{ "best_metric": null, "best_model_checkpoint": null, "epoch": 0.9973149213655543, "eval_steps": 50, "global_step": 1300, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.007671653241273495, "grad_norm": 1.4047255516052246, "learning_rate": 4.9616270145817344e-05, "logits/chosen": -0.5156614184379578, "logits/rejected": -0.5155973434448242, "logps/chosen": -1.6896766424179077, "logps/rejected": -3.5133140087127686, "loss": 0.6952, "rewards/accuracies": 0.5, "rewards/chosen": 0.00019091162539552897, "rewards/margins": 0.001644002622924745, "rewards/rejected": -0.0014530897606164217, "step": 10 }, { "epoch": 0.01534330648254699, "grad_norm": 1.5815409421920776, "learning_rate": 4.923254029163469e-05, "logits/chosen": -0.5282754302024841, "logits/rejected": -0.5348799824714661, "logps/chosen": -2.51069974899292, "logps/rejected": -2.873567819595337, "loss": 0.6911, "rewards/accuracies": 0.550000011920929, "rewards/chosen": 0.003554452210664749, "rewards/margins": 0.004838407970964909, "rewards/rejected": -0.0012839583214372396, "step": 20 }, { "epoch": 0.023014959723820484, "grad_norm": 1.5293035507202148, "learning_rate": 4.8848810437452035e-05, "logits/chosen": -0.46481943130493164, "logits/rejected": -0.4620364308357239, "logps/chosen": -2.9464457035064697, "logps/rejected": -4.222188472747803, "loss": 0.6923, "rewards/accuracies": 0.6000000238418579, "rewards/chosen": -0.02544591762125492, "rewards/margins": 0.06909430772066116, "rewards/rejected": -0.09454023092985153, "step": 30 }, { "epoch": 0.03068661296509398, "grad_norm": 2.4810216426849365, "learning_rate": 4.846508058326938e-05, "logits/chosen": -0.4796638488769531, "logits/rejected": -0.4740970730781555, "logps/chosen": -4.077430725097656, "logps/rejected": -2.8998451232910156, "loss": 0.6951, "rewards/accuracies": 0.4000000059604645, "rewards/chosen": -0.0723290666937828, "rewards/margins": -0.027712786570191383, "rewards/rejected": -0.044616274535655975, "step": 40 }, { "epoch": 0.038358266206367474, "grad_norm": 1.7016780376434326, "learning_rate": 4.8081350729086726e-05, "logits/chosen": -0.4345604479312897, "logits/rejected": -0.4407508969306946, "logps/chosen": -2.6753084659576416, "logps/rejected": -3.9175124168395996, "loss": 0.6948, "rewards/accuracies": 0.6000000238418579, "rewards/chosen": -0.022549040615558624, "rewards/margins": 0.023919526487588882, "rewards/rejected": -0.046468570828437805, "step": 50 }, { "epoch": 0.038358266206367474, "eval_logits/chosen": -0.2606842815876007, "eval_logits/rejected": -0.2579291760921478, "eval_logps/chosen": -3.532228946685791, "eval_logps/rejected": -3.719329833984375, "eval_loss": 0.6931471824645996, "eval_rewards/accuracies": 0.47187501192092896, "eval_rewards/chosen": -0.04993545264005661, "eval_rewards/margins": -0.011583473533391953, "eval_rewards/rejected": -0.03835199028253555, "eval_runtime": 92.1681, "eval_samples_per_second": 27.71, "eval_steps_per_second": 0.868, "step": 50 }, { "epoch": 0.04602991944764097, "grad_norm": 1.5512512922286987, "learning_rate": 4.769762087490407e-05, "logits/chosen": -0.41122546792030334, "logits/rejected": -0.4166368544101715, "logps/chosen": -3.745007276535034, "logps/rejected": -2.612656593322754, "loss": 0.6999, "rewards/accuracies": 0.5, "rewards/chosen": -0.02822841703891754, "rewards/margins": -0.011006011627614498, "rewards/rejected": -0.01722240075469017, "step": 60 }, { "epoch": 0.05370157268891446, "grad_norm": 1.307566523551941, "learning_rate": 4.731389102072141e-05, "logits/chosen": -0.45976758003234863, "logits/rejected": -0.4562222361564636, "logps/chosen": -2.7820112705230713, "logps/rejected": -3.3871638774871826, "loss": 0.694, "rewards/accuracies": 0.4749999940395355, "rewards/chosen": -0.015000693500041962, "rewards/margins": 0.0025637741200625896, "rewards/rejected": -0.01756446808576584, "step": 70 }, { "epoch": 0.06137322593018796, "grad_norm": 1.6361109018325806, "learning_rate": 4.693016116653876e-05, "logits/chosen": -0.3749166429042816, "logits/rejected": -0.37954169511795044, "logps/chosen": -2.275099992752075, "logps/rejected": -4.181614398956299, "loss": 0.6907, "rewards/accuracies": 0.550000011920929, "rewards/chosen": -0.020970512181520462, "rewards/margins": 0.025806229561567307, "rewards/rejected": -0.04677674546837807, "step": 80 }, { "epoch": 0.06904487917146145, "grad_norm": 1.83301842212677, "learning_rate": 4.65464313123561e-05, "logits/chosen": -0.5111284255981445, "logits/rejected": -0.5044411420822144, "logps/chosen": -2.0928683280944824, "logps/rejected": -5.220086574554443, "loss": 0.6899, "rewards/accuracies": 0.7250000238418579, "rewards/chosen": -0.017647314816713333, "rewards/margins": 0.07800659537315369, "rewards/rejected": -0.09565390646457672, "step": 90 }, { "epoch": 0.07671653241273495, "grad_norm": 2.6800460815429688, "learning_rate": 4.6162701458173444e-05, "logits/chosen": -0.44164401292800903, "logits/rejected": -0.44124874472618103, "logps/chosen": -3.620197296142578, "logps/rejected": -3.795258045196533, "loss": 0.7065, "rewards/accuracies": 0.550000011920929, "rewards/chosen": -0.06498938798904419, "rewards/margins": 0.0013415880966931581, "rewards/rejected": -0.06633097678422928, "step": 100 }, { "epoch": 0.07671653241273495, "eval_logits/chosen": -0.25962817668914795, "eval_logits/rejected": -0.25671660900115967, "eval_logps/chosen": -3.3411738872528076, "eval_logps/rejected": -3.666893720626831, "eval_loss": 0.6914116144180298, "eval_rewards/accuracies": 0.53125, "eval_rewards/chosen": -0.030829980969429016, "eval_rewards/margins": 0.002278365660458803, "eval_rewards/rejected": -0.03310834616422653, "eval_runtime": 87.6826, "eval_samples_per_second": 29.128, "eval_steps_per_second": 0.912, "step": 100 }, { "epoch": 0.08438818565400844, "grad_norm": 1.4465539455413818, "learning_rate": 4.577897160399079e-05, "logits/chosen": -0.32688966393470764, "logits/rejected": -0.3221715986728668, "logps/chosen": -3.515720844268799, "logps/rejected": -4.333282470703125, "loss": 0.6929, "rewards/accuracies": 0.550000011920929, "rewards/chosen": -0.0163725558668375, "rewards/margins": 0.029236530885100365, "rewards/rejected": -0.04560908302664757, "step": 110 }, { "epoch": 0.09205983889528194, "grad_norm": 1.4560517072677612, "learning_rate": 4.5395241749808135e-05, "logits/chosen": -0.48749834299087524, "logits/rejected": -0.4916568696498871, "logps/chosen": -3.262430191040039, "logps/rejected": -3.4948978424072266, "loss": 0.6837, "rewards/accuracies": 0.5, "rewards/chosen": -0.01899554952979088, "rewards/margins": 0.03163793310523033, "rewards/rejected": -0.05063347890973091, "step": 120 }, { "epoch": 0.09973149213655542, "grad_norm": 1.2213283777236938, "learning_rate": 4.501151189562548e-05, "logits/chosen": -0.4065719544887543, "logits/rejected": -0.4011871814727783, "logps/chosen": -3.5878148078918457, "logps/rejected": -4.147099494934082, "loss": 0.703, "rewards/accuracies": 0.4749999940395355, "rewards/chosen": -0.047407474368810654, "rewards/margins": 0.005889671854674816, "rewards/rejected": -0.05329713970422745, "step": 130 }, { "epoch": 0.10740314537782893, "grad_norm": 2.1541309356689453, "learning_rate": 4.4627782041442826e-05, "logits/chosen": -0.48593592643737793, "logits/rejected": -0.4823059141635895, "logps/chosen": -3.6085586547851562, "logps/rejected": -3.899576187133789, "loss": 0.7012, "rewards/accuracies": 0.4749999940395355, "rewards/chosen": -0.06346064805984497, "rewards/margins": -0.018461909145116806, "rewards/rejected": -0.044998738914728165, "step": 140 }, { "epoch": 0.11507479861910241, "grad_norm": 1.4447782039642334, "learning_rate": 4.424405218726017e-05, "logits/chosen": -0.5145712494850159, "logits/rejected": -0.5148640275001526, "logps/chosen": -4.518651962280273, "logps/rejected": -3.0175552368164062, "loss": 0.6934, "rewards/accuracies": 0.42500001192092896, "rewards/chosen": -0.06713158637285233, "rewards/margins": -0.025863245129585266, "rewards/rejected": -0.04126833751797676, "step": 150 }, { "epoch": 0.11507479861910241, "eval_logits/chosen": -0.2573604881763458, "eval_logits/rejected": -0.25465065240859985, "eval_logps/chosen": -3.439135789871216, "eval_logps/rejected": -3.6858487129211426, "eval_loss": 0.691031813621521, "eval_rewards/accuracies": 0.512499988079071, "eval_rewards/chosen": -0.04062613099813461, "eval_rewards/margins": -0.005622314754873514, "eval_rewards/rejected": -0.035003818571567535, "eval_runtime": 87.701, "eval_samples_per_second": 29.122, "eval_steps_per_second": 0.912, "step": 150 }, { "epoch": 0.12274645186037592, "grad_norm": 2.1208548545837402, "learning_rate": 4.386032233307752e-05, "logits/chosen": -0.4267118573188782, "logits/rejected": -0.42528724670410156, "logps/chosen": -2.49307918548584, "logps/rejected": -5.034300804138184, "loss": 0.6927, "rewards/accuracies": 0.574999988079071, "rewards/chosen": -0.02940777875483036, "rewards/margins": 0.01965685375034809, "rewards/rejected": -0.04906462877988815, "step": 160 }, { "epoch": 0.13041810510164942, "grad_norm": 1.4776760339736938, "learning_rate": 4.347659247889486e-05, "logits/chosen": -0.5030231475830078, "logits/rejected": -0.5042341947555542, "logps/chosen": -2.0654220581054688, "logps/rejected": -4.4005961418151855, "loss": 0.6875, "rewards/accuracies": 0.7250000238418579, "rewards/chosen": 0.0012591822305694222, "rewards/margins": 0.07784555107355118, "rewards/rejected": -0.07658636569976807, "step": 170 }, { "epoch": 0.1380897583429229, "grad_norm": 1.282079815864563, "learning_rate": 4.309286262471221e-05, "logits/chosen": -0.45819562673568726, "logits/rejected": -0.4634513258934021, "logps/chosen": -3.5271224975585938, "logps/rejected": -3.9698429107666016, "loss": 0.6878, "rewards/accuracies": 0.5249999761581421, "rewards/chosen": -0.026808280497789383, "rewards/margins": 0.02008042298257351, "rewards/rejected": -0.04688870906829834, "step": 180 }, { "epoch": 0.1457614115841964, "grad_norm": 1.368600606918335, "learning_rate": 4.270913277052955e-05, "logits/chosen": -0.384149432182312, "logits/rejected": -0.3903164267539978, "logps/chosen": -3.522733211517334, "logps/rejected": -3.272055149078369, "loss": 0.7013, "rewards/accuracies": 0.4749999940395355, "rewards/chosen": -0.037452831864356995, "rewards/margins": -0.00283648818731308, "rewards/rejected": -0.034616343677043915, "step": 190 }, { "epoch": 0.1534330648254699, "grad_norm": 1.9209519624710083, "learning_rate": 4.23254029163469e-05, "logits/chosen": -0.5128501653671265, "logits/rejected": -0.5161421298980713, "logps/chosen": -2.724151134490967, "logps/rejected": -3.3391120433807373, "loss": 0.6932, "rewards/accuracies": 0.5, "rewards/chosen": -0.008422130718827248, "rewards/margins": 0.02306986413896084, "rewards/rejected": -0.03149199113249779, "step": 200 }, { "epoch": 0.1534330648254699, "eval_logits/chosen": -0.2552909255027771, "eval_logits/rejected": -0.25272154808044434, "eval_logps/chosen": -3.351407527923584, "eval_logps/rejected": -3.6925933361053467, "eval_loss": 0.6919052600860596, "eval_rewards/accuracies": 0.528124988079071, "eval_rewards/chosen": -0.031853340566158295, "eval_rewards/margins": 0.0038249720819294453, "eval_rewards/rejected": -0.03567831218242645, "eval_runtime": 87.7075, "eval_samples_per_second": 29.12, "eval_steps_per_second": 0.912, "step": 200 }, { "epoch": 0.1611047180667434, "grad_norm": 1.575499415397644, "learning_rate": 4.194167306216424e-05, "logits/chosen": -0.45738959312438965, "logits/rejected": -0.45437225699424744, "logps/chosen": -4.087313175201416, "logps/rejected": -2.974914073944092, "loss": 0.7002, "rewards/accuracies": 0.4749999940395355, "rewards/chosen": -0.051591504365205765, "rewards/margins": -0.020506689324975014, "rewards/rejected": -0.0310848206281662, "step": 210 }, { "epoch": 0.16877637130801687, "grad_norm": 1.5426521301269531, "learning_rate": 4.155794320798158e-05, "logits/chosen": -0.4632241725921631, "logits/rejected": -0.4625597596168518, "logps/chosen": -2.200758457183838, "logps/rejected": -3.836440324783325, "loss": 0.6852, "rewards/accuracies": 0.6499999761581421, "rewards/chosen": 0.030624931678175926, "rewards/margins": 0.11931898444890976, "rewards/rejected": -0.08869405090808868, "step": 220 }, { "epoch": 0.17644802454929037, "grad_norm": 1.3270576000213623, "learning_rate": 4.117421335379893e-05, "logits/chosen": -0.4626352787017822, "logits/rejected": -0.4678708016872406, "logps/chosen": -2.704134464263916, "logps/rejected": -3.5658984184265137, "loss": 0.6898, "rewards/accuracies": 0.6000000238418579, "rewards/chosen": -0.010677630081772804, "rewards/margins": 0.02222255989909172, "rewards/rejected": -0.032900188118219376, "step": 230 }, { "epoch": 0.18411967779056387, "grad_norm": 1.4585351943969727, "learning_rate": 4.0790483499616274e-05, "logits/chosen": -0.32787808775901794, "logits/rejected": -0.3283527195453644, "logps/chosen": -3.4351840019226074, "logps/rejected": -3.090237617492676, "loss": 0.6953, "rewards/accuracies": 0.4749999940395355, "rewards/chosen": -0.02141396515071392, "rewards/margins": -0.00582413375377655, "rewards/rejected": -0.01558983325958252, "step": 240 }, { "epoch": 0.19179133103183735, "grad_norm": 1.5329344272613525, "learning_rate": 4.0406753645433616e-05, "logits/chosen": -0.4365290105342865, "logits/rejected": -0.431086927652359, "logps/chosen": -4.269787788391113, "logps/rejected": -2.5051932334899902, "loss": 0.7021, "rewards/accuracies": 0.42500001192092896, "rewards/chosen": -0.03114442527294159, "rewards/margins": -0.03054250404238701, "rewards/rejected": -0.0006019207648932934, "step": 250 }, { "epoch": 0.19179133103183735, "eval_logits/chosen": -0.25592881441116333, "eval_logits/rejected": -0.2537047863006592, "eval_logps/chosen": -2.992680311203003, "eval_logps/rejected": -3.351801633834839, "eval_loss": 0.6899347901344299, "eval_rewards/accuracies": 0.5249999761581421, "eval_rewards/chosen": 0.00401937635615468, "eval_rewards/margins": 0.005618534050881863, "eval_rewards/rejected": -0.0015991575783118606, "eval_runtime": 87.6512, "eval_samples_per_second": 29.138, "eval_steps_per_second": 0.913, "step": 250 }, { "epoch": 0.19946298427311085, "grad_norm": 2.1084399223327637, "learning_rate": 4.0023023791250965e-05, "logits/chosen": -0.4519824981689453, "logits/rejected": -0.4398951530456543, "logps/chosen": -2.484307050704956, "logps/rejected": -3.2058215141296387, "loss": 0.6891, "rewards/accuracies": 0.6000000238418579, "rewards/chosen": 0.010583693161606789, "rewards/margins": 0.03147741034626961, "rewards/rejected": -0.02089371532201767, "step": 260 }, { "epoch": 0.20713463751438435, "grad_norm": 1.2669779062271118, "learning_rate": 3.963929393706831e-05, "logits/chosen": -0.4561208188533783, "logits/rejected": -0.46242856979370117, "logps/chosen": -2.3247485160827637, "logps/rejected": -4.493135929107666, "loss": 0.6886, "rewards/accuracies": 0.550000011920929, "rewards/chosen": -0.0006832152721472085, "rewards/margins": 0.01231946051120758, "rewards/rejected": -0.013002678751945496, "step": 270 }, { "epoch": 0.21480629075565785, "grad_norm": 2.359649181365967, "learning_rate": 3.925556408288565e-05, "logits/chosen": -0.41936391592025757, "logits/rejected": -0.4227515161037445, "logps/chosen": -3.846536636352539, "logps/rejected": -2.6847872734069824, "loss": 0.6942, "rewards/accuracies": 0.42500001192092896, "rewards/chosen": -0.0359012708067894, "rewards/margins": -0.027397548779845238, "rewards/rejected": -0.00850372388958931, "step": 280 }, { "epoch": 0.22247794399693133, "grad_norm": 2.732644557952881, "learning_rate": 3.8871834228703e-05, "logits/chosen": -0.41848206520080566, "logits/rejected": -0.42105674743652344, "logps/chosen": -3.4439868927001953, "logps/rejected": -3.4636013507843018, "loss": 0.692, "rewards/accuracies": 0.5249999761581421, "rewards/chosen": -0.038081876933574677, "rewards/margins": -0.0026010065339505672, "rewards/rejected": -0.03548087924718857, "step": 290 }, { "epoch": 0.23014959723820483, "grad_norm": 1.7410922050476074, "learning_rate": 3.848810437452034e-05, "logits/chosen": -0.41233521699905396, "logits/rejected": -0.4056723117828369, "logps/chosen": -2.7670276165008545, "logps/rejected": -4.538814544677734, "loss": 0.6859, "rewards/accuracies": 0.6000000238418579, "rewards/chosen": -0.03271554037928581, "rewards/margins": 0.008648352697491646, "rewards/rejected": -0.04136389121413231, "step": 300 }, { "epoch": 0.23014959723820483, "eval_logits/chosen": -0.25867313146591187, "eval_logits/rejected": -0.2568739354610443, "eval_logps/chosen": -3.2079968452453613, "eval_logps/rejected": -3.6974151134490967, "eval_loss": 0.6905180215835571, "eval_rewards/accuracies": 0.543749988079071, "eval_rewards/chosen": -0.01751227304339409, "eval_rewards/margins": 0.018648196011781693, "eval_rewards/rejected": -0.03616046532988548, "eval_runtime": 87.8864, "eval_samples_per_second": 29.06, "eval_steps_per_second": 0.91, "step": 300 }, { "epoch": 0.23782125047947833, "grad_norm": 1.6948188543319702, "learning_rate": 3.810437452033768e-05, "logits/chosen": -0.4763808250427246, "logits/rejected": -0.47943514585494995, "logps/chosen": -3.1502089500427246, "logps/rejected": -4.5027756690979, "loss": 0.6943, "rewards/accuracies": 0.6499999761581421, "rewards/chosen": -0.018290940672159195, "rewards/margins": 0.01174425520002842, "rewards/rejected": -0.030035194009542465, "step": 310 }, { "epoch": 0.24549290372075183, "grad_norm": 1.2756513357162476, "learning_rate": 3.772064466615503e-05, "logits/chosen": -0.5276287198066711, "logits/rejected": -0.5248805284500122, "logps/chosen": -3.315579891204834, "logps/rejected": -2.8239669799804688, "loss": 0.689, "rewards/accuracies": 0.6000000238418579, "rewards/chosen": 0.0013980341609567404, "rewards/margins": 0.03776642307639122, "rewards/rejected": -0.036368388682603836, "step": 320 }, { "epoch": 0.25316455696202533, "grad_norm": 1.7512315511703491, "learning_rate": 3.733691481197237e-05, "logits/chosen": -0.4925498366355896, "logits/rejected": -0.49201974272727966, "logps/chosen": -2.7408628463745117, "logps/rejected": -3.967554807662964, "loss": 0.6947, "rewards/accuracies": 0.550000011920929, "rewards/chosen": 0.011951248161494732, "rewards/margins": 0.022660424932837486, "rewards/rejected": -0.01070917397737503, "step": 330 }, { "epoch": 0.26083621020329883, "grad_norm": 2.073042869567871, "learning_rate": 3.6953184957789715e-05, "logits/chosen": -0.4371699392795563, "logits/rejected": -0.4345782697200775, "logps/chosen": -2.061657190322876, "logps/rejected": -3.6886394023895264, "loss": 0.691, "rewards/accuracies": 0.5249999761581421, "rewards/chosen": 0.018956389278173447, "rewards/margins": 0.023976778611540794, "rewards/rejected": -0.005020388402044773, "step": 340 }, { "epoch": 0.2685078634445723, "grad_norm": 1.3207100629806519, "learning_rate": 3.6569455103607064e-05, "logits/chosen": -0.3427134156227112, "logits/rejected": -0.3307674825191498, "logps/chosen": -2.7634048461914062, "logps/rejected": -3.556063413619995, "loss": 0.6961, "rewards/accuracies": 0.5, "rewards/chosen": -0.006752232555299997, "rewards/margins": 0.004960778169333935, "rewards/rejected": -0.01171301119029522, "step": 350 }, { "epoch": 0.2685078634445723, "eval_logits/chosen": -0.2583015561103821, "eval_logits/rejected": -0.2559833824634552, "eval_logps/chosen": -3.264441728591919, "eval_logps/rejected": -3.69817852973938, "eval_loss": 0.6912429928779602, "eval_rewards/accuracies": 0.5375000238418579, "eval_rewards/chosen": -0.023156724870204926, "eval_rewards/margins": 0.013080140575766563, "eval_rewards/rejected": -0.03623687103390694, "eval_runtime": 87.903, "eval_samples_per_second": 29.055, "eval_steps_per_second": 0.91, "step": 350 }, { "epoch": 0.2761795166858458, "grad_norm": 1.3935655355453491, "learning_rate": 3.6185725249424406e-05, "logits/chosen": -0.34501805901527405, "logits/rejected": -0.3470131456851959, "logps/chosen": -3.49810791015625, "logps/rejected": -3.8505330085754395, "loss": 0.689, "rewards/accuracies": 0.4749999940395355, "rewards/chosen": -0.05252550169825554, "rewards/margins": -0.009586334228515625, "rewards/rejected": -0.042939163744449615, "step": 360 }, { "epoch": 0.2838511699271193, "grad_norm": 1.6383577585220337, "learning_rate": 3.580199539524175e-05, "logits/chosen": -0.48913854360580444, "logits/rejected": -0.4845835566520691, "logps/chosen": -3.775874376296997, "logps/rejected": -2.7326130867004395, "loss": 0.6899, "rewards/accuracies": 0.574999988079071, "rewards/chosen": -0.04597536846995354, "rewards/margins": -0.018767457455396652, "rewards/rejected": -0.027207905426621437, "step": 370 }, { "epoch": 0.2915228231683928, "grad_norm": 1.5284889936447144, "learning_rate": 3.54182655410591e-05, "logits/chosen": -0.3916495442390442, "logits/rejected": -0.3948409855365753, "logps/chosen": -3.680704116821289, "logps/rejected": -3.731623411178589, "loss": 0.6865, "rewards/accuracies": 0.5249999761581421, "rewards/chosen": -0.02748117782175541, "rewards/margins": 0.013354331254959106, "rewards/rejected": -0.04083551466464996, "step": 380 }, { "epoch": 0.2991944764096663, "grad_norm": 1.75348699092865, "learning_rate": 3.503453568687644e-05, "logits/chosen": -0.3796854615211487, "logits/rejected": -0.38789474964141846, "logps/chosen": -2.4654390811920166, "logps/rejected": -5.180593490600586, "loss": 0.6917, "rewards/accuracies": 0.6499999761581421, "rewards/chosen": -0.019467497244477272, "rewards/margins": 0.08367259800434113, "rewards/rejected": -0.10314009338617325, "step": 390 }, { "epoch": 0.3068661296509398, "grad_norm": 1.9211453199386597, "learning_rate": 3.465080583269378e-05, "logits/chosen": -0.42692264914512634, "logits/rejected": -0.42641615867614746, "logps/chosen": -2.586097240447998, "logps/rejected": -4.711082458496094, "loss": 0.6978, "rewards/accuracies": 0.5249999761581421, "rewards/chosen": -0.04758203774690628, "rewards/margins": 0.018220845609903336, "rewards/rejected": -0.06580288708209991, "step": 400 }, { "epoch": 0.3068661296509398, "eval_logits/chosen": -0.25191357731819153, "eval_logits/rejected": -0.24964801967144012, "eval_logps/chosen": -3.459468126296997, "eval_logps/rejected": -3.9026706218719482, "eval_loss": 0.6893244385719299, "eval_rewards/accuracies": 0.5249999761581421, "eval_rewards/chosen": -0.04265937954187393, "eval_rewards/margins": 0.014026666060090065, "eval_rewards/rejected": -0.05668604373931885, "eval_runtime": 87.8645, "eval_samples_per_second": 29.068, "eval_steps_per_second": 0.91, "step": 400 }, { "epoch": 0.3145377828922133, "grad_norm": 1.5234233140945435, "learning_rate": 3.426707597851113e-05, "logits/chosen": -0.28339290618896484, "logits/rejected": -0.27566584944725037, "logps/chosen": -3.7383079528808594, "logps/rejected": -3.349996566772461, "loss": 0.6824, "rewards/accuracies": 0.4749999940395355, "rewards/chosen": -0.04696209728717804, "rewards/margins": -0.0003062840551137924, "rewards/rejected": -0.0466558113694191, "step": 410 }, { "epoch": 0.3222094361334868, "grad_norm": 1.5943760871887207, "learning_rate": 3.388334612432847e-05, "logits/chosen": -0.440364271402359, "logits/rejected": -0.4411134719848633, "logps/chosen": -2.43839430809021, "logps/rejected": -4.264366626739502, "loss": 0.6874, "rewards/accuracies": 0.625, "rewards/chosen": -0.002678838325664401, "rewards/margins": 0.06886979192495346, "rewards/rejected": -0.07154864072799683, "step": 420 }, { "epoch": 0.32988108937476024, "grad_norm": 2.3275911808013916, "learning_rate": 3.3499616270145814e-05, "logits/chosen": -0.4382743239402771, "logits/rejected": -0.4452611804008484, "logps/chosen": -4.034234523773193, "logps/rejected": -3.718547821044922, "loss": 0.6962, "rewards/accuracies": 0.550000011920929, "rewards/chosen": -0.05849426984786987, "rewards/margins": 0.03669988363981247, "rewards/rejected": -0.09519414603710175, "step": 430 }, { "epoch": 0.33755274261603374, "grad_norm": 1.7006807327270508, "learning_rate": 3.311588641596316e-05, "logits/chosen": -0.488353431224823, "logits/rejected": -0.4902741014957428, "logps/chosen": -2.8092918395996094, "logps/rejected": -4.1550493240356445, "loss": 0.6961, "rewards/accuracies": 0.550000011920929, "rewards/chosen": -0.017056407406926155, "rewards/margins": 0.026486407965421677, "rewards/rejected": -0.04354281350970268, "step": 440 }, { "epoch": 0.34522439585730724, "grad_norm": 2.1540279388427734, "learning_rate": 3.2732156561780505e-05, "logits/chosen": -0.46933096647262573, "logits/rejected": -0.47831687331199646, "logps/chosen": -3.4161593914031982, "logps/rejected": -3.7839865684509277, "loss": 0.6963, "rewards/accuracies": 0.4749999940395355, "rewards/chosen": -0.062318772077560425, "rewards/margins": -0.014327460899949074, "rewards/rejected": -0.0479913167655468, "step": 450 }, { "epoch": 0.34522439585730724, "eval_logits/chosen": -0.2543574869632721, "eval_logits/rejected": -0.25201791524887085, "eval_logps/chosen": -3.321566343307495, "eval_logps/rejected": -3.75457763671875, "eval_loss": 0.6880484819412231, "eval_rewards/accuracies": 0.5249999761581421, "eval_rewards/chosen": -0.028869222849607468, "eval_rewards/margins": 0.01300754863768816, "eval_rewards/rejected": -0.04187677055597305, "eval_runtime": 87.6848, "eval_samples_per_second": 29.127, "eval_steps_per_second": 0.912, "step": 450 }, { "epoch": 0.35289604909858074, "grad_norm": 2.41361665725708, "learning_rate": 3.2348426707597854e-05, "logits/chosen": -0.517076849937439, "logits/rejected": -0.5212947130203247, "logps/chosen": -3.707463026046753, "logps/rejected": -2.6402366161346436, "loss": 0.6878, "rewards/accuracies": 0.42500001192092896, "rewards/chosen": -0.03469429910182953, "rewards/margins": -0.0030276328325271606, "rewards/rejected": -0.03166666626930237, "step": 460 }, { "epoch": 0.36056770233985425, "grad_norm": 1.4909201860427856, "learning_rate": 3.1964696853415196e-05, "logits/chosen": -0.44358164072036743, "logits/rejected": -0.4401718080043793, "logps/chosen": -2.663454532623291, "logps/rejected": -4.2893218994140625, "loss": 0.6871, "rewards/accuracies": 0.699999988079071, "rewards/chosen": 0.005995375569909811, "rewards/margins": 0.09306643158197403, "rewards/rejected": -0.08707105368375778, "step": 470 }, { "epoch": 0.36823935558112775, "grad_norm": 1.5664546489715576, "learning_rate": 3.1580966999232545e-05, "logits/chosen": -0.5029574632644653, "logits/rejected": -0.5032514333724976, "logps/chosen": -3.0805327892303467, "logps/rejected": -3.4728941917419434, "loss": 0.7022, "rewards/accuracies": 0.44999998807907104, "rewards/chosen": -0.03778943419456482, "rewards/margins": 0.011032382026314735, "rewards/rejected": -0.048821814358234406, "step": 480 }, { "epoch": 0.37591100882240125, "grad_norm": 1.4422327280044556, "learning_rate": 3.119723714504989e-05, "logits/chosen": -0.4989328384399414, "logits/rejected": -0.49760347604751587, "logps/chosen": -3.063786745071411, "logps/rejected": -3.134028434753418, "loss": 0.6919, "rewards/accuracies": 0.5, "rewards/chosen": -0.03893281891942024, "rewards/margins": -0.007891304790973663, "rewards/rejected": -0.03104151412844658, "step": 490 }, { "epoch": 0.3835826620636747, "grad_norm": 1.7234938144683838, "learning_rate": 3.0813507290867236e-05, "logits/chosen": -0.4708559513092041, "logits/rejected": -0.4755636155605316, "logps/chosen": -3.4675452709198, "logps/rejected": -4.40737247467041, "loss": 0.6909, "rewards/accuracies": 0.4749999940395355, "rewards/chosen": -0.02458677999675274, "rewards/margins": 0.0017846006667241454, "rewards/rejected": -0.02637138031423092, "step": 500 }, { "epoch": 0.3835826620636747, "eval_logits/chosen": -0.25855669379234314, "eval_logits/rejected": -0.2564457058906555, "eval_logps/chosen": -3.37125825881958, "eval_logps/rejected": -3.7394096851348877, "eval_loss": 0.6893431544303894, "eval_rewards/accuracies": 0.515625, "eval_rewards/chosen": -0.03383838012814522, "eval_rewards/margins": 0.006521561183035374, "eval_rewards/rejected": -0.04035994037985802, "eval_runtime": 87.8159, "eval_samples_per_second": 29.084, "eval_steps_per_second": 0.911, "step": 500 }, { "epoch": 0.3912543153049482, "grad_norm": 1.5501247644424438, "learning_rate": 3.0429777436684575e-05, "logits/chosen": -0.4411280155181885, "logits/rejected": -0.4376070499420166, "logps/chosen": -3.4258601665496826, "logps/rejected": -3.10990309715271, "loss": 0.6846, "rewards/accuracies": 0.5, "rewards/chosen": -0.0510549321770668, "rewards/margins": -0.015095305629074574, "rewards/rejected": -0.035959623754024506, "step": 510 }, { "epoch": 0.3989259685462217, "grad_norm": 1.5012112855911255, "learning_rate": 3.0046047582501917e-05, "logits/chosen": -0.4509127140045166, "logits/rejected": -0.4517589509487152, "logps/chosen": -2.4196524620056152, "logps/rejected": -3.98419451713562, "loss": 0.6872, "rewards/accuracies": 0.625, "rewards/chosen": 0.0005872026085853577, "rewards/margins": 0.0763540267944336, "rewards/rejected": -0.07576681673526764, "step": 520 }, { "epoch": 0.4065976217874952, "grad_norm": 2.1596052646636963, "learning_rate": 2.9662317728319266e-05, "logits/chosen": -0.41014641523361206, "logits/rejected": -0.4043218195438385, "logps/chosen": -3.8312087059020996, "logps/rejected": -2.908543109893799, "loss": 0.6886, "rewards/accuracies": 0.44999998807907104, "rewards/chosen": -0.02407076209783554, "rewards/margins": 0.02513458952307701, "rewards/rejected": -0.04920535162091255, "step": 530 }, { "epoch": 0.4142692750287687, "grad_norm": 1.4137400388717651, "learning_rate": 2.9278587874136608e-05, "logits/chosen": -0.4017483592033386, "logits/rejected": -0.39557427167892456, "logps/chosen": -3.8227367401123047, "logps/rejected": -3.195641279220581, "loss": 0.6934, "rewards/accuracies": 0.4749999940395355, "rewards/chosen": -0.03194689378142357, "rewards/margins": -0.01945153810083866, "rewards/rejected": -0.012495351955294609, "step": 540 }, { "epoch": 0.4219409282700422, "grad_norm": 1.4084391593933105, "learning_rate": 2.8894858019953953e-05, "logits/chosen": -0.4596850872039795, "logits/rejected": -0.4715039134025574, "logps/chosen": -3.1471800804138184, "logps/rejected": -3.90801739692688, "loss": 0.6944, "rewards/accuracies": 0.5, "rewards/chosen": -0.019990995526313782, "rewards/margins": 0.018570415675640106, "rewards/rejected": -0.03856140747666359, "step": 550 }, { "epoch": 0.4219409282700422, "eval_logits/chosen": -0.26123708486557007, "eval_logits/rejected": -0.25891199707984924, "eval_logps/chosen": -3.1744627952575684, "eval_logps/rejected": -3.591878890991211, "eval_loss": 0.6874387264251709, "eval_rewards/accuracies": 0.515625, "eval_rewards/chosen": -0.014158882200717926, "eval_rewards/margins": 0.011447983793914318, "eval_rewards/rejected": -0.02560686506330967, "eval_runtime": 90.5755, "eval_samples_per_second": 28.197, "eval_steps_per_second": 0.883, "step": 550 }, { "epoch": 0.4296125815113157, "grad_norm": 3.040137529373169, "learning_rate": 2.85111281657713e-05, "logits/chosen": -0.41563597321510315, "logits/rejected": -0.41134652495384216, "logps/chosen": -2.954399824142456, "logps/rejected": -2.869027853012085, "loss": 0.7005, "rewards/accuracies": 0.699999988079071, "rewards/chosen": 0.04915647208690643, "rewards/margins": 0.0781073272228241, "rewards/rejected": -0.02895086072385311, "step": 560 }, { "epoch": 0.4372842347525892, "grad_norm": 1.730731725692749, "learning_rate": 2.8127398311588644e-05, "logits/chosen": -0.43522629141807556, "logits/rejected": -0.42925986647605896, "logps/chosen": -3.231485366821289, "logps/rejected": -3.533066511154175, "loss": 0.6926, "rewards/accuracies": 0.6499999761581421, "rewards/chosen": 0.012996886856853962, "rewards/margins": 0.05765851214528084, "rewards/rejected": -0.04466162249445915, "step": 570 }, { "epoch": 0.44495588799386265, "grad_norm": 1.4563325643539429, "learning_rate": 2.7743668457405987e-05, "logits/chosen": -0.4720652103424072, "logits/rejected": -0.46499529480934143, "logps/chosen": -3.880791187286377, "logps/rejected": -3.5834224224090576, "loss": 0.7003, "rewards/accuracies": 0.5, "rewards/chosen": -0.055101800709962845, "rewards/margins": -0.01127576269209385, "rewards/rejected": -0.043826036155223846, "step": 580 }, { "epoch": 0.45262754123513615, "grad_norm": 1.680908441543579, "learning_rate": 2.7359938603223335e-05, "logits/chosen": -0.2684255540370941, "logits/rejected": -0.2682047188282013, "logps/chosen": -2.8263630867004395, "logps/rejected": -3.5271003246307373, "loss": 0.6887, "rewards/accuracies": 0.699999988079071, "rewards/chosen": 0.018410243093967438, "rewards/margins": 0.04611648991703987, "rewards/rejected": -0.027706250548362732, "step": 590 }, { "epoch": 0.46029919447640966, "grad_norm": 1.4875671863555908, "learning_rate": 2.6976208749040678e-05, "logits/chosen": -0.5006714463233948, "logits/rejected": -0.5000745058059692, "logps/chosen": -2.8757147789001465, "logps/rejected": -3.4219863414764404, "loss": 0.6837, "rewards/accuracies": 0.675000011920929, "rewards/chosen": 0.022861991077661514, "rewards/margins": 0.0631416067481041, "rewards/rejected": -0.04027962684631348, "step": 600 }, { "epoch": 0.46029919447640966, "eval_logits/chosen": -0.25836092233657837, "eval_logits/rejected": -0.2553294599056244, "eval_logps/chosen": -3.034031867980957, "eval_logps/rejected": -3.487445116043091, "eval_loss": 0.6917591691017151, "eval_rewards/accuracies": 0.515625, "eval_rewards/chosen": -0.00011577569239307195, "eval_rewards/margins": 0.015047714114189148, "eval_rewards/rejected": -0.015163486823439598, "eval_runtime": 95.32, "eval_samples_per_second": 26.794, "eval_steps_per_second": 0.839, "step": 600 }, { "epoch": 0.46797084771768316, "grad_norm": 1.0080013275146484, "learning_rate": 2.659247889485802e-05, "logits/chosen": -0.4055057466030121, "logits/rejected": -0.4117734432220459, "logps/chosen": -2.3307080268859863, "logps/rejected": -3.9688384532928467, "loss": 0.6911, "rewards/accuracies": 0.6000000238418579, "rewards/chosen": -0.004725713282823563, "rewards/margins": -4.6133995965647046e-06, "rewards/rejected": -0.004721099976450205, "step": 610 }, { "epoch": 0.47564250095895666, "grad_norm": 1.2446097135543823, "learning_rate": 2.620874904067537e-05, "logits/chosen": -0.3564910292625427, "logits/rejected": -0.3598317801952362, "logps/chosen": -2.8816721439361572, "logps/rejected": -3.8162410259246826, "loss": 0.7045, "rewards/accuracies": 0.4749999940395355, "rewards/chosen": -0.04238850623369217, "rewards/margins": -0.050994087010622025, "rewards/rejected": 0.00860558170825243, "step": 620 }, { "epoch": 0.48331415420023016, "grad_norm": 2.1006405353546143, "learning_rate": 2.582501918649271e-05, "logits/chosen": -0.4763752520084381, "logits/rejected": -0.4724522531032562, "logps/chosen": -3.6634361743927, "logps/rejected": -3.523616313934326, "loss": 0.6876, "rewards/accuracies": 0.550000011920929, "rewards/chosen": -0.015630219131708145, "rewards/margins": 0.03971021622419357, "rewards/rejected": -0.05534043163061142, "step": 630 }, { "epoch": 0.49098580744150366, "grad_norm": 1.9049338102340698, "learning_rate": 2.5441289332310053e-05, "logits/chosen": -0.46114760637283325, "logits/rejected": -0.4514070451259613, "logps/chosen": -3.082925796508789, "logps/rejected": -3.305220365524292, "loss": 0.6923, "rewards/accuracies": 0.44999998807907104, "rewards/chosen": -0.03719641640782356, "rewards/margins": -0.010646897368133068, "rewards/rejected": -0.02654952183365822, "step": 640 }, { "epoch": 0.49865746068277716, "grad_norm": 1.5372164249420166, "learning_rate": 2.50575594781274e-05, "logits/chosen": -0.4374101758003235, "logits/rejected": -0.4314614236354828, "logps/chosen": -3.3405063152313232, "logps/rejected": -3.743143081665039, "loss": 0.6894, "rewards/accuracies": 0.6000000238418579, "rewards/chosen": -0.011367046274244785, "rewards/margins": 0.05027226731181145, "rewards/rejected": -0.061639316380023956, "step": 650 }, { "epoch": 0.49865746068277716, "eval_logits/chosen": -0.25022387504577637, "eval_logits/rejected": -0.24710026383399963, "eval_logps/chosen": -3.5461456775665283, "eval_logps/rejected": -3.8798840045928955, "eval_loss": 0.688687801361084, "eval_rewards/accuracies": 0.49687498807907104, "eval_rewards/chosen": -0.0513271763920784, "eval_rewards/margins": 0.0030801701359450817, "eval_rewards/rejected": -0.05440734699368477, "eval_runtime": 89.1213, "eval_samples_per_second": 28.658, "eval_steps_per_second": 0.898, "step": 650 }, { "epoch": 0.5063291139240507, "grad_norm": 1.4859964847564697, "learning_rate": 2.4673829623944744e-05, "logits/chosen": -0.4470909535884857, "logits/rejected": -0.4404575228691101, "logps/chosen": -3.57057523727417, "logps/rejected": -4.062138080596924, "loss": 0.6839, "rewards/accuracies": 0.550000011920929, "rewards/chosen": -0.05643466114997864, "rewards/margins": 0.012388080358505249, "rewards/rejected": -0.06882274150848389, "step": 660 }, { "epoch": 0.5140007671653242, "grad_norm": 2.348560333251953, "learning_rate": 2.429009976976209e-05, "logits/chosen": -0.4324890971183777, "logits/rejected": -0.4288514256477356, "logps/chosen": -3.4150989055633545, "logps/rejected": -3.761260509490967, "loss": 0.6905, "rewards/accuracies": 0.574999988079071, "rewards/chosen": -0.047535546123981476, "rewards/margins": 0.00843095313757658, "rewards/rejected": -0.05596649646759033, "step": 670 }, { "epoch": 0.5216724204065977, "grad_norm": 2.114100933074951, "learning_rate": 2.390636991557943e-05, "logits/chosen": -0.4224484860897064, "logits/rejected": -0.41605424880981445, "logps/chosen": -3.8406059741973877, "logps/rejected": -3.227407455444336, "loss": 0.6923, "rewards/accuracies": 0.4749999940395355, "rewards/chosen": -0.09835998713970184, "rewards/margins": -0.028456950560212135, "rewards/rejected": -0.06990303844213486, "step": 680 }, { "epoch": 0.5293440736478712, "grad_norm": 2.143270254135132, "learning_rate": 2.3522640061396777e-05, "logits/chosen": -0.35925137996673584, "logits/rejected": -0.35858094692230225, "logps/chosen": -4.149369239807129, "logps/rejected": -4.207783222198486, "loss": 0.6901, "rewards/accuracies": 0.574999988079071, "rewards/chosen": -0.07230309396982193, "rewards/margins": 0.022937068715691566, "rewards/rejected": -0.09524016082286835, "step": 690 }, { "epoch": 0.5370157268891446, "grad_norm": 1.3052890300750732, "learning_rate": 2.3138910207214122e-05, "logits/chosen": -0.4177284836769104, "logits/rejected": -0.4176989197731018, "logps/chosen": -2.371631622314453, "logps/rejected": -5.967281341552734, "loss": 0.6972, "rewards/accuracies": 0.699999988079071, "rewards/chosen": -0.025050634518265724, "rewards/margins": 0.0979982241988182, "rewards/rejected": -0.12304886430501938, "step": 700 }, { "epoch": 0.5370157268891446, "eval_logits/chosen": -0.2624206840991974, "eval_logits/rejected": -0.259426087141037, "eval_logps/chosen": -3.501861572265625, "eval_logps/rejected": -3.8878912925720215, "eval_loss": 0.6871370673179626, "eval_rewards/accuracies": 0.48124998807907104, "eval_rewards/chosen": -0.046898726373910904, "eval_rewards/margins": 0.008309418335556984, "eval_rewards/rejected": -0.05520814657211304, "eval_runtime": 91.2498, "eval_samples_per_second": 27.989, "eval_steps_per_second": 0.877, "step": 700 }, { "epoch": 0.5446873801304181, "grad_norm": 1.2912267446517944, "learning_rate": 2.2755180353031468e-05, "logits/chosen": -0.45912307500839233, "logits/rejected": -0.4561159014701843, "logps/chosen": -3.2371792793273926, "logps/rejected": -3.8025176525115967, "loss": 0.6983, "rewards/accuracies": 0.574999988079071, "rewards/chosen": -0.024336762726306915, "rewards/margins": 0.016023293137550354, "rewards/rejected": -0.040360044687986374, "step": 710 }, { "epoch": 0.5523590333716916, "grad_norm": 1.5830397605895996, "learning_rate": 2.2371450498848813e-05, "logits/chosen": -0.382514089345932, "logits/rejected": -0.3864609897136688, "logps/chosen": -1.6761538982391357, "logps/rejected": -4.973038673400879, "loss": 0.6851, "rewards/accuracies": 0.6499999761581421, "rewards/chosen": -0.012824362143874168, "rewards/margins": 0.0575835183262825, "rewards/rejected": -0.07040788233280182, "step": 720 }, { "epoch": 0.5600306866129651, "grad_norm": 1.851820707321167, "learning_rate": 2.198772064466616e-05, "logits/chosen": -0.30062246322631836, "logits/rejected": -0.3018190860748291, "logps/chosen": -2.90508770942688, "logps/rejected": -3.7549374103546143, "loss": 0.6959, "rewards/accuracies": 0.625, "rewards/chosen": -0.02952618896961212, "rewards/margins": 0.014450429007411003, "rewards/rejected": -0.04397661238908768, "step": 730 }, { "epoch": 0.5677023398542386, "grad_norm": 1.2601155042648315, "learning_rate": 2.16039907904835e-05, "logits/chosen": -0.5282914042472839, "logits/rejected": -0.5290226936340332, "logps/chosen": -2.747199535369873, "logps/rejected": -3.397148609161377, "loss": 0.6875, "rewards/accuracies": 0.550000011920929, "rewards/chosen": 0.023777935653924942, "rewards/margins": 0.060780882835388184, "rewards/rejected": -0.03700294345617294, "step": 740 }, { "epoch": 0.5753739930955121, "grad_norm": 1.9561665058135986, "learning_rate": 2.1220260936300846e-05, "logits/chosen": -0.46929964423179626, "logits/rejected": -0.4689921736717224, "logps/chosen": -3.4335849285125732, "logps/rejected": -3.539712429046631, "loss": 0.6861, "rewards/accuracies": 0.6499999761581421, "rewards/chosen": -0.01373897772282362, "rewards/margins": 0.052022386342287064, "rewards/rejected": -0.06576136499643326, "step": 750 }, { "epoch": 0.5753739930955121, "eval_logits/chosen": -0.26118698716163635, "eval_logits/rejected": -0.25837644934654236, "eval_logps/chosen": -3.3246681690216064, "eval_logps/rejected": -3.706329345703125, "eval_loss": 0.6877872943878174, "eval_rewards/accuracies": 0.512499988079071, "eval_rewards/chosen": -0.029179394245147705, "eval_rewards/margins": 0.007872538641095161, "eval_rewards/rejected": -0.03705193102359772, "eval_runtime": 95.7364, "eval_samples_per_second": 26.677, "eval_steps_per_second": 0.836, "step": 750 }, { "epoch": 0.5830456463367856, "grad_norm": 1.3404334783554077, "learning_rate": 2.0836531082118192e-05, "logits/chosen": -0.37314945459365845, "logits/rejected": -0.3766701817512512, "logps/chosen": -2.7383177280426025, "logps/rejected": -2.9454519748687744, "loss": 0.693, "rewards/accuracies": 0.4749999940395355, "rewards/chosen": -0.04017211124300957, "rewards/margins": -0.045864250510931015, "rewards/rejected": 0.005692136473953724, "step": 760 }, { "epoch": 0.5907172995780591, "grad_norm": 1.4485934972763062, "learning_rate": 2.0452801227935534e-05, "logits/chosen": -0.46926283836364746, "logits/rejected": -0.46992430090904236, "logps/chosen": -3.8600451946258545, "logps/rejected": -3.410430431365967, "loss": 0.6925, "rewards/accuracies": 0.5, "rewards/chosen": -0.031888850033283234, "rewards/margins": 0.010654223151504993, "rewards/rejected": -0.042543068528175354, "step": 770 }, { "epoch": 0.5983889528193326, "grad_norm": 1.8931002616882324, "learning_rate": 2.006907137375288e-05, "logits/chosen": -0.3966008424758911, "logits/rejected": -0.4044922888278961, "logps/chosen": -2.1048827171325684, "logps/rejected": -4.816119194030762, "loss": 0.6858, "rewards/accuracies": 0.625, "rewards/chosen": 0.002983979182317853, "rewards/margins": 0.08909803628921509, "rewards/rejected": -0.0861140564084053, "step": 780 }, { "epoch": 0.6060606060606061, "grad_norm": 2.2583765983581543, "learning_rate": 1.9685341519570225e-05, "logits/chosen": -0.4961305558681488, "logits/rejected": -0.4885970652103424, "logps/chosen": -3.6672680377960205, "logps/rejected": -3.291787624359131, "loss": 0.692, "rewards/accuracies": 0.5249999761581421, "rewards/chosen": -0.02681976929306984, "rewards/margins": 0.014080104418098927, "rewards/rejected": -0.04089986905455589, "step": 790 }, { "epoch": 0.6137322593018796, "grad_norm": 1.8438897132873535, "learning_rate": 1.9301611665387567e-05, "logits/chosen": -0.46609607338905334, "logits/rejected": -0.4627665877342224, "logps/chosen": -3.5119385719299316, "logps/rejected": -3.595712184906006, "loss": 0.6866, "rewards/accuracies": 0.6000000238418579, "rewards/chosen": -0.025941628962755203, "rewards/margins": 0.030908849090337753, "rewards/rejected": -0.056850481778383255, "step": 800 }, { "epoch": 0.6137322593018796, "eval_logits/chosen": -0.2616404891014099, "eval_logits/rejected": -0.25945189595222473, "eval_logps/chosen": -3.2828381061553955, "eval_logps/rejected": -3.7320716381073, "eval_loss": 0.6877512335777283, "eval_rewards/accuracies": 0.518750011920929, "eval_rewards/chosen": -0.024996396154165268, "eval_rewards/margins": 0.014629754237830639, "eval_rewards/rejected": -0.03962615132331848, "eval_runtime": 96.9287, "eval_samples_per_second": 26.349, "eval_steps_per_second": 0.825, "step": 800 }, { "epoch": 0.6214039125431531, "grad_norm": 1.5462312698364258, "learning_rate": 1.8917881811204912e-05, "logits/chosen": -0.5312341451644897, "logits/rejected": -0.5319844484329224, "logps/chosen": -3.227830410003662, "logps/rejected": -3.9378840923309326, "loss": 0.6881, "rewards/accuracies": 0.625, "rewards/chosen": -0.031698137521743774, "rewards/margins": 0.033506251871585846, "rewards/rejected": -0.06520438939332962, "step": 810 }, { "epoch": 0.6290755657844266, "grad_norm": 1.644169807434082, "learning_rate": 1.8534151957022258e-05, "logits/chosen": -0.42604535818099976, "logits/rejected": -0.42680415511131287, "logps/chosen": -3.025153875350952, "logps/rejected": -3.5833792686462402, "loss": 0.6951, "rewards/accuracies": 0.6000000238418579, "rewards/chosen": -0.017264336347579956, "rewards/margins": 0.03092106617987156, "rewards/rejected": -0.048185400664806366, "step": 820 }, { "epoch": 0.6367472190257001, "grad_norm": 2.567456007003784, "learning_rate": 1.81504221028396e-05, "logits/chosen": -0.3624061048030853, "logits/rejected": -0.3603518605232239, "logps/chosen": -2.727677583694458, "logps/rejected": -4.618671417236328, "loss": 0.6831, "rewards/accuracies": 0.675000011920929, "rewards/chosen": 0.013222327455878258, "rewards/margins": 0.04956119880080223, "rewards/rejected": -0.03633887320756912, "step": 830 }, { "epoch": 0.6444188722669736, "grad_norm": 1.6993279457092285, "learning_rate": 1.7766692248656946e-05, "logits/chosen": -0.43364423513412476, "logits/rejected": -0.43153029680252075, "logps/chosen": -3.7280707359313965, "logps/rejected": -3.3984951972961426, "loss": 0.6892, "rewards/accuracies": 0.6000000238418579, "rewards/chosen": -0.003762350883334875, "rewards/margins": 0.04601341485977173, "rewards/rejected": -0.04977576062083244, "step": 840 }, { "epoch": 0.652090525508247, "grad_norm": 1.2924683094024658, "learning_rate": 1.738296239447429e-05, "logits/chosen": -0.40035882592201233, "logits/rejected": -0.39904215931892395, "logps/chosen": -2.473344087600708, "logps/rejected": -4.267284393310547, "loss": 0.6901, "rewards/accuracies": 0.574999988079071, "rewards/chosen": -0.02397814765572548, "rewards/margins": 0.036180444061756134, "rewards/rejected": -0.060158587992191315, "step": 850 }, { "epoch": 0.652090525508247, "eval_logits/chosen": -0.264896422624588, "eval_logits/rejected": -0.2623136341571808, "eval_logps/chosen": -3.455109119415283, "eval_logps/rejected": -3.909381866455078, "eval_loss": 0.6873416304588318, "eval_rewards/accuracies": 0.503125011920929, "eval_rewards/chosen": -0.04222346469759941, "eval_rewards/margins": 0.015133677050471306, "eval_rewards/rejected": -0.057357143610715866, "eval_runtime": 87.969, "eval_samples_per_second": 29.033, "eval_steps_per_second": 0.909, "step": 850 }, { "epoch": 0.6597621787495205, "grad_norm": 1.6212762594223022, "learning_rate": 1.6999232540291633e-05, "logits/chosen": -0.41735363006591797, "logits/rejected": -0.4102698266506195, "logps/chosen": -3.7062439918518066, "logps/rejected": -3.9006316661834717, "loss": 0.6915, "rewards/accuracies": 0.5249999761581421, "rewards/chosen": -0.0529264435172081, "rewards/margins": 0.005926959216594696, "rewards/rejected": -0.0588533990085125, "step": 860 }, { "epoch": 0.667433831990794, "grad_norm": 2.2773942947387695, "learning_rate": 1.661550268610898e-05, "logits/chosen": -0.37981244921684265, "logits/rejected": -0.3839191496372223, "logps/chosen": -2.9031481742858887, "logps/rejected": -4.830349922180176, "loss": 0.6897, "rewards/accuracies": 0.625, "rewards/chosen": -0.02680964395403862, "rewards/margins": 0.06183745339512825, "rewards/rejected": -0.08864708989858627, "step": 870 }, { "epoch": 0.6751054852320675, "grad_norm": 1.763682246208191, "learning_rate": 1.6231772831926324e-05, "logits/chosen": -0.34375470876693726, "logits/rejected": -0.3387607932090759, "logps/chosen": -3.3046181201934814, "logps/rejected": -4.335677146911621, "loss": 0.6843, "rewards/accuracies": 0.574999988079071, "rewards/chosen": -0.04779592901468277, "rewards/margins": 0.02044987678527832, "rewards/rejected": -0.06824581325054169, "step": 880 }, { "epoch": 0.682777138473341, "grad_norm": 2.387399196624756, "learning_rate": 1.584804297774367e-05, "logits/chosen": -0.4725160598754883, "logits/rejected": -0.4763157367706299, "logps/chosen": -2.8794379234313965, "logps/rejected": -4.024861812591553, "loss": 0.6943, "rewards/accuracies": 0.5, "rewards/chosen": -0.046050556004047394, "rewards/margins": 0.01775188557803631, "rewards/rejected": -0.06380244344472885, "step": 890 }, { "epoch": 0.6904487917146145, "grad_norm": 1.5189565420150757, "learning_rate": 1.5464313123561015e-05, "logits/chosen": -0.31694844365119934, "logits/rejected": -0.3208076059818268, "logps/chosen": -2.9370594024658203, "logps/rejected": -4.222288608551025, "loss": 0.6835, "rewards/accuracies": 0.6499999761581421, "rewards/chosen": -0.04458096995949745, "rewards/margins": 0.028453299775719643, "rewards/rejected": -0.07303427159786224, "step": 900 }, { "epoch": 0.6904487917146145, "eval_logits/chosen": -0.2631658613681793, "eval_logits/rejected": -0.2605777382850647, "eval_logps/chosen": -3.5630836486816406, "eval_logps/rejected": -4.000800132751465, "eval_loss": 0.6874542832374573, "eval_rewards/accuracies": 0.5093749761581421, "eval_rewards/chosen": -0.053020983934402466, "eval_rewards/margins": 0.013478020206093788, "eval_rewards/rejected": -0.0664990097284317, "eval_runtime": 87.9229, "eval_samples_per_second": 29.048, "eval_steps_per_second": 0.91, "step": 900 }, { "epoch": 0.698120444955888, "grad_norm": 1.9298661947250366, "learning_rate": 1.5080583269378359e-05, "logits/chosen": -0.4393790364265442, "logits/rejected": -0.43957772850990295, "logps/chosen": -3.8378818035125732, "logps/rejected": -3.808051586151123, "loss": 0.6941, "rewards/accuracies": 0.4749999940395355, "rewards/chosen": -0.08017073571681976, "rewards/margins": -0.014708010479807854, "rewards/rejected": -0.06546272337436676, "step": 910 }, { "epoch": 0.7057920981971615, "grad_norm": 1.7689329385757446, "learning_rate": 1.4696853415195703e-05, "logits/chosen": -0.46744298934936523, "logits/rejected": -0.4742264747619629, "logps/chosen": -3.0718374252319336, "logps/rejected": -3.5021071434020996, "loss": 0.6854, "rewards/accuracies": 0.574999988079071, "rewards/chosen": -0.03193381056189537, "rewards/margins": 0.04194425791501999, "rewards/rejected": -0.07387807220220566, "step": 920 }, { "epoch": 0.713463751438435, "grad_norm": 1.4333994388580322, "learning_rate": 1.4313123561013048e-05, "logits/chosen": -0.3608981966972351, "logits/rejected": -0.3596799373626709, "logps/chosen": -4.751004219055176, "logps/rejected": -3.0089735984802246, "loss": 0.6926, "rewards/accuracies": 0.4000000059604645, "rewards/chosen": -0.0744011402130127, "rewards/margins": -0.03302520513534546, "rewards/rejected": -0.04137593135237694, "step": 930 }, { "epoch": 0.7211354046797085, "grad_norm": 1.605447769165039, "learning_rate": 1.3929393706830394e-05, "logits/chosen": -0.3481396734714508, "logits/rejected": -0.3457208275794983, "logps/chosen": -4.311774253845215, "logps/rejected": -3.5297610759735107, "loss": 0.6945, "rewards/accuracies": 0.44999998807907104, "rewards/chosen": -0.06438607722520828, "rewards/margins": -0.021948862820863724, "rewards/rejected": -0.04243721812963486, "step": 940 }, { "epoch": 0.728807057920982, "grad_norm": 0.9545491337776184, "learning_rate": 1.3545663852647736e-05, "logits/chosen": -0.45011836290359497, "logits/rejected": -0.4549293518066406, "logps/chosen": -2.901266098022461, "logps/rejected": -3.653916835784912, "loss": 0.6907, "rewards/accuracies": 0.550000011920929, "rewards/chosen": -0.004453243222087622, "rewards/margins": 0.06575339287519455, "rewards/rejected": -0.07020662724971771, "step": 950 }, { "epoch": 0.728807057920982, "eval_logits/chosen": -0.2597009241580963, "eval_logits/rejected": -0.25726523995399475, "eval_logps/chosen": -3.299370288848877, "eval_logps/rejected": -3.774941921234131, "eval_loss": 0.6873298287391663, "eval_rewards/accuracies": 0.53125, "eval_rewards/chosen": -0.0266495980322361, "eval_rewards/margins": 0.01726352609694004, "eval_rewards/rejected": -0.04391312599182129, "eval_runtime": 87.9968, "eval_samples_per_second": 29.024, "eval_steps_per_second": 0.909, "step": 950 }, { "epoch": 0.7364787111622555, "grad_norm": 1.472712755203247, "learning_rate": 1.3161933998465081e-05, "logits/chosen": -0.43601498007774353, "logits/rejected": -0.4370816648006439, "logps/chosen": -2.949425220489502, "logps/rejected": -3.2113566398620605, "loss": 0.6916, "rewards/accuracies": 0.4000000059604645, "rewards/chosen": -0.03191310912370682, "rewards/margins": -0.030944937840104103, "rewards/rejected": -0.0009681672090664506, "step": 960 }, { "epoch": 0.744150364403529, "grad_norm": 1.710877537727356, "learning_rate": 1.2778204144282427e-05, "logits/chosen": -0.46664342284202576, "logits/rejected": -0.466810941696167, "logps/chosen": -3.9124972820281982, "logps/rejected": -3.6439247131347656, "loss": 0.6928, "rewards/accuracies": 0.550000011920929, "rewards/chosen": -0.05568718910217285, "rewards/margins": -0.004351111128926277, "rewards/rejected": -0.05133608728647232, "step": 970 }, { "epoch": 0.7518220176448025, "grad_norm": 1.8609956502914429, "learning_rate": 1.239447429009977e-05, "logits/chosen": -0.5233210325241089, "logits/rejected": -0.5286127328872681, "logps/chosen": -3.5049729347229004, "logps/rejected": -2.9099135398864746, "loss": 0.6877, "rewards/accuracies": 0.5249999761581421, "rewards/chosen": -0.02429554983973503, "rewards/margins": 0.04005228728055954, "rewards/rejected": -0.06434784084558487, "step": 980 }, { "epoch": 0.759493670886076, "grad_norm": 2.3433878421783447, "learning_rate": 1.2010744435917114e-05, "logits/chosen": -0.5355673432350159, "logits/rejected": -0.5309178233146667, "logps/chosen": -2.8020567893981934, "logps/rejected": -4.379094123840332, "loss": 0.6869, "rewards/accuracies": 0.44999998807907104, "rewards/chosen": -0.034182898700237274, "rewards/margins": 0.029533013701438904, "rewards/rejected": -0.06371591240167618, "step": 990 }, { "epoch": 0.7671653241273494, "grad_norm": 1.5511538982391357, "learning_rate": 1.162701458173446e-05, "logits/chosen": -0.41299349069595337, "logits/rejected": -0.4154202342033386, "logps/chosen": -2.535586357116699, "logps/rejected": -4.728541851043701, "loss": 0.6938, "rewards/accuracies": 0.6499999761581421, "rewards/chosen": -0.01752440631389618, "rewards/margins": 0.10527894645929337, "rewards/rejected": -0.12280334532260895, "step": 1000 }, { "epoch": 0.7671653241273494, "eval_logits/chosen": -0.25718361139297485, "eval_logits/rejected": -0.2549431622028351, "eval_logps/chosen": -3.5508739948272705, "eval_logps/rejected": -3.9987282752990723, "eval_loss": 0.687543511390686, "eval_rewards/accuracies": 0.512499988079071, "eval_rewards/chosen": -0.051799990236759186, "eval_rewards/margins": 0.014491766691207886, "eval_rewards/rejected": -0.06629175692796707, "eval_runtime": 87.5816, "eval_samples_per_second": 29.161, "eval_steps_per_second": 0.913, "step": 1000 }, { "epoch": 0.7748369773686229, "grad_norm": 1.9154669046401978, "learning_rate": 1.1243284727551805e-05, "logits/chosen": -0.5070533156394958, "logits/rejected": -0.5113304853439331, "logps/chosen": -3.3207507133483887, "logps/rejected": -4.22388219833374, "loss": 0.6931, "rewards/accuracies": 0.4749999940395355, "rewards/chosen": -0.0673786848783493, "rewards/margins": -0.0010321959853172302, "rewards/rejected": -0.06634648889303207, "step": 1010 }, { "epoch": 0.7825086306098964, "grad_norm": 1.3908665180206299, "learning_rate": 1.0859554873369149e-05, "logits/chosen": -0.45152488350868225, "logits/rejected": -0.4454428553581238, "logps/chosen": -4.341331481933594, "logps/rejected": -2.536820650100708, "loss": 0.688, "rewards/accuracies": 0.42500001192092896, "rewards/chosen": -0.0593075156211853, "rewards/margins": -0.003714442951604724, "rewards/rejected": -0.05559306591749191, "step": 1020 }, { "epoch": 0.7901802838511699, "grad_norm": 1.4446948766708374, "learning_rate": 1.0475825019186493e-05, "logits/chosen": -0.4947948455810547, "logits/rejected": -0.4966978132724762, "logps/chosen": -3.387843608856201, "logps/rejected": -3.6044819355010986, "loss": 0.6969, "rewards/accuracies": 0.4749999940395355, "rewards/chosen": -0.06083371490240097, "rewards/margins": -0.006257879547774792, "rewards/rejected": -0.0545758493244648, "step": 1030 }, { "epoch": 0.7978519370924434, "grad_norm": 1.610182523727417, "learning_rate": 1.0092095165003838e-05, "logits/chosen": -0.46212753653526306, "logits/rejected": -0.46821293234825134, "logps/chosen": -4.950307846069336, "logps/rejected": -3.0810484886169434, "loss": 0.69, "rewards/accuracies": 0.375, "rewards/chosen": -0.04846534878015518, "rewards/margins": -0.025759827345609665, "rewards/rejected": -0.022705521434545517, "step": 1040 }, { "epoch": 0.8055235903337169, "grad_norm": 2.2419750690460205, "learning_rate": 9.708365310821182e-06, "logits/chosen": -0.5252290964126587, "logits/rejected": -0.5231043100357056, "logps/chosen": -3.5570881366729736, "logps/rejected": -3.7880332469940186, "loss": 0.6942, "rewards/accuracies": 0.5, "rewards/chosen": -0.06180233880877495, "rewards/margins": -0.026278842240571976, "rewards/rejected": -0.03552350401878357, "step": 1050 }, { "epoch": 0.8055235903337169, "eval_logits/chosen": -0.25850772857666016, "eval_logits/rejected": -0.25608521699905396, "eval_logps/chosen": -3.3484771251678467, "eval_logps/rejected": -3.820146083831787, "eval_loss": 0.6879977583885193, "eval_rewards/accuracies": 0.515625, "eval_rewards/chosen": -0.031560301780700684, "eval_rewards/margins": 0.01687331683933735, "eval_rewards/rejected": -0.04843362048268318, "eval_runtime": 87.9503, "eval_samples_per_second": 29.039, "eval_steps_per_second": 0.91, "step": 1050 }, { "epoch": 0.8131952435749904, "grad_norm": 1.5563366413116455, "learning_rate": 9.324635456638526e-06, "logits/chosen": -0.4408280849456787, "logits/rejected": -0.44397348165512085, "logps/chosen": -3.351261854171753, "logps/rejected": -3.4453811645507812, "loss": 0.6788, "rewards/accuracies": 0.5, "rewards/chosen": -0.03186880797147751, "rewards/margins": 0.015346929430961609, "rewards/rejected": -0.04721573740243912, "step": 1060 }, { "epoch": 0.8208668968162639, "grad_norm": 1.7316789627075195, "learning_rate": 8.940905602455871e-06, "logits/chosen": -0.4371892511844635, "logits/rejected": -0.43460506200790405, "logps/chosen": -3.015501022338867, "logps/rejected": -4.129702091217041, "loss": 0.6886, "rewards/accuracies": 0.625, "rewards/chosen": 0.010989089496433735, "rewards/margins": 0.06226798892021179, "rewards/rejected": -0.05127890035510063, "step": 1070 }, { "epoch": 0.8285385500575374, "grad_norm": 1.842515230178833, "learning_rate": 8.557175748273217e-06, "logits/chosen": -0.43732374906539917, "logits/rejected": -0.4389157295227051, "logps/chosen": -3.6136844158172607, "logps/rejected": -4.016729354858398, "loss": 0.6864, "rewards/accuracies": 0.5, "rewards/chosen": -0.029665043577551842, "rewards/margins": 0.018349073827266693, "rewards/rejected": -0.04801411181688309, "step": 1080 }, { "epoch": 0.8362102032988109, "grad_norm": 1.4809995889663696, "learning_rate": 8.17344589409056e-06, "logits/chosen": -0.39032843708992004, "logits/rejected": -0.3862852156162262, "logps/chosen": -3.43218731880188, "logps/rejected": -3.326885223388672, "loss": 0.7062, "rewards/accuracies": 0.550000011920929, "rewards/chosen": -0.03662601485848427, "rewards/margins": 2.0000337826786563e-05, "rewards/rejected": -0.03664601594209671, "step": 1090 }, { "epoch": 0.8438818565400844, "grad_norm": 1.7941354513168335, "learning_rate": 7.789716039907906e-06, "logits/chosen": -0.3972841799259186, "logits/rejected": -0.4001404643058777, "logps/chosen": -2.768026828765869, "logps/rejected": -3.8738293647766113, "loss": 0.6784, "rewards/accuracies": 0.5249999761581421, "rewards/chosen": -0.028935182839632034, "rewards/margins": -0.008622194640338421, "rewards/rejected": -0.020312989130616188, "step": 1100 }, { "epoch": 0.8438818565400844, "eval_logits/chosen": -0.25824981927871704, "eval_logits/rejected": -0.2558724284172058, "eval_logps/chosen": -3.438009262084961, "eval_logps/rejected": -3.9306366443634033, "eval_loss": 0.6873763203620911, "eval_rewards/accuracies": 0.5218750238418579, "eval_rewards/chosen": -0.04051350802183151, "eval_rewards/margins": 0.018969083204865456, "eval_rewards/rejected": -0.05948259308934212, "eval_runtime": 103.4451, "eval_samples_per_second": 24.689, "eval_steps_per_second": 0.773, "step": 1100 }, { "epoch": 0.8515535097813579, "grad_norm": 1.5570868253707886, "learning_rate": 7.40598618572525e-06, "logits/chosen": -0.4041014611721039, "logits/rejected": -0.4015308916568756, "logps/chosen": -3.7765114307403564, "logps/rejected": -3.0345726013183594, "loss": 0.6972, "rewards/accuracies": 0.4749999940395355, "rewards/chosen": -0.04268242418766022, "rewards/margins": -0.022908544167876244, "rewards/rejected": -0.019773881882429123, "step": 1110 }, { "epoch": 0.8592251630226314, "grad_norm": 1.0964603424072266, "learning_rate": 7.022256331542594e-06, "logits/chosen": -0.4172094464302063, "logits/rejected": -0.41948825120925903, "logps/chosen": -3.422609329223633, "logps/rejected": -3.803391933441162, "loss": 0.6869, "rewards/accuracies": 0.42500001192092896, "rewards/chosen": -0.06300051510334015, "rewards/margins": -0.010609099641442299, "rewards/rejected": -0.0523914210498333, "step": 1120 }, { "epoch": 0.8668968162639049, "grad_norm": 1.906252145767212, "learning_rate": 6.638526477359939e-06, "logits/chosen": -0.44013237953186035, "logits/rejected": -0.4369295537471771, "logps/chosen": -3.6771011352539062, "logps/rejected": -3.834157943725586, "loss": 0.6724, "rewards/accuracies": 0.4749999940395355, "rewards/chosen": -0.033518821001052856, "rewards/margins": 0.04495871439576149, "rewards/rejected": -0.07847753912210464, "step": 1130 }, { "epoch": 0.8745684695051784, "grad_norm": 1.4220091104507446, "learning_rate": 6.254796623177284e-06, "logits/chosen": -0.5218509435653687, "logits/rejected": -0.527131199836731, "logps/chosen": -2.6902661323547363, "logps/rejected": -3.8175301551818848, "loss": 0.6918, "rewards/accuracies": 0.5249999761581421, "rewards/chosen": -0.05095089226961136, "rewards/margins": 0.008214332163333893, "rewards/rejected": -0.059165216982364655, "step": 1140 }, { "epoch": 0.8822401227464519, "grad_norm": 1.280017375946045, "learning_rate": 5.871066768994628e-06, "logits/chosen": -0.3222968280315399, "logits/rejected": -0.32637038826942444, "logps/chosen": -3.7984070777893066, "logps/rejected": -3.347282886505127, "loss": 0.6897, "rewards/accuracies": 0.5, "rewards/chosen": -0.050425935536623, "rewards/margins": 0.01578127220273018, "rewards/rejected": -0.06620721518993378, "step": 1150 }, { "epoch": 0.8822401227464519, "eval_logits/chosen": -0.25547298789024353, "eval_logits/rejected": -0.25304776430130005, "eval_logps/chosen": -3.5789942741394043, "eval_logps/rejected": -4.070747375488281, "eval_loss": 0.6874914169311523, "eval_rewards/accuracies": 0.5218750238418579, "eval_rewards/chosen": -0.05461201071739197, "eval_rewards/margins": 0.01888173446059227, "eval_rewards/rejected": -0.07349374145269394, "eval_runtime": 94.5303, "eval_samples_per_second": 27.018, "eval_steps_per_second": 0.846, "step": 1150 }, { "epoch": 0.8899117759877253, "grad_norm": 1.3330215215682983, "learning_rate": 5.487336914811972e-06, "logits/chosen": -0.4289553165435791, "logits/rejected": -0.42481836676597595, "logps/chosen": -4.021355628967285, "logps/rejected": -2.860283374786377, "loss": 0.6836, "rewards/accuracies": 0.675000011920929, "rewards/chosen": -0.00559669965878129, "rewards/margins": 0.03529663756489754, "rewards/rejected": -0.04089333862066269, "step": 1160 }, { "epoch": 0.8975834292289988, "grad_norm": 2.0986618995666504, "learning_rate": 5.103607060629317e-06, "logits/chosen": -0.4019424021244049, "logits/rejected": -0.39785686135292053, "logps/chosen": -2.8404293060302734, "logps/rejected": -5.334328651428223, "loss": 0.6977, "rewards/accuracies": 0.550000011920929, "rewards/chosen": -0.03890450671315193, "rewards/margins": 0.015505614690482616, "rewards/rejected": -0.05441011115908623, "step": 1170 }, { "epoch": 0.9052550824702723, "grad_norm": 1.6253297328948975, "learning_rate": 4.719877206446662e-06, "logits/chosen": -0.2909601032733917, "logits/rejected": -0.2908070385456085, "logps/chosen": -2.6695282459259033, "logps/rejected": -4.1958699226379395, "loss": 0.6873, "rewards/accuracies": 0.6499999761581421, "rewards/chosen": -0.02150137722492218, "rewards/margins": 0.053162820637226105, "rewards/rejected": -0.07466419786214828, "step": 1180 }, { "epoch": 0.9129267357115458, "grad_norm": 1.6267880201339722, "learning_rate": 4.336147352264006e-06, "logits/chosen": -0.49156007170677185, "logits/rejected": -0.49580636620521545, "logps/chosen": -2.781247854232788, "logps/rejected": -4.9970293045043945, "loss": 0.6777, "rewards/accuracies": 0.574999988079071, "rewards/chosen": -0.03273757919669151, "rewards/margins": 0.08275122940540314, "rewards/rejected": -0.11548881232738495, "step": 1190 }, { "epoch": 0.9205983889528193, "grad_norm": 1.53132164478302, "learning_rate": 3.952417498081351e-06, "logits/chosen": -0.4613899290561676, "logits/rejected": -0.45741963386535645, "logps/chosen": -3.1849541664123535, "logps/rejected": -4.181215763092041, "loss": 0.6808, "rewards/accuracies": 0.5249999761581421, "rewards/chosen": -0.048704423010349274, "rewards/margins": 0.04184052720665932, "rewards/rejected": -0.09054495394229889, "step": 1200 }, { "epoch": 0.9205983889528193, "eval_logits/chosen": -0.25466716289520264, "eval_logits/rejected": -0.2519235908985138, "eval_logps/chosen": -3.6220946311950684, "eval_logps/rejected": -4.108763694763184, "eval_loss": 0.6875774264335632, "eval_rewards/accuracies": 0.5062500238418579, "eval_rewards/chosen": -0.05892205238342285, "eval_rewards/margins": 0.018373344093561172, "eval_rewards/rejected": -0.07729540020227432, "eval_runtime": 87.859, "eval_samples_per_second": 29.069, "eval_steps_per_second": 0.911, "step": 1200 }, { "epoch": 0.9282700421940928, "grad_norm": 1.241278052330017, "learning_rate": 3.568687643898695e-06, "logits/chosen": -0.4009276032447815, "logits/rejected": -0.40318647027015686, "logps/chosen": -4.127172470092773, "logps/rejected": -3.866589069366455, "loss": 0.6829, "rewards/accuracies": 0.42500001192092896, "rewards/chosen": -0.09695043414831161, "rewards/margins": -0.046161942183971405, "rewards/rejected": -0.05078849941492081, "step": 1210 }, { "epoch": 0.9359416954353663, "grad_norm": 1.5091854333877563, "learning_rate": 3.18495778971604e-06, "logits/chosen": -0.4826277196407318, "logits/rejected": -0.478055864572525, "logps/chosen": -3.1896512508392334, "logps/rejected": -4.594376564025879, "loss": 0.6881, "rewards/accuracies": 0.574999988079071, "rewards/chosen": -0.04869696497917175, "rewards/margins": 0.07252007722854614, "rewards/rejected": -0.12121704965829849, "step": 1220 }, { "epoch": 0.9436133486766398, "grad_norm": 1.8513944149017334, "learning_rate": 2.8012279355333845e-06, "logits/chosen": -0.44737452268600464, "logits/rejected": -0.4507928788661957, "logps/chosen": -2.7748920917510986, "logps/rejected": -4.892573356628418, "loss": 0.6837, "rewards/accuracies": 0.675000011920929, "rewards/chosen": -0.0341179221868515, "rewards/margins": 0.06964664161205292, "rewards/rejected": -0.10376457124948502, "step": 1230 }, { "epoch": 0.9512850019179133, "grad_norm": 1.2888293266296387, "learning_rate": 2.417498081350729e-06, "logits/chosen": -0.42681989073753357, "logits/rejected": -0.4308759570121765, "logps/chosen": -3.6465072631835938, "logps/rejected": -4.087521553039551, "loss": 0.6922, "rewards/accuracies": 0.4749999940395355, "rewards/chosen": -0.08686716109514236, "rewards/margins": -0.01433576364070177, "rewards/rejected": -0.07253140956163406, "step": 1240 }, { "epoch": 0.9589566551591868, "grad_norm": 1.8460888862609863, "learning_rate": 2.0337682271680737e-06, "logits/chosen": -0.4428413510322571, "logits/rejected": -0.4398593008518219, "logps/chosen": -3.1560444831848145, "logps/rejected": -4.46377420425415, "loss": 0.6783, "rewards/accuracies": 0.550000011920929, "rewards/chosen": -0.02867904305458069, "rewards/margins": 0.08603222668170929, "rewards/rejected": -0.11471126228570938, "step": 1250 }, { "epoch": 0.9589566551591868, "eval_logits/chosen": -0.257295697927475, "eval_logits/rejected": -0.25472259521484375, "eval_logps/chosen": -3.5708394050598145, "eval_logps/rejected": -4.090200424194336, "eval_loss": 0.6876680850982666, "eval_rewards/accuracies": 0.5249999761581421, "eval_rewards/chosen": -0.053796518594026566, "eval_rewards/margins": 0.021642494946718216, "eval_rewards/rejected": -0.07543901354074478, "eval_runtime": 88.4028, "eval_samples_per_second": 28.89, "eval_steps_per_second": 0.905, "step": 1250 }, { "epoch": 0.9666283084004603, "grad_norm": 2.0357563495635986, "learning_rate": 1.6500383729854182e-06, "logits/chosen": -0.40101394057273865, "logits/rejected": -0.40456658601760864, "logps/chosen": -3.5583319664001465, "logps/rejected": -4.374687194824219, "loss": 0.6872, "rewards/accuracies": 0.550000011920929, "rewards/chosen": -0.056288547813892365, "rewards/margins": 0.033787138760089874, "rewards/rejected": -0.09007568657398224, "step": 1260 }, { "epoch": 0.9742999616417338, "grad_norm": 1.2816194295883179, "learning_rate": 1.266308518802763e-06, "logits/chosen": -0.45148950815200806, "logits/rejected": -0.4437050223350525, "logps/chosen": -4.794127464294434, "logps/rejected": -2.5109446048736572, "loss": 0.6883, "rewards/accuracies": 0.42500001192092896, "rewards/chosen": -0.08158718794584274, "rewards/margins": -0.058057576417922974, "rewards/rejected": -0.023529618978500366, "step": 1270 }, { "epoch": 0.9819716148830073, "grad_norm": 1.3781119585037231, "learning_rate": 8.825786646201075e-07, "logits/chosen": -0.37359991669654846, "logits/rejected": -0.3782901167869568, "logps/chosen": -3.121410369873047, "logps/rejected": -4.8245930671691895, "loss": 0.6868, "rewards/accuracies": 0.6499999761581421, "rewards/chosen": -0.03528415039181709, "rewards/margins": 0.05396044999361038, "rewards/rejected": -0.08924460411071777, "step": 1280 }, { "epoch": 0.9896432681242808, "grad_norm": 1.4439499378204346, "learning_rate": 4.988488104374521e-07, "logits/chosen": -0.4285728335380554, "logits/rejected": -0.43194809556007385, "logps/chosen": -4.162619113922119, "logps/rejected": -3.23869252204895, "loss": 0.687, "rewards/accuracies": 0.375, "rewards/chosen": -0.0661448985338211, "rewards/margins": -0.029600393027067184, "rewards/rejected": -0.036544498056173325, "step": 1290 }, { "epoch": 0.9973149213655543, "grad_norm": 1.6206567287445068, "learning_rate": 1.1511895625479663e-07, "logits/chosen": -0.4840824604034424, "logits/rejected": -0.4765002727508545, "logps/chosen": -3.34405255317688, "logps/rejected": -3.7568817138671875, "loss": 0.6942, "rewards/accuracies": 0.5249999761581421, "rewards/chosen": -0.05117567628622055, "rewards/margins": 0.012036425992846489, "rewards/rejected": -0.06321209669113159, "step": 1300 }, { "epoch": 0.9973149213655543, "eval_logits/chosen": -0.2575662434101105, "eval_logits/rejected": -0.2548866271972656, "eval_logps/chosen": -3.5700440406799316, "eval_logps/rejected": -4.087610721588135, "eval_loss": 0.6874552369117737, "eval_rewards/accuracies": 0.528124988079071, "eval_rewards/chosen": -0.05371702462434769, "eval_rewards/margins": 0.021463003009557724, "eval_rewards/rejected": -0.07518003135919571, "eval_runtime": 87.8745, "eval_samples_per_second": 29.064, "eval_steps_per_second": 0.91, "step": 1300 } ], "logging_steps": 10, "max_steps": 1303, "num_input_tokens_seen": 0, "num_train_epochs": 1, "save_steps": 50, "total_flos": 0.0, "train_batch_size": 2, "trial_name": null, "trial_params": null }