gemma-2-2b-it-mt-dpo-full_sardine / trainer_state.json
martimfasantos's picture
Model save
7ae2afd verified
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 2.9897455374097985,
"eval_steps": 800,
"global_step": 738,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.004051145714647424,
"grad_norm": 23.354017564608814,
"learning_rate": 1.3513513513513514e-09,
"logits/chosen": -8.401385307312012,
"logits/rejected": -7.887872695922852,
"logps/chosen": -22.588991165161133,
"logps/rejected": -32.73207473754883,
"loss": 0.6931,
"rewards/accuracies": 0.0,
"rewards/chosen": 0.0,
"rewards/margins": 0.0,
"rewards/rejected": 0.0,
"step": 1
},
{
"epoch": 0.040511457146474236,
"grad_norm": 23.005760986532493,
"learning_rate": 1.3513513513513514e-08,
"logits/chosen": -8.023005485534668,
"logits/rejected": -7.215701580047607,
"logps/chosen": -26.691545486450195,
"logps/rejected": -35.018611907958984,
"loss": 0.6932,
"rewards/accuracies": 0.3958333432674408,
"rewards/chosen": -0.00048815118498168886,
"rewards/margins": -0.001463304040953517,
"rewards/rejected": 0.0009751527686603367,
"step": 10
},
{
"epoch": 0.08102291429294847,
"grad_norm": 25.76788201635075,
"learning_rate": 2.7027027027027028e-08,
"logits/chosen": -8.08788776397705,
"logits/rejected": -7.492005825042725,
"logps/chosen": -28.649845123291016,
"logps/rejected": -38.781715393066406,
"loss": 0.6934,
"rewards/accuracies": 0.4937500059604645,
"rewards/chosen": 0.0005952938226982951,
"rewards/margins": 0.00045213793055154383,
"rewards/rejected": 0.0001431556447641924,
"step": 20
},
{
"epoch": 0.12153437143942271,
"grad_norm": 24.285898108451498,
"learning_rate": 4.054054054054054e-08,
"logits/chosen": -8.020319938659668,
"logits/rejected": -7.39060115814209,
"logps/chosen": -26.554407119750977,
"logps/rejected": -39.03184509277344,
"loss": 0.6916,
"rewards/accuracies": 0.518750011920929,
"rewards/chosen": 0.0008592296508140862,
"rewards/margins": 0.003104406874626875,
"rewards/rejected": -0.002245177049189806,
"step": 30
},
{
"epoch": 0.16204582858589694,
"grad_norm": 26.673301885189623,
"learning_rate": 5.4054054054054056e-08,
"logits/chosen": -8.143043518066406,
"logits/rejected": -7.433783531188965,
"logps/chosen": -25.945232391357422,
"logps/rejected": -37.73704147338867,
"loss": 0.6882,
"rewards/accuracies": 0.574999988079071,
"rewards/chosen": 9.569372195983306e-05,
"rewards/margins": 0.007951222360134125,
"rewards/rejected": -0.007855528965592384,
"step": 40
},
{
"epoch": 0.2025572857323712,
"grad_norm": 26.497438005541635,
"learning_rate": 6.756756756756756e-08,
"logits/chosen": -8.0227632522583,
"logits/rejected": -7.415746212005615,
"logps/chosen": -28.301464080810547,
"logps/rejected": -38.43585205078125,
"loss": 0.6843,
"rewards/accuracies": 0.6468750238418579,
"rewards/chosen": 0.0006854601087979972,
"rewards/margins": 0.0169823057949543,
"rewards/rejected": -0.01629684492945671,
"step": 50
},
{
"epoch": 0.24306874287884542,
"grad_norm": 24.919397084889898,
"learning_rate": 8.108108108108108e-08,
"logits/chosen": -8.055706024169922,
"logits/rejected": -7.392535209655762,
"logps/chosen": -26.291152954101562,
"logps/rejected": -37.135032653808594,
"loss": 0.6751,
"rewards/accuracies": 0.734375,
"rewards/chosen": 0.0014377154875546694,
"rewards/margins": 0.03630940988659859,
"rewards/rejected": -0.03487168997526169,
"step": 60
},
{
"epoch": 0.28358020002531964,
"grad_norm": 20.09272984199583,
"learning_rate": 9.45945945945946e-08,
"logits/chosen": -8.232405662536621,
"logits/rejected": -7.535184383392334,
"logps/chosen": -27.17940902709961,
"logps/rejected": -36.42096710205078,
"loss": 0.6635,
"rewards/accuracies": 0.824999988079071,
"rewards/chosen": 0.0027192712295800447,
"rewards/margins": 0.06449674069881439,
"rewards/rejected": -0.06177746504545212,
"step": 70
},
{
"epoch": 0.3240916571717939,
"grad_norm": 23.04151710177534,
"learning_rate": 9.997985455197112e-08,
"logits/chosen": -8.351137161254883,
"logits/rejected": -7.78473424911499,
"logps/chosen": -26.988977432250977,
"logps/rejected": -36.72052001953125,
"loss": 0.6537,
"rewards/accuracies": 0.753125011920929,
"rewards/chosen": 0.00015634755254723132,
"rewards/margins": 0.08445531874895096,
