{ "best_metric": null, "best_model_checkpoint": null, "epoch": 0.9988623435722411, "eval_steps": 10000000, "global_step": 439, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "grad_norm": 22.786040256127208, "learning_rate": 2.2727272727272727e-09, "logits/chosen": -1.6768856048583984, "logits/rejected": -1.7259055376052856, "logps/chosen": -1.2793102264404297, "logps/rejected": -1.2162058353424072, "loss": 0.6931, "rewards/accuracies": 0.0, "rewards/chosen": 0.0, "rewards/margins": 0.0, "rewards/rejected": 0.0, "step": 1 }, { "epoch": 0.02, "grad_norm": 25.6836964844408, "learning_rate": 2.2727272727272725e-08, "logits/chosen": -1.7025811672210693, "logits/rejected": -1.6681201457977295, "logps/chosen": -1.2131770849227905, "logps/rejected": -1.2202790975570679, "loss": 0.6931, "rewards/accuracies": 0.4166666567325592, "rewards/chosen": 0.00025717043899931014, "rewards/margins": -0.0006778165698051453, "rewards/rejected": 0.0009349870379082859, "step": 10 }, { "epoch": 0.05, "grad_norm": 26.405806920127038, "learning_rate": 4.545454545454545e-08, "logits/chosen": -1.7794504165649414, "logits/rejected": -1.734635591506958, "logps/chosen": -1.144683599472046, "logps/rejected": -1.1852140426635742, "loss": 0.6925, "rewards/accuracies": 0.5375000238418579, "rewards/chosen": -0.0029223989695310593, "rewards/margins": 0.0005865652346983552, "rewards/rejected": -0.0035089640878140926, "step": 20 }, { "epoch": 0.07, "grad_norm": 28.068883928310647, "learning_rate": 6.818181818181817e-08, "logits/chosen": -1.7446922063827515, "logits/rejected": -1.676011085510254, "logps/chosen": -1.1949479579925537, "logps/rejected": -1.2484194040298462, "loss": 0.6882, "rewards/accuracies": 0.6625000238418579, "rewards/chosen": -0.024093259125947952, "rewards/margins": 0.012483139522373676, "rewards/rejected": -0.036576397716999054, "step": 30 }, { "epoch": 0.09, "grad_norm": 19.593533440172358, "learning_rate": 9.09090909090909e-08, "logits/chosen": -1.7312889099121094, "logits/rejected": -1.6649024486541748, "logps/chosen": -1.2515684366226196, "logps/rejected": -1.3352997303009033, "loss": 0.6797, "rewards/accuracies": 0.6625000238418579, "rewards/chosen": -0.08923971652984619, "rewards/margins": 0.06281541287899017, "rewards/rejected": -0.15205512940883636, "step": 40 }, { "epoch": 0.11, "grad_norm": 24.263100694250465, "learning_rate": 9.994307990108962e-08, "logits/chosen": -1.6921930313110352, "logits/rejected": -1.6271032094955444, "logps/chosen": -1.298782467842102, "logps/rejected": -1.3565475940704346, "loss": 0.6637, "rewards/accuracies": 0.7250000238418579, "rewards/chosen": -0.1813907027244568, "rewards/margins": 0.07534531503915787, "rewards/rejected": -0.25673604011535645, "step": 50 }, { "epoch": 0.14, "grad_norm": 28.402662229181594, "learning_rate": 9.959570405988094e-08, "logits/chosen": -1.7207958698272705, "logits/rejected": -1.6402089595794678, "logps/chosen": -1.2883288860321045, "logps/rejected": -1.378244161605835, "loss": 0.6599, "rewards/accuracies": 0.6499999761581421, "rewards/chosen": -0.35846152901649475, "rewards/margins": 0.07597381621599197, "rewards/rejected": -0.4344353675842285, "step": 60 }, { "epoch": 0.16, "grad_norm": 21.201003636828116, "learning_rate": 9.893476820924666e-08, "logits/chosen": -1.801526665687561, "logits/rejected": -1.7130231857299805, "logps/chosen": -1.4759318828582764, "logps/rejected": -1.5954515933990479, "loss": 0.6498, "rewards/accuracies": 0.699999988079071, "rewards/chosen": -0.5261799097061157, "rewards/margins": 0.14570936560630798, "rewards/rejected": -0.6718892455101013, "step": 70 }, { "epoch": 0.18, "grad_norm": 26.86998789465821, "learning_rate": 9.796445099843647e-08, "logits/chosen": -1.7849477529525757, "logits/rejected": -1.6990400552749634, "logps/chosen": -1.5336757898330688, "logps/rejected": -1.6886417865753174, "loss": 0.654, "rewards/accuracies": 0.668749988079071, "rewards/chosen": -0.6534146070480347, "rewards/margins": 0.18887227773666382, "rewards/rejected": -0.8422868847846985, "step": 80 }, { "epoch": 0.2, "grad_norm": 25.145260129960253, "learning_rate": 9.669088708527066e-08, "logits/chosen": -1.7331278324127197, "logits/rejected": -1.6643707752227783, "logps/chosen": -1.604683518409729, "logps/rejected": -1.6745525598526, "loss": 0.6397, "rewards/accuracies": 0.643750011920929, "rewards/chosen": -0.7379721999168396, "rewards/margins": 0.12564878165721893, "rewards/rejected": -0.8636210560798645, "step": 90 }, { "epoch": 0.23, "grad_norm": 28.208412906348087, "learning_rate": 9.512212835085849e-08, "logits/chosen": -1.770500898361206, "logits/rejected": -1.6813795566558838, "logps/chosen": -1.557369351387024, "logps/rejected": -1.7003123760223389, "loss": 0.623, "rewards/accuracies": 0.6937500238418579, "rewards/chosen": -0.7778279781341553, "rewards/margins": 0.200466588139534, "rewards/rejected": -0.9782945513725281, "step": 100 }, { "epoch": 0.25, "grad_norm": 25.857658698377822, "learning_rate": 9.326809299301306e-08, "logits/chosen": -1.7706959247589111, "logits/rejected": -1.666680932044983, "logps/chosen": -1.6019655466079712, "logps/rejected": -1.7736393213272095, "loss": 0.6212, "rewards/accuracies": 0.7124999761581421, "rewards/chosen": -0.8026469945907593, "rewards/margins": 0.240141361951828, "rewards/rejected": -1.0427882671356201, "step": 110 }, { "epoch": 0.27, "grad_norm": 27.113925141570313, "learning_rate": 9.114050282021158e-08, "logits/chosen": -1.7661209106445312, "logits/rejected": -1.7044270038604736, "logps/chosen": -1.5668163299560547, "logps/rejected": -1.7313486337661743, "loss": 0.6144, "rewards/accuracies": 0.731249988079071, "rewards/chosen": -0.8385075330734253, "rewards/margins": 0.2342797964811325, "rewards/rejected": -1.0727872848510742, "step": 120 }, { "epoch": 0.3, "grad_norm": 25.192318558632277, "learning_rate": 8.875280914254802e-08, "logits/chosen": -1.7651408910751343, "logits/rejected": -1.6746151447296143, "logps/chosen": -1.7618343830108643, "logps/rejected": -1.9648230075836182, "loss": 0.5995, "rewards/accuracies": 0.668749988079071, "rewards/chosen": -1.1380090713500977, "rewards/margins": 0.34912705421447754, "rewards/rejected": -1.4871362447738647, "step": 130 }, { "epoch": 0.32, "grad_norm": 24.555074276571457, "learning_rate": 8.612010772821971e-08, "logits/chosen": -1.7977707386016846, "logits/rejected": -1.7540991306304932, "logps/chosen": -1.8272440433502197, "logps/rejected": -1.9743133783340454, "loss": 0.5993, "rewards/accuracies": 0.737500011920929, "rewards/chosen": -1.2267296314239502, "rewards/margins": 0.34395888447761536, "rewards/rejected": -1.5706886053085327, "step": 140 }, { "epoch": 0.34, "grad_norm": 29.798327176387467, "learning_rate": 8.325904336322055e-08, "logits/chosen": -1.7762537002563477, "logits/rejected": -1.722198247909546, "logps/chosen": -1.9330543279647827, "logps/rejected": -2.1491074562072754, "loss": 0.6088, "rewards/accuracies": 0.6812499761581421, "rewards/chosen": -1.5313791036605835, "rewards/margins": 0.3377588093280792, "rewards/rejected": -1.8691380023956299, "step": 150 }, { "epoch": 0.36, "grad_norm": 22.379243172505653, "learning_rate": 8.01877046176447e-08, "logits/chosen": -1.7237749099731445, "logits/rejected": -1.6540390253067017, "logps/chosen": -1.9900233745574951, "logps/rejected": -2.1875462532043457, "loss": 0.5903, "rewards/accuracies": 0.675000011920929, "rewards/chosen": -1.6234468221664429, "rewards/margins": 0.29227370023727417, "rewards/rejected": -1.9157205820083618, "step": 160 }, { "epoch": 0.39, "grad_norm": 19.935363442551864, "learning_rate": 7.692550948392249e-08, "logits/chosen": -1.7721624374389648, "logits/rejected": -1.7196706533432007, "logps/chosen": -1.960934042930603, "logps/rejected": -2.164457321166992, "loss": 0.5921, "rewards/accuracies": 0.706250011920929, "rewards/chosen": -1.501357078552246, "rewards/margins": 0.38615694642066956, "rewards/rejected": -1.887514352798462, "step": 170 }, { "epoch": 0.41, "grad_norm": 27.99675949527233, "learning_rate": 7.349308261002021e-08, "logits/chosen": -1.7341270446777344, "logits/rejected": -1.6832844018936157, "logps/chosen": -1.9921613931655884, "logps/rejected": -2.2205045223236084, "loss": 0.5854, "rewards/accuracies": 0.699999988079071, "rewards/chosen": -1.5901505947113037, "rewards/margins": 0.35738545656204224, "rewards/rejected": -1.9475358724594116, "step": 180 }, { "epoch": 0.43, "grad_norm": 24.529175323127856, "learning_rate": 6.991212490377531e-08, "logits/chosen": -1.7890892028808594, "logits/rejected": -1.746469259262085, "logps/chosen": -2.087700843811035, "logps/rejected": -2.3541736602783203, "loss": 0.5608, "rewards/accuracies": 0.75, "rewards/chosen": -1.7151954174041748, "rewards/margins": 0.4898623526096344, "rewards/rejected": -2.2050578594207764, "step": 190 }, { "epoch": 0.46, "grad_norm": 30.651962115004057, "learning_rate": 6.620527633276978e-08, "logits/chosen": -1.7287298440933228, "logits/rejected": -1.6635551452636719, "logps/chosen": -2.149373769760132, "logps/rejected": -2.5112643241882324, "loss": 0.5803, "rewards/accuracies": 0.71875, "rewards/chosen": -1.909841537475586, "rewards/margins": 0.5502029657363892, "rewards/rejected": -2.4600441455841064, "step": 200 }, { "epoch": 0.48, "grad_norm": 27.408811645376396, "learning_rate": 6.239597278716581e-08, "logits/chosen": -1.7830358743667603, "logits/rejected": -1.7272449731826782, "logps/chosen": -2.303990125656128, "logps/rejected": -2.549621105194092, "loss": 0.5627, "rewards/accuracies": 0.7250000238418579, "rewards/chosen": -2.1173198223114014, "rewards/margins": 0.5325894355773926, "rewards/rejected": -2.649909496307373, "step": 210 }, { "epoch": 0.5, "grad_norm": 25.141892376540234, "learning_rate": 5.8508297910462456e-08, "logits/chosen": -1.729365587234497, "logits/rejected": -1.6552003622055054, "logps/chosen": -2.227386713027954, "logps/rejected": -2.558889627456665, "loss": 0.5666, "rewards/accuracies": 0.6937500238418579, "rewards/chosen": -2.1277427673339844, "rewards/margins": 0.5474593043327332, "rewards/rejected": -2.675201892852783, "step": 220 }, { "epoch": 0.52, "grad_norm": 25.203206141016988, "learning_rate": 5.456683083494731e-08, "logits/chosen": -1.7191989421844482, "logits/rejected": -1.6832917928695679, "logps/chosen": -2.193159580230713, "logps/rejected": -2.4023752212524414, "loss": 0.5847, "rewards/accuracies": 0.65625, "rewards/chosen": -1.9987010955810547, "rewards/margins": 0.3709481358528137, "rewards/rejected": -2.3696494102478027, "step": 230 }, { "epoch": 0.55, "grad_norm": 31.30257665339357, "learning_rate": 5.059649078450834e-08, "logits/chosen": -1.7015504837036133, "logits/rejected": -1.662376046180725, "logps/chosen": -2.2371976375579834, "logps/rejected": -2.5139238834381104, "loss": 0.5533, "rewards/accuracies": 0.6625000238418579, "rewards/chosen": -2.137688398361206, "rewards/margins": 0.4293746054172516, "rewards/rejected": -2.567063570022583, "step": 240 }, { "epoch": 0.57, "grad_norm": 24.832978144124045, "learning_rate": 4.6622379527277186e-08, "logits/chosen": -1.7127622365951538, "logits/rejected": -1.6652097702026367, "logps/chosen": -2.328359603881836, "logps/rejected": -2.558023452758789, "loss": 0.5645, "rewards/accuracies": 0.6187499761581421, "rewards/chosen": -2.378941535949707, "rewards/margins": 0.3931688666343689, "rewards/rejected": -2.7721104621887207, "step": 250 }, { "epoch": 0.59, "grad_norm": 28.702758237373775, "learning_rate": 4.26696226741691e-08, "logits/chosen": -1.725927710533142, "logits/rejected": -1.6632360219955444, "logps/chosen": -2.4254117012023926, "logps/rejected": -2.7066590785980225, "loss": 0.5648, "rewards/accuracies": 0.65625, "rewards/chosen": -2.385176181793213, "rewards/margins": 0.522560715675354, "rewards/rejected": -2.9077370166778564, "step": 260 }, { "epoch": 0.61, "grad_norm": 24.57643832571065, "learning_rate": 3.876321082668098e-08, "logits/chosen": -1.7822297811508179, "logits/rejected": -1.7207348346710205, "logps/chosen": -2.3284859657287598, "logps/rejected": -2.6171371936798096, "loss": 0.5581, "rewards/accuracies": 0.7250000238418579, "rewards/chosen": -2.179990530014038, "rewards/margins": 0.5586847066879272, "rewards/rejected": -2.7386748790740967, "step": 270 }, { "epoch": 0.64, "grad_norm": 26.182300162855046, "learning_rate": 3.492784157826244e-08, "logits/chosen": -1.721687912940979, "logits/rejected": -1.6329882144927979, "logps/chosen": -2.33394193649292, "logps/rejected": -2.6748275756835938, "loss": 0.5564, "rewards/accuracies": 0.7124999761581421, "rewards/chosen": -2.2221310138702393, "rewards/margins": 0.610991895198822, "rewards/rejected": -2.833122730255127, "step": 280 }, { "epoch": 0.66, "grad_norm": 22.815010055443274, "learning_rate": 3.118776336817812e-08, "logits/chosen": -1.755319595336914, "logits/rejected": -1.7038185596466064, "logps/chosen": -2.277468204498291, "logps/rejected": -2.616783618927002, "loss": 0.5473, "rewards/accuracies": 0.731249988079071, "rewards/chosen": -2.1484389305114746, "rewards/margins": 0.6641989946365356, "rewards/rejected": -2.8126378059387207, "step": 290 }, { "epoch": 0.68, "grad_norm": 23.65537068910306, "learning_rate": 2.7566622175067443e-08, "logits/chosen": -1.7356256246566772, "logits/rejected": -1.6781995296478271, "logps/chosen": -2.3750412464141846, "logps/rejected": -2.7629873752593994, "loss": 0.5497, "rewards/accuracies": 0.7562500238418579, "rewards/chosen": -2.3329153060913086, "rewards/margins": 0.6199190616607666, "rewards/rejected": -2.9528346061706543, "step": 300 }, { "epoch": 0.71, "grad_norm": 27.623729083981935, "learning_rate": 2.408731201945432e-08, "logits/chosen": -1.7345046997070312, "logits/rejected": -1.6898084878921509, "logps/chosen": -2.304429531097412, "logps/rejected": -2.555560827255249, "loss": 0.5555, "rewards/accuracies": 0.675000011920929, "rewards/chosen": -2.2210681438446045, "rewards/margins": 0.46195346117019653, "rewards/rejected": -2.6830217838287354, "step": 310 }, { "epoch": 0.73, "grad_norm": 26.649151417142534, "learning_rate": 2.0771830220378112e-08, "logits/chosen": -1.68793523311615, "logits/rejected": -1.6301701068878174, "logps/chosen": -2.3206005096435547, "logps/rejected": -2.559023380279541, "loss": 0.5553, "rewards/accuracies": 0.625, "rewards/chosen": -2.195432424545288, "rewards/margins": 0.4925463795661926, "rewards/rejected": -2.687978744506836, "step": 320 }, { "epoch": 0.75, "grad_norm": 24.875498436093228, "learning_rate": 1.7641138321260257e-08, "logits/chosen": -1.722652792930603, "logits/rejected": -1.6583467721939087, "logps/chosen": -2.2422232627868652, "logps/rejected": -2.6411871910095215, "loss": 0.5479, "rewards/accuracies": 0.831250011920929, "rewards/chosen": -2.116737127304077, "rewards/margins": 0.7428576946258545, "rewards/rejected": -2.8595950603485107, "step": 330 }, { "epoch": 0.77, "grad_norm": 26.307918213457313, "learning_rate": 1.4715029564277793e-08, "logits/chosen": -1.7855126857757568, "logits/rejected": -1.7428696155548096, "logps/chosen": -2.2222747802734375, "logps/rejected": -2.578981399536133, "loss": 0.5581, "rewards/accuracies": 0.75, "rewards/chosen": -2.0352282524108887, "rewards/margins": 0.6392861604690552, "rewards/rejected": -2.6745142936706543, "step": 340 }, { "epoch": 0.8, "grad_norm": 25.341211676404754, "learning_rate": 1.2012003751113343e-08, "logits/chosen": -1.7763115167617798, "logits/rejected": -1.7206395864486694, "logps/chosen": -2.328850269317627, "logps/rejected": -2.70664119720459, "loss": 0.5366, "rewards/accuracies": 0.71875, "rewards/chosen": -2.3333606719970703, "rewards/margins": 0.6552005410194397, "rewards/rejected": -2.9885611534118652, "step": 350 }, { "epoch": 0.82, "grad_norm": 31.147010925293188, "learning_rate": 9.549150281252633e-09, "logits/chosen": -1.725211501121521, "logits/rejected": -1.6820595264434814, "logps/chosen": -2.2657017707824707, "logps/rejected": -2.6082816123962402, "loss": 0.5395, "rewards/accuracies": 0.6937500238418579, "rewards/chosen": -2.1871337890625, "rewards/margins": 0.5816982388496399, "rewards/rejected": -2.768831729888916, "step": 360 }, { "epoch": 0.84, "grad_norm": 24.868596217446985, "learning_rate": 7.3420401072985306e-09, "logits/chosen": -1.7670882940292358, "logits/rejected": -1.7162415981292725, "logps/chosen": -2.3288838863372803, "logps/rejected": -2.694032669067383, "loss": 0.5377, "rewards/accuracies": 0.6812499761581421, "rewards/chosen": -2.2285408973693848, "rewards/margins": 0.6153696179389954, "rewards/rejected": -2.8439104557037354, "step": 370 }, { "epoch": 0.86, "grad_norm": 28.264277931849893, "learning_rate": 5.404627290395369e-09, "logits/chosen": -1.7286021709442139, "logits/rejected": -1.6685943603515625, "logps/chosen": -2.2551302909851074, "logps/rejected": -2.6050727367401123, "loss": 0.537, "rewards/accuracies": 0.737500011920929, "rewards/chosen": -2.1310482025146484, "rewards/margins": 0.619892954826355, "rewards/rejected": -2.750941514968872, "step": 380 }, { "epoch": 0.89, "grad_norm": 32.571402526944446, "learning_rate": 3.74916077816162e-09, "logits/chosen": -1.734569787979126, "logits/rejected": -1.678733468055725, "logps/chosen": -2.2776572704315186, "logps/rejected": -2.573931932449341, "loss": 0.5505, "rewards/accuracies": 0.6812499761581421, "rewards/chosen": -2.255725145339966, "rewards/margins": 0.5602914690971375, "rewards/rejected": -2.816016674041748, "step": 390 }, { "epoch": 0.91, "grad_norm": 30.76786815731744, "learning_rate": 2.386106962899165e-09, "logits/chosen": -1.6618547439575195, "logits/rejected": -1.599841594696045, "logps/chosen": -2.417301654815674, "logps/rejected": -2.7463181018829346, "loss": 0.5418, "rewards/accuracies": 0.6937500238418579, "rewards/chosen": -2.381761312484741, "rewards/margins": 0.5453529357910156, "rewards/rejected": -2.927114725112915, "step": 400 }, { "epoch": 0.93, "grad_norm": 31.25431184934112, "learning_rate": 1.3240835096913706e-09, "logits/chosen": -1.6889197826385498, "logits/rejected": -1.5975219011306763, "logps/chosen": -2.266620397567749, "logps/rejected": -2.7067856788635254, "loss": 0.5607, "rewards/accuracies": 0.7124999761581421, "rewards/chosen": -2.190139055252075, "rewards/margins": 0.7415019273757935, "rewards/rejected": -2.931641101837158, "step": 410 }, { "epoch": 0.96, "grad_norm": 23.39234165392022, "learning_rate": 5.698048727497462e-10, "logits/chosen": -1.7234703302383423, "logits/rejected": -1.65776789188385, "logps/chosen": -2.363593578338623, "logps/rejected": -2.7682502269744873, "loss": 0.5431, "rewards/accuracies": 0.7250000238418579, "rewards/chosen": -2.3445918560028076, "rewards/margins": 0.6753993034362793, "rewards/rejected": -3.019990921020508, "step": 420 }, { "epoch": 0.98, "grad_norm": 24.387678586587597, "learning_rate": 1.2803984447259387e-10, "logits/chosen": -1.7296276092529297, "logits/rejected": -1.68634033203125, "logps/chosen": -2.347764730453491, "logps/rejected": -2.731468677520752, "loss": 0.5333, "rewards/accuracies": 0.7250000238418579, "rewards/chosen": -2.330850601196289, "rewards/margins": 0.6721019744873047, "rewards/rejected": -3.002953052520752, "step": 430 }, { "epoch": 1.0, "step": 439, "total_flos": 0.0, "train_loss": 0.4350618153878388, "train_runtime": 5287.1692, "train_samples_per_second": 10.636, "train_steps_per_second": 0.083 } ], "logging_steps": 10, "max_steps": 439, "num_input_tokens_seen": 0, "num_train_epochs": 1, "save_steps": 100, "total_flos": 0.0, "train_batch_size": 4, "trial_name": null, "trial_params": null }