats/lm_head/lr_effective_mean": -8.6207041505304e-11, "adam_stats/lm_head/lr_effective_min": -2.6612095098244026e-05, "adam_stats/lm_head/lr_effective_std": 1.7013837805279763e-06, "adam_stats/lr_effective_max": 2.7220710762776434e-05, "adam_stats/lr_effective_mean": 1.8766373177658835e-10, "adam_stats/lr_effective_min": -2.710056105570402e-05, "adam_stats/m_t_max": 0.0017060228856280446, "adam_stats/m_t_mean": -3.094752579202442e-11, "adam_stats/m_t_min": -0.002005667192861438, "adam_stats/v_t_max": 1.6379750377382152e-05, "adam_stats/v_t_mean": 1.914823334614457e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -7.823601722717285, "all_logprobs/max": -6.5, "all_logprobs/median": -7.90625, "all_logprobs/min": -8.5, "all_logprobs/p1": -8.25, "all_logprobs/p10": -8.125, "all_logprobs/p25": -8.0625, "all_logprobs/p5": -8.1875, "all_logprobs/p75": -7.65625, "all_logprobs/var": 0.10488373786211014, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.028, "feature_vector_variance/max_squared_error": 467.70489501953125, "feature_vector_variance/metric": 32.35719299316406, "generated_tokens/total": 1393167.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.4453878909250906e-05, "loss": 0.0, "mean_logprobs": -7.8125, "mean_logprobs/var": 0.00048828125, "num_completions/total": 1680, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 130.39434814453125, "per_token_feature_norm/max": 132.0, "per_token_feature_norm/median": 130.0, "per_token_feature_norm/min": 129.0, "per_token_feature_norm/p25": 130.0, "per_token_feature_norm/p75": 131.0, "per_token_feature_norm/var": 0.24345459043979645, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0233933925628662, "per_token_policy_error_norm/max": 1.0234375, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 3.4301692153348995e-07, "policy_entropy": 3.866800308227539, "policy_entropy/max": 3.90625, "policy_entropy/median": 3.875, "policy_entropy/min": 3.8125, "policy_entropy/p25": 3.859375, "policy_entropy/p75": 3.875, "policy_entropy/var": 0.00014553137589246035, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.01604730449616909, "policy_sharpness/max": 0.02205207385122776, "policy_sharpness/median": 0.01590110920369625, "policy_sharpness/min": 0.011505852453410625, "policy_sharpness/p25": 0.015286615118384361, "policy_sharpness/p75": 0.016636405140161514, "policy_sharpness/var": 1.1849225529658725e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 21, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 2.3854834580561146e-05, "adam_stats/lm_head/lr_effective_mean": -7.70701349961378e-11, "adam_stats/lm_head/lr_effective_min": -2.3793265427229926e-05, "adam_stats/lm_head/lr_effective_std": 1.5210897572615067e-06, "adam_stats/lr_effective_max": 2.4337430659215897e-05, "adam_stats/lr_effective_mean": 1.677863820104264e-10, "adam_stats/lr_effective_min": -2.4229922928498127e-05, "adam_stats/m_t_max": 0.0015354205388575792, "adam_stats/m_t_mean": -2.7852761763647038e-11, "adam_stats/m_t_min": -0.001805100473575294, "adam_stats/v_t_max": 1.636337037780322e-05, "adam_stats/v_t_mean": 1.91290776621611e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -7.736799716949463, "all_logprobs/max": -6.28125, "all_logprobs/median": -7.84375, "all_logprobs/min": -8.4375, "all_logprobs/p1": -8.1875, "all_logprobs/p10": -8.125, "all_logprobs/p25": -8.0, "all_logprobs/p5": -8.125, "all_logprobs/p75": -7.5625, "all_logprobs/var": 0.12411930412054062, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.029333333333333333, "feature_vector_variance/max_squared_error": 494.19805908203125, "feature_vector_variance/metric": 33.846160888671875, "generated_tokens/total": 1475087.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.4351590932319506e-05, "loss": 0.0, "mean_logprobs": -7.75, "mean_logprobs/var": 0.000614166259765625, "num_completions/total": 1760, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 130.9550323486328, "per_token_feature_norm/max": 132.0, "per_token_feature_norm/median": 131.0, "per_token_feature_norm/min": 129.0, "per_token_feature_norm/p25": 131.0, "per_token_feature_norm/p75": 131.0, "per_token_feature_norm/var": 0.14573395252227783, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.023293375968933, "per_token_policy_error_norm/max": 1.0234375, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.1057605888709077e-06, "policy_entropy": 3.8606202602386475, "policy_entropy/max": 3.890625, "policy_entropy/median": 3.859375, "policy_entropy/min": 3.796875, "policy_entropy/p25": 3.859375, "policy_entropy/p75": 3.875, "policy_entropy/var": 0.00019104905368294567, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.017141375690698624, "policy_sharpness/max": 0.024323228746652603, "policy_sharpness/median": 0.01695753075182438, "policy_sharpness/min": 0.012167616747319698, "policy_sharpness/p25": 0.01593487523496151, "policy_sharpness/p75": 0.018143007531762123, "policy_sharpness/var": 2.5691379050840624e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 22, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 2.1314475816325285e-05, "adam_stats/lm_head/lr_effective_mean": -6.885775977183428e-11, "adam_stats/lm_head/lr_effective_min": -2.1259473214740865e-05, "adam_stats/lm_head/lr_effective_std": 1.3590355365522555e-06, "adam_stats/lr_effective_max": 2.1745705453213304e-05, "adam_stats/lr_effective_mean": 1.499189522524702e-10, "adam_stats/lr_effective_min": -2.16495682252571e-05, "adam_stats/m_t_max": 0.0013818783918395638, "adam_stats/m_t_mean": -2.5067486281171725e-11, "adam_stats/m_t_min": -0.0016245903680101037, "adam_stats/v_t_max": 1.6347006749128923e-05, "adam_stats/v_t_mean": 1.910994583062542e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -7.678443431854248, "all_logprobs/max": -6.15625, "all_logprobs/median": -7.78125, "all_logprobs/min": -8.8125, "all_logprobs/p1": -8.25, "all_logprobs/p10": -8.0625, "all_logprobs/p25": -7.96875, "all_logprobs/p5": -8.125, "all_logprobs/p75": -7.46875, "all_logprobs/var": 0.15004484355449677, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.030666666666666665, "feature_vector_variance/max_squared_error": 2452.583251953125, "feature_vector_variance/metric": 64.67431640625, "generated_tokens/total": 1557007.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.4240955347243754e-05, "loss": 0.0, "mean_logprobs": -7.6875, "mean_logprobs/var": 0.0074462890625, "num_completions/total": 1840, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 131.2245330810547, "per_token_feature_norm/max": 133.0, "per_token_feature_norm/median": 131.0, "per_token_feature_norm/min": 128.0, "per_token_feature_norm/p25": 131.0, "per_token_feature_norm/p75": 131.0, "per_token_feature_norm/var": 0.20690195262432098, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0231860876083374, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.9134452031721594e-06, "policy_entropy": 3.852975606918335, "policy_entropy/max": 3.890625, "policy_entropy/median": 3.859375, "policy_entropy/min": 3.71875, "policy_entropy/p25": 3.84375, "policy_entropy/p75": 3.859375, "policy_entropy/var": 0.00024987253709696233, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.01848132722079754, "policy_sharpness/max": 0.026126200333237648, "policy_sharpness/median": 0.01827055774629116, "policy_sharpness/min": 0.013004202395677567, "policy_sharpness/p25": 0.01696578599512577, "policy_sharpness/p75": 0.01974114589393139, "policy_sharpness/var": 4.069193892064504e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 23, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 1.903243173728697e-05, "adam_stats/lm_head/lr_effective_mean": -6.148098002922708e-11, "adam_stats/lm_head/lr_effective_min": -1.898332266137004e-05, "adam_stats/lm_head/lr_effective_std": 1.213467385241529e-06, "adam_stats/lr_effective_max": 1.941750815603882e-05, "adam_stats/lr_effective_mean": 1.3386834696316186e-10, "adam_stats/lr_effective_min": -1.9331595467519946e-05, "adam_stats/m_t_max": 0.0012436904944479465, "adam_stats/m_t_mean": -2.2560734530552295e-11, "adam_stats/m_t_min": -0.0014621312730014324, "adam_stats/v_t_max": 1.6330659491359256e-05, "adam_stats/v_t_mean": 1.90908313463245e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -7.544188022613525, "all_logprobs/max": -5.9375, "all_logprobs/median": -7.625, "all_logprobs/min": -8.375, "all_logprobs/p1": -8.125, "all_logprobs/p10": -8.0, "all_logprobs/p25": -7.90625, "all_logprobs/p5": -8.0625, "all_logprobs/p75": -7.28125, "all_logprobs/var": 0.174672469496727, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.032, "feature_vector_variance/max_squared_error": 621.6149291992188, "feature_vector_variance/metric": 38.66765213012695, "generated_tokens/total": 1638927.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.4122106946441953e-05, "loss": 0.0, "mean_logprobs": -7.53125, "mean_logprobs/var": 0.001129150390625, "num_completions/total": 1920, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 131.82382202148438, "per_token_feature_norm/max": 133.0, "per_token_feature_norm/median": 132.0, "per_token_feature_norm/min": 130.0, "per_token_feature_norm/p25": 132.0, "per_token_feature_norm/p75": 132.0, "per_token_feature_norm/var": 0.2671683132648468, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0231562852859497, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 2.1747594018961536e-06, "policy_entropy": 3.8412277698516846, "policy_entropy/max": 3.890625, "policy_entropy/median": 3.84375, "policy_entropy/min": 3.75, "policy_entropy/p25": 3.828125, "policy_entropy/p75": 3.859375, "policy_entropy/var": 0.0003557268937584013, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.020849402993917465, "policy_sharpness/max": 0.028182772919535637, "policy_sharpness/median": 0.020700782537460327, "policy_sharpness/min": 0.012456483207643032, "policy_sharpness/p25": 0.01933148130774498, "policy_sharpness/p75": 0.022236647084355354, "policy_sharpness/var": 4.387046828924213e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 24, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 1.69837185239885e-05, "adam_stats/lm_head/lr_effective_mean": -5.485902704327472e-11, "adam_stats/lm_head/lr_effective_min": -1.6939902707235888e-05, "adam_stats/lm_head/lr_effective_std": 1.0827901633092551e-06, "adam_stats/lr_effective_max": 1.7327363821095787e-05, "adam_stats/lr_effective_mean": 1.194588733488544e-10, "adam_stats/lr_effective_min": -1.725063702906482e-05, "adam_stats/m_t_max": 0.001119321445003152, "adam_stats/m_t_mean": -2.0304662465275847e-11, "adam_stats/m_t_min": -0.0013159181689843535, "adam_stats/v_t_max": 1.631432860449422e-05, "adam_stats/v_t_mean": 1.9071749388088755e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -7.455016613006592, "all_logprobs/max": -5.875, "all_logprobs/median": -7.53125, "all_logprobs/min": -8.375, "all_logprobs/p1": -8.0625, "all_logprobs/p10": -7.96875, "all_logprobs/p25": -7.84375, "all_logprobs/p5": -8.0, "all_logprobs/p75": -7.15625, "all_logprobs/var": 0.20046751201152802, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.03333333333333333, "feature_vector_variance/max_squared_error": 634.2110595703125, "feature_vector_variance/metric": 41.3685188293457, "generated_tokens/total": 1720847.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.3995190528383292e-05, "loss": 0.0, "mean_logprobs": -7.46875, "mean_logprobs/var": 0.0014801025390625, "num_completions/total": 2000, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 132.22142028808594, "per_token_feature_norm/max": 134.0, "per_token_feature_norm/median": 132.0, "per_token_feature_norm/min": 130.0, "per_token_feature_norm/p25": 132.0, "per_token_feature_norm/p75": 132.0, "per_token_feature_norm/var": 0.2269372045993805, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0232313871383667, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 2.0124091406614752e-06, "policy_entropy": 3.8300774097442627, "policy_entropy/max": 3.890625, "policy_entropy/median": 3.828125, "policy_entropy/min": 3.734375, "policy_entropy/p25": 3.8125, "policy_entropy/p75": 3.84375, "policy_entropy/var": 0.00039466816815547645, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.022686608135700226, "policy_sharpness/max": 0.029558073729276657, "policy_sharpness/median": 0.022638054564595222, "policy_sharpness/min": 0.012832382693886757, "policy_sharpness/p25": 0.02129201963543892, "policy_sharpness/p75": 0.02400153875350952, "policy_sharpness/var": 4.165754489804385e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 25, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 1.5145641555136535e-05, "adam_stats/lm_head/lr_effective_mean": -4.8918320783020164e-11, "adam_stats/lm_head/lr_effective_min": -1.5106576029211283e-05, "adam_stats/lm_head/lr_effective_std": 9.655548183218343e-07, "adam_stats/lr_effective_max": 1.5452113075298257e-05, "adam_stats/lr_effective_mean": 1.0653071480515308e-10, "adam_stats/lr_effective_min": -1.5383633581222966e-05, "adam_stats/m_t_max": 0.0010073892772197723, "adam_stats/m_t_mean": -1.8274194657497134e-11, "adam_stats/m_t_min": -0.0011843263637274504, "adam_stats/v_t_max": 1.629801408853382e-05, "adam_stats/v_t_mean": 1.905268044027908e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -7.351210117340088, "all_logprobs/max": -5.75, "all_logprobs/median": -7.4375, "all_logprobs/min": -8.3125, "all_logprobs/p1": -8.0, "all_logprobs/p10": -7.90625, "all_logprobs/p25": -7.78125, "all_logprobs/p5": -7.9375, "all_logprobs/p75": -7.0, "all_logprobs/var": 0.22951819002628326, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.034666666666666665, "feature_vector_variance/max_squared_error": 822.9774780273438, "feature_vector_variance/metric": 44.32963180541992, "generated_tokens/total": 1802767.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.3860360721173195e-05, "loss": 0.0, "mean_logprobs": -7.34375, "mean_logprobs/var": 0.00151824951171875, "num_completions/total": 2080, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 132.6728973388672, "per_token_feature_norm/max": 134.0, "per_token_feature_norm/median": 133.0, "per_token_feature_norm/min": 130.0, "per_token_feature_norm/p25": 132.0, "per_token_feature_norm/p75": 133.0, "per_token_feature_norm/var": 0.3398904800415039, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.023445963859558, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.6331288179571857e-06, "policy_entropy": 3.8176071643829346, "policy_entropy/max": 3.890625, "policy_entropy/median": 3.8125, "policy_entropy/min": 3.734375, "policy_entropy/p25": 3.8125, "policy_entropy/p75": 3.828125, "policy_entropy/var": 0.0004168625164311379, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.02445732057094574, "policy_sharpness/max": 0.03008628450334072, "policy_sharpness/median": 0.024371249601244926, "policy_sharpness/min": 0.012915666215121746, "policy_sharpness/p25": 0.023194028064608574, "policy_sharpness/p75": 0.025739796459674835, "policy_sharpness/var": 3.7027764392405516e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 26, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 1.3497590771294199e-05, "adam_stats/lm_head/lr_effective_mean": -4.359219235583467e-11, "adam_stats/lm_head/lr_effective_min": -1.3462780771078542e-05, "adam_stats/lm_head/lr_effective_std": 8.604450840721256e-07, "adam_stats/lr_effective_max": 1.3770728401141241e-05, "adam_stats/lr_effective_mean": 9.493902886070771e-11, "adam_stats/lr_effective_min": -1.3709650374948978e-05, "adam_stats/m_t_max": 0.0009066503262147307, "adam_stats/m_t_mean": -1.6446776232581506e-11, "adam_stats/m_t_min": -0.0010658936807885766, "adam_stats/v_t_max": 1.6281715943478048e-05, "adam_stats/v_t_mean": 1.903362667129982e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -7.288339138031006, "all_logprobs/max": -5.6875, "all_logprobs/median": -7.375, "all_logprobs/min": -8.25, "all_logprobs/p1": -8.0, "all_logprobs/p10": -7.875, "all_logprobs/p25": -7.71875, "all_logprobs/p5": -7.90625, "all_logprobs/p75": -6.9375, "all_logprobs/var": 0.24497288465499878, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.036, "feature_vector_variance/max_squared_error": 840.2994384765625, "feature_vector_variance/metric": 48.50019073486328, "generated_tokens/total": 1884687.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.3717781794162813e-05, "loss": 0.0, "mean_logprobs": -7.28125, "mean_logprobs/var": 0.00445556640625, "num_completions/total": 2160, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 132.98924255371094, "per_token_feature_norm/max": 135.0, "per_token_feature_norm/median": 133.0, "per_token_feature_norm/min": 131.0, "per_token_feature_norm/p25": 133.0, "per_token_feature_norm/p75": 133.0, "per_token_feature_norm/var": 0.3179914355278015, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235522985458374, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.6811218301882036e-06, "policy_entropy": 3.8118278980255127, "policy_entropy/max": 3.890625, "policy_entropy/median": 3.8125, "policy_entropy/min": 3.734375, "policy_entropy/p25": 3.796875, "policy_entropy/p75": 3.828125, "policy_entropy/var": 0.00042007918818853796, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.025262553244829178, "policy_sharpness/max": 0.030916115269064903, "policy_sharpness/median": 0.02524835243821144, "policy_sharpness/min": 0.014143156819045544, "policy_sharpness/p25": 0.024026861414313316, "policy_sharpness/p75": 0.026562124490737915, "policy_sharpness/var": 3.475262474239571e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 27, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 1.202085695695132e-05, "adam_stats/lm_head/lr_effective_mean": -3.8820075626277983e-11, "adam_stats/lm_head/lr_effective_min": -1.1989861377514899e-05, "adam_stats/lm_head/lr_effective_std": 7.6626685086012e-07, "adam_stats/lr_effective_max": 1.2264124961802736e-05, "adam_stats/lr_effective_mean": 8.455237898719048e-11, "adam_stats/lr_effective_min": -1.2209685337438714e-05, "adam_stats/m_t_max": 0.0008159852586686611, "adam_stats/m_t_mean": -1.480210398696613e-11, "adam_stats/m_t_min": -0.0009593042777851224, "adam_stats/v_t_max": 1.626543416932691e-05, "adam_stats/v_t_mean": 1.9014583744342284e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -7.236064434051514, "all_logprobs/max": -5.59375, "all_logprobs/median": -7.3125, "all_logprobs/min": -8.4375, "all_logprobs/p1": -8.125, "all_logprobs/p10": -7.84375, "all_logprobs/p25": -7.6875, "all_logprobs/p5": -7.9375, "all_logprobs/p75": -6.84375, "all_logprobs/var": 0.2817250192165375, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.037333333333333336, "feature_vector_variance/max_squared_error": 2211.006103515625, "feature_vector_variance/metric": 106.8323745727539, "generated_tokens/total": 1966607.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.3567627457812107e-05, "loss": 0.0, "mean_logprobs": -7.25, "mean_logprobs/var": 0.028076171875, "num_completions/total": 2240, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 133.2335205078125, "per_token_feature_norm/max": 135.0, "per_token_feature_norm/median": 133.0, "per_token_feature_norm/min": 129.0, "per_token_feature_norm/p25": 133.0, "per_token_feature_norm/p75": 134.0, "per_token_feature_norm/var": 0.583903431892395, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0236777067184448, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 2.306392389073153e-06, "policy_entropy": 3.806265354156494, "policy_entropy/max": 3.875, "policy_entropy/median": 3.8125, "policy_entropy/min": 3.71875, "policy_entropy/p25": 3.796875, "policy_entropy/p75": 3.828125, "policy_entropy/var": 0.00048449760652147233, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.025886720046401024, "policy_sharpness/max": 0.03137604892253876, "policy_sharpness/median": 0.026327284052968025, "policy_sharpness/min": 0.015773558989167213, "policy_sharpness/p25": 0.0250750333070755, "policy_sharpness/p75": 0.02743041142821312, "policy_sharpness/var": 5.692143076885259e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 28, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 1.0698482583393343e-05, "adam_stats/lm_head/lr_effective_mean": -3.4547090882464104e-11, "adam_stats/lm_head/lr_effective_min": -1.0670899428077973e-05, "adam_stats/lm_head/lr_effective_std": 6.819372515565192e-07, "adam_stats/lr_effective_max": 1.0914999620581511e-05, "adam_stats/lr_effective_mean": 7.525128437047002e-11, "adam_stats/lr_effective_min": -1.0866509910556488e-05, "adam_stats/m_t_max": 0.000734386732801795, "adam_stats/m_t_mean": -1.3321887863682047e-11, "adam_stats/m_t_min": -0.0008633738034404814, "adam_stats/v_t_max": 1.6249168766080402e-05, "adam_stats/v_t_mean": 1.8995577680258613e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -7.125840663909912, "all_logprobs/max": -5.5625, "all_logprobs/median": -7.1875, "all_logprobs/min": -8.25, "all_logprobs/p1": -7.90625, "all_logprobs/p10": -7.75, "all_logprobs/p25": -7.5625, "all_logprobs/p5": -7.8125, "all_logprobs/p75": -6.71875, "all_logprobs/var": 0.2707185447216034, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.03866666666666667, "feature_vector_variance/max_squared_error": 1168.984130859375, "feature_vector_variance/metric": 56.83901596069336, "generated_tokens/total": 2048527.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.3410080652050414e-05, "loss": 0.0, "mean_logprobs": -7.125, "mean_logprobs/var": 0.006500244140625, "num_completions/total": 2320, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 133.7236785888672, "per_token_feature_norm/max": 136.0, "per_token_feature_norm/median": 134.0, "per_token_feature_norm/min": 131.0, "per_token_feature_norm/p25": 133.0, "per_token_feature_norm/p75": 134.0, "per_token_feature_norm/var": 0.4071057438850403, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.023715853691101, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 2.516880385883269e-06, "policy_entropy": 3.800950765609741, "policy_entropy/max": 3.890625, "policy_entropy/median": 3.796875, "policy_entropy/min": 3.71875, "policy_entropy/p25": 3.78125, "policy_entropy/p75": 3.8125, "policy_entropy/var": 0.00039798521902412176, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.026834500953555107, "policy_sharpness/max": 0.03165366128087044, "policy_sharpness/median": 0.02685849368572235, "policy_sharpness/min": 0.0159663874655962, "policy_sharpness/p25": 0.025944627821445465, "policy_sharpness/p75": 0.027866145595908165, "policy_sharpness/var": 2.5586498395568924e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 29, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 9.515089004707988e-06, "adam_stats/lm_head/lr_effective_mean": -3.072351401067763e-11, "adam_stats/lm_head/lr_effective_min": -9.490561751590576e-06, "adam_stats/lm_head/lr_effective_std": 6.064748276912724e-07, "adam_stats/lr_effective_max": 9.707667231850792e-06, "adam_stats/lr_effective_mean": 6.692785908812837e-11, "adam_stats/lr_effective_min": -9.664506251283456e-06, "adam_stats/m_t_max": 0.0006609480478800833, "adam_stats/m_t_mean": -1.198969742932654e-11, "adam_stats/m_t_min": -0.0007770364172756672, "adam_stats/v_t_max": 1.6232919733738527e-05, "adam_stats/v_t_mean": 1.8976582458196667e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -7.062556743621826, "all_logprobs/max": -5.5, "all_logprobs/median": -7.09375, "all_logprobs/min": -8.25, "all_logprobs/p1": -7.84375, "all_logprobs/p10": -7.6875, "all_logprobs/p25": -7.53125, "all_logprobs/p5": -7.75, "all_logprobs/p75": -6.65625, "all_logprobs/var": 0.2736203968524933, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.04, "feature_vector_variance/max_squared_error": 1323.1572265625, "feature_vector_variance/metric": 50.542049407958984, "generated_tokens/total": 2130447.