| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 5.0, | |
| "eval_steps": 500, | |
| "global_step": 30, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "completion_length": 316.15625, | |
| "epoch": 0.3333333333333333, | |
| "grad_norm": 21.52219009399414, | |
| "kl": 0.0030751565063837916, | |
| "learning_rate": 4.985344892885899e-07, | |
| "loss": 0.0, | |
| "reward": 0.7553889602422714, | |
| "reward_std": 0.8830663803964853, | |
| "rewards/concensus_correctness_reward_func": 0.12793750315904617, | |
| "rewards/consensus_reward_func": 0.125, | |
| "rewards/cumulative_reward_2": 0.0, | |
| "rewards/final_correctness_reward_func": 0.0, | |
| "rewards/question_recreation_reward_func": 0.3625452145934105, | |
| "rewards/soft_format_reward_func": 0.0, | |
| "rewards/strict_format_reward_func": 0.0, | |
| "rewards/xmlcount_reward_func": 0.13990625739097595, | |
| "step": 2 | |
| }, | |
| { | |
| "completion_length": 300.5625, | |
| "epoch": 0.6666666666666666, | |
| "grad_norm": 1.9447593688964844, | |
| "kl": 0.0016620483074802905, | |
| "learning_rate": 4.869132927957006e-07, | |
| "loss": 0.0, | |
| "reward": 0.46813742350786924, | |
| "reward_std": 0.5111670058686286, | |
| "rewards/concensus_correctness_reward_func": 0.0390625, | |
| "rewards/consensus_reward_func": 0.0, | |
| "rewards/cumulative_reward_2": 0.0, | |
| "rewards/final_correctness_reward_func": 0.0, | |
| "rewards/question_recreation_reward_func": 0.25704367691650987, | |
| "rewards/soft_format_reward_func": 0.0, | |
| "rewards/strict_format_reward_func": 0.0, | |
| "rewards/xmlcount_reward_func": 0.1720312472898513, | |
| "step": 4 | |
| }, | |
| { | |
| "completion_length": 356.59375, | |
| "epoch": 1.0, | |
| "grad_norm": 6.489116668701172, | |
| "kl": 0.002859230269677937, | |
| "learning_rate": 4.642142940418973e-07, | |
| "loss": 0.0, | |
| "reward": 0.12191218649968505, | |
| "reward_std": 0.35503741540014744, | |
| "rewards/concensus_correctness_reward_func": 0.0, | |
| "rewards/consensus_reward_func": 0.0, | |
| "rewards/cumulative_reward_2": 0.0, | |
| "rewards/final_correctness_reward_func": 0.0625, | |
| "rewards/question_recreation_reward_func": 0.24194344179704785, | |
| "rewards/soft_format_reward_func": 0.0, | |
| "rewards/strict_format_reward_func": 0.0, | |
| "rewards/xmlcount_reward_func": -0.1825312376022339, | |
| "step": 6 | |
| }, | |
| { | |
| "completion_length": 345.09375, | |
| "epoch": 1.3333333333333333, | |
| "grad_norm": 2.4612724781036377, | |
| "kl": 0.0013941506476840004, | |
| "learning_rate": 4.314988729807827e-07, | |
| "loss": 0.0, | |
| "reward": 0.33836980583146214, | |
| "reward_std": 0.2510320176370442, | |
| "rewards/concensus_correctness_reward_func": 0.0, | |
| "rewards/consensus_reward_func": 0.0, | |
| "rewards/cumulative_reward_2": 0.0, | |
| "rewards/final_correctness_reward_func": 0.0625, | |
| "rewards/question_recreation_reward_func": 0.1744322977028787, | |
| "rewards/soft_format_reward_func": 0.0, | |
| "rewards/strict_format_reward_func": 0.0, | |
| "rewards/xmlcount_reward_func": 0.10143750347197056, | |
| "step": 8 | |
| }, | |
| { | |
| "completion_length": 339.03125, | |
| "epoch": 1.6666666666666665, | |
| "grad_norm": 1.9401979446411133, | |
| "kl": 0.0020870345033472404, | |
| "learning_rate": 3.902967663405956e-07, | |
| "loss": 0.0, | |
| "reward": 0.46584965474903584, | |
| "reward_std": 0.5341893364675343, | |
| "rewards/concensus_correctness_reward_func": 0.049937501549720764, | |
| "rewards/consensus_reward_func": 0.0, | |
| "rewards/cumulative_reward_2": 0.0, | |
| "rewards/final_correctness_reward_func": 0.0, | |
| "rewards/question_recreation_reward_func": 0.33859966415911913, | |
| "rewards/soft_format_reward_func": 0.0, | |
| "rewards/strict_format_reward_func": 0.0, | |
| "rewards/xmlcount_reward_func": 0.07731248997151852, | |
| "step": 10 | |
| }, | |
| { | |
| "completion_length": 375.65625, | |
| "epoch": 2.0, | |
| "grad_norm": 6.633189678192139, | |
| "kl": 0.0027604274946497753, | |
