| { | |
| "best_global_step": 52, | |
| "best_metric": 1.0, | |
| "best_model_checkpoint": "/home/jovyan/RLHFBackdoor/alignment-handbook/save/llama-8b-instruct-medical-reward-10epochs-2e-5-64-128-1024/checkpoint-52", | |
| "epoch": 9.347826086956522, | |
| "eval_steps": 4, | |
| "global_step": 140, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.06956521739130435, | |
| "grad_norm": 16.051664352416992, | |
| "learning_rate": 0.0, | |
| "loss": 0.5155, | |
| "step": 1 | |
| }, | |
| { | |
| "epoch": 0.1391304347826087, | |
| "grad_norm": 16.695077896118164, | |
| "learning_rate": 1.3333333333333334e-06, | |
| "loss": 0.5624, | |
| "step": 2 | |
| }, | |
| { | |
| "epoch": 0.20869565217391303, | |
| "grad_norm": 20.069669723510742, | |
| "learning_rate": 2.666666666666667e-06, | |
| "loss": 0.4644, | |
| "step": 3 | |
| }, | |
| { | |
| "epoch": 0.2782608695652174, | |
| "grad_norm": 18.166921615600586, | |
| "learning_rate": 4.000000000000001e-06, | |
| "loss": 0.52, | |
| "step": 4 | |
| }, | |
| { | |
| "epoch": 0.2782608695652174, | |
| "eval_accuracy": 0.82, | |
| "eval_chosen_reward_mean": 0.7202807664871216, | |
| "eval_loss": 0.44523465633392334, | |
| "eval_rejected_reward_mean": 0.2797192633152008, | |
| "eval_reward_diff_mean": 0.4405614733695984, | |
| "eval_reward_diff_std": 0.48559921979904175, | |
| "eval_runtime": 4.4816, | |
| "eval_samples_per_second": 22.983, | |
| "eval_steps_per_second": 5.802, | |
| "step": 4 | |
| }, | |
| { | |
| "epoch": 0.34782608695652173, | |
| "grad_norm": 18.02265167236328, | |
| "learning_rate": 5.333333333333334e-06, | |
| "loss": 0.3831, | |
| "step": 5 | |
| }, | |
| { | |
| "epoch": 0.41739130434782606, | |
| "grad_norm": 19.39177894592285, | |
| "learning_rate": 6.666666666666667e-06, | |
| "loss": 0.5606, | |
| "step": 6 | |
| }, | |
| { | |
| "epoch": 0.48695652173913045, | |
| "grad_norm": 21.10468292236328, | |
| "learning_rate": 8.000000000000001e-06, | |
| "loss": 0.4886, | |
| "step": 7 | |
| }, | |
| { | |
| "epoch": 0.5565217391304348, | |
| "grad_norm": 15.2007417678833, | |
| "learning_rate": 9.333333333333334e-06, | |
| "loss": 0.432, | |
| "step": 8 | |
| }, | |
| { | |
| "epoch": 0.5565217391304348, | |
| "eval_accuracy": 0.86, | |
| "eval_chosen_reward_mean": 0.7534962296485901, | |
| "eval_loss": 0.39489009976387024, | |
| "eval_rejected_reward_mean": 0.2465037852525711, | |
| "eval_reward_diff_mean": 0.5069924592971802, | |
| "eval_reward_diff_std": 0.48188310861587524, | |
| "eval_runtime": 4.5265, | |
| "eval_samples_per_second": 22.755, | |
| "eval_steps_per_second": 5.744, | |
| "step": 8 | |
| }, | |
| { | |
| "epoch": 0.6260869565217392, | |
| "grad_norm": 14.061135292053223, | |
| "learning_rate": 1.0666666666666667e-05, | |
| "loss": 0.3212, | |
| "step": 9 | |
| }, | |
| { | |
| "epoch": 0.6956521739130435, | |
| "grad_norm": 16.22977638244629, | |
| "learning_rate": 1.2e-05, | |
| "loss": 0.4953, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.7652173913043478, | |
| "grad_norm": 9.790153503417969, | |
| "learning_rate": 1.3333333333333333e-05, | |
| "loss": 0.314, | |
| "step": 11 | |
| }, | |
| { | |
| "epoch": 0.8347826086956521, | |
| "grad_norm": 9.781067848205566, | |
| "learning_rate": 1.4666666666666666e-05, | |
| "loss": 0.3011, | |
| "step": 12 | |
| }, | |
| { | |
| "epoch": 0.8347826086956521, | |
| "eval_accuracy": 0.91, | |
| "eval_chosen_reward_mean": 0.8568271398544312, | |
| "eval_loss": 0.2553744912147522, | |
| "eval_rejected_reward_mean": 0.14317283034324646, | |
| "eval_reward_diff_mean": 0.7136543989181519, | |
| "eval_reward_diff_std": 0.46591508388519287, | |
| "eval_runtime": 4.4565, | |
| "eval_samples_per_second": 23.113, | |
| "eval_steps_per_second": 5.834, | |
| "step": 12 | |
| }, | |
| { | |
| "epoch": 0.9043478260869565, | |
| "grad_norm": 6.0656208992004395, | |
| "learning_rate": 1.6000000000000003e-05, | |
| "loss": 0.2098, | |
| "step": 13 | |
| }, | |
| { | |
| "epoch": 0.9739130434782609, | |
| "grad_norm": 4.201033115386963, | |
| "learning_rate": 1.7333333333333336e-05, | |
| "loss": 0.1864, | |
| "step": 14 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "grad_norm": 3.1622679233551025, | |
| "learning_rate": 1.866666666666667e-05, | |
| "loss": 0.1906, | |
| "step": 15 | |
| }, | |
| { | |
| "epoch": 1.0695652173913044, | |
| "grad_norm": 2.8981573581695557, | |
| "learning_rate": 2e-05, | |
| "loss": 0.1935, | |
| "step": 16 | |
| }, | |
| { | |
| "epoch": 1.0695652173913044, | |
| "eval_accuracy": 0.91, | |
| "eval_chosen_reward_mean": 0.9146673083305359, | |
| "eval_loss": 0.18809932470321655, | |
| "eval_rejected_reward_mean": 0.08533263951539993, | |
| "eval_reward_diff_mean": 0.8293347358703613, | |
| "eval_reward_diff_std": 0.4708438813686371, | |
| "eval_runtime": 4.4838, | |
| "eval_samples_per_second": 22.972, | |
| "eval_steps_per_second": 5.799, | |
| "step": 16 | |
| }, | |
| { | |
| "epoch": 1.1391304347826088, | |
| "grad_norm": 1.4406570196151733, | |
| "learning_rate": 1.999729241179462e-05, | |
| "loss": 0.0381, | |
| "step": 17 | |
| }, | |
| { | |
| "epoch": 1.208695652173913, | |
| "grad_norm": 1.6988517045974731, | |
| "learning_rate": 1.998917111338525e-05, | |
| "loss": 0.0896, | |
| "step": 18 | |
| }, | |
| { | |
| "epoch": 1.2782608695652173, | |
| "grad_norm": 1.7537212371826172, | |
| "learning_rate": 1.9975640502598243e-05, | |
| "loss": 0.0829, | |
| "step": 19 | |
| }, | |
| { | |
| "epoch": 1.3478260869565217, | |
| "grad_norm": 2.655623435974121, | |
| "learning_rate": 1.9956707906498046e-05, | |
| "loss": 0.1333, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 1.3478260869565217, | |
| "eval_accuracy": 0.91, | |
| "eval_chosen_reward_mean": 0.9230502247810364, | |
| "eval_loss": 0.15013982355594635, | |
| "eval_rejected_reward_mean": 0.07694970071315765, | |
| "eval_reward_diff_mean": 0.8461005687713623, | |
| "eval_reward_diff_std": 0.4364628195762634, | |
| "eval_runtime": 4.4752, | |
| "eval_samples_per_second": 23.016, | |
| "eval_steps_per_second": 5.81, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 1.4173913043478261, | |
| "grad_norm": 2.136287212371826, | |
| "learning_rate": 1.9932383577419432e-05, | |
| "loss": 0.1375, | |
| "step": 21 | |
| }, | |
| { | |
| "epoch": 1.4869565217391305, | |
| "grad_norm": 1.613239049911499, | |
| "learning_rate": 1.9902680687415704e-05, | |
| "loss": 0.0605, | |
| "step": 22 | |
| }, | |
| { | |
| "epoch": 1.5565217391304347, | |
| "grad_norm": 2.315011978149414, | |
| "learning_rate": 1.9867615321125796e-05, | |
| "loss": 0.0821, | |
| "step": 23 | |
| }, | |
| { | |
| "epoch": 1.626086956521739, | |
| "grad_norm": 4.691528797149658, | |
| "learning_rate": 1.9827206467064133e-05, | |
| "loss": 0.141, | |
| "step": 24 | |
| }, | |
| { | |
| "epoch": 1.626086956521739, | |
| "eval_accuracy": 0.95, | |
| "eval_chosen_reward_mean": 0.9383553862571716, | |
| "eval_loss": 0.09369319677352905, | |
| "eval_rejected_reward_mean": 0.0616445392370224, | |
| "eval_reward_diff_mean": 0.8767109513282776, | |
| "eval_reward_diff_std": 0.3437209129333496, | |
| "eval_runtime": 4.4716, | |
| "eval_samples_per_second": 23.034, | |
| "eval_steps_per_second": 5.814, | |
| "step": 24 | |
| }, | |
| { | |
| "epoch": 1.6956521739130435, | |
| "grad_norm": 0.7377265691757202, | |
| "learning_rate": 1.9781476007338058e-05, | |
| "loss": 0.0161, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 1.7652173913043478, | |
| "grad_norm": 2.014756202697754, | |
| "learning_rate": 1.973044870579824e-05, | |
| "loss": 0.1014, | |
| "step": 26 | |
| }, | |
| { | |
| "epoch": 1.8347826086956522, | |
| "grad_norm": 3.497450590133667, | |
| "learning_rate": 1.967415219462864e-05, | |
| "loss": 0.1077, | |
| "step": 27 | |
| }, | |
| { | |
| "epoch": 1.9043478260869566, | |
| "grad_norm": 2.395733118057251, | |
| "learning_rate": 1.961261695938319e-05, | |
| "loss": 0.0922, | |
| "step": 28 | |
| }, | |
| { | |
| "epoch": 1.9043478260869566, | |
| "eval_accuracy": 0.96, | |
| "eval_chosen_reward_mean": 0.9490196108818054, | |
| "eval_loss": 0.07248755544424057, | |
| "eval_rejected_reward_mean": 0.050980377942323685, | |
| "eval_reward_diff_mean": 0.8980393409729004, | |
| "eval_reward_diff_std": 0.2919844388961792, | |
| "eval_runtime": 4.469, | |
| "eval_samples_per_second": 23.048, | |
| "eval_steps_per_second": 5.818, | |
| "step": 28 | |
| }, | |
| { | |
| "epoch": 1.973913043478261, | |
| "grad_norm": 3.5625710487365723, | |
| "learning_rate": 1.954587632247732e-05, | |
| "loss": 0.163, | |
| "step": 29 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 3.5020124912261963, | |
| "learning_rate": 1.9473966425143292e-05, | |
| "loss": 0.1025, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 2.0695652173913044, | |
| "grad_norm": 2.214233875274658, | |
| "learning_rate": 1.9396926207859085e-05, | |
| "loss": 0.0769, | |
| "step": 31 | |
| }, | |
| { | |
| "epoch": 2.139130434782609, | |
| "grad_norm": 1.038892149925232, | |
| "learning_rate": 1.9314797389261426e-05, | |
| "loss": 0.0322, | |
| "step": 32 | |
| }, | |
| { | |
| "epoch": 2.139130434782609, | |
| "eval_accuracy": 0.95, | |
| "eval_chosen_reward_mean": 0.9458566308021545, | |
| "eval_loss": 0.0769653245806694, | |
| "eval_rejected_reward_mean": 0.05414345860481262, | |
| "eval_reward_diff_mean": 0.8917130827903748, | |
| "eval_reward_diff_std": 0.30557724833488464, | |
| "eval_runtime": 4.467, | |
| "eval_samples_per_second": 23.058, | |
| "eval_steps_per_second": 5.821, | |
| "step": 32 | |
| }, | |
| { | |
| "epoch": 2.208695652173913, | |
| "grad_norm": 1.216461181640625, | |
| "learning_rate": 1.9227624443554425e-05, | |
| "loss": 0.0345, | |
| "step": 33 | |
| }, | |
| { | |
| "epoch": 2.2782608695652176, | |
| "grad_norm": 1.8307952880859375, | |
| "learning_rate": 1.913545457642601e-05, | |
| "loss": 0.0863, | |
| "step": 34 | |
| }, | |
| { | |
| "epoch": 2.3478260869565215, | |
| "grad_norm": 1.3736193180084229, | |
| "learning_rate": 1.9038337699485207e-05, | |
| "loss": 0.0789, | |
| "step": 35 | |
| }, | |
| { | |
| "epoch": 2.417391304347826, | |
| "grad_norm": 2.5011188983917236, | |
| "learning_rate": 1.8936326403234125e-05, | |
| "loss": 0.0864, | |
| "step": 36 | |
| }, | |
| { | |
| "epoch": 2.417391304347826, | |
| "eval_accuracy": 0.96, | |
| "eval_chosen_reward_mean": 0.9428873658180237, | |
| "eval_loss": 0.08493011444807053, | |
| "eval_rejected_reward_mean": 0.05711263790726662, | |
| "eval_reward_diff_mean": 0.8857746720314026, | |
| "eval_reward_diff_std": 0.3248811662197113, | |
| "eval_runtime": 4.493, | |
| "eval_samples_per_second": 22.925, | |
| "eval_steps_per_second": 5.787, | |
| "step": 36 | |
| }, | |
| { | |
| "epoch": 2.4869565217391303, | |
| "grad_norm": 1.3938976526260376, | |
| "learning_rate": 1.8829475928589272e-05, | |
| "loss": 0.0321, | |
| "step": 37 | |
| }, | |
| { | |
| "epoch": 2.5565217391304347, | |
| "grad_norm": 2.110109567642212, | |
| "learning_rate": 1.8717844136967626e-05, | |
| "loss": 0.0744, | |
| "step": 38 | |
| }, | |
| { | |
| "epoch": 2.626086956521739, | |
| "grad_norm": 2.0099246501922607, | |
| "learning_rate": 1.860149147895366e-05, | |
| "loss": 0.0811, | |
| "step": 39 | |
| }, | |
| { | |
| "epoch": 2.6956521739130435, | |
| "grad_norm": 1.074698567390442, | |
| "learning_rate": 1.848048096156426e-05, | |
| "loss": 0.028, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 2.6956521739130435, | |
| "eval_accuracy": 0.97, | |
| "eval_chosen_reward_mean": 0.9522405862808228, | |
| "eval_loss": 0.0649605244398117, | |
| "eval_rejected_reward_mean": 0.04775938019156456, | |
| "eval_reward_diff_mean": 0.9044812917709351, | |
| "eval_reward_diff_std": 0.2758881151676178, | |
| "eval_runtime": 4.514, | |
| "eval_samples_per_second": 22.818, | |
| "eval_steps_per_second": 5.76, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 2.765217391304348, | |
| "grad_norm": 0.8246477246284485, | |
| "learning_rate": 1.8354878114129368e-05, | |
| "loss": 0.0278, | |
| "step": 41 | |
| }, | |
| { | |
| "epoch": 2.8347826086956522, | |
| "grad_norm": 0.8497962355613708, | |
| "learning_rate": 1.8224750952806626e-05, | |
| "loss": 0.015, | |
| "step": 42 | |
| }, | |
| { | |
| "epoch": 2.9043478260869566, | |
| "grad_norm": 0.3047845959663391, | |
| "learning_rate": 1.8090169943749477e-05, | |
| "loss": 0.006, | |
| "step": 43 | |
| }, | |
| { | |
| "epoch": 2.973913043478261, | |
| "grad_norm": 0.9201741814613342, | |
| "learning_rate": 1.795120796494848e-05, | |
| "loss": 0.0325, | |
| "step": 44 | |
| }, | |
| { | |
| "epoch": 2.973913043478261, | |
| "eval_accuracy": 0.98, | |
| "eval_chosen_reward_mean": 0.9637383222579956, | |
| "eval_loss": 0.0462159626185894, | |
| "eval_rejected_reward_mean": 0.03626168891787529, | |
| "eval_reward_diff_mean": 0.9274765849113464, | |
| "eval_reward_diff_std": 0.221104234457016, | |
| "eval_runtime": 4.5409, | |
| "eval_samples_per_second": 22.683, | |
| "eval_steps_per_second": 5.726, | |
| "step": 44 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 4.132351398468018, | |
| "learning_rate": 1.7807940266766595e-05, | |
| "loss": 0.0737, | |
| "step": 45 | |
| }, | |
| { | |
| "epoch": 3.0695652173913044, | |
| "grad_norm": 0.9027279615402222, | |
| "learning_rate": 1.766044443118978e-05, | |
| "loss": 0.0225, | |
| "step": 46 | |
| }, | |
| { | |
| "epoch": 3.139130434782609, | |
| "grad_norm": 0.35845667123794556, | |
| "learning_rate": 1.7508800329814993e-05, | |
| "loss": 0.0078, | |
| "step": 47 | |
| }, | |
| { | |
| "epoch": 3.208695652173913, | |
| "grad_norm": 0.2947898209095001, | |
| "learning_rate": 1.735309008059829e-05, | |
| "loss": 0.0059, | |
| "step": 48 | |
| }, | |
| { | |
| "epoch": 3.208695652173913, | |
| "eval_accuracy": 0.99, | |
| "eval_chosen_reward_mean": 0.9699317216873169, | |
| "eval_loss": 0.037224188446998596, | |
| "eval_rejected_reward_mean": 0.030068309977650642, | |
| "eval_reward_diff_mean": 0.9398634433746338, | |
| "eval_reward_diff_std": 0.191928893327713, | |
| "eval_runtime": 4.5026, | |
| "eval_samples_per_second": 22.876, | |
| "eval_steps_per_second": 5.774, | |
| "step": 48 | |
| }, | |
| { | |
| "epoch": 3.2782608695652176, | |
| "grad_norm": 0.519256055355072, | |
| "learning_rate": 1.7193398003386514e-05, | |
| "loss": 0.0106, | |
| "step": 49 | |
| }, | |
| { | |
| "epoch": 3.3478260869565215, | |
| "grad_norm": 0.2725602984428406, | |
| "learning_rate": 1.702981057425662e-05, | |
| "loss": 0.0072, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 3.417391304347826, | |
| "grad_norm": 0.8170077204704285, | |
| "learning_rate": 1.686241637868734e-05, | |
| "loss": 0.0118, | |
| "step": 51 | |
| }, | |
| { | |
| "epoch": 3.4869565217391303, | |
| "grad_norm": 0.6010556817054749, | |
| "learning_rate": 1.6691306063588583e-05, | |
| "loss": 0.011, | |
| "step": 52 | |
| }, | |
| { | |
| "epoch": 3.4869565217391303, | |
| "eval_accuracy": 1.0, | |
| "eval_chosen_reward_mean": 0.9803745746612549, | |
| "eval_loss": 0.02279542200267315, | |
| "eval_rejected_reward_mean": 0.019625453278422356, | |
| "eval_reward_diff_mean": 0.9607490301132202, | |
| "eval_reward_diff_std": 0.13550853729248047, | |
| "eval_runtime": 4.5356, | |
| "eval_samples_per_second": 22.709, | |
| "eval_steps_per_second": 5.732, | |
| "step": 52 | |
| }, | |
| { | |
| "epoch": 3.5565217391304347, | |
| "grad_norm": 0.6597536206245422, | |
| "learning_rate": 1.6516572288214555e-05, | |
| "loss": 0.0152, | |
| "step": 53 | |
| }, | |
| { | |
| "epoch": 3.626086956521739, | |
| "grad_norm": 0.24238398671150208, | |
| "learning_rate": 1.63383096739871e-05, | |
| "loss": 0.0041, | |
| "step": 54 | |
| }, | |
| { | |
| "epoch": 3.6956521739130435, | |
| "grad_norm": 0.042787156999111176, | |
| "learning_rate": 1.6156614753256583e-05, | |
| "loss": 0.0004, | |
| "step": 55 | |
| }, | |
| { | |
| "epoch": 3.765217391304348, | |
| "grad_norm": 0.5205024480819702, | |
| "learning_rate": 1.5971585917027864e-05, | |
| "loss": 0.0058, | |
| "step": 56 | |
| }, | |
| { | |
| "epoch": 3.765217391304348, | |
| "eval_accuracy": 1.0, | |
| "eval_chosen_reward_mean": 0.9882292151451111, | |
| "eval_loss": 0.012967481277883053, | |
| "eval_rejected_reward_mean": 0.011770866811275482, | |
| "eval_reward_diff_mean": 0.9764581322669983, | |
| "eval_reward_diff_std": 0.087115578353405, | |
| "eval_runtime": 4.5199, | |
| "eval_samples_per_second": 22.788, | |
| "eval_steps_per_second": 5.752, | |
| "step": 56 | |
| }, | |
| { | |
| "epoch": 3.8347826086956522, | |
| "grad_norm": 0.09940219670534134, | |
| "learning_rate": 1.5783323361679865e-05, | |
| "loss": 0.0009, | |
| "step": 57 | |
| }, | |
| { | |
| "epoch": 3.9043478260869566, | |
| "grad_norm": 0.14570941030979156, | |
| "learning_rate": 1.5591929034707468e-05, | |
| "loss": 0.0028, | |
| "step": 58 | |
| }, | |
| { | |
| "epoch": 3.973913043478261, | |
| "grad_norm": 0.1424681395292282, | |
| "learning_rate": 1.539750657951513e-05, | |
| "loss": 0.0026, | |
| "step": 59 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 0.47905346751213074, | |
| "learning_rate": 1.5200161279292154e-05, | |
| "loss": 0.0052, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 1.0, | |
| "eval_chosen_reward_mean": 0.9917762875556946, | |
| "eval_loss": 0.008945059031248093, | |
| "eval_rejected_reward_mean": 0.008223675191402435, | |
| "eval_reward_diff_mean": 0.9835525751113892, | |
| "eval_reward_diff_std": 0.06846771389245987, | |
| "eval_runtime": 4.546, | |
| "eval_samples_per_second": 22.657, | |
| "eval_steps_per_second": 5.719, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 4.069565217391304, | |
| "grad_norm": 0.19887447357177734, | |
| "learning_rate": 1.5000000000000002e-05, | |
| "loss": 0.0026, | |
| "step": 61 | |
| }, | |
| { | |
| "epoch": 4.139130434782609, | |
| "grad_norm": 0.14599338173866272, | |
| "learning_rate": 1.4797131132502464e-05, | |
| "loss": 0.0024, | |
| "step": 62 | |
| }, | |
| { | |
| "epoch": 4.208695652173913, | |
| "grad_norm": 0.034447040408849716, | |
| "learning_rate": 1.4591664533870118e-05, | |
| "loss": 0.0006, | |
| "step": 63 | |
| }, | |
| { | |
| "epoch": 4.278260869565218, | |
| "grad_norm": 0.09839093685150146, | |
| "learning_rate": 1.4383711467890776e-05, | |
| "loss": 0.0015, | |
| "step": 64 | |
| }, | |
| { | |
| "epoch": 4.278260869565218, | |
| "eval_accuracy": 1.0, | |
| "eval_chosen_reward_mean": 0.994703471660614, | |
| "eval_loss": 0.005607697181403637, | |
| "eval_rejected_reward_mean": 0.005296613555401564, | |
| "eval_reward_diff_mean": 0.9894067645072937, | |
| "eval_reward_diff_std": 0.04620999097824097, | |
| "eval_runtime": 4.4844, | |
| "eval_samples_per_second": 22.969, | |
| "eval_steps_per_second": 5.798, | |
| "step": 64 | |
| }, | |
| { | |
| "epoch": 4.3478260869565215, | |
| "grad_norm": 0.034802138805389404, | |
| "learning_rate": 1.417338454481818e-05, | |
| "loss": 0.0006, | |
| "step": 65 | |
| }, | |
| { | |
| "epoch": 4.417391304347826, | |
| "grad_norm": 0.04835907369852066, | |
| "learning_rate": 1.396079766039157e-05, | |
| "loss": 0.0007, | |
| "step": 66 | |
| }, | |
| { | |
| "epoch": 4.48695652173913, | |
| "grad_norm": 0.035097040235996246, | |
| "learning_rate": 1.3746065934159123e-05, | |
| "loss": 0.0007, | |
| "step": 67 | |
| }, | |
| { | |
| "epoch": 4.556521739130435, | |
| "grad_norm": 0.07052204012870789, | |
| "learning_rate": 1.3529305647138689e-05, | |
| "loss": 0.0009, | |
| "step": 68 | |
| }, | |
| { | |
| "epoch": 4.556521739130435, | |
| "eval_accuracy": 1.0, | |
| "eval_chosen_reward_mean": 0.9965487718582153, | |
| "eval_loss": 0.003575300332158804, | |
| "eval_rejected_reward_mean": 0.0034511922858655453, | |
| "eval_reward_diff_mean": 0.9930976033210754, | |
| "eval_reward_diff_std": 0.029713084921240807, | |
| "eval_runtime": 4.4807, | |
| "eval_samples_per_second": 22.988, | |
| "eval_steps_per_second": 5.803, | |
| "step": 68 | |
| }, | |
| { | |
| "epoch": 4.626086956521739, | |
| "grad_norm": 0.029114194214344025, | |
| "learning_rate": 1.3310634178849583e-05, | |
| "loss": 0.0005, | |
| "step": 69 | |
| }, | |
| { | |
| "epoch": 4.695652173913043, | |
| "grad_norm": 0.008723776787519455, | |
| "learning_rate": 1.3090169943749475e-05, | |
| "loss": 0.0001, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 4.765217391304348, | |
| "grad_norm": 0.022413769736886024, | |
| "learning_rate": 1.2868032327110904e-05, | |
| "loss": 0.0005, | |
| "step": 71 | |
| }, | |
| { | |
| "epoch": 4.834782608695652, | |
| "grad_norm": 0.006887242663651705, | |
| "learning_rate": 1.2644341620372025e-05, | |
| "loss": 0.0001, | |
| "step": 72 | |
| }, | |
| { | |
| "epoch": 4.834782608695652, | |
| "eval_accuracy": 1.0, | |
| "eval_chosen_reward_mean": 0.997631311416626, | |
| "eval_loss": 0.0024212903808802366, | |
| "eval_rejected_reward_mean": 0.002368778921663761, | |
| "eval_reward_diff_mean": 0.9952623844146729, | |
| "eval_reward_diff_std": 0.019513826817274094, | |
| "eval_runtime": 4.6718, | |
| "eval_samples_per_second": 22.047, | |
| "eval_steps_per_second": 5.565, | |
| "step": 72 | |
| }, | |
| { | |
| "epoch": 4.904347826086957, | |
| "grad_norm": 0.03990592807531357, | |
| "learning_rate": 1.2419218955996677e-05, | |
| "loss": 0.0008, | |
| "step": 73 | |
| }, | |
| { | |
| "epoch": 4.973913043478261, | |
| "grad_norm": 0.01335582323372364, | |
| "learning_rate": 1.2192786241879033e-05, | |
| "loss": 0.0002, | |
| "step": 74 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 0.004690114408731461, | |
| "learning_rate": 1.1965166095328302e-05, | |
| "loss": 0.0001, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 5.069565217391304, | |
| "grad_norm": 0.00890415720641613, | |
| "learning_rate": 1.1736481776669307e-05, | |
| "loss": 0.0001, | |
| "step": 76 | |
| }, | |
| { | |
| "epoch": 5.069565217391304, | |
| "eval_accuracy": 1.0, | |
| "eval_chosen_reward_mean": 0.9981046915054321, | |
| "eval_loss": 0.0019279808038845658, | |
| "eval_rejected_reward_mean": 0.001895188121125102, | |
| "eval_reward_diff_mean": 0.9962095618247986, | |
| "eval_reward_diff_std": 0.015480460599064827, | |
| "eval_runtime": 4.5162, | |
| "eval_samples_per_second": 22.807, | |
| "eval_steps_per_second": 5.757, | |
| "step": 76 | |
| }, | |
| { | |
| "epoch": 5.139130434782609, | |
| "grad_norm": 0.03652049973607063, | |
| "learning_rate": 1.1506857122494832e-05, | |
| "loss": 0.0007, | |
| "step": 77 | |
| }, | |
| { | |
| "epoch": 5.208695652173913, | |
| "grad_norm": 0.0075807636603713036, | |
| "learning_rate": 1.127641647860595e-05, | |
| "loss": 0.0001, | |
| "step": 78 | |
| }, | |
| { | |
| "epoch": 5.278260869565218, | |
| "grad_norm": 0.0193843524903059, | |
| "learning_rate": 1.1045284632676535e-05, | |
| "loss": 0.0004, | |
| "step": 79 | |
| }, | |
| { | |
| "epoch": 5.3478260869565215, | |
| "grad_norm": 0.025827020406723022, | |
| "learning_rate": 1.0813586746678584e-05, | |
| "loss": 0.0004, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 5.3478260869565215, | |
| "eval_accuracy": 1.0, | |
| "eval_chosen_reward_mean": 0.998360276222229, | |
| "eval_loss": 0.0016629365272819996, | |
| "eval_rejected_reward_mean": 0.0016397041035816073, | |
| "eval_reward_diff_mean": 0.996720552444458, | |
| "eval_reward_diff_std": 0.013053392991423607, | |
| "eval_runtime": 4.4789, | |
| "eval_samples_per_second": 22.997, | |
| "eval_steps_per_second": 5.805, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 5.417391304347826, | |
| "grad_norm": 0.01660292036831379, | |
| "learning_rate": 1.0581448289104759e-05, | |
| "loss": 0.0003, | |
| "step": 81 | |
| }, | |
| { | |
| "epoch": 5.48695652173913, | |
| "grad_norm": 0.01648598164319992, | |
| "learning_rate": 1.0348994967025012e-05, | |
| "loss": 0.0003, | |
| "step": 82 | |
| }, | |
| { | |
| "epoch": 5.556521739130435, | |
| "grad_norm": 0.010060160420835018, | |
| "learning_rate": 1.0116352658013973e-05, | |
| "loss": 0.0002, | |
| "step": 83 | |
| }, | |
| { | |
| "epoch": 5.626086956521739, | |
| "grad_norm": 0.014060553163290024, | |
| "learning_rate": 9.883647341986032e-06, | |
| "loss": 0.0002, | |
| "step": 84 | |
| }, | |
| { | |
| "epoch": 5.626086956521739, | |
| "eval_accuracy": 1.0, | |
| "eval_chosen_reward_mean": 0.9985561370849609, | |
| "eval_loss": 0.00146092486102134, | |
| "eval_rejected_reward_mean": 0.0014438106445595622, | |
| "eval_reward_diff_mean": 0.9971124529838562, | |
| "eval_reward_diff_std": 0.011214913800358772, | |
| "eval_runtime": 4.4801, | |
| "eval_samples_per_second": 22.99, | |
| "eval_steps_per_second": 5.803, | |
| "step": 84 | |
| }, | |
| { | |
| "epoch": 5.695652173913043, | |
| "grad_norm": 0.06678416579961777, | |
| "learning_rate": 9.651005032974994e-06, | |
| "loss": 0.0007, | |
| "step": 85 | |
| }, | |
| { | |
| "epoch": 5.765217391304348, | |
| "grad_norm": 0.013254191726446152, | |
| "learning_rate": 9.418551710895243e-06, | |
| "loss": 0.0001, | |
| "step": 86 | |
| }, | |
| { | |
| "epoch": 5.834782608695652, | |
| "grad_norm": 0.009798450395464897, | |
| "learning_rate": 9.18641325332142e-06, | |
| "loss": 0.0002, | |
| "step": 87 | |
| }, | |
| { | |
| "epoch": 5.904347826086957, | |
| "grad_norm": 0.024214737117290497, | |
| "learning_rate": 8.954715367323468e-06, | |
| "loss": 0.0004, | |
| "step": 88 | |
| }, | |
| { | |
| "epoch": 5.904347826086957, | |
| "eval_accuracy": 1.0, | |
| "eval_chosen_reward_mean": 0.998676061630249, | |
| "eval_loss": 0.001338434056378901, | |
| "eval_rejected_reward_mean": 0.0013239399995654821, | |
| "eval_reward_diff_mean": 0.9973520636558533, | |
| "eval_reward_diff_std": 0.010331124067306519, | |
| "eval_runtime": 4.5542, | |
| "eval_samples_per_second": 22.616, | |
| "eval_steps_per_second": 5.709, | |
| "step": 88 | |
| }, | |
| { | |
| "epoch": 5.973913043478261, | |
| "grad_norm": 0.007471485063433647, | |
| "learning_rate": 8.723583521394054e-06, | |
| "loss": 0.0001, | |
| "step": 89 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "grad_norm": 0.004855886101722717, | |
| "learning_rate": 8.49314287750517e-06, | |
| "loss": 0.0001, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 6.069565217391304, | |
| "grad_norm": 0.0033697374165058136, | |
| "learning_rate": 8.263518223330698e-06, | |
| "loss": 0.0001, | |
| "step": 91 | |
| }, | |
| { | |
| "epoch": 6.139130434782609, | |
| "grad_norm": 0.014584217220544815, | |
| "learning_rate": 8.034833904671698e-06, | |
| "loss": 0.0003, | |
| "step": 92 | |
| }, | |
| { | |
| "epoch": 6.139130434782609, | |
| "eval_accuracy": 1.0, | |
| "eval_chosen_reward_mean": 0.9988161325454712, | |
| "eval_loss": 0.001194782555103302, | |
| "eval_rejected_reward_mean": 0.0011839327635243535, | |
| "eval_reward_diff_mean": 0.9976321458816528, | |
| "eval_reward_diff_std": 0.008936475962400436, | |
| "eval_runtime": 4.4959, | |
| "eval_samples_per_second": 22.91, | |
| "eval_steps_per_second": 5.783, | |
| "step": 92 | |
| }, | |
| { | |
| "epoch": 6.208695652173913, | |
| "grad_norm": 0.022683285176753998, | |
| "learning_rate": 7.807213758120965e-06, | |
| "loss": 0.0004, | |
| "step": 93 | |
| }, | |
| { | |
| "epoch": 6.278260869565218, | |
| "grad_norm": 0.006901457905769348, | |
| "learning_rate": 7.580781044003324e-06, | |
| "loss": 0.0001, | |
| "step": 94 | |
| }, | |
| { | |
| "epoch": 6.3478260869565215, | |
| "grad_norm": 0.008371288888156414, | |
| "learning_rate": 7.355658379627981e-06, | |
| "loss": 0.0001, | |
| "step": 95 | |
| }, | |
| { | |
| "epoch": 6.417391304347826, | |
| "grad_norm": 0.014148000627756119, | |
| "learning_rate": 7.131967672889101e-06, | |
| "loss": 0.0002, | |
| "step": 96 | |
| }, | |
| { | |
| "epoch": 6.417391304347826, | |
| "eval_accuracy": 1.0, | |
| "eval_chosen_reward_mean": 0.9988046884536743, | |
| "eval_loss": 0.0012071639066562057, | |
| "eval_rejected_reward_mean": 0.0011953354114666581, | |
| "eval_reward_diff_mean": 0.9976092576980591, | |
| "eval_reward_diff_std": 0.009342298842966557, | |
| "eval_runtime": 4.4909, | |
| "eval_samples_per_second": 22.935, | |
| "eval_steps_per_second": 5.79, | |
| "step": 96 | |
| }, | |
| { | |
| "epoch": 6.48695652173913, | |
| "grad_norm": 0.011246114037930965, | |
| "learning_rate": 6.909830056250527e-06, | |
| "loss": 0.0002, | |
| "step": 97 | |
| }, | |
| { | |
| "epoch": 6.556521739130435, | |
| "grad_norm": 0.00854424573481083, | |
| "learning_rate": 6.689365821150421e-06, | |
| "loss": 0.0001, | |
| "step": 98 | |
| }, | |
| { | |
| "epoch": 6.626086956521739, | |
| "grad_norm": 0.023023126646876335, | |
| "learning_rate": 6.4706943528613135e-06, | |
| "loss": 0.0005, | |
| "step": 99 | |
| }, | |
| { | |
| "epoch": 6.695652173913043, | |
| "grad_norm": 0.006646462716162205, | |
| "learning_rate": 6.25393406584088e-06, | |
| "loss": 0.0001, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 6.695652173913043, | |
| "eval_accuracy": 1.0, | |
| "eval_chosen_reward_mean": 0.9988515973091125, | |
| "eval_loss": 0.001159390085376799, | |
| "eval_rejected_reward_mean": 0.0011484603164717555, | |
| "eval_reward_diff_mean": 0.9977031946182251, | |
| "eval_reward_diff_std": 0.008984483778476715, | |
| "eval_runtime": 4.5157, | |
| "eval_samples_per_second": 22.809, | |
| "eval_steps_per_second": 5.758, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 6.765217391304348, | |
| "grad_norm": 0.011894858442246914, | |
| "learning_rate": 6.039202339608432e-06, | |
| "loss": 0.0002, | |
| "step": 101 | |
| }, | |
| { | |
| "epoch": 6.834782608695652, | |
| "grad_norm": 0.010595357045531273, | |
| "learning_rate": 5.8266154551818225e-06, | |
| "loss": 0.0001, | |
| "step": 102 | |
| }, | |
| { | |
| "epoch": 6.904347826086957, | |
| "grad_norm": 0.011542735621333122, | |
| "learning_rate": 5.616288532109225e-06, | |
| "loss": 0.0002, | |
| "step": 103 | |
| }, | |
| { | |
| "epoch": 6.973913043478261, | |
| "grad_norm": 0.051797788590192795, | |
| "learning_rate": 5.4083354661298816e-06, | |
| "loss": 0.0005, | |
| "step": 104 | |
| }, | |
| { | |
| "epoch": 6.973913043478261, | |
| "eval_accuracy": 1.0, | |
| "eval_chosen_reward_mean": 0.9989070892333984, | |
| "eval_loss": 0.0011026657884940505, | |
| "eval_rejected_reward_mean": 0.0010928683914244175, | |
| "eval_reward_diff_mean": 0.9978142380714417, | |
| "eval_reward_diff_std": 0.008508885279297829, | |
| "eval_runtime": 4.4792, | |
| "eval_samples_per_second": 22.995, | |
| "eval_steps_per_second": 5.805, | |
| "step": 104 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "grad_norm": 0.0011278060264885426, | |
| "learning_rate": 5.202868867497542e-06, | |
| "loss": 0.0, | |
| "step": 105 | |
| }, | |
| { | |
| "epoch": 7.069565217391304, | |
| "grad_norm": 0.004797390196472406, | |
| "learning_rate": 5.000000000000003e-06, | |
| "loss": 0.0001, | |
| "step": 106 | |
| }, | |
| { | |
| "epoch": 7.139130434782609, | |
| "grad_norm": 0.05033459514379501, | |
| "learning_rate": 4.799838720707847e-06, | |
| "loss": 0.0007, | |
| "step": 107 | |
| }, | |
| { | |
| "epoch": 7.208695652173913, | |
| "grad_norm": 0.006659619510173798, | |
| "learning_rate": 4.6024934204848745e-06, | |
| "loss": 0.0001, | |
| "step": 108 | |
| }, | |
| { | |
| "epoch": 7.208695652173913, | |
| "eval_accuracy": 1.0, | |
| "eval_chosen_reward_mean": 0.9988996982574463, | |
| "eval_loss": 0.001110102399252355, | |
| "eval_rejected_reward_mean": 0.00110039790160954, | |
| "eval_reward_diff_mean": 0.9977990984916687, | |
| "eval_reward_diff_std": 0.008465510793030262, | |
| "eval_runtime": 4.4603, | |
| "eval_samples_per_second": 23.093, | |
| "eval_steps_per_second": 5.829, | |
| "step": 108 | |
| }, | |
| { | |
| "epoch": 7.278260869565218, | |
| "grad_norm": 0.0025184813421219587, | |
| "learning_rate": 4.408070965292534e-06, | |
| "loss": 0.0, | |
| "step": 109 | |
| }, | |
| { | |
| "epoch": 7.3478260869565215, | |
| "grad_norm": 0.001780255581252277, | |
| "learning_rate": 4.216676638320135e-06, | |
| "loss": 0.0, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 7.417391304347826, | |
| "grad_norm": 0.005449565593153238, | |
| "learning_rate": 4.028414082972141e-06, | |
| "loss": 0.0001, | |
| "step": 111 | |
| }, | |
| { | |
| "epoch": 7.48695652173913, | |
| "grad_norm": 0.01474890485405922, | |
| "learning_rate": 3.8433852467434175e-06, | |
| "loss": 0.0002, | |
| "step": 112 | |
| }, | |
| { | |
| "epoch": 7.48695652173913, | |
| "eval_accuracy": 1.0, | |
| "eval_chosen_reward_mean": 0.9989209175109863, | |
| "eval_loss": 0.0010885810479521751, | |
| "eval_rejected_reward_mean": 0.001079102628864348, | |
| "eval_reward_diff_mean": 0.9978418946266174, | |
| "eval_reward_diff_std": 0.008369901217520237, | |
| "eval_runtime": 4.4683, | |
| "eval_samples_per_second": 23.052, | |
| "eval_steps_per_second": 5.819, | |
| "step": 112 | |
| }, | |
| { | |
| "epoch": 7.556521739130435, | |
| "grad_norm": 0.01239398680627346, | |
| "learning_rate": 3.661690326012897e-06, | |
| "loss": 0.0002, | |
| "step": 113 | |
| }, | |
| { | |
| "epoch": 7.626086956521739, | |
| "grad_norm": 0.012043354101479053, | |
| "learning_rate": 3.483427711785449e-06, | |
| "loss": 0.0002, | |
| "step": 114 | |
| }, | |
| { | |
| "epoch": 7.695652173913043, | |
| "grad_norm": 0.007400067988783121, | |
| "learning_rate": 3.308693936411421e-06, | |
| "loss": 0.0001, | |
| "step": 115 | |
| }, | |
| { | |
| "epoch": 7.765217391304348, | |
| "grad_norm": 0.015701085329055786, | |
| "learning_rate": 3.1375836213126653e-06, | |
| "loss": 0.0003, | |
| "step": 116 | |
| }, | |
| { | |
| "epoch": 7.765217391304348, | |
| "eval_accuracy": 1.0, | |
| "eval_chosen_reward_mean": 0.9989361763000488, | |
| "eval_loss": 0.0010728331981226802, | |
| "eval_rejected_reward_mean": 0.0010637931991368532, | |
| "eval_reward_diff_mean": 0.9978724718093872, | |
| "eval_reward_diff_std": 0.008171496912837029, | |
| "eval_runtime": 4.489, | |
| "eval_samples_per_second": 22.945, | |
| "eval_steps_per_second": 5.792, | |
| "step": 116 | |
| }, | |
| { | |
| "epoch": 7.834782608695652, | |
| "grad_norm": 0.023439966142177582, | |
| "learning_rate": 2.970189425743383e-06, | |
| "loss": 0.0002, | |
| "step": 117 | |
| }, | |
| { | |
| "epoch": 7.904347826086957, | |
| "grad_norm": 0.014678899198770523, | |
| "learning_rate": 2.8066019966134907e-06, | |
| "loss": 0.0002, | |
| "step": 118 | |
| }, | |
| { | |
| "epoch": 7.973913043478261, | |
| "grad_norm": 0.009870647452771664, | |
| "learning_rate": 2.6469099194017144e-06, | |
| "loss": 0.0002, | |
| "step": 119 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "grad_norm": 0.025675436481833458, | |
| "learning_rate": 2.4911996701850083e-06, | |
| "loss": 0.0004, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 1.0, | |
| "eval_chosen_reward_mean": 0.9989721775054932, | |
| "eval_loss": 0.0010361903114244342, | |
| "eval_rejected_reward_mean": 0.0010278514819219708, | |
| "eval_reward_diff_mean": 0.9979442358016968, | |
| "eval_reward_diff_std": 0.007848705165088177, | |
| "eval_runtime": 4.4876, | |
| "eval_samples_per_second": 22.952, | |
| "eval_steps_per_second": 5.794, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 8.069565217391304, | |
| "grad_norm": 0.010908635333180428, | |
| "learning_rate": 2.339555568810221e-06, | |
| "loss": 0.0001, | |
| "step": 121 | |
| }, | |
| { | |
| "epoch": 8.139130434782608, | |
| "grad_norm": 0.020105818286538124, | |
| "learning_rate": 2.192059733233408e-06, | |
| "loss": 0.0003, | |
| "step": 122 | |
| }, | |
| { | |
| "epoch": 8.208695652173914, | |
| "grad_norm": 0.008475644513964653, | |
| "learning_rate": 2.048792035051521e-06, | |
| "loss": 0.0001, | |
| "step": 123 | |
| }, | |
| { | |
| "epoch": 8.278260869565218, | |
| "grad_norm": 0.012479428201913834, | |
| "learning_rate": 1.9098300562505266e-06, | |
| "loss": 0.0002, | |
| "step": 124 | |
| }, | |
| { | |
| "epoch": 8.278260869565218, | |
| "eval_accuracy": 1.0, | |
| "eval_chosen_reward_mean": 0.9989458322525024, | |
| "eval_loss": 0.0010632032062858343, | |
| "eval_rejected_reward_mean": 0.0010542016243562102, | |
| "eval_reward_diff_mean": 0.9978914856910706, | |
| "eval_reward_diff_std": 0.008157290518283844, | |
| "eval_runtime": 4.5475, | |
| "eval_samples_per_second": 22.65, | |
| "eval_steps_per_second": 5.717, | |
| "step": 124 | |
| }, | |
| { | |
| "epoch": 8.347826086956522, | |
| "grad_norm": 0.00540159922093153, | |
| "learning_rate": 1.7752490471933769e-06, | |
| "loss": 0.0001, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 8.417391304347825, | |
| "grad_norm": 0.007868447341024876, | |
| "learning_rate": 1.6451218858706374e-06, | |
| "loss": 0.0001, | |
| "step": 126 | |
| }, | |
| { | |
| "epoch": 8.486956521739131, | |
| "grad_norm": 0.04023267701268196, | |
| "learning_rate": 1.5195190384357405e-06, | |
| "loss": 0.0005, | |
| "step": 127 | |
| }, | |
| { | |
| "epoch": 8.556521739130435, | |
| "grad_norm": 0.014718690887093544, | |
| "learning_rate": 1.3985085210463479e-06, | |
| "loss": 0.0002, | |
| "step": 128 | |
| }, | |
| { | |
| "epoch": 8.556521739130435, | |
| "eval_accuracy": 1.0, | |
| "eval_chosen_reward_mean": 0.9989545941352844, | |
| "eval_loss": 0.0010542994132265449, | |
| "eval_rejected_reward_mean": 0.0010454322909936309, | |
| "eval_reward_diff_mean": 0.9979091882705688, | |
| "eval_reward_diff_std": 0.008097290061414242, | |
| "eval_runtime": 4.5176, | |
| "eval_samples_per_second": 22.8, | |
| "eval_steps_per_second": 5.755, | |
| "step": 128 | |
| }, | |
| { | |
| "epoch": 8.626086956521739, | |
| "grad_norm": 0.006092917174100876, | |
| "learning_rate": 1.282155863032377e-06, | |
| "loss": 0.0001, | |
| "step": 129 | |
| }, | |
| { | |
| "epoch": 8.695652173913043, | |
| "grad_norm": 0.015089413151144981, | |
| "learning_rate": 1.1705240714107301e-06, | |
| "loss": 0.0002, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 8.765217391304347, | |
| "grad_norm": 0.008863482624292374, | |
| "learning_rate": 1.0636735967658785e-06, | |
| "loss": 0.0001, | |
| "step": 131 | |
| }, | |
| { | |
| "epoch": 8.834782608695653, | |
| "grad_norm": 0.02049126662313938, | |
| "learning_rate": 9.616623005147952e-07, | |
| "loss": 0.0004, | |
| "step": 132 | |
| }, | |
| { | |
| "epoch": 8.834782608695653, | |
| "eval_accuracy": 1.0, | |
| "eval_chosen_reward_mean": 0.9989612102508545, | |
| "eval_loss": 0.0010475697927176952, | |
| "eval_rejected_reward_mean": 0.0010388661175966263, | |
| "eval_reward_diff_mean": 0.9979221820831299, | |
| "eval_reward_diff_std": 0.00802167784422636, | |
| "eval_runtime": 4.4869, | |
| "eval_samples_per_second": 22.956, | |
| "eval_steps_per_second": 5.795, | |
| "step": 132 | |
| }, | |
| { | |
| "epoch": 8.904347826086957, | |
| "grad_norm": 0.0050810459069907665, | |
| "learning_rate": 8.645454235739903e-07, | |
| "loss": 0.0001, | |
| "step": 133 | |
| }, | |
| { | |
| "epoch": 8.97391304347826, | |
| "grad_norm": 0.0044286418706178665, | |
| "learning_rate": 7.723755564455771e-07, | |
| "loss": 0.0001, | |
| "step": 134 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "grad_norm": 0.0036248955875635147, | |
| "learning_rate": 6.852026107385756e-07, | |
| "loss": 0.0, | |
| "step": 135 | |
| }, | |
| { | |
| "epoch": 9.069565217391304, | |
| "grad_norm": 0.010801658034324646, | |
| "learning_rate": 6.030737921409169e-07, | |
| "loss": 0.0002, | |
| "step": 136 | |
| }, | |
| { | |
| "epoch": 9.069565217391304, | |
| "eval_accuracy": 1.0, | |
| "eval_chosen_reward_mean": 0.9989877939224243, | |
| "eval_loss": 0.0010204218560829759, | |
| "eval_rejected_reward_mean": 0.0010121165541931987, | |
| "eval_reward_diff_mean": 0.997975766658783, | |
| "eval_reward_diff_std": 0.007839000783860683, | |
| "eval_runtime": 4.5032, | |
| "eval_samples_per_second": 22.872, | |
| "eval_steps_per_second": 5.774, | |
| "step": 136 | |
| }, | |
| { | |
| "epoch": 9.139130434782608, | |
| "grad_norm": 0.012117389589548111, | |
| "learning_rate": 5.26033574856708e-07, | |
| "loss": 0.0002, | |
| "step": 137 | |
| }, | |
| { | |
| "epoch": 9.208695652173914, | |
| "grad_norm": 0.003176574595272541, | |
| "learning_rate": 4.5412367752268094e-07, | |
| "loss": 0.0001, | |
| "step": 138 | |
| }, | |
| { | |
| "epoch": 9.278260869565218, | |
| "grad_norm": 0.006320824846625328, | |
| "learning_rate": 3.8738304061681107e-07, | |
| "loss": 0.0001, | |
| "step": 139 | |
| }, | |
| { | |
| "epoch": 9.347826086956522, | |
| "grad_norm": 0.0101704066619277, | |
| "learning_rate": 3.2584780537136206e-07, | |
| "loss": 0.0002, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 9.347826086956522, | |
| "eval_accuracy": 1.0, | |
| "eval_chosen_reward_mean": 0.9989598989486694, | |
| "eval_loss": 0.0010490770218893886, | |
| "eval_rejected_reward_mean": 0.0010401182807981968, | |
| "eval_reward_diff_mean": 0.9979196786880493, | |
| "eval_reward_diff_std": 0.008142856881022453, | |
| "eval_runtime": 4.5214, | |
| "eval_samples_per_second": 22.78, | |
| "eval_steps_per_second": 5.75, | |
| "step": 140 | |
| } | |
| ], | |
| "logging_steps": 1, | |
| "max_steps": 150, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 10, | |
| "save_steps": 4, | |
| "stateful_callbacks": { | |
| "TrainerControl": { | |
| "args": { | |
| "should_epoch_stop": false, | |
| "should_evaluate": false, | |
| "should_log": false, | |
| "should_save": true, | |
| "should_training_stop": false | |
| }, | |
| "attributes": {} | |
| } | |
| }, | |
| "total_flos": 0.0, | |
| "train_batch_size": 8, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |