{ "best_metric": null, "best_model_checkpoint": null, "epoch": 1.0, "eval_steps": 100, "global_step": 478, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0020920502092050207, "grad_norm": 10.133265544600611, "learning_rate": 1.0416666666666666e-08, "logits/chosen": -3.21875, "logits/rejected": -3.21875, "logps/chosen": -420.0, "logps/rejected": -552.0, "loss": 0.6914, "rewards/accuracies": 0.0, "rewards/chosen": 0.0, "rewards/margins": 0.0, "rewards/rejected": 0.0, "step": 1 }, { "epoch": 0.02092050209205021, "grad_norm": 9.184965981316628, "learning_rate": 1.0416666666666667e-07, "logits/chosen": -3.265625, "logits/rejected": -3.28125, "logps/chosen": -484.0, "logps/rejected": -516.0, "loss": 0.6919, "rewards/accuracies": 0.1111111119389534, "rewards/chosen": 0.001251220703125, "rewards/margins": 0.00055694580078125, "rewards/rejected": 0.00069427490234375, "step": 10 }, { "epoch": 0.04184100418410042, "grad_norm": 10.105256126223187, "learning_rate": 2.0833333333333333e-07, "logits/chosen": -3.21875, "logits/rejected": -3.234375, "logps/chosen": -520.0, "logps/rejected": -540.0, "loss": 0.6902, "rewards/accuracies": 0.3125, "rewards/chosen": 0.00799560546875, "rewards/margins": 0.004119873046875, "rewards/rejected": 0.003875732421875, "step": 20 }, { "epoch": 0.06276150627615062, "grad_norm": 8.97965579353216, "learning_rate": 3.1249999999999997e-07, "logits/chosen": -3.28125, "logits/rejected": -3.265625, "logps/chosen": -468.0, "logps/rejected": -516.0, "loss": 0.6829, "rewards/accuracies": 0.5375000238418579, "rewards/chosen": 0.040771484375, "rewards/margins": 0.0198974609375, "rewards/rejected": 0.0208740234375, "step": 30 }, { "epoch": 0.08368200836820083, "grad_norm": 9.227660720825863, "learning_rate": 4.1666666666666667e-07, "logits/chosen": -3.25, "logits/rejected": -3.328125, "logps/chosen": -506.0, "logps/rejected": -520.0, "loss": 0.6627, "rewards/accuracies": 0.675000011920929, "rewards/chosen": 0.09130859375, "rewards/margins": 0.059814453125, "rewards/rejected": 0.03173828125, "step": 40 }, { "epoch": 0.10460251046025104, "grad_norm": 9.819087605132305, "learning_rate": 4.999733114418725e-07, "logits/chosen": -3.359375, "logits/rejected": -3.390625, "logps/chosen": -498.0, "logps/rejected": -544.0, "loss": 0.6301, "rewards/accuracies": 0.737500011920929, "rewards/chosen": 0.0257568359375, "rewards/margins": 0.1376953125, "rewards/rejected": -0.11181640625, "step": 50 }, { "epoch": 0.12552301255230125, "grad_norm": 12.048088131076954, "learning_rate": 4.990398100856366e-07, "logits/chosen": -3.53125, "logits/rejected": -3.484375, "logps/chosen": -568.0, "logps/rejected": -604.0, "loss": 0.5911, "rewards/accuracies": 0.7749999761581421, "rewards/chosen": -0.236328125, "rewards/margins": 0.267578125, "rewards/rejected": -0.50390625, "step": 60 }, { "epoch": 0.14644351464435146, "grad_norm": 14.528700694610064, "learning_rate": 4.967775735898179e-07, "logits/chosen": -3.703125, "logits/rejected": -3.65625, "logps/chosen": -576.0, "logps/rejected": -660.0, "loss": 0.54, "rewards/accuracies": 0.78125, "rewards/chosen": -0.6015625, "rewards/margins": 0.478515625, "rewards/rejected": -1.078125, "step": 70 }, { "epoch": 0.16736401673640167, "grad_norm": 14.89693527880136, "learning_rate": 4.931986719649298e-07, "logits/chosen": -3.78125, "logits/rejected": -3.734375, "logps/chosen": -680.0, "logps/rejected": -792.0, "loss": 0.468, "rewards/accuracies": 0.7875000238418579, "rewards/chosen": -1.4921875, "rewards/margins": 0.75, "rewards/rejected": -2.234375, "step": 80 }, { "epoch": 0.18828451882845187, "grad_norm": 24.010753148040607, "learning_rate": 4.883222001996351e-07, "logits/chosen": -3.71875, "logits/rejected": -3.734375, "logps/chosen": -704.0, "logps/rejected": -880.0, "loss": 0.3987, "rewards/accuracies": 0.84375, "rewards/chosen": -2.03125, "rewards/margins": 1.3125, "rewards/rejected": -3.34375, "step": 90 }, { "epoch": 0.20920502092050208, "grad_norm": 20.988032239641004, "learning_rate": 4.821741763807186e-07, "logits/chosen": -3.71875, "logits/rejected": -3.703125, "logps/chosen": -732.0, "logps/rejected": -912.0, "loss": 0.3407, "rewards/accuracies": 0.8374999761581421, "rewards/chosen": -2.46875, "rewards/margins": 1.5859375, "rewards/rejected": -4.0625, "step": 100 }, { "epoch": 0.20920502092050208, "eval_logits/chosen": -3.75, "eval_logits/rejected": -3.75, "eval_logps/chosen": -724.0, "eval_logps/rejected": -912.0, "eval_loss": 0.3028574287891388, "eval_rewards/accuracies": 0.8828125, "eval_rewards/chosen": -2.25, "eval_rewards/margins": 1.75, "eval_rewards/rejected": -4.0, "eval_runtime": 89.7294, "eval_samples_per_second": 22.289, "eval_steps_per_second": 0.357, "step": 100 }, { "epoch": 0.2301255230125523, "grad_norm": 19.53117460181654, "learning_rate": 4.747874028753375e-07, "logits/chosen": -3.703125, "logits/rejected": -3.65625, "logps/chosen": -756.0, "logps/rejected": -1024.0, "loss": 0.2921, "rewards/accuracies": 0.862500011920929, "rewards/chosen": -2.578125, "rewards/margins": 2.390625, "rewards/rejected": -4.96875, "step": 110 }, { "epoch": 0.2510460251046025, "grad_norm": 26.848552029083983, "learning_rate": 4.662012913161997e-07, "logits/chosen": -3.765625, "logits/rejected": -3.75, "logps/chosen": -692.0, "logps/rejected": -1040.0, "loss": 0.2388, "rewards/accuracies": 0.918749988079071, "rewards/chosen": -2.15625, "rewards/margins": 3.140625, "rewards/rejected": -5.3125, "step": 120 }, { "epoch": 0.2719665271966527, "grad_norm": 49.46963494787925, "learning_rate": 4.5646165232345103e-07, "logits/chosen": -3.75, "logits/rejected": -3.671875, "logps/chosen": -820.0, "logps/rejected": -1376.0, "loss": 0.2143, "rewards/accuracies": 0.8812500238418579, "rewards/chosen": -3.265625, "rewards/margins": 5.125, "rewards/rejected": -8.375, "step": 130 }, { "epoch": 0.2928870292887029, "grad_norm": 27.856656903315532, "learning_rate": 4.456204510851956e-07, "logits/chosen": -3.75, "logits/rejected": -3.625, "logps/chosen": -852.0, "logps/rejected": -1416.0, "loss": 0.196, "rewards/accuracies": 0.875, "rewards/chosen": -3.40625, "rewards/margins": 5.25, "rewards/rejected": -8.625, "step": 140 }, { "epoch": 0.3138075313807531, "grad_norm": 23.325737015073706, "learning_rate": 4.337355301007335e-07, "logits/chosen": -3.796875, "logits/rejected": -3.71875, "logps/chosen": -740.0, "logps/rejected": -1360.0, "loss": 0.1871, "rewards/accuracies": 0.925000011920929, "rewards/chosen": -2.453125, "rewards/margins": 5.78125, "rewards/rejected": -8.25, "step": 150 }, { "epoch": 0.33472803347280333, "grad_norm": 40.59499239982402, "learning_rate": 4.2087030056579986e-07, "logits/chosen": -3.765625, "logits/rejected": -3.65625, "logps/chosen": -708.0, "logps/rejected": -1296.0, "loss": 0.1955, "rewards/accuracies": 0.8999999761581421, "rewards/chosen": -2.453125, "rewards/margins": 5.5, "rewards/rejected": -7.9375, "step": 160 }, { "epoch": 0.35564853556485354, "grad_norm": 32.891772625367125, "learning_rate": 4.070934040463998e-07, "logits/chosen": -3.703125, "logits/rejected": -3.734375, "logps/chosen": -716.0, "logps/rejected": -1320.0, "loss": 0.2043, "rewards/accuracies": 0.893750011920929, "rewards/chosen": -2.359375, "rewards/margins": 5.375, "rewards/rejected": -7.75, "step": 170 }, { "epoch": 0.37656903765690375, "grad_norm": 21.73830270472661, "learning_rate": 3.9247834624635404e-07, "logits/chosen": -3.765625, "logits/rejected": -3.71875, "logps/chosen": -780.0, "logps/rejected": -1376.0, "loss": 0.1924, "rewards/accuracies": 0.90625, "rewards/chosen": -2.578125, "rewards/margins": 5.84375, "rewards/rejected": -8.4375, "step": 180 }, { "epoch": 0.39748953974895396, "grad_norm": 16.372098376222763, "learning_rate": 3.7710310482256523e-07, "logits/chosen": -3.8125, "logits/rejected": -3.734375, "logps/chosen": -696.0, "logps/rejected": -1344.0, "loss": 0.1693, "rewards/accuracies": 0.893750011920929, "rewards/chosen": -2.234375, "rewards/margins": 5.96875, "rewards/rejected": -8.1875, "step": 190 }, { "epoch": 0.41841004184100417, "grad_norm": 41.66546161927783, "learning_rate": 3.610497133404795e-07, "logits/chosen": -3.703125, "logits/rejected": -3.703125, "logps/chosen": -844.0, "logps/rejected": -1624.0, "loss": 0.1574, "rewards/accuracies": 0.893750011920929, "rewards/chosen": -3.328125, "rewards/margins": 7.625, "rewards/rejected": -11.0, "step": 200 }, { "epoch": 0.41841004184100417, "eval_logits/chosen": -3.765625, "eval_logits/rejected": -3.71875, "eval_logps/chosen": -780.0, "eval_logps/rejected": -1536.0, "eval_loss": 0.15842990577220917, "eval_rewards/accuracies": 0.9296875, "eval_rewards/chosen": -2.796875, "eval_rewards/margins": 7.40625, "eval_rewards/rejected": -10.1875, "eval_runtime": 90.1471, "eval_samples_per_second": 22.186, "eval_steps_per_second": 0.355, "step": 200 }, { "epoch": 0.4393305439330544, "grad_norm": 38.17786401364724, "learning_rate": 3.4440382358952115e-07, "logits/chosen": -3.71875, "logits/rejected": -3.625, "logps/chosen": -728.0, "logps/rejected": -1584.0, "loss": 0.169, "rewards/accuracies": 0.925000011920929, "rewards/chosen": -2.578125, "rewards/margins": 8.125, "rewards/rejected": -10.75, "step": 210 }, { "epoch": 0.4602510460251046, "grad_norm": 28.36201506292477, "learning_rate": 3.272542485937368e-07, "logits/chosen": -3.734375, "logits/rejected": -3.703125, "logps/chosen": -732.0, "logps/rejected": -1560.0, "loss": 0.1756, "rewards/accuracies": 0.8999999761581421, "rewards/chosen": -2.359375, "rewards/margins": 8.0, "rewards/rejected": -10.375, "step": 220 }, { "epoch": 0.4811715481171548, "grad_norm": 37.321244440982746, "learning_rate": 3.096924887558854e-07, "logits/chosen": -3.8125, "logits/rejected": -3.703125, "logps/chosen": -800.0, "logps/rejected": -1656.0, "loss": 0.1563, "rewards/accuracies": 0.90625, "rewards/chosen": -2.78125, "rewards/margins": 7.84375, "rewards/rejected": -10.625, "step": 230 }, { "epoch": 0.502092050209205, "grad_norm": 29.25849606336902, "learning_rate": 2.9181224366319943e-07, "logits/chosen": -3.71875, "logits/rejected": -3.578125, "logps/chosen": -716.0, "logps/rejected": -1440.0, "loss": 0.1714, "rewards/accuracies": 0.918749988079071, "rewards/chosen": -2.328125, "rewards/margins": 7.1875, "rewards/rejected": -9.5625, "step": 240 }, { "epoch": 0.5230125523012552, "grad_norm": 50.281378871196466, "learning_rate": 2.7370891215954565e-07, "logits/chosen": -3.734375, "logits/rejected": -3.65625, "logps/chosen": -760.0, "logps/rejected": -1576.0, "loss": 0.1625, "rewards/accuracies": 0.949999988079071, "rewards/chosen": -2.6875, "rewards/margins": 7.8125, "rewards/rejected": -10.5, "step": 250 }, { "epoch": 0.5439330543933054, "grad_norm": 35.703635573374726, "learning_rate": 2.55479083351317e-07, "logits/chosen": -3.734375, "logits/rejected": -3.640625, "logps/chosen": -800.0, "logps/rejected": -1600.0, "loss": 0.1357, "rewards/accuracies": 0.90625, "rewards/chosen": -3.078125, "rewards/margins": 7.625, "rewards/rejected": -10.6875, "step": 260 }, { "epoch": 0.5648535564853556, "grad_norm": 24.94230756271867, "learning_rate": 2.3722002126275822e-07, "logits/chosen": -3.65625, "logits/rejected": -3.65625, "logps/chosen": -804.0, "logps/rejected": -1544.0, "loss": 0.1701, "rewards/accuracies": 0.90625, "rewards/chosen": -3.296875, "rewards/margins": 7.1875, "rewards/rejected": -10.5, "step": 270 }, { "epoch": 0.5857740585774058, "grad_norm": 28.862452074602317, "learning_rate": 2.19029145890313e-07, "logits/chosen": -3.65625, "logits/rejected": -3.578125, "logps/chosen": -832.0, "logps/rejected": -1520.0, "loss": 0.1453, "rewards/accuracies": 0.9437500238418579, "rewards/chosen": -3.15625, "rewards/margins": 6.625, "rewards/rejected": -9.75, "step": 280 }, { "epoch": 0.606694560669456, "grad_norm": 33.07346655704146, "learning_rate": 2.0100351342479216e-07, "logits/chosen": -3.59375, "logits/rejected": -3.5625, "logps/chosen": -840.0, "logps/rejected": -1592.0, "loss": 0.1752, "rewards/accuracies": 0.9125000238418579, "rewards/chosen": -3.4375, "rewards/margins": 7.3125, "rewards/rejected": -10.75, "step": 290 }, { "epoch": 0.6276150627615062, "grad_norm": 28.81151134715954, "learning_rate": 1.8323929841460178e-07, "logits/chosen": -3.625, "logits/rejected": -3.578125, "logps/chosen": -720.0, "logps/rejected": -1408.0, "loss": 0.1551, "rewards/accuracies": 0.918749988079071, "rewards/chosen": -2.359375, "rewards/margins": 6.5, "rewards/rejected": -8.875, "step": 300 }, { "epoch": 0.6276150627615062, "eval_logits/chosen": -3.640625, "eval_logits/rejected": -3.59375, "eval_logps/chosen": -744.0, "eval_logps/rejected": -1472.0, "eval_loss": 0.13853245973587036, "eval_rewards/accuracies": 0.93359375, "eval_rewards/chosen": -2.453125, "eval_rewards/margins": 7.15625, "eval_rewards/rejected": -9.625, "eval_runtime": 90.005, "eval_samples_per_second": 22.221, "eval_steps_per_second": 0.356, "step": 300 }, { "epoch": 0.6485355648535565, "grad_norm": 39.90299619186908, "learning_rate": 1.6583128063291573e-07, "logits/chosen": -3.59375, "logits/rejected": -3.53125, "logps/chosen": -780.0, "logps/rejected": -1672.0, "loss": 0.1403, "rewards/accuracies": 0.949999988079071, "rewards/chosen": -2.921875, "rewards/margins": 8.5, "rewards/rejected": -11.4375, "step": 310 }, { "epoch": 0.6694560669456067, "grad_norm": 30.707942119385958, "learning_rate": 1.488723393865766e-07, "logits/chosen": -3.609375, "logits/rejected": -3.578125, "logps/chosen": -864.0, "logps/rejected": -1776.0, "loss": 0.1436, "rewards/accuracies": 0.9375, "rewards/chosen": -3.296875, "rewards/margins": 9.0, "rewards/rejected": -12.3125, "step": 320 }, { "epoch": 0.6903765690376569, "grad_norm": 22.500572393034663, "learning_rate": 1.3245295796480788e-07, "logits/chosen": -3.5625, "logits/rejected": -3.484375, "logps/chosen": -888.0, "logps/rejected": -1808.0, "loss": 0.1576, "rewards/accuracies": 0.918749988079071, "rewards/chosen": -3.828125, "rewards/margins": 9.0, "rewards/rejected": -12.8125, "step": 330 }, { "epoch": 0.7112970711297071, "grad_norm": 27.937514221124733, "learning_rate": 1.1666074087171627e-07, "logits/chosen": -3.53125, "logits/rejected": -3.46875, "logps/chosen": -960.0, "logps/rejected": -1920.0, "loss": 0.1881, "rewards/accuracies": 0.9437500238418579, "rewards/chosen": -4.5625, "rewards/margins": 9.3125, "rewards/rejected": -13.875, "step": 340 }, { "epoch": 0.7322175732217573, "grad_norm": 21.646791307337544, "learning_rate": 1.0157994641835734e-07, "logits/chosen": -3.609375, "logits/rejected": -3.515625, "logps/chosen": -808.0, "logps/rejected": -1488.0, "loss": 0.1677, "rewards/accuracies": 0.893750011920929, "rewards/chosen": -3.421875, "rewards/margins": 6.4375, "rewards/rejected": -9.875, "step": 350 }, { "epoch": 0.7531380753138075, "grad_norm": 26.51641238698488, "learning_rate": 8.729103716819111e-08, "logits/chosen": -3.515625, "logits/rejected": -3.4375, "logps/chosen": -804.0, "logps/rejected": -1520.0, "loss": 0.154, "rewards/accuracies": 0.9437500238418579, "rewards/chosen": -2.9375, "rewards/margins": 6.875, "rewards/rejected": -9.8125, "step": 360 }, { "epoch": 0.7740585774058577, "grad_norm": 32.61897967589747, "learning_rate": 7.387025063449081e-08, "logits/chosen": -3.59375, "logits/rejected": -3.5, "logps/chosen": -792.0, "logps/rejected": -1480.0, "loss": 0.1439, "rewards/accuracies": 0.8999999761581421, "rewards/chosen": -3.28125, "rewards/margins": 6.46875, "rewards/rejected": -9.75, "step": 370 }, { "epoch": 0.7949790794979079, "grad_norm": 24.073566416356307, "learning_rate": 6.138919252022435e-08, "logits/chosen": -3.59375, "logits/rejected": -3.53125, "logps/chosen": -852.0, "logps/rejected": -1696.0, "loss": 0.1474, "rewards/accuracies": 0.9437500238418579, "rewards/chosen": -3.5, "rewards/margins": 7.875, "rewards/rejected": -11.375, "step": 380 }, { "epoch": 0.8158995815899581, "grad_norm": 22.024234097844488, "learning_rate": 4.991445467064689e-08, "logits/chosen": -3.625, "logits/rejected": -3.515625, "logps/chosen": -820.0, "logps/rejected": -1672.0, "loss": 0.155, "rewards/accuracies": 0.956250011920929, "rewards/chosen": -3.421875, "rewards/margins": 7.90625, "rewards/rejected": -11.375, "step": 390 }, { "epoch": 0.8368200836820083, "grad_norm": 22.5208730955551, "learning_rate": 3.9507259776993954e-08, "logits/chosen": -3.53125, "logits/rejected": -3.46875, "logps/chosen": -816.0, "logps/rejected": -1592.0, "loss": 0.1511, "rewards/accuracies": 0.9125000238418579, "rewards/chosen": -3.421875, "rewards/margins": 7.3125, "rewards/rejected": -10.6875, "step": 400 }, { "epoch": 0.8368200836820083, "eval_logits/chosen": -3.5625, "eval_logits/rejected": -3.5, "eval_logps/chosen": -836.0, "eval_logps/rejected": -1592.0, "eval_loss": 0.13080458343029022, "eval_rewards/accuracies": 0.94921875, "eval_rewards/chosen": -3.34375, "eval_rewards/margins": 7.46875, "eval_rewards/rejected": -10.8125, "eval_runtime": 90.1024, "eval_samples_per_second": 22.197, "eval_steps_per_second": 0.355, "step": 400 }, { "epoch": 0.8577405857740585, "grad_norm": 33.05531795103649, "learning_rate": 3.022313472693447e-08, "logits/chosen": -3.609375, "logits/rejected": -3.53125, "logps/chosen": -836.0, "logps/rejected": -1672.0, "loss": 0.1302, "rewards/accuracies": 0.9312499761581421, "rewards/chosen": -3.5, "rewards/margins": 7.78125, "rewards/rejected": -11.25, "step": 410 }, { "epoch": 0.8786610878661087, "grad_norm": 22.39150376227694, "learning_rate": 2.2111614344599684e-08, "logits/chosen": -3.515625, "logits/rejected": -3.4375, "logps/chosen": -820.0, "logps/rejected": -1640.0, "loss": 0.1296, "rewards/accuracies": 0.925000011920929, "rewards/chosen": -3.3125, "rewards/margins": 7.78125, "rewards/rejected": -11.0625, "step": 420 }, { "epoch": 0.899581589958159, "grad_norm": 34.532509934361165, "learning_rate": 1.521597710086439e-08, "logits/chosen": -3.625, "logits/rejected": -3.5, "logps/chosen": -804.0, "logps/rejected": -1648.0, "loss": 0.138, "rewards/accuracies": 0.925000011920929, "rewards/chosen": -3.125, "rewards/margins": 7.875, "rewards/rejected": -11.0, "step": 430 }, { "epoch": 0.9205020920502092, "grad_norm": 33.11707475610062, "learning_rate": 9.57301420397924e-09, "logits/chosen": -3.609375, "logits/rejected": -3.5625, "logps/chosen": -844.0, "logps/rejected": -1632.0, "loss": 0.1543, "rewards/accuracies": 0.9125000238418579, "rewards/chosen": -3.46875, "rewards/margins": 7.5625, "rewards/rejected": -11.0625, "step": 440 }, { "epoch": 0.9414225941422594, "grad_norm": 46.292590138732955, "learning_rate": 5.212833302556258e-09, "logits/chosen": -3.515625, "logits/rejected": -3.5, "logps/chosen": -824.0, "logps/rejected": -1632.0, "loss": 0.1439, "rewards/accuracies": 0.9312499761581421, "rewards/chosen": -3.53125, "rewards/margins": 7.75, "rewards/rejected": -11.25, "step": 450 }, { "epoch": 0.9623430962343096, "grad_norm": 13.872523653141515, "learning_rate": 2.158697848236607e-09, "logits/chosen": -3.546875, "logits/rejected": -3.46875, "logps/chosen": -864.0, "logps/rejected": -1792.0, "loss": 0.1534, "rewards/accuracies": 0.9437500238418579, "rewards/chosen": -3.34375, "rewards/margins": 9.0, "rewards/rejected": -12.3125, "step": 460 }, { "epoch": 0.9832635983263598, "grad_norm": 40.836340138155315, "learning_rate": 4.269029751107489e-10, "logits/chosen": -3.609375, "logits/rejected": -3.46875, "logps/chosen": -784.0, "logps/rejected": -1576.0, "loss": 0.1607, "rewards/accuracies": 0.90625, "rewards/chosen": -3.234375, "rewards/margins": 7.625, "rewards/rejected": -10.875, "step": 470 }, { "epoch": 1.0, "step": 478, "total_flos": 0.0, "train_loss": 0.251720839215123, "train_runtime": 7290.7659, "train_samples_per_second": 8.385, "train_steps_per_second": 0.066 } ], "logging_steps": 10, "max_steps": 478, "num_input_tokens_seen": 0, "num_train_epochs": 1, "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": 0.0, "train_batch_size": 8, "trial_name": null, "trial_params": null }