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.3245333323392335e-05, "loss": 0.0, "mean_logprobs": -7.0625, "mean_logprobs/var": 0.0019073486328125, "num_completions/total": 2400, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 134.01026916503906, "per_token_feature_norm/max": 136.0, "per_token_feature_norm/median": 134.0, "per_token_feature_norm/min": 131.0, "per_token_feature_norm/p25": 134.0, "per_token_feature_norm/p75": 134.0, "per_token_feature_norm/var": 0.31617045402526855, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0238069295883179, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 3.176885911670979e-06, "policy_entropy": 3.7966179847717285, "policy_entropy/max": 3.890625, "policy_entropy/median": 3.796875, "policy_entropy/min": 3.71875, "policy_entropy/p25": 3.78125, "policy_entropy/p75": 3.8125, "policy_entropy/var": 0.00037642818642780185, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.027334293350577354, "policy_sharpness/max": 0.03211734816431999, "policy_sharpness/median": 0.027259154245257378, "policy_sharpness/min": 0.015044267289340496, "policy_sharpness/p25": 0.026468336582183838, "policy_sharpness/p75": 0.028204532340168953, "policy_sharpness/var": 1.8150210507883457e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 30, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 8.456757313979324e-06, "adam_stats/lm_head/lr_effective_mean": -2.7304254343607504e-11, "adam_stats/lm_head/lr_effective_min": -8.434961273451336e-06, "adam_stats/lm_head/lr_effective_std": 5.389909460973286e-07, "adam_stats/lr_effective_max": 8.627924216852989e-06, "adam_stats/lr_effective_mean": 5.948389003584964e-11, "adam_stats/lr_effective_min": -8.589533535996452e-06, "adam_stats/m_t_max": 0.0005948532489128411, "adam_stats/m_t_mean": -1.0790722482223458e-11, "adam_stats/m_t_min": -0.0006993327406235039, "adam_stats/v_t_max": 1.6216687072301283e-05, "adam_stats/v_t_mean": 1.8957619762199895e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.992227077484131, "all_logprobs/max": -5.5, "all_logprobs/median": -7.03125, "all_logprobs/min": -8.125, "all_logprobs/p1": -7.84375, "all_logprobs/p10": -7.625, "all_logprobs/p25": -7.46875, "all_logprobs/p5": -7.6875, "all_logprobs/p75": -6.5625, "all_logprobs/var": 0.2801041305065155, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.04133333333333333, "feature_vector_variance/max_squared_error": 1293.361083984375, "feature_vector_variance/metric": 60.47905349731445, "generated_tokens/total": 2212367.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.3073586191080456e-05, "loss": 0.0, "mean_logprobs": -7.0, "mean_logprobs/var": 0.0059814453125, "num_completions/total": 2480, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 134.313232421875, "per_token_feature_norm/max": 136.0, "per_token_feature_norm/median": 134.0, "per_token_feature_norm/min": 131.0, "per_token_feature_norm/p25": 134.0, "per_token_feature_norm/p75": 135.0, "per_token_feature_norm/var": 0.41885802149772644, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0237780809402466, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 3.458091214270098e-06, "policy_entropy": 3.795395612716675, "policy_entropy/max": 3.875, "policy_entropy/median": 3.796875, "policy_entropy/min": 3.71875, "policy_entropy/p25": 3.78125, "policy_entropy/p75": 3.8125, "policy_entropy/var": 0.00039928953628987074, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.027586717158555984, "policy_sharpness/max": 0.03244597837328911, "policy_sharpness/median": 0.027451230213046074, "policy_sharpness/min": 0.016027159988880157, "policy_sharpness/p25": 0.026570377871394157, "policy_sharpness/p75": 0.028659213334321976, "policy_sharpness/var": 2.220700025645783e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 31, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 7.510889190598391e-06, "adam_stats/lm_head/lr_effective_mean": -2.4248578839314305e-11, "adam_stats/lm_head/lr_effective_min": -7.4915337791026104e-06, "adam_stats/lm_head/lr_effective_std": 4.786814429280639e-07, "adam_stats/lr_effective_max": 7.662919415452052e-06, "adam_stats/lr_effective_mean": 5.283094448027548e-11, "adam_stats/lr_effective_min": -7.628795174241532e-06, "adam_stats/m_t_max": 0.0005353679298423231, "adam_stats/m_t_mean": -9.711656825950321e-12, "adam_stats/m_t_min": -0.0006293994374573231, "adam_stats/v_t_max": 1.6200470781768672e-05, "adam_stats/v_t_mean": 1.8938639718968364e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.944955348968506, "all_logprobs/max": -5.46875, "all_logprobs/median": -6.96875, "all_logprobs/min": -8.125, "all_logprobs/p1": -7.78125, "all_logprobs/p10": -7.59375, "all_logprobs/p25": -7.40625, "all_logprobs/p5": -7.65625, "all_logprobs/p75": -6.5, "all_logprobs/var": 0.2847624719142914, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.042666666666666665, "feature_vector_variance/max_squared_error": 1142.3133544921875, "feature_vector_variance/metric": 59.76356887817383, "generated_tokens/total": 2294287.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.2895048502539883e-05, "loss": 0.0, "mean_logprobs": -6.9375, "mean_logprobs/var": 0.004241943359375, "num_completions/total": 2560, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 134.53854370117188, "per_token_feature_norm/max": 136.0, "per_token_feature_norm/median": 135.0, "per_token_feature_norm/min": 131.0, "per_token_feature_norm/p25": 134.0, "per_token_feature_norm/p75": 135.0, "per_token_feature_norm/var": 0.41187360882759094, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0238072872161865, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 4.060235823999392e-06, "policy_entropy": 3.7928483486175537, "policy_entropy/max": 3.875, "policy_entropy/median": 3.796875, "policy_entropy/min": 3.71875, "policy_entropy/p25": 3.78125, "policy_entropy/p75": 3.8125, "policy_entropy/var": 0.0003861950244754553, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.027865424752235413, "policy_sharpness/max": 0.0326305516064167, "policy_sharpness/median": 0.02771683782339096, "policy_sharpness/min": 0.017575779929757118, "policy_sharpness/p25": 0.026871249079704285, "policy_sharpness/p75": 0.0288490392267704, "policy_sharpness/var": 1.935283989951131e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 32, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 6.666086846962571e-06, "adam_stats/lm_head/lr_effective_mean": -2.1519610746145723e-11, "adam_stats/lm_head/lr_effective_min": -6.64891149426694e-06, "adam_stats/lm_head/lr_effective_std": 4.2481894979573553e-07, "adam_stats/lr_effective_max": 6.8010244831384625e-06, "adam_stats/lr_effective_mean": 4.6888860849580993e-11, "adam_stats/lr_effective_min": -6.770713753212476e-06, "adam_stats/m_t_max": 0.0004818311135750264, "adam_stats/m_t_mean": -8.740486633074251e-12, "adam_stats/m_t_min": -0.0005664594937115908, "adam_stats/v_t_max": 1.6184270862140693e-05, "adam_stats/v_t_mean": 1.8919713885845457e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.902143955230713, "all_logprobs/max": -5.46875, "all_logprobs/median": -6.9375, "all_logprobs/min": -8.0625, "all_logprobs/p1": -7.75, "all_logprobs/p10": -7.5625, "all_logprobs/p25": -7.375, "all_logprobs/p5": -7.625, "all_logprobs/p75": -6.4375, "all_logprobs/var": 0.2840985059738159, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.044, "feature_vector_variance/max_squared_error": 1347.928466796875, "feature_vector_variance/metric": 54.27275466918945, "generated_tokens/total": 2376207.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.270993777844248e-05, "loss": 0.0, "mean_logprobs": -6.90625, "mean_logprobs/var": 0.001739501953125, "num_completions/total": 2640, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 134.74041748046875, "per_token_feature_norm/max": 136.0, "per_token_feature_norm/median": 135.0, "per_token_feature_norm/min": 131.0, "per_token_feature_norm/p25": 134.0, "per_token_feature_norm/p75": 135.0, "per_token_feature_norm/var": 0.3783612549304962, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0238310098648071, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 4.492882908380125e-06, "policy_entropy": 3.790802478790283, "policy_entropy/max": 3.875, "policy_entropy/median": 3.796875, "policy_entropy/min": 3.71875, "policy_entropy/p25": 3.78125, "policy_entropy/p75": 3.8125, "policy_entropy/var": 0.0003822858852799982, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.028179896995425224, "policy_sharpness/max": 0.03273409232497215, "policy_sharpness/median": 0.028095737099647522, "policy_sharpness/min": 0.018281061202287674, "policy_sharpness/p25": 0.027191627770662308, "policy_sharpness/p75": 0.029137905687093735, "policy_sharpness/var": 1.7840193322626874e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 33, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 5.912053893553093e-06, "adam_stats/lm_head/lr_effective_mean": -1.908403123029867e-11, "adam_stats/lm_head/lr_effective_min": -5.896823495277204e-06, "adam_stats/lm_head/lr_effective_std": 3.767462146697653e-07, "adam_stats/lr_effective_max": 6.031733846612042e-06, "adam_stats/lr_effective_mean": 4.1585172805280735e-11, "adam_stats/lr_effective_min": -6.004830538586248e-06, "adam_stats/m_t_max": 0.0004336480051279068, "adam_stats/m_t_mean": -7.86644100553291e-12, "adam_stats/m_t_min": -0.0005098135443404317, "adam_stats/v_t_max": 1.6168087313417345e-05, "adam_stats/v_t_mean": 1.890078805272255e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.850996494293213, "all_logprobs/max": -5.46875, "all_logprobs/median": -6.875, "all_logprobs/min": -8.0625, "all_logprobs/p1": -7.6875, "all_logprobs/p10": -7.5, "all_logprobs/p25": -7.34375, "all_logprobs/p5": -7.5625, "all_logprobs/p75": -6.375, "all_logprobs/var": 0.2884191572666168, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.04533333333333334, "feature_vector_variance/max_squared_error": 1260.9957275390625, "feature_vector_variance/metric": 55.33992385864258, "generated_tokens/total": 2458127.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.2518479547691437e-05, "loss": 0.0, "mean_logprobs": -6.84375, "mean_logprobs/var": 0.0013885498046875, "num_completions/total": 2720, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 135.0316619873047, "per_token_feature_norm/max": 137.0, "per_token_feature_norm/median": 135.0, "per_token_feature_norm/min": 132.0, "per_token_feature_norm/p25": 135.0, "per_token_feature_norm/p75": 135.0, "per_token_feature_norm/var": 0.33298563957214355, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0238289833068848, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 5.076349680166459e-06, "policy_entropy": 3.7887227535247803, "policy_entropy/max": 3.859375, "policy_entropy/median": 3.796875, "policy_entropy/min": 3.71875, "policy_entropy/p25": 3.78125, "policy_entropy/p75": 3.796875, "policy_entropy/var": 0.0003903232573065907, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.02843744121491909, "policy_sharpness/max": 0.033052969723939896, "policy_sharpness/median": 0.028413865715265274, "policy_sharpness/min": 0.02025435119867325, "policy_sharpness/p25": 0.02739795856177807, "policy_sharpness/p75": 0.02943502366542816, "policy_sharpness/var": 1.8555530232333695e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 34, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 5.239485744823469e-06, "adam_stats/lm_head/lr_effective_mean": -1.6911756717252757e-11, "adam_stats/lm_head/lr_effective_min": -5.225990207691211e-06, "adam_stats/lm_head/lr_effective_std": 3.3386959330528043e-07, "adam_stats/lr_effective_max": 5.345556473912438e-06, "adam_stats/lr_effective_mean": 3.685449168067123e-11, "adam_stats/lr_effective_min": -5.321694061422022e-06, "adam_stats/m_t_max": 0.00039028318133205175, "adam_stats/m_t_mean": -7.0797942161582306e-12, "adam_stats/m_t_min": -0.0004588321899063885, "adam_stats/v_t_max": 1.615192013559863e-05, "adam_stats/v_t_mean": 1.8881892577260473e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.813439846038818, "all_logprobs/max": -5.4375, "all_logprobs/median": -6.875, "all_logprobs/min": -8.0625, "all_logprobs/p1": -7.65625, "all_logprobs/p10": -7.46875, "all_logprobs/p25": -7.3125, "all_logprobs/p5": -7.53125, "all_logprobs/p75": -6.34375, "all_logprobs/var": 0.2889094948768616, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.04666666666666667, "feature_vector_variance/max_squared_error": 1334.7513427734375, "feature_vector_variance/metric": 56.707069396972656, "generated_tokens/total": 2540047.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.2320907072649045e-05, "loss": 0.0, "mean_logprobs": -6.8125, "mean_logprobs/var": 0.00115966796875, "num_completions/total": 2800, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 135.21690368652344, "per_token_feature_norm/max": 137.0, "per_token_feature_norm/median": 135.0, "per_token_feature_norm/min": 132.0, "per_token_feature_norm/p25": 135.0, "per_token_feature_norm/p75": 136.0, "per_token_feature_norm/var": 0.30413156747817993, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0238150358200073, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 5.39633310836507e-06, "policy_entropy": 3.787435531616211, "policy_entropy/max": 3.859375, "policy_entropy/median": 3.796875, "policy_entropy/min": 3.71875, "policy_entropy/p25": 3.78125, "policy_entropy/p75": 3.796875, "policy_entropy/var": 0.00038829795084893703, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.028664732351899147, "policy_sharpness/max": 0.03366221487522125, "policy_sharpness/median": 0.028685474768280983, "policy_sharpness/min": 0.020102793350815773, "policy_sharpness/p25": 0.027629802003502846, "policy_sharpness/p75": 0.029646608978509903, "policy_sharpness/var": 1.8604372371555655e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 35, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 4.6399873099289834e-06, "adam_stats/lm_head/lr_effective_mean": -1.4975643988779375e-11, "adam_stats/lm_head/lr_effective_min": -4.628037459042389e-06, "adam_stats/lm_head/lr_effective_std": 2.9565322279268e-07, "adam_stats/lr_effective_max": 4.733926289191004e-06, "adam_stats/lr_effective_mean": 3.2637757080289376e-11, "adam_stats/lr_effective_min": -4.712777354143327e-06, "adam_stats/m_t_max": 0.0003512548573780805, "adam_stats/m_t_mean": -6.371815835376493e-12, "adam_stats/m_t_min": -0.0004129489534534514, "adam_stats/v_t_max": 1.6135769328684546e-05, "adam_stats/v_t_mean": 1.886300360701143e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.796523571014404, "all_logprobs/max": -5.4375, "all_logprobs/median": -6.84375, "all_logprobs/min": -8.0625, "all_logprobs/p1": -7.65625, "all_logprobs/p10": -7.46875, "all_logprobs/p25": -7.28125, "all_logprobs/p5": -7.53125, "all_logprobs/p75": -6.34375, "all_logprobs/var": 0.29236066341400146, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.048, "feature_vector_variance/max_squared_error": 1206.734130859375, "feature_vector_variance/metric": 60.1046257019043, "generated_tokens/total": 2621967.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.2117461064942437e-05, "loss": 0.0, "mean_logprobs": -6.8125, "mean_logprobs/var": 0.00244140625, "num_completions/total": 2880, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 135.2681884765625, "per_token_feature_norm/max": 137.0, "per_token_feature_norm/median": 135.0, "per_token_feature_norm/min": 132.0, "per_token_feature_norm/p25": 135.0, "per_token_feature_norm/p75": 136.0, "per_token_feature_norm/var": 0.3449479341506958, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0237634181976318, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 5.28287500856095e-06, "policy_entropy": 3.7868735790252686, "policy_entropy/max": 3.859375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.71875, "policy_entropy/p25": 3.78125, "policy_entropy/p75": 3.796875, "policy_entropy/var": 0.0003675985208246857, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.02874922566115856, "policy_sharpness/max": 0.03324954956769943, "policy_sharpness/median": 0.028762755915522575, "policy_sharpness/min": 0.02087710238993168, "policy_sharpness/p25": 0.027763355523347855, "policy_sharpness/p75": 0.029687125235795975, "policy_sharpness/var": 1.741543314892624e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 36, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 4.105985226487974e-06, "adam_stats/lm_head/lr_effective_mean": -1.3251182269524708e-11, "adam_stats/lm_head/lr_effective_min": -4.095412805327214e-06, "adam_stats/lm_head/lr_effective_std": 2.6161393407164724e-07, "adam_stats/lr_effective_max": 4.189117589703528e-06, "adam_stats/lr_effective_mean": 2.888169564618792e-11, "adam_stats/lr_effective_min": -4.1703874558152165e-06, "adam_stats/m_t_max": 0.0003161293570883572, "adam_stats/m_t_mean": -5.734634121734583e-12, "adam_stats/m_t_min": -0.00037165405228734016, "adam_stats/v_t_max": 1.611963307368569e-05, "adam_stats/v_t_mean": 1.8844140657614528e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.779452800750732, "all_logprobs/max": -5.4375, "all_logprobs/median": -6.84375, "all_logprobs/min": -8.125, "all_logprobs/p1": -7.65625, "all_logprobs/p10": -7.4375, "all_logprobs/p25": -7.28125, "all_logprobs/p5": -7.5, "all_logprobs/p75": -6.3125, "all_logprobs/var": 0.2931441068649292, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.04933333333333333, "feature_vector_variance/max_squared_error": 2266.069580078125, "feature_vector_variance/metric": 71.86038208007812, "generated_tokens/total": 2703887.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.1908389392193549e-05, "loss": 0.0, "mean_logprobs": -6.78125, "mean_logprobs/var": 0.0036773681640625, "num_completions/total": 2960, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 135.3329315185547, "per_token_feature_norm/max": 137.0, "per_token_feature_norm/median": 135.0, "per_token_feature_norm/min": 131.0, "per_token_feature_norm/p25": 135.0, "per_token_feature_norm/p75": 136.0, "per_token_feature_norm/var": 0.4500691890716553, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0236921310424805, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 5.223639163887128e-06, "policy_entropy": 3.7873191833496094, "policy_entropy/max": 3.875, "policy_entropy/median": 3.796875, "policy_entropy/min": 3.71875, "policy_entropy/p25": 3.78125, "policy_entropy/p75": 3.796875, "policy_entropy/var": 0.0003754021890927106, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.028764594346284866, "policy_sharpness/max": 0.03329081833362579, "policy_sharpness/median": 0.028765005990862846, "policy_sharpness/min": 0.016439074650406837, "policy_sharpness/p25": 0.02777761034667492, "policy_sharpness/p75": 0.029749399051070213, "policy_sharpness/var": 1.9769202026509447e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 37, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 3.630652372521581e-06, "adam_stats/lm_head/lr_effective_mean": -1.1716301540065643e-11, "adam_stats/lm_head/lr_effective_min": -3.62130504072411e-06, "adam_stats/lm_head/lr_effective_std": 2.3131603654746868e-07, "adam_stats/lr_effective_max": 3.704164328155457e-06, "adam_stats/lr_effective_mean": 2.5538278089487498e-11, "adam_stats/lr_effective_min": -3.6875899240840226e-06, "adam_stats/m_t_max": 0.0002845164272002876, "adam_stats/m_t_mean": -5.161172270812253e-12, "adam_stats/m_t_min": -0.00033448863541707397, "adam_stats/v_t_max": 1.6103513189591467e-05, "adam_stats/v_t_mean": 1.882530156066542e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.796367645263672, "all_logprobs/max": -5.4375, "all_logprobs/median": -6.84375, "all_logprobs/min": -8.3125, "all_logprobs/p1": -7.90625, "all_logprobs/p10": -7.46875, "all_logprobs/p25": -7.28125, "all_logprobs/p5": -7.59375, "all_logprobs/p75": -6.3125, "all_logprobs/var": 0.3226962983608246, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.050666666666666665, "feature_vector_variance/max_squared_error": 3205.17724609375, "feature_vector_variance/metric": 149.5718536376953, "generated_tokens/total": 2785807.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.1693946776030601e-05, "loss": 0.0, "mean_logprobs": -6.8125, "mean_logprobs/var": 0.03173828125, "num_completions/total": 3040, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 135.28558349609375, "per_token_feature_norm/max": 137.0, "per_token_feature_norm/median": 135.0, "per_token_feature_norm/min": 130.0, "per_token_feature_norm/p25": 135.0, "per_token_feature_norm/p75": 136.0, "per_token_feature_norm/var": 1.0274038314819336, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0237358808517456, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 5.498989139596233e-06, "policy_entropy": 3.7855629920959473, "policy_entropy/max": 3.875, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.796875, "policy_entropy/var": 0.00038159574614837766, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.028804799541831017, "policy_sharpness/max": 0.033742498606443405, "policy_sharpness/median": 0.029005100950598717, "policy_sharpness/min": 0.017187874764204025, "policy_sharpness/p25": 0.027902161702513695, "policy_sharpness/p75": 0.02997449040412903, "policy_sharpness/var": 3.0350377073773416e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 38, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 3.2078389722300926e-06, "adam_stats/lm_head/lr_effective_mean": -1.035111146102663e-11, "adam_stats/lm_head/lr_effective_min": -3.1995814424590208e-06, "adam_stats/lm_head/lr_effective_std": 2.043672253648765e-07, "adam_stats/lr_effective_max": 3.2727932648413116e-06, "adam_stats/lr_effective_mean": 2.256425948865548e-11, "adam_stats/lr_effective_min": -3.2581369850959163e-06, "adam_stats/m_t_max": 0.0002560647844802588, "adam_stats/m_t_mean": -4.6450556075161575e-12, "adam_stats/m_t_min": -0.00030103977769613266, "adam_stats/v_t_max": 1.6087409676401876e-05, "adam_stats/v_t_mean": 1.8806477642546726e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.710330486297607, "all_logprobs/max": -5.375, "all_logprobs/median": -6.75, "all_logprobs/min": -8.0625, "all_logprobs/p1": -7.5625, "all_logprobs/p10": -7.40625, "all_logprobs/p25": -7.21875, "all_logprobs/p5": -7.46875, "all_logprobs/p75": -6.25, "all_logprobs/var": 0.29730358719825745, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.052, "feature_vector_variance/max_squared_error": 1459.47119140625, "feature_vector_variance/metric": 59.78212356567383, "generated_tokens/total": 2867727.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.1474394481749037e-05, "loss": 0.0, "mean_logprobs": -6.71875, "mean_logprobs/var": 0.001312255859375, "num_completions/total": 3120, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 135.71676635742188, "per_token_feature_norm/max": 137.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 132.0, "per_token_feature_norm/p25": 135.0, "per_token_feature_norm/p75": 136.0, "per_token_feature_norm/var": 0.3857848048210144, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0237420797348022, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 6.44295596430311e-06, "policy_entropy": 3.7826459407806396, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.796875, "policy_entropy/var": 0.0003658164932858199, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.0293432530015707, "policy_sharpness/max": 0.033915817737579346, "policy_sharpness/median": 0.029342735186219215, "policy_sharpness/min": 0.021382803097367287, "policy_sharpness/p25": 0.02847013808786869, "policy_sharpness/p75": 0.03018457256257534, "policy_sharpness/var": 1.5749817521282239e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 39, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 2.8320084766164655e-06, "adam_stats/lm_head/lr_effective_mean": -9.137709686135587e-12, "adam_stats/lm_head/lr_effective_min": -2.8247193313291064e-06, "adam_stats/lm_head/lr_effective_std": 1.804142470973602e-07, "adam_stats/lr_effective_max": 2.8893555281683803e-06, "adam_stats/lr_effective_mean": 1.992068662803881e-11, "adam_stats/lr_effective_min": -2.876405915230862e-06, "adam_stats/m_t_max": 0.0002304583031218499, "adam_stats/m_t_mean": -4.180547401311241e-12, "adam_stats/m_t_min": -0.0002709357941057533, "adam_stats/v_t_max": 1.6071322534116916e-05, "adam_stats/v_t_mean": 1.8787668903258448e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.756722450256348, "all_logprobs/max": -5.375, "all_logprobs/median": -6.8125, "all_logprobs/min": -8.4375, "all_logprobs/p1": -7.8125, "all_logprobs/p10": -7.4375, "all_logprobs/p25": -7.25, "all_logprobs/p5": -7.53125, "all_logprobs/p75": -6.28125, "all_logprobs/var": 0.3196379244327545, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.05333333333333334, "feature_vector_variance/max_squared_error": 4135.36572265625, "feature_vector_variance/metric": 134.29254150390625, "generated_tokens/total": 2949647.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.125e-05, "loss": 0.0, "mean_logprobs": -6.75, "mean_logprobs/var": 0.027099609375, "num_completions/total": 3200, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 135.51449584960938, "per_token_feature_norm/max": 137.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 130.0, "per_token_feature_norm/p25": 135.0, "per_token_feature_norm/p75": 136.0, "per_token_feature_norm/var": 0.9817370772361755, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0237056016921997, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 6.109962669142988e-06, "policy_entropy": 3.7846341133117676, "policy_entropy/max": 3.875, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.796875, "policy_entropy/var": 0.0004097261989954859, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.02898579277098179, "policy_sharpness/max": 0.033599939197301865, "policy_sharpness/median": 0.029204681515693665, "policy_sharpness/min": 0.018942827358841896, "policy_sharpness/p25": 0.0281932782381773, "policy_sharpness/p75": 0.030109543353319168, "policy_sharpness/var": 3.0816181606496684e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 40, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 2.4981795831990894e-06, "adam_stats/lm_head/lr_effective_mean": -8.05999798153545e-12, "adam_stats/lm_head/lr_effective_min": -2.491750592525932e-06, "adam_stats/lm_head/lr_effective_std": 1.5913934703348787e-07, "adam_stats/lr_effective_max": 2.5487693164905068e-06, "adam_stats/lr_effective_mean": 1.75725579920627e-11, "adam_stats/lr_effective_min": -2.5373371954628965e-06, "adam_stats/m_t_max": 0.00020741246407851577, "adam_stats/m_t_mean": -3.762492487707769e-12, "adam_stats/m_t_min": -0.00024384220887441188, "adam_stats/v_t_max": 1.605525176273659e-05, "adam_stats/v_t_mean": 1.8768884016417964e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.690103054046631, "all_logprobs/max": -5.375, "all_logprobs/median": -6.75, "all_logprobs/min": -8.0, "all_logprobs/p1": -7.5625, "all_logprobs/p10": -7.375, "all_logprobs/p25": -7.1875, "all_logprobs/p5": -7.4375, "all_logprobs/p75": -6.21875, "all_logprobs/var": 0.30065926909446716, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.05466666666666667, "feature_vector_variance/max_squared_error": 1594.9493408203125, "feature_vector_variance/metric": 63.918426513671875, "generated_tokens/total": 3031567.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.1021036720894182e-05, "loss": 0.0, "mean_logprobs": -6.6875, "mean_logprobs/var": 0.0022125244140625, "num_completions/total": 3280, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 135.8619842529297, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 132.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 136.0, "per_token_feature_norm/var": 0.3921639919281006, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.023705005645752, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 6.938790193089517e-06, "policy_entropy": 3.7813503742218018, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.796875, "policy_entropy/var": 0.0003607870312407613, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.029507959261536598, "policy_sharpness/max": 0.03382803127169609, "policy_sharpness/median": 0.029489045962691307, "policy_sharpness/min": 0.020829832181334496, "policy_sharpness/p25": 0.028658462688326836, "policy_sharpness/p75": 0.030333133414387703, "policy_sharpness/var": 1.5433649878104916e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 41, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 2.2018748495611362e-06, "adam_stats/lm_head/lr_effective_mean": -7.103501380861799e-12, "adam_stats/lm_head/lr_effective_min": -2.196209379690117e-06, "adam_stats/lm_head/lr_effective_std": 1.4025687278262922e-07, "adam_stats/lr_effective_max": 2.2464666926680366e-06, "adam_stats/lr_effective_mean": 1.548837620657384e-11, "adam_stats/lr_effective_min": -2.2363824427884538e-06, "adam_stats/m_t_max": 0.0001866712118498981, "adam_stats/m_t_mean": -3.386242891992297e-12, "adam_stats/m_t_min": -0.00021945798653177917, "adam_stats/v_t_max": 1.6039197362260893e-05, "adam_stats/v_t_mean": 1.8750114308407895e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.682619571685791, "all_logprobs/max": -5.28125, "all_logprobs/median": -6.75, "all_logprobs/min": -8.0625, "all_logprobs/p1": -7.53125, "all_logprobs/p10": -7.375, "all_logprobs/p25": -7.1875, "all_logprobs/p5": -7.4375, "all_logprobs/p75": -6.21875, "all_logprobs/var": 0.3016487658023834, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.056, "feature_vector_variance/max_squared_error": 1574.0350341796875, "feature_vector_variance/metric": 60.7067985534668, "generated_tokens/total": 3113487.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.078778360091808e-05, "loss": 0.0, "mean_logprobs": -6.6875, "mean_logprobs/var": 0.00151824951171875, "num_completions/total": 3360, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 135.9196014404297, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 132.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 136.0, "per_token_feature_norm/var": 0.3621789216995239, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0236741304397583, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 7.1048921199690085e-06, "policy_entropy": 3.781285047531128, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.796875, "policy_entropy/var": 0.0003546706575434655, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.02950943075120449, "policy_sharpness/max": 0.03423319384455681, "policy_sharpness/median": 0.029478542506694794, "policy_sharpness/min": 0.02056947909295559, "policy_sharpness/p25": 0.02867721952497959, "policy_sharpness/p75": 0.030305372551083565, "policy_sharpness/var": 1.4671953749711975e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 42, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 1.939070898515638e-06, "adam_stats/lm_head/lr_effective_mean": -6.255213794853498e-12, "adam_stats/lm_head/lr_effective_min": -1.9340823200764135e-06, "adam_stats/lm_head/lr_effective_std": 1.2351019051948242e-07, "adam_stats/lr_effective_max": 1.9783421976171667e-06, "adam_stats/lr_effective_mean": 1.3639815839294744e-11, "adam_stats/lr_effective_min": -1.9694546153914416e-06, "adam_stats/m_t_max": 0.00016800408775452524, "adam_stats/m_t_mean": -3.0476187979494584e-12, "adam_stats/m_t_min": -0.0001975121849682182, "adam_stats/v_t_max": 1.6023157513700426e-05, "adam_stats/v_t_mean": 1.873136845284562e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.659846782684326, "all_logprobs/max": -5.3125, "all_logprobs/median": -6.71875, "all_logprobs/min": -7.90625, "all_logprobs/p1": -7.53125, "all_logprobs/p10": -7.34375, "all_logprobs/p25": -7.15625, "all_logprobs/p5": -7.40625, "all_logprobs/p75": -6.1875, "all_logprobs/var": 0.3031567335128784, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.05733333333333333, "feature_vector_variance/max_squared_error": 1711.9827880859375, "feature_vector_variance/metric": 59.71718978881836, "generated_tokens/total": 3195407.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.0550524823068504e-05, "loss": 0.0, "mean_logprobs": -6.65625, "mean_logprobs/var": 0.00077056884765625, "num_completions/total": 3440, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.07057189941406, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 132.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 136.0, "per_token_feature_norm/var": 0.3200112581253052, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.023677945137024, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 7.895831913629081e-06, "policy_entropy": 3.7793240547180176, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.796875, "policy_entropy/var": 0.00035442886292003095, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.029746538028120995, "policy_sharpness/max": 0.03406437486410141, "policy_sharpness/median": 0.029711885377764702, "policy_sharpness/min": 0.021866749972105026, "policy_sharpness/p25": 0.02894657850265503, "policy_sharpness/p75": 0.030504953116178513, "policy_sharpness/var": 1.3959291891296743e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 43, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 1.7061549897334771e-06, "adam_stats/lm_head/lr_effective_mean": -5.50345359562332e-12, "adam_stats/lm_head/lr_effective_min": -1.7017661093632341e-06, "adam_stats/lm_head/lr_effective_std": 1.0866887834026784e-07, "adam_stats/lr_effective_max": 1.7407111272405018e-06, "adam_stats/lr_effective_mean": 1.2001476201728423e-11, "adam_stats/lr_effective_min": -1.7328845842712326e-06, "adam_stats/m_t_max": 0.00015120366879273206, "adam_stats/m_t_mean": -2.742857807200294e-12, "adam_stats/m_t_min": -0.00017776095774024725, "adam_stats/v_t_max": 1.600713403604459e-05, "adam_stats/v_t_mean": 1.8712631270900726e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.663926124572754, "all_logprobs/max": -5.28125, "all_logprobs/median": -6.71875, "all_logprobs/min": -8.0625, "all_logprobs/p1": -7.5625, "all_logprobs/p10": -7.34375, "all_logprobs/p25": -7.15625, "all_logprobs/p5": -7.4375, "all_logprobs/p75": -6.1875, "all_logprobs/var": 0.30497485399246216, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.058666666666666666, "feature_vector_variance/max_squared_error": 1811.744384765625, "feature_vector_variance/metric": 67.27422332763672, "generated_tokens/total": 3277327.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.0309549450619342e-05, "loss": 0.0, "mean_logprobs": -6.65625, "mean_logprobs/var": 0.00286865234375, "num_completions/total": 3520, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.0216827392578, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 132.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 136.0, "per_token_feature_norm/var": 0.38173824548721313, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0236215591430664, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 7.5703183028963394e-06, "policy_entropy": 3.7799370288848877, "policy_entropy/max": 3.859375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.796875, "policy_entropy/var": 0.0003509391681291163, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.02967517450451851, "policy_sharpness/max": 0.03419867902994156, "policy_sharpness/median": 0.029644358903169632, "policy_sharpness/min": 0.020411916077136993, "policy_sharpness/p25": 0.028865545988082886, "policy_sharpness/p75": 0.030471939593553543, "policy_sharpness/var": 1.4785582607146353e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 44, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 1.4998862525317236e-06, "adam_stats/lm_head/lr_effective_mean": -4.837756931164172e-12, "adam_stats/lm_head/lr_effective_min": -1.4960286307541537e-06, "adam_stats/lm_head/lr_effective_std": 9.552622515229814e-08, "adam_stats/lr_effective_max": 1.5302662177418824e-06, "adam_stats/lr_effective_mean": 1.0550582976720513e-11, "adam_stats/lr_effective_min": -1.5233803196679219e-06, "adam_stats/m_t_max": 0.00013608329754788429, "adam_stats/m_t_mean": -2.468571137434483e-12, "adam_stats/m_t_min": -0.00015998486196622252, "adam_stats/v_t_max": 1.5991126929293387e-05, "adam_stats/v_t_mean": 1.8693922278212316e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.6511383056640625, "all_logprobs/max": -5.21875, "all_logprobs/median": -6.71875, "all_logprobs/min": -8.0625, "all_logprobs/p1": -7.53125, "all_logprobs/p10": -7.34375, "all_logprobs/p25": -7.15625, "all_logprobs/p5": -7.40625, "all_logprobs/p75": -6.1875, "all_logprobs/var": 0.3028331398963928, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.06, "feature_vector_variance/max_squared_error": 1658.0821533203125, "feature_vector_variance/metric": 63.193580627441406, "generated_tokens/total": 3359247.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.0065151074942516e-05, "loss": 0.0, "mean_logprobs": -6.65625, "mean_logprobs/var": 0.00130462646484375, "num_completions/total": 3600, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.09593200683594, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 136.0, "per_token_feature_norm/var": 0.33118465542793274, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0236014127731323, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 7.759151230857242e-06, "policy_entropy": 3.7792325019836426, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.796875, "policy_entropy/var": 0.00033346787677146494, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.029812544584274292, "policy_sharpness/max": 0.034047119319438934, "policy_sharpness/median": 0.029762905091047287, "policy_sharpness/min": 0.021607141941785812, "policy_sharpness/p25": 0.029038114473223686, "policy_sharpness/p75": 0.03055165894329548, "policy_sharpness/var": 1.3525527720048558e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 45, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 1.317357941843511e-06, "adam_stats/lm_head/lr_effective_mean": -4.2487159623849635e-12, "adam_stats/lm_head/lr_effective_min": -1.3139703014530824e-06, "adam_stats/lm_head/lr_effective_std": 8.389685035581351e-08, "adam_stats/lr_effective_max": 1.3440421753330156e-06, "adam_stats/lr_effective_mean": 9.266660355722323e-12, "adam_stats/lr_effective_min": -1.3379894880927168e-06, "adam_stats/m_t_max": 0.0001224749576067552, "adam_stats/m_t_mean": -2.221714132111252e-12, "adam_stats/m_t_min": -0.00014398636994883418, "adam_stats/v_t_max": 1.5975136193446815e-05, "adam_stats/v_t_mean": 1.867522846435432e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.643501281738281, "all_logprobs/max": -5.15625, "all_logprobs/median": -6.71875, "all_logprobs/min": -8.0, "all_logprobs/p1": -7.53125, "all_logprobs/p10": -7.34375, "all_logprobs/p25": -7.15625, "all_logprobs/p5": -7.40625, "all_logprobs/p75": -6.15625, "all_logprobs/var": 0.30812159180641174, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.06133333333333333, "feature_vector_variance/max_squared_error": 1510.7662353515625, "feature_vector_variance/metric": 65.26241302490234, "generated_tokens/total": 3441167.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 9.817627457812105e-06, "loss": 0.0, "mean_logprobs": -6.65625, "mean_logprobs/var": 0.0018768310546875, "num_completions/total": 3680, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.11732482910156, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 136.0, "per_token_feature_norm/var": 0.3477512300014496, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.023630976676941, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 8.373663149541244e-06, "policy_entropy": 3.777808427810669, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.796875, "policy_entropy/var": 0.0003382944269105792, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.02984047494828701, "policy_sharpness/max": 0.034078631550073624, "policy_sharpness/median": 0.029786914587020874, "policy_sharpness/min": 0.022295920178294182, "policy_sharpness/p25": 0.02905312180519104, "policy_sharpness/p75": 0.03059573844075203, "policy_sharpness/var": 1.3866531389794545e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 46, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 1.155965605903475e-06, "adam_stats/lm_head/lr_effective_mean": -3.727927255087193e-12, "adam_stats/lm_head/lr_effective_min": -1.1529933772180812e-06, "adam_stats/lm_head/lr_effective_std": 7.36146716917574e-08, "adam_stats/lr_effective_max": 1.179381797555834e-06, "adam_stats/lr_effective_mean": 8.1314139449562e-12, "adam_stats/lr_effective_min": -1.1740664831449976e-06, "adam_stats/m_t_max": 0.00011022746184607968, "adam_stats/m_t_mean": -1.999544410255516e-12, "adam_stats/m_t_min": -0.00012958773004356772, "adam_stats/v_t_max": 1.5959161828504875e-05, "adam_stats/v_t_mean": 1.8656551997731086e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.707104682922363, "all_logprobs/max": -5.125, "all_logprobs/median": -6.78125, "all_logprobs/min": -8.25, "all_logprobs/p1": -7.90625, "all_logprobs/p10": -7.40625, "all_logprobs/p25": -7.1875, "all_logprobs/p5": -7.59375, "all_logprobs/p75": -6.21875, "all_logprobs/var": 0.34897592663764954, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.06266666666666666, "feature_vector_variance/max_squared_error": 3324.17333984375, "feature_vector_variance/metric": 187.7640838623047, "generated_tokens/total": 3523087.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 9.567280168627493e-06, "loss": 0.0, "mean_logprobs": -6.71875, "mean_logprobs/var": 0.0556640625, "num_completions/total": 3760, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 135.85614013671875, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 131.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 136.0, "per_token_feature_norm/var": 1.3342400789260864, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235379934310913, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 7.945764991745818e-06, "policy_entropy": 3.7825469970703125, "policy_entropy/max": 3.875, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.796875, "policy_entropy/var": 0.0005174307152628899, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.029234493151307106, "policy_sharpness/max": 0.034290216863155365, "policy_sharpness/median": 0.029692376032471657, "policy_sharpness/min": 0.017904412001371384, "policy_sharpness/p25": 0.028826531022787094, "policy_sharpness/p75": 0.03052971139550209, "policy_sharpness/var": 5.245845386525616e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 47, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 1.0133773002962698e-06, "adam_stats/lm_head/lr_effective_mean": -3.267849315799487e-12, "adam_stats/lm_head/lr_effective_min": -1.010772166409879e-06, "adam_stats/lm_head/lr_effective_std": 6.45309725655352e-08, "adam_stats/lr_effective_max": 1.033906187331013e-06, "adam_stats/lr_effective_mean": 7.128428056168978e-12, "adam_stats/lr_effective_min": -1.0292427532476722e-06, "adam_stats/m_t_max": 9.920471347868443e-05, "adam_stats/m_t_mean": -1.7995887982916181e-12, "adam_stats/m_t_min": -0.00011662895121844485, "adam_stats/v_t_max": 1.5943202015478164e-05, "adam_stats/v_t_mean": 1.86378972151513e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.654760837554932, "all_logprobs/max": -5.25, "all_logprobs/median": -6.71875, "all_logprobs/min": -8.0, "all_logprobs/p1": -7.53125, "all_logprobs/p10": -7.34375, "all_logprobs/p25": -7.15625, "all_logprobs/p5": -7.40625, "all_logprobs/p75": -6.1875, "all_logprobs/var": 0.3073756694793701, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.064, "feature_vector_variance/max_squared_error": 1657.1343994140625, "feature_vector_variance/metric": 67.9100570678711, "generated_tokens/total": 3605007.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 9.314414216997507e-06, "loss": 0.0, "mean_logprobs": -6.65625, "mean_logprobs/var": 0.00244140625, "num_completions/total": 3840, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.06861877441406, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 132.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 136.0, "per_token_feature_norm/var": 0.36979687213897705, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0236488580703735, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 8.25839742901735e-06, "policy_entropy": 3.7783501148223877, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.796875, "policy_entropy/var": 0.0003340212278999388, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.029740775004029274, "policy_sharpness/max": 0.03412364795804024, "policy_sharpness/median": 0.029695378616452217, "policy_sharpness/min": 0.02164391055703163, "policy_sharpness/p25": 0.028942076489329338, "policy_sharpness/p75": 0.030519956722855568, "policy_sharpness/var": 1.4383924735739129e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 48, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 8.87505905211583e-07, "adam_stats/lm_head/lr_effective_mean": -2.861744695381585e-12, "adam_stats/lm_head/lr_effective_min": -8.852246651258611e-07, "adam_stats/lm_head/lr_effective_std": 5.6512675428166403e-08, "adam_stats/lr_effective_max": 9.054858765011886e-07, "adam_stats/lr_effective_mean": 6.243026495073023e-12, "adam_stats/lr_effective_min": -9.013984367811645e-07, "adam_stats/m_t_max": 8.928423631004989e-05, "adam_stats/m_t_mean": -1.6196292354150876e-12, "adam_stats/m_t_min": -0.00010496605682419613, "adam_stats/v_t_max": 1.5927258573356085e-05, "adam_stats/v_t_mean": 1.8619259779806274e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.617442607879639, "all_logprobs/max": -5.25, "all_logprobs/median": -6.6875, "all_logprobs/min": -7.90625, "all_logprobs/p1": -7.5, "all_logprobs/p10": -7.3125, "all_logprobs/p25": -7.125, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.15625, "all_logprobs/var": 0.3108353018760681, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.06533333333333333, "feature_vector_variance/max_squared_error": 1600.9105224609375, "feature_vector_variance/metric": 63.243896484375, "generated_tokens/total": 3686927.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 9.059337681133194e-06, "loss": 0.0, "mean_logprobs": -6.625, "mean_logprobs/var": 0.00127410888671875, "num_completions/total": 3920, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.2517547607422, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.3104918599128723, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.02359938621521, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 9.624685844755732e-06, "policy_entropy": 3.776050567626953, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.796875, "policy_entropy/var": 0.00033603375777602196, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.030014891177415848, "policy_sharpness/max": 0.03427445888519287, "policy_sharpness/median": 0.029939977452158928, "policy_sharpness/min": 0.02173694595694542, "policy_sharpness/p25": 0.029246509075164795, "policy_sharpness/p75": 0.030745048075914383, "policy_sharpness/var": 1.2996597433811985e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 49, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 7.764852512082143e-07, "adam_stats/lm_head/lr_effective_mean": -2.5035772066583917e-12, "adam_stats/lm_head/lr_effective_min": -7.744897629891057e-07, "adam_stats/lm_head/lr_effective_std": 4.944078924040696e-08, "adam_stats/lr_effective_max": 7.922168379081995e-07, "adam_stats/lr_effective_mean": 5.462091716423867e-12, "adam_stats/lr_effective_min": -7.88637976256723e-07, "adam_stats/m_t_max": 8.035581413423643e-05, "adam_stats/m_t_mean": -1.4576665612400785e-12, "adam_stats/m_t_min": -9.446944750379771e-05, "adam_stats/v_t_max": 1.5911331502138637e-05, "adam_stats/v_t_mean": 1.8600633186482973e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.6422882080078125, "all_logprobs/max": -5.21875, "all_logprobs/median": -6.71875, "all_logprobs/min": -8.0625, "all_logprobs/p1": -7.53125, "all_logprobs/p10": -7.34375, "all_logprobs/p25": -7.15625, "all_logprobs/p5": -7.40625, "all_logprobs/p75": -6.15625, "all_logprobs/var": 0.30937686562538147, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.06666666666666667, "feature_vector_variance/max_squared_error": 1854.1571044921875, "feature_vector_variance/metric": 67.53270721435547, "generated_tokens/total": 3768847.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 8.80236133250198e-06, "loss": 0.0, "mean_logprobs": -6.65625, "mean_logprobs/var": 0.0030517578125, "num_completions/total": 4000, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.124755859375, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 132.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 136.0, "per_token_feature_norm/var": 0.361490398645401, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.023566722869873, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 8.530024388164748e-06, "policy_entropy": 3.7778682708740234, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.796875, "policy_entropy/var": 0.0003241777594666928, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.029788916930556297, "policy_sharpness/max": 0.034180670976638794, "policy_sharpness/median": 0.029744146391749382, "policy_sharpness/min": 0.021078934893012047, "policy_sharpness/p25": 0.029006602242588997, "policy_sharpness/p75": 0.030547719448804855, "policy_sharpness/var": 1.3810976042805123e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 50, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 6.786474386899499e-07, "adam_stats/lm_head/lr_effective_mean": -2.187965737407427e-12, "adam_stats/lm_head/lr_effective_min": -6.769036531295569e-07, "adam_stats/lm_head/lr_effective_std": 4.3208974176423e-08, "adam_stats/lr_effective_max": 6.923975774952851e-07, "adam_stats/lr_effective_mean": 4.773879642289147e-12, "adam_stats/lr_effective_min": -6.892670398883638e-07, "adam_stats/m_t_max": 7.232023199321702e-05, "adam_stats/m_t_mean": -1.3118990485963544e-12, "adam_stats/m_t_min": -8.502249693265185e-05, "adam_stats/v_t_max": 1.589542080182582e-05, "adam_stats/v_t_mean": 1.858204128762919e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.623703956604004, "all_logprobs/max": -5.125, "all_logprobs/median": -6.6875, "all_logprobs/min": -7.90625, "all_logprobs/p1": -7.5, "all_logprobs/p10": -7.3125, "all_logprobs/p25": -7.125, "all_logprobs/p5": -7.40625, "all_logprobs/p75": -6.15625, "all_logprobs/var": 0.3113349378108978, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.068, "feature_vector_variance/max_squared_error": 2213.154296875, "feature_vector_variance/metric": 67.66040802001953, "generated_tokens/total": 3850767.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 8.543798257200491e-06, "loss": 0.0, "mean_logprobs": -6.625, "mean_logprobs/var": 0.00173187255859375, "num_completions/total": 4080, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.21469116210938, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 132.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.3402784764766693, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235751867294312, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 9.159616638498846e-06, "policy_entropy": 3.776813268661499, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.796875, "policy_entropy/var": 0.00033293318119831383, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.02996680699288845, "policy_sharpness/max": 0.03433973714709282, "policy_sharpness/median": 0.029898708686232567, "policy_sharpness/min": 0.020727792754769325, "policy_sharpness/p25": 0.029200179502367973, "policy_sharpness/p75": 0.030691025778651237, "policy_sharpness/var": 1.3210965335019864e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 51, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 5.925030563957989e-07, "adam_stats/lm_head/lr_effective_mean": -1.910098164706331e-12, "adam_stats/lm_head/lr_effective_min": -5.909808464821253e-07, "adam_stats/lm_head/lr_effective_std": 3.772227330500755e-08, "adam_stats/lr_effective_max": 6.045085001460393e-07, "adam_stats/lr_effective_mean": 4.167923384895689e-12, "adam_stats/lr_effective_min": -6.017730811436195e-07, "adam_stats/m_t_max": 6.508820661110803e-05, "adam_stats/m_t_mean": -1.1807097183638704e-12, "adam_stats/m_t_min": -7.652024214621633e-05, "adam_stats/v_t_max": 1.5879526472417638e-05, "adam_stats/v_t_mean": 1.856345806239279e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.614305019378662, "all_logprobs/max": -5.125, "all_logprobs/median": -6.6875, "all_logprobs/min": -8.0, "all_logprobs/p1": -7.5, "all_logprobs/p10": -7.3125, "all_logprobs/p25": -7.125, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.3158220052719116, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.06933333333333333, "feature_vector_variance/max_squared_error": 2062.398193359375, "feature_vector_variance/metric": 67.2181167602539, "generated_tokens/total": 3932687.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 8.283963474507402e-06, "loss": 0.0, "mean_logprobs": -6.625, "mean_logprobs/var": 0.0021209716796875, "num_completions/total": 4160, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.2650909423828, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 132.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.33952251076698303, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.023630976676941, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.0066447430290282e-05, "policy_entropy": 3.7751526832580566, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.0003329580067656934, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.030019259080290794, "policy_sharpness/max": 0.03466236591339111, "policy_sharpness/median": 0.029956484213471413, "policy_sharpness/min": 0.020623499527573586, "policy_sharpness/p25": 0.02925645187497139, "policy_sharpness/p75": 0.03075724095106125, "policy_sharpness/var": 1.315354438702343e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 52, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 5.167229915059579e-07, "adam_stats/lm_head/lr_effective_mean": -1.6656785542870645e-12, "adam_stats/lm_head/lr_effective_min": -5.15395697675558e-07, "adam_stats/lm_head/lr_effective_std": 3.289596861577593e-08, "adam_stats/lr_effective_max": 5.271935492601187e-07, "adam_stats/lr_effective_mean": 3.6348641110905966e-12, "adam_stats/lr_effective_min": -5.248060688245459e-07, "adam_stats/m_t_max": 5.857938594999723e-05, "adam_stats/m_t_mean": -1.0626384754769402e-12, "adam_stats/m_t_min": -6.886821938678622e-05, "adam_stats/v_t_max": 1.5863646694924682e-05, "adam_stats/v_t_mean": 1.854489652119984e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.596192359924316, "all_logprobs/max": -5.15625, "all_logprobs/median": -6.6875, "all_logprobs/min": -7.9375, "all_logprobs/p1": -7.46875, "all_logprobs/p10": -7.3125, "all_logprobs/p25": -7.09375, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.3149842917919159, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.07066666666666667, "feature_vector_variance/max_squared_error": 1484.1505126953125, "feature_vector_variance/metric": 61.373207092285156, "generated_tokens/total": 4014607.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 8.02317355308094e-06, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.000888824462890625, "num_completions/total": 4240, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.31741333007812, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.30743408203125, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235460996627808, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.0516066140553448e-05, "policy_entropy": 3.7748401165008545, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.00034057869925163686, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.03012160025537014, "policy_sharpness/max": 0.034498799592256546, "policy_sharpness/median": 0.030040517449378967, "policy_sharpness/min": 0.022127099335193634, "policy_sharpness/p25": 0.02936449460685253, "policy_sharpness/p75": 0.030828332528471947, "policy_sharpness/var": 1.255763436347479e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 53, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 4.5012200189376017e-07, "adam_stats/lm_head/lr_effective_mean": -1.4508821661457505e-12, "adam_stats/lm_head/lr_effective_min": -4.489659772843879e-07, "adam_stats/lm_head/lr_effective_std": 2.8654486783352695e-08, "adam_stats/lr_effective_max": 4.592434379446786e-07, "adam_stats/lr_effective_mean": 3.1663727629444027e-12, "adam_stats/lr_effective_min": -4.571620024762524e-07, "adam_stats/m_t_max": 5.272144699119963e-05, "adam_stats/m_t_mean": -9.56374953189898e-13, "adam_stats/m_t_min": -6.198139453772455e-05, "adam_stats/v_t_max": 1.584778328833636e-05, "adam_stats/v_t_mean": 1.8526347990432956e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.672000408172607, "all_logprobs/max": -5.15625, "all_logprobs/median": -6.75, "all_logprobs/min": -8.25, "all_logprobs/p1": -7.84375, "all_logprobs/p10": -7.375, "all_logprobs/p25": -7.15625, "all_logprobs/p5": -7.5, "all_logprobs/p75": -6.1875, "all_logprobs/var": 0.3398115038871765, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.072, "feature_vector_variance/max_squared_error": 4211.44873046875, "feature_vector_variance/metric": 163.57119750976562, "generated_tokens/total": 4096527.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 7.76174622526876e-06, "loss": 0.0, "mean_logprobs": -6.6875, "mean_logprobs/var": 0.040283203125, "num_completions/total": 4320, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 135.95547485351562, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 130.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 1.1386786699295044, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235109329223633, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 8.667248039273545e-06, "policy_entropy": 3.7801568508148193, "policy_entropy/max": 3.875, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.796875, "policy_entropy/var": 0.0004476238100323826, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.029557382687926292, "policy_sharpness/max": 0.03402610868215561, "policy_sharpness/median": 0.029767407104372978, "policy_sharpness/min": 0.018737994134426117, "policy_sharpness/p25": 0.028964586555957794, "policy_sharpness/p75": 0.03058842569589615, "policy_sharpness/var": 2.931126118710381e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 54, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 3.916437663065153e-07, "adam_stats/lm_head/lr_effective_mean": -1.2622973413062288e-12, "adam_stats/lm_head/lr_effective_min": -3.9063812096173933e-07, "adam_stats/lm_head/lr_effective_std": 2.4930512765308777e-08, "adam_stats/lr_effective_max": 3.9958055708666507e-07, "adam_stats/lr_effective_mean": 2.7550208208521054e-12, "adam_stats/lr_effective_min": -3.9776816151970706e-07, "adam_stats/m_t_max": 4.7449302655877545e-05, "adam_stats/m_t_mean": -8.607371000841912e-13, "adam_stats/m_t_min": -5.5783253628760576e-05, "adam_stats/v_t_max": 1.5831936252652667e-05, "adam_stats/v_t_mean": 1.850782764892256e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.616007328033447, "all_logprobs/max": -5.125, "all_logprobs/median": -6.6875, "all_logprobs/min": -8.0625, "all_logprobs/p1": -7.5, "all_logprobs/p10": -7.3125, "all_logprobs/p25": -7.125, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.15625, "all_logprobs/var": 0.3119471073150635, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.07333333333333333, "feature_vector_variance/max_squared_error": 2051.100830078125, "feature_vector_variance/metric": 73.4501953125, "generated_tokens/total": 4178447.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 7.5e-06, "loss": 0.0, "mean_logprobs": -6.625, "mean_logprobs/var": 0.003509521484375, "num_completions/total": 4400, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.2475128173828, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 132.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.3910871148109436, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235037803649902, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 9.342516023025382e-06, "policy_entropy": 3.776630401611328, "policy_entropy/max": 3.859375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.796875, "policy_entropy/var": 0.00033178317244164646, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.02995959296822548, "policy_sharpness/max": 0.03443952649831772, "policy_sharpness/median": 0.029903961345553398, "policy_sharpness/min": 0.020390156656503677, "policy_sharpness/p25": 0.02919398993253708, "policy_sharpness/p75": 0.030708283185958862, "policy_sharpness/var": 1.3774068747807178e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 55, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 3.403478103791713e-07, "adam_stats/lm_head/lr_effective_mean": -1.0968872294833676e-12, "adam_stats/lm_head/lr_effective_min": -3.394739849227335e-07, "adam_stats/lm_head/lr_effective_std": 2.1664090965600735e-08, "adam_stats/lr_effective_max": 3.4724547504083603e-07, "adam_stats/lr_effective_mean": 2.3941883631889427e-12, "adam_stats/lr_effective_min": -3.4566915019240696e-07, "adam_stats/m_t_max": 4.270437057130039e-05, "adam_stats/m_t_mean": -7.746635364430654e-13, "adam_stats/m_t_min": -5.0204926083097234e-05, "adam_stats/v_t_max": 1.5816103768884204e-05, "adam_stats/v_t_mean": 1.848931598102954e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.594099521636963, "all_logprobs/max": -5.15625, "all_logprobs/median": -6.6875, "all_logprobs/min": -7.9375, "all_logprobs/p1": -7.46875, "all_logprobs/p10": -7.28125, "all_logprobs/p25": -7.09375, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.311859667301178, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.07466666666666667, "feature_vector_variance/max_squared_error": 1811.9180908203125, "feature_vector_variance/metric": 63.817073822021484, "generated_tokens/total": 4260367.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 7.238253774731245e-06, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.00115966796875, "num_completions/total": 4480, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.36891174316406, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.32532328367233276, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235240459442139, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.0503223165869713e-05, "policy_entropy": 3.7750351428985596, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.00034020296880044043, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.030129684135317802, "policy_sharpness/max": 0.03476140275597572, "policy_sharpness/median": 0.03004651889204979, "policy_sharpness/min": 0.022277163341641426, "policy_sharpness/p25": 0.029384754598140717, "policy_sharpness/p75": 0.030822327360510826, "policy_sharpness/var": 1.214828898810083e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 56, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 2.953971431907121e-07, "adam_stats/lm_head/lr_effective_mean": -9.519495651824639e-13, "adam_stats/lm_head/lr_effective_min": -2.946388519831089e-07, "adam_stats/lm_head/lr_effective_std": 1.88018827174119e-08, "adam_stats/lr_effective_max": 3.0138411943880783e-07, "adam_stats/lr_effective_mean": 2.0779905574030977e-12, "adam_stats/lr_effective_min": -3.000149035869981e-07, "adam_stats/m_t_max": 3.843393278657459e-05, "adam_stats/m_t_mean": -6.971969117482157e-13, "adam_stats/m_t_min": -4.518443165579811e-05, "adam_stats/v_t_max": 1.5800287656020373e-05, "adam_stats/v_t_mean": 1.847082382877563e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.617790222167969, "all_logprobs/max": -5.1875, "all_logprobs/median": -6.6875, "all_logprobs/min": -8.0, "all_logprobs/p1": -7.5, "all_logprobs/p10": -7.3125, "all_logprobs/p25": -7.125, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.15625, "all_logprobs/var": 0.3119657337665558, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.076, "feature_vector_variance/max_squared_error": 1576.56201171875, "feature_vector_variance/metric": 69.23607635498047, "generated_tokens/total": 4342287.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 6.976826446919061e-06, "loss": 0.0, "mean_logprobs": -6.625, "mean_logprobs/var": 0.00250244140625, "num_completions/total": 4560, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.23590087890625, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.3636297881603241, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235484838485718, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 9.596885320206638e-06, "policy_entropy": 3.7761924266815186, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.796875, "policy_entropy/var": 0.0003289585292804986, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.02993767522275448, "policy_sharpness/max": 0.034180670976638794, "policy_sharpness/median": 0.029890457168221474, "policy_sharpness/min": 0.021523859351873398, "policy_sharpness/p25": 0.029170168563723564, "policy_sharpness/p75": 0.030690276995301247, "policy_sharpness/var": 1.3734183994529303e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 57, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 2.560475422797026e-07, "adam_stats/lm_head/lr_effective_mean": -8.250812684983122e-13, "adam_stats/lm_head/lr_effective_min": -2.553903755142528e-07, "adam_stats/lm_head/lr_effective_std": 1.6296457516773444e-08, "adam_stats/lr_effective_max": 2.612372611565661e-07, "adam_stats/lr_effective_mean": 1.8011873459747307e-12, "adam_stats/lr_effective_min": -2.600495179194695e-07, "adam_stats/m_t_max": 3.459053914411925e-05, "adam_stats/m_t_mean": -6.274775620970785e-13, "adam_stats/m_t_min": -4.066598921781406e-05, "adam_stats/v_t_max": 1.5784487914061174e-05, "adam_stats/v_t_mean": 1.84523598657782e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.584883213043213, "all_logprobs/max": -5.125, "all_logprobs/median": -6.65625, "all_logprobs/min": -7.9375, "all_logprobs/p1": -7.46875, "all_logprobs/p10": -7.28125, "all_logprobs/p25": -7.09375, "all_logprobs/p5": -7.34375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.31593164801597595, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.07733333333333334, "feature_vector_variance/max_squared_error": 1915.662353515625, "feature_vector_variance/metric": 64.9424057006836, "generated_tokens/total": 4424207.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 6.7160365254926005e-06, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.0011138916015625, "num_completions/total": 4640, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.40347290039062, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.33633941411972046, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235679149627686, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.125815379054984e-05, "policy_entropy": 3.7738595008850098, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.00033711694413796067, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.03011305443942547, "policy_sharpness/max": 0.034495800733566284, "policy_sharpness/median": 0.030040517449378967, "policy_sharpness/min": 0.022274907678365707, "policy_sharpness/p25": 0.029363557696342468, "policy_sharpness/p75": 0.030818767845630646, "policy_sharpness/var": 1.2442003480828134e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 58, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 2.2163780499795394e-07, "adam_stats/lm_head/lr_effective_mean": -7.14148358504918e-13, "adam_stats/lm_head/lr_effective_min": -2.210690439596874e-07, "adam_stats/lm_head/lr_effective_std": 1.4105677514919535e-08, "adam_stats/lr_effective_max": 2.261303251316349e-07, "adam_stats/lr_effective_mean": 1.559135307839521e-12, "adam_stats/lr_effective_min": -2.2510137398512597e-07, "adam_stats/m_t_max": 3.113148341071792e-05, "adam_stats/m_t_mean": -5.647296866251317e-13, "adam_stats/m_t_min": -3.6599391023628414e-05, "adam_stats/v_t_max": 1.5768702724017203e-05, "adam_stats/v_t_mean": 1.8433902407993807e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.603366374969482, "all_logprobs/max": -5.15625, "all_logprobs/median": -6.6875, "all_logprobs/min": -7.90625, "all_logprobs/p1": -7.46875, "all_logprobs/p10": -7.3125, "all_logprobs/p25": -7.125, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.310659795999527, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.07866666666666666, "feature_vector_variance/max_squared_error": 1531.677734375, "feature_vector_variance/metric": 62.79999923706055, "generated_tokens/total": 4506127.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 6.456201742799511e-06, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.00115203857421875, "num_completions/total": 4720, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.3009796142578, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.3135417699813843, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.023533821105957, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.0192240551987197e-05, "policy_entropy": 3.7754693031311035, "policy_entropy/max": 3.828125, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.0003244230756536126, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.030036717653274536, "policy_sharpness/max": 0.03433223068714142, "policy_sharpness/median": 0.029962483793497086, "policy_sharpness/min": 0.02270258404314518, "policy_sharpness/p25": 0.0292864628136158, "policy_sharpness/p75": 0.030754800885915756, "policy_sharpness/var": 1.2494442671595607e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 59, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 1.915809093588905e-07, "adam_stats/lm_head/lr_effective_mean": -6.172557582249938e-13, "adam_stats/lm_head/lr_effective_min": -1.9108934168343694e-07, "adam_stats/lm_head/lr_effective_std": 1.2192137077704501e-08, "adam_stats/lr_effective_max": 1.9546438068118732e-07, "adam_stats/lr_effective_mean": 1.3477009222148673e-12, "adam_stats/lr_effective_min": -1.9457425537439121e-07, "adam_stats/m_t_max": 2.8018333978252485e-05, "adam_stats/m_t_mean": -5.08256739646662e-13, "adam_stats/m_t_min": -3.293944973847829e-05, "adam_stats/v_t_max": 1.5752933904877864e-05, "adam_stats/v_t_mean": 1.8415466634252864e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.601814270019531, "all_logprobs/max": -5.09375, "all_logprobs/median": -6.6875, "all_logprobs/min": -7.9375, "all_logprobs/p1": -7.5, "all_logprobs/p10": -7.3125, "all_logprobs/p25": -7.125, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.31268110871315, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.08, "feature_vector_variance/max_squared_error": 1805.6593017578125, "feature_vector_variance/metric": 65.9982681274414, "generated_tokens/total": 4588047.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 6.197638667498023e-06, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.002410888671875, "num_completions/total": 4800, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.3099365234375, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 132.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.33870929479599, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.023549199104309, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.0139128789887764e-05, "policy_entropy": 3.775456190109253, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.0003335291112307459, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.030032331123948097, "policy_sharpness/max": 0.03435549512505531, "policy_sharpness/median": 0.029963985085487366, "policy_sharpness/min": 0.020482441410422325, "policy_sharpness/p25": 0.029263203963637352, "policy_sharpness/p75": 0.03076530620455742, "policy_sharpness/var": 1.3237272469268646e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 60, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 1.6535601332634542e-07, "adam_stats/lm_head/lr_effective_mean": -5.327228458709699e-13, "adam_stats/lm_head/lr_effective_min": -1.6493181931309664e-07, "adam_stats/lm_head/lr_effective_std": 1.0522651194833088e-08, "adam_stats/lr_effective_max": 1.6870808394742198e-07, "adam_stats/lr_effective_mean": 1.1632226215238517e-12, "adam_stats/lr_effective_min": -1.6793919144220126e-07, "adam_stats/m_t_max": 2.5216499125235714e-05, "adam_stats/m_t_mean": -4.574309138936916e-13, "adam_stats/m_t_min": -2.9645503673236817e-05, "adam_stats/v_t_max": 1.5737181456643157e-05, "adam_stats/v_t_mean": 1.8397059049768405e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.598233222961426, "all_logprobs/max": -5.125, "all_logprobs/median": -6.6875, "all_logprobs/min": -7.90625, "all_logprobs/p1": -7.5, "all_logprobs/p10": -7.3125, "all_logprobs/p25": -7.125, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.31602099537849426, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.08133333333333333, "feature_vector_variance/max_squared_error": 1714.7489013671875, "feature_vector_variance/metric": 68.05635833740234, "generated_tokens/total": 4669967.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 5.9406623188668065e-06, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.0025634765625, "num_completions/total": 4880, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.30039978027344, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.3692229390144348, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235624313354492, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.0397507139714435e-05, "policy_entropy": 3.7747910022735596, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.0003403381851967424, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.030083125457167625, "policy_sharpness/max": 0.034461282193660736, "policy_sharpness/median": 0.030024010688066483, "policy_sharpness/min": 0.02190275676548481, "policy_sharpness/p25": 0.02933148294687271, "policy_sharpness/p75": 0.030819326639175415, "policy_sharpness/var": 1.3361640185394208e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 61, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 1.4250153412831423e-07, "adam_stats/lm_head/lr_effective_mean": -4.590599276578511e-13, "adam_stats/lm_head/lr_effective_min": -1.4213601673418452e-07, "adam_stats/lm_head/lr_effective_std": 9.067805173401666e-09, "adam_stats/lr_effective_max": 1.4539044457251293e-07, "adam_stats/lr_effective_mean": 1.002452244477925e-12, "adam_stats/lr_effective_min": -1.4472729503722803e-07, "adam_stats/m_t_max": 2.269484866701532e-05, "adam_stats/m_t_mean": -4.116879634506049e-13, "adam_stats/m_t_min": -2.6680952942115255e-05, "adam_stats/v_t_max": 1.572144537931308e-05, "adam_stats/v_t_mean": 1.8378655802092636e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.609178066253662, "all_logprobs/max": -5.09375, "all_logprobs/median": -6.6875, "all_logprobs/min": -7.9375, "all_logprobs/p1": -7.53125, "all_logprobs/p10": -7.3125, "all_logprobs/p25": -7.125, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.3151797652244568, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.08266666666666667, "feature_vector_variance/max_squared_error": 2085.67626953125, "feature_vector_variance/metric": 74.93392181396484, "generated_tokens/total": 4751887.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 5.685585783002493e-06, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.00494384765625, "num_completions/total": 4960, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.2446746826172, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 132.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.4260997474193573, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0234819650650024, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 9.561755177855957e-06, "policy_entropy": 3.776501417160034, "policy_entropy/max": 3.875, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.796875, "policy_entropy/var": 0.0003418064152356237, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.029938746243715286, "policy_sharpness/max": 0.034362245351076126, "policy_sharpness/median": 0.029920468106865883, "policy_sharpness/min": 0.01940051093697548, "policy_sharpness/p25": 0.029179921373724937, "policy_sharpness/p75": 0.030721787363290787, "policy_sharpness/var": 1.483550022385316e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 62, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 1.2260856863122171e-07, "adam_stats/lm_head/lr_effective_mean": -3.9494752954172263e-13, "adam_stats/lm_head/lr_effective_min": -1.2229412504893844e-07, "adam_stats/lm_head/lr_effective_std": 7.80155140489569e-09, "adam_stats/lr_effective_max": 1.2509431712715013e-07, "adam_stats/lr_effective_mean": 8.625148663864157e-13, "adam_stats/lr_effective_min": -1.245233107738386e-07, "adam_stats/m_t_max": 2.0425362890819088e-05, "adam_stats/m_t_mean": -3.705190451328e-13, "adam_stats/m_t_min": -2.401285746600479e-05, "adam_stats/v_t_max": 1.5705723853898235e-05, "adam_stats/v_t_mean": 1.836028508048204e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.609012126922607, "all_logprobs/max": -5.21875, "all_logprobs/median": -6.6875, "all_logprobs/min": -7.9375, "all_logprobs/p1": -7.5, "all_logprobs/p10": -7.3125, "all_logprobs/p25": -7.125, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.314659059047699, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.084, "feature_vector_variance/max_squared_error": 1751.0753173828125, "feature_vector_variance/metric": 68.21553039550781, "generated_tokens/total": 4833807.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 5.432719831372507e-06, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.002105712890625, "num_completions/total": 5040, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.27194213867188, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.3484311103820801, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235713720321655, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.0128454050573055e-05, "policy_entropy": 3.7752602100372314, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.0003226120606996119, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.030042892321944237, "policy_sharpness/max": 0.034215185791254044, "policy_sharpness/median": 0.02997749112546444, "policy_sharpness/min": 0.022166118025779724, "policy_sharpness/p25": 0.0292902160435915, "policy_sharpness/p75": 0.03077130764722824, "policy_sharpness/var": 1.2931285482409294e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 63, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 1.0531526584145467e-07, "adam_stats/lm_head/lr_effective_mean": -3.392175592070029e-13, "adam_stats/lm_head/lr_effective_min": -1.0504522407472905e-07, "adam_stats/lm_head/lr_effective_std": 6.7008363302534235e-09, "adam_stats/lr_effective_max": 1.0745053913296942e-07, "adam_stats/lr_effective_mean": 7.408642384472419e-13, "adam_stats/lr_effective_min": -1.0695966778939692e-07, "adam_stats/m_t_max": 1.838282696553506e-05, "adam_stats/m_t_mean": -3.3346714061952e-13, "adam_stats/m_t_min": -2.161157135560643e-05, "adam_stats/v_t_max": 1.569001869938802e-05, "adam_stats/v_t_mean": 1.8341923032488827e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.578835964202881, "all_logprobs/max": -5.0625, "all_logprobs/median": -6.65625, "all_logprobs/min": -7.84375, "all_logprobs/p1": -7.46875, "all_logprobs/p10": -7.28125, "all_logprobs/p25": -7.09375, "all_logprobs/p5": -7.34375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.3155173659324646, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.08533333333333333, "feature_vector_variance/max_squared_error": 1568.193359375, "feature_vector_variance/metric": 64.15740203857422, "generated_tokens/total": 4915727.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 5.182372542187895e-06, "loss": 0.0, "mean_logprobs": -6.5625, "mean_logprobs/var": 0.00157928466796875, "num_completions/total": 5120, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.4332733154297, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.33920562267303467, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235248804092407, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.1210155207663774e-05, "policy_entropy": 3.7737622261047363, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.0003406048926990479, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.030190763995051384, "policy_sharpness/max": 0.034664615988731384, "policy_sharpness/median": 0.030109543353319168, "policy_sharpness/min": 0.02279786951839924, "policy_sharpness/p25": 0.029438026249408722, "policy_sharpness/p75": 0.030892856419086456, "policy_sharpness/var": 1.212914867210202e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 64, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 9.03016754705277e-08, "adam_stats/lm_head/lr_effective_mean": -2.908382085511918e-13, "adam_stats/lm_head/lr_effective_min": -9.007015222550763e-08, "adam_stats/lm_head/lr_effective_std": 5.745276698831958e-09, "adam_stats/lr_effective_max": 9.21326233083164e-08, "adam_stats/lr_effective_mean": 6.352499906311926e-13, "adam_stats/lr_effective_min": -9.171140646913045e-08, "adam_stats/m_t_max": 1.6544543541385792e-05, "adam_stats/m_t_mean": -3.0012038861049195e-13, "adam_stats/m_t_min": -1.9450413674348965e-05, "adam_stats/v_t_max": 1.5674328096793033e-05, "adam_stats/v_t_mean": 1.8323573994921682e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.588417053222656, "all_logprobs/max": -5.09375, "all_logprobs/median": -6.65625, "all_logprobs/min": -7.9375, "all_logprobs/p1": -7.46875, "all_logprobs/p10": -7.28125, "all_logprobs/p25": -7.09375, "all_logprobs/p5": -7.34375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.31519269943237305, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.08666666666666667, "feature_vector_variance/max_squared_error": 1570.3988037109375, "feature_vector_variance/metric": 65.93915557861328, "generated_tokens/total": 4997647.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 4.934848925057485e-06, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.001495361328125, "num_completions/total": 5200, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.3943634033203, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.34370288252830505, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235329866409302, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.1207926036149729e-05, "policy_entropy": 3.7741305828094482, "policy_entropy/max": 3.828125, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.0003324282297398895, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.03013364039361477, "policy_sharpness/max": 0.03441851958632469, "policy_sharpness/median": 0.030058523640036583, "policy_sharpness/min": 0.022899160161614418, "policy_sharpness/p25": 0.029389256611466408, "policy_sharpness/p75": 0.03083883412182331, "policy_sharpness/var": 1.238551817550615e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 65, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 7.728508677473656e-08, "adam_stats/lm_head/lr_effective_mean": -2.4889714855119704e-13, "adam_stats/lm_head/lr_effective_min": -7.708696614372457e-08, "adam_stats/lm_head/lr_effective_std": 4.916865581350294e-09, "adam_stats/lr_effective_max": 7.885219588388281e-08, "adam_stats/lr_effective_mean": 5.436830998427344e-13, "adam_stats/lr_effective_min": -7.849140359894591e-08, "adam_stats/m_t_max": 1.4890089005348273e-05, "adam_stats/m_t_mean": -2.701083985385405e-13, "adam_stats/m_t_min": -1.7505371943116188e-05, "adam_stats/v_t_max": 1.565865386510268e-05, "adam_stats/v_t_mean": 1.8305250978206677e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.601449489593506, "all_logprobs/max": -5.125, "all_logprobs/median": -6.6875, "all_logprobs/min": -7.9375, "all_logprobs/p1": -7.5, "all_logprobs/p10": -7.3125, "all_logprobs/p25": -7.125, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.3141447901725769, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.088, "feature_vector_variance/max_squared_error": 1619.9930419921875, "feature_vector_variance/metric": 69.89456939697266, "generated_tokens/total": 5079567.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 4.6904505493806595e-06, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.00250244140625, "num_completions/total": 5280, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.31443786621094, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.36947372555732727, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.023571491241455, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.052705556503497e-05, "policy_entropy": 3.7744903564453125, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.00033157935831695795, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.030090490356087685, "policy_sharpness/max": 0.03431647643446922, "policy_sharpness/median": 0.030028510838747025, "policy_sharpness/min": 0.022408463060855865, "policy_sharpness/p25": 0.02932623028755188, "policy_sharpness/p75": 0.030823830515146255, "policy_sharpness/var": 1.3060770243100706e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 66, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 6.601597135613702e-08, "adam_stats/lm_head/lr_effective_mean": -2.1258937264647532e-13, "adam_stats/lm_head/lr_effective_min": -6.584677691989782e-08, "adam_stats/lm_head/lr_effective_std": 4.1997090249878966e-09, "adam_stats/lr_effective_max": 6.735466229201847e-08, "adam_stats/lr_effective_mean": 4.644094353870032e-13, "adam_stats/lr_effective_min": -6.704622990127973e-08, "adam_stats/m_t_max": 1.3401079741015565e-05, "adam_stats/m_t_mean": -2.4309750718508327e-13, "adam_stats/m_t_min": -1.575483474880457e-05, "adam_stats/v_t_max": 1.5642996004316956e-05, "adam_stats/v_t_mean": 1.8286953982343812e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.591681957244873, "all_logprobs/max": -5.1875, "all_logprobs/median": -6.65625, "all_logprobs/min": -7.96875, "all_logprobs/p1": -7.46875, "all_logprobs/p10": -7.28125, "all_logprobs/p25": -7.09375, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.31606385111808777, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.08933333333333333, "feature_vector_variance/max_squared_error": 2103.80859375, "feature_vector_variance/metric": 65.06818389892578, "generated_tokens/total": 5161487.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 4.4494751769315e-06, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.0016326904296875, "num_completions/total": 5360, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.3517608642578, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 132.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.336158812046051, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.023503303527832, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.0688941074477043e-05, "policy_entropy": 3.774672031402588, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.0003250479931011796, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.030128294602036476, "policy_sharpness/max": 0.03436299413442612, "policy_sharpness/median": 0.030054772272706032, "policy_sharpness/min": 0.021865246817469597, "policy_sharpness/p25": 0.029381753876805305, "policy_sharpness/p75": 0.03083508461713791, "policy_sharpness/var": 1.2289754067751346e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 67, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 5.6274306103887284e-08, "adam_stats/lm_head/lr_effective_mean": -1.8120548735556935e-13, "adam_stats/lm_head/lr_effective_min": -5.613010145566477e-08, "adam_stats/lm_head/lr_effective_std": 3.5797920183711085e-09, "adam_stats/lr_effective_max": 5.741550879179158e-08, "adam_stats/lr_effective_mean": 3.9587988919995154e-13, "adam_stats/lr_effective_min": -5.71523806058849e-08, "adam_stats/m_t_max": 1.2060971130267717e-05, "adam_stats/m_t_mean": -2.1878778357162926e-13, "adam_stats/m_t_min": -1.4179350728227291e-05, "adam_stats/v_t_max": 1.562735269544646e-05, "adam_stats/v_t_mean": 1.8268661323289637e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.585099697113037, "all_logprobs/max": -5.15625, "all_logprobs/median": -6.65625, "all_logprobs/min": -7.9375, "all_logprobs/p1": -7.46875, "all_logprobs/p10": -7.28125, "all_logprobs/p25": -7.09375, "all_logprobs/p5": -7.34375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.3151557445526123, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.09066666666666667, "feature_vector_variance/max_squared_error": 1706.2054443359375, "feature_vector_variance/metric": 63.61259078979492, "generated_tokens/total": 5243407.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 4.212216399081919e-06, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.000946044921875, "num_completions/total": 5440, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.38661193847656, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.32650187611579895, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235732793807983, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.123587935580872e-05, "policy_entropy": 3.7736313343048096, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.0003337883681524545, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.03019077517092228, "policy_sharpness/max": 0.03478541225194931, "policy_sharpness/median": 0.03011329658329487, "policy_sharpness/min": 0.02252926118671894, "policy_sharpness/p25": 0.029439525678753853, "policy_sharpness/p75": 0.030897358432412148, "policy_sharpness/var": 1.241254835804284e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 68, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 4.7866166141830035e-08, "adam_stats/lm_head/lr_effective_mean": -1.5411960986935758e-13, "adam_stats/lm_head/lr_effective_min": -4.774352646563784e-08, "adam_stats/lm_head/lr_effective_std": 3.0447646626186042e-09, "adam_stats/lr_effective_max": 4.8836909627425484e-08, "adam_stats/lr_effective_mean": 3.3673145652043934e-13, "adam_stats/lr_effective_min": -4.861291102997711e-08, "adam_stats/m_t_max": 1.0854873835342005e-05, "adam_stats/m_t_mean": -1.96908949649105e-13, "adam_stats/m_t_min": -1.2761415746354032e-05, "adam_stats/v_t_max": 1.56117257574806e-05, "adam_stats/v_t_mean": 1.825039902189629e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.6070756912231445, "all_logprobs/max": -5.125, "all_logprobs/median": -6.6875, "all_logprobs/min": -7.96875, "all_logprobs/p1": -7.5, "all_logprobs/p10": -7.3125, "all_logprobs/p25": -7.125, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.3138926923274994, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.092, "feature_vector_variance/max_squared_error": 1624.0460205078125, "feature_vector_variance/metric": 72.10182189941406, "generated_tokens/total": 5325327.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 3.978963279105821e-06, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.0028839111328125, "num_completions/total": 5520, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.28521728515625, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.3824124336242676, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0234911441802979, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 9.863378181762528e-06, "policy_entropy": 3.7758119106292725, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.00032703299075365067, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.029994552955031395, "policy_sharpness/max": 0.034526560455560684, "policy_sharpness/median": 0.029934724792838097, "policy_sharpness/min": 0.02204832062125206, "policy_sharpness/p25": 0.029237695038318634, "policy_sharpness/p75": 0.03073679469525814, "policy_sharpness/var": 1.341354845862952e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 69, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 4.06208613412673e-08, "adam_stats/lm_head/lr_effective_mean": -1.3078177863584672e-13, "adam_stats/lm_head/lr_effective_min": -4.0516798804901555e-08, "adam_stats/lm_head/lr_effective_std": 2.5837572081854887e-09, "adam_stats/lr_effective_max": 4.1444710774385385e-08, "adam_stats/lr_effective_mean": 2.8576259739586385e-13, "adam_stats/lr_effective_min": -4.1254470062312976e-08, "adam_stats/m_t_max": 9.769386451807804e-06, "adam_stats/m_t_mean": -1.7721806281571079e-13, "adam_stats/m_t_min": -1.1485273716971278e-05, "adam_stats/v_t_max": 1.5596113371429965e-05, "adam_stats/v_t_mean": 1.823213888890729e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.583874702453613, "all_logprobs/max": -5.0625, "all_logprobs/median": -6.65625, "all_logprobs/min": -7.90625, "all_logprobs/p1": -7.5, "all_logprobs/p10": -7.28125, "all_logprobs/p25": -7.09375, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.3195229172706604, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.09333333333333334, "feature_vector_variance/max_squared_error": 1580.0604248046875, "feature_vector_variance/metric": 71.66973114013672, "generated_tokens/total": 5407247.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 3.750000000000002e-06, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.0034637451171875, "num_completions/total": 5600, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.4146728515625, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.3864253759384155, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235483646392822, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.1272547453700099e-05, "policy_entropy": 3.7732796669006348, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.0003388439945410937, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.030186036601662636, "policy_sharpness/max": 0.034351740032434464, "policy_sharpness/median": 0.030131302773952484, "policy_sharpness/min": 0.022569777444005013, "policy_sharpness/p25": 0.0294357743114233, "policy_sharpness/p75": 0.030921367928385735, "policy_sharpness/var": 1.3229314390628133e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 70, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 3.438830376012447e-08, "adam_stats/lm_head/lr_effective_mean": -1.1070758567689742e-13, "adam_stats/lm_head/lr_effective_min": -3.430022488259965e-08, "adam_stats/lm_head/lr_effective_std": 2.187211745052764e-09, "adam_stats/lr_effective_max": 3.50857867204013e-08, "adam_stats/lr_effective_mean": 2.419182475871251e-13, "adam_stats/lr_effective_min": -3.492461075893516e-08, "adam_stats/m_t_max": 8.792447260930203e-06, "adam_stats/m_t_mean": -1.5949624975787613e-13, "adam_stats/m_t_min": -1.0336745617678389e-05, "adam_stats/v_t_max": 1.5580517356283963e-05, "adam_stats/v_t_mean": 1.8213911281983464e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.5927042961120605, "all_logprobs/max": -5.09375, "all_logprobs/median": -6.6875, "all_logprobs/min": -7.9375, "all_logprobs/p1": -7.46875, "all_logprobs/p10": -7.28125, "all_logprobs/p25": -7.09375, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.3159146308898926, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.09466666666666666, "feature_vector_variance/max_squared_error": 1438.659423828125, "feature_vector_variance/metric": 65.45097351074219, "generated_tokens/total": 5489167.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 3.525605518250964e-06, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.00153350830078125, "num_completions/total": 5680, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.34950256347656, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.3365570306777954, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.023535132408142, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.0779105650726706e-05, "policy_entropy": 3.7740724086761475, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.00032912520691752434, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.030104169622063637, "policy_sharpness/max": 0.03469537943601608, "policy_sharpness/median": 0.030039764940738678, "policy_sharpness/min": 0.02266206592321396, "policy_sharpness/p25": 0.029351741075515747, "policy_sharpness/p75": 0.030823830515146255, "policy_sharpness/var": 1.279726006941928e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 71, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 2.9036618443001316e-08, "adam_stats/lm_head/lr_effective_mean": -9.347189499188738e-14, "adam_stats/lm_head/lr_effective_min": -2.896226192206086e-08, "adam_stats/lm_head/lr_effective_std": 1.8467308793290726e-09, "adam_stats/lr_effective_max": 2.9625589093029703e-08, "adam_stats/lr_effective_mean": 2.0427031648204835e-13, "adam_stats/lr_effective_min": -2.948938693236869e-08, "adam_stats/m_t_max": 7.913202352938242e-06, "adam_stats/m_t_mean": -1.435466329136048e-13, "adam_stats/m_t_min": -9.30307123780949e-06, "adam_stats/v_t_max": 1.5564937712042592e-05, "adam_stats/v_t_mean": 1.8195692348677017e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.579585552215576, "all_logprobs/max": -5.03125, "all_logprobs/median": -6.65625, "all_logprobs/min": -7.96875, "all_logprobs/p1": -7.46875, "all_logprobs/p10": -7.28125, "all_logprobs/p25": -7.09375, "all_logprobs/p5": -7.34375, "all_logprobs/p75": -6.09375, "all_logprobs/var": 0.31798455119132996, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.096, "feature_vector_variance/max_squared_error": 1751.6123046875, "feature_vector_variance/metric": 67.36102294921875, "generated_tokens/total": 5571087.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 3.3060532239694e-06, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.001708984375, "num_completions/total": 5760, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.42483520507812, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.3532153069972992, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235451459884644, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.1282226296316367e-05, "policy_entropy": 3.77348256111145, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.0003378770488779992, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.03025258518755436, "policy_sharpness/max": 0.034494295716285706, "policy_sharpness/median": 0.030170317739248276, "policy_sharpness/min": 0.021418068557977676, "policy_sharpness/p25": 0.02950161322951317, "policy_sharpness/p75": 0.030953630805015564, "policy_sharpness/var": 1.2164433655925677e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 72, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 2.4450073965454067e-08, "adam_stats/lm_head/lr_effective_mean": -7.87016393895279e-14, "adam_stats/lm_head/lr_effective_min": -2.4387471597719923e-08, "adam_stats/lm_head/lr_effective_std": 1.5549455056884653e-09, "adam_stats/lr_effective_max": 2.4946038124085135e-08, "adam_stats/lr_effective_mean": 1.7200501548249314e-13, "adam_stats/lr_effective_min": -2.4831258826907288e-08, "adam_stats/m_t_max": 7.121881935745478e-06, "adam_stats/m_t_mean": -1.2919201299033123e-13, "adam_stats/m_t_min": -8.37276365928119e-06, "adam_stats/v_t_max": 1.554937261971645e-05, "adam_stats/v_t_mean": 1.81775037730314e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.585881233215332, "all_logprobs/max": -5.0625, "all_logprobs/median": -6.65625, "all_logprobs/min": -7.9375, "all_logprobs/p1": -7.46875, "all_logprobs/p10": -7.28125, "all_logprobs/p25": -7.09375, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.31680771708488464, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.09733333333333333, "feature_vector_variance/max_squared_error": 1532.70703125, "feature_vector_variance/metric": 64.43016815185547, "generated_tokens/total": 5653007.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 3.0916106078064522e-06, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.00133514404296875, "num_completions/total": 5840, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.3801727294922, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.32827043533325195, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235717296600342, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.1111133062513545e-05, "policy_entropy": 3.773587465286255, "policy_entropy/max": 3.828125, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.00032255571568384767, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.03013637661933899, "policy_sharpness/max": 0.03423244133591652, "policy_sharpness/median": 0.030061524361371994, "policy_sharpness/min": 0.02225315384566784, "policy_sharpness/p25": 0.029389256611466408, "policy_sharpness/p75": 0.030847840011119843, "policy_sharpness/var": 1.2227590104885167e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 73, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 2.0527201272102502e-08, "adam_stats/lm_head/lr_effective_mean": -6.60696269829536e-14, "adam_stats/lm_head/lr_effective_min": -2.0474653084079364e-08, "adam_stats/lm_head/lr_effective_std": 1.3053957914621606e-09, "adam_stats/lr_effective_max": 2.094361484239471e-08, "adam_stats/lr_effective_mean": 1.444082619326062e-13, "adam_stats/lr_effective_min": -2.084717465322683e-08, "adam_stats/m_t_max": 6.409693469322519e-06, "adam_stats/m_t_mean": -1.1627280898079267e-13, "adam_stats/m_t_min": -7.535486929555191e-06, "adam_stats/v_t_max": 1.553382389829494e-05, "adam_stats/v_t_mean": 1.815932712360968e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.5663275718688965, "all_logprobs/max": -5.0625, "all_logprobs/median": -6.65625, "all_logprobs/min": -7.90625, "all_logprobs/p1": -7.46875, "all_logprobs/p10": -7.28125, "all_logprobs/p25": -7.09375, "all_logprobs/p5": -7.34375, "all_logprobs/p75": -6.09375, "all_logprobs/var": 0.31763166189193726, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.09866666666666667, "feature_vector_variance/max_squared_error": 1833.416748046875, "feature_vector_variance/metric": 65.08281707763672, "generated_tokens/total": 5734927.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 2.882538935057563e-06, "loss": 0.0, "mean_logprobs": -6.5625, "mean_logprobs/var": 0.0012969970703125, "num_completions/total": 5920, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.48304748535156, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 137.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.3421729505062103, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.023574709892273, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.1844197615573648e-05, "policy_entropy": 3.772475481033325, "policy_entropy/max": 3.828125, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.0003392958897165954, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.030277907848358154, "policy_sharpness/max": 0.03450329974293709, "policy_sharpness/median": 0.030196577310562134, "policy_sharpness/min": 0.020622748881578445, "policy_sharpness/p25": 0.02952881157398224, "policy_sharpness/p75": 0.03097539022564888, "policy_sharpness/var": 1.2088603398296982e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 74, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 1.7179145217482983e-08, "adam_stats/lm_head/lr_effective_mean": -5.52894454395117e-14, "adam_stats/lm_head/lr_effective_min": -1.713517328028047e-08, "adam_stats/lm_head/lr_effective_std": 1.0924245952281808e-09, "adam_stats/lr_effective_max": 1.7527655771232276e-08, "adam_stats/lr_effective_mean": 1.2085523012038413e-13, "adam_stats/lr_effective_min": -1.7446883049387907e-08, "adam_stats/m_t_max": 5.768723895016592e-06, "adam_stats/m_t_mean": -1.0464551317493354e-13, "adam_stats/m_t_min": -6.781938282074407e-06, "adam_stats/v_t_max": 1.551828972878866e-05, "adam_stats/v_t_mean": 1.8141159147805341e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.587189674377441, "all_logprobs/max": -5.15625, "all_logprobs/median": -6.65625, "all_logprobs/min": -7.9375, "all_logprobs/p1": -7.46875, "all_logprobs/p10": -7.28125, "all_logprobs/p25": -7.09375, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.31642600893974304, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.1, "feature_vector_variance/max_squared_error": 1812.1451416015625, "feature_vector_variance/metric": 68.83256530761719, "generated_tokens/total": 5816847.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 2.6790929273509547e-06, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.00213623046875, "num_completions/total": 6000, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.3862762451172, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 132.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.3658563792705536, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235726833343506, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.102888745663222e-05, "policy_entropy": 3.7739205360412598, "policy_entropy/max": 3.859375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.0003350348270032555, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.030149126425385475, "policy_sharpness/max": 0.034468039870262146, "policy_sharpness/median": 0.03007803112268448, "policy_sharpness/min": 0.021532859653234482, "policy_sharpness/p25": 0.02939525805413723, "policy_sharpness/p75": 0.03086734563112259, "policy_sharpness/var": 1.2892280665255385e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 75, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 1.4328173314481774e-08, "adam_stats/lm_head/lr_effective_mean": -4.611051530835006e-14, "adam_stats/lm_head/lr_effective_min": -1.4291504868424454e-08, "adam_stats/lm_head/lr_effective_std": 9.110834309211668e-10, "adam_stats/lr_effective_max": 1.4618863453108588e-08, "adam_stats/lr_effective_mean": 1.0079898070738946e-13, "adam_stats/lr_effective_min": -1.4551439164733893e-08, "adam_stats/m_t_max": 5.191851414565463e-06, "adam_stats/m_t_mean": -9.418089545005712e-14, "adam_stats/m_t_min": -6.103744453866966e-06, "adam_stats/v_t_max": 1.550277193018701e-05, "adam_stats/v_t_mean": 1.8123028034874866e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.585793972015381, "all_logprobs/max": -5.0625, "all_logprobs/median": -6.65625, "all_logprobs/min": -7.90625, "all_logprobs/p1": -7.46875, "all_logprobs/p10": -7.28125, "all_logprobs/p25": -7.09375, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.3194011151790619, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.10133333333333333, "feature_vector_variance/max_squared_error": 1476.0880126953125, "feature_vector_variance/metric": 68.16045379638672, "generated_tokens/total": 5898767.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 2.4815204523085656e-06, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.0023651123046875, "num_completions/total": 6080, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.3944854736328, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.3546168804168701, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235857963562012, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.1284452739346307e-05, "policy_entropy": 3.7733840942382812, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.0003301930264569819, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.030154744163155556, "policy_sharpness/max": 0.034315723925828934, "policy_sharpness/median": 0.030090035870671272, "policy_sharpness/min": 0.022981693968176842, "policy_sharpness/p25": 0.02939225547015667, "policy_sharpness/p75": 0.030885355547070503, "policy_sharpness/var": 1.27299915675394e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 76, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 1.1906371000236504e-08, "adam_stats/lm_head/lr_effective_mean": -3.831395988776638e-14, "adam_stats/lm_head/lr_effective_min": -1.187590559226237e-08, "adam_stats/lm_head/lr_effective_std": 7.570494231501357e-10, "adam_stats/lr_effective_max": 1.2147941319540223e-08, "adam_stats/lr_effective_mean": 8.37618413214794e-14, "adam_stats/lr_effective_min": -1.2091868839547715e-08, "adam_stats/m_t_max": 4.6726663640583865e-06, "adam_stats/m_t_mean": -8.476287841107169e-14, "adam_stats/m_t_min": -5.493369826581329e-06, "adam_stats/v_t_max": 1.5487268683500588e-05, "adam_stats/v_t_mean": 1.8104904511359599e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.629826545715332, "all_logprobs/max": -5.125, "all_logprobs/median": -6.6875, "all_logprobs/min": -7.9375, "all_logprobs/p1": -7.53125, "all_logprobs/p10": -7.34375, "all_logprobs/p25": -7.125, "all_logprobs/p5": -7.40625, "all_logprobs/p75": -6.15625, "all_logprobs/var": 0.3141312599182129, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.10266666666666667, "feature_vector_variance/max_squared_error": 1867.837646484375, "feature_vector_variance/metric": 82.34205627441406, "generated_tokens/total": 5980687.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 2.29006222155752e-06, "loss": 0.0, "mean_logprobs": -6.625, "mean_logprobs/var": 0.00701904296875, "num_completions/total": 6160, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.17308044433594, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 132.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.48028144240379333, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0234942436218262, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 8.819165486784186e-06, "policy_entropy": 3.7774364948272705, "policy_entropy/max": 3.859375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.796875, "policy_entropy/var": 0.0003345465811435133, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.02984699420630932, "policy_sharpness/max": 0.03444628044962883, "policy_sharpness/median": 0.029822930693626404, "policy_sharpness/min": 0.02073529362678528, "policy_sharpness/p25": 0.029046369716525078, "policy_sharpness/p75": 0.030652010813355446, "policy_sharpness/var": 1.5453047126356978e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 77, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 9.854473681514264e-09, "adam_stats/lm_head/lr_effective_mean": -3.170880035320914e-14, "adam_stats/lm_head/lr_effective_min": -9.829262737071076e-09, "adam_stats/lm_head/lr_effective_std": 6.265499230551086e-10, "adam_stats/lr_effective_max": 1.0054423071892415e-08, "adam_stats/lr_effective_mean": 6.932684813590675e-14, "adam_stats/lr_effective_min": -1.0007978445969457e-08, "adam_stats/m_t_max": 4.205399818602018e-06, "adam_stats/m_t_mean": -7.628659937910717e-14, "adam_stats/m_t_min": -4.944032752973726e-06, "adam_stats/v_t_max": 1.54717818077188e-05, "adam_stats/v_t_mean": 1.8086800503483436e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.592912197113037, "all_logprobs/max": -5.125, "all_logprobs/median": -6.65625, "all_logprobs/min": -7.90625, "all_logprobs/p1": -7.5, "all_logprobs/p10": -7.3125, "all_logprobs/p25": -7.09375, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.31759241223335266, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.104, "feature_vector_variance/max_squared_error": 2048.390869140625, "feature_vector_variance/metric": 67.53319549560547, "generated_tokens/total": 6062607.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 2.104951497460118e-06, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.0019378662109375, "num_completions/total": 6240, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.35899353027344, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 132.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.3486977815628052, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235365629196167, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.0784777259686962e-05, "policy_entropy": 3.774231433868408, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.000324415770592168, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.03009038046002388, "policy_sharpness/max": 0.034387003630399704, "policy_sharpness/median": 0.030016504228115082, "policy_sharpness/min": 0.021045921370387077, "policy_sharpness/p25": 0.029334483668208122, "policy_sharpness/p75": 0.030812012031674385, "policy_sharpness/var": 1.2639794704227825e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 78, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 8.120824901425294e-09, "adam_stats/lm_head/lr_effective_mean": -2.6128516803511706e-14, "adam_stats/lm_head/lr_effective_min": -8.100052184545348e-09, "adam_stats/lm_head/lr_effective_std": 5.162972271932631e-10, "adam_stats/lr_effective_max": 8.285605979097e-09, "adam_stats/lr_effective_mean": 5.713073921278025e-14, "adam_stats/lr_effective_min": -8.247301508390592e-09, "adam_stats/m_t_max": 3.784859700317611e-06, "adam_stats/m_t_mean": -6.865791572427393e-14, "adam_stats/m_t_min": -4.449629159353208e-06, "adam_stats/v_t_max": 1.5456309483852237e-05, "adam_stats/v_t_mean": 1.806871709544855e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.570004940032959, "all_logprobs/max": -5.125, "all_logprobs/median": -6.65625, "all_logprobs/min": -7.875, "all_logprobs/p1": -7.46875, "all_logprobs/p10": -7.28125, "all_logprobs/p25": -7.09375, "all_logprobs/p5": -7.34375, "all_logprobs/p75": -6.09375, "all_logprobs/var": 0.3194361627101898, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.10533333333333333, "feature_vector_variance/max_squared_error": 1518.0723876953125, "feature_vector_variance/metric": 63.574825286865234, "generated_tokens/total": 6144527.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.9264138089195424e-06, "loss": 0.0, "mean_logprobs": -6.5625, "mean_logprobs/var": 0.000789642333984375, "num_completions/total": 6320, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.4622039794922, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.3353680968284607, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235745906829834, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.1862872270285152e-05, "policy_entropy": 3.7725212574005127, "policy_entropy/max": 3.828125, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.0003358829126227647, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.030294854193925858, "policy_sharpness/max": 0.0345025509595871, "policy_sharpness/median": 0.03022058866918087, "policy_sharpness/min": 0.022338686510920525, "policy_sharpness/p25": 0.029552070423960686, "policy_sharpness/p75": 0.030985895544290543, "policy_sharpness/var": 1.1947057600991684e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 79, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 6.660462403829115e-09, "adam_stats/lm_head/lr_effective_mean": -2.1428297097879703e-14, "adam_stats/lm_head/lr_effective_min": -6.6434271417392665e-09, "adam_stats/lm_head/lr_effective_std": 4.234297912297791e-10, "adam_stats/lr_effective_max": 6.7956178462225125e-09, "adam_stats/lr_effective_mean": 4.685709353619179e-14, "adam_stats/lr_effective_min": -6.764176774254338e-09, "adam_stats/m_t_max": 3.4063737075484823e-06, "adam_stats/m_t_mean": -6.179211059931938e-14, "adam_stats/m_t_min": -4.004666152468417e-06, "adam_stats/v_t_max": 1.544085353089031e-05, "adam_stats/v_t_mean": 1.8050639108424527e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.573646068572998, "all_logprobs/max": -5.125, "all_logprobs/median": -6.65625, "all_logprobs/min": -8.0, "all_logprobs/p1": -7.46875, "all_logprobs/p10": -7.28125, "all_logprobs/p25": -7.09375, "all_logprobs/p5": -7.34375, "all_logprobs/p75": -6.09375, "all_logprobs/var": 0.31826385855674744, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.10666666666666667, "feature_vector_variance/max_squared_error": 1866.6055908203125, "feature_vector_variance/metric": 63.986610412597656, "generated_tokens/total": 6226447.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.7546666766076658e-06, "loss": 0.0, "mean_logprobs": -6.5625, "mean_logprobs/var": 0.00109100341796875, "num_completions/total": 6400, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.44346618652344, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.33218467235565186, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235294103622437, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.1624370017671026e-05, "policy_entropy": 3.7732772827148438, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.00033521887962706387, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.030230853706598282, "policy_sharpness/max": 0.03444552794098854, "policy_sharpness/median": 0.03014555759727955, "policy_sharpness/min": 0.022375449538230896, "policy_sharpness/p25": 0.029483793303370476, "policy_sharpness/p75": 0.030924370512366295, "policy_sharpness/var": 1.2014810408800258e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 80, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 5.434305450080501e-09, "adam_stats/lm_head/lr_effective_mean": -1.748218235485873e-14, "adam_stats/lm_head/lr_effective_min": -5.420408566436663e-09, "adam_stats/lm_head/lr_effective_std": 3.454605768116181e-10, "adam_stats/lr_effective_max": 5.5445847912949375e-09, "adam_stats/lr_effective_mean": 3.823109634860239e-14, "adam_stats/lr_effective_min": -5.5189115499842956e-09, "adam_stats/m_t_max": 3.065736336793634e-06, "adam_stats/m_t_mean": -5.5612886664486647e-14, "adam_stats/m_t_min": -3.60419949174684e-06, "adam_stats/v_t_max": 1.5425412129843608e-05, "adam_stats/v_t_mean": 1.803259906847654e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.577882289886475, "all_logprobs/max": -5.125, "all_logprobs/median": -6.65625, "all_logprobs/min": -7.90625, "all_logprobs/p1": -7.46875, "all_logprobs/p10": -7.28125, "all_logprobs/p25": -7.09375, "all_logprobs/p5": -7.34375, "all_logprobs/p75": -6.09375, "all_logprobs/var": 0.3171963691711426, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.108, "feature_vector_variance/max_squared_error": 1962.2003173828125, "feature_vector_variance/metric": 64.94758605957031, "generated_tokens/total": 6308367.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.5899193479495858e-06, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.00133514404296875, "num_completions/total": 6480, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.4412384033203, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 132.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.3391050398349762, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235605239868164, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.1505161637614947e-05, "policy_entropy": 3.7729904651641846, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.0003348955069668591, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.030253922566771507, "policy_sharpness/max": 0.03478466346859932, "policy_sharpness/median": 0.030175570398569107, "policy_sharpness/min": 0.021504350006580353, "policy_sharpness/p25": 0.029507802799344063, "policy_sharpness/p75": 0.030951378867030144, "policy_sharpness/var": 1.2167141676400206e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 81, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 4.408431841795846e-09, "adam_stats/lm_head/lr_effective_mean": -1.418091972381309e-14, "adam_stats/lm_head/lr_effective_min": -4.397159969471431e-09, "adam_stats/lm_head/lr_effective_std": 2.8023086628969907e-10, "adam_stats/lr_effective_max": 4.497898054012239e-09, "adam_stats/lr_effective_mean": 3.1014019884157903e-14, "adam_stats/lr_effective_min": -4.4770551710371365e-09, "adam_stats/m_t_max": 2.7591627258516382e-06, "adam_stats/m_t_mean": -5.0051596981598445e-14, "adam_stats/m_t_min": -3.2437794743600534e-06, "adam_stats/v_t_max": 1.540998709970154e-05, "adam_stats/v_t_mean": 1.8014553607517692e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.598460674285889, "all_logprobs/max": -5.03125, "all_logprobs/median": -6.6875, "all_logprobs/min": -8.0, "all_logprobs/p1": -7.5, "all_logprobs/p10": -7.3125, "all_logprobs/p25": -7.09375, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.3143330216407776, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.10933333333333334, "feature_vector_variance/max_squared_error": 1867.274169921875, "feature_vector_variance/metric": 71.74384307861328, "generated_tokens/total": 6390287.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.432372542187895e-06, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.0029144287109375, "num_completions/total": 6560, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.34300231933594, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 132.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.3802904486656189, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235480070114136, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.0305549949407578e-05, "policy_entropy": 3.774808645248413, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.0003297295479569584, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.03004644624888897, "policy_sharpness/max": 0.034489795565605164, "policy_sharpness/median": 0.02998049184679985, "policy_sharpness/min": 0.02232292853295803, "policy_sharpness/p25": 0.029282713308930397, "policy_sharpness/p75": 0.030785564333200455, "policy_sharpness/var": 1.3185763236833736e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 82, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 3.55344220714926e-09, "adam_stats/lm_head/lr_effective_mean": -1.1429779581020698e-14, "adam_stats/lm_head/lr_effective_min": -3.544358584406382e-09, "adam_stats/lm_head/lr_effective_std": 2.2587000325646756e-10, "adam_stats/lr_effective_max": 3.6255609625612806e-09, "adam_stats/lr_effective_mean": 2.499911187076303e-14, "adam_stats/lr_effective_min": -3.6087473009871474e-09, "adam_stats/m_t_max": 2.4832463623170042e-06, "adam_stats/m_t_mean": -4.5046438638691316e-14, "adam_stats/m_t_min": -2.9194013677624753e-06, "adam_stats/v_t_max": 1.5394576621474698e-05, "adam_stats/v_t_mean": 1.7996549346241397e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.60500955581665, "all_logprobs/max": -5.0625, "all_logprobs/median": -6.6875, "all_logprobs/min": -8.0, "all_logprobs/p1": -7.53125, "all_logprobs/p10": -7.3125, "all_logprobs/p25": -7.125, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.31698697805404663, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.11066666666666666, "feature_vector_variance/max_squared_error": 1934.6044921875, "feature_vector_variance/metric": 78.90301513671875, "generated_tokens/total": 6472207.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.282218205837188e-06, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.006011962890625, "num_completions/total": 6640, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.28866577148438, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 132.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.4552677869796753, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235459804534912, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.0212850611424074e-05, "policy_entropy": 3.775162935256958, "policy_entropy/max": 3.859375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.00034181022783741355, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.030042096972465515, "policy_sharpness/max": 0.03467737138271332, "policy_sharpness/median": 0.03001200407743454, "policy_sharpness/min": 0.0210136529058218, "policy_sharpness/p25": 0.029299220070242882, "policy_sharpness/p75": 0.0308088231831789, "policy_sharpness/var": 1.4596281516787712e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 83, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 2.843897339843693e-09, "adam_stats/lm_head/lr_effective_mean": -9.146837746856068e-15, "adam_stats/lm_head/lr_effective_min": -2.836628487656867e-09, "adam_stats/lm_head/lr_effective_std": 1.8075928254202012e-10, "adam_stats/lr_effective_max": 2.9016185010277695e-09, "adam_stats/lr_effective_mean": 2.0007402716992403e-14, "adam_stats/lr_effective_min": -2.8881517177836713e-09, "adam_stats/m_t_max": 2.234921566923731e-06, "adam_stats/m_t_mean": -4.0541781899921386e-14, "adam_stats/m_t_min": -2.627461071824655e-06, "adam_stats/v_t_max": 1.537918251415249e-05, "adam_stats/v_t_mean": 1.797854833757162e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.606201171875, "all_logprobs/max": -5.125, "all_logprobs/median": -6.6875, "all_logprobs/min": -8.0, "all_logprobs/p1": -7.53125, "all_logprobs/p10": -7.3125, "all_logprobs/p25": -7.125, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.31819337606430054, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.112, "feature_vector_variance/max_squared_error": 1844.352294921875, "feature_vector_variance/metric": 75.85870361328125, "generated_tokens/total": 6554127.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.1396392788268054e-06, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.004119873046875, "num_completions/total": 6720, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.27932739257812, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 132.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.41334113478660583, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235854387283325, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.0265315722790547e-05, "policy_entropy": 3.7745628356933594, "policy_entropy/max": 3.859375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.0003252574533689767, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.03004641644656658, "policy_sharpness/max": 0.03454381600022316, "policy_sharpness/median": 0.029992496594786644, "policy_sharpness/min": 0.021751200780272484, "policy_sharpness/p25": 0.02929171547293663, "policy_sharpness/p75": 0.030777310952544212, "policy_sharpness/var": 1.342150426353328e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 84, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 2.2578203751066894e-09, "adam_stats/lm_head/lr_effective_mean": -7.261310219016e-15, "adam_stats/lm_head/lr_effective_min": -2.2520507680923174e-09, "adam_stats/lm_head/lr_effective_std": 1.4350053090250725e-10, "adam_stats/lr_effective_max": 2.3036490492955863e-09, "adam_stats/lr_effective_mean": 1.58842834989837e-14, "adam_stats/lr_effective_min": -2.292948719784249e-09, "adam_stats/m_t_max": 2.0114293874939904e-06, "adam_stats/m_t_mean": -3.648763149260992e-14, "adam_stats/m_t_min": -2.3647148736927193e-06, "adam_stats/v_t_max": 1.536380295874551e-05, "adam_stats/v_t_mean": 1.79605766023605e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.594354152679443, "all_logprobs/max": -5.125, "all_logprobs/median": -6.6875, "all_logprobs/min": -7.96875, "all_logprobs/p1": -7.5, "all_logprobs/p10": -7.28125, "all_logprobs/p25": -7.09375, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.3173149526119232, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.11333333333333333, "feature_vector_variance/max_squared_error": 1951.5074462890625, "feature_vector_variance/metric": 73.73546600341797, "generated_tokens/total": 6636047.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.0048094716167097e-06, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.003997802734375, "num_completions/total": 6800, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.36451721191406, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 132.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.4160228371620178, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235698223114014, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.0891112651734147e-05, "policy_entropy": 3.7741219997406006, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.00034329359186813235, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.03009648434817791, "policy_sharpness/max": 0.03443277254700661, "policy_sharpness/median": 0.03004351817071438, "policy_sharpness/min": 0.02125149965286255, "policy_sharpness/p25": 0.02932548150420189, "policy_sharpness/p75": 0.030850838869810104, "policy_sharpness/var": 1.3768006965619861e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 85, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 1.7762598059078982e-09, "adam_stats/lm_head/lr_effective_mean": -5.7121640384860846e-15, "adam_stats/lm_head/lr_effective_min": -1.771721325205533e-09, "adam_stats/lm_head/lr_effective_std": 1.1288805307207994e-10, "adam_stats/lr_effective_max": 1.8123155198779273e-09, "adam_stats/lr_effective_mean": 1.2496426995674357e-14, "adam_stats/lr_effective_min": -1.8038909255224667e-09, "adam_stats/m_t_max": 1.8102864487445913e-06, "adam_stats/m_t_mean": -3.283886427759078e-14, "adam_stats/m_t_min": -2.128243295373977e-06, "adam_stats/v_t_max": 1.534843977424316e-05, "adam_stats/v_t_mean": 1.7942610288160243e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.650525093078613, "all_logprobs/max": -5.0625, "all_logprobs/median": -6.71875, "all_logprobs/min": -8.25, "all_logprobs/p1": -7.84375, "all_logprobs/p10": -7.34375, "all_logprobs/p25": -7.15625, "all_logprobs/p5": -7.5, "all_logprobs/p75": -6.15625, "all_logprobs/var": 0.35743987560272217, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.11466666666666667, "feature_vector_variance/max_squared_error": 3549.874755859375, "feature_vector_variance/metric": 201.3330535888672, "generated_tokens/total": 6717967.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 8.778930535580476e-07, "loss": 0.0, "mean_logprobs": -6.65625, "mean_logprobs/var": 0.053466796875, "num_completions/total": 6880, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.0996551513672, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 131.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 1.3451988697052002, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235708951950073, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.0684415428841021e-05, "policy_entropy": 3.7779762744903564, "policy_entropy/max": 3.875, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.796875, "policy_entropy/var": 0.0005676210857927799, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.029551107436418533, "policy_sharpness/max": 0.03472613915801048, "policy_sharpness/median": 0.03001275658607483, "policy_sharpness/min": 0.018046217039227486, "policy_sharpness/p25": 0.029218187555670738, "policy_sharpness/p75": 0.030843336135149002, "policy_sharpness/var": 5.454020993056474e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 86, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 1.3829020151234772e-09, "adam_stats/lm_head/lr_effective_mean": -4.446868339459979e-15, "adam_stats/lm_head/lr_effective_min": -1.3793691744368175e-09, "adam_stats/lm_head/lr_effective_std": 8.788408889515154e-11, "adam_stats/lr_effective_max": 1.4109744483903341e-09, "adam_stats/lr_effective_mean": 9.729097512161094e-15, "adam_stats/lr_effective_min": -1.404410254757238e-09, "adam_stats/m_t_max": 1.6292577811327646e-06, "adam_stats/m_t_mean": -2.955498191558985e-14, "adam_stats/m_t_min": -1.9154188066750066e-06, "adam_stats/v_t_max": 1.533309114165604e-05, "adam_stats/v_t_mean": 1.7924665658003436e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.5902862548828125, "all_logprobs/max": -5.0625, "all_logprobs/median": -6.65625, "all_logprobs/min": -7.96875, "all_logprobs/p1": -7.46875, "all_logprobs/p10": -7.28125, "all_logprobs/p25": -7.09375, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.31583359837532043, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.116, "feature_vector_variance/max_squared_error": 1520.3980712890625, "feature_vector_variance/metric": 67.00997161865234, "generated_tokens/total": 6799887.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 7.59044652756249e-07, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.00162506103515625, "num_completions/total": 6960, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.3749542236328, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.348477840423584, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.023553729057312, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.1053773960156832e-05, "policy_entropy": 3.773895263671875, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.00032805572845973074, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.0301497969776392, "policy_sharpness/max": 0.034602340310811996, "policy_sharpness/median": 0.030075781047344208, "policy_sharpness/min": 0.022920168936252594, "policy_sharpness/p25": 0.029401259496808052, "policy_sharpness/p75": 0.0308665968477726, "policy_sharpness/var": 1.261657189388643e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 87, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 1.0637322134243732e-09, "adam_stats/lm_head/lr_effective_mean": -3.4202955358485785e-15, "adam_stats/lm_head/lr_effective_min": -1.0610152756385105e-09, "adam_stats/lm_head/lr_effective_std": 6.759717785520536e-11, "adam_stats/lr_effective_max": 1.0853268284094497e-09, "adam_stats/lr_effective_mean": 7.483673308329199e-15, "adam_stats/lr_effective_min": -1.0802737593351708e-09, "adam_stats/m_t_max": 1.4663319234387018e-06, "adam_stats/m_t_mean": -2.6599463734053308e-14, "adam_stats/m_t_min": -1.7238768350580358e-06, "adam_stats/v_t_max": 1.5317758879973553e-05, "adam_stats/v_t_mean": 1.790674379609225e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.645834445953369, "all_logprobs/max": -5.0625, "all_logprobs/median": -6.71875, "all_logprobs/min": -8.25, "all_logprobs/p1": -7.8125, "all_logprobs/p10": -7.34375, "all_logprobs/p25": -7.15625, "all_logprobs/p5": -7.46875, "all_logprobs/p75": -6.15625, "all_logprobs/var": 0.34887173771858215, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.11733333333333333, "feature_vector_variance/max_squared_error": 3666.05810546875, "feature_vector_variance/metric": 160.95382690429688, "generated_tokens/total": 6881807.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 6.484090676804927e-07, "loss": 0.0, "mean_logprobs": -6.65625, "mean_logprobs/var": 0.042724609375, "num_completions/total": 7040, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.1322784423828, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 131.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 1.1786832809448242, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235689878463745, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.032280852086842e-05, "policy_entropy": 3.7771530151367188, "policy_entropy/max": 3.875, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.796875, "policy_entropy/var": 0.00046942709013819695, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.029760723933577538, "policy_sharpness/max": 0.03457533195614815, "policy_sharpness/median": 0.02997598983347416, "policy_sharpness/min": 0.017157113179564476, "policy_sharpness/p25": 0.02919267676770687, "policy_sharpness/p75": 0.03080732375383377, "policy_sharpness/var": 3.051084149774397e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 88, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 8.067358447938489e-10, "adam_stats/lm_head/lr_effective_mean": -2.5937673504231734e-15, "adam_stats/lm_head/lr_effective_min": -8.046757149493544e-10, "adam_stats/lm_head/lr_effective_std": 5.12631187499224e-11, "adam_stats/lr_effective_max": 8.231141324088753e-10, "adam_stats/lr_effective_mean": 5.6756476010970044e-15, "adam_stats/lr_effective_min": -8.192788114591565e-10, "adam_stats/m_t_max": 1.3196986401453614e-06, "adam_stats/m_t_mean": -2.393951939352705e-14, "adam_stats/m_t_min": -1.551489162920916e-06, "adam_stats/v_t_max": 1.5302441170206293e-05, "adam_stats/v_t_mean": 1.7888844702426687e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.648637294769287, "all_logprobs/max": -5.0625, "all_logprobs/median": -6.71875, "all_logprobs/min": -8.3125, "all_logprobs/p1": -7.90625, "all_logprobs/p10": -7.34375, "all_logprobs/p25": -7.15625, "all_logprobs/p5": -7.5625, "all_logprobs/p75": -6.15625, "all_logprobs/var": 0.36876800656318665, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.11866666666666667, "feature_vector_variance/max_squared_error": 3954.766845703125, "feature_vector_variance/metric": 209.1328125, "generated_tokens/total": 6963727.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 5.461210907490952e-07, "loss": 0.0, "mean_logprobs": -6.65625, "mean_logprobs/var": 0.06494140625, "num_completions/total": 7120, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.13970947265625, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 131.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 1.4718419313430786, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.02349054813385, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.0852876584976912e-05, "policy_entropy": 3.778719425201416, "policy_entropy/max": 3.859375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.796875, "policy_entropy/var": 0.0006220846553333104, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.029482845216989517, "policy_sharpness/max": 0.034561075270175934, "policy_sharpness/median": 0.0300150066614151, "policy_sharpness/min": 0.01717361994087696, "policy_sharpness/p25": 0.029245197772979736, "policy_sharpness/p75": 0.03083733469247818, "policy_sharpness/var": 6.481476702902e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 89, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 6.01635297137193e-10, "adam_stats/lm_head/lr_effective_mean": -1.934201640879326e-15, "adam_stats/lm_head/lr_effective_min": -6.000991370491704e-10, "adam_stats/lm_head/lr_effective_std": 3.822824695687288e-11, "adam_stats/lr_effective_max": 6.138503594321776e-10, "adam_stats/lr_effective_mean": 4.232710881823306e-15, "adam_stats/lr_effective_min": -6.109878158966353e-10, "adam_stats/m_t_max": 1.1877287988681928e-06, "adam_stats/m_t_mean": -2.1545567115361167e-14, "adam_stats/m_t_min": -1.396340167048038e-06, "adam_stats/v_t_max": 1.5287139831343666e-05, "adam_stats/v_t_mean": 1.7870948861367641e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.566693305969238, "all_logprobs/max": -5.03125, "all_logprobs/median": -6.65625, "all_logprobs/min": -7.875, "all_logprobs/p1": -7.46875, "all_logprobs/p10": -7.28125, "all_logprobs/p25": -7.09375, "all_logprobs/p5": -7.34375, "all_logprobs/p75": -6.09375, "all_logprobs/var": 0.32024049758911133, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.12, "feature_vector_variance/max_squared_error": 1679.2772216796875, "feature_vector_variance/metric": 66.56009674072266, "generated_tokens/total": 7045647.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 4.5230534410568764e-07, "loss": 0.0, "mean_logprobs": -6.5625, "mean_logprobs/var": 0.0013580322265625, "num_completions/total": 7200, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.5047607421875, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 137.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.3520081043243408, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0236049890518188, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.2264131328265648e-05, "policy_entropy": 3.771977186203003, "policy_entropy/max": 3.828125, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.0003418262640479952, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.030248314142227173, "policy_sharpness/max": 0.03460384160280228, "policy_sharpness/median": 0.030178571119904518, "policy_sharpness/min": 0.022025059908628464, "policy_sharpness/p25": 0.02949804812669754, "policy_sharpness/p75": 0.03094237670302391, "policy_sharpness/var": 1.2117139931433485e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 90, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 4.3965994733774494e-10, "adam_stats/lm_head/lr_effective_mean": -1.4133633050741434e-15, "adam_stats/lm_head/lr_effective_min": -4.3853756737100014e-10, "adam_stats/lm_head/lr_effective_std": 2.793478469376165e-11, "adam_stats/lr_effective_max": 4.485868621006972e-10, "adam_stats/lr_effective_mean": 3.0931669646959946e-15, "adam_stats/lr_effective_min": -4.464933700543128e-10, "adam_stats/m_t_max": 1.0689558394005871e-06, "adam_stats/m_t_mean": -1.9391021923473133e-14, "adam_stats/m_t_min": -1.2567061276058666e-06, "adam_stats/v_t_max": 1.5271853044396266e-05, "adam_stats/v_t_mean": 1.7853074704352045e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.588511943817139, "all_logprobs/max": -5.0625, "all_logprobs/median": -6.65625, "all_logprobs/min": -7.90625, "all_logprobs/p1": -7.46875, "all_logprobs/p10": -7.28125, "all_logprobs/p25": -7.09375, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.31357359886169434, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.12133333333333333, "feature_vector_variance/max_squared_error": 1533.0802001953125, "feature_vector_variance/metric": 65.87471771240234, "generated_tokens/total": 7127567.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 3.6707612778634855e-07, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.00157928466796875, "num_completions/total": 7280, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.3827362060547, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.3396228849887848, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235114097595215, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.0728782399382908e-05, "policy_entropy": 3.774404525756836, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.00033019634429365396, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.030148077756166458, "policy_sharpness/max": 0.034627851098775864, "policy_sharpness/median": 0.030072027817368507, "policy_sharpness/min": 0.022819627076387405, "policy_sharpness/p25": 0.02939525805413723, "policy_sharpness/p75": 0.030853841453790665, "policy_sharpness/var": 1.2413288459356409e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 91, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 3.1334429495544214e-10, "adam_stats/lm_head/lr_effective_mean": -1.0072265710961122e-15, "adam_stats/lm_head/lr_effective_min": -3.125445457996534e-10, "adam_stats/lm_head/lr_effective_std": 1.9907998860535514e-11, "adam_stats/lr_effective_max": 3.1970687208726645e-10, "adam_stats/lr_effective_mean": 2.2045000187435404e-15, "adam_stats/lr_effective_min": -3.182136221191456e-10, "adam_stats/m_t_max": 9.620601986171096e-07, "adam_stats/m_t_mean": -1.7451914479521546e-14, "adam_stats/m_t_min": -1.1310354466331773e-06, "adam_stats/v_t_max": 1.5256581718858797e-05, "adam_stats/v_t_mean": 1.78352222313799e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.595637798309326, "all_logprobs/max": -5.09375, "all_logprobs/median": -6.6875, "all_logprobs/min": -7.9375, "all_logprobs/p1": -7.46875, "all_logprobs/p10": -7.28125, "all_logprobs/p25": -7.09375, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.3137800097465515, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.12266666666666666, "feature_vector_variance/max_squared_error": 1718.290771484375, "feature_vector_variance/metric": 69.3204574584961, "generated_tokens/total": 7209487.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 2.905372804626083e-07, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.00244140625, "num_completions/total": 7360, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.36180114746094, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.36173874139785767, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0234946012496948, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.037781657942105e-05, "policy_entropy": 3.775061845779419, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.0003308948071207851, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.030071739107370377, "policy_sharpness/max": 0.03466236591339111, "policy_sharpness/median": 0.03001350536942482, "policy_sharpness/min": 0.021763956174254417, "policy_sharpness/p25": 0.029316477477550507, "policy_sharpness/p75": 0.030793817713856697, "policy_sharpness/var": 1.2663418829106377e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 92, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 2.163512147212998e-10, "adam_stats/lm_head/lr_effective_mean": -6.953979560697827e-16, "adam_stats/lm_head/lr_effective_min": -2.1579908693336591e-10, "adam_stats/lm_head/lr_effective_std": 1.3744927938874607e-11, "adam_stats/lr_effective_max": 2.2074451988540744e-10, "adam_stats/lr_effective_mean": 1.5221194767653986e-15, "adam_stats/lr_effective_min": -2.1971270636189644e-10, "adam_stats/m_t_max": 8.658541332806635e-07, "adam_stats/m_t_mean": -1.5706721676316676e-14, "adam_stats/m_t_min": -1.0179318223890732e-06, "adam_stats/v_t_max": 1.5241324945236556e-05, "adam_stats/v_t_mean": 1.781738602144034e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.591836452484131, "all_logprobs/max": -5.0, "all_logprobs/median": -6.6875, "all_logprobs/min": -7.84375, "all_logprobs/p1": -7.46875, "all_logprobs/p10": -7.28125, "all_logprobs/p25": -7.09375, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.31552836298942566, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.124, "feature_vector_variance/max_squared_error": 1402.1151123046875, "feature_vector_variance/metric": 65.72899627685547, "generated_tokens/total": 7291407.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 2.2278205293002645e-07, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.0016632080078125, "num_completions/total": 7440, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.37158203125, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.3378578722476959, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235815048217773, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.0942224434984382e-05, "policy_entropy": 3.7737977504730225, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.00032699061557650566, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.030130773782730103, "policy_sharpness/max": 0.03431347385048866, "policy_sharpness/median": 0.030055521056056023, "policy_sharpness/min": 0.023228541016578674, "policy_sharpness/p25": 0.029374249279499054, "policy_sharpness/p75": 0.030848028138279915, "policy_sharpness/var": 1.2603550203493796e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 93, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 1.4331731634786848e-10, "adam_stats/lm_head/lr_effective_mean": -4.606181578432238e-16, "adam_stats/lm_head/lr_effective_min": -1.429516643947082e-10, "adam_stats/lm_head/lr_effective_std": 9.104565192041836e-12, "adam_stats/lr_effective_max": 1.4622773825134772e-10, "adam_stats/lr_effective_mean": 1.0082995500820466e-15, "adam_stats/lr_effective_min": -1.4554370209030054e-10, "adam_stats/m_t_max": 7.792687028995715e-07, "adam_stats/m_t_mean": -1.4136048153432293e-14, "adam_stats/m_t_min": -9.161386174127983e-07, "adam_stats/v_t_max": 1.5226083633024246e-05, "adam_stats/v_t_mean": 1.7799574748150748e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.589755535125732, "all_logprobs/max": -5.03125, "all_logprobs/median": -6.65625, "all_logprobs/min": -7.9375, "all_logprobs/p1": -7.5, "all_logprobs/p10": -7.28125, "all_logprobs/p25": -7.09375, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.3161739408969879, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.12533333333333332, "feature_vector_variance/max_squared_error": 1659.2044677734375, "feature_vector_variance/metric": 68.93157958984375, "generated_tokens/total": 7373327.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.6389299449645734e-07, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.0027618408203125, "num_completions/total": 7520, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.36553955078125, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.3762362599372864, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235228538513184, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.0683736945793498e-05, "policy_entropy": 3.7747292518615723, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.0003355664375703782, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.030132034793496132, "policy_sharpness/max": 0.03468262404203415, "policy_sharpness/median": 0.030064526945352554, "policy_sharpness/min": 0.02174144610762596, "policy_sharpness/p25": 0.029375748708844185, "policy_sharpness/p75": 0.030852342024445534, "policy_sharpness/var": 1.2818652521673357e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 94, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 8.971823978187743e-11, "adam_stats/lm_head/lr_effective_mean": -2.883312328552255e-16, "adam_stats/lm_head/lr_effective_min": -8.948937424424486e-11, "adam_stats/lm_head/lr_effective_std": 5.699262242697678e-12, "adam_stats/lr_effective_max": 9.154028923763491e-11, "adam_stats/lr_effective_mean": 6.312084758378718e-16, "adam_stats/lr_effective_min": -9.11117362112357e-11, "adam_stats/m_t_max": 7.013418326096144e-07, "adam_stats/m_t_mean": -1.2722440796990222e-14, "adam_stats/m_t_min": -8.245247613558604e-07, "adam_stats/v_t_max": 1.5210857782221865e-05, "adam_stats/v_t_mean": 1.7781768895872019e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.609070777893066, "all_logprobs/max": -5.125, "all_logprobs/median": -6.6875, "all_logprobs/min": -7.9375, "all_logprobs/p1": -7.5, "all_logprobs/p10": -7.3125, "all_logprobs/p25": -7.125, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.31489917635917664, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.12666666666666668, "feature_vector_variance/max_squared_error": 2040.0054931640625, "feature_vector_variance/metric": 76.77559661865234, "generated_tokens/total": 7455247.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.1394185240843985e-07, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.004913330078125, "num_completions/total": 7600, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.28370666503906, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 132.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.42411988973617554, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235238075256348, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 9.94298989098752e-06, "policy_entropy": 3.775475263595581, "policy_entropy/max": 3.859375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.0003326121950522065, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.030002404004335403, "policy_sharpness/max": 0.03420243039727211, "policy_sharpness/median": 0.029957232996821404, "policy_sharpness/min": 0.02039765752851963, "policy_sharpness/p25": 0.02923000231385231, "policy_sharpness/p75": 0.03076530620455742, "policy_sharpness/var": 1.3888023886465817e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 95, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 5.1750780744042046e-11, "adam_stats/lm_head/lr_effective_mean": -1.663015122500136e-16, "adam_stats/lm_head/lr_effective_min": -5.161879951254278e-11, "adam_stats/lm_head/lr_effective_std": 3.287245909505687e-12, "adam_stats/lr_effective_max": 5.280182541200773e-11, "adam_stats/lr_effective_mean": 3.6409147431420584e-16, "adam_stats/lr_effective_min": -5.255443996654563e-11, "adam_stats/m_t_max": 6.312076266112854e-07, "adam_stats/m_t_mean": -1.1450203239444894e-14, "adam_stats/m_t_min": -7.420722454298811e-07, "adam_stats/v_t_max": 1.5195647392829414e-05, "adam_stats/v_t_mean": 1.7763994485456291e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.5699920654296875, "all_logprobs/max": -5.09375, "all_logprobs/median": -6.65625, "all_logprobs/min": -7.90625, "all_logprobs/p1": -7.46875, "all_logprobs/p10": -7.28125, "all_logprobs/p25": -7.09375, "all_logprobs/p5": -7.34375, "all_logprobs/p75": -6.09375, "all_logprobs/var": 0.3185133934020996, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.128, "feature_vector_variance/max_squared_error": 1725.9617919921875, "feature_vector_variance/metric": 65.43387603759766, "generated_tokens/total": 7537167.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 7.298948443822229e-08, "loss": 0.0, "mean_logprobs": -6.5625, "mean_logprobs/var": 0.00113677978515625, "num_completions/total": 7680, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.4490509033203, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.34184184670448303, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235536098480225, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.162527496489929e-05, "policy_entropy": 3.7729060649871826, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.00033687593531794846, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.030220985412597656, "policy_sharpness/max": 0.03447328880429268, "policy_sharpness/median": 0.030141055583953857, "policy_sharpness/min": 0.022167615592479706, "policy_sharpness/p25": 0.029475539922714233, "policy_sharpness/p75": 0.030922116711735725, "policy_sharpness/var": 1.2330575600572047e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 96, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 2.6230548982875135e-11, "adam_stats/lm_head/lr_effective_mean": -8.428595232789498e-17, "adam_stats/lm_head/lr_effective_min": -2.616366151508842e-11, "adam_stats/lm_head/lr_effective_std": 1.6660958637032541e-12, "adam_stats/lr_effective_max": 2.676331378737018e-11, "adam_stats/lr_effective_mean": 1.8454504021687474e-16, "adam_stats/lr_effective_min": -2.6637822156394542e-11, "adam_stats/m_t_max": 5.680868753188406e-07, "adam_stats/m_t_mean": -1.0305179866181794e-14, "adam_stats/m_t_min": -6.678649810964998e-07, "adam_stats/v_t_max": 1.5180451555352192e-05, "adam_stats/v_t_mean": 1.7746225496051427e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.593055725097656, "all_logprobs/max": -5.125, "all_logprobs/median": -6.6875, "all_logprobs/min": -7.9375, "all_logprobs/p1": -7.46875, "all_logprobs/p10": -7.28125, "all_logprobs/p25": -7.09375, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.316000759601593, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.12933333333333333, "feature_vector_variance/max_squared_error": 1667.2105712890625, "feature_vector_variance/metric": 67.78489685058594, "generated_tokens/total": 7619087.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 4.108578473795033e-08, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.001739501953125, "num_completions/total": 7760, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.3777313232422, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.34425947070121765, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0236115455627441, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.0998991456290241e-05, "policy_entropy": 3.7736191749572754, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.0003356586385052651, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.030125712975859642, "policy_sharpness/max": 0.034764405339956284, "policy_sharpness/median": 0.030059274286031723, "policy_sharpness/min": 0.022265907377004623, "policy_sharpness/p25": 0.02936599776148796, "policy_sharpness/p75": 0.030848588794469833, "policy_sharpness/var": 1.3049788094576797e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 97, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 1.0502787008148662e-11, "adam_stats/lm_head/lr_effective_mean": -3.374588195411748e-17, "adam_stats/lm_head/lr_effective_min": -1.0476008949211746e-11, "adam_stats/lm_head/lr_effective_std": 6.670747396304855e-13, "adam_stats/lr_effective_max": 1.0716118097053862e-11, "adam_stats/lr_effective_mean": 7.38926264545132e-17, "adam_stats/lr_effective_min": -1.0665831932932246e-11, "adam_stats/m_t_max": 5.112781877869566e-07, "adam_stats/m_t_mean": -9.274657559695584e-15, "adam_stats/m_t_min": -6.010784545651404e-07, "adam_stats/v_t_max": 1.51652711792849e-05, "adam_stats/v_t_mean": 1.7728483611700874e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.570683479309082, "all_logprobs/max": -5.09375, "all_logprobs/median": -6.65625, "all_logprobs/min": -7.96875, "all_logprobs/p1": -7.46875, "all_logprobs/p10": -7.28125, "all_logprobs/p25": -7.09375, "all_logprobs/p5": -7.34375, "all_logprobs/p75": -6.09375, "all_logprobs/var": 0.31679004430770874, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.13066666666666665, "feature_vector_variance/max_squared_error": 1894.5213623046875, "feature_vector_variance/metric": 64.33052825927734, "generated_tokens/total": 7701007.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 1.8269623051318517e-08, "loss": 0.0, "mean_logprobs": -6.5625, "mean_logprobs/var": 0.00109100341796875, "num_completions/total": 7840, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.47064208984375, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.3381803333759308, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235804319381714, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.1679416275001131e-05, "policy_entropy": 3.772650957107544, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.00034236256033182144, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.030263496562838554, "policy_sharpness/max": 0.03437650203704834, "policy_sharpness/median": 0.03018007054924965, "policy_sharpness/min": 0.021316029131412506, "policy_sharpness/p25": 0.029512304812669754, "policy_sharpness/p75": 0.030958132818341255, "policy_sharpness/var": 1.215773067997361e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 98, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 2.365027011036469e-12, "adam_stats/lm_head/lr_effective_mean": -7.59838019087738e-18, "adam_stats/lm_head/lr_effective_min": -2.3589975459148427e-12, "adam_stats/lm_head/lr_effective_std": 1.5020463713967536e-13, "adam_stats/lr_effective_max": 2.4130675756184328e-12, "adam_stats/lr_effective_mean": 1.6639257490928835e-17, "adam_stats/lr_effective_min": -2.4017348439903108e-12, "adam_stats/m_t_max": 4.6015034627089335e-07, "adam_stats/m_t_mean": -8.347194598934751e-15, "adam_stats/m_t_min": -5.409705750025751e-07, "adam_stats/v_t_max": 1.5150106264627539e-05, "adam_stats/v_t_mean": 1.771075473777639e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.588897705078125, "all_logprobs/max": -5.0625, "all_logprobs/median": -6.65625, "all_logprobs/min": -7.90625, "all_logprobs/p1": -7.46875, "all_logprobs/p10": -7.28125, "all_logprobs/p25": -7.09375, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.31346559524536133, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.132, "feature_vector_variance/max_squared_error": 1743.3577880859375, "feature_vector_variance/metric": 66.40225982666016, "generated_tokens/total": 7782927.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 4.568797356781784e-09, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.001495361328125, "num_completions/total": 7920, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.38694763183594, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 133.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.3394205868244171, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.0235345363616943, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.0708444278861862e-05, "policy_entropy": 3.774251699447632, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.0003312909684609622, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.030133787542581558, "policy_sharpness/max": 0.0345790795981884, "policy_sharpness/median": 0.030060024932026863, "policy_sharpness/min": 0.021935025230050087, "policy_sharpness/p25": 0.029378751292824745, "policy_sharpness/p75": 0.030843336135149002, "policy_sharpness/var": 1.257069811799738e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 99, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "accuracy_reward": 0.0, "accuracy_reward/incorrect": 0.0, "accuracy_reward/incorrect/max": 0.0, "accuracy_reward/incorrect/median": 0.0, "accuracy_reward/incorrect/min": 0.0, "accuracy_reward/incorrect/p25": 0.0, "accuracy_reward/incorrect/p75": 0.0, "accuracy_reward/incorrect/var": 0.0, "accuracy_reward/max": 0.0, "accuracy_reward/median": 0.0, "accuracy_reward/min": 0.0, "accuracy_reward/p25": 0.0, "accuracy_reward/p75": 0.0, "accuracy_reward/var": 0.0, "adam_stats/lm_head/lr_effective_max": 0.0, "adam_stats/lm_head/lr_effective_mean": 0.0, "adam_stats/lm_head/lr_effective_min": 0.0, "adam_stats/lm_head/lr_effective_std": 0.0, "adam_stats/lr_effective_max": 0.0, "adam_stats/lr_effective_mean": 0.0, "adam_stats/lr_effective_min": 0.0, "adam_stats/m_t_max": 4.1413531448597496e-07, "adam_stats/m_t_mean": -7.512477510733528e-15, "adam_stats/m_t_min": -4.868734890806081e-07, "adam_stats/v_t_max": 1.5134955901885405e-05, "adam_stats/v_t_mean": 1.769304429528884e-12, "adam_stats/v_t_min": 0.0, "advantages": 0.0, "advantages/max": 0.0, "advantages/median": 0.0, "advantages/min": 0.0, "advantages/p25": 0.0, "advantages/p75": 0.0, "advantages/var": 0.0, "all_logprobs": -6.604489803314209, "all_logprobs/max": -5.0625, "all_logprobs/median": -6.6875, "all_logprobs/min": -8.0, "all_logprobs/p1": -7.5, "all_logprobs/p10": -7.3125, "all_logprobs/p25": -7.125, "all_logprobs/p5": -7.375, "all_logprobs/p75": -6.125, "all_logprobs/var": 0.31601253151893616, "clip_ratio": 0.0, "completion_length": 1024.0, "completion_length/incorrect": 1024.0, "completion_length/incorrect/max": 1024.0, "completion_length/incorrect/median": 1024.0, "completion_length/incorrect/min": 1024.0, "completion_length/incorrect/p25": 1024.0, "completion_length/incorrect/p75": 1024.0, "completion_length/incorrect/var": 0.0, "completion_length/max": 1024.0, "completion_length/median": 1024.0, "completion_length/min": 1024.0, "completion_length/p25": 1024.0, "completion_length/p75": 1024.0, "completion_length/var": 0.0, "epoch": 0.13333333333333333, "feature_vector_variance/max_squared_error": 1890.3560791015625, "feature_vector_variance/metric": 71.27523040771484, "generated_tokens/total": 7864847.0, "global_fisher_curvature": 0.0, "global_fisher_curvature/max": 0.0, "global_fisher_curvature/median": 0.0, "global_fisher_curvature/min": 0.0, "global_fisher_curvature/p25": 0.0, "global_fisher_curvature/p75": 0.0, "global_fisher_curvature/p85": 0.0, "global_fisher_curvature/p90": 0.0, "global_fisher_curvature/p95": 0.0, "global_fisher_curvature/p99": 0.0, "global_fisher_curvature/var": NaN, "global_fisher_kl_divergence": 0.0, "global_fisher_kl_divergence/max": 0.0, "global_fisher_kl_divergence/median": 0.0, "global_fisher_kl_divergence/min": 0.0, "global_fisher_kl_divergence/p25": 0.0, "global_fisher_kl_divergence/p75": 0.0, "global_fisher_kl_divergence/p85": 0.0, "global_fisher_kl_divergence/p90": 0.0, "global_fisher_kl_divergence/p95": 0.0, "global_fisher_kl_divergence/p99": 0.0, "global_fisher_kl_divergence/var": NaN, "global_full_update_term": 0.0, "global_full_update_term/max": 0.0, "global_full_update_term/median": 0.0, "global_full_update_term/min": 0.0, "global_full_update_term/p25": 0.0, "global_full_update_term/p75": 0.0, "global_full_update_term/p85": 0.0, "global_full_update_term/p90": 0.0, "global_full_update_term/p95": 0.0, "global_full_update_term/p99": 0.0, "global_full_update_term/var": NaN, "global_hessian_coeff": 0.0, "global_hessian_coeff/max": 0.0, "global_hessian_coeff/median": 0.0, "global_hessian_coeff/min": 0.0, "global_hessian_coeff/p25": 0.0, "global_hessian_coeff/p75": 0.0, "global_hessian_coeff/p99": 0.0, "global_hessian_coeff/var": NaN, "global_hessian_coeff_abs": 0.0, "global_hessian_coeff_abs/max": 0.0, "global_hessian_coeff_abs/median": 0.0, "global_hessian_coeff_abs/min": 0.0, "global_hessian_coeff_abs/p25": 0.0, "global_hessian_coeff_abs/p75": 0.0, "global_hessian_coeff_abs/p99": 0.0, "global_hessian_coeff_abs/var": NaN, "grad_norm": 0.0, "grouped_std_rewards": 0.0, "learning_rate": 0.0, "loss": 0.0, "mean_logprobs": -6.59375, "mean_logprobs/var": 0.0038604736328125, "num_completions/total": 8000, "per_sentence_gradient_norm": 0.0, "per_sentence_gradient_norm/max": 0.0, "per_sentence_gradient_norm/median": 0.0, "per_sentence_gradient_norm/min": 0.0, "per_sentence_gradient_norm/p25": 0.0, "per_sentence_gradient_norm/p75": 0.0, "per_sentence_gradient_norm/var": 0.0, "per_token_feature_norm": 136.3182830810547, "per_token_feature_norm/max": 138.0, "per_token_feature_norm/median": 136.0, "per_token_feature_norm/min": 132.0, "per_token_feature_norm/p25": 136.0, "per_token_feature_norm/p75": 137.0, "per_token_feature_norm/var": 0.39796245098114014, "per_token_gradient_norm": 0.0, "per_token_gradient_norm/max": 0.0, "per_token_gradient_norm/median": 0.0, "per_token_gradient_norm/min": 0.0, "per_token_gradient_norm/p1": 0.0, "per_token_gradient_norm/p10": 0.0, "per_token_gradient_norm/p25": 0.0, "per_token_gradient_norm/p5": 0.0, "per_token_gradient_norm/p75": 0.0, "per_token_gradient_norm/var": 0.0, "per_token_policy_error_norm": 1.023533821105957, "per_token_policy_error_norm/max": 1.03125, "per_token_policy_error_norm/median": 1.0234375, "per_token_policy_error_norm/min": 1.015625, "per_token_policy_error_norm/p25": 1.0234375, "per_token_policy_error_norm/p75": 1.0234375, "per_token_policy_error_norm/var": 1.0207129889749922e-05, "policy_entropy": 3.775257110595703, "policy_entropy/max": 3.84375, "policy_entropy/median": 3.78125, "policy_entropy/min": 3.703125, "policy_entropy/p25": 3.765625, "policy_entropy/p75": 3.78125, "policy_entropy/var": 0.00032903661485761404, "policy_loss": 0.0, "policy_loss/max": 0.0, "policy_loss/median": 0.0, "policy_loss/min": 0.0, "policy_loss/p25": 0.0, "policy_loss/p75": 0.0, "policy_loss/var": 0.0, "policy_sharpness": 0.030023498460650444, "policy_sharpness/max": 0.03438175097107887, "policy_sharpness/median": 0.029965486377477646, "policy_sharpness/min": 0.021648408845067024, "policy_sharpness/p25": 0.029268456622958183, "policy_sharpness/p75": 0.030753301456570625, "policy_sharpness/var": 1.2996247278351802e-06, "reward": 0.0, "reward/max": 0.0, "reward/median": 0.0, "reward/min": 0.0, "reward/p25": 0.0, "reward/p75": 0.0, "reward/var": 0.0, "rewards/accuracy_reward": 0.0, "rewards/accuracy_reward/max": 0.0, "rewards/accuracy_reward/median": 0.0, "rewards/accuracy_reward/min": 0.0, "rewards/accuracy_reward/p25": 0.0, "rewards/accuracy_reward/p75": 0.0, "rewards/accuracy_reward/var": 0.0, "sentence_fisher_curvature": 0.0, "sentence_fisher_curvature/max": 0.0, "sentence_fisher_curvature/median": 0.0, "sentence_fisher_curvature/min": 0.0, "sentence_fisher_curvature/p25": 0.0, "sentence_fisher_curvature/p75": 0.0, "sentence_fisher_curvature/p85": 0.0, "sentence_fisher_curvature/p90": 0.0, "sentence_fisher_curvature/p95": 0.0, "sentence_fisher_curvature/p99": 0.0, "sentence_fisher_curvature/var": 0.0, "sentence_fisher_kl_divergence": 0.0, "sentence_fisher_kl_divergence/max": 0.0, "sentence_fisher_kl_divergence/median": 0.0, "sentence_fisher_kl_divergence/min": 0.0, "sentence_fisher_kl_divergence/p25": 0.0, "sentence_fisher_kl_divergence/p75": 0.0, "sentence_fisher_kl_divergence/p85": 0.0, "sentence_fisher_kl_divergence/p90": 0.0, "sentence_fisher_kl_divergence/p95": 0.0, "sentence_fisher_kl_divergence/p99": 0.0, "sentence_fisher_kl_divergence/var": 0.0, "sentence_full_gradient_variance/max_squared_error": 0.0, "sentence_full_gradient_variance/metric": 0.0, "sentence_full_gradient_variance/p75": 0.0, "sentence_full_gradient_variance/p90": 0.0, "sentence_full_gradient_variance/p95": 0.0, "sentence_full_gradient_variance/p99": 0.0, "sentence_full_update_term": 0.0, "sentence_full_update_term/max": 0.0, "sentence_full_update_term/median": 0.0, "sentence_full_update_term/min": 0.0, "sentence_full_update_term/p25": 0.0, "sentence_full_update_term/p75": 0.0, "sentence_full_update_term/p85": 0.0, "sentence_full_update_term/p90": 0.0, "sentence_full_update_term/p95": 0.0, "sentence_full_update_term/p99": 0.0, "sentence_full_update_term/var": 0.0, "sentence_hessian_coeff": 0.0, "sentence_hessian_coeff/max": 0.0, "sentence_hessian_coeff/median": 0.0, "sentence_hessian_coeff/min": 0.0, "sentence_hessian_coeff/p25": 0.0, "sentence_hessian_coeff/p75": 0.0, "sentence_hessian_coeff/p99": 0.0, "sentence_hessian_coeff/var": 0.0, "sentence_hessian_coeff_abs": 0.0, "sentence_hessian_coeff_abs/max": 0.0, "sentence_hessian_coeff_abs/median": 0.0, "sentence_hessian_coeff_abs/min": 0.0, "sentence_hessian_coeff_abs/p25": 0.0, "sentence_hessian_coeff_abs/p75": 0.0, "sentence_hessian_coeff_abs/p99": 0.0, "sentence_hessian_coeff_abs/var": 0.0, "step": 100, "token_fisher_curvature": 0.0, "token_fisher_curvature/max": 0.0, "token_fisher_curvature/median": 0.0, "token_fisher_curvature/min": 0.0, "token_fisher_curvature/p25": 0.0, "token_fisher_curvature/p75": 0.0, "token_fisher_curvature/p85": 0.0, "token_fisher_curvature/p90": 0.0, "token_fisher_curvature/p95": 0.0, "token_fisher_curvature/p99": 0.0, "token_fisher_curvature/var": 0.0, "token_fisher_kl_divergence": 0.0, "token_fisher_kl_divergence/max": 0.0, "token_fisher_kl_divergence/median": 0.0, "token_fisher_kl_divergence/min": 0.0, "token_fisher_kl_divergence/p25": 0.0, "token_fisher_kl_divergence/p75": 0.0, "token_fisher_kl_divergence/p85": 0.0, "token_fisher_kl_divergence/p90": 0.0, "token_fisher_kl_divergence/p95": 0.0, "token_fisher_kl_divergence/p99": 0.0, "token_fisher_kl_divergence/var": 0.0, "token_full_update_term": 0.0, "token_full_update_term/max": 0.0, "token_full_update_term/median": 0.0, "token_full_update_term/min": 0.0, "token_full_update_term/p25": 0.0, "token_full_update_term/p75": 0.0, "token_full_update_term/p85": 0.0, "token_full_update_term/p90": 0.0, "token_full_update_term/p95": 0.0, "token_full_update_term/p99": 0.0, "token_full_update_term/var": 0.0, "token_hessian_coeff": 0.0, "token_hessian_coeff/max": 0.0, "token_hessian_coeff/median": 0.0, "token_hessian_coeff/min": 0.0, "token_hessian_coeff/p25": 0.0, "token_hessian_coeff/p75": 0.0, "token_hessian_coeff/p99": 0.0, "token_hessian_coeff/var": 0.0, "token_hessian_coeff_abs": 0.0, "token_hessian_coeff_abs/max": 0.0, "token_hessian_coeff_abs/median": 0.0, "token_hessian_coeff_abs/min": 0.0, "token_hessian_coeff_abs/p25": 0.0, "token_hessian_coeff_abs/p75": 0.0, "token_hessian_coeff_abs/p99": 0.0, "token_hessian_coeff_abs/var": 0.0 }, { "adam_stats/lm_head/lr_effective_max": 0.0, "adam_stats/lm_head/lr_effective_mean": 0.0, "adam_stats/lm_head/lr_effective_min": 0.0, "adam_stats/lm_head/lr_effective_std": 0.0, "adam_stats/lr_effective_max": 0.0, "adam_stats/lr_effective_mean": 0.0, "adam_stats/lr_effective_min": 0.0, "adam_stats/m_t_max": 4.1413531448597496e-07, "adam_stats/m_t_mean": -7.512477510733528e-15, "adam_stats/m_t_min": -4.868734890806081e-07, "adam_stats/v_t_max": 1.5134955901885405e-05, "adam_stats/v_t_mean": 1.769304429528884e-12, "adam_stats/v_t_min": 0.0, "epoch": 0.13333333333333333, "step": 100, "total_flos": 0.0, "train_loss": 2.980232949312267e-11, "train_runtime": 14681.8696, "train_samples_per_second": 0.545, "train_steps_per_second": 0.007 } ], "logging_steps": 1, "max_steps": 100, "num_input_tokens_seen": 0, "num_train_epochs": 1, "save_steps": 10, "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": 20, "trial_name": null, "trial_params": null }