| "learning_rate": 3.4253453883497864e-07, | |
| "loss": 0.0, | |
| "reward": 0.49713892582803965, | |
| "reward_std": 0.6947258803993464, | |
| "rewards/concensus_correctness_reward_func": 0.07800000160932541, | |
| "rewards/consensus_reward_func": 0.0, | |
| "rewards/cumulative_reward_2": 0.0, | |
| "rewards/final_correctness_reward_func": 0.0625, | |
| "rewards/question_recreation_reward_func": 0.300607668235898, | |
| "rewards/soft_format_reward_func": 0.0, | |
| "rewards/strict_format_reward_func": 0.0, | |
| "rewards/xmlcount_reward_func": 0.05603125458583236, | |
| "step": 12 | |
| }, | |
| { | |
| "completion_length": 317.78125, | |
| "epoch": 2.3333333333333335, | |
| "grad_norm": 22.7032413482666, | |
| "kl": 0.004153315079747699, | |
| "learning_rate": 2.9044549913819124e-07, | |
| "loss": 0.0, | |
| "reward": 0.25570555380545557, | |
| "reward_std": 0.4283748477464542, | |
| "rewards/concensus_correctness_reward_func": 0.0, | |
| "rewards/consensus_reward_func": 0.0, | |
| "rewards/cumulative_reward_2": 0.0, | |
| "rewards/final_correctness_reward_func": 0.0, | |
| "rewards/question_recreation_reward_func": 0.18479929422028363, | |
| "rewards/soft_format_reward_func": 0.0, | |
| "rewards/strict_format_reward_func": 0.0, | |
| "rewards/xmlcount_reward_func": 0.07090624794363976, | |
| "step": 14 | |
| }, | |
| { | |
| "completion_length": 462.53125, | |
| "epoch": 2.6666666666666665, | |
| "grad_norm": 1.7028603553771973, | |
| "kl": 0.0016381583554903045, | |
| "learning_rate": 2.3646527285364563e-07, | |
| "loss": 0.0, | |
| "reward": 0.3854512879624963, | |
| "reward_std": 0.6286731185391545, | |
| "rewards/concensus_correctness_reward_func": 0.0078125, | |
| "rewards/consensus_reward_func": 0.0625, | |
| "rewards/cumulative_reward_2": 0.0, | |
| "rewards/final_correctness_reward_func": 0.125, | |
| "rewards/question_recreation_reward_func": 0.2168888095766306, | |
| "rewards/soft_format_reward_func": 0.0, | |
| "rewards/strict_format_reward_func": 0.0, | |
| "rewards/xmlcount_reward_func": -0.0267499890178442, | |
| "step": 16 | |
| }, | |
| { | |
| "completion_length": 428.8125, | |
| "epoch": 3.0, | |
| "grad_norm": 7.420510292053223, | |
| "kl": 0.0014154443051666021, | |
| "learning_rate": 1.8311791536769483e-07, | |
| "loss": 0.0, | |
| "reward": 1.9474243009462953, | |
| "reward_std": 2.0522735831327736, | |
| "rewards/concensus_correctness_reward_func": 1.25, | |
| "rewards/consensus_reward_func": 0.0, | |
| "rewards/cumulative_reward_2": 0.0, | |
| "rewards/final_correctness_reward_func": 0.125, | |
| "rewards/question_recreation_reward_func": 0.3706743409857154, | |
| "rewards/soft_format_reward_func": 0.0, | |
| "rewards/strict_format_reward_func": 0.0, | |
| "rewards/xmlcount_reward_func": 0.2017499953508377, | |
| "step": 18 | |
| }, | |
| { | |
| "completion_length": 335.375, | |
| "epoch": 3.3333333333333335, | |
| "grad_norm": 1.8302611112594604, | |
| "kl": 0.0017498906381661072, | |
| "learning_rate": 1.328978898250525e-07, | |
| "loss": 0.0, | |
| "reward": 0.29314853716641665, | |
| "reward_std": 0.4034676030278206, | |
| "rewards/concensus_correctness_reward_func": 0.0, | |
| "rewards/consensus_reward_func": 0.0, | |
| "rewards/cumulative_reward_2": 0.0, | |
| "rewards/final_correctness_reward_func": 0.0, | |
| "rewards/question_recreation_reward_func": 0.22377353068441153, | |
| "rewards/soft_format_reward_func": 0.0, | |
| "rewards/strict_format_reward_func": 0.0, | |
| "rewards/xmlcount_reward_func": 0.0693750036880374, | |
| "step": 20 | |
| }, | |
| { | |
| "completion_length": 375.75, | |
| "epoch": 3.6666666666666665, | |
| "grad_norm": 2.6359448432922363, | |
| "kl": 0.001499448495451361, | |
| "learning_rate": 8.81534288045431e-08, | |
| "loss": 0.0, | |
| "reward": 0.40015758108347654, | |
| "reward_std": 0.5831130938604474, | |
| "rewards/concensus_correctness_reward_func": 0.0, | |
| "rewards/consensus_reward_func": 0.0, | |
| "rewards/cumulative_reward_2": 0.0, | |
| "rewards/final_correctness_reward_func": 0.0625, | |
| "rewards/question_recreation_reward_func": 0.2579075903631747, | |
| "rewards/soft_format_reward_func": 0.0, | |
| "rewards/strict_format_reward_func": 0.0, | |
| "rewards/xmlcount_reward_func": 0.0797500004991889, | |
| "step": 22 | |
| }, | |
| { | |
| "completion_length": 384.8125, | |
| "epoch": 4.0, | |
| "grad_norm": 1.4408454895019531, | |
| "kl": 0.002384145453106612, | |
| "learning_rate": 5.097673357358906e-08, | |
| "loss": 0.0, | |
| "reward": 0.5307273641228676, | |
| "reward_std": 0.6833547167479992, | |
| "rewards/concensus_correctness_reward_func": 0.0, | |
| "rewards/consensus_reward_func": 0.0, | |
| "rewards/cumulative_reward_2": 0.0, | |
| "rewards/final_correctness_reward_func": 0.125, | |
| "rewards/question_recreation_reward_func": 0.37400862388312817, | |
| "rewards/soft_format_reward_func": 0.0, | |
| "rewards/strict_format_reward_func": 0.0, | |
| "rewards/xmlcount_reward_func": 0.031718747690320015, | |
| "step": 24 | |
| }, | |
| { | |
| "completion_length": 345.78125, | |
| "epoch": 4.333333333333333, | |
| "grad_norm": 2.0862185955047607, | |
| "kl": 0.002453506240271963, | |
| "learning_rate": 2.3106145082260774e-08, | |
| "loss": 0.0, | |
| "reward": 0.30404016049578786, | |
| "reward_std": 0.45155784906819463, | |
| "rewards/concensus_correctness_reward_func": 0.0, | |
| "rewards/consensus_reward_func": 0.0, | |
| "rewards/cumulative_reward_2": 0.0, | |
| "rewards/final_correctness_reward_func": 0.0, | |
| "rewards/question_recreation_reward_func": 0.21141517208889127, | |
| "rewards/soft_format_reward_func": 0.0, | |
| "rewards/strict_format_reward_func": 0.0, | |
| "rewards/xmlcount_reward_func": 0.09262499958276749, | |
| "step": 26 | |
| }, | |
| { | |
| "completion_length": 450.09375, | |
| "epoch": 4.666666666666667, | |
| "grad_norm": 1.684699535369873, | |
| "kl": 0.0021325907873688266, | |
| "learning_rate": 5.844861072478335e-09, | |
| "loss": 0.0, | |
| "reward": 0.05113972723484039, | |
| "reward_std": 1.3334482870995998, | |
| "rewards/concensus_correctness_reward_func": 0.016124999150633812, | |
| "rewards/consensus_reward_func": 0.0, | |
| "rewards/cumulative_reward_2": 0.0, | |
| "rewards/final_correctness_reward_func": 0.0, | |
| "rewards/question_recreation_reward_func": 0.4341397238895297, | |
| "rewards/soft_format_reward_func": 0.0, | |
| "rewards/strict_format_reward_func": 0.0, | |
| "rewards/xmlcount_reward_func": -0.39912501722574234, | |
| "step": 28 | |
| }, | |
| { | |
| "completion_length": 348.8125, | |
| "epoch": 5.0, | |
| "grad_norm": 20.362266540527344, | |
| "kl": 0.001994746329728514, | |
| "learning_rate": 0.0, | |
| "loss": 0.0, | |
| "reward": 1.1518285004422069, | |
| "reward_std": 2.1777954126009718, | |
| "rewards/concensus_correctness_reward_func": 0.6850625015795231, | |
| "rewards/consensus_reward_func": 0.0625, | |
| "rewards/cumulative_reward_2": 0.0, | |
| "rewards/final_correctness_reward_func": 0.0625, | |
| "rewards/question_recreation_reward_func": 0.20567227923311293, | |
| "rewards/soft_format_reward_func": 0.0, | |
| "rewards/strict_format_reward_func": 0.0, | |
| "rewards/xmlcount_reward_func": 0.13609375664964318, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "step": 30, | |
| "total_flos": 0.0, | |
| "train_loss": 2.2229773397460423e-06, | |
| "train_runtime": 1482.5235, | |
| "train_samples_per_second": 0.324, | |
| "train_steps_per_second": 0.02 | |
| } | |
| ], | |
| "logging_steps": 2, | |
| "max_steps": 30, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 5, | |
| "save_steps": 25, | |
| "stateful_callbacks": { | |
| "TrainerControl": { | |
| "args": { | |
| "should_epoch_stop": false, | |
| "should_evaluate": false, | |
| "should_log": false, | |
| "should_save": true, | |
| "should_training_stop": true | |
| }, | |
| "attributes": {} | |
| } | |
| }, | |
| "total_flos": 0.0, | |
| "train_batch_size": 4, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |