| { | |
| "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": 621.5997104644775, | |
| "epoch": 0.005598320503848845, | |
| "grad_norm": 0.01685222238302231, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.21130952710518613, | |
| "reward_std": 0.1508616954088211, | |
| "rewards/accuracy_reward": 0.21130952710518613, | |
| "step": 1 | |
| }, | |
| { | |
| "completion_length": 593.0461368560791, | |
| "epoch": 0.01119664100769769, | |
| "grad_norm": 0.017173435539007187, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.21502976398915052, | |
| "reward_std": 0.1765604354441166, | |
| "rewards/accuracy_reward": 0.21502976398915052, | |
| "step": 2 | |
| }, | |
| { | |
| "completion_length": 625.5074501037598, | |
| "epoch": 0.016794961511546535, | |
| "grad_norm": 0.013760975562036037, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.15401786100119352, | |
| "reward_std": 0.13073960039764643, | |
| "rewards/accuracy_reward": 0.15401786100119352, | |
| "step": 3 | |
| }, | |
| { | |
| "completion_length": 647.694206237793, | |
| "epoch": 0.02239328201539538, | |
| "grad_norm": 0.015759378671646118, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1808035772992298, | |
| "reward_std": 0.1654246775433421, | |
| "rewards/accuracy_reward": 0.1808035772992298, | |
| "step": 4 | |
| }, | |
| { | |
| "completion_length": 629.4486751556396, | |
| "epoch": 0.02799160251924423, | |
| "grad_norm": 0.014322967268526554, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.23883929010480642, | |
| "reward_std": 0.15154671855270863, | |
| "rewards/accuracy_reward": 0.23883929010480642, | |
| "step": 5 | |
| }, | |
| { | |
| "completion_length": 604.1823024749756, | |
| "epoch": 0.03358992302309307, | |
| "grad_norm": 0.014635809697210789, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.2194940543267876, | |
| "reward_std": 0.1607616674154997, | |
| "rewards/accuracy_reward": 0.2194940543267876, | |
| "step": 6 | |
| }, | |
| { | |
| "completion_length": 659.3504657745361, | |
| "epoch": 0.03918824352694192, | |
| "grad_norm": 0.015097221359610558, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.16666666860692203, | |
| "reward_std": 0.13727260008454323, | |
| "rewards/accuracy_reward": 0.16666666860692203, | |
| "step": 7 | |
| }, | |
| { | |
| "completion_length": 587.1049137115479, | |
| "epoch": 0.04478656403079076, | |
| "grad_norm": 0.01467145700007677, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.16815476398915052, | |
| "reward_std": 0.15577294118702412, | |
| "rewards/accuracy_reward": 0.16815476398915052, | |
| "step": 8 | |
| }, | |
| { | |
| "completion_length": 675.1614761352539, | |
| "epoch": 0.05038488453463961, | |
| "grad_norm": 0.015891388058662415, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1755952414823696, | |
| "reward_std": 0.16377168311737478, | |
| "rewards/accuracy_reward": 0.1755952414823696, | |
| "step": 9 | |
| }, | |
| { | |
| "completion_length": 651.4523887634277, | |
| "epoch": 0.05598320503848846, | |
| "grad_norm": 0.014163345098495483, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.18080357555299997, | |
| "reward_std": 0.13789755944162607, | |
| "rewards/accuracy_reward": 0.18080357555299997, | |
| "step": 10 | |
| }, | |
| { | |
| "completion_length": 652.5982303619385, | |
| "epoch": 0.0615815255423373, | |
| "grad_norm": 0.013874271884560585, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.17187500558793545, | |
| "reward_std": 0.1334617892280221, | |
| "rewards/accuracy_reward": 0.17187500558793545, | |
| "step": 11 | |
| }, | |
| { | |
| "completion_length": 629.9330501556396, | |
| "epoch": 0.06717984604618614, | |
| "grad_norm": 0.015252678655087948, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.15997024276293814, | |
| "reward_std": 0.13702514953911304, | |
| "rewards/accuracy_reward": 0.15997024276293814, | |
| "step": 12 | |
| }, | |
| { | |
| "completion_length": 635.7953948974609, | |
| "epoch": 0.072778166550035, | |
| "grad_norm": 0.014308245852589607, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1398809556849301, | |
| "reward_std": 0.13629459426738322, | |
| "rewards/accuracy_reward": 0.1398809556849301, | |
| "step": 13 | |
| }, | |
| { | |
| "completion_length": 655.9814014434814, | |
| "epoch": 0.07837648705388384, | |
| "grad_norm": 0.014390479773283005, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.17485119425691664, | |
| "reward_std": 0.14875314245000482, | |
| "rewards/accuracy_reward": 0.17485119425691664, | |
| "step": 14 | |
| }, | |
| { | |
| "completion_length": 656.0022430419922, | |
| "epoch": 0.08397480755773268, | |
| "grad_norm": 0.017086399719119072, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.17633928690338507, | |
| "reward_std": 0.13736709533259273, | |
| "rewards/accuracy_reward": 0.17633928690338507, | |
| "step": 15 | |
| }, | |
| { | |
| "completion_length": 578.8586463928223, | |
| "epoch": 0.08957312806158152, | |
| "grad_norm": 0.015446292236447334, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.21205357555299997, | |
| "reward_std": 0.15961862821131945, | |
| "rewards/accuracy_reward": 0.21205357555299997, | |
| "step": 16 | |
| }, | |
| { | |
| "completion_length": 624.680814743042, | |
| "epoch": 0.09517144856543037, | |
| "grad_norm": 0.016500283032655716, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.18601190892513841, | |
| "reward_std": 0.14123531756922603, | |
| "rewards/accuracy_reward": 0.18601190892513841, | |
| "step": 17 | |
| }, | |
| { | |
| "completion_length": 661.7500152587891, | |
| "epoch": 0.10076976906927922, | |
| "grad_norm": 0.014794868417084217, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.18229167046956718, | |
| "reward_std": 0.15161667275242507, | |
| "rewards/accuracy_reward": 0.18229167046956718, | |
| "step": 18 | |
| }, | |
| { | |
| "completion_length": 632.3244209289551, | |
| "epoch": 0.10636808957312806, | |
| "grad_norm": 0.01493909489363432, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1636904807528481, | |
| "reward_std": 0.1566695892252028, | |
| "rewards/accuracy_reward": 0.1636904807528481, | |
| "step": 19 | |
| }, | |
| { | |
| "completion_length": 627.7477836608887, | |
| "epoch": 0.11196641007697691, | |
| "grad_norm": 0.01570167765021324, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.2046130986418575, | |
| "reward_std": 0.14942223858088255, | |
| "rewards/accuracy_reward": 0.2046130986418575, | |
| "step": 20 | |
| }, | |
| { | |
| "completion_length": 625.4613227844238, | |
| "epoch": 0.11756473058082575, | |
| "grad_norm": 0.016287673264741898, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.22991071827709675, | |
| "reward_std": 0.1482053459621966, | |
| "rewards/accuracy_reward": 0.22991071827709675, | |
| "step": 21 | |
| }, | |
| { | |
| "completion_length": 695.4895973205566, | |
| "epoch": 0.1231630510846746, | |
| "grad_norm": 0.013338581658899784, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.129464287601877, | |
| "reward_std": 0.13608421152457595, | |
| "rewards/accuracy_reward": 0.129464287601877, | |
| "step": 22 | |
| }, | |
| { | |
| "completion_length": 666.1235198974609, | |
| "epoch": 0.12876137158852344, | |
| "grad_norm": 0.015841346234083176, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.20386905269697309, | |
| "reward_std": 0.14255746873095632, | |
| "rewards/accuracy_reward": 0.20386905269697309, | |
| "step": 23 | |
| }, | |
| { | |
| "completion_length": 649.9732265472412, | |
| "epoch": 0.13435969209237228, | |
| "grad_norm": 0.016053833067417145, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.17038690764456987, | |
| "reward_std": 0.12253910815343261, | |
| "rewards/accuracy_reward": 0.17038690764456987, | |
| "step": 24 | |
| }, | |
| { | |
| "completion_length": 613.6108741760254, | |
| "epoch": 0.13995801259622112, | |
| "grad_norm": 0.020091773942112923, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.22470238478854299, | |
| "reward_std": 0.13420080952346325, | |
| "rewards/accuracy_reward": 0.22470238478854299, | |
| "step": 25 | |
| }, | |
| { | |
| "completion_length": 629.3846836090088, | |
| "epoch": 0.14555633310007, | |
| "grad_norm": 0.016363132745027542, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.19642857508733869, | |
| "reward_std": 0.14178543630987406, | |
| "rewards/accuracy_reward": 0.19642857508733869, | |
| "step": 26 | |
| }, | |
| { | |
| "completion_length": 641.9903411865234, | |
| "epoch": 0.15115465360391883, | |
| "grad_norm": 0.016413498669862747, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.19642857648432255, | |
| "reward_std": 0.16177992545999587, | |
| "rewards/accuracy_reward": 0.19642857648432255, | |
| "step": 27 | |
| }, | |
| { | |
| "completion_length": 647.2522392272949, | |
| "epoch": 0.15675297410776767, | |
| "grad_norm": 0.015543129295110703, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.19345238601090387, | |
| "reward_std": 0.13141631055623293, | |
| "rewards/accuracy_reward": 0.19345238601090387, | |
| "step": 28 | |
| }, | |
| { | |
| "completion_length": 635.9523849487305, | |
| "epoch": 0.16235129461161651, | |
| "grad_norm": 0.02027256041765213, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.22023809660458937, | |
| "reward_std": 0.12436413252726197, | |
| "rewards/accuracy_reward": 0.22023809660458937, | |
| "step": 29 | |
| }, | |
| { | |
| "completion_length": 658.7291736602783, | |
| "epoch": 0.16794961511546536, | |
| "grad_norm": 0.019577788189053535, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.19642857584403828, | |
| "reward_std": 0.161903060041368, | |
| "rewards/accuracy_reward": 0.19642857584403828, | |
| "step": 30 | |
| }, | |
| { | |
| "completion_length": 666.1145973205566, | |
| "epoch": 0.1735479356193142, | |
| "grad_norm": 0.016409732401371002, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.19494048005435616, | |
| "reward_std": 0.12808324676007032, | |
| "rewards/accuracy_reward": 0.19494048005435616, | |
| "step": 31 | |
| }, | |
| { | |
| "completion_length": 662.9114723205566, | |
| "epoch": 0.17914625612316304, | |
| "grad_norm": 0.02575683780014515, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1733630993985571, | |
| "reward_std": 0.14043171424418688, | |
| "rewards/accuracy_reward": 0.1733630993985571, | |
| "step": 32 | |
| }, | |
| { | |
| "completion_length": 672.6763496398926, | |
| "epoch": 0.1847445766270119, | |
| "grad_norm": 0.02134682983160019, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1979166700039059, | |
| "reward_std": 0.1591111128218472, | |
| "rewards/accuracy_reward": 0.1979166700039059, | |
| "step": 33 | |
| }, | |
| { | |
| "completion_length": 640.2604217529297, | |
| "epoch": 0.19034289713086075, | |
| "grad_norm": 0.023335738107562065, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.22172619309276342, | |
| "reward_std": 0.16011474281549454, | |
| "rewards/accuracy_reward": 0.22172619309276342, | |
| "step": 34 | |
| }, | |
| { | |
| "completion_length": 679.5119152069092, | |
| "epoch": 0.1959412176347096, | |
| "grad_norm": 0.026544196531176567, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.21651785960420966, | |
| "reward_std": 0.1384496740065515, | |
| "rewards/accuracy_reward": 0.21651785960420966, | |
| "step": 35 | |
| }, | |
| { | |
| "completion_length": 644.0104331970215, | |
| "epoch": 0.20153953813855843, | |
| "grad_norm": 0.02789580263197422, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.23065476911142468, | |
| "reward_std": 0.14074094174429774, | |
| "rewards/accuracy_reward": 0.23065476911142468, | |
| "step": 36 | |
| }, | |
| { | |
| "completion_length": 686.8742713928223, | |
| "epoch": 0.20713785864240727, | |
| "grad_norm": 0.02931089699268341, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.2165178587893024, | |
| "reward_std": 0.1497473274357617, | |
| "rewards/accuracy_reward": 0.2165178587893024, | |
| "step": 37 | |
| }, | |
| { | |
| "completion_length": 707.8519477844238, | |
| "epoch": 0.21273617914625612, | |
| "grad_norm": 0.02348598837852478, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.17857143143191934, | |
| "reward_std": 0.12875390285626054, | |
| "rewards/accuracy_reward": 0.17857143143191934, | |
| "step": 38 | |
| }, | |
| { | |
| "completion_length": 739.4196586608887, | |
| "epoch": 0.21833449965010496, | |
| "grad_norm": 0.025258367881178856, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.16294643242144957, | |
| "reward_std": 0.10602654609829187, | |
| "rewards/accuracy_reward": 0.16294643242144957, | |
| "step": 39 | |
| }, | |
| { | |
| "completion_length": 781.2157859802246, | |
| "epoch": 0.22393282015395383, | |
| "grad_norm": 0.02543455921113491, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.17038690770277753, | |
| "reward_std": 0.1598060461692512, | |
| "rewards/accuracy_reward": 0.17038690770277753, | |
| "step": 40 | |
| }, | |
| { | |
| "completion_length": 806.7745704650879, | |
| "epoch": 0.22953114065780267, | |
| "grad_norm": 0.02464163675904274, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1443452417734079, | |
| "reward_std": 0.13860655017197132, | |
| "rewards/accuracy_reward": 0.1443452417734079, | |
| "step": 41 | |
| }, | |
| { | |
| "completion_length": 797.2678718566895, | |
| "epoch": 0.2351294611616515, | |
| "grad_norm": 0.022209392860531807, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1465773843228817, | |
| "reward_std": 0.1135863265953958, | |
| "rewards/accuracy_reward": 0.1465773843228817, | |
| "step": 42 | |
| }, | |
| { | |
| "completion_length": 800.4032897949219, | |
| "epoch": 0.24072778166550035, | |
| "grad_norm": 0.027684131637215614, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1502976210322231, | |
| "reward_std": 0.14116185018792748, | |
| "rewards/accuracy_reward": 0.1502976210322231, | |
| "step": 43 | |
| }, | |
| { | |
| "completion_length": 802.7254638671875, | |
| "epoch": 0.2463261021693492, | |
| "grad_norm": 0.02655387856066227, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.16294643009314314, | |
| "reward_std": 0.11001868033781648, | |
| "rewards/accuracy_reward": 0.16294643009314314, | |
| "step": 44 | |
| }, | |
| { | |
| "completion_length": 825.2031326293945, | |
| "epoch": 0.25192442267319803, | |
| "grad_norm": 0.02530819922685623, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.16889881389215589, | |
| "reward_std": 0.12393559236079454, | |
| "rewards/accuracy_reward": 0.16889881389215589, | |
| "step": 45 | |
| }, | |
| { | |
| "completion_length": 809.6086502075195, | |
| "epoch": 0.2575227431770469, | |
| "grad_norm": 0.027012314647436142, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1339285749127157, | |
| "reward_std": 0.12476211739704013, | |
| "rewards/accuracy_reward": 0.1339285749127157, | |
| "step": 46 | |
| }, | |
| { | |
| "completion_length": 822.8742713928223, | |
| "epoch": 0.2631210636808957, | |
| "grad_norm": 0.027949512004852295, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1458333374466747, | |
| "reward_std": 0.12774629099294543, | |
| "rewards/accuracy_reward": 0.1458333374466747, | |
| "step": 47 | |
| }, | |
| { | |
| "completion_length": 830.1458473205566, | |
| "epoch": 0.26871938418474456, | |
| "grad_norm": 0.040896009653806686, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.09077381109818816, | |
| "reward_std": 0.09958610264584422, | |
| "rewards/accuracy_reward": 0.09077381109818816, | |
| "step": 48 | |
| }, | |
| { | |
| "completion_length": 858.4375228881836, | |
| "epoch": 0.2743177046885934, | |
| "grad_norm": 0.025837421417236328, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.12276786024449393, | |
| "reward_std": 0.10703055281192064, | |
| "rewards/accuracy_reward": 0.12276786024449393, | |
| "step": 49 | |
| }, | |
| { | |
| "completion_length": 837.8504638671875, | |
| "epoch": 0.27991602519244224, | |
| "grad_norm": 0.029402956366539, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.10639881080714986, | |
| "reward_std": 0.09327335562556982, | |
| "rewards/accuracy_reward": 0.10639881080714986, | |
| "step": 50 | |
| }, | |
| { | |
| "completion_length": 795.3244209289551, | |
| "epoch": 0.28551434569629114, | |
| "grad_norm": 0.035774268209934235, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1502976214978844, | |
| "reward_std": 0.15580066060647368, | |
| "rewards/accuracy_reward": 0.1502976214978844, | |
| "step": 51 | |
| }, | |
| { | |
| "completion_length": 804.2485198974609, | |
| "epoch": 0.29111266620014, | |
| "grad_norm": 0.028347313404083252, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.13764881319366395, | |
| "reward_std": 0.10927183693274856, | |
| "rewards/accuracy_reward": 0.13764881319366395, | |
| "step": 52 | |
| }, | |
| { | |
| "completion_length": 806.589298248291, | |
| "epoch": 0.2967109867039888, | |
| "grad_norm": 0.04255390539765358, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.16815476428018883, | |
| "reward_std": 0.11745391692966223, | |
| "rewards/accuracy_reward": 0.16815476428018883, | |
| "step": 53 | |
| }, | |
| { | |
| "completion_length": 827.7753105163574, | |
| "epoch": 0.30230930720783766, | |
| "grad_norm": 0.028443966060876846, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.13690476393094286, | |
| "reward_std": 0.09341412922367454, | |
| "rewards/accuracy_reward": 0.13690476393094286, | |
| "step": 54 | |
| }, | |
| { | |
| "completion_length": 793.2686157226562, | |
| "epoch": 0.3079076277116865, | |
| "grad_norm": 0.06588064879179001, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.14434524159878492, | |
| "reward_std": 0.14150456665083766, | |
| "rewards/accuracy_reward": 0.14434524159878492, | |
| "step": 55 | |
| }, | |
| { | |
| "completion_length": 808.3698043823242, | |
| "epoch": 0.31350594821553535, | |
| "grad_norm": 0.07028539478778839, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.15476190706249326, | |
| "reward_std": 0.08559148758649826, | |
| "rewards/accuracy_reward": 0.15476190706249326, | |
| "step": 56 | |
| }, | |
| { | |
| "completion_length": 750.5506172180176, | |
| "epoch": 0.3191042687193842, | |
| "grad_norm": 0.20404821634292603, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.25000000395812094, | |
| "reward_std": 0.15004885476082563, | |
| "rewards/accuracy_reward": 0.25000000395812094, | |
| "step": 57 | |
| }, | |
| { | |
| "completion_length": 802.5141525268555, | |
| "epoch": 0.32470258922323303, | |
| "grad_norm": 0.30963581800460815, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.15848214778816327, | |
| "reward_std": 0.13095595594495535, | |
| "rewards/accuracy_reward": 0.15848214778816327, | |
| "step": 58 | |
| }, | |
| { | |
| "completion_length": 817.5855865478516, | |
| "epoch": 0.33030090972708187, | |
| "grad_norm": 0.4757382571697235, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.08258928690338507, | |
| "reward_std": 0.09397146827541292, | |
| "rewards/accuracy_reward": 0.08258928690338507, | |
| "step": 59 | |
| }, | |
| { | |
| "completion_length": 780.9456977844238, | |
| "epoch": 0.3358992302309307, | |
| "grad_norm": 0.9979881644248962, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.14136905007762834, | |
| "reward_std": 0.10418214229866862, | |
| "rewards/accuracy_reward": 0.14136905007762834, | |
| "step": 60 | |
| }, | |
| { | |
| "completion_length": 806.7462959289551, | |
| "epoch": 0.34149755073477955, | |
| "grad_norm": 2.2850308418273926, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.16741071856813505, | |
| "reward_std": 0.12904326571151614, | |
| "rewards/accuracy_reward": 0.16741071856813505, | |
| "step": 61 | |
| }, | |
| { | |
| "completion_length": 796.9077606201172, | |
| "epoch": 0.3470958712386284, | |
| "grad_norm": 3.0867838859558105, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1555059568490833, | |
| "reward_std": 0.1193866366520524, | |
| "rewards/accuracy_reward": 0.1555059568490833, | |
| "step": 62 | |
| }, | |
| { | |
| "completion_length": 815.139892578125, | |
| "epoch": 0.35269419174247724, | |
| "grad_norm": 9.1638822555542, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.12425595556851476, | |
| "reward_std": 0.11769626522436738, | |
| "rewards/accuracy_reward": 0.12425595556851476, | |
| "step": 63 | |
| }, | |
| { | |
| "completion_length": 811.334831237793, | |
| "epoch": 0.3582925122463261, | |
| "grad_norm": 9.835746765136719, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.13467262114863843, | |
| "reward_std": 0.10124374250881374, | |
| "rewards/accuracy_reward": 0.13467262114863843, | |
| "step": 64 | |
| }, | |
| { | |
| "completion_length": 816.058048248291, | |
| "epoch": 0.363890832750175, | |
| "grad_norm": 12.197829246520996, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.13467262114863843, | |
| "reward_std": 0.11615834524855018, | |
| "rewards/accuracy_reward": 0.13467262114863843, | |
| "step": 65 | |
| }, | |
| { | |
| "completion_length": 778.9769439697266, | |
| "epoch": 0.3694891532540238, | |
| "grad_norm": 17.09342384338379, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1324404794140719, | |
| "reward_std": 0.1296369112096727, | |
| "rewards/accuracy_reward": 0.1324404794140719, | |
| "step": 66 | |
| }, | |
| { | |
| "completion_length": 791.2901954650879, | |
| "epoch": 0.37508747375787266, | |
| "grad_norm": 5.410295009613037, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.10342262109043077, | |
| "reward_std": 0.10896145971491933, | |
| "rewards/accuracy_reward": 0.10342262109043077, | |
| "step": 67 | |
| }, | |
| { | |
| "completion_length": 745.5297737121582, | |
| "epoch": 0.3806857942617215, | |
| "grad_norm": 13.46410846710205, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.21428571629803628, | |
| "reward_std": 0.14342379802837968, | |
| "rewards/accuracy_reward": 0.21428571629803628, | |
| "step": 68 | |
| }, | |
| { | |
| "completion_length": 776.0439109802246, | |
| "epoch": 0.38628411476557034, | |
| "grad_norm": 8.488734245300293, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.11830357467988506, | |
| "reward_std": 0.11542791035026312, | |
| "rewards/accuracy_reward": 0.11830357467988506, | |
| "step": 69 | |
| }, | |
| { | |
| "completion_length": 797.398078918457, | |
| "epoch": 0.3918824352694192, | |
| "grad_norm": 8.824621200561523, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.11383928917348385, | |
| "reward_std": 0.09247074788436294, | |
| "rewards/accuracy_reward": 0.11383928917348385, | |
| "step": 70 | |
| }, | |
| { | |
| "completion_length": 765.494800567627, | |
| "epoch": 0.397480755773268, | |
| "grad_norm": 7.611891746520996, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.15178571932483464, | |
| "reward_std": 0.10585443489253521, | |
| "rewards/accuracy_reward": 0.15178571932483464, | |
| "step": 71 | |
| }, | |
| { | |
| "completion_length": 778.1890029907227, | |
| "epoch": 0.40307907627711687, | |
| "grad_norm": 14.848072052001953, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.18452381214592606, | |
| "reward_std": 0.16349461674690247, | |
| "rewards/accuracy_reward": 0.18452381214592606, | |
| "step": 72 | |
| }, | |
| { | |
| "completion_length": 774.4456939697266, | |
| "epoch": 0.4086773967809657, | |
| "grad_norm": 5.173851490020752, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.20014881325187162, | |
| "reward_std": 0.12936217105016112, | |
| "rewards/accuracy_reward": 0.20014881325187162, | |
| "step": 73 | |
| }, | |
| { | |
| "completion_length": 841.1287307739258, | |
| "epoch": 0.41427571728481455, | |
| "grad_norm": 2.6559793949127197, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.13839285890571773, | |
| "reward_std": 0.13339613983407617, | |
| "rewards/accuracy_reward": 0.13839285890571773, | |
| "step": 74 | |
| }, | |
| { | |
| "completion_length": 818.7284393310547, | |
| "epoch": 0.4198740377886634, | |
| "grad_norm": 1.2601547241210938, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.15699405129998922, | |
| "reward_std": 0.147474380210042, | |
| "rewards/accuracy_reward": 0.15699405129998922, | |
| "step": 75 | |
| }, | |
| { | |
| "completion_length": 806.4762115478516, | |
| "epoch": 0.42547235829251223, | |
| "grad_norm": 0.6955339312553406, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.14434524072567, | |
| "reward_std": 0.11006892891600728, | |
| "rewards/accuracy_reward": 0.14434524072567, | |
| "step": 76 | |
| }, | |
| { | |
| "completion_length": 829.7209968566895, | |
| "epoch": 0.4310706787963611, | |
| "grad_norm": 0.3638227880001068, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.16294643277069554, | |
| "reward_std": 0.12348859198391438, | |
| "rewards/accuracy_reward": 0.16294643277069554, | |
| "step": 77 | |
| }, | |
| { | |
| "completion_length": 825.5610313415527, | |
| "epoch": 0.4366689993002099, | |
| "grad_norm": 0.26862311363220215, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1272321459837258, | |
| "reward_std": 0.10739197209477425, | |
| "rewards/accuracy_reward": 0.1272321459837258, | |
| "step": 78 | |
| }, | |
| { | |
| "completion_length": 789.9360198974609, | |
| "epoch": 0.44226731980405876, | |
| "grad_norm": 0.2580260634422302, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.16964286100119352, | |
| "reward_std": 0.12814528262242675, | |
| "rewards/accuracy_reward": 0.16964286100119352, | |
| "step": 79 | |
| }, | |
| { | |
| "completion_length": 812.0885543823242, | |
| "epoch": 0.44786564030790765, | |
| "grad_norm": 0.21123050153255463, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.15252976340707392, | |
| "reward_std": 0.11390745500102639, | |
| "rewards/accuracy_reward": 0.15252976340707392, | |
| "step": 80 | |
| }, | |
| { | |
| "completion_length": 827.6890068054199, | |
| "epoch": 0.4534639608117565, | |
| "grad_norm": 0.21955974400043488, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.13392857264261693, | |
| "reward_std": 0.11876308592036366, | |
| "rewards/accuracy_reward": 0.13392857264261693, | |
| "step": 81 | |
| }, | |
| { | |
| "completion_length": 790.8415336608887, | |
| "epoch": 0.45906228131560534, | |
| "grad_norm": 0.15459412336349487, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.21205357508733869, | |
| "reward_std": 0.1082809790968895, | |
| "rewards/accuracy_reward": 0.21205357508733869, | |
| "step": 82 | |
| }, | |
| { | |
| "completion_length": 799.1346817016602, | |
| "epoch": 0.4646606018194542, | |
| "grad_norm": 0.23524288833141327, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1569940506014973, | |
| "reward_std": 0.11971638561226428, | |
| "rewards/accuracy_reward": 0.1569940506014973, | |
| "step": 83 | |
| }, | |
| { | |
| "completion_length": 757.0647468566895, | |
| "epoch": 0.470258922323303, | |
| "grad_norm": 0.19839145243167877, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.2038690508925356, | |
| "reward_std": 0.13258925126865506, | |
| "rewards/accuracy_reward": 0.2038690508925356, | |
| "step": 84 | |
| }, | |
| { | |
| "completion_length": 808.5535926818848, | |
| "epoch": 0.47585724282715186, | |
| "grad_norm": 0.32531610131263733, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1666666684905067, | |
| "reward_std": 0.13159588165581226, | |
| "rewards/accuracy_reward": 0.1666666684905067, | |
| "step": 85 | |
| }, | |
| { | |
| "completion_length": 813.6451110839844, | |
| "epoch": 0.4814555633310007, | |
| "grad_norm": 0.18594929575920105, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.144345240900293, | |
| "reward_std": 0.1254395372234285, | |
| "rewards/accuracy_reward": 0.144345240900293, | |
| "step": 86 | |
| }, | |
| { | |
| "completion_length": 774.9620704650879, | |
| "epoch": 0.48705388383484954, | |
| "grad_norm": 0.18389521539211273, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.19270833674818277, | |
| "reward_std": 0.15590714011341333, | |
| "rewards/accuracy_reward": 0.19270833674818277, | |
| "step": 87 | |
| }, | |
| { | |
| "completion_length": 769.0573043823242, | |
| "epoch": 0.4926522043386984, | |
| "grad_norm": 0.2218749225139618, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.2105654807528481, | |
| "reward_std": 0.12664200831204653, | |
| "rewards/accuracy_reward": 0.2105654807528481, | |
| "step": 88 | |
| }, | |
| { | |
| "completion_length": 792.6644401550293, | |
| "epoch": 0.4982505248425472, | |
| "grad_norm": 1.4022585153579712, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.15476190886693075, | |
| "reward_std": 0.11095877178013325, | |
| "rewards/accuracy_reward": 0.15476190886693075, | |
| "step": 89 | |
| }, | |
| { | |
| "completion_length": 782.799861907959, | |
| "epoch": 0.5038488453463961, | |
| "grad_norm": 28.66908073425293, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.15104167041135952, | |
| "reward_std": 0.15992113715037704, | |
| "rewards/accuracy_reward": 0.15104167041135952, | |
| "step": 90 | |
| }, | |
| { | |
| "completion_length": 795.1547698974609, | |
| "epoch": 0.509447165850245, | |
| "grad_norm": 8.09515380859375, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1569940495537594, | |
| "reward_std": 0.12339623901061714, | |
| "rewards/accuracy_reward": 0.1569940495537594, | |
| "step": 91 | |
| }, | |
| { | |
| "completion_length": 823.6614723205566, | |
| "epoch": 0.5150454863540938, | |
| "grad_norm": 4.843145847320557, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1197916705859825, | |
| "reward_std": 0.11639799829572439, | |
| "rewards/accuracy_reward": 0.1197916705859825, | |
| "step": 92 | |
| }, | |
| { | |
| "completion_length": 814.1956977844238, | |
| "epoch": 0.5206438068579426, | |
| "grad_norm": 3.3891513347625732, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.12276785861467943, | |
| "reward_std": 0.11077928310260177, | |
| "rewards/accuracy_reward": 0.12276785861467943, | |
| "step": 93 | |
| }, | |
| { | |
| "completion_length": 832.5439147949219, | |
| "epoch": 0.5262421273617914, | |
| "grad_norm": 3.9651668071746826, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.12797619338380173, | |
| "reward_std": 0.13058934407308698, | |
| "rewards/accuracy_reward": 0.12797619338380173, | |
| "step": 94 | |
| }, | |
| { | |
| "completion_length": 845.0171318054199, | |
| "epoch": 0.5318404478656403, | |
| "grad_norm": 2.0770602226257324, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.11235119315097108, | |
| "reward_std": 0.08664647908881307, | |
| "rewards/accuracy_reward": 0.11235119315097108, | |
| "step": 95 | |
| }, | |
| { | |
| "completion_length": 790.6480865478516, | |
| "epoch": 0.5374387683694891, | |
| "grad_norm": 1.959823727607727, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.12351190787740052, | |
| "reward_std": 0.10759956203401089, | |
| "rewards/accuracy_reward": 0.12351190787740052, | |
| "step": 96 | |
| }, | |
| { | |
| "completion_length": 816.3586463928223, | |
| "epoch": 0.543037088873338, | |
| "grad_norm": 1.8126938343048096, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1688988117966801, | |
| "reward_std": 0.13670154195278883, | |
| "rewards/accuracy_reward": 0.1688988117966801, | |
| "step": 97 | |
| }, | |
| { | |
| "completion_length": 787.2901878356934, | |
| "epoch": 0.5486354093771868, | |
| "grad_norm": 1.0017977952957153, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.15104166907258332, | |
| "reward_std": 0.12999275140464306, | |
| "rewards/accuracy_reward": 0.15104166907258332, | |
| "step": 98 | |
| }, | |
| { | |
| "completion_length": 755.874267578125, | |
| "epoch": 0.5542337298810357, | |
| "grad_norm": 1.0542266368865967, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.24479167215758935, | |
| "reward_std": 0.15221792878583074, | |
| "rewards/accuracy_reward": 0.24479167215758935, | |
| "step": 99 | |
| }, | |
| { | |
| "completion_length": 791.4122161865234, | |
| "epoch": 0.5598320503848845, | |
| "grad_norm": 0.7178149223327637, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.17038690840126947, | |
| "reward_std": 0.13499723048880696, | |
| "rewards/accuracy_reward": 0.17038690840126947, | |
| "step": 100 | |
| }, | |
| { | |
| "completion_length": 744.4151840209961, | |
| "epoch": 0.5654303708887334, | |
| "grad_norm": 0.7150797843933105, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.20610119280172512, | |
| "reward_std": 0.1412361622788012, | |
| "rewards/accuracy_reward": 0.20610119280172512, | |
| "step": 101 | |
| }, | |
| { | |
| "completion_length": 789.238109588623, | |
| "epoch": 0.5710286913925823, | |
| "grad_norm": 0.9325894117355347, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.16443452634848654, | |
| "reward_std": 0.11632257583551109, | |
| "rewards/accuracy_reward": 0.16443452634848654, | |
| "step": 102 | |
| }, | |
| { | |
| "completion_length": 768.8869132995605, | |
| "epoch": 0.5766270118964311, | |
| "grad_norm": 7.234724998474121, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.125744050310459, | |
| "reward_std": 0.10205298336222768, | |
| "rewards/accuracy_reward": 0.125744050310459, | |
| "step": 103 | |
| }, | |
| { | |
| "completion_length": 751.9308242797852, | |
| "epoch": 0.58222533240028, | |
| "grad_norm": 3.879939079284668, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1755952414823696, | |
| "reward_std": 0.12294519320130348, | |
| "rewards/accuracy_reward": 0.1755952414823696, | |
| "step": 104 | |
| }, | |
| { | |
| "completion_length": 627.394359588623, | |
| "epoch": 0.5878236529041287, | |
| "grad_norm": 3.7597320079803467, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.2224702457897365, | |
| "reward_std": 0.15529246046207845, | |
| "rewards/accuracy_reward": 0.2224702457897365, | |
| "step": 105 | |
| }, | |
| { | |
| "completion_length": 617.4479293823242, | |
| "epoch": 0.5934219734079776, | |
| "grad_norm": 1.7537007331848145, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.2157738134264946, | |
| "reward_std": 0.14103552885353565, | |
| "rewards/accuracy_reward": 0.2157738134264946, | |
| "step": 106 | |
| }, | |
| { | |
| "completion_length": 654.3489608764648, | |
| "epoch": 0.5990202939118264, | |
| "grad_norm": 1.367697834968567, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.17485119379125535, | |
| "reward_std": 0.14419334661215544, | |
| "rewards/accuracy_reward": 0.17485119379125535, | |
| "step": 107 | |
| }, | |
| { | |
| "completion_length": 674.4724807739258, | |
| "epoch": 0.6046186144156753, | |
| "grad_norm": 0.4473717212677002, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.17857143364381045, | |
| "reward_std": 0.11048215441405773, | |
| "rewards/accuracy_reward": 0.17857143364381045, | |
| "step": 108 | |
| }, | |
| { | |
| "completion_length": 725.7738227844238, | |
| "epoch": 0.6102169349195241, | |
| "grad_norm": 3.6793627738952637, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.14806547824991867, | |
| "reward_std": 0.08863410213962197, | |
| "rewards/accuracy_reward": 0.14806547824991867, | |
| "step": 109 | |
| }, | |
| { | |
| "completion_length": 677.5327529907227, | |
| "epoch": 0.615815255423373, | |
| "grad_norm": 1.7502440214157104, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.24627976631745696, | |
| "reward_std": 0.16675466718152165, | |
| "rewards/accuracy_reward": 0.24627976631745696, | |
| "step": 110 | |
| }, | |
| { | |
| "completion_length": 738.4843883514404, | |
| "epoch": 0.6214135759272218, | |
| "grad_norm": 5.396337509155273, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.17485119419870898, | |
| "reward_std": 0.1276370887644589, | |
| "rewards/accuracy_reward": 0.17485119419870898, | |
| "step": 111 | |
| }, | |
| { | |
| "completion_length": 669.8638534545898, | |
| "epoch": 0.6270118964310707, | |
| "grad_norm": 13.67143726348877, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.21726190828485414, | |
| "reward_std": 0.13963406695984304, | |
| "rewards/accuracy_reward": 0.21726190828485414, | |
| "step": 112 | |
| }, | |
| { | |
| "completion_length": 665.2529907226562, | |
| "epoch": 0.6326102169349195, | |
| "grad_norm": 44.409034729003906, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.19940476591000333, | |
| "reward_std": 0.14031004393473268, | |
| "rewards/accuracy_reward": 0.19940476591000333, | |
| "step": 113 | |
| }, | |
| { | |
| "completion_length": 632.8326091766357, | |
| "epoch": 0.6382085374387684, | |
| "grad_norm": 10.954506874084473, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.2090773865347728, | |
| "reward_std": 0.13506855070590973, | |
| "rewards/accuracy_reward": 0.2090773865347728, | |
| "step": 114 | |
| }, | |
| { | |
| "completion_length": 684.7834930419922, | |
| "epoch": 0.6438068579426172, | |
| "grad_norm": 3.232715606689453, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.18601190974004567, | |
| "reward_std": 0.14560958743095398, | |
| "rewards/accuracy_reward": 0.18601190974004567, | |
| "step": 115 | |
| }, | |
| { | |
| "completion_length": 707.04465675354, | |
| "epoch": 0.6494051784464661, | |
| "grad_norm": 0.662769079208374, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.21875000512227416, | |
| "reward_std": 0.1376757239922881, | |
| "rewards/accuracy_reward": 0.21875000512227416, | |
| "step": 116 | |
| }, | |
| { | |
| "completion_length": 669.0446586608887, | |
| "epoch": 0.655003498950315, | |
| "grad_norm": 0.33869776129722595, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.21056548063643277, | |
| "reward_std": 0.14122196286916733, | |
| "rewards/accuracy_reward": 0.21056548063643277, | |
| "step": 117 | |
| }, | |
| { | |
| "completion_length": 684.3393020629883, | |
| "epoch": 0.6606018194541637, | |
| "grad_norm": 0.7248935103416443, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1614583363989368, | |
| "reward_std": 0.11945095984265208, | |
| "rewards/accuracy_reward": 0.1614583363989368, | |
| "step": 118 | |
| }, | |
| { | |
| "completion_length": 666.3139991760254, | |
| "epoch": 0.6662001399580126, | |
| "grad_norm": 43.89714431762695, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.22470238665118814, | |
| "reward_std": 0.10662061581388116, | |
| "rewards/accuracy_reward": 0.22470238665118814, | |
| "step": 119 | |
| }, | |
| { | |
| "completion_length": 679.056568145752, | |
| "epoch": 0.6717984604618614, | |
| "grad_norm": 8.356165885925293, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.21279762336052954, | |
| "reward_std": 0.124493746785447, | |
| "rewards/accuracy_reward": 0.21279762336052954, | |
| "step": 120 | |
| }, | |
| { | |
| "completion_length": 687.572925567627, | |
| "epoch": 0.6773967809657103, | |
| "grad_norm": 7.832132816314697, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.18750000436557457, | |
| "reward_std": 0.14207694586366415, | |
| "rewards/accuracy_reward": 0.18750000436557457, | |
| "step": 121 | |
| }, | |
| { | |
| "completion_length": 689.9590873718262, | |
| "epoch": 0.6829951014695591, | |
| "grad_norm": 2.498194932937622, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.15773809858364984, | |
| "reward_std": 0.10648334585130215, | |
| "rewards/accuracy_reward": 0.15773809858364984, | |
| "step": 122 | |
| }, | |
| { | |
| "completion_length": 704.2939109802246, | |
| "epoch": 0.688593421973408, | |
| "grad_norm": 6.899082183837891, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.18377976468764246, | |
| "reward_std": 0.1484173396602273, | |
| "rewards/accuracy_reward": 0.18377976468764246, | |
| "step": 123 | |
| }, | |
| { | |
| "completion_length": 651.5878105163574, | |
| "epoch": 0.6941917424772568, | |
| "grad_norm": 7.5064544677734375, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.2500000016298145, | |
| "reward_std": 0.16986942291259766, | |
| "rewards/accuracy_reward": 0.2500000016298145, | |
| "step": 124 | |
| }, | |
| { | |
| "completion_length": 695.3638496398926, | |
| "epoch": 0.6997900629811057, | |
| "grad_norm": 4.8156867027282715, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.20238095673266798, | |
| "reward_std": 0.13348530931398273, | |
| "rewards/accuracy_reward": 0.20238095673266798, | |
| "step": 125 | |
| }, | |
| { | |
| "completion_length": 702.0327548980713, | |
| "epoch": 0.7053883834849545, | |
| "grad_norm": 7.651693344116211, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.21875000500585884, | |
| "reward_std": 0.16874144971370697, | |
| "rewards/accuracy_reward": 0.21875000500585884, | |
| "step": 126 | |
| }, | |
| { | |
| "completion_length": 671.8259048461914, | |
| "epoch": 0.7109867039888034, | |
| "grad_norm": 6.942368507385254, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.24330357555299997, | |
| "reward_std": 0.1387807228602469, | |
| "rewards/accuracy_reward": 0.24330357555299997, | |
| "step": 127 | |
| }, | |
| { | |
| "completion_length": 694.0915260314941, | |
| "epoch": 0.7165850244926522, | |
| "grad_norm": 4.375088691711426, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.20833333634072915, | |
| "reward_std": 0.12121187802404165, | |
| "rewards/accuracy_reward": 0.20833333634072915, | |
| "step": 128 | |
| }, | |
| { | |
| "completion_length": 703.304328918457, | |
| "epoch": 0.722183344996501, | |
| "grad_norm": 3.8942065238952637, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1785714317811653, | |
| "reward_std": 0.1311384318396449, | |
| "rewards/accuracy_reward": 0.1785714317811653, | |
| "step": 129 | |
| }, | |
| { | |
| "completion_length": 681.0907897949219, | |
| "epoch": 0.72778166550035, | |
| "grad_norm": 2.7340543270111084, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1979166711680591, | |
| "reward_std": 0.12608638824895024, | |
| "rewards/accuracy_reward": 0.1979166711680591, | |
| "step": 130 | |
| }, | |
| { | |
| "completion_length": 695.2366180419922, | |
| "epoch": 0.7333799860041987, | |
| "grad_norm": 1.5587713718414307, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.15922619513003156, | |
| "reward_std": 0.11638409737497568, | |
| "rewards/accuracy_reward": 0.15922619513003156, | |
| "step": 131 | |
| }, | |
| { | |
| "completion_length": 709.5788822174072, | |
| "epoch": 0.7389783065080476, | |
| "grad_norm": 3.3193564414978027, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.159226194024086, | |
| "reward_std": 0.13807840831577778, | |
| "rewards/accuracy_reward": 0.159226194024086, | |
| "step": 132 | |
| }, | |
| { | |
| "completion_length": 698.4427185058594, | |
| "epoch": 0.7445766270118964, | |
| "grad_norm": 1.9349188804626465, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.19717262242920697, | |
| "reward_std": 0.12813474284484982, | |
| "rewards/accuracy_reward": 0.19717262242920697, | |
| "step": 133 | |
| }, | |
| { | |
| "completion_length": 688.2210006713867, | |
| "epoch": 0.7501749475157453, | |
| "grad_norm": 2.0050582885742188, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.20758929062867537, | |
| "reward_std": 0.1788321421481669, | |
| "rewards/accuracy_reward": 0.20758929062867537, | |
| "step": 134 | |
| }, | |
| { | |
| "completion_length": 696.2515087127686, | |
| "epoch": 0.7557732680195941, | |
| "grad_norm": 0.6444657444953918, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.19642857677536085, | |
| "reward_std": 0.12122054304927588, | |
| "rewards/accuracy_reward": 0.19642857677536085, | |
| "step": 135 | |
| }, | |
| { | |
| "completion_length": 694.3489761352539, | |
| "epoch": 0.761371588523443, | |
| "grad_norm": 0.5755271315574646, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.2113095311797224, | |
| "reward_std": 0.15177863789722323, | |
| "rewards/accuracy_reward": 0.2113095311797224, | |
| "step": 136 | |
| }, | |
| { | |
| "completion_length": 703.0320091247559, | |
| "epoch": 0.7669699090272918, | |
| "grad_norm": 0.382269024848938, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.19419643247965723, | |
| "reward_std": 0.11693665431812406, | |
| "rewards/accuracy_reward": 0.19419643247965723, | |
| "step": 137 | |
| }, | |
| { | |
| "completion_length": 705.675615310669, | |
| "epoch": 0.7725682295311407, | |
| "grad_norm": 0.23526860773563385, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.17187500471482053, | |
| "reward_std": 0.12707782164216042, | |
| "rewards/accuracy_reward": 0.17187500471482053, | |
| "step": 138 | |
| }, | |
| { | |
| "completion_length": 673.7269477844238, | |
| "epoch": 0.7781665500349895, | |
| "grad_norm": 0.22726382315158844, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.2581845277454704, | |
| "reward_std": 0.1626861086115241, | |
| "rewards/accuracy_reward": 0.2581845277454704, | |
| "step": 139 | |
| }, | |
| { | |
| "completion_length": 701.4620704650879, | |
| "epoch": 0.7837648705388384, | |
| "grad_norm": 0.19594760239124298, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.21726190811023116, | |
| "reward_std": 0.15826332941651344, | |
| "rewards/accuracy_reward": 0.21726190811023116, | |
| "step": 140 | |
| }, | |
| { | |
| "completion_length": 726.7105808258057, | |
| "epoch": 0.7893631910426872, | |
| "grad_norm": 0.21776525676250458, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.19122024084208533, | |
| "reward_std": 0.13169277971610427, | |
| "rewards/accuracy_reward": 0.19122024084208533, | |
| "step": 141 | |
| }, | |
| { | |
| "completion_length": 709.7366180419922, | |
| "epoch": 0.794961511546536, | |
| "grad_norm": 0.17518886923789978, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.19642857578583062, | |
| "reward_std": 0.13794927392154932, | |
| "rewards/accuracy_reward": 0.19642857578583062, | |
| "step": 142 | |
| }, | |
| { | |
| "completion_length": 700.7433128356934, | |
| "epoch": 0.8005598320503848, | |
| "grad_norm": 0.1962345391511917, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.2261904808692634, | |
| "reward_std": 0.17080192361027002, | |
| "rewards/accuracy_reward": 0.2261904808692634, | |
| "step": 143 | |
| }, | |
| { | |
| "completion_length": 749.479175567627, | |
| "epoch": 0.8061581525542337, | |
| "grad_norm": 0.09322026371955872, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.14434524165699258, | |
| "reward_std": 0.07725285878404975, | |
| "rewards/accuracy_reward": 0.14434524165699258, | |
| "step": 144 | |
| }, | |
| { | |
| "completion_length": 726.3467464447021, | |
| "epoch": 0.8117564730580826, | |
| "grad_norm": 0.1386028379201889, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.2031250053551048, | |
| "reward_std": 0.14093447849154472, | |
| "rewards/accuracy_reward": 0.2031250053551048, | |
| "step": 145 | |
| }, | |
| { | |
| "completion_length": 738.7120704650879, | |
| "epoch": 0.8173547935619314, | |
| "grad_norm": 0.12930598855018616, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.17857143195578828, | |
| "reward_std": 0.12077544536441565, | |
| "rewards/accuracy_reward": 0.17857143195578828, | |
| "step": 146 | |
| }, | |
| { | |
| "completion_length": 765.6473388671875, | |
| "epoch": 0.8229531140657803, | |
| "grad_norm": 0.13917317986488342, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.12425595498643816, | |
| "reward_std": 0.12115136627107859, | |
| "rewards/accuracy_reward": 0.12425595498643816, | |
| "step": 147 | |
| }, | |
| { | |
| "completion_length": 741.0476341247559, | |
| "epoch": 0.8285514345696291, | |
| "grad_norm": 0.5330355763435364, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.18080357613507658, | |
| "reward_std": 0.12820188701152802, | |
| "rewards/accuracy_reward": 0.18080357613507658, | |
| "step": 148 | |
| }, | |
| { | |
| "completion_length": 715.349723815918, | |
| "epoch": 0.834149755073478, | |
| "grad_norm": 0.1551419347524643, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.18452381377574056, | |
| "reward_std": 0.13737005554139614, | |
| "rewards/accuracy_reward": 0.18452381377574056, | |
| "step": 149 | |
| }, | |
| { | |
| "completion_length": 748.0156364440918, | |
| "epoch": 0.8397480755773268, | |
| "grad_norm": 0.17706456780433655, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.19047619315097108, | |
| "reward_std": 0.10885073570534587, | |
| "rewards/accuracy_reward": 0.19047619315097108, | |
| "step": 150 | |
| }, | |
| { | |
| "completion_length": 772.0535888671875, | |
| "epoch": 0.8453463960811757, | |
| "grad_norm": 0.3694287836551666, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.17336309852544218, | |
| "reward_std": 0.11179409665055573, | |
| "rewards/accuracy_reward": 0.17336309852544218, | |
| "step": 151 | |
| }, | |
| { | |
| "completion_length": 774.1912307739258, | |
| "epoch": 0.8509447165850245, | |
| "grad_norm": 0.2840059697628021, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.15252976468764246, | |
| "reward_std": 0.11831123428419232, | |
| "rewards/accuracy_reward": 0.15252976468764246, | |
| "step": 152 | |
| }, | |
| { | |
| "completion_length": 760.9918327331543, | |
| "epoch": 0.8565430370888734, | |
| "grad_norm": 0.2995321452617645, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1778273859526962, | |
| "reward_std": 0.15920937061309814, | |
| "rewards/accuracy_reward": 0.1778273859526962, | |
| "step": 153 | |
| }, | |
| { | |
| "completion_length": 763.8556671142578, | |
| "epoch": 0.8621413575927221, | |
| "grad_norm": 1.4188278913497925, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1696428614668548, | |
| "reward_std": 0.12167660798877478, | |
| "rewards/accuracy_reward": 0.1696428614668548, | |
| "step": 154 | |
| }, | |
| { | |
| "completion_length": 771.9032936096191, | |
| "epoch": 0.867739678096571, | |
| "grad_norm": 0.15566886961460114, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.20684524130774662, | |
| "reward_std": 0.13866402814164758, | |
| "rewards/accuracy_reward": 0.20684524130774662, | |
| "step": 155 | |
| }, | |
| { | |
| "completion_length": 797.9323081970215, | |
| "epoch": 0.8733379986004198, | |
| "grad_norm": 0.21956497430801392, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.11979166796663776, | |
| "reward_std": 0.08592891087755561, | |
| "rewards/accuracy_reward": 0.11979166796663776, | |
| "step": 156 | |
| }, | |
| { | |
| "completion_length": 803.9062576293945, | |
| "epoch": 0.8789363191042687, | |
| "grad_norm": 0.4193490743637085, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.14955357427243143, | |
| "reward_std": 0.13559656590223312, | |
| "rewards/accuracy_reward": 0.14955357427243143, | |
| "step": 157 | |
| }, | |
| { | |
| "completion_length": 734.5416793823242, | |
| "epoch": 0.8845346396081175, | |
| "grad_norm": 0.32442227005958557, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.20312500384170562, | |
| "reward_std": 0.13238779548555613, | |
| "rewards/accuracy_reward": 0.20312500384170562, | |
| "step": 158 | |
| }, | |
| { | |
| "completion_length": 780.0461502075195, | |
| "epoch": 0.8901329601119664, | |
| "grad_norm": 0.2237851321697235, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.15773809980601072, | |
| "reward_std": 0.11771008232608438, | |
| "rewards/accuracy_reward": 0.15773809980601072, | |
| "step": 159 | |
| }, | |
| { | |
| "completion_length": 725.6488285064697, | |
| "epoch": 0.8957312806158153, | |
| "grad_norm": 0.20033074915409088, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.19122024055104703, | |
| "reward_std": 0.13583081704564393, | |
| "rewards/accuracy_reward": 0.19122024055104703, | |
| "step": 160 | |
| }, | |
| { | |
| "completion_length": 790.0461349487305, | |
| "epoch": 0.9013296011196641, | |
| "grad_norm": 0.2317737340927124, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.19345238577807322, | |
| "reward_std": 0.17005419172346592, | |
| "rewards/accuracy_reward": 0.19345238577807322, | |
| "step": 161 | |
| }, | |
| { | |
| "completion_length": 770.9166851043701, | |
| "epoch": 0.906927921623513, | |
| "grad_norm": 0.19784215092658997, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.16592262306949124, | |
| "reward_std": 0.12148957094177604, | |
| "rewards/accuracy_reward": 0.16592262306949124, | |
| "step": 162 | |
| }, | |
| { | |
| "completion_length": 801.5565567016602, | |
| "epoch": 0.9125262421273618, | |
| "grad_norm": 0.16940854489803314, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.14806547743501142, | |
| "reward_std": 0.14629250299185514, | |
| "rewards/accuracy_reward": 0.14806547743501142, | |
| "step": 163 | |
| }, | |
| { | |
| "completion_length": 758.389892578125, | |
| "epoch": 0.9181245626312107, | |
| "grad_norm": 0.251857727766037, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1979166735545732, | |
| "reward_std": 0.13391391886398196, | |
| "rewards/accuracy_reward": 0.1979166735545732, | |
| "step": 164 | |
| }, | |
| { | |
| "completion_length": 778.1622200012207, | |
| "epoch": 0.9237228831350595, | |
| "grad_norm": 0.2375626116991043, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.17038691049674526, | |
| "reward_std": 0.12663642317056656, | |
| "rewards/accuracy_reward": 0.17038691049674526, | |
| "step": 165 | |
| }, | |
| { | |
| "completion_length": 788.4985237121582, | |
| "epoch": 0.9293212036389084, | |
| "grad_norm": 0.2811802327632904, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1428571462747641, | |
| "reward_std": 0.09295102627947927, | |
| "rewards/accuracy_reward": 0.1428571462747641, | |
| "step": 166 | |
| }, | |
| { | |
| "completion_length": 706.411470413208, | |
| "epoch": 0.9349195241427571, | |
| "grad_norm": 0.75251305103302, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.2306547665502876, | |
| "reward_std": 0.14719042740762234, | |
| "rewards/accuracy_reward": 0.2306547665502876, | |
| "step": 167 | |
| }, | |
| { | |
| "completion_length": 744.7879638671875, | |
| "epoch": 0.940517844646606, | |
| "grad_norm": 0.21151037514209747, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.16964286006987095, | |
| "reward_std": 0.13533117901533842, | |
| "rewards/accuracy_reward": 0.16964286006987095, | |
| "step": 168 | |
| }, | |
| { | |
| "completion_length": 734.4300689697266, | |
| "epoch": 0.9461161651504548, | |
| "grad_norm": 0.24678246676921844, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.17336309817619622, | |
| "reward_std": 0.14309520740061998, | |
| "rewards/accuracy_reward": 0.17336309817619622, | |
| "step": 169 | |
| }, | |
| { | |
| "completion_length": 724.0357284545898, | |
| "epoch": 0.9517144856543037, | |
| "grad_norm": 0.36960089206695557, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.22023809846723452, | |
| "reward_std": 0.16868228651583195, | |
| "rewards/accuracy_reward": 0.22023809846723452, | |
| "step": 170 | |
| }, | |
| { | |
| "completion_length": 681.9323043823242, | |
| "epoch": 0.9573128061581525, | |
| "grad_norm": 0.3980717062950134, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.25297619425691664, | |
| "reward_std": 0.1777943717315793, | |
| "rewards/accuracy_reward": 0.25297619425691664, | |
| "step": 171 | |
| }, | |
| { | |
| "completion_length": 706.1183166503906, | |
| "epoch": 0.9629111266620014, | |
| "grad_norm": 0.1279979944229126, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.18601190811023116, | |
| "reward_std": 0.12339989701285958, | |
| "rewards/accuracy_reward": 0.18601190811023116, | |
| "step": 172 | |
| }, | |
| { | |
| "completion_length": 696.6547718048096, | |
| "epoch": 0.9685094471658502, | |
| "grad_norm": 0.10279275476932526, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.17187500558793545, | |
| "reward_std": 0.1345887309871614, | |
| "rewards/accuracy_reward": 0.17187500558793545, | |
| "step": 173 | |
| }, | |
| { | |
| "completion_length": 659.0766487121582, | |
| "epoch": 0.9741077676696991, | |
| "grad_norm": 0.14932937920093536, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.23511904943734407, | |
| "reward_std": 0.12750320974737406, | |
| "rewards/accuracy_reward": 0.23511904943734407, | |
| "step": 174 | |
| }, | |
| { | |
| "completion_length": 710.5974884033203, | |
| "epoch": 0.979706088173548, | |
| "grad_norm": 0.15650224685668945, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.20089286181610078, | |
| "reward_std": 0.1395609532482922, | |
| "rewards/accuracy_reward": 0.20089286181610078, | |
| "step": 175 | |
| }, | |
| { | |
| "completion_length": 752.3147430419922, | |
| "epoch": 0.9853044086773968, | |
| "grad_norm": 0.14169637858867645, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1659226217889227, | |
| "reward_std": 0.14061464229598641, | |
| "rewards/accuracy_reward": 0.1659226217889227, | |
| "step": 176 | |
| }, | |
| { | |
| "completion_length": 718.4464435577393, | |
| "epoch": 0.9909027291812457, | |
| "grad_norm": 0.1935679018497467, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.20907738618552685, | |
| "reward_std": 0.13396611297503114, | |
| "rewards/accuracy_reward": 0.20907738618552685, | |
| "step": 177 | |
| }, | |
| { | |
| "completion_length": 753.9546356201172, | |
| "epoch": 0.9965010496850945, | |
| "grad_norm": 0.23016896843910217, | |
| "learning_rate": 3e-07, | |
| "loss": 0.0, | |
| "reward": 0.1510416681994684, | |
| "reward_std": 0.11266338312998414, | |
| "rewards/accuracy_reward": 0.1510416681994684, | |
| "step": 178 | |
| }, | |
| { | |
| "epoch": 0.9965010496850945, | |
| "step": 178, | |
| "total_flos": 0.0, | |
| "train_loss": 1.002405131310157e-08, | |
| "train_runtime": 79864.5906, | |
| "train_samples_per_second": 0.25, | |
| "train_steps_per_second": 0.002 | |
| } | |
| ], | |
| "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 | |
| } | |