| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 0.9965010496850945, | |
| "eval_steps": 100, | |
| "global_step": 178, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "completion_length": 548.2455501556396, | |
| "epoch": 0.005598320503848845, | |
| "grad_norm": 0.003676735097542405, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.03125000128056854, | |
| "reward_std": 0.035294653847813606, | |
| "rewards/accuracy_reward": 0.03125000128056854, | |
| "step": 1 | |
| }, | |
| { | |
| "completion_length": 521.3355770111084, | |
| "epoch": 0.01119664100769769, | |
| "grad_norm": 0.001673316117376089, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.005952381179668009, | |
| "reward_std": 0.008138235658407211, | |
| "rewards/accuracy_reward": 0.005952381179668009, | |
| "step": 2 | |
| }, | |
| { | |
| "completion_length": 565.3147449493408, | |
| "epoch": 0.016794961511546535, | |
| "grad_norm": 0.002517723012715578, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.009672619285993278, | |
| "reward_std": 0.01636804547160864, | |
| "rewards/accuracy_reward": 0.009672619285993278, | |
| "step": 3 | |
| }, | |
| { | |
| "completion_length": 577.5029792785645, | |
| "epoch": 0.02239328201539538, | |
| "grad_norm": 0.003224034095183015, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.029761905316263437, | |
| "reward_std": 0.02953372267074883, | |
| "rewards/accuracy_reward": 0.029761905316263437, | |
| "step": 4 | |
| }, | |
| { | |
| "completion_length": 564.710578918457, | |
| "epoch": 0.02799160251924423, | |
| "grad_norm": 0.0038144055288285017, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.020833333779592067, | |
| "reward_std": 0.022426264360547066, | |
| "rewards/accuracy_reward": 0.020833333779592067, | |
| "step": 5 | |
| }, | |
| { | |
| "completion_length": 534.8861694335938, | |
| "epoch": 0.03358992302309307, | |
| "grad_norm": 0.003479737089946866, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.026785714842844754, | |
| "reward_std": 0.031758472323417664, | |
| "rewards/accuracy_reward": 0.026785714842844754, | |
| "step": 6 | |
| }, | |
| { | |
| "completion_length": 580.7991256713867, | |
| "epoch": 0.03918824352694192, | |
| "grad_norm": 0.0025944788940250874, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.014880952483508736, | |
| "reward_std": 0.010090996511280537, | |
| "rewards/accuracy_reward": 0.014880952483508736, | |
| "step": 7 | |
| }, | |
| { | |
| "completion_length": 515.3296279907227, | |
| "epoch": 0.04478656403079076, | |
| "grad_norm": 0.003000692930072546, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.021577381354290992, | |
| "reward_std": 0.019644177984446287, | |
| "rewards/accuracy_reward": 0.021577381354290992, | |
| "step": 8 | |
| }, | |
| { | |
| "completion_length": 571.0290279388428, | |
| "epoch": 0.05038488453463961, | |
| "grad_norm": 0.003332852851599455, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.02901785750873387, | |
| "reward_std": 0.026531722396612167, | |
| "rewards/accuracy_reward": 0.02901785750873387, | |
| "step": 9 | |
| }, | |
| { | |
| "completion_length": 580.1770935058594, | |
| "epoch": 0.05598320503848846, | |
| "grad_norm": 0.0018766584107652307, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.0074404762126505375, | |
| "reward_std": 0.003475441597402096, | |
| "rewards/accuracy_reward": 0.0074404762126505375, | |
| "step": 10 | |
| }, | |
| { | |
| "completion_length": 576.3244209289551, | |
| "epoch": 0.0615815255423373, | |
| "grad_norm": 0.004447213374078274, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.03497023892123252, | |
| "reward_std": 0.03141464572399855, | |
| "rewards/accuracy_reward": 0.03497023892123252, | |
| "step": 11 | |
| }, | |
| { | |
| "completion_length": 555.8616218566895, | |
| "epoch": 0.06717984604618614, | |
| "grad_norm": 0.003440001280978322, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.033482144062872976, | |
| "reward_std": 0.03174867480993271, | |
| "rewards/accuracy_reward": 0.033482144062872976, | |
| "step": 12 | |
| }, | |
| { | |
| "completion_length": 555.1770973205566, | |
| "epoch": 0.072778166550035, | |
| "grad_norm": 0.004410188645124435, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.02604166732635349, | |
| "reward_std": 0.03677397267892957, | |
| "rewards/accuracy_reward": 0.02604166732635349, | |
| "step": 13 | |
| }, | |
| { | |
| "completion_length": 582.8259029388428, | |
| "epoch": 0.07837648705388384, | |
| "grad_norm": 0.0032948441803455353, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.019345238630194217, | |
| "reward_std": 0.02020683465525508, | |
| "rewards/accuracy_reward": 0.019345238630194217, | |
| "step": 14 | |
| }, | |
| { | |
| "completion_length": 573.2150402069092, | |
| "epoch": 0.08397480755773268, | |
| "grad_norm": 0.004153064452111721, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.02083333395421505, | |
| "reward_std": 0.024796947836875916, | |
| "rewards/accuracy_reward": 0.02083333395421505, | |
| "step": 15 | |
| }, | |
| { | |
| "completion_length": 511.9151840209961, | |
| "epoch": 0.08957312806158152, | |
| "grad_norm": 0.0052322824485599995, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.04017857229337096, | |
| "reward_std": 0.04290085076354444, | |
| "rewards/accuracy_reward": 0.04017857229337096, | |
| "step": 16 | |
| }, | |
| { | |
| "completion_length": 538.8735218048096, | |
| "epoch": 0.09517144856543037, | |
| "grad_norm": 0.005073050037026405, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.0349702388048172, | |
| "reward_std": 0.03157654171809554, | |
| "rewards/accuracy_reward": 0.0349702388048172, | |
| "step": 17 | |
| }, | |
| { | |
| "completion_length": 594.4494113922119, | |
| "epoch": 0.10076976906927922, | |
| "grad_norm": 0.004708003718405962, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.030505953240208328, | |
| "reward_std": 0.035523281432688236, | |
| "rewards/accuracy_reward": 0.030505953240208328, | |
| "step": 18 | |
| }, | |
| { | |
| "completion_length": 557.1242733001709, | |
| "epoch": 0.10636808957312806, | |
| "grad_norm": 0.0028041391633450985, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.02827381028328091, | |
| "reward_std": 0.01928615104407072, | |
| "rewards/accuracy_reward": 0.02827381028328091, | |
| "step": 19 | |
| }, | |
| { | |
| "completion_length": 560.6994209289551, | |
| "epoch": 0.11196641007697691, | |
| "grad_norm": 0.004033979959785938, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.029761905258055776, | |
| "reward_std": 0.03157439874485135, | |
| "rewards/accuracy_reward": 0.029761905258055776, | |
| "step": 20 | |
| }, | |
| { | |
| "completion_length": 565.3564147949219, | |
| "epoch": 0.11756473058082575, | |
| "grad_norm": 0.0020209557842463255, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.004464285797439516, | |
| "reward_std": 0.00787156680598855, | |
| "rewards/accuracy_reward": 0.004464285797439516, | |
| "step": 21 | |
| }, | |
| { | |
| "completion_length": 618.4695091247559, | |
| "epoch": 0.1231630510846746, | |
| "grad_norm": 0.002670794492587447, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.021577381703536958, | |
| "reward_std": 0.01789072621613741, | |
| "rewards/accuracy_reward": 0.021577381703536958, | |
| "step": 22 | |
| }, | |
| { | |
| "completion_length": 584.6480770111084, | |
| "epoch": 0.12876137158852344, | |
| "grad_norm": 0.005253300536423922, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.03720238187815994, | |
| "reward_std": 0.03154852241277695, | |
| "rewards/accuracy_reward": 0.03720238187815994, | |
| "step": 23 | |
| }, | |
| { | |
| "completion_length": 581.5640029907227, | |
| "epoch": 0.13435969209237228, | |
| "grad_norm": 0.0021275475155562162, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.005952381179668009, | |
| "reward_std": 0.008138235658407211, | |
| "rewards/accuracy_reward": 0.005952381179668009, | |
| "step": 24 | |
| }, | |
| { | |
| "completion_length": 545.2701072692871, | |
| "epoch": 0.13995801259622112, | |
| "grad_norm": 0.002449671970680356, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.015625000465661287, | |
| "reward_std": 0.017629378009587526, | |
| "rewards/accuracy_reward": 0.015625000465661287, | |
| "step": 25 | |
| }, | |
| { | |
| "completion_length": 545.0669708251953, | |
| "epoch": 0.14555633310007, | |
| "grad_norm": 0.0028808764182031155, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.0200892859720625, | |
| "reward_std": 0.018164015375077724, | |
| "rewards/accuracy_reward": 0.0200892859720625, | |
| "step": 26 | |
| }, | |
| { | |
| "completion_length": 561.1131038665771, | |
| "epoch": 0.15115465360391883, | |
| "grad_norm": 0.004521294496953487, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.043898809934034944, | |
| "reward_std": 0.027857428416609764, | |
| "rewards/accuracy_reward": 0.043898809934034944, | |
| "step": 27 | |
| }, | |
| { | |
| "completion_length": 573.0468807220459, | |
| "epoch": 0.15675297410776767, | |
| "grad_norm": 0.004434187430888414, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.03050595335662365, | |
| "reward_std": 0.03036304796114564, | |
| "rewards/accuracy_reward": 0.03050595335662365, | |
| "step": 28 | |
| }, | |
| { | |
| "completion_length": 573.6540298461914, | |
| "epoch": 0.16235129461161651, | |
| "grad_norm": 0.0023913481272757053, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.03422619198681787, | |
| "reward_std": 0.01964203454554081, | |
| "rewards/accuracy_reward": 0.03422619198681787, | |
| "step": 29 | |
| }, | |
| { | |
| "completion_length": 582.3474769592285, | |
| "epoch": 0.16794961511546536, | |
| "grad_norm": 0.0027055193204432726, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.027529762592166662, | |
| "reward_std": 0.02218207810074091, | |
| "rewards/accuracy_reward": 0.027529762592166662, | |
| "step": 30 | |
| }, | |
| { | |
| "completion_length": 583.9047756195068, | |
| "epoch": 0.1735479356193142, | |
| "grad_norm": 0.003626331454142928, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.03720238181995228, | |
| "reward_std": 0.028277710545808077, | |
| "rewards/accuracy_reward": 0.03720238181995228, | |
| "step": 31 | |
| }, | |
| { | |
| "completion_length": 576.6346855163574, | |
| "epoch": 0.17914625612316304, | |
| "grad_norm": 0.0032239772845059633, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.015625000291038305, | |
| "reward_std": 0.022000661585479975, | |
| "rewards/accuracy_reward": 0.015625000291038305, | |
| "step": 32 | |
| }, | |
| { | |
| "completion_length": 568.7916736602783, | |
| "epoch": 0.1847445766270119, | |
| "grad_norm": 0.0029333089478313923, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.025297619693446904, | |
| "reward_std": 0.020201513543725014, | |
| "rewards/accuracy_reward": 0.025297619693446904, | |
| "step": 33 | |
| }, | |
| { | |
| "completion_length": 555.8154830932617, | |
| "epoch": 0.19034289713086075, | |
| "grad_norm": 0.003082014387473464, | |
| "learning_rate": 3e-07, | |
| "loss": -0.0, | |
| "reward": 0.01264880975941196, | |
| "reward_std": 0.019351367838680744, | |
| "rewards/accuracy_reward": 0.01264880975941196, | |
| "step": 34 | |
| }, | |
| { | |
| "completion_length": 593.186767578125, | |
| "epoch": 0.1959412176347096, | |
| "grad_norm": 0.003542139893397689, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.029017857741564512, | |
| "reward_std": 0.03906076308339834, | |
| "rewards/accuracy_reward": 0.029017857741564512, | |
| "step": 35 | |
| }, | |
| { | |
| "completion_length": 539.3273983001709, | |
| "epoch": 0.20153953813855843, | |
| "grad_norm": 0.0035734642297029495, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.04092262004269287, | |
| "reward_std": 0.02163945697247982, | |
| "rewards/accuracy_reward": 0.04092262004269287, | |
| "step": 36 | |
| }, | |
| { | |
| "completion_length": 528.0818519592285, | |
| "epoch": 0.20713785864240727, | |
| "grad_norm": 0.004155021160840988, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.04241071513388306, | |
| "reward_std": 0.031058762688189745, | |
| "rewards/accuracy_reward": 0.04241071513388306, | |
| "step": 37 | |
| }, | |
| { | |
| "completion_length": 572.0044708251953, | |
| "epoch": 0.21273617914625612, | |
| "grad_norm": 0.0043424940668046474, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.031250000349245965, | |
| "reward_std": 0.03391032665967941, | |
| "rewards/accuracy_reward": 0.031250000349245965, | |
| "step": 38 | |
| }, | |
| { | |
| "completion_length": 561.321439743042, | |
| "epoch": 0.21833449965010496, | |
| "grad_norm": 0.0013517189072445035, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.014136905316263437, | |
| "reward_std": 0.004597577266395092, | |
| "rewards/accuracy_reward": 0.014136905316263437, | |
| "step": 39 | |
| }, | |
| { | |
| "completion_length": 604.6815567016602, | |
| "epoch": 0.22393282015395383, | |
| "grad_norm": 0.004611727315932512, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.04464285826543346, | |
| "reward_std": 0.042770151514559984, | |
| "rewards/accuracy_reward": 0.04464285826543346, | |
| "step": 40 | |
| }, | |
| { | |
| "completion_length": 591.9910755157471, | |
| "epoch": 0.22953114065780267, | |
| "grad_norm": 0.003368583507835865, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.03720238246023655, | |
| "reward_std": 0.022582839708775282, | |
| "rewards/accuracy_reward": 0.03720238246023655, | |
| "step": 41 | |
| }, | |
| { | |
| "completion_length": 585.0193538665771, | |
| "epoch": 0.2351294611616515, | |
| "grad_norm": 0.0034278561361134052, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.04315476305782795, | |
| "reward_std": 0.02897424390539527, | |
| "rewards/accuracy_reward": 0.04315476305782795, | |
| "step": 42 | |
| }, | |
| { | |
| "completion_length": 588.648078918457, | |
| "epoch": 0.24072778166550035, | |
| "grad_norm": 0.003546294756233692, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.04092262068297714, | |
| "reward_std": 0.02937500481493771, | |
| "rewards/accuracy_reward": 0.04092262068297714, | |
| "step": 43 | |
| }, | |
| { | |
| "completion_length": 591.8616180419922, | |
| "epoch": 0.2463261021693492, | |
| "grad_norm": 0.0032863786909729242, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.029017857741564512, | |
| "reward_std": 0.02016195748001337, | |
| "rewards/accuracy_reward": 0.029017857741564512, | |
| "step": 44 | |
| }, | |
| { | |
| "completion_length": 595.4494113922119, | |
| "epoch": 0.25192442267319803, | |