"rewards/rejected": -0.08429896831512451,
"step": 80
},
{
"epoch": 0.36460311431826814,
"grad_norm": 24.86811296840432,
"learning_rate": 9.98568022639826e-08,
"logits/chosen": -8.316564559936523,
"logits/rejected": -7.610459327697754,
"logps/chosen": -28.134241104125977,
"logps/rejected": -40.095638275146484,
"loss": 0.6307,
"rewards/accuracies": 0.862500011920929,
"rewards/chosen": 0.005406993441283703,
"rewards/margins": 0.13842421770095825,
"rewards/rejected": -0.13301721215248108,
"step": 90
},
{
"epoch": 0.4051145714647424,
"grad_norm": 23.930696194698072,
"learning_rate": 9.962216467480192e-08,
"logits/chosen": -8.399057388305664,
"logits/rejected": -7.728206634521484,
"logps/chosen": -27.389541625976562,
"logps/rejected": -42.16829299926758,
"loss": 0.6168,
"rewards/accuracies": 0.796875,
"rewards/chosen": -0.0014130814233794808,
"rewards/margins": 0.17133907973766327,
"rewards/rejected": -0.17275215685367584,
"step": 100
},
{
"epoch": 0.4456260286112166,
"grad_norm": 25.12250587863022,
"learning_rate": 9.927646693054496e-08,
"logits/chosen": -8.546191215515137,
"logits/rejected": -7.818883419036865,
"logps/chosen": -26.858484268188477,
"logps/rejected": -42.138423919677734,
"loss": 0.5877,
"rewards/accuracies": 0.846875011920929,
"rewards/chosen": -0.002900606021285057,
"rewards/margins": 0.24653196334838867,
"rewards/rejected": -0.24943256378173828,
"step": 110
},
{
"epoch": 0.48613748575769083,
"grad_norm": 23.993302988236724,
"learning_rate": 9.882048274282503e-08,
"logits/chosen": -8.554715156555176,
"logits/rejected": -7.8766679763793945,
"logps/chosen": -26.874399185180664,
"logps/rejected": -40.70722961425781,
"loss": 0.5763,
"rewards/accuracies": 0.8031250238418579,
"rewards/chosen": -0.026281490921974182,
"rewards/margins": 0.26420071721076965,
"rewards/rejected": -0.29048219323158264,
"step": 120
},
{
"epoch": 0.5266489429041651,
"grad_norm": 19.760387337865797,
"learning_rate": 9.825523265709666e-08,
"logits/chosen": -8.872574806213379,
"logits/rejected": -8.05439281463623,
"logps/chosen": -29.766094207763672,
"logps/rejected": -39.4333381652832,
"loss": 0.5627,
"rewards/accuracies": 0.84375,
"rewards/chosen": -0.01587745174765587,
"rewards/margins": 0.3382584750652313,
"rewards/rejected": -0.3541359305381775,
"step": 130
},
{
"epoch": 0.5671604000506393,
"grad_norm": 20.405644698254697,
"learning_rate": 9.758198176855646e-08,
"logits/chosen": -8.83372974395752,
"logits/rejected": -8.060525894165039,
"logps/chosen": -27.360458374023438,
"logps/rejected": -41.453636169433594,
"loss": 0.5348,
"rewards/accuracies": 0.846875011920929,
"rewards/chosen": -0.028619933873414993,
"rewards/margins": 0.39799413084983826,
"rewards/rejected": -0.42661410570144653,
"step": 140
},
{
"epoch": 0.6076718571971136,
"grad_norm": 24.42451138080484,
"learning_rate": 9.680223689071363e-08,
"logits/chosen": -9.05488109588623,
"logits/rejected": -8.361872673034668,
"logps/chosen": -27.39214515686035,
"logps/rejected": -44.14860916137695,
"loss": 0.5253,
"rewards/accuracies": 0.856249988079071,
"rewards/chosen": -0.041482653468847275,
"rewards/margins": 0.43668460845947266,
"rewards/rejected": -0.47816720604896545,
"step": 150
},
{
"epoch": 0.6481833143435878,
"grad_norm": 22.08758150384141,
"learning_rate": 9.59177431829666e-08,
"logits/chosen": -9.038153648376465,
"logits/rejected": -8.328339576721191,
"logps/chosen": -27.219493865966797,
"logps/rejected": -43.879791259765625,
"loss": 0.5219,
"rewards/accuracies": 0.828125,
"rewards/chosen": -0.039076704531908035,
"rewards/margins": 0.47056007385253906,
"rewards/rejected": -0.5096367597579956,
"step": 160
},
{
"epoch": 0.6886947714900621,
"grad_norm": 22.525750174655215,
"learning_rate": 9.493048024473411e-08,
"logits/chosen": -9.239489555358887,
"logits/rejected": -8.590790748596191,
"logps/chosen": -27.6241455078125,
"logps/rejected": -43.37400817871094,
"loss": 0.5152,
"rewards/accuracies": 0.8187500238418579,
"rewards/chosen": -0.07299178838729858,
"rewards/margins": 0.4755212664604187,
"rewards/rejected": -0.5485130548477173,
"step": 170
},
{
"epoch": 0.7292062286365363,
"grad_norm": 21.336294616647535,
"learning_rate": 9.384265768488224e-08,
"logits/chosen": -9.427377700805664,
"logits/rejected": -8.607305526733398,
"logps/chosen": -27.55912208557129,
"logps/rejected": -42.556678771972656,
"loss": 0.4916,
"rewards/accuracies": 0.840624988079071,
"rewards/chosen": -0.07397119700908661,
"rewards/margins": 0.5928912162780762,
"rewards/rejected": -0.666862428188324,
"step": 180
},
{
"epoch": 0.7697176857830105,
"grad_norm": 20.933312614297105,
"learning_rate": 9.265671017636383e-08,
"logits/chosen": -9.61644458770752,
"logits/rejected": -8.950950622558594,
"logps/chosen": -28.78060531616211,
"logps/rejected": -45.048828125,
"loss": 0.4693,
"rewards/accuracies": 0.840624988079071,
"rewards/chosen": -0.06900081783533096,
"rewards/margins": 0.6354625821113586,
"rewards/rejected": -0.7044633030891418,
"step": 190
},
{
"epoch": 0.8102291429294848,
"grad_norm": 20.036505723069673,
"learning_rate": 9.137529200713809e-08,
"logits/chosen": -9.64765739440918,
"logits/rejected": -8.868057250976562,
"logps/chosen": -26.3610782623291,
"logps/rejected": -43.371124267578125,
"loss": 0.4555,
"rewards/accuracies": 0.8031250238418579,
"rewards/chosen": -0.10404930263757706,
"rewards/margins": 0.7085543870925903,
"rewards/rejected": -0.812603771686554,
"step": 200
},
{
"epoch": 0.850740600075959,
"grad_norm": 21.18417249103817,
"learning_rate": 9.000127113956672e-08,
"logits/chosen": -9.870735168457031,
"logits/rejected": -9.190484046936035,
"logps/chosen": -26.790430068969727,
"logps/rejected": -43.78266525268555,
"loss": 0.4639,
"rewards/accuracies": 0.8343750238418579,
"rewards/chosen": -0.10198328644037247,
"rewards/margins": 0.7101414799690247,
"rewards/rejected": -0.8121248483657837,
"step": 210
},
{
"epoch": 0.8912520572224332,
"grad_norm": 18.99507288099157,
"learning_rate": 8.853772279158165e-08,
"logits/chosen": -9.834332466125488,
"logits/rejected": -9.097946166992188,
"logps/chosen": -27.970256805419922,
"logps/rejected": -44.3878059387207,
"loss": 0.4372,
"rewards/accuracies": 0.84375,
"rewards/chosen": -0.1515013873577118,
"rewards/margins": 0.78831946849823,
"rewards/rejected": -0.9398208856582642,
"step": 220
},
{
"epoch": 0.9317635143689075,
"grad_norm": 22.643159035198256,
"learning_rate": 8.698792255399104e-08,
"logits/chosen": -10.250986099243164,
"logits/rejected": -9.542135238647461,
"logps/chosen": -28.57198715209961,
"logps/rejected": -47.67482376098633,
"loss": 0.4174,
"rewards/accuracies": 0.846875011920929,
"rewards/chosen": -0.16280151903629303,
"rewards/margins": 0.9155281186103821,
"rewards/rejected": -1.0783296823501587,
"step": 230
},
{
"epoch": 0.9722749715153817,
"grad_norm": 17.009463712911476,
"learning_rate": 8.535533905932736e-08,
"logits/chosen": -10.252339363098145,
"logits/rejected": -9.567429542541504,
"logps/chosen": -29.68218421936035,
"logps/rejected": -49.98845291137695,
"loss": 0.4311,
"rewards/accuracies": 0.840624988079071,
"rewards/chosen": -0.21189716458320618,
"rewards/margins": 0.9156039953231812,
"rewards/rejected": -1.1275012493133545,
"step": 240
},
{
"epoch": 1.012786428661856,
"grad_norm": 21.283801315155564,
"learning_rate": 8.364362621864594e-08,
"logits/chosen": -10.185906410217285,
"logits/rejected": -9.48774528503418,
"logps/chosen": -28.432941436767578,
"logps/rejected": -50.117576599121094,
"loss": 0.4013,
"rewards/accuracies": 0.859375,
"rewards/chosen": -0.21948130428791046,
"rewards/margins": 0.9736725687980652,
"rewards/rejected": -1.1931538581848145,
"step": 250
},
{
"epoch": 1.0532978858083302,
"grad_norm": 22.754490059016796,
"learning_rate": 8.185661504364844e-08,
"logits/chosen": -10.365915298461914,
"logits/rejected": -9.623150825500488,
"logps/chosen": -28.134822845458984,
"logps/rejected": -48.77445983886719,
"loss": 0.3799,
"rewards/accuracies": 0.8843749761581421,
"rewards/chosen": -0.22865962982177734,
"rewards/margins": 1.057023286819458,
"rewards/rejected": -1.2856829166412354,
"step": 260
},
{
"epoch": 1.0938093429548044,
"grad_norm": 29.101808968686573,
"learning_rate": 7.999830507243477e-08,
"logits/chosen": -10.540546417236328,
"logits/rejected": -9.749776840209961,
"logps/chosen": -32.03081512451172,
"logps/rejected": -50.996429443359375,
"loss": 0.4043,
"rewards/accuracies": 0.846875011920929,
"rewards/chosen": -0.30270159244537354,
"rewards/margins": 1.1776001453399658,
"rewards/rejected": -1.480301856994629,
"step": 270
},
{
"epoch": 1.1343208001012786,
"grad_norm": 22.52674198964978,
"learning_rate": 7.80728554180734e-08,
"logits/chosen": -10.482057571411133,
"logits/rejected": -9.802620887756348,
"logps/chosen": -29.116191864013672,
"logps/rejected": -51.501319885253906,
"loss": 0.3695,
"rewards/accuracies": 0.856249988079071,
"rewards/chosen": -0.2759430706501007,
"rewards/margins": 1.228001356124878,
"rewards/rejected": -1.5039442777633667,
"step": 280
},
{
"epoch": 1.174832257247753,
"grad_norm": 23.744363106936603,
"learning_rate": 7.608457546002422e-08,
"logits/chosen": -10.582446098327637,
"logits/rejected": -9.793972969055176,
"logps/chosen": -30.690120697021484,
"logps/rejected": -52.65431594848633,
"loss": 0.3608,
"rewards/accuracies": 0.846875011920929,
"rewards/chosen": -0.35969388484954834,
"rewards/margins": 1.3402436971664429,
"rewards/rejected": -1.6999378204345703,
"step": 290
},
{
"epoch": 1.2153437143942272,
"grad_norm": 16.96872058311971,
"learning_rate": 7.403791519924793e-08,
"logits/chosen": -10.636259078979492,
"logits/rejected": -9.861931800842285,
"logps/chosen": -30.6519718170166,
"logps/rejected": -55.83306884765625,
"loss": 0.3609,
"rewards/accuracies": 0.856249988079071,
"rewards/chosen": -0.34394341707229614,
"rewards/margins": 1.4830596446990967,
"rewards/rejected": -1.8270031213760376,
"step": 300
},
{
"epoch": 1.2558551715407014,
"grad_norm": 23.80487547845051,
"learning_rate": 7.193745529858826e-08,
"logits/chosen": -10.673724174499512,
"logits/rejected": -9.968966484069824,
"logps/chosen": -32.3521614074707,
"logps/rejected": -55.50031661987305,
"loss": 0.3796,
"rewards/accuracies": 0.878125011920929,
"rewards/chosen": -0.4293053150177002,
"rewards/margins": 1.2904064655303955,
"rewards/rejected": -1.7197116613388062,
"step": 310
},
{
"epoch": 1.2963666286871756,
"grad_norm": 26.02880426926903,
"learning_rate": 6.978789683071759e-08,
"logits/chosen": -10.829614639282227,
"logits/rejected": -10.158222198486328,
"logps/chosen": -30.30867576599121,
"logps/rejected": -54.475677490234375,
"loss": 0.3895,
"rewards/accuracies": 0.84375,
"rewards/chosen": -0.41875213384628296,
"rewards/margins": 1.3132567405700684,
"rewards/rejected": -1.732008695602417,
"step": 320
},
{
"epoch": 1.3368780858336498,
"grad_norm": 23.11525326850835,
"learning_rate": 6.759405075659165e-08,
"logits/chosen": -10.891769409179688,
"logits/rejected": -10.190674781799316,
"logps/chosen": -30.676244735717773,
"logps/rejected": -58.19831466674805,
"loss": 0.3621,
"rewards/accuracies": 0.8687499761581421,
"rewards/chosen": -0.4654254913330078,
"rewards/margins": 1.587354302406311,
"rewards/rejected": -2.0527796745300293,
"step": 330
},
{
"epoch": 1.377389542980124,
"grad_norm": 19.61508445315914,
"learning_rate": 6.536082715796124e-08,
"logits/chosen": -11.017595291137695,
"logits/rejected": -10.286384582519531,
"logps/chosen": -29.53006362915039,
"logps/rejected": -54.037132263183594,
"loss": 0.3455,
"rewards/accuracies": 0.8656250238418579,
"rewards/chosen": -0.3892853260040283,
"rewards/margins": 1.4755113124847412,
"rewards/rejected": -1.8647968769073486,
"step": 340
},
{
"epoch": 1.4179010001265984,
"grad_norm": 24.546782869534564,
"learning_rate": 6.309322424804034e-08,
"logits/chosen": -10.974300384521484,
"logits/rejected": -10.224813461303711,
"logps/chosen": -33.27606201171875,
"logps/rejected": -58.134117126464844,
"loss": 0.3673,
"rewards/accuracies": 0.8500000238418579,
"rewards/chosen": -0.5354558229446411,
"rewards/margins": 1.5486235618591309,
"rewards/rejected": -2.0840792655944824,
"step": 350
},
{
"epoch": 1.4584124572730726,
"grad_norm": 20.799419761461635,
"learning_rate": 6.079631718492568e-08,
"logits/chosen": -11.121304512023926,
"logits/rejected": -10.479646682739258,
"logps/chosen": -30.873401641845703,
"logps/rejected": -58.193626403808594,
"loss": 0.3602,
"rewards/accuracies": 0.875,
"rewards/chosen": -0.4879949986934662,
"rewards/margins": 1.5607774257659912,
"rewards/rejected": -2.0487725734710693,
"step": 360
},
{
"epoch": 1.4989239144195468,
"grad_norm": 22.15106589586126,
"learning_rate": 5.847524671280484e-08,
"logits/chosen": -11.006221771240234,
"logits/rejected": -10.338632583618164,
"logps/chosen": -31.7496395111084,
"logps/rejected": -56.75434112548828,
"loss": 0.355,
"rewards/accuracies": 0.846875011920929,
"rewards/chosen": -0.5273393392562866,
"rewards/margins": 1.4817734956741333,
"rewards/rejected": -2.00911283493042,
"step": 370
},
{
"epoch": 1.5394353715660212,
"grad_norm": 27.06131590311837,
"learning_rate": 5.6135207656374886e-08,
"logits/chosen": -11.114490509033203,
"logits/rejected": -10.422750473022461,
"logps/chosen": -31.69234848022461,
"logps/rejected": -55.545692443847656,
"loss": 0.3597,
"rewards/accuracies": 0.8531249761581421,
"rewards/chosen": -0.5423402190208435,
"rewards/margins": 1.4755096435546875,
"rewards/rejected": -2.017849922180176,
"step": 380
},
{
"epoch": 1.5799468287124951,
"grad_norm": 25.089579946286218,
"learning_rate": 5.3781437294222844e-08,
"logits/chosen": -11.122465133666992,
"logits/rejected": -10.4068603515625,
"logps/chosen": -32.763519287109375,
"logps/rejected": -58.76951217651367,
"loss": 0.3421,
"rewards/accuracies": 0.8500000238418579,
"rewards/chosen": -0.5654906034469604,
"rewards/margins": 1.569490671157837,
"rewards/rejected": -2.134981155395508,
"step": 390
},
{
"epoch": 1.6204582858589696,
"grad_norm": 25.42503001610947,
"learning_rate": 5.141920363718916e-08,
"logits/chosen": -11.278289794921875,
"logits/rejected": -10.557046890258789,
"logps/chosen": -33.54096221923828,
"logps/rejected": -57.26496505737305,
"loss": 0.3619,
"rewards/accuracies": 0.796875,
"rewards/chosen": -0.638907253742218,
"rewards/margins": 1.6008431911468506,
"rewards/rejected": -2.239750385284424,
"step": 400
},
{
"epoch": 1.6609697430054438,
"grad_norm": 19.761132187041504,
"learning_rate": 4.905379363794906e-08,
"logits/chosen": -11.1757230758667,
"logits/rejected": -10.361800193786621,
"logps/chosen": -31.939849853515625,
"logps/rejected": -58.90076446533203,
"loss": 0.3557,
"rewards/accuracies": 0.8500000238418579,
"rewards/chosen": -0.5620619058609009,
"rewards/margins": 1.6297252178192139,
"rewards/rejected": -2.1917872428894043,
"step": 410
},
{
"epoch": 1.701481200151918,
"grad_norm": 21.623030847191835,
"learning_rate": 4.669050135819966e-08,
"logits/chosen": -11.290115356445312,
"logits/rejected": -10.585660934448242,
"logps/chosen": -31.6041316986084,
"logps/rejected": -54.78083419799805,
"loss": 0.3602,
"rewards/accuracies": 0.8374999761581421,
"rewards/chosen": -0.6143931150436401,
"rewards/margins": 1.5332918167114258,
"rewards/rejected": -2.1476848125457764,
"step": 420
},
{
"epoch": 1.7419926572983924,
"grad_norm": 24.475959531037525,
"learning_rate": 4.4334616119936507e-08,
"logits/chosen": -11.401643753051758,
"logits/rejected": -10.660685539245605,
"logps/chosen": -31.317325592041016,
"logps/rejected": -57.25482940673828,
"loss": 0.3347,
"rewards/accuracies": 0.8687499761581421,
"rewards/chosen": -0.557376503944397,
"rewards/margins": 1.634526014328003,
"rewards/rejected": -2.1919026374816895,
"step": 430
},
{
"epoch": 1.7825041144448663,
"grad_norm": 19.486205802493238,
"learning_rate": 4.199141066733789e-08,
"logits/chosen": -11.200121879577637,
"logits/rejected": -10.525270462036133,
"logps/chosen": -29.716289520263672,
"logps/rejected": -55.837684631347656,
"loss": 0.3402,
"rewards/accuracies": 0.8500000238418579,
"rewards/chosen": -0.5918062925338745,
"rewards/margins": 1.5346314907073975,
"rewards/rejected": -2.1264379024505615,
"step": 440
},
{
"epoch": 1.8230155715913408,
"grad_norm": 33.35679620452287,
"learning_rate": 3.966612936575235e-08,
"logits/chosen": -11.320077896118164,
"logits/rejected": -10.705958366394043,
"logps/chosen": -31.9770450592041,
"logps/rejected": -57.85862350463867,
"loss": 0.3475,
"rewards/accuracies": 0.84375,
"rewards/chosen": -0.6092299818992615,
"rewards/margins": 1.5435864925384521,
"rewards/rejected": -2.1528162956237793,
"step": 450
},
{
"epoch": 1.863527028737815,
"grad_norm": 31.567941489065486,
"learning_rate": 3.736397646420134e-08,
"logits/chosen": -11.576532363891602,
"logits/rejected": -10.939656257629395,
"logps/chosen": -33.84613037109375,
"logps/rejected": -59.848854064941406,
"loss": 0.3416,
"rewards/accuracies": 0.871874988079071,
"rewards/chosen": -0.6238502860069275,
"rewards/margins": 1.6080996990203857,
"rewards/rejected": -2.231950283050537,
"step": 460
},
{
"epoch": 1.9040384858842891,
"grad_norm": 26.954070410698666,
"learning_rate": 3.509010444766674e-08,
"logits/chosen": -11.481412887573242,
"logits/rejected": -10.886090278625488,
"logps/chosen": -33.668968200683594,
"logps/rejected": -61.31293869018555,
"loss": 0.3405,
"rewards/accuracies": 0.8374999761581421,
"rewards/chosen": -0.6743445992469788,
"rewards/margins": 1.5814064741134644,
"rewards/rejected": -2.255751132965088,
"step": 470
},
{
"epoch": 1.9445499430307633,
"grad_norm": 26.90008640112988,
"learning_rate": 3.284960250523237e-08,
"logits/chosen": -11.358150482177734,
"logits/rejected": -10.692336082458496,
"logps/chosen": -32.4575309753418,
"logps/rejected": -58.795753479003906,
"loss": 0.3619,
"rewards/accuracies": 0.8812500238418579,
"rewards/chosen": -0.6383572816848755,
"rewards/margins": 1.5735241174697876,
"rewards/rejected": -2.211881160736084,
"step": 480
},
{
"epoch": 1.9850614001772375,
"grad_norm": 22.343884141895376,
"learning_rate": 3.064748513988914e-08,
"logits/chosen": -11.427927017211914,
"logits/rejected": -10.837053298950195,
"logps/chosen": -32.22171401977539,
"logps/rejected": -59.157142639160156,
"loss": 0.3589,
"rewards/accuracies": 0.887499988079071,
"rewards/chosen": -0.6551386117935181,
"rewards/margins": 1.6512985229492188,
"rewards/rejected": -2.3064370155334473,
"step": 490
},
{
"epoch": 2.025572857323712,
"grad_norm": 21.328409409312936,
"learning_rate": 2.8488680945496145e-08,
"logits/chosen": -11.407200813293457,
"logits/rejected": -10.8077974319458,
"logps/chosen": -32.2757453918457,
"logps/rejected": -58.08913040161133,
"loss": 0.3406,
"rewards/accuracies": 0.8500000238418579,
"rewards/chosen": -0.6142057776451111,
"rewards/margins": 1.6251834630966187,
"rewards/rejected": -2.239389419555664,
"step": 500
},
{
"epoch": 2.066084314470186,
"grad_norm": 21.92232136811071,
"learning_rate": 2.6378021576016464e-08,
"logits/chosen": -11.57470703125,
"logits/rejected": -10.841165542602539,
"logps/chosen": -33.64252471923828,
"logps/rejected": -60.10270309448242,
"loss": 0.3445,
"rewards/accuracies": 0.8500000238418579,
"rewards/chosen": -0.6759551763534546,
"rewards/margins": 1.7165590524673462,
"rewards/rejected": -2.3925139904022217,
"step": 510
},
{
"epoch": 2.1065957716166603,
"grad_norm": 23.376464913490395,
"learning_rate": 2.4320230931715697e-08,
"logits/chosen": -11.309526443481445,
"logits/rejected": -10.582537651062012,
"logps/chosen": -33.03705596923828,
"logps/rejected": -59.86212158203125,
"loss": 0.3228,
"rewards/accuracies": 0.871874988079071,
"rewards/chosen": -0.7096266746520996,
"rewards/margins": 1.673158884048462,
"rewards/rejected": -2.3827853202819824,
"step": 520
},
{
"epoch": 2.1471072287631348,
"grad_norm": 18.289595804333104,
"learning_rate": 2.2319914586525774e-08,
"logits/chosen": -11.554033279418945,
"logits/rejected": -10.901694297790527,
"logps/chosen": -34.2962646484375,
"logps/rejected": -61.11516189575195,
"loss": 0.3355,
"rewards/accuracies": 0.859375,
"rewards/chosen": -0.6974956393241882,
"rewards/margins": 1.66302490234375,
"rewards/rejected": -2.360520601272583,
"step": 530
},
{
"epoch": 2.1876186859096087,
"grad_norm": 23.20641172076809,
"learning_rate": 2.038154948023668e-08,
"logits/chosen": -11.543357849121094,
"logits/rejected": -10.89212703704834,
"logps/chosen": -35.47829055786133,
"logps/rejected": -60.59331130981445,
"loss": 0.3243,
"rewards/accuracies": 0.859375,
"rewards/chosen": -0.7092735171318054,
"rewards/margins": 1.7081466913223267,
"rewards/rejected": -2.4174201488494873,
"step": 540
},
{
"epoch": 2.228130143056083,
"grad_norm": 25.247728097262844,
"learning_rate": 1.8509473898586432e-08,
"logits/chosen": -11.410323143005371,
"logits/rejected": -10.694147109985352,
"logps/chosen": -32.15620803833008,
"logps/rejected": -62.000328063964844,
"loss": 0.2968,
"rewards/accuracies": 0.8999999761581421,
"rewards/chosen": -0.6627341508865356,
"rewards/margins": 1.8608602285385132,
"rewards/rejected": -2.523594617843628,
"step": 550
},
{
"epoch": 2.268641600202557,
"grad_norm": 24.11504934805817,
"learning_rate": 1.6707877763674887e-08,
"logits/chosen": -11.610152244567871,
"logits/rejected": -11.053690910339355,
"logps/chosen": -33.62152862548828,
"logps/rejected": -62.07065963745117,
"loss": 0.3314,
"rewards/accuracies": 0.846875011920929,
"rewards/chosen": -0.7055118083953857,
"rewards/margins": 1.7053735256195068,
"rewards/rejected": -2.4108853340148926,
"step": 560
},
{
"epoch": 2.3091530573490315,
"grad_norm": 24.563417003466505,
"learning_rate": 1.4980793256432472e-08,
"logits/chosen": -11.563300132751465,
"logits/rejected": -10.760797500610352,
"logps/chosen": -34.131736755371094,
"logps/rejected": -62.216285705566406,
"loss": 0.3008,
"rewards/accuracies": 0.893750011920929,
"rewards/chosen": -0.7754543423652649,
"rewards/margins": 1.8263394832611084,
"rewards/rejected": -2.6017937660217285,
"step": 570
},
{
"epoch": 2.349664514495506,
"grad_norm": 32.15938788887828,
"learning_rate": 1.3332085792131965e-08,
"logits/chosen": -11.64229679107666,
"logits/rejected": -10.900683403015137,
"logps/chosen": -35.07176971435547,
"logps/rejected": -61.276954650878906,
"loss": 0.3318,
"rewards/accuracies": 0.862500011920929,
"rewards/chosen": -0.7460302114486694,
"rewards/margins": 1.772905707359314,
"rewards/rejected": -2.5189361572265625,
"step": 580
},
{
"epoch": 2.39017597164198,
"grad_norm": 30.52452011331439,
"learning_rate": 1.1765445369141274e-08,
"logits/chosen": -11.567706108093262,
"logits/rejected": -10.895853042602539,
"logps/chosen": -33.69668960571289,
"logps/rejected": -62.25682830810547,
"loss": 0.3143,
"rewards/accuracies": 0.8656250238418579,
"rewards/chosen": -0.7219974398612976,
"rewards/margins": 1.7591053247451782,
"rewards/rejected": -2.481102466583252,
"step": 590
},
{
"epoch": 2.4306874287884543,
"grad_norm": 19.90507599157542,
"learning_rate": 1.0284378310279368e-08,
"logits/chosen": -11.508050918579102,
"logits/rejected": -10.74337387084961,
"logps/chosen": -32.797637939453125,
"logps/rejected": -63.94108200073242,
"loss": 0.303,
"rewards/accuracies": 0.8812500238418579,
"rewards/chosen": -0.6646064519882202,
"rewards/margins": 2.0007808208465576,
"rewards/rejected": -2.6653873920440674,
"step": 600
},
{
"epoch": 2.4711988859349283,
"grad_norm": 21.194773980773512,
"learning_rate": 8.892199415259499e-09,
"logits/chosen": -11.653422355651855,
"logits/rejected": -11.00390911102295,
"logps/chosen": -35.742061614990234,
"logps/rejected": -65.51417541503906,
"loss": 0.3164,
"rewards/accuracies": 0.8812500238418579,
"rewards/chosen": -0.7325908541679382,
"rewards/margins": 1.797716498374939,
"rewards/rejected": -2.5303072929382324,
"step": 610
},
{
"epoch": 2.5117103430814027,
"grad_norm": 32.194708817268754,
"learning_rate": 7.592024541783343e-09,
"logits/chosen": -11.667333602905273,
"logits/rejected": -10.975776672363281,
"logps/chosen": -34.44770812988281,
"logps/rejected": -59.715660095214844,
"loss": 0.3359,
"rewards/accuracies": 0.862500011920929,
"rewards/chosen": -0.762362003326416,
"rewards/margins": 1.7372121810913086,
"rewards/rejected": -2.4995741844177246,
"step": 620
},
{
"epoch": 2.552221800227877,
"grad_norm": 22.079251389305,
"learning_rate": 6.386763631890313e-09,
"logits/chosen": -11.638243675231934,
"logits/rejected": -10.932289123535156,
"logps/chosen": -32.0985107421875,
"logps/rejected": -59.693931579589844,
"loss": 0.3138,
"rewards/accuracies": 0.878125011920929,
"rewards/chosen": -0.6655905842781067,
"rewards/margins": 1.7755588293075562,
"rewards/rejected": -2.4411492347717285,
"step": 630
},
{
"epoch": 2.592733257374351,
"grad_norm": 27.178193998967842,
"learning_rate": 5.279114199170093e-09,
"logits/chosen": -11.623285293579102,
"logits/rejected": -10.849055290222168,
"logps/chosen": -34.02157211303711,
"logps/rejected": -63.82293701171875,
"loss": 0.2965,
"rewards/accuracies": 0.8812500238418579,
"rewards/chosen": -0.7236124873161316,
"rewards/margins": 1.9638967514038086,
"rewards/rejected": -2.687509059906006,
"step": 640
},
{
"epoch": 2.6332447145208255,
"grad_norm": 23.237750807734354,
"learning_rate": 4.271555291414636e-09,
"logits/chosen": -11.59123420715332,
"logits/rejected": -10.88310432434082,
"logps/chosen": -35.09485626220703,
"logps/rejected": -63.60078811645508,
"loss": 0.3114,
"rewards/accuracies": 0.8968750238418579,
"rewards/chosen": -0.6854656934738159,
"rewards/margins": 1.8889812231063843,
"rewards/rejected": -2.5744469165802,
"step": 650
},
{
"epoch": 2.6737561716672995,
"grad_norm": 23.37299143104811,
"learning_rate": 3.3663419422218677e-09,
"logits/chosen": -11.671598434448242,
"logits/rejected": -10.90255355834961,
"logps/chosen": -34.43938446044922,
"logps/rejected": -63.91756057739258,
"loss": 0.295,
"rewards/accuracies": 0.887499988079071,
"rewards/chosen": -0.690475344657898,
"rewards/margins": 2.0302391052246094,
"rewards/rejected": -2.7207143306732178,
"step": 660
},
{
"epoch": 2.714267628813774,
"grad_norm": 21.948999572300334,
"learning_rate": 2.565500123969183e-09,
"logits/chosen": -11.691490173339844,
"logits/rejected": -11.033750534057617,
"logps/chosen": -34.22254180908203,
"logps/rejected": -60.50248336791992,
"loss": 0.3374,
"rewards/accuracies": 0.840624988079071,
"rewards/chosen": -0.8206373453140259,
"rewards/margins": 1.7287733554840088,
"rewards/rejected": -2.549410343170166,
"step": 670
},
{
"epoch": 2.754779085960248,
"grad_norm": 29.12118466657181,
"learning_rate": 1.8708222134525163e-09,
"logits/chosen": -11.480539321899414,
"logits/rejected": -10.761662483215332,
"logps/chosen": -34.17140197753906,
"logps/rejected": -62.354156494140625,
"loss": 0.3059,
"rewards/accuracies": 0.887499988079071,
"rewards/chosen": -0.7504395246505737,
"rewards/margins": 1.792453408241272,
"rewards/rejected": -2.5428929328918457,
"step": 680
},
{
"epoch": 2.7952905431067223,
"grad_norm": 28.127183360242164,
"learning_rate": 1.283862980339334e-09,
"logits/chosen": -11.640631675720215,
"logits/rejected": -11.045839309692383,
"logps/chosen": -36.884315490722656,
"logps/rejected": -64.67533874511719,
"loss": 0.3354,
"rewards/accuracies": 0.846875011920929,
"rewards/chosen": -0.8064830899238586,
"rewards/margins": 1.7638835906982422,
"rewards/rejected": -2.570366859436035,
"step": 690
},
{
"epoch": 2.8358020002531967,
"grad_norm": 27.365840296062363,
"learning_rate": 8.059361074139293e-10,
"logits/chosen": -11.53732967376709,
"logits/rejected": -10.982316970825195,
"logps/chosen": -34.8204231262207,
"logps/rejected": -60.4800910949707,
"loss": 0.3765,
"rewards/accuracies": 0.875,
"rewards/chosen": -0.77055424451828,
"rewards/margins": 1.6154954433441162,
"rewards/rejected": -2.386049747467041,
"step": 700
},
{
"epoch": 2.8763134573996707,
"grad_norm": 23.555963752594412,
"learning_rate": 4.381112504031337e-10,
"logits/chosen": -11.630104064941406,
"logits/rejected": -10.956541061401367,
"logps/chosen": -34.54252243041992,
"logps/rejected": -60.827423095703125,
"loss": 0.3183,
"rewards/accuracies": 0.871874988079071,
"rewards/chosen": -0.7591537833213806,
"rewards/margins": 1.6892755031585693,
"rewards/rejected": -2.4484293460845947,
"step": 710
},
{
"epoch": 2.916824914546145,
"grad_norm": 21.502729719070256,
"learning_rate": 1.8121164396283638e-10,
"logits/chosen": -11.620569229125977,
"logits/rejected": -10.932313919067383,
"logps/chosen": -36.305519104003906,
"logps/rejected": -66.49055480957031,
"loss": 0.3241,
"rewards/accuracies": 0.859375,
"rewards/chosen": -0.7203236222267151,
"rewards/margins": 1.99014413356781,
"rewards/rejected": -2.71046781539917,
"step": 720
},
{
"epoch": 2.9573363716926195,
"grad_norm": 22.58072518400253,
"learning_rate": 3.581225918342645e-11,
"logits/chosen": -11.6244535446167,
"logits/rejected": -10.896235466003418,
"logps/chosen": -31.406940460205078,
"logps/rejected": -62.654396057128906,
"loss": 0.3115,
"rewards/accuracies": 0.887499988079071,
"rewards/chosen": -0.7061868906021118,
"rewards/margins": 1.8940362930297852,
"rewards/rejected": -2.6002233028411865,
"step": 730
},
{
"epoch": 2.9897455374097985,
"step": 738,
"total_flos": 0.0,
"train_loss": 0.4158425409457871,
"train_runtime": 20368.7971,
"train_samples_per_second": 2.327,
"train_steps_per_second": 0.036
}
],
"logging_steps": 10,
"max_steps": 738,
"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": 0.0,
"train_batch_size": 1,
"trial_name": null,
"trial_params": null
}