{ "best_metric": null, "best_model_checkpoint": null, "epoch": 3.0, "eval_steps": 50, "global_step": 414, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.036231884057971016, "grad_norm": 61.75757328159282, "learning_rate": 5e-07, "logits/chosen": -2.732090473175049, "logits/rejected": -2.7100460529327393, "logps/chosen": -182.59107971191406, "logps/rejected": -189.5584716796875, "loss": 0.6889, "rewards/accuracies": 0.30000001192092896, "rewards/chosen": -0.00281245238147676, "rewards/margins": 0.0058334446512162685, "rewards/rejected": -0.008645896799862385, "step": 5 }, { "epoch": 0.07246376811594203, "grad_norm": 44.951594498596215, "learning_rate": 1e-06, "logits/chosen": -2.754081964492798, "logits/rejected": -2.752152919769287, "logps/chosen": -197.337158203125, "logps/rejected": -184.00933837890625, "loss": 0.6274, "rewards/accuracies": 0.6625000238418579, "rewards/chosen": 0.03147688880562782, "rewards/margins": 0.1896156221628189, "rewards/rejected": -0.1581387221813202, "step": 10 }, { "epoch": 0.10869565217391304, "grad_norm": 51.34158391398985, "learning_rate": 9.996221126793764e-07, "logits/chosen": -2.694983959197998, "logits/rejected": -2.692361831665039, "logps/chosen": -203.20387268066406, "logps/rejected": -204.64244079589844, "loss": 0.5838, "rewards/accuracies": 0.75, "rewards/chosen": 0.6150370836257935, "rewards/margins": 0.9413955807685852, "rewards/rejected": -0.32635849714279175, "step": 15 }, { "epoch": 0.14492753623188406, "grad_norm": 34.76477183019994, "learning_rate": 9.984890219128145e-07, "logits/chosen": -2.612672805786133, "logits/rejected": -2.5829074382781982, "logps/chosen": -188.62716674804688, "logps/rejected": -192.87452697753906, "loss": 0.5142, "rewards/accuracies": 0.6937500238418579, "rewards/chosen": 0.8545471429824829, "rewards/margins": 1.280996561050415, "rewards/rejected": -0.4264492094516754, "step": 20 }, { "epoch": 0.18115942028985507, "grad_norm": 36.75278346647978, "learning_rate": 9.966024404228493e-07, "logits/chosen": -2.450106143951416, "logits/rejected": -2.4297895431518555, "logps/chosen": -179.98348999023438, "logps/rejected": -179.38925170898438, "loss": 0.5032, "rewards/accuracies": 0.7562500238418579, "rewards/chosen": 0.38695499300956726, "rewards/margins": 0.8900691866874695, "rewards/rejected": -0.5031141638755798, "step": 25 }, { "epoch": 0.21739130434782608, "grad_norm": 31.781918105397544, "learning_rate": 9.939652198703783e-07, "logits/chosen": -2.324214458465576, "logits/rejected": -2.325657367706299, "logps/chosen": -188.5428466796875, "logps/rejected": -193.8271942138672, "loss": 0.4995, "rewards/accuracies": 0.762499988079071, "rewards/chosen": 0.6558719873428345, "rewards/margins": 1.2207121849060059, "rewards/rejected": -0.5648401975631714, "step": 30 }, { "epoch": 0.2536231884057971, "grad_norm": 39.36776247005876, "learning_rate": 9.905813465442354e-07, "logits/chosen": -2.236240863800049, "logits/rejected": -2.2105681896209717, "logps/chosen": -203.98277282714844, "logps/rejected": -194.84640502929688, "loss": 0.5091, "rewards/accuracies": 0.7124999761581421, "rewards/chosen": 0.8834564089775085, "rewards/margins": 1.2675695419311523, "rewards/rejected": -0.3841131329536438, "step": 35 }, { "epoch": 0.2898550724637681, "grad_norm": 30.817630358317576, "learning_rate": 9.864559353357187e-07, "logits/chosen": -2.068774700164795, "logits/rejected": -2.0603950023651123, "logps/chosen": -182.76817321777344, "logps/rejected": -185.9797821044922, "loss": 0.4873, "rewards/accuracies": 0.78125, "rewards/chosen": 1.03325617313385, "rewards/margins": 1.0384714603424072, "rewards/rejected": -0.005215352866798639, "step": 40 }, { "epoch": 0.32608695652173914, "grad_norm": 29.09268118121073, "learning_rate": 9.815952220071804e-07, "logits/chosen": -1.8718488216400146, "logits/rejected": -1.8250553607940674, "logps/chosen": -195.60968017578125, "logps/rejected": -221.5565643310547, "loss": 0.4597, "rewards/accuracies": 0.78125, "rewards/chosen": 1.3850222826004028, "rewards/margins": 1.8469291925430298, "rewards/rejected": -0.4619070589542389, "step": 45 }, { "epoch": 0.36231884057971014, "grad_norm": 29.526743630011346, "learning_rate": 9.76006553766365e-07, "logits/chosen": -1.653713583946228, "logits/rejected": -1.6171553134918213, "logps/chosen": -198.85989379882812, "logps/rejected": -203.60678100585938, "loss": 0.4516, "rewards/accuracies": 0.793749988079071, "rewards/chosen": 0.8292047381401062, "rewards/margins": 1.6851797103881836, "rewards/rejected": -0.8559748530387878, "step": 50 }, { "epoch": 0.36231884057971014, "eval_logits/chosen": -1.7065542936325073, "eval_logits/rejected": -1.630993127822876, "eval_logps/chosen": -192.20655822753906, "eval_logps/rejected": -206.51295471191406, "eval_loss": 0.4420754015445709, "eval_rewards/accuracies": 0.7903226017951965, "eval_rewards/chosen": 0.8112886548042297, "eval_rewards/margins": 1.641775369644165, "eval_rewards/rejected": -0.8304866552352905, "eval_runtime": 247.7543, "eval_samples_per_second": 15.83, "eval_steps_per_second": 0.25, "step": 50 }, { "epoch": 0.39855072463768115, "grad_norm": 30.94859785748943, "learning_rate": 9.696983781607415e-07, "logits/chosen": -1.7253024578094482, "logits/rejected": -1.6905288696289062, "logps/chosen": -182.9173126220703, "logps/rejected": -171.9159698486328, "loss": 0.4573, "rewards/accuracies": 0.793749988079071, "rewards/chosen": 0.6172864437103271, "rewards/margins": 1.648385763168335, "rewards/rejected": -1.031099557876587, "step": 55 }, { "epoch": 0.43478260869565216, "grad_norm": 40.75469044830845, "learning_rate": 9.626802303086209e-07, "logits/chosen": -1.87893807888031, "logits/rejected": -1.8299003839492798, "logps/chosen": -186.30145263671875, "logps/rejected": -193.9145965576172, "loss": 0.4264, "rewards/accuracies": 0.8374999761581421, "rewards/chosen": 0.4657188057899475, "rewards/margins": 1.7288262844085693, "rewards/rejected": -1.2631075382232666, "step": 60 }, { "epoch": 0.47101449275362317, "grad_norm": 35.556274541495966, "learning_rate": 9.549627184863528e-07, "logits/chosen": -2.016784906387329, "logits/rejected": -1.9150521755218506, "logps/chosen": -191.3840789794922, "logps/rejected": -192.66639709472656, "loss": 0.4289, "rewards/accuracies": 0.793749988079071, "rewards/chosen": 0.0974711924791336, "rewards/margins": 1.6010549068450928, "rewards/rejected": -1.5035837888717651, "step": 65 }, { "epoch": 0.5072463768115942, "grad_norm": 26.46585227154451, "learning_rate": 9.465575080933957e-07, "logits/chosen": -1.853308916091919, "logits/rejected": -1.7947351932525635, "logps/chosen": -172.3099822998047, "logps/rejected": -208.057373046875, "loss": 0.3948, "rewards/accuracies": 0.824999988079071, "rewards/chosen": 0.21238946914672852, "rewards/margins": 1.8403332233428955, "rewards/rejected": -1.627943754196167, "step": 70 }, { "epoch": 0.5434782608695652, "grad_norm": 31.533541728553253, "learning_rate": 9.374773040194878e-07, "logits/chosen": -1.8850362300872803, "logits/rejected": -1.8103622198104858, "logps/chosen": -205.5053253173828, "logps/rejected": -210.96981811523438, "loss": 0.4364, "rewards/accuracies": 0.8187500238418579, "rewards/chosen": 0.17310531437397003, "rewards/margins": 1.8103282451629639, "rewards/rejected": -1.6372228860855103, "step": 75 }, { "epoch": 0.5797101449275363, "grad_norm": 29.780905727815526, "learning_rate": 9.277358314405818e-07, "logits/chosen": -1.7906593084335327, "logits/rejected": -1.742597222328186, "logps/chosen": -188.9757080078125, "logps/rejected": -205.398193359375, "loss": 0.3987, "rewards/accuracies": 0.800000011920929, "rewards/chosen": -0.19728976488113403, "rewards/margins": 1.879663109779358, "rewards/rejected": -2.0769529342651367, "step": 80 }, { "epoch": 0.6159420289855072, "grad_norm": 34.4646468352745, "learning_rate": 9.173478150725651e-07, "logits/chosen": -1.7377640008926392, "logits/rejected": -1.6257518529891968, "logps/chosen": -210.00320434570312, "logps/rejected": -215.84835815429688, "loss": 0.4258, "rewards/accuracies": 0.824999988079071, "rewards/chosen": 0.08612661063671112, "rewards/margins": 2.4435980319976807, "rewards/rejected": -2.357471227645874, "step": 85 }, { "epoch": 0.6521739130434783, "grad_norm": 29.12537980218493, "learning_rate": 9.063289569141251e-07, "logits/chosen": -1.7976572513580322, "logits/rejected": -1.739854097366333, "logps/chosen": -214.8435821533203, "logps/rejected": -224.52005004882812, "loss": 0.4147, "rewards/accuracies": 0.84375, "rewards/chosen": 0.46363013982772827, "rewards/margins": 2.330965518951416, "rewards/rejected": -1.867335557937622, "step": 90 }, { "epoch": 0.6884057971014492, "grad_norm": 35.00421638148543, "learning_rate": 8.946959125124051e-07, "logits/chosen": -1.861108422279358, "logits/rejected": -1.780923843383789, "logps/chosen": -207.5733184814453, "logps/rejected": -193.34400939941406, "loss": 0.4121, "rewards/accuracies": 0.8125, "rewards/chosen": 0.4294491708278656, "rewards/margins": 2.142913341522217, "rewards/rejected": -1.7134641408920288, "step": 95 }, { "epoch": 0.7246376811594203, "grad_norm": 31.611698501726103, "learning_rate": 8.824662657873238e-07, "logits/chosen": -1.8221423625946045, "logits/rejected": -1.802095651626587, "logps/chosen": -173.2090301513672, "logps/rejected": -206.5529327392578, "loss": 0.3759, "rewards/accuracies": 0.793749988079071, "rewards/chosen": -0.04721298813819885, "rewards/margins": 1.9821780920028687, "rewards/rejected": -2.029391050338745, "step": 100 }, { "epoch": 0.7246376811594203, "eval_logits/chosen": -1.8523844480514526, "eval_logits/rejected": -1.7929590940475464, "eval_logps/chosen": -200.7910614013672, "eval_logps/rejected": -220.96961975097656, "eval_loss": 0.4121003746986389, "eval_rewards/accuracies": 0.8145161271095276, "eval_rewards/chosen": -0.047160252928733826, "eval_rewards/margins": 2.2289960384368896, "eval_rewards/rejected": -2.276156187057495, "eval_runtime": 247.371, "eval_samples_per_second": 15.855, "eval_steps_per_second": 0.251, "step": 100 }, { "epoch": 0.7608695652173914, "grad_norm": 30.01063089972391, "learning_rate": 8.696585024526135e-07, "logits/chosen": -1.7823431491851807, "logits/rejected": -1.7234203815460205, "logps/chosen": -189.0630340576172, "logps/rejected": -224.55642700195312, "loss": 0.3969, "rewards/accuracies": 0.824999988079071, "rewards/chosen": -0.0058825016021728516, "rewards/margins": 2.5169529914855957, "rewards/rejected": -2.5228357315063477, "step": 105 }, { "epoch": 0.7971014492753623, "grad_norm": 35.65348267869082, "learning_rate": 8.562919820737535e-07, "logits/chosen": -1.7099103927612305, "logits/rejected": -1.6304385662078857, "logps/chosen": -206.9807586669922, "logps/rejected": -209.36962890625, "loss": 0.3755, "rewards/accuracies": 0.84375, "rewards/chosen": -0.26569992303848267, "rewards/margins": 2.464618444442749, "rewards/rejected": -2.730318546295166, "step": 110 }, { "epoch": 0.8333333333333334, "grad_norm": 28.250647507886438, "learning_rate": 8.423869088050315e-07, "logits/chosen": -1.7219148874282837, "logits/rejected": -1.677403450012207, "logps/chosen": -195.88735961914062, "logps/rejected": -222.36581420898438, "loss": 0.3912, "rewards/accuracies": 0.8187500238418579, "rewards/chosen": -0.14722302556037903, "rewards/margins": 2.4208686351776123, "rewards/rejected": -2.568091630935669, "step": 115 }, { "epoch": 0.8695652173913043, "grad_norm": 36.27157250663838, "learning_rate": 8.2796430084997e-07, "logits/chosen": -1.6080610752105713, "logits/rejected": -1.521059513092041, "logps/chosen": -197.2279510498047, "logps/rejected": -208.6706085205078, "loss": 0.3668, "rewards/accuracies": 0.8812500238418579, "rewards/chosen": 0.07672278583049774, "rewards/margins": 2.720585584640503, "rewards/rejected": -2.64386248588562, "step": 120 }, { "epoch": 0.9057971014492754, "grad_norm": 28.694980241284195, "learning_rate": 8.130459586912753e-07, "logits/chosen": -1.4262475967407227, "logits/rejected": -1.3733441829681396, "logps/chosen": -219.4936981201172, "logps/rejected": -217.61599731445312, "loss": 0.4582, "rewards/accuracies": 0.800000011920929, "rewards/chosen": -0.8654589653015137, "rewards/margins": 1.6804126501083374, "rewards/rejected": -2.5458714962005615, "step": 125 }, { "epoch": 0.9420289855072463, "grad_norm": 29.710262188798424, "learning_rate": 7.97654432138333e-07, "logits/chosen": -1.4053936004638672, "logits/rejected": -1.336163878440857, "logps/chosen": -210.55026245117188, "logps/rejected": -243.9113311767578, "loss": 0.3921, "rewards/accuracies": 0.862500011920929, "rewards/chosen": 0.20443829894065857, "rewards/margins": 2.7204947471618652, "rewards/rejected": -2.516056537628174, "step": 130 }, { "epoch": 0.9782608695652174, "grad_norm": 26.71701106117664, "learning_rate": 7.81812986242061e-07, "logits/chosen": -1.423004388809204, "logits/rejected": -1.2980186939239502, "logps/chosen": -193.02523803710938, "logps/rejected": -232.86788940429688, "loss": 0.3631, "rewards/accuracies": 0.8500000238418579, "rewards/chosen": 0.15479817986488342, "rewards/margins": 3.0325751304626465, "rewards/rejected": -2.877776622772217, "step": 135 }, { "epoch": 1.0144927536231885, "grad_norm": 18.847111481815627, "learning_rate": 7.655455661286375e-07, "logits/chosen": -1.3630199432373047, "logits/rejected": -1.3213447332382202, "logps/chosen": -193.20803833007812, "logps/rejected": -237.5965118408203, "loss": 0.2543, "rewards/accuracies": 0.8999999761581421, "rewards/chosen": -0.20990173518657684, "rewards/margins": 3.6189141273498535, "rewards/rejected": -3.8288159370422363, "step": 140 }, { "epoch": 1.0507246376811594, "grad_norm": 18.388157966842616, "learning_rate": 7.488767608052628e-07, "logits/chosen": -1.543648362159729, "logits/rejected": -1.399395227432251, "logps/chosen": -190.61196899414062, "logps/rejected": -237.07424926757812, "loss": 0.1744, "rewards/accuracies": 0.9312499761581421, "rewards/chosen": 0.6787484884262085, "rewards/margins": 4.023434638977051, "rewards/rejected": -3.3446857929229736, "step": 145 }, { "epoch": 1.0869565217391304, "grad_norm": 15.923928842240379, "learning_rate": 7.318317659926636e-07, "logits/chosen": -1.6209495067596436, "logits/rejected": -1.5568897724151611, "logps/chosen": -172.939697265625, "logps/rejected": -233.3376007080078, "loss": 0.149, "rewards/accuracies": 0.9312499761581421, "rewards/chosen": 1.1221749782562256, "rewards/margins": 3.8970863819122314, "rewards/rejected": -2.774911403656006, "step": 150 }, { "epoch": 1.0869565217391304, "eval_logits/chosen": -1.7425010204315186, "eval_logits/rejected": -1.674597978591919, "eval_logps/chosen": -194.48468017578125, "eval_logps/rejected": -217.0243682861328, "eval_loss": 0.42049291729927063, "eval_rewards/accuracies": 0.8205645084381104, "eval_rewards/chosen": 0.5834774374961853, "eval_rewards/margins": 2.4651052951812744, "eval_rewards/rejected": -1.881628155708313, "eval_runtime": 247.5785, "eval_samples_per_second": 15.841, "eval_steps_per_second": 0.25, "step": 150 }, { "epoch": 1.1231884057971016, "grad_norm": 14.18220461970911, "learning_rate": 7.144363460405189e-07, "logits/chosen": -1.7796205282211304, "logits/rejected": -1.6700912714004517, "logps/chosen": -190.59030151367188, "logps/rejected": -233.08151245117188, "loss": 0.1482, "rewards/accuracies": 0.96875, "rewards/chosen": 1.356343150138855, "rewards/margins": 4.483328342437744, "rewards/rejected": -3.1269848346710205, "step": 155 }, { "epoch": 1.1594202898550725, "grad_norm": 12.199643576270322, "learning_rate": 6.967167949833762e-07, "logits/chosen": -1.7053067684173584, "logits/rejected": -1.613364577293396, "logps/chosen": -192.91790771484375, "logps/rejected": -245.5927734375, "loss": 0.143, "rewards/accuracies": 0.9624999761581421, "rewards/chosen": 0.4924169182777405, "rewards/margins": 4.924551963806152, "rewards/rejected": -4.432135105133057, "step": 160 }, { "epoch": 1.1956521739130435, "grad_norm": 16.84620648534237, "learning_rate": 6.786998967959219e-07, "logits/chosen": -1.649950385093689, "logits/rejected": -1.558600664138794, "logps/chosen": -199.79678344726562, "logps/rejected": -227.9362030029297, "loss": 0.1491, "rewards/accuracies": 0.9375, "rewards/chosen": 0.5148376226425171, "rewards/margins": 4.6056809425354, "rewards/rejected": -4.090843200683594, "step": 165 }, { "epoch": 1.2318840579710144, "grad_norm": 16.743277828937153, "learning_rate": 6.604128849076838e-07, "logits/chosen": -1.687930703163147, "logits/rejected": -1.5980262756347656, "logps/chosen": -199.6280517578125, "logps/rejected": -237.2197265625, "loss": 0.1514, "rewards/accuracies": 0.9437500238418579, "rewards/chosen": 1.0777654647827148, "rewards/margins": 4.6541428565979, "rewards/rejected": -3.5763778686523438, "step": 170 }, { "epoch": 1.2681159420289856, "grad_norm": 13.4419999910089, "learning_rate": 6.418834010383609e-07, "logits/chosen": -1.7620418071746826, "logits/rejected": -1.6492313146591187, "logps/chosen": -170.84674072265625, "logps/rejected": -228.17239379882812, "loss": 0.1461, "rewards/accuracies": 0.9375, "rewards/chosen": 0.7329138517379761, "rewards/margins": 4.496224403381348, "rewards/rejected": -3.763310670852661, "step": 175 }, { "epoch": 1.3043478260869565, "grad_norm": 15.060085944373489, "learning_rate": 6.231394534160007e-07, "logits/chosen": -1.8257992267608643, "logits/rejected": -1.7924093008041382, "logps/chosen": -183.6071319580078, "logps/rejected": -224.40194702148438, "loss": 0.142, "rewards/accuracies": 0.96875, "rewards/chosen": 1.3987538814544678, "rewards/margins": 4.50449275970459, "rewards/rejected": -3.105739116668701, "step": 180 }, { "epoch": 1.3405797101449275, "grad_norm": 15.268514575865197, "learning_rate": 6.042093744411828e-07, "logits/chosen": -1.853198766708374, "logits/rejected": -1.799068808555603, "logps/chosen": -184.3455047607422, "logps/rejected": -228.256591796875, "loss": 0.1444, "rewards/accuracies": 0.9437500238418579, "rewards/chosen": 1.1724538803100586, "rewards/margins": 4.359854698181152, "rewards/rejected": -3.1874005794525146, "step": 185 }, { "epoch": 1.3768115942028984, "grad_norm": 14.506206484787588, "learning_rate": 5.851217778611993e-07, "logits/chosen": -1.8662179708480835, "logits/rejected": -1.8571313619613647, "logps/chosen": -198.0624542236328, "logps/rejected": -219.442626953125, "loss": 0.1349, "rewards/accuracies": 0.925000011920929, "rewards/chosen": 0.6407368779182434, "rewards/margins": 4.615514278411865, "rewards/rejected": -3.974777936935425, "step": 190 }, { "epoch": 1.4130434782608696, "grad_norm": 20.861341598868098, "learning_rate": 5.659055155189651e-07, "logits/chosen": -1.9783111810684204, "logits/rejected": -1.8647491931915283, "logps/chosen": -189.13699340820312, "logps/rejected": -227.8821563720703, "loss": 0.1536, "rewards/accuracies": 0.9437500238418579, "rewards/chosen": 0.45683974027633667, "rewards/margins": 4.89407205581665, "rewards/rejected": -4.437232971191406, "step": 195 }, { "epoch": 1.4492753623188406, "grad_norm": 12.518481247292005, "learning_rate": 5.465896337420358e-07, "logits/chosen": -1.964616060256958, "logits/rejected": -1.8386001586914062, "logps/chosen": -203.31442260742188, "logps/rejected": -265.9437561035156, "loss": 0.1474, "rewards/accuracies": 0.9437500238418579, "rewards/chosen": 0.7917453050613403, "rewards/margins": 5.376971244812012, "rewards/rejected": -4.585226535797119, "step": 200 }, { "epoch": 1.4492753623188406, "eval_logits/chosen": -1.8598568439483643, "eval_logits/rejected": -1.7946782112121582, "eval_logps/chosen": -205.7305908203125, "eval_logps/rejected": -235.58175659179688, "eval_loss": 0.42740947008132935, "eval_rewards/accuracies": 0.8306451439857483, "eval_rewards/chosen": -0.5411156415939331, "eval_rewards/margins": 3.1962532997131348, "eval_rewards/rejected": -3.737368583679199, "eval_runtime": 247.3295, "eval_samples_per_second": 15.857, "eval_steps_per_second": 0.251, "step": 200 }, { "epoch": 1.4855072463768115, "grad_norm": 21.916596761757884, "learning_rate": 5.272033294376521e-07, "logits/chosen": -1.813153862953186, "logits/rejected": -1.768066644668579, "logps/chosen": -194.30654907226562, "logps/rejected": -225.20693969726562, "loss": 0.1401, "rewards/accuracies": 0.9375, "rewards/chosen": 0.010840868577361107, "rewards/margins": 4.742875099182129, "rewards/rejected": -4.732035160064697, "step": 205 }, { "epoch": 1.5217391304347827, "grad_norm": 15.855755657784645, "learning_rate": 5.077759059601755e-07, "logits/chosen": -1.7765309810638428, "logits/rejected": -1.7269878387451172, "logps/chosen": -208.30252075195312, "logps/rejected": -228.03353881835938, "loss": 0.1563, "rewards/accuracies": 0.956250011920929, "rewards/chosen": 0.508411705493927, "rewards/margins": 5.39729118347168, "rewards/rejected": -4.888879299163818, "step": 210 }, { "epoch": 1.5579710144927537, "grad_norm": 16.97962064526418, "learning_rate": 4.883367288176238e-07, "logits/chosen": -1.7922019958496094, "logits/rejected": -1.8073742389678955, "logps/chosen": -179.53660583496094, "logps/rejected": -229.8582763671875, "loss": 0.1482, "rewards/accuracies": 0.9437500238418579, "rewards/chosen": 0.9332772493362427, "rewards/margins": 5.133488178253174, "rewards/rejected": -4.2002105712890625, "step": 215 }, { "epoch": 1.5942028985507246, "grad_norm": 18.956232551415866, "learning_rate": 4.6891518128425974e-07, "logits/chosen": -1.9554294347763062, "logits/rejected": -1.8807146549224854, "logps/chosen": -198.28443908691406, "logps/rejected": -239.2546844482422, "loss": 0.1527, "rewards/accuracies": 0.9312499761581421, "rewards/chosen": 1.2289119958877563, "rewards/margins": 5.14874792098999, "rewards/rejected": -3.9198365211486816, "step": 220 }, { "epoch": 1.6304347826086958, "grad_norm": 18.38495079748664, "learning_rate": 4.495406199863217e-07, "logits/chosen": -1.990740418434143, "logits/rejected": -1.9738916158676147, "logps/chosen": -177.74075317382812, "logps/rejected": -256.86663818359375, "loss": 0.1242, "rewards/accuracies": 0.9437500238418579, "rewards/chosen": 1.0406591892242432, "rewards/margins": 5.407895088195801, "rewards/rejected": -4.367236614227295, "step": 225 }, { "epoch": 1.6666666666666665, "grad_norm": 21.41394990672278, "learning_rate": 4.302423305280385e-07, "logits/chosen": -2.0460195541381836, "logits/rejected": -1.9780528545379639, "logps/chosen": -177.8385009765625, "logps/rejected": -264.10272216796875, "loss": 0.1277, "rewards/accuracies": 0.949999988079071, "rewards/chosen": 0.6652169227600098, "rewards/margins": 5.778229713439941, "rewards/rejected": -5.113012313842773, "step": 230 }, { "epoch": 1.7028985507246377, "grad_norm": 18.775710144238932, "learning_rate": 4.1104948322499386e-07, "logits/chosen": -2.060439109802246, "logits/rejected": -1.9934184551239014, "logps/chosen": -184.96261596679688, "logps/rejected": -243.159423828125, "loss": 0.1271, "rewards/accuracies": 0.9437500238418579, "rewards/chosen": 0.6158139109611511, "rewards/margins": 5.492893218994141, "rewards/rejected": -4.877079010009766, "step": 235 }, { "epoch": 1.7391304347826086, "grad_norm": 17.40854531155315, "learning_rate": 3.919910890117584e-07, "logits/chosen": -2.0762391090393066, "logits/rejected": -2.0377309322357178, "logps/chosen": -180.35549926757812, "logps/rejected": -232.47412109375, "loss": 0.1345, "rewards/accuracies": 0.9312499761581421, "rewards/chosen": 0.7253020405769348, "rewards/margins": 5.309723854064941, "rewards/rejected": -4.584421634674072, "step": 240 }, { "epoch": 1.7753623188405796, "grad_norm": 16.818208369109737, "learning_rate": 3.7309595559042973e-07, "logits/chosen": -2.0712027549743652, "logits/rejected": -2.053870916366577, "logps/chosen": -186.7015380859375, "logps/rejected": -233.06875610351562, "loss": 0.1329, "rewards/accuracies": 0.9375, "rewards/chosen": 0.42518267035484314, "rewards/margins": 5.10868501663208, "rewards/rejected": -4.683502674102783, "step": 245 }, { "epoch": 1.8115942028985508, "grad_norm": 17.857229947247134, "learning_rate": 3.54392643886374e-07, "logits/chosen": -2.1355109214782715, "logits/rejected": -2.082988739013672, "logps/chosen": -187.5356903076172, "logps/rejected": -232.5768585205078, "loss": 0.1268, "rewards/accuracies": 0.956250011920929, "rewards/chosen": 0.629612147808075, "rewards/margins": 5.511849403381348, "rewards/rejected": -4.882236957550049, "step": 250 }, { "epoch": 1.8115942028985508, "eval_logits/chosen": -2.1449685096740723, "eval_logits/rejected": -2.0992848873138428, "eval_logps/chosen": -200.98956298828125, "eval_logps/rejected": -231.31536865234375, "eval_loss": 0.4333266615867615, "eval_rewards/accuracies": 0.8205645084381104, "eval_rewards/chosen": -0.06701094657182693, "eval_rewards/margins": 3.2437193393707275, "eval_rewards/rejected": -3.310730457305908, "eval_runtime": 247.4867, "eval_samples_per_second": 15.847, "eval_steps_per_second": 0.251, "step": 250 }, { "epoch": 1.8478260869565217, "grad_norm": 19.685959893776015, "learning_rate": 3.3590942487697765e-07, "logits/chosen": -2.168308734893799, "logits/rejected": -2.149319648742676, "logps/chosen": -185.67611694335938, "logps/rejected": -222.47714233398438, "loss": 0.1592, "rewards/accuracies": 0.9375, "rewards/chosen": 0.9020439386367798, "rewards/margins": 4.887114524841309, "rewards/rejected": -3.9850707054138184, "step": 255 }, { "epoch": 1.8840579710144927, "grad_norm": 17.22446921219621, "learning_rate": 3.176742368586725e-07, "logits/chosen": -2.170022964477539, "logits/rejected": -2.1088662147521973, "logps/chosen": -195.13963317871094, "logps/rejected": -216.1388702392578, "loss": 0.1331, "rewards/accuracies": 0.9312499761581421, "rewards/chosen": 0.9453809857368469, "rewards/margins": 5.353398323059082, "rewards/rejected": -4.408017158508301, "step": 260 }, { "epoch": 1.9202898550724639, "grad_norm": 15.391628097816579, "learning_rate": 2.997146432168236e-07, "logits/chosen": -2.1484408378601074, "logits/rejected": -2.0970280170440674, "logps/chosen": -191.478271484375, "logps/rejected": -245.2235870361328, "loss": 0.1396, "rewards/accuracies": 0.918749988079071, "rewards/chosen": 0.7697240114212036, "rewards/margins": 6.003005027770996, "rewards/rejected": -5.233281135559082, "step": 265 }, { "epoch": 1.9565217391304348, "grad_norm": 21.110077363875014, "learning_rate": 2.8205779076231446e-07, "logits/chosen": -2.118835926055908, "logits/rejected": -2.0954811573028564, "logps/chosen": -189.69805908203125, "logps/rejected": -235.75991821289062, "loss": 0.1295, "rewards/accuracies": 0.9437500238418579, "rewards/chosen": 1.1907222270965576, "rewards/margins": 5.428012847900391, "rewards/rejected": -4.237290382385254, "step": 270 }, { "epoch": 1.9927536231884058, "grad_norm": 20.754559500185525, "learning_rate": 2.647303686978035e-07, "logits/chosen": -2.063872814178467, "logits/rejected": -2.017089366912842, "logps/chosen": -182.68482971191406, "logps/rejected": -217.0849151611328, "loss": 0.1488, "rewards/accuracies": 0.9375, "rewards/chosen": 0.8351422548294067, "rewards/margins": 4.959425926208496, "rewards/rejected": -4.124283790588379, "step": 275 }, { "epoch": 2.028985507246377, "grad_norm": 9.164991188020005, "learning_rate": 2.4775856827568014e-07, "logits/chosen": -2.0497758388519287, "logits/rejected": -2.0113561153411865, "logps/chosen": -185.25265502929688, "logps/rejected": -223.13851928710938, "loss": 0.0883, "rewards/accuracies": 0.956250011920929, "rewards/chosen": 1.0435569286346436, "rewards/margins": 5.537060737609863, "rewards/rejected": -4.493502616882324, "step": 280 }, { "epoch": 2.0652173913043477, "grad_norm": 8.010646253617603, "learning_rate": 2.3116804320869464e-07, "logits/chosen": -2.0255563259124756, "logits/rejected": -1.9762938022613525, "logps/chosen": -188.28140258789062, "logps/rejected": -224.3779296875, "loss": 0.0703, "rewards/accuracies": 0.96875, "rewards/chosen": 0.6650876402854919, "rewards/margins": 5.7544169425964355, "rewards/rejected": -5.089330196380615, "step": 285 }, { "epoch": 2.101449275362319, "grad_norm": 8.650482101201316, "learning_rate": 2.1498387089310865e-07, "logits/chosen": -2.0082168579101562, "logits/rejected": -2.001406192779541, "logps/chosen": -194.86776733398438, "logps/rejected": -250.7891082763672, "loss": 0.0694, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": 1.0825541019439697, "rewards/margins": 6.063149452209473, "rewards/rejected": -4.980595588684082, "step": 290 }, { "epoch": 2.13768115942029, "grad_norm": 9.639312957872246, "learning_rate": 1.9923051450297336e-07, "logits/chosen": -2.034083843231201, "logits/rejected": -1.9722801446914673, "logps/chosen": -193.46327209472656, "logps/rejected": -229.68417358398438, "loss": 0.0724, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": 1.1218969821929932, "rewards/margins": 6.115258693695068, "rewards/rejected": -4.993361949920654, "step": 295 }, { "epoch": 2.1739130434782608, "grad_norm": 11.415587223465, "learning_rate": 1.839317860128368e-07, "logits/chosen": -2.004582166671753, "logits/rejected": -1.9546029567718506, "logps/chosen": -191.2842559814453, "logps/rejected": -250.6446075439453, "loss": 0.064, "rewards/accuracies": 0.9937499761581421, "rewards/chosen": 0.9089698791503906, "rewards/margins": 6.477193355560303, "rewards/rejected": -5.568222999572754, "step": 300 }, { "epoch": 2.1739130434782608, "eval_logits/chosen": -1.9908860921859741, "eval_logits/rejected": -1.9326670169830322, "eval_logps/chosen": -205.4859619140625, "eval_logps/rejected": -239.16648864746094, "eval_loss": 0.4331871271133423, "eval_rewards/accuracies": 0.8306451439857483, "eval_rewards/chosen": -0.5166527032852173, "eval_rewards/margins": 3.5791897773742676, "eval_rewards/rejected": -4.095842361450195, "eval_runtime": 247.4217, "eval_samples_per_second": 15.851, "eval_steps_per_second": 0.251, "step": 300 }, { "epoch": 2.210144927536232, "grad_norm": 13.069163886945839, "learning_rate": 1.6911081020477176e-07, "logits/chosen": -1.9845168590545654, "logits/rejected": -1.9513275623321533, "logps/chosen": -188.0749969482422, "logps/rejected": -253.936279296875, "loss": 0.0617, "rewards/accuracies": 0.96875, "rewards/chosen": 0.8729268908500671, "rewards/margins": 6.1785569190979, "rewards/rejected": -5.305630683898926, "step": 305 }, { "epoch": 2.246376811594203, "grad_norm": 12.409574297740958, "learning_rate": 1.5478998971412666e-07, "logits/chosen": -1.9543142318725586, "logits/rejected": -1.9092079401016235, "logps/chosen": -190.325439453125, "logps/rejected": -254.7587432861328, "loss": 0.0669, "rewards/accuracies": 0.981249988079071, "rewards/chosen": 0.8444012403488159, "rewards/margins": 6.521615028381348, "rewards/rejected": -5.677213668823242, "step": 310 }, { "epoch": 2.282608695652174, "grad_norm": 17.672030401926676, "learning_rate": 1.4099097116683873e-07, "logits/chosen": -1.996664047241211, "logits/rejected": -1.9541898965835571, "logps/chosen": -210.48422241210938, "logps/rejected": -274.6320495605469, "loss": 0.0611, "rewards/accuracies": 0.987500011920929, "rewards/chosen": 1.3748170137405396, "rewards/margins": 7.120137691497803, "rewards/rejected": -5.7453203201293945, "step": 315 }, { "epoch": 2.318840579710145, "grad_norm": 11.067029428009443, "learning_rate": 1.2773461245949247e-07, "logits/chosen": -1.9853408336639404, "logits/rejected": -1.930381417274475, "logps/chosen": -203.17752075195312, "logps/rejected": -247.21945190429688, "loss": 0.0587, "rewards/accuracies": 0.981249988079071, "rewards/chosen": 0.9430925250053406, "rewards/margins": 6.5651397705078125, "rewards/rejected": -5.622047424316406, "step": 320 }, { "epoch": 2.355072463768116, "grad_norm": 11.355849520152098, "learning_rate": 1.1504095123158014e-07, "logits/chosen": -1.9925590753555298, "logits/rejected": -1.9830715656280518, "logps/chosen": -195.1802520751953, "logps/rejected": -257.0408020019531, "loss": 0.0646, "rewards/accuracies": 0.987500011920929, "rewards/chosen": 0.7049287557601929, "rewards/margins": 6.576409339904785, "rewards/rejected": -5.871480464935303, "step": 325 }, { "epoch": 2.391304347826087, "grad_norm": 10.31665774657082, "learning_rate": 1.0292917457762323e-07, "logits/chosen": -1.981650948524475, "logits/rejected": -1.9037120342254639, "logps/chosen": -188.784912109375, "logps/rejected": -247.7895965576172, "loss": 0.0585, "rewards/accuracies": 0.981249988079071, "rewards/chosen": 0.8972769975662231, "rewards/margins": 6.589321136474609, "rewards/rejected": -5.692043781280518, "step": 330 }, { "epoch": 2.427536231884058, "grad_norm": 15.155495148000533, "learning_rate": 9.141759004493282e-08, "logits/chosen": -1.9452364444732666, "logits/rejected": -1.9261302947998047, "logps/chosen": -179.8898162841797, "logps/rejected": -243.61328125, "loss": 0.055, "rewards/accuracies": 0.9937499761581421, "rewards/chosen": 0.39632946252822876, "rewards/margins": 6.330525875091553, "rewards/rejected": -5.934195518493652, "step": 335 }, { "epoch": 2.463768115942029, "grad_norm": 13.686346123613403, "learning_rate": 8.052359796084951e-08, "logits/chosen": -1.983902931213379, "logits/rejected": -1.8964077234268188, "logps/chosen": -190.7554473876953, "logps/rejected": -248.05477905273438, "loss": 0.0633, "rewards/accuracies": 0.987500011920929, "rewards/chosen": 1.1571325063705444, "rewards/margins": 6.923959255218506, "rewards/rejected": -5.766826629638672, "step": 340 }, { "epoch": 2.5, "grad_norm": 12.547226987434398, "learning_rate": 7.026366513129139e-08, "logits/chosen": -1.9592113494873047, "logits/rejected": -1.8749430179595947, "logps/chosen": -184.05401611328125, "logps/rejected": -228.3144989013672, "loss": 0.0603, "rewards/accuracies": 0.987500011920929, "rewards/chosen": 0.9612242579460144, "rewards/margins": 5.984218597412109, "rewards/rejected": -5.0229949951171875, "step": 345 }, { "epoch": 2.536231884057971, "grad_norm": 10.429640146690803, "learning_rate": 6.065329995036572e-08, "logits/chosen": -1.9933052062988281, "logits/rejected": -1.9089053869247437, "logps/chosen": -190.28915405273438, "logps/rejected": -235.51040649414062, "loss": 0.056, "rewards/accuracies": 0.981249988079071, "rewards/chosen": 0.5995205640792847, "rewards/margins": 6.035165786743164, "rewards/rejected": -5.43564510345459, "step": 350 }, { "epoch": 2.536231884057971, "eval_logits/chosen": -1.9756227731704712, "eval_logits/rejected": -1.916344165802002, "eval_logps/chosen": -205.54391479492188, "eval_logps/rejected": -239.3422393798828, "eval_loss": 0.44805780053138733, "eval_rewards/accuracies": 0.8185483813285828, "eval_rewards/chosen": -0.5224470496177673, "eval_rewards/margins": 3.5909695625305176, "eval_rewards/rejected": -4.1134161949157715, "eval_runtime": 247.8226, "eval_samples_per_second": 15.826, "eval_steps_per_second": 0.25, "step": 350 }, { "epoch": 2.572463768115942, "grad_norm": 14.703238617144146, "learning_rate": 5.170702895866591e-08, "logits/chosen": -1.9757484197616577, "logits/rejected": -1.877820611000061, "logps/chosen": -181.67088317871094, "logps/rejected": -234.2239227294922, "loss": 0.0539, "rewards/accuracies": 0.9937499761581421, "rewards/chosen": 0.7643092274665833, "rewards/margins": 6.472746849060059, "rewards/rejected": -5.708437919616699, "step": 355 }, { "epoch": 2.608695652173913, "grad_norm": 16.11683384920457, "learning_rate": 4.343837488569057e-08, "logits/chosen": -1.9882529973983765, "logits/rejected": -1.937097191810608, "logps/chosen": -187.87106323242188, "logps/rejected": -243.0420379638672, "loss": 0.0668, "rewards/accuracies": 0.987500011920929, "rewards/chosen": 0.925794780254364, "rewards/margins": 6.516719818115234, "rewards/rejected": -5.590925216674805, "step": 360 }, { "epoch": 2.644927536231884, "grad_norm": 11.830853634227745, "learning_rate": 3.585983620957112e-08, "logits/chosen": -1.9911915063858032, "logits/rejected": -1.9083200693130493, "logps/chosen": -183.7674560546875, "logps/rejected": -238.542236328125, "loss": 0.0632, "rewards/accuracies": 0.9937499761581421, "rewards/chosen": 0.6369873285293579, "rewards/margins": 6.4096479415893555, "rewards/rejected": -5.772660732269287, "step": 365 }, { "epoch": 2.681159420289855, "grad_norm": 12.756089452099605, "learning_rate": 2.8982868265005454e-08, "logits/chosen": -1.9761543273925781, "logits/rejected": -1.9331614971160889, "logps/chosen": -189.85000610351562, "logps/rejected": -237.2536163330078, "loss": 0.0563, "rewards/accuracies": 0.987500011920929, "rewards/chosen": 0.8311630487442017, "rewards/margins": 6.6239190101623535, "rewards/rejected": -5.792755603790283, "step": 370 }, { "epoch": 2.717391304347826, "grad_norm": 14.499306986468172, "learning_rate": 2.2817865927956092e-08, "logits/chosen": -1.9935853481292725, "logits/rejected": -1.9374994039535522, "logps/chosen": -182.0215606689453, "logps/rejected": -235.44271850585938, "loss": 0.0611, "rewards/accuracies": 0.981249988079071, "rewards/chosen": 0.9944933652877808, "rewards/margins": 6.3055338859558105, "rewards/rejected": -5.31104040145874, "step": 375 }, { "epoch": 2.753623188405797, "grad_norm": 12.558302938263997, "learning_rate": 1.7374147903282176e-08, "logits/chosen": -1.9671310186386108, "logits/rejected": -1.9008516073226929, "logps/chosen": -196.24166870117188, "logps/rejected": -243.8437957763672, "loss": 0.0483, "rewards/accuracies": 0.987500011920929, "rewards/chosen": 0.6471545696258545, "rewards/margins": 6.556443691253662, "rewards/rejected": -5.909289360046387, "step": 380 }, { "epoch": 2.789855072463768, "grad_norm": 7.550248384042512, "learning_rate": 1.2659942639057952e-08, "logits/chosen": -1.9696018695831299, "logits/rejected": -1.9180386066436768, "logps/chosen": -197.0672607421875, "logps/rejected": -251.66244506835938, "loss": 0.0548, "rewards/accuracies": 0.987500011920929, "rewards/chosen": 0.7232016921043396, "rewards/margins": 6.699126243591309, "rewards/rejected": -5.975924968719482, "step": 385 }, { "epoch": 2.8260869565217392, "grad_norm": 8.880479703177947, "learning_rate": 8.682375888868166e-09, "logits/chosen": -1.975227952003479, "logits/rejected": -1.9457371234893799, "logps/chosen": -191.05844116210938, "logps/rejected": -256.9841003417969, "loss": 0.0488, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": 0.38350382447242737, "rewards/margins": 6.125405311584473, "rewards/rejected": -5.741901874542236, "step": 390 }, { "epoch": 2.86231884057971, "grad_norm": 10.914718704369829, "learning_rate": 5.447459940880084e-09, "logits/chosen": -1.978607416152954, "logits/rejected": -1.921014428138733, "logps/chosen": -183.47064208984375, "logps/rejected": -249.2570037841797, "loss": 0.0512, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": 0.4558964669704437, "rewards/margins": 6.345686912536621, "rewards/rejected": -5.8897905349731445, "step": 395 }, { "epoch": 2.898550724637681, "grad_norm": 12.495149398792217, "learning_rate": 2.9600845299737053e-09, "logits/chosen": -1.958653450012207, "logits/rejected": -1.9010944366455078, "logps/chosen": -177.4004364013672, "logps/rejected": -237.0882110595703, "loss": 0.0721, "rewards/accuracies": 0.9624999761581421, "rewards/chosen": 0.2784881889820099, "rewards/margins": 6.083136558532715, "rewards/rejected": -5.804648399353027, "step": 400 }, { "epoch": 2.898550724637681, "eval_logits/chosen": -1.9731168746948242, "eval_logits/rejected": -1.9129306077957153, "eval_logps/chosen": -207.34255981445312, "eval_logps/rejected": -242.29010009765625, "eval_loss": 0.4506772756576538, "eval_rewards/accuracies": 0.8185483813285828, "eval_rewards/chosen": -0.7023105025291443, "eval_rewards/margins": 3.7058920860290527, "eval_rewards/rejected": -4.408202171325684, "eval_runtime": 247.9558, "eval_samples_per_second": 15.817, "eval_steps_per_second": 0.25, "step": 400 }, { "epoch": 2.9347826086956523, "grad_norm": 10.380155420725053, "learning_rate": 1.2240094466668404e-09, "logits/chosen": -2.010136127471924, "logits/rejected": -1.8942158222198486, "logps/chosen": -191.6996612548828, "logps/rejected": -268.3369445800781, "loss": 0.0597, "rewards/accuracies": 0.987500011920929, "rewards/chosen": 0.43750476837158203, "rewards/margins": 6.982994079589844, "rewards/rejected": -6.5454888343811035, "step": 405 }, { "epoch": 2.971014492753623, "grad_norm": 18.203204347608896, "learning_rate": 2.418588540059607e-10, "logits/chosen": -1.9836390018463135, "logits/rejected": -1.9380347728729248, "logps/chosen": -186.40122985839844, "logps/rejected": -238.04464721679688, "loss": 0.0554, "rewards/accuracies": 0.9937499761581421, "rewards/chosen": 0.8866332769393921, "rewards/margins": 6.659180641174316, "rewards/rejected": -5.772547721862793, "step": 410 } ], "logging_steps": 5, "max_steps": 414, "num_input_tokens_seen": 0, "num_train_epochs": 3, "save_steps": 100, "stateful_callbacks": { "TrainerControl": { "args": { "should_epoch_stop": false, "should_evaluate": false, "should_log": false, "should_save": true, "should_training_stop": true }, "attributes": {} } }, "total_flos": 4881795388538880.0, "train_batch_size": 8, "trial_name": null, "trial_params": null }