| "grad_norm": 0.003891779575496912, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.0513392873108387, | |
| "reward_std": 0.021664298605173826, | |
| "rewards/accuracy_reward": 0.0513392873108387, | |
| "step": 45 | |
| }, | |
| { | |
| "completion_length": 576.5751647949219, | |
| "epoch": 0.2575227431770469, | |
| "grad_norm": 0.004738082177937031, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.04092262021731585, | |
| "reward_std": 0.03612527949735522, | |
| "rewards/accuracy_reward": 0.04092262021731585, | |
| "step": 46 | |
| }, | |
| { | |
| "completion_length": 582.7514915466309, | |
| "epoch": 0.2631210636808957, | |
| "grad_norm": 0.005118933971971273, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.044642857857979834, | |
| "reward_std": 0.02991444803774357, | |
| "rewards/accuracy_reward": 0.044642857857979834, | |
| "step": 47 | |
| }, | |
| { | |
| "completion_length": 597.6398983001709, | |
| "epoch": 0.26871938418474456, | |
| "grad_norm": 0.003335759276524186, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.017857143247965723, | |
| "reward_std": 0.02291816775687039, | |
| "rewards/accuracy_reward": 0.017857143247965723, | |
| "step": 48 | |
| }, | |
| { | |
| "completion_length": 632.6994171142578, | |
| "epoch": 0.2743177046885934, | |
| "grad_norm": 0.002710092579945922, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.026785714959260076, | |
| "reward_std": 0.018883246928453445, | |
| "rewards/accuracy_reward": 0.026785714959260076, | |
| "step": 49 | |
| }, | |
| { | |
| "completion_length": 614.4695072174072, | |
| "epoch": 0.27991602519244224, | |
| "grad_norm": 0.004074351862072945, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.034970239736139774, | |
| "reward_std": 0.0352512919344008, | |
| "rewards/accuracy_reward": 0.034970239736139774, | |
| "step": 50 | |
| }, | |
| { | |
| "completion_length": 564.6369132995605, | |
| "epoch": 0.28551434569629114, | |
| "grad_norm": 0.00433464627712965, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.04761904897168279, | |
| "reward_std": 0.036211316008120775, | |
| "rewards/accuracy_reward": 0.04761904897168279, | |
| "step": 51 | |
| }, | |
| { | |
| "completion_length": 574.7514991760254, | |
| "epoch": 0.29111266620014, | |
| "grad_norm": 0.0021451774518936872, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.02232142904540524, | |
| "reward_std": 0.01412591733969748, | |
| "rewards/accuracy_reward": 0.02232142904540524, | |
| "step": 52 | |
| }, | |
| { | |
| "completion_length": 587.8154850006104, | |
| "epoch": 0.2967109867039888, | |
| "grad_norm": 0.004314142279326916, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.04017857281723991, | |
| "reward_std": 0.03161609871312976, | |
| "rewards/accuracy_reward": 0.04017857281723991, | |
| "step": 53 | |
| }, | |
| { | |
| "completion_length": 612.7529888153076, | |
| "epoch": 0.30230930720783766, | |
| "grad_norm": 0.003322584554553032, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.023809524485841393, | |
| "reward_std": 0.029193072579801083, | |
| "rewards/accuracy_reward": 0.023809524485841393, | |
| "step": 54 | |
| }, | |
| { | |
| "completion_length": 602.6979274749756, | |
| "epoch": 0.3079076277116865, | |
| "grad_norm": 0.0024285970721393824, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.014136904967017472, | |
| "reward_std": 0.017268173396587372, | |
| "rewards/accuracy_reward": 0.014136904967017472, | |
| "step": 55 | |
| }, | |
| { | |
| "completion_length": 594.776798248291, | |
| "epoch": 0.31350594821553535, | |
| "grad_norm": 0.002405191073194146, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.031250000931322575, | |
| "reward_std": 0.008076196536421776, | |
| "rewards/accuracy_reward": 0.031250000931322575, | |
| "step": 56 | |
| }, | |
| { | |
| "completion_length": 569.1354274749756, | |
| "epoch": 0.3191042687193842, | |
| "grad_norm": 0.004245223011821508, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.056547619809862226, | |
| "reward_std": 0.03459409927017987, | |
| "rewards/accuracy_reward": 0.056547619809862226, | |
| "step": 57 | |
| }, | |
| { | |
| "completion_length": 610.0952472686768, | |
| "epoch": 0.32470258922323303, | |
| "grad_norm": 0.003974903374910355, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.035714287078008056, | |
| "reward_std": 0.02040082309395075, | |
| "rewards/accuracy_reward": 0.035714287078008056, | |
| "step": 58 | |
| }, | |
| { | |
| "completion_length": 645.4836387634277, | |
| "epoch": 0.33030090972708187, | |
| "grad_norm": 0.0026164394803345203, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.011160714784637094, | |
| "reward_std": 0.011207811534404755, | |
| "rewards/accuracy_reward": 0.011160714784637094, | |
| "step": 59 | |
| }, | |
| { | |
| "completion_length": 626.8445014953613, | |
| "epoch": 0.3358992302309307, | |
| "grad_norm": 0.0031215217895805836, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.024553572467993945, | |
| "reward_std": 0.02602896187454462, | |
| "rewards/accuracy_reward": 0.024553572467993945, | |
| "step": 60 | |
| }, | |
| { | |
| "completion_length": 602.254472732544, | |
| "epoch": 0.34149755073477955, | |
| "grad_norm": 0.0025609612930566072, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.03273809637175873, | |
| "reward_std": 0.02431014971807599, | |
| "rewards/accuracy_reward": 0.03273809637175873, | |
| "step": 61 | |
| }, | |
| { | |
| "completion_length": 605.808048248291, | |
| "epoch": 0.3470958712386284, | |
| "grad_norm": 0.0041216155514121056, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.0223214291036129, | |
| "reward_std": 0.027785591781139374, | |
| "rewards/accuracy_reward": 0.0223214291036129, | |
| "step": 62 | |
| }, | |
| { | |
| "completion_length": 620.4248657226562, | |
| "epoch": 0.35269419174247724, | |
| "grad_norm": 0.0027605677023530006, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.020089286204893142, | |
| "reward_std": 0.02404030319303274, | |
| "rewards/accuracy_reward": 0.020089286204893142, | |
| "step": 63 | |
| }, | |
| { | |
| "completion_length": 616.587064743042, | |
| "epoch": 0.3582925122463261, | |
| "grad_norm": 0.006433432921767235, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.034970238571986556, | |
| "reward_std": 0.0240207826718688, | |
| "rewards/accuracy_reward": 0.034970238571986556, | |
| "step": 64 | |
| }, | |
| { | |
| "completion_length": 656.1890068054199, | |
| "epoch": 0.363890832750175, | |
| "grad_norm": 0.002678812015801668, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.038690476736519486, | |
| "reward_std": 0.026529580354690552, | |
| "rewards/accuracy_reward": 0.038690476736519486, | |
| "step": 65 | |
| }, | |
| { | |
| "completion_length": 615.4241199493408, | |
| "epoch": 0.3694891532540238, | |
| "grad_norm": 0.004133144393563271, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.055059525300748646, | |
| "reward_std": 0.030434885527938604, | |
| "rewards/accuracy_reward": 0.055059525300748646, | |
| "step": 66 | |
| }, | |
| { | |
| "completion_length": 631.1212978363037, | |
| "epoch": 0.37508747375787266, | |
| "grad_norm": 0.004037661012262106, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.038690477376803756, | |
| "reward_std": 0.028071781154721975, | |
| "rewards/accuracy_reward": 0.038690477376803756, | |
| "step": 67 | |
| }, | |
| { | |
| "completion_length": 577.1934661865234, | |
| "epoch": 0.3806857942617215, | |
| "grad_norm": 0.0031959640327841043, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.023809524427633733, | |
| "reward_std": 0.0161665934138, | |
| "rewards/accuracy_reward": 0.023809524427633733, | |
| "step": 68 | |
| }, | |
| { | |
| "completion_length": 645.9285850524902, | |
| "epoch": 0.38628411476557034, | |
| "grad_norm": 0.004407494328916073, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.030505953065585345, | |
| "reward_std": 0.03433060785755515, | |
| "rewards/accuracy_reward": 0.030505953065585345, | |
| "step": 69 | |
| }, | |
| { | |
| "completion_length": 629.8266506195068, | |
| "epoch": 0.3918824352694192, | |
| "grad_norm": 0.0024215257726609707, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.019345238513778895, | |
| "reward_std": 0.02312279725447297, | |
| "rewards/accuracy_reward": 0.019345238513778895, | |
| "step": 70 | |
| }, | |
| { | |
| "completion_length": 604.0178699493408, | |
| "epoch": 0.397480755773268, | |
| "grad_norm": 0.003344905562698841, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.02306547691114247, | |
| "reward_std": 0.019507942255586386, | |
| "rewards/accuracy_reward": 0.02306547691114247, | |
| "step": 71 | |
| }, | |
| { | |
| "completion_length": 590.7827472686768, | |
| "epoch": 0.40307907627711687, | |
| "grad_norm": 0.0028942637145519257, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.05803571501746774, | |
| "reward_std": 0.03128055343404412, | |
| "rewards/accuracy_reward": 0.05803571501746774, | |
| "step": 72 | |
| }, | |
| { | |
| "completion_length": 599.7842330932617, | |
| "epoch": 0.4086773967809657, | |
| "grad_norm": 0.0034949486143887043, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.05580357339931652, | |
| "reward_std": 0.015248052775859833, | |
| "rewards/accuracy_reward": 0.05580357339931652, | |
| "step": 73 | |
| }, | |
| { | |
| "completion_length": 635.7009048461914, | |
| "epoch": 0.41427571728481455, | |
| "grad_norm": 0.0022625280544161797, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.02678571466822177, | |
| "reward_std": 0.021662155631929636, | |
| "rewards/accuracy_reward": 0.02678571466822177, | |
| "step": 74 | |
| }, | |
| { | |
| "completion_length": 631.5788726806641, | |
| "epoch": 0.4198740377886634, | |
| "grad_norm": 0.0037094622384756804, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.02604166726814583, | |
| "reward_std": 0.02264805557206273, | |
| "rewards/accuracy_reward": 0.02604166726814583, | |
| "step": 75 | |
| }, | |
| { | |
| "completion_length": 636.7358722686768, | |
| "epoch": 0.42547235829251223, | |
| "grad_norm": 0.003503567073494196, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.03273809503298253, | |
| "reward_std": 0.02112219762057066, | |
| "rewards/accuracy_reward": 0.03273809503298253, | |
| "step": 76 | |
| }, | |
| { | |
| "completion_length": 635.9628219604492, | |
| "epoch": 0.4310706787963611, | |
| "grad_norm": 0.0047474331222474575, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.04166666779201478, | |
| "reward_std": 0.03789610881358385, | |
| "rewards/accuracy_reward": 0.04166666779201478, | |
| "step": 77 | |
| }, | |
| { | |
| "completion_length": 638.1703987121582, | |
| "epoch": 0.4366689993002099, | |
| "grad_norm": 0.0018300635274499655, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.05282738315872848, | |
| "reward_std": 0.01760453707538545, | |
| "rewards/accuracy_reward": 0.05282738315872848, | |
| "step": 78 | |
| }, | |
| { | |
| "completion_length": 617.7492637634277, | |
| "epoch": 0.44226731980405876, | |
| "grad_norm": 0.0036325210239738226, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.04092262010090053, | |
| "reward_std": 0.03621981432661414, | |
| "rewards/accuracy_reward": 0.04092262010090053, | |
| "step": 79 | |
| }, | |
| { | |
| "completion_length": 621.1123561859131, | |
| "epoch": 0.44786564030790765, | |
| "grad_norm": 0.0025254676584154367, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.04092262004269287, | |
| "reward_std": 0.02054001996293664, | |
| "rewards/accuracy_reward": 0.04092262004269287, | |
| "step": 80 | |
| }, | |
| { | |
| "completion_length": 670.346004486084, | |
| "epoch": 0.4534639608117565, | |
| "grad_norm": 0.0038352429401129484, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.03497023903764784, | |
| "reward_std": 0.03099754173308611, | |
| "rewards/accuracy_reward": 0.03497023903764784, | |
| "step": 81 | |
| }, | |
| { | |
| "completion_length": 632.6949424743652, | |
| "epoch": 0.45906228131560534, | |
| "grad_norm": 0.005097648594528437, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.05059524026000872, | |
| "reward_std": 0.02744494192302227, | |
| "rewards/accuracy_reward": 0.05059524026000872, | |
| "step": 82 | |
| }, | |
| { | |
| "completion_length": 630.6183204650879, | |
| "epoch": 0.4646606018194542, | |
| "grad_norm": 0.0024996348656713963, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.012648809934034944, | |
| "reward_std": 0.014683252666145563, | |
| "rewards/accuracy_reward": 0.012648809934034944, | |
| "step": 83 | |
| }, | |
| { | |
| "completion_length": 582.6726245880127, | |
| "epoch": 0.470258922323303, | |
| "grad_norm": 0.0026008612476289272, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.037202382169198245, | |
| "reward_std": 0.021662155631929636, | |
| "rewards/accuracy_reward": 0.037202382169198245, | |
| "step": 84 | |
| }, | |
| { | |
| "completion_length": 665.8355827331543, | |
| "epoch": 0.47585724282715186, | |
| "grad_norm": 0.002596538746729493, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.04389881080714986, | |
| "reward_std": 0.02897106483578682, | |
| "rewards/accuracy_reward": 0.04389881080714986, | |
| "step": 85 | |
| }, | |
| { | |
| "completion_length": 635.4821529388428, | |
| "epoch": 0.4814555633310007, | |
| "grad_norm": 0.0027116115670651197, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.024553571769502014, | |
| "reward_std": 0.025856829015538096, | |
| "rewards/accuracy_reward": 0.024553571769502014, | |
| "step": 86 | |
| }, | |
| { | |
| "completion_length": 642.225456237793, | |
| "epoch": 0.48705388383484954, | |
| "grad_norm": 0.00335879810154438, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.0535714304423891, | |
| "reward_std": 0.04018289828673005, | |
| "rewards/accuracy_reward": 0.0535714304423891, | |
| "step": 87 | |
| }, | |
| { | |
| "completion_length": 611.2046222686768, | |
| "epoch": 0.4926522043386984, | |
| "grad_norm": 0.004874168895184994, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.06994047714397311, | |
| "reward_std": 0.030564499087631702, | |
| "rewards/accuracy_reward": 0.06994047714397311, | |
| "step": 88 | |
| }, | |
| { | |
| "completion_length": 634.81325340271, | |
| "epoch": 0.4982505248425472, | |
| "grad_norm": 0.0030458923429250717, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.03273809637175873, | |
| "reward_std": 0.025404266081750393, | |
| "rewards/accuracy_reward": 0.03273809637175873, | |
| "step": 89 | |
| }, | |
| { | |
| "completion_length": 606.618314743042, | |
| "epoch": 0.5038488453463961, | |
| "grad_norm": 0.0029052915051579475, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.034226191812194884, | |
| "reward_std": 0.018181392922997475, | |
| "rewards/accuracy_reward": 0.034226191812194884, | |
| "step": 90 | |
| }, | |
| { | |
| "completion_length": 649.934534072876, | |
| "epoch": 0.509447165850245, | |
| "grad_norm": 0.0023663390893489122, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.028273810050450265, | |
| "reward_std": 0.014822450000792742, | |
| "rewards/accuracy_reward": 0.028273810050450265, | |
| "step": 91 | |
| }, | |
| { | |
| "completion_length": 645.944206237793, | |
| "epoch": 0.5150454863540938, | |
| "grad_norm": 0.002700702054426074, | |
| "learning_rate": 3e-07, | |
| "loss": -0.0, | |
| "reward": 0.026041666977107525, | |
| "reward_std": 0.023188014514744282, | |
| "rewards/accuracy_reward": 0.026041666977107525, | |
| "step": 92 | |
| }, | |
| { | |
| "completion_length": 606.7001571655273, | |
| "epoch": 0.5206438068579426, | |
| "grad_norm": 0.002376874443143606, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.02455357206054032, | |
| "reward_std": 0.02123525319620967, | |
| "rewards/accuracy_reward": 0.02455357206054032, | |
| "step": 93 | |
| }, | |
| { | |
| "completion_length": 635.0580425262451, | |
| "epoch": 0.5262421273617914, | |
| "grad_norm": 0.0022008493542671204, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.02678571513388306, | |
| "reward_std": 0.016507241874933243, | |
| "rewards/accuracy_reward": 0.02678571513388306, | |
| "step": 94 | |
| }, | |
| { | |
| "completion_length": 651.9226360321045, | |
| "epoch": 0.5318404478656403, | |
| "grad_norm": 0.003217194229364395, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.04464285826543346, | |
| "reward_std": 0.018883246928453445, | |
| "rewards/accuracy_reward": 0.04464285826543346, | |
| "step": 95 | |
| }, | |
| { | |
| "completion_length": 621.9144458770752, | |
| "epoch": 0.5374387683694891, | |
| "grad_norm": 0.0026905853301286697, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.04166666849050671, | |
| "reward_std": 0.022358688060194254, | |
| "rewards/accuracy_reward": 0.04166666849050671, | |
| "step": 96 | |
| }, | |
| { | |
| "completion_length": 630.8422737121582, | |
| "epoch": 0.543037088873338, | |
| "grad_norm": 0.00430277269333601, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.03422619169577956, | |
| "reward_std": 0.03490746580064297, | |
| "rewards/accuracy_reward": 0.03422619169577956, | |
| "step": 97 | |
| }, | |
| { | |
| "completion_length": 635.1577453613281, | |
| "epoch": 0.5486354093771868, | |
| "grad_norm": 0.0027177336160093546, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.024553572409786284, | |
| "reward_std": 0.01907807867974043, | |
| "rewards/accuracy_reward": 0.024553572409786284, | |
| "step": 98 | |
| }, | |
| { | |
| "completion_length": 621.9509105682373, | |
| "epoch": 0.5542337298810357, | |
| "grad_norm": 0.0029670181684195995, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.04761904920451343, | |
| "reward_std": 0.024105519521981478, | |
| "rewards/accuracy_reward": 0.04761904920451343, | |
| "step": 99 | |
| }, | |
| { | |
| "completion_length": 645.11012840271, | |
| "epoch": 0.5598320503848845, | |
| "grad_norm": 0.004191585350781679, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.03720238246023655, | |
| "reward_std": 0.027021698653697968, | |
| "rewards/accuracy_reward": 0.03720238246023655, | |
| "step": 100 | |
| }, | |
| { | |
| "completion_length": 628.647331237793, | |
| "epoch": 0.5654303708887334, | |
| "grad_norm": 0.005200605373829603, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.04017857293365523, | |
| "reward_std": 0.032400546595454216, | |
| "rewards/accuracy_reward": 0.04017857293365523, | |
| "step": 101 | |
| }, | |
| { | |
| "completion_length": 674.1391506195068, | |
| "epoch": 0.5710286913925823, | |
| "grad_norm": 0.00332686142064631, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.039434525242540985, | |
| "reward_std": 0.02585682924836874, | |
| "rewards/accuracy_reward": 0.039434525242540985, | |
| "step": 102 | |
| }, | |
| { | |
| "completion_length": 635.6376628875732, | |
| "epoch": 0.5766270118964311, | |
| "grad_norm": 0.002928570844233036, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.03348214377183467, | |
| "reward_std": 0.017823366448283195, | |
| "rewards/accuracy_reward": 0.03348214377183467, | |
| "step": 103 | |
| }, | |
| { | |
| "completion_length": 667.8132553100586, | |
| "epoch": 0.58222533240028, | |
| "grad_norm": 0.0027422241400927305, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.05059524020180106, | |
| "reward_std": 0.02381271030753851, | |
| "rewards/accuracy_reward": 0.05059524020180106, | |
| "step": 104 | |
| }, | |
| { | |
| "completion_length": 607.5372142791748, | |
| "epoch": 0.5878236529041287, | |
| "grad_norm": 0.0026503645349293947, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.023809524660464376, | |
| "reward_std": 0.021662155631929636, | |
| "rewards/accuracy_reward": 0.023809524660464376, | |
| "step": 105 | |
| }, | |
| { | |
| "completion_length": 622.7715854644775, | |
| "epoch": 0.5934219734079776, | |
| "grad_norm": 0.003517127363011241, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.04836309718666598, | |
| "reward_std": 0.03329852968454361, | |
| "rewards/accuracy_reward": 0.04836309718666598, | |
| "step": 106 | |
| }, | |
| { | |
| "completion_length": 647.0959930419922, | |
| "epoch": 0.5990202939118264, | |
| "grad_norm": 0.0036508163902908564, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.061755954287946224, | |
| "reward_std": 0.03932873113080859, | |
| "rewards/accuracy_reward": 0.061755954287946224, | |
| "step": 107 | |
| }, | |
| { | |
| "completion_length": 623.8727836608887, | |
| "epoch": 0.6046186144156753, | |
| "grad_norm": 0.0034824141766875982, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.05282738321693614, | |
| "reward_std": 0.023543079383671284, | |
| "rewards/accuracy_reward": 0.05282738321693614, | |
| "step": 108 | |
| }, | |
| { | |
| "completion_length": 646.765645980835, | |
| "epoch": 0.6102169349195241, | |
| "grad_norm": 0.0026973283383995295, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.017113095498643816, | |
| "reward_std": 0.017985261976718903, | |
| "rewards/accuracy_reward": 0.017113095498643816, | |
| "step": 109 | |
| }, | |
| { | |
| "completion_length": 625.6376571655273, | |
| "epoch": 0.615815255423373, | |
| "grad_norm": 0.0037007054779678583, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.06547619227785617, | |
| "reward_std": 0.040159355383366346, | |
| "rewards/accuracy_reward": 0.06547619227785617, | |
| "step": 110 | |
| }, | |
| { | |
| "completion_length": 657.3631019592285, | |
| "epoch": 0.6214135759272218, | |
| "grad_norm": 0.0027858198154717684, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.027529762883204967, | |
| "reward_std": 0.016143894754350185, | |
| "rewards/accuracy_reward": 0.027529762883204967, | |
| "step": 111 | |
| }, | |
| { | |
| "completion_length": 627.3422756195068, | |
| "epoch": 0.6270118964310707, | |
| "grad_norm": 0.0035545658320188522, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.05059523967793211, | |
| "reward_std": 0.03195094550028443, | |
| "rewards/accuracy_reward": 0.05059523967793211, | |
| "step": 112 | |
| }, | |
| { | |
| "completion_length": 629.8177185058594, | |
| "epoch": 0.6326102169349195, | |
| "grad_norm": 0.0040982505306601524, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.05357142956927419, | |
| "reward_std": 0.03337568882852793, | |
| "rewards/accuracy_reward": 0.05357142956927419, | |
| "step": 113 | |
| }, | |
| { | |
| "completion_length": 650.6034278869629, | |
| "epoch": 0.6382085374387684, | |
| "grad_norm": 0.0028056029696017504, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.05208333529299125, | |
| "reward_std": 0.01970939477905631, | |
| "rewards/accuracy_reward": 0.05208333529299125, | |
| "step": 114 | |
| }, | |
| { | |
| "completion_length": 649.5796318054199, | |
| "epoch": 0.6438068579426172, | |
| "grad_norm": 0.0023333376739174128, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.014880952483508736, | |
| "reward_std": 0.010090996511280537, | |
| "rewards/accuracy_reward": 0.014880952483508736, | |
| "step": 115 | |
| }, | |
| { | |
| "completion_length": 645.5312652587891, | |
| "epoch": 0.6494051784464661, | |
| "grad_norm": 0.0031304731965065002, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.022321428870782256, | |
| "reward_std": 0.015947763342410326, | |
| "rewards/accuracy_reward": 0.022321428870782256, | |
| "step": 116 | |
| }, | |
| { | |
| "completion_length": 606.377986907959, | |
| "epoch": 0.655003498950315, | |
| "grad_norm": 0.004076777026057243, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.030505952949170023, | |
| "reward_std": 0.028484483249485493, | |
| "rewards/accuracy_reward": 0.030505952949170023, | |
| "step": 117 | |
| }, | |
| { | |
| "completion_length": 633.9389953613281, | |
| "epoch": 0.6606018194541637, | |
| "grad_norm": 0.0019400623859837651, | |
| "learning_rate": 3e-07, | |
| "loss": -0.0, | |
| "reward": 0.01934523874660954, | |
| "reward_std": 0.014487121719866991, | |
| "rewards/accuracy_reward": 0.01934523874660954, | |
| "step": 118 | |
| }, | |
| { | |
| "completion_length": 626.8236713409424, | |
| "epoch": 0.6662001399580126, | |
| "grad_norm": 0.0019256924279034138, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.040178572409786284, | |
| "reward_std": 0.013790588825941086, | |
| "rewards/accuracy_reward": 0.040178572409786284, | |
| "step": 119 | |
| }, | |
| { | |
| "completion_length": 633.7790336608887, | |
| "epoch": 0.6717984604618614, | |
| "grad_norm": 0.0031517883762717247, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.04092261998448521, | |
| "reward_std": 0.022219491191208363, | |
| "rewards/accuracy_reward": 0.04092261998448521, | |
| "step": 120 | |
| }, | |
| { | |
| "completion_length": 648.1837844848633, | |
| "epoch": 0.6773967809657103, | |
| "grad_norm": 0.0026536902878433466, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.032738096197135746, | |
| "reward_std": 0.02146284654736519, | |
| "rewards/accuracy_reward": 0.032738096197135746, | |
| "step": 121 | |
| }, | |
| { | |
| "completion_length": 652.9933204650879, | |
| "epoch": 0.6829951014695591, | |
| "grad_norm": 0.0033722377847880125, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.023809524544049054, | |
| "reward_std": 0.020562718622386456, | |
| "rewards/accuracy_reward": 0.023809524544049054, | |
| "step": 122 | |
| }, | |
| { | |
| "completion_length": 656.3534355163574, | |
| "epoch": 0.688593421973408, | |
| "grad_norm": 0.002776005771011114, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.034970239794347435, | |
| "reward_std": 0.01705795805901289, | |
| "rewards/accuracy_reward": 0.034970239794347435, | |
| "step": 123 | |
| }, | |
| { | |
| "completion_length": 610.5788822174072, | |
| "epoch": 0.6941917424772568, | |
| "grad_norm": 0.0037796536926180124, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.07142857392318547, | |
| "reward_std": 0.03704408532939851, | |
| "rewards/accuracy_reward": 0.07142857392318547, | |
| "step": 124 | |
| }, | |
| { | |
| "completion_length": 639.4814128875732, | |
| "epoch": 0.6997900629811057, | |
| "grad_norm": 0.003291892819106579, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.05133928789291531, | |
| "reward_std": 0.033701435662806034, | |
| "rewards/accuracy_reward": 0.05133928789291531, | |
| "step": 125 | |
| }, | |
| { | |
| "completion_length": 627.3214416503906, | |
| "epoch": 0.7053883834849545, | |
| "grad_norm": 0.002914518816396594, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.0386904776096344, | |
| "reward_std": 0.02166215470060706, | |
| "rewards/accuracy_reward": 0.0386904776096344, | |
| "step": 126 | |
| }, | |
| { | |
| "completion_length": 614.4308052062988, | |
| "epoch": 0.7109867039888034, | |
| "grad_norm": 0.003878154093399644, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.03348214388824999, | |
| "reward_std": 0.018164014909416437, | |
| "rewards/accuracy_reward": 0.03348214388824999, | |
| "step": 127 | |
| }, | |
| { | |
| "completion_length": 637.3244171142578, | |
| "epoch": 0.7165850244926522, | |
| "grad_norm": 0.002592942677438259, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.026785714901052415, | |
| "reward_std": 0.01852204231545329, | |
| "rewards/accuracy_reward": 0.026785714901052415, | |
| "step": 128 | |
| }, | |
| { | |
| "completion_length": 662.0282936096191, | |
| "epoch": 0.722183344996501, | |
| "grad_norm": 0.0030688210390508175, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.0424107150756754, | |
| "reward_std": 0.029373969649896026, | |
| "rewards/accuracy_reward": 0.0424107150756754, | |
| "step": 129 | |
| }, | |
| { | |
| "completion_length": 609.4747123718262, | |
| "epoch": 0.72778166550035, | |
| "grad_norm": 0.0026889187283813953, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.03869047720218077, | |
| "reward_std": 0.021642634645104408, | |
| "rewards/accuracy_reward": 0.03869047720218077, | |
| "step": 130 | |
| }, | |
| { | |
| "completion_length": 628.831859588623, | |
| "epoch": 0.7333799860041987, | |
| "grad_norm": 0.0032478254288434982, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.019345238571986556, | |
| "reward_std": 0.01975191291421652, | |
| "rewards/accuracy_reward": 0.019345238571986556, | |
| "step": 131 | |
| }, | |
| { | |
| "completion_length": 642.3846893310547, | |
| "epoch": 0.7389783065080476, | |
| "grad_norm": 0.0019529862329363823, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.01116071455180645, | |
| "reward_std": 0.013233252801001072, | |
| "rewards/accuracy_reward": 0.01116071455180645, | |
| "step": 132 | |
| }, | |
| { | |
| "completion_length": 634.9181728363037, | |
| "epoch": 0.7445766270118964, | |
| "grad_norm": 0.004057868849486113, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.031250000873114914, | |
| "reward_std": 0.040879431180655956, | |
| "rewards/accuracy_reward": 0.031250000873114914, | |
| "step": 133 | |
| }, | |
| { | |
| "completion_length": 609.6331977844238, | |
| "epoch": 0.7501749475157453, | |
| "grad_norm": 0.003426521783694625, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.038690477260388434, | |
| "reward_std": 0.029601082671433687, | |
| "rewards/accuracy_reward": 0.038690477260388434, | |
| "step": 134 | |
| }, | |
| { | |
| "completion_length": 639.4613227844238, | |
| "epoch": 0.7557732680195941, | |
| "grad_norm": 0.003970231860876083, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.054315477260388434, | |
| "reward_std": 0.029604259878396988, | |
| "rewards/accuracy_reward": 0.054315477260388434, | |
| "step": 135 | |
| }, | |
| { | |
| "completion_length": 614.8199501037598, | |
| "epoch": 0.761371588523443, | |
| "grad_norm": 0.0031252307817339897, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.05654762062476948, | |
| "reward_std": 0.02807308081537485, | |
| "rewards/accuracy_reward": 0.05654762062476948, | |
| "step": 136 | |
| }, | |
| { | |
| "completion_length": 640.7924194335938, | |
| "epoch": 0.7669699090272918, | |
| "grad_norm": 0.0020789685659110546, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.019345238397363573, | |
| "reward_std": 0.015046600718051195, | |
| "rewards/accuracy_reward": 0.019345238397363573, | |
| "step": 137 | |
| }, | |
| { | |
| "completion_length": 640.4226322174072, | |
| "epoch": 0.7725682295311407, | |
| "grad_norm": 0.0019399580778554082, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.012648809934034944, | |
| "reward_std": 0.013656496535986662, | |
| "rewards/accuracy_reward": 0.012648809934034944, | |
| "step": 138 | |
| }, | |
| { | |
| "completion_length": 594.6205444335938, | |
| "epoch": 0.7781665500349895, | |
| "grad_norm": 0.0063768248073756695, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.052083334885537624, | |
| "reward_std": 0.023117476608604193, | |
| "rewards/accuracy_reward": 0.052083334885537624, | |
| "step": 139 | |
| }, | |
| { | |
| "completion_length": 613.655517578125, | |
| "epoch": 0.7837648705388384, | |
| "grad_norm": 0.0031261774711310863, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.039434525300748646, | |
| "reward_std": 0.020764170680195093, | |
| "rewards/accuracy_reward": 0.039434525300748646, | |
| "step": 140 | |
| }, | |
| { | |
| "completion_length": 645.6837902069092, | |
| "epoch": 0.7893631910426872, | |
| "grad_norm": 0.0014134430093690753, | |
| "learning_rate": 3e-07, | |
| "loss": -0.0, | |
| "reward": 0.02157738187815994, | |
| "reward_std": 0.007513539865612984, | |
| "rewards/accuracy_reward": 0.02157738187815994, | |
| "step": 141 | |
| }, | |
| { | |
| "completion_length": 630.043176651001, | |
| "epoch": 0.794961511546536, | |
| "grad_norm": 0.004378010053187609, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.04241071571595967, | |
| "reward_std": 0.03774271160364151, | |
| "rewards/accuracy_reward": 0.04241071571595967, | |
| "step": 142 | |
| }, | |
| { | |
| "completion_length": 622.1555137634277, | |
| "epoch": 0.8005598320503848, | |
| "grad_norm": 0.0029741383623331785, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.00892857153667137, | |
| "reward_std": 0.016726072411984205, | |
| "rewards/accuracy_reward": 0.00892857153667137, | |
| "step": 143 | |
| }, | |
| { | |
| "completion_length": 646.7626647949219, | |
| "epoch": 0.8061581525542337, | |
| "grad_norm": 0.0028938716277480125, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.021577381528913975, | |
| "reward_std": 0.022778970655053854, | |
| "rewards/accuracy_reward": 0.021577381528913975, | |
| "step": 144 | |
| }, | |
| { | |
| "completion_length": 651.0178699493408, | |
| "epoch": 0.8117564730580826, | |
| "grad_norm": 0.003098010318353772, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.023809524718672037, | |
| "reward_std": 0.00955103849992156, | |
| "rewards/accuracy_reward": 0.023809524718672037, | |
| "step": 145 | |
| }, | |
| { | |
| "completion_length": 655.3891487121582, | |
| "epoch": 0.8173547935619314, | |
| "grad_norm": 0.002411956200376153, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.028273810166865587, | |
| "reward_std": 0.024912146851420403, | |
| "rewards/accuracy_reward": 0.028273810166865587, | |
| "step": 146 | |
| }, | |
| { | |
| "completion_length": 660.3393001556396, | |
| "epoch": 0.8229531140657803, | |
| "grad_norm": 0.0024566147476434708, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.03199404838960618, | |
| "reward_std": 0.021460703574121, | |
| "rewards/accuracy_reward": 0.03199404838960618, | |
| "step": 147 | |
| }, | |
| { | |
| "completion_length": 660.4471855163574, | |
| "epoch": 0.8285514345696291, | |
| "grad_norm": 0.0033500257413834333, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.03199404838960618, | |
| "reward_std": 0.026453721337020397, | |
| "rewards/accuracy_reward": 0.03199404838960618, | |
| "step": 148 | |
| }, | |
| { | |
| "completion_length": 616.9553737640381, | |
| "epoch": 0.834149755073478, | |
| "grad_norm": 0.0035977945663034916, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.04613095358945429, | |
| "reward_std": 0.03803316270932555, | |
| "rewards/accuracy_reward": 0.04613095358945429, | |
| "step": 149 | |
| }, | |
| { | |
| "completion_length": 624.0811023712158, | |
| "epoch": 0.8397480755773268, | |
| "grad_norm": 0.0024934441316872835, | |
| "learning_rate": 3e-07, | |
| "loss": -0.0, | |
| "reward": 0.020089286321308464, | |
| "reward_std": 0.014129094779491425, | |
| "rewards/accuracy_reward": 0.020089286321308464, | |
| "step": 150 | |
| }, | |
| { | |
| "completion_length": 649.3125076293945, | |
| "epoch": 0.8453463960811757, | |
| "grad_norm": 0.0040410468354821205, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.08035714528523386, | |
| "reward_std": 0.037069959565997124, | |
| "rewards/accuracy_reward": 0.08035714528523386, | |
| "step": 151 | |
| }, | |
| { | |
| "completion_length": 655.87575340271, | |
| "epoch": 0.8509447165850245, | |
| "grad_norm": 0.0036378325894474983, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.020089286554139107, | |
| "reward_std": 0.025742772268131375, | |
| "rewards/accuracy_reward": 0.020089286554139107, | |
| "step": 152 | |
| }, | |
| { | |
| "completion_length": 608.526050567627, | |
| "epoch": 0.8565430370888734, | |
| "grad_norm": 0.0019003109773620963, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.019345238339155912, | |
| "reward_std": 0.013455044478178024, | |
| "rewards/accuracy_reward": 0.019345238339155912, | |
| "step": 153 | |
| }, | |
| { | |
| "completion_length": 605.8340892791748, | |
| "epoch": 0.8621413575927221, | |
| "grad_norm": 0.002737295813858509, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.04538690613117069, | |
| "reward_std": 0.025694933719933033, | |
| "rewards/accuracy_reward": 0.04538690613117069, | |
| "step": 154 | |
| }, | |
| { | |
| "completion_length": 649.8913841247559, | |
| "epoch": 0.867739678096571, | |
| "grad_norm": 0.0013863355852663517, | |
| "learning_rate": 3e-07, | |
| "loss": -0.0, | |
| "reward": 0.012648810050450265, | |
| "reward_std": 0.007513539865612984, | |
| "rewards/accuracy_reward": 0.012648810050450265, | |
| "step": 155 | |
| }, | |
| { | |
| "completion_length": 649.567720413208, | |
| "epoch": 0.8733379986004198, | |
| "grad_norm": 0.003731328761205077, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.020089286379516125, | |
| "reward_std": 0.022558840923011303, | |
| "rewards/accuracy_reward": 0.020089286379516125, | |
| "step": 156 | |
| }, | |
| { | |
| "completion_length": 661.0640029907227, | |
| "epoch": 0.8789363191042687, | |
| "grad_norm": 0.0034664925187826157, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.05505952559178695, | |
| "reward_std": 0.035452743992209435, | |
| "rewards/accuracy_reward": 0.05505952559178695, | |
| "step": 157 | |
| }, | |
| { | |
| "completion_length": 616.1681728363037, | |
| "epoch": 0.8845346396081175, | |
| "grad_norm": 0.003077705856412649, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.020833333721384406, | |
| "reward_std": 0.023389466106891632, | |
| "rewards/accuracy_reward": 0.020833333721384406, | |
| "step": 158 | |
| }, | |
| { | |
| "completion_length": 645.6845321655273, | |
| "epoch": 0.8901329601119664, | |
| "grad_norm": 0.0034521420020610094, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.02380952425301075, | |
| "reward_std": 0.01677391119301319, | |
| "rewards/accuracy_reward": 0.02380952425301075, | |
| "step": 159 | |
| }, | |
| { | |
| "completion_length": 621.7961502075195, | |
| "epoch": 0.8957312806158153, | |
| "grad_norm": 0.0036637301091104746, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.04092262004269287, | |
| "reward_std": 0.02751574432477355, | |
| "rewards/accuracy_reward": 0.04092262004269287, | |
| "step": 160 | |
| }, | |
| { | |
| "completion_length": 661.4003067016602, | |
| "epoch": 0.9013296011196641, | |
| "grad_norm": 0.0030420024413615465, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.031250000989530236, | |
| "reward_std": 0.022196792997419834, | |
| "rewards/accuracy_reward": 0.031250000989530236, | |
| "step": 161 | |
| }, | |
| { | |
| "completion_length": 636.8690567016602, | |
| "epoch": 0.906927921623513, | |
| "grad_norm": 0.003414291888475418, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.05357143084984273, | |
| "reward_std": 0.021321506705135107, | |
| "rewards/accuracy_reward": 0.05357143084984273, | |
| "step": 162 | |
| }, | |
| { | |
| "completion_length": 641.8973331451416, | |
| "epoch": 0.9125262421273618, | |
| "grad_norm": 0.0033480250276625156, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.03720238176174462, | |
| "reward_std": 0.029198394622653723, | |
| "rewards/accuracy_reward": 0.03720238176174462, | |
| "step": 163 | |
| }, | |
| { | |
| "completion_length": 623.6264972686768, | |
| "epoch": 0.9181245626312107, | |
| "grad_norm": 0.003853818401694298, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.05729166802484542, | |
| "reward_std": 0.037587220780551434, | |
| "rewards/accuracy_reward": 0.05729166802484542, | |
| "step": 164 | |
| }, | |
| { | |
| "completion_length": 649.7075996398926, | |
| "epoch": 0.9237228831350595, | |
| "grad_norm": 0.0024723373353481293, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.019345238513778895, | |
| "reward_std": 0.01928615104407072, | |
| "rewards/accuracy_reward": 0.019345238513778895, | |
| "step": 165 | |
| }, | |
| { | |
| "completion_length": 680.0818481445312, | |
| "epoch": 0.9293212036389084, | |
| "grad_norm": 0.00249727675691247, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.03869047743501142, | |
| "reward_std": 0.015805388800799847, | |
| "rewards/accuracy_reward": 0.03869047743501142, | |
| "step": 166 | |
| }, | |
| { | |
| "completion_length": 606.4137096405029, | |
| "epoch": 0.9349195241427571, | |
| "grad_norm": 0.0031826056074351072, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.04166666814126074, | |
| "reward_std": 0.02047350350767374, | |
| "rewards/accuracy_reward": 0.04166666814126074, | |
| "step": 167 | |
| }, | |
| { | |
| "completion_length": 620.4821529388428, | |
| "epoch": 0.940517844646606, | |
| "grad_norm": 0.0025734296068549156, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.02380952431121841, | |
| "reward_std": 0.0181671935133636, | |
| "rewards/accuracy_reward": 0.02380952431121841, | |
| "step": 168 | |
| }, | |
| { | |
| "completion_length": 617.4836406707764, | |
| "epoch": 0.9461161651504548, | |
| "grad_norm": 0.0036193837877362967, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.03199404873885214, | |
| "reward_std": 0.018744049593806267, | |
| "rewards/accuracy_reward": 0.03199404873885214, | |
| "step": 169 | |
| }, | |
| { | |
| "completion_length": 630.1562633514404, | |
| "epoch": 0.9517144856543037, | |
| "grad_norm": 0.0025704463478177786, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.03497023897944018, | |
| "reward_std": 0.01300378143787384, | |
| "rewards/accuracy_reward": 0.03497023897944018, | |
| "step": 170 | |
| }, | |
| { | |
| "completion_length": 608.0171241760254, | |
| "epoch": 0.9573128061581525, | |
| "grad_norm": 0.003403712995350361, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.05357143055880442, | |
| "reward_std": 0.030789734097197652, | |
| "rewards/accuracy_reward": 0.05357143055880442, | |
| "step": 171 | |
| }, | |
| { | |
| "completion_length": 624.5669803619385, | |
| "epoch": 0.9629111266620014, | |
| "grad_norm": 0.00244723167270422, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.018601191113702953, | |
| "reward_std": 0.014688573777675629, | |
| "rewards/accuracy_reward": 0.018601191113702953, | |
| "step": 172 | |
| }, | |
| { | |
| "completion_length": 619.2753105163574, | |
| "epoch": 0.9685094471658502, | |
| "grad_norm": 0.003216799348592758, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.015625000465661287, | |
| "reward_std": 0.011909665539860725, | |
| "rewards/accuracy_reward": 0.015625000465661287, | |
| "step": 173 | |
| }, | |
| { | |
| "completion_length": 603.6629619598389, | |
| "epoch": 0.9741077676696991, | |
| "grad_norm": 0.004921557381749153, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1026785732829012, | |
| "reward_std": 0.035813949070870876, | |
| "rewards/accuracy_reward": 0.1026785732829012, | |
| "step": 174 | |
| }, | |
| { | |
| "completion_length": 619.9709911346436, | |
| "epoch": 0.979706088173548, | |
| "grad_norm": 0.0021367412991821766, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.03943452483508736, | |
| "reward_std": 0.017064577899873257, | |
| "rewards/accuracy_reward": 0.03943452483508736, | |
| "step": 175 | |
| }, | |
| { | |
| "completion_length": 661.3519515991211, | |
| "epoch": 0.9853044086773968, | |
| "grad_norm": 0.003553919028490782, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.03422619169577956, | |
| "reward_std": 0.02724563330411911, | |
| "rewards/accuracy_reward": 0.03422619169577956, | |
| "step": 176 | |
| }, | |
| { | |
| "completion_length": 638.8660793304443, | |
| "epoch": 0.9909027291812457, | |
| "grad_norm": 0.0027842505369335413, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.05803571594879031, | |
| "reward_std": 0.021928824484348297, | |
| "rewards/accuracy_reward": 0.05803571594879031, | |
| "step": 177 | |
| }, | |
| { | |
| "completion_length": 666.0945014953613, | |
| "epoch": 0.9965010496850945, | |
| "grad_norm": 0.004953757394105196, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.03571428661234677, | |
| "reward_std": 0.03300916403532028, | |
| "rewards/accuracy_reward": 0.03571428661234677, | |
| "step": 178 | |
| }, | |
| { | |
| "epoch": 0.9965010496850945, | |
| "step": 178, | |
| "total_flos": 0.0, | |
| "train_loss": 1.5679008615386834e-09, | |
| "train_runtime": 33567.1333, | |
| "train_samples_per_second": 0.596, | |
| "train_steps_per_second": 0.005 | |
| } | |
| ], | |
| "logging_steps": 1, | |
| "max_steps": 178, | |
| "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": 1, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |