| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 0.9993002099370188, | |
| "eval_steps": 500, | |
| "global_step": 357, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 750.0178985595703, | |
| "epoch": 0.0027991602519244225, | |
| "grad_norm": 0.30192625522613525, | |
| "kl": 0.0, | |
| "learning_rate": 5.555555555555555e-07, | |
| "loss": 0.0293, | |
| "reward": 0.3750000186264515, | |
| "reward_std": 0.3078143782913685, | |
| "rewards/accuracy_reward": 0.3750000186264515, | |
| "step": 1 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 731.9107496473524, | |
| "epoch": 0.02799160251924423, | |
| "grad_norm": 0.2259158194065094, | |
| "kl": 0.00010485781563652886, | |
| "learning_rate": 5.555555555555557e-06, | |
| "loss": 0.0128, | |
| "reward": 0.3670635099212329, | |
| "reward_std": 0.26350560991300476, | |
| "rewards/accuracy_reward": 0.3670635099212329, | |
| "step": 10 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 716.9598579406738, | |
| "epoch": 0.05598320503848846, | |
| "grad_norm": 0.3309926688671112, | |
| "kl": 0.0035821676254272463, | |
| "learning_rate": 1.1111111111111113e-05, | |
| "loss": 0.0246, | |
| "reward": 0.40535716228187085, | |
| "reward_std": 0.2991223815828562, | |
| "rewards/accuracy_reward": 0.40535716228187085, | |
| "step": 20 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 743.5920013427734, | |
| "epoch": 0.08397480755773268, | |
| "grad_norm": 0.30614855885505676, | |
| "kl": 0.01827545166015625, | |
| "learning_rate": 1.6666666666666667e-05, | |
| "loss": 0.015, | |
| "reward": 0.48482145443558694, | |
| "reward_std": 0.2562454042956233, | |
| "rewards/accuracy_reward": 0.48482145443558694, | |
| "step": 30 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 675.5830673217773, | |
| "epoch": 0.11196641007697691, | |
| "grad_norm": 0.2699500024318695, | |
| "kl": 0.03763885498046875, | |
| "learning_rate": 1.999310448492752e-05, | |
| "loss": 0.0166, | |
| "reward": 0.4187500230967999, | |
| "reward_std": 0.28400791343301535, | |
| "rewards/accuracy_reward": 0.4187500230967999, | |
| "step": 40 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 724.2911056518554, | |
| "epoch": 0.13995801259622112, | |
| "grad_norm": 0.28290238976478577, | |
| "kl": 0.054229736328125, | |
| "learning_rate": 1.9915651236017307e-05, | |
| "loss": 0.0087, | |
| "reward": 0.41517859101295473, | |
| "reward_std": 0.2900457665324211, | |
| "rewards/accuracy_reward": 0.41517859101295473, | |
| "step": 50 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 648.35092086792, | |
| "epoch": 0.16794961511546536, | |
| "grad_norm": 0.26611465215682983, | |
| "kl": 0.063665771484375, | |
| "learning_rate": 1.975286910165463e-05, | |
| "loss": 0.016, | |
| "reward": 0.46607145220041274, | |
| "reward_std": 0.23811200838536023, | |
| "rewards/accuracy_reward": 0.46607145220041274, | |
| "step": 60 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 653.6018226623535, | |
| "epoch": 0.1959412176347096, | |
| "grad_norm": 0.2749662697315216, | |
| "kl": 0.077093505859375, | |
| "learning_rate": 1.95063160182963e-05, | |
| "loss": 0.0167, | |
| "reward": 0.46250002086162567, | |
| "reward_std": 0.26145450249314306, | |
| "rewards/accuracy_reward": 0.46250002086162567, | |
| "step": 70 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 849.4866439819336, | |
| "epoch": 0.22393282015395383, | |
| "grad_norm": 0.19597865641117096, | |
| "kl": 0.08182373046875, | |
| "learning_rate": 1.917835166772562e-05, | |
| "loss": 0.0064, | |
| "reward": 0.5196428813040257, | |
| "reward_std": 0.26366451028734444, | |
| "rewards/accuracy_reward": 0.5196428813040257, | |
| "step": 80 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 893.0500381469726, | |
| "epoch": 0.25192442267319803, | |
| "grad_norm": 0.2700682282447815, | |
| "kl": 0.07144775390625, | |
| "learning_rate": 1.877211489328239e-05, | |
| "loss": 0.0087, | |
| "reward": 0.45446430817246436, | |
| "reward_std": 0.2592643300071359, | |
| "rewards/accuracy_reward": 0.45446430817246436, | |
| "step": 90 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 749.3607467651367, | |
| "epoch": 0.27991602519244224, | |
| "grad_norm": 0.2792201042175293, | |
| "kl": 0.079168701171875, | |
| "learning_rate": 1.829149365898355e-05, | |
| "loss": 0.0121, | |
| "reward": 0.4687500227242708, | |
| "reward_std": 0.2029303913936019, | |
| "rewards/accuracy_reward": 0.4687500227242708, | |
| "step": 100 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 666.171460723877, | |
| "epoch": 0.3079076277116865, | |
| "grad_norm": 0.24913422763347626, | |
| "kl": 0.0840087890625, | |
| "learning_rate": 1.7741087839045992e-05, | |
| "loss": -0.0013, | |
| "reward": 0.47857145592570305, | |
| "reward_std": 0.2354975413531065, | |
| "rewards/accuracy_reward": 0.47857145592570305, | |
| "step": 110 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 733.9053924560546, | |
| "epoch": 0.3358992302309307, | |
| "grad_norm": 0.23493996262550354, | |
| "kl": 0.078680419921875, | |
| "learning_rate": 1.712616519394157e-05, | |
| "loss": 0.0031, | |
| "reward": 0.4696428798139095, | |
| "reward_std": 0.246359870582819, | |
| "rewards/accuracy_reward": 0.4696428798139095, | |
| "step": 120 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 966.488444519043, | |
| "epoch": 0.363890832750175, | |
| "grad_norm": 0.24625414609909058, | |
| "kl": 0.087646484375, | |
| "learning_rate": 1.6452610954323337e-05, | |
| "loss": 0.0079, | |
| "reward": 0.42410715818405154, | |
| "reward_std": 0.2372832555323839, | |
| "rewards/accuracy_reward": 0.42410715818405154, | |
| "step": 130 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 932.5196838378906, | |
| "epoch": 0.3918824352694192, | |
| "grad_norm": 0.3386368751525879, | |
| "kl": 0.0993896484375, | |
| "learning_rate": 1.5726871495339563e-05, | |
| "loss": 0.0242, | |
| "reward": 0.43839287767186763, | |
| "reward_std": 0.23617825247347354, | |
| "rewards/accuracy_reward": 0.43839287767186763, | |
| "step": 140 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 888.8741516113281, | |
| "epoch": 0.4198740377886634, | |
| "grad_norm": 0.4019307792186737, | |
| "kl": 0.1232421875, | |
| "learning_rate": 1.4955892640410717e-05, | |
| "loss": 0.0233, | |
| "reward": 0.4660714529454708, | |
| "reward_std": 0.2454029094427824, | |
| "rewards/accuracy_reward": 0.4660714529454708, | |
| "step": 150 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 801.5536071777344, | |
| "epoch": 0.44786564030790765, | |
| "grad_norm": 0.3658553957939148, | |
| "kl": 0.1235595703125, | |
| "learning_rate": 1.4147053184944674e-05, | |
| "loss": 0.016, | |
| "reward": 0.4410714492201805, | |
| "reward_std": 0.25928416270762683, | |
| "rewards/accuracy_reward": 0.4410714492201805, | |
| "step": 160 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 689.7009254455567, | |
| "epoch": 0.47585724282715186, | |
| "grad_norm": 0.5175524950027466, | |
| "kl": 0.17607421875, | |
| "learning_rate": 1.3308094276213557e-05, | |
| "loss": 0.0262, | |
| "reward": 0.4401785898953676, | |
| "reward_std": 0.23054485712200404, | |
| "rewards/accuracy_reward": 0.4401785898953676, | |
| "step": 170 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 613.9518119812012, | |
| "epoch": 0.5038488453463961, | |
| "grad_norm": 0.38585546612739563, | |
| "kl": 0.1792724609375, | |
| "learning_rate": 1.2447045325275215e-05, | |
| "loss": 0.0113, | |
| "reward": 0.4428571630269289, | |
| "reward_std": 0.2558211121708155, | |
| "rewards/accuracy_reward": 0.4428571630269289, | |
| "step": 180 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 767.7786056518555, | |
| "epoch": 0.5318404478656403, | |
| "grad_norm": 0.22975799441337585, | |
| "kl": 0.10023193359375, | |
| "learning_rate": 1.1572147160012956e-05, | |
| "loss": 0.0294, | |
| "reward": 0.45357144996523857, | |
| "reward_std": 0.2570741597563028, | |
| "rewards/accuracy_reward": 0.45357144996523857, | |
| "step": 190 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 1036.6339736938476, | |
| "epoch": 0.5598320503848845, | |
| "grad_norm": 0.3325261175632477, | |
| "kl": 0.1287353515625, | |
| "learning_rate": 1.0691773154771508e-05, | |
| "loss": 0.0119, | |
| "reward": 0.4821428835391998, | |
| "reward_std": 0.2324786176905036, | |
| "rewards/accuracy_reward": 0.4821428835391998, | |
| "step": 200 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 868.197361755371, | |
| "epoch": 0.5878236529041287, | |
| "grad_norm": 0.6180446743965149, | |
| "kl": 0.1508056640625, | |
| "learning_rate": 9.814349091432634e-06, | |
| "loss": 0.0277, | |
| "reward": 0.4821428798139095, | |
| "reward_std": 0.2132600512355566, | |
| "rewards/accuracy_reward": 0.4821428798139095, | |
| "step": 210 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 818.4768188476562, | |
| "epoch": 0.615815255423373, | |
| "grad_norm": 0.2782590687274933, | |
| "kl": 0.127099609375, | |
| "learning_rate": 8.948272518914737e-06, | |
| "loss": 0.0071, | |
| "reward": 0.4758928822353482, | |
| "reward_std": 0.22520754840224982, | |
| "rewards/accuracy_reward": 0.4758928822353482, | |
| "step": 220 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 785.9794929504394, | |
| "epoch": 0.6438068579426172, | |
| "grad_norm": 0.2999161183834076, | |
| "kl": 0.10723876953125, | |
| "learning_rate": 8.101832382881249e-06, | |
| "loss": 0.0243, | |
| "reward": 0.4767857313156128, | |
| "reward_std": 0.2523977212607861, | |
| "rewards/accuracy_reward": 0.4767857313156128, | |
| "step": 230 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 833.9795013427735, | |
| "epoch": 0.6717984604618614, | |
| "grad_norm": 0.28707244992256165, | |
| "kl": 0.1548828125, | |
| "learning_rate": 7.283129694856508e-06, | |
| "loss": 0.0181, | |
| "reward": 0.420535734295845, | |
| "reward_std": 0.24855004157871008, | |
| "rewards/accuracy_reward": 0.420535734295845, | |
| "step": 240 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 825.6973556518554, | |
| "epoch": 0.6997900629811057, | |
| "grad_norm": 0.2718600630760193, | |
| "kl": 0.2059814453125, | |
| "learning_rate": 6.500000000000003e-06, | |
| "loss": 0.0192, | |
| "reward": 0.46696431189775467, | |
| "reward_std": 0.2507600516080856, | |
| "rewards/accuracy_reward": 0.46696431189775467, | |
| "step": 250 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 729.3732452392578, | |
| "epoch": 0.72778166550035, | |
| "grad_norm": 0.327871173620224, | |
| "kl": 0.1481689453125, | |
| "learning_rate": 5.759938385575454e-06, | |
| "loss": 0.0284, | |
| "reward": 0.48750002318993213, | |
| "reward_std": 0.25541665144264697, | |
| "rewards/accuracy_reward": 0.48750002318993213, | |
| "step": 260 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 769.4178924560547, | |
| "epoch": 0.7557732680195941, | |
| "grad_norm": 0.3249359428882599, | |
| "kl": 0.23892822265625, | |
| "learning_rate": 5.070027747835002e-06, | |
| "loss": 0.024, | |
| "reward": 0.48839287757873534, | |
| "reward_std": 0.2081593234091997, | |
| "rewards/accuracy_reward": 0.48839287757873534, | |
| "step": 270 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 738.6937805175781, | |
| "epoch": 0.7837648705388384, | |
| "grad_norm": 0.2741276025772095, | |
| "kl": 0.20296630859375, | |
| "learning_rate": 4.436871003853553e-06, | |
| "loss": 0.0376, | |
| "reward": 0.47500001937150954, | |
| "reward_std": 0.26311200819909575, | |
| "rewards/accuracy_reward": 0.47500001937150954, | |
| "step": 280 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 706.5491378784179, | |
| "epoch": 0.8117564730580826, | |
| "grad_norm": 0.29115381836891174, | |
| "kl": 0.1580322265625, | |
| "learning_rate": 3.866527897092401e-06, | |
| "loss": 0.0139, | |
| "reward": 0.4812500238418579, | |
| "reward_std": 0.2547359408810735, | |
| "rewards/accuracy_reward": 0.4812500238418579, | |
| "step": 290 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 751.8768203735351, | |
| "epoch": 0.8397480755773268, | |
| "grad_norm": 0.26477473974227905, | |
| "kl": 0.1802001953125, | |
| "learning_rate": 3.364457001506166e-06, | |
| "loss": 0.0253, | |
| "reward": 0.4580357393249869, | |
| "reward_std": 0.2625793442130089, | |
| "rewards/accuracy_reward": 0.4580357393249869, | |
| "step": 300 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 732.6241462707519, | |
| "epoch": 0.867739678096571, | |
| "grad_norm": 0.3064480125904083, | |
| "kl": 0.19808349609375, | |
| "learning_rate": 2.935463479253442e-06, | |
| "loss": 0.02, | |
| "reward": 0.40446430407464506, | |
| "reward_std": 0.2393452214077115, | |
| "rewards/accuracy_reward": 0.40446430407464506, | |
| "step": 310 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 715.6580657958984, | |
| "epoch": 0.8957312806158153, | |
| "grad_norm": 0.37576210498809814, | |
| "kl": 0.19476318359375, | |
| "learning_rate": 2.5836530920055976e-06, | |
| "loss": 0.0327, | |
| "reward": 0.40625001918524506, | |
| "reward_std": 0.23536933306604624, | |
| "rewards/accuracy_reward": 0.40625001918524506, | |
| "step": 320 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 731.3232460021973, | |
| "epoch": 0.9237228831350595, | |
| "grad_norm": 0.3001156151294708, | |
| "kl": 0.3147216796875, | |
| "learning_rate": 2.3123929059970286e-06, | |
| "loss": 0.0557, | |
| "reward": 0.4071428745985031, | |
| "reward_std": 0.22367824967950584, | |
| "rewards/accuracy_reward": 0.4071428745985031, | |
| "step": 330 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 669.9786071777344, | |
| "epoch": 0.9517144856543037, | |
| "grad_norm": 0.504320502281189, | |
| "kl": 0.3324951171875, | |
| "learning_rate": 2.1242790668964046e-06, | |
| "loss": 0.0244, | |
| "reward": 0.42232145071029664, | |
| "reward_std": 0.2736980877816677, | |
| "rewards/accuracy_reward": 0.42232145071029664, | |
| "step": 340 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 650.3339591979981, | |
| "epoch": 0.979706088173548, | |
| "grad_norm": 0.5142662525177002, | |
| "kl": 0.21497802734375, | |
| "learning_rate": 2.021111952915447e-06, | |
| "loss": 0.0222, | |
| "reward": 0.43214288018643854, | |
| "reward_std": 0.2758882647380233, | |
| "rewards/accuracy_reward": 0.43214288018643854, | |
| "step": 350 | |
| }, | |
| { | |
| "clip_ratio": 0.0, | |
| "completion_length": 621.5370178222656, | |
| "epoch": 0.9993002099370188, | |
| "kl": 0.20172991071428573, | |
| "reward": 0.45025512549493996, | |
| "reward_std": 0.22983166708477906, | |
| "rewards/accuracy_reward": 0.45025512549493996, | |
| "step": 357, | |
| "total_flos": 0.0, | |
| "train_loss": 0.01889443341694793, | |
| "train_runtime": 24438.0983, | |
| "train_samples_per_second": 0.409, | |
| "train_steps_per_second": 0.015 | |
| } | |
| ], | |
| "logging_steps": 10, | |
| "max_steps": 357, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 1, | |
| "save_steps": 500, | |
| "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 | |
| } | |