silviasapora's picture
Model save
a2a488a verified
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 2.9765925925925925,
"eval_steps": 500,
"global_step": 315,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.047407407407407405,
"grad_norm": 107.5957260131836,
"learning_rate": 7.8125e-06,
"log_odds_chosen": 0.26999545097351074,
"log_odds_ratio": -0.6788457632064819,
"logits/chosen": -2.960033655166626,
"logits/rejected": -2.8769211769104004,
"logps/chosen": -1.1197593212127686,
"logps/rejected": -1.3242356777191162,
"loss": 47.1372,
"nll_loss": 1.4214212894439697,
"rewards/accuracies": 0.574999988079071,
"rewards/chosen": -0.11197594553232193,
"rewards/margins": 0.020447632297873497,
"rewards/rejected": -0.13242356479167938,
"step": 5
},
{
"epoch": 0.09481481481481481,
"grad_norm": 59.963645935058594,
"learning_rate": 1.5625e-05,
"log_odds_chosen": 0.17921434342861176,
"log_odds_ratio": -0.6778210401535034,
"logits/chosen": -2.9659476280212402,
"logits/rejected": -2.8464667797088623,
"logps/chosen": -0.9941017031669617,
"logps/rejected": -1.1148567199707031,
"loss": 41.0221,
"nll_loss": 1.266223669052124,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.09941016137599945,
"rewards/margins": 0.012075509876012802,
"rewards/rejected": -0.11148568242788315,
"step": 10
},
{
"epoch": 0.14222222222222222,
"grad_norm": 54.22554016113281,
"learning_rate": 2.34375e-05,
"log_odds_chosen": 0.19334463775157928,
"log_odds_ratio": -0.6705215573310852,
"logits/chosen": -2.8042972087860107,
"logits/rejected": -2.72182035446167,
"logps/chosen": -0.7811649441719055,
"logps/rejected": -0.8896393775939941,
"loss": 38.9195,
"nll_loss": 1.2349504232406616,
"rewards/accuracies": 0.581250011920929,
"rewards/chosen": -0.07811649888753891,
"rewards/margins": 0.01084743719547987,
"rewards/rejected": -0.08896392583847046,
"step": 15
},
{
"epoch": 0.18962962962962962,
"grad_norm": 46.67384338378906,
"learning_rate": 3.125e-05,
"log_odds_chosen": 0.16569766402244568,
"log_odds_ratio": -0.6850992441177368,
"logits/chosen": -2.8097779750823975,
"logits/rejected": -2.758016347885132,
"logps/chosen": -0.8054255247116089,
"logps/rejected": -0.9155207872390747,
"loss": 36.5441,
"nll_loss": 1.1204253435134888,
"rewards/accuracies": 0.581250011920929,
"rewards/chosen": -0.08054256439208984,
"rewards/margins": 0.011009521782398224,
"rewards/rejected": -0.09155207872390747,
"step": 20
},
{
"epoch": 0.23703703703703705,
"grad_norm": 32.77006530761719,
"learning_rate": 3.90625e-05,
"log_odds_chosen": 0.11824393272399902,
"log_odds_ratio": -0.7369982004165649,
"logits/chosen": -2.8846487998962402,
"logits/rejected": -2.8371191024780273,
"logps/chosen": -0.7971025705337524,
"logps/rejected": -0.8806389570236206,
"loss": 35.5873,
"nll_loss": 1.1432932615280151,
"rewards/accuracies": 0.518750011920929,
"rewards/chosen": -0.07971025258302689,
"rewards/margins": 0.008353643119335175,
"rewards/rejected": -0.08806388825178146,
"step": 25
},
{
"epoch": 0.28444444444444444,
"grad_norm": 44.49482345581055,
"learning_rate": 4.6875e-05,
"log_odds_chosen": 0.11238924413919449,
"log_odds_ratio": -0.7262482643127441,
"logits/chosen": -2.8552589416503906,
"logits/rejected": -2.7313339710235596,
"logps/chosen": -0.760868489742279,
"logps/rejected": -0.8262151479721069,
"loss": 34.6256,
"nll_loss": 1.0859296321868896,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.0760868489742279,
"rewards/margins": 0.006534676067531109,
"rewards/rejected": -0.08262152969837189,
"step": 30
},
{
"epoch": 0.33185185185185184,
"grad_norm": 35.2032356262207,
"learning_rate": 4.998613757348784e-05,
"log_odds_chosen": 0.11870062351226807,
"log_odds_ratio": -0.7236028909683228,
"logits/chosen": -2.8732895851135254,
"logits/rejected": -2.9246368408203125,
"logps/chosen": -0.7803806066513062,
"logps/rejected": -0.8395193219184875,
"loss": 33.8984,
"nll_loss": 1.0961161851882935,
"rewards/accuracies": 0.574999988079071,
"rewards/chosen": -0.0780380591750145,
"rewards/margins": 0.005913878325372934,
"rewards/rejected": -0.083951935172081,
"step": 35
},
{
"epoch": 0.37925925925925924,
"grad_norm": 31.074262619018555,
"learning_rate": 4.990147841143462e-05,
"log_odds_chosen": 0.16566646099090576,
"log_odds_ratio": -0.6969919204711914,
"logits/chosen": -2.8940587043762207,
"logits/rejected": -2.846470832824707,
"logps/chosen": -0.7750462293624878,
"logps/rejected": -0.8606101870536804,
"loss": 33.9215,
"nll_loss": 1.0579229593276978,
"rewards/accuracies": 0.5687500238418579,
"rewards/chosen": -0.07750462740659714,
"rewards/margins": 0.008556395769119263,
"rewards/rejected": -0.086061030626297,
"step": 40
},
{
"epoch": 0.4266666666666667,
"grad_norm": 26.24691390991211,
"learning_rate": 4.97401218720448e-05,
"log_odds_chosen": 0.15471526980400085,
"log_odds_ratio": -0.724504828453064,
"logits/chosen": -2.8084166049957275,
"logits/rejected": -2.8175244331359863,
"logps/chosen": -0.7267907857894897,
"logps/rejected": -0.8097001314163208,
"loss": 33.6756,
"nll_loss": 1.0580939054489136,
"rewards/accuracies": 0.512499988079071,
"rewards/chosen": -0.07267908751964569,
"rewards/margins": 0.00829092413187027,
"rewards/rejected": -0.08097001165151596,
"step": 45
},
{
"epoch": 0.4740740740740741,
"grad_norm": 33.051918029785156,
"learning_rate": 4.9502564938797946e-05,
"log_odds_chosen": 0.11409230530261993,
"log_odds_ratio": -0.7364336848258972,
"logits/chosen": -2.795748472213745,
"logits/rejected": -2.805166006088257,
"logps/chosen": -0.7951005697250366,
"logps/rejected": -0.8444055318832397,
"loss": 35.2252,
"nll_loss": 1.1173629760742188,
"rewards/accuracies": 0.53125,
"rewards/chosen": -0.07951006293296814,
"rewards/margins": 0.004930487833917141,
"rewards/rejected": -0.08444055169820786,
"step": 50
},
{
"epoch": 0.5214814814814814,
"grad_norm": 25.374826431274414,
"learning_rate": 4.918953929490768e-05,
"log_odds_chosen": 0.09201577305793762,
"log_odds_ratio": -0.7416759729385376,
"logits/chosen": -2.7740275859832764,
"logits/rejected": -2.732271194458008,
"logps/chosen": -0.7545334100723267,
"logps/rejected": -0.8210805058479309,
"loss": 32.9599,
"nll_loss": 1.0351916551589966,
"rewards/accuracies": 0.550000011920929,
"rewards/chosen": -0.07545334100723267,
"rewards/margins": 0.006654707249253988,
"rewards/rejected": -0.0821080431342125,
"step": 55
},
{
"epoch": 0.5688888888888889,
"grad_norm": 27.035547256469727,
"learning_rate": 4.88020090697132e-05,
"log_odds_chosen": 0.17979878187179565,
"log_odds_ratio": -0.6883889436721802,
"logits/chosen": -2.917654037475586,
"logits/rejected": -2.8189804553985596,
"logps/chosen": -0.7465445399284363,
"logps/rejected": -0.8283771276473999,
"loss": 32.1903,
"nll_loss": 1.0064656734466553,
"rewards/accuracies": 0.5562499761581421,
"rewards/chosen": -0.0746544599533081,
"rewards/margins": 0.008183250203728676,
"rewards/rejected": -0.08283771574497223,
"step": 60
},
{
"epoch": 0.6162962962962963,
"grad_norm": 29.824989318847656,
"learning_rate": 4.834116786912897e-05,
"log_odds_chosen": 0.14272919297218323,
"log_odds_ratio": -0.7043147087097168,
"logits/chosen": -2.823390007019043,
"logits/rejected": -2.815788984298706,
"logps/chosen": -0.7772132754325867,
"logps/rejected": -0.8756623268127441,
"loss": 32.3031,
"nll_loss": 1.006659984588623,
"rewards/accuracies": 0.5687500238418579,
"rewards/chosen": -0.07772133499383926,
"rewards/margins": 0.00984489731490612,
"rewards/rejected": -0.08756622672080994,
"step": 65
},
{
"epoch": 0.6637037037037037,
"grad_norm": 25.48737907409668,
"learning_rate": 4.7808435099299045e-05,
"log_odds_chosen": 0.10067082941532135,
"log_odds_ratio": -0.7191606760025024,
"logits/chosen": -2.91566801071167,
"logits/rejected": -2.7849342823028564,
"logps/chosen": -0.7710874676704407,
"logps/rejected": -0.8389760255813599,
"loss": 31.6335,
"nll_loss": 1.0086121559143066,
"rewards/accuracies": 0.574999988079071,
"rewards/chosen": -0.07710875570774078,
"rewards/margins": 0.0067888544872403145,
"rewards/rejected": -0.08389760553836823,
"step": 70
},
{
"epoch": 0.7111111111111111,
"grad_norm": 26.371871948242188,
"learning_rate": 4.720545159477922e-05,
"log_odds_chosen": 0.32483047246932983,
"log_odds_ratio": -0.6155678033828735,
"logits/chosen": -2.8636772632598877,
"logits/rejected": -2.8909075260162354,
"logps/chosen": -0.692024827003479,
"logps/rejected": -0.8595916628837585,
"loss": 32.0981,
"nll_loss": 1.0088964700698853,
"rewards/accuracies": 0.6812499761581421,
"rewards/chosen": -0.0692024827003479,
"rewards/margins": 0.016756689175963402,
"rewards/rejected": -0.08595915883779526,
"step": 75
},
{
"epoch": 0.7585185185185185,
"grad_norm": 25.50819206237793,
"learning_rate": 4.653407456471222e-05,
"log_odds_chosen": 0.19090376794338226,
"log_odds_ratio": -0.6875820159912109,
"logits/chosen": -2.970930814743042,
"logits/rejected": -2.879568338394165,
"logps/chosen": -0.784750759601593,
"logps/rejected": -0.8799096941947937,
"loss": 32.0035,
"nll_loss": 1.0407371520996094,
"rewards/accuracies": 0.637499988079071,
"rewards/chosen": -0.07847508043050766,
"rewards/margins": 0.009515885263681412,
"rewards/rejected": -0.08799097687005997,
"step": 80
},
{
"epoch": 0.8059259259259259,
"grad_norm": 21.803564071655273,
"learning_rate": 4.579637187256222e-05,
"log_odds_chosen": 0.15773186087608337,
"log_odds_ratio": -0.6912397742271423,
"logits/chosen": -2.9440784454345703,
"logits/rejected": -2.7553608417510986,
"logps/chosen": -0.7397053241729736,
"logps/rejected": -0.8279611468315125,
"loss": 31.9112,
"nll_loss": 1.0011701583862305,
"rewards/accuracies": 0.581250011920929,
"rewards/chosen": -0.07397053390741348,
"rewards/margins": 0.00882558710873127,
"rewards/rejected": -0.0827961266040802,
"step": 85
},
{
"epoch": 0.8533333333333334,
"grad_norm": 23.784290313720703,
"learning_rate": 4.499461566702685e-05,
"log_odds_chosen": 0.22803974151611328,
"log_odds_ratio": -0.6771184206008911,
"logits/chosen": -2.9111487865448,
"logits/rejected": -2.8253490924835205,
"logps/chosen": -0.7062662839889526,
"logps/rejected": -0.8198903799057007,
"loss": 32.583,
"nll_loss": 0.9986779093742371,
"rewards/accuracies": 0.612500011920929,
"rewards/chosen": -0.0706266313791275,
"rewards/margins": 0.011362412944436073,
"rewards/rejected": -0.08198904246091843,
"step": 90
},
{
"epoch": 0.9007407407407407,
"grad_norm": 24.204864501953125,
"learning_rate": 4.413127538374411e-05,
"log_odds_chosen": 0.25104185938835144,
"log_odds_ratio": -0.6683576703071594,
"logits/chosen": -2.8897457122802734,
"logits/rejected": -2.7754015922546387,
"logps/chosen": -0.6870729923248291,
"logps/rejected": -0.8187867403030396,
"loss": 31.9808,
"nll_loss": 0.9533413052558899,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.06870730221271515,
"rewards/margins": 0.013171376660466194,
"rewards/rejected": -0.08187868446111679,
"step": 95
},
{
"epoch": 0.9481481481481482,
"grad_norm": 25.084274291992188,
"learning_rate": 4.320901013934887e-05,
"log_odds_chosen": 0.054086942225694656,
"log_odds_ratio": -0.7604089379310608,
"logits/chosen": -2.862680435180664,
"logits/rejected": -2.7483227252960205,
"logps/chosen": -0.7729731798171997,
"logps/rejected": -0.8076101541519165,
"loss": 32.02,
"nll_loss": 1.0369198322296143,
"rewards/accuracies": 0.53125,
"rewards/chosen": -0.07729730755090714,
"rewards/margins": 0.003463698085397482,
"rewards/rejected": -0.08076101541519165,
"step": 100
},
{
"epoch": 0.9955555555555555,
"grad_norm": 24.100692749023438,
"learning_rate": 4.223066054130568e-05,
"log_odds_chosen": 0.15314075350761414,
"log_odds_ratio": -0.70216304063797,
"logits/chosen": -2.9259018898010254,
"logits/rejected": -2.785470724105835,
"logps/chosen": -0.7330957055091858,
"logps/rejected": -0.851188063621521,
"loss": 31.8368,
"nll_loss": 0.9923515319824219,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.07330957055091858,
"rewards/margins": 0.011809232644736767,
"rewards/rejected": -0.08511880785226822,
"step": 105
},
{
"epoch": 1.037925925925926,
"grad_norm": 24.946685791015625,
"learning_rate": 4.1199239938743797e-05,
"log_odds_chosen": 0.4960819184780121,
"log_odds_ratio": -0.5754141807556152,
"logits/chosen": -2.847261667251587,
"logits/rejected": -2.7735049724578857,
"logps/chosen": -0.5829983949661255,
"logps/rejected": -0.8058205246925354,
"loss": 24.0306,
"nll_loss": 0.8102088570594788,
"rewards/accuracies": 0.7132866978645325,
"rewards/chosen": -0.05829983949661255,
"rewards/margins": 0.022282211109995842,
"rewards/rejected": -0.08058205991983414,
"step": 110
},
{
"epoch": 1.0853333333333333,
"grad_norm": 26.976266860961914,
"learning_rate": 4.0117925141242174e-05,
"log_odds_chosen": 0.654797375202179,
"log_odds_ratio": -0.5052435994148254,
"logits/chosen": -2.877814769744873,
"logits/rejected": -2.680081844329834,
"logps/chosen": -0.5720535516738892,
"logps/rejected": -0.8802263140678406,
"loss": 25.9472,
"nll_loss": 0.7921805381774902,
"rewards/accuracies": 0.793749988079071,
"rewards/chosen": -0.057205356657505035,
"rewards/margins": 0.03081728145480156,
"rewards/rejected": -0.0880226343870163,
"step": 115
},
{
"epoch": 1.1327407407407408,
"grad_norm": 24.159303665161133,
"learning_rate": 3.899004663415084e-05,
"log_odds_chosen": 0.6985219120979309,
"log_odds_ratio": -0.5000964403152466,
"logits/chosen": -2.8363966941833496,
"logits/rejected": -2.77884840965271,
"logps/chosen": -0.5254793167114258,
"logps/rejected": -0.8278259038925171,
"loss": 25.8696,
"nll_loss": 0.7871843576431274,
"rewards/accuracies": 0.762499988079071,
"rewards/chosen": -0.05254793167114258,
"rewards/margins": 0.03023465909063816,
"rewards/rejected": -0.08278258144855499,
"step": 120
},
{
"epoch": 1.1801481481481482,
"grad_norm": 25.176586151123047,
"learning_rate": 3.781907832058587e-05,
"log_odds_chosen": 0.7422643303871155,
"log_odds_ratio": -0.49143218994140625,
"logits/chosen": -2.8330469131469727,
"logits/rejected": -2.794886827468872,
"logps/chosen": -0.5751325488090515,
"logps/rejected": -0.9076019525527954,
"loss": 24.3083,
"nll_loss": 0.7706891298294067,
"rewards/accuracies": 0.7875000238418579,
"rewards/chosen": -0.05751325935125351,
"rewards/margins": 0.033246930688619614,
"rewards/rejected": -0.09076018631458282,
"step": 125
},
{
"epoch": 1.2275555555555555,
"grad_norm": 24.129467010498047,
"learning_rate": 3.660862682169282e-05,
"log_odds_chosen": 0.6279364824295044,
"log_odds_ratio": -0.52507483959198,
"logits/chosen": -2.8697004318237305,
"logits/rejected": -2.7813589572906494,
"logps/chosen": -0.5734705328941345,
"logps/rejected": -0.8611089587211609,
"loss": 24.637,
"nll_loss": 0.7875975966453552,
"rewards/accuracies": 0.737500011920929,
"rewards/chosen": -0.05734705924987793,
"rewards/margins": 0.028763839974999428,
"rewards/rejected": -0.0861109048128128,
"step": 130
},
{
"epoch": 1.274962962962963,
"grad_norm": 22.866924285888672,
"learning_rate": 3.5362420368134356e-05,
"log_odds_chosen": 0.6941056251525879,
"log_odds_ratio": -0.52723228931427,
"logits/chosen": -2.845369577407837,
"logits/rejected": -2.7874202728271484,
"logps/chosen": -0.508094310760498,
"logps/rejected": -0.7907238006591797,
"loss": 24.905,
"nll_loss": 0.7590480446815491,
"rewards/accuracies": 0.7250000238418579,
"rewards/chosen": -0.050809431821107864,
"rewards/margins": 0.028262948617339134,
"rewards/rejected": -0.07907238602638245,
"step": 135
},
{
"epoch": 1.3223703703703704,
"grad_norm": 26.00486183166504,
"learning_rate": 3.408429731701635e-05,
"log_odds_chosen": 0.6756368279457092,
"log_odds_ratio": -0.5294371247291565,
"logits/chosen": -2.7922794818878174,
"logits/rejected": -2.8006014823913574,
"logps/chosen": -0.6041140556335449,
"logps/rejected": -0.9101554751396179,
"loss": 26.2335,
"nll_loss": 0.8219060897827148,
"rewards/accuracies": 0.7437499761581421,
"rewards/chosen": -0.06041140481829643,
"rewards/margins": 0.03060414455831051,
"rewards/rejected": -0.09101554751396179,
"step": 140
},
{
"epoch": 1.3697777777777778,
"grad_norm": 25.439064025878906,
"learning_rate": 3.2778194329621104e-05,
"log_odds_chosen": 0.5306147933006287,
"log_odds_ratio": -0.5561264157295227,
"logits/chosen": -2.901402235031128,
"logits/rejected": -2.760891914367676,
"logps/chosen": -0.6078349351882935,
"logps/rejected": -0.8788266181945801,
"loss": 25.9841,
"nll_loss": 0.8179038166999817,
"rewards/accuracies": 0.6812499761581421,
"rewards/chosen": -0.06078348681330681,
"rewards/margins": 0.027099166065454483,
"rewards/rejected": -0.08788266032934189,
"step": 145
},
{
"epoch": 1.417185185185185,
"grad_norm": 29.468650817871094,
"learning_rate": 3.144813424636031e-05,
"log_odds_chosen": 0.5843684673309326,
"log_odds_ratio": -0.549953043460846,
"logits/chosen": -2.8449625968933105,
"logits/rejected": -2.8639121055603027,
"logps/chosen": -0.5666799545288086,
"logps/rejected": -0.8146460652351379,
"loss": 25.1356,
"nll_loss": 0.7769128084182739,
"rewards/accuracies": 0.6875,
"rewards/chosen": -0.0566679947078228,
"rewards/margins": 0.024796610698103905,
"rewards/rejected": -0.08146460354328156,
"step": 150
},
{
"epoch": 1.4645925925925927,
"grad_norm": 25.020612716674805,
"learning_rate": 3.0098213696293542e-05,
"log_odds_chosen": 0.6784299612045288,
"log_odds_ratio": -0.5230305790901184,
"logits/chosen": -2.895751476287842,
"logits/rejected": -2.7628910541534424,
"logps/chosen": -0.5524539351463318,
"logps/rejected": -0.8418664932250977,
"loss": 25.5094,
"nll_loss": 0.8113977313041687,
"rewards/accuracies": 0.737500011920929,
"rewards/chosen": -0.055245399475097656,
"rewards/margins": 0.028941255062818527,
"rewards/rejected": -0.08418664336204529,
"step": 155
},
{
"epoch": 1.512,
"grad_norm": 26.339784622192383,
"learning_rate": 2.8732590479375165e-05,
"log_odds_chosen": 0.5665224194526672,
"log_odds_ratio": -0.5301405191421509,
"logits/chosen": -2.8651890754699707,
"logits/rejected": -2.8648974895477295,
"logps/chosen": -0.628059983253479,
"logps/rejected": -0.9070053100585938,
"loss": 26.3028,
"nll_loss": 0.8419073224067688,
"rewards/accuracies": 0.7250000238418579,
"rewards/chosen": -0.06280599534511566,
"rewards/margins": 0.02789454162120819,
"rewards/rejected": -0.09070053696632385,
"step": 160
},
{
"epoch": 1.5594074074074074,
"grad_norm": 24.364530563354492,
"learning_rate": 2.7355470760292956e-05,
"log_odds_chosen": 0.8036432266235352,
"log_odds_ratio": -0.4671328663825989,
"logits/chosen": -2.807288408279419,
"logits/rejected": -2.860252857208252,
"logps/chosen": -0.5591052174568176,
"logps/rejected": -0.9000210762023926,
"loss": 25.9433,
"nll_loss": 0.7956122756004333,
"rewards/accuracies": 0.793749988079071,
"rewards/chosen": -0.05591052025556564,
"rewards/margins": 0.03409158065915108,
"rewards/rejected": -0.09000210464000702,
"step": 165
},
{
"epoch": 1.6068148148148147,
"grad_norm": 26.519641876220703,
"learning_rate": 2.597109611334169e-05,
"log_odds_chosen": 0.7449638843536377,
"log_odds_ratio": -0.4777608811855316,
"logits/chosen": -2.812837600708008,
"logits/rejected": -2.7875418663024902,
"logps/chosen": -0.5420358180999756,
"logps/rejected": -0.8753792643547058,
"loss": 25.1331,
"nll_loss": 0.7836862802505493,
"rewards/accuracies": 0.793749988079071,
"rewards/chosen": -0.054203588515520096,
"rewards/margins": 0.03333434462547302,
"rewards/rejected": -0.08753792196512222,
"step": 170
},
{
"epoch": 1.6542222222222223,
"grad_norm": 24.914644241333008,
"learning_rate": 2.458373045823404e-05,
"log_odds_chosen": 0.6300590634346008,
"log_odds_ratio": -0.5320771932601929,
"logits/chosen": -2.8087375164031982,
"logits/rejected": -2.713311195373535,
"logps/chosen": -0.5621276497840881,
"logps/rejected": -0.8645404577255249,
"loss": 25.1251,
"nll_loss": 0.8012740015983582,
"rewards/accuracies": 0.75,
"rewards/chosen": -0.056212764233350754,
"rewards/margins": 0.030241286382079124,
"rewards/rejected": -0.08645405620336533,
"step": 175
},
{
"epoch": 1.7016296296296296,
"grad_norm": 24.421783447265625,
"learning_rate": 2.3197646927086697e-05,
"log_odds_chosen": 0.511154294013977,
"log_odds_ratio": -0.58769690990448,
"logits/chosen": -2.733468770980835,
"logits/rejected": -2.773117780685425,
"logps/chosen": -0.5480560064315796,
"logps/rejected": -0.7600927352905273,
"loss": 25.3565,
"nll_loss": 0.7882987260818481,
"rewards/accuracies": 0.675000011920929,
"rewards/chosen": -0.054805606603622437,
"rewards/margins": 0.021203670650720596,
"rewards/rejected": -0.07600927352905273,
"step": 180
},
{
"epoch": 1.749037037037037,
"grad_norm": 24.164962768554688,
"learning_rate": 2.1817114703032176e-05,
"log_odds_chosen": 0.6179710030555725,
"log_odds_ratio": -0.5148657560348511,
"logits/chosen": -2.845136880874634,
"logits/rejected": -2.8214142322540283,
"logps/chosen": -0.5659202337265015,
"logps/rejected": -0.8389551043510437,
"loss": 25.83,
"nll_loss": 0.8041189312934875,
"rewards/accuracies": 0.762499988079071,
"rewards/chosen": -0.056592028588056564,
"rewards/margins": 0.027303475886583328,
"rewards/rejected": -0.08389550447463989,
"step": 185
},
{
"epoch": 1.7964444444444445,
"grad_norm": 24.201217651367188,
"learning_rate": 2.0446385870993467e-05,
"log_odds_chosen": 0.7095943689346313,
"log_odds_ratio": -0.4986342489719391,
"logits/chosen": -2.864686965942383,
"logits/rejected": -2.7122364044189453,
"logps/chosen": -0.5929415225982666,
"logps/rejected": -0.9149606823921204,
"loss": 25.6013,
"nll_loss": 0.7932403087615967,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": -0.05929415300488472,
"rewards/margins": 0.032201915979385376,
"rewards/rejected": -0.0914960727095604,
"step": 190
},
{
"epoch": 1.8438518518518519,
"grad_norm": 25.97276496887207,
"learning_rate": 1.9089682321121834e-05,
"log_odds_chosen": 0.7197151184082031,
"log_odds_ratio": -0.5103024840354919,
"logits/chosen": -2.8166863918304443,
"logits/rejected": -2.6799378395080566,
"logps/chosen": -0.5752886533737183,
"logps/rejected": -0.8961560130119324,
"loss": 25.1918,
"nll_loss": 0.7882518172264099,
"rewards/accuracies": 0.768750011920929,
"rewards/chosen": -0.05752886086702347,
"rewards/margins": 0.03208674117922783,
"rewards/rejected": -0.089615598320961,
"step": 195
},
{
"epoch": 1.8912592592592592,
"grad_norm": 24.435470581054688,
"learning_rate": 1.775118274523545e-05,
"log_odds_chosen": 0.6891900300979614,
"log_odds_ratio": -0.5204278230667114,
"logits/chosen": -2.885361909866333,
"logits/rejected": -2.828627109527588,
"logps/chosen": -0.5744878053665161,
"logps/rejected": -0.892475962638855,
"loss": 26.1889,
"nll_loss": 0.7850193977355957,
"rewards/accuracies": 0.7437499761581421,
"rewards/chosen": -0.05744878575205803,
"rewards/margins": 0.031798820942640305,
"rewards/rejected": -0.08924759924411774,
"step": 200
},
{
"epoch": 1.9386666666666668,
"grad_norm": 22.373199462890625,
"learning_rate": 1.643500976631037e-05,
"log_odds_chosen": 0.554446816444397,
"log_odds_ratio": -0.5625575184822083,
"logits/chosen": -2.8687710762023926,
"logits/rejected": -2.728229284286499,
"logps/chosen": -0.5928065776824951,
"logps/rejected": -0.8653701543807983,
"loss": 25.113,
"nll_loss": 0.7887372374534607,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": -0.05928065627813339,
"rewards/margins": 0.027256358414888382,
"rewards/rejected": -0.08653701841831207,
"step": 205
},
{
"epoch": 1.986074074074074,
"grad_norm": 23.288532257080078,
"learning_rate": 1.514521724066537e-05,
"log_odds_chosen": 0.6083322763442993,
"log_odds_ratio": -0.5458379983901978,
"logits/chosen": -2.854628086090088,
"logits/rejected": -2.8980631828308105,
"logps/chosen": -0.5750885605812073,
"logps/rejected": -0.8365706205368042,
"loss": 23.8154,
"nll_loss": 0.7530295252799988,
"rewards/accuracies": 0.731249988079071,
"rewards/chosen": -0.05750884860754013,
"rewards/margins": 0.02614821121096611,
"rewards/rejected": -0.08365706354379654,
"step": 210
},
{
"epoch": 2.0284444444444443,
"grad_norm": 24.61016082763672,
"learning_rate": 1.3885777771950348e-05,
"log_odds_chosen": 1.0653361082077026,
"log_odds_ratio": -0.4205915629863739,
"logits/chosen": -2.783123254776001,
"logits/rejected": -2.719882011413574,
"logps/chosen": -0.4712402820587158,
"logps/rejected": -0.8792357444763184,
"loss": 19.6217,
"nll_loss": 0.6721786856651306,
"rewards/accuracies": 0.8181818127632141,
"rewards/chosen": -0.04712402820587158,
"rewards/margins": 0.040799543261528015,
"rewards/rejected": -0.08792358636856079,
"step": 215
},
{
"epoch": 2.075851851851852,
"grad_norm": 33.874610900878906,
"learning_rate": 1.2660570475395683e-05,
"log_odds_chosen": 1.245701789855957,
"log_odds_ratio": -0.36852705478668213,
"logits/chosen": -2.7865612506866455,
"logits/rejected": -2.7448060512542725,
"logps/chosen": -0.4370403289794922,
"logps/rejected": -0.9250956773757935,
"loss": 19.9741,
"nll_loss": 0.6481715440750122,
"rewards/accuracies": 0.84375,
"rewards/chosen": -0.04370403289794922,
"rewards/margins": 0.04880553111433983,
"rewards/rejected": -0.09250956773757935,
"step": 220
},
{
"epoch": 2.1232592592592594,
"grad_norm": 26.486011505126953,
"learning_rate": 1.1473369030008974e-05,
"log_odds_chosen": 1.2085529565811157,
"log_odds_ratio": -0.38652700185775757,
"logits/chosen": -2.754307270050049,
"logits/rejected": -2.792224168777466,
"logps/chosen": -0.4022933840751648,
"logps/rejected": -0.8532126545906067,
"loss": 20.6782,
"nll_loss": 0.6350698471069336,
"rewards/accuracies": 0.8500000238418579,
"rewards/chosen": -0.04022934287786484,
"rewards/margins": 0.045091934502124786,
"rewards/rejected": -0.08532126247882843,
"step": 225
},
{
"epoch": 2.1706666666666665,
"grad_norm": 26.554258346557617,
"learning_rate": 1.0327830055518842e-05,
"log_odds_chosen": 1.3051682710647583,
"log_odds_ratio": -0.3686615526676178,
"logits/chosen": -2.7639107704162598,
"logits/rejected": -2.693751811981201,
"logps/chosen": -0.41990265250205994,
"logps/rejected": -0.8988162279129028,
"loss": 19.327,
"nll_loss": 0.5870728492736816,
"rewards/accuracies": 0.8500000238418579,
"rewards/chosen": -0.04199026897549629,
"rewards/margins": 0.04789134860038757,
"rewards/rejected": -0.08988161385059357,
"step": 230
},
{
"epoch": 2.218074074074074,
"grad_norm": 31.239700317382812,
"learning_rate": 9.227481849865235e-06,
"log_odds_chosen": 1.1520278453826904,
"log_odds_ratio": -0.41394537687301636,
"logits/chosen": -2.778120756149292,
"logits/rejected": -2.7755844593048096,
"logps/chosen": -0.4497902989387512,
"logps/rejected": -0.8696093559265137,
"loss": 20.2714,
"nll_loss": 0.6394407153129578,
"rewards/accuracies": 0.8374999761581421,
"rewards/chosen": -0.044979028403759,
"rewards/margins": 0.04198190197348595,
"rewards/rejected": -0.08696093410253525,
"step": 235
},
{
"epoch": 2.2654814814814817,
"grad_norm": 26.99662971496582,
"learning_rate": 8.175713521924978e-06,
"log_odds_chosen": 1.287632703781128,
"log_odds_ratio": -0.37112313508987427,
"logits/chosen": -2.7816336154937744,
"logits/rejected": -2.8271846771240234,
"logps/chosen": -0.43638721108436584,
"logps/rejected": -0.9052692651748657,
"loss": 20.5114,
"nll_loss": 0.6593940854072571,
"rewards/accuracies": 0.862500011920929,
"rewards/chosen": -0.04363872855901718,
"rewards/margins": 0.046888213604688644,
"rewards/rejected": -0.09052693098783493,
"step": 240
},
{
"epoch": 2.3128888888888888,
"grad_norm": 27.01542854309082,
"learning_rate": 7.1757645529443665e-06,
"log_odds_chosen": 1.3749712705612183,
"log_odds_ratio": -0.3259935975074768,
"logits/chosen": -2.788048028945923,
"logits/rejected": -2.701140880584717,
"logps/chosen": -0.4096066951751709,
"logps/rejected": -0.91987144947052,
"loss": 19.5069,
"nll_loss": 0.62156742811203,
"rewards/accuracies": 0.8999999761581421,
"rewards/chosen": -0.04096066951751709,
"rewards/margins": 0.05102648213505745,
"rewards/rejected": -0.09198714792728424,
"step": 245
},
{
"epoch": 2.3602962962962963,
"grad_norm": 27.530044555664062,
"learning_rate": 6.230714818829733e-06,
"log_odds_chosen": 1.3079187870025635,
"log_odds_ratio": -0.33900976181030273,
"logits/chosen": -2.7988991737365723,
"logits/rejected": -2.7709498405456543,
"logps/chosen": -0.4449501931667328,
"logps/rejected": -0.9620338678359985,
"loss": 19.171,
"nll_loss": 0.6168783903121948,
"rewards/accuracies": 0.8812500238418579,
"rewards/chosen": -0.04449502006173134,
"rewards/margins": 0.05170837789773941,
"rewards/rejected": -0.09620338678359985,
"step": 250
},
{
"epoch": 2.407703703703704,
"grad_norm": 28.68503189086914,
"learning_rate": 5.343475104027743e-06,
"log_odds_chosen": 1.423537254333496,
"log_odds_ratio": -0.328805148601532,
"logits/chosen": -2.7419135570526123,
"logits/rejected": -2.7263638973236084,
"logps/chosen": -0.3865131735801697,
"logps/rejected": -0.9120645523071289,
"loss": 19.2,
"nll_loss": 0.6022146344184875,
"rewards/accuracies": 0.887499988079071,
"rewards/chosen": -0.03865131735801697,
"rewards/margins": 0.0525551438331604,
"rewards/rejected": -0.09120644629001617,
"step": 255
},
{
"epoch": 2.455111111111111,
"grad_norm": 26.74730110168457,
"learning_rate": 4.516778136213037e-06,
"log_odds_chosen": 1.4979491233825684,
"log_odds_ratio": -0.3211529850959778,
"logits/chosen": -2.7689156532287598,
"logits/rejected": -2.6690585613250732,
"logps/chosen": -0.3924681544303894,
"logps/rejected": -0.9140318632125854,
"loss": 19.3033,
"nll_loss": 0.594567060470581,
"rewards/accuracies": 0.875,
"rewards/chosen": -0.039246816188097,
"rewards/margins": 0.052156370133161545,
"rewards/rejected": -0.09140319377183914,
"step": 260
},
{
"epoch": 2.5025185185185186,
"grad_norm": 28.366756439208984,
"learning_rate": 3.7531701693965554e-06,
"log_odds_chosen": 1.323599100112915,
"log_odds_ratio": -0.37130218744277954,
"logits/chosen": -2.7441766262054443,
"logits/rejected": -2.7618346214294434,
"logps/chosen": -0.4224226474761963,
"logps/rejected": -0.9127100110054016,
"loss": 19.0019,
"nll_loss": 0.5947035551071167,
"rewards/accuracies": 0.8687499761581421,
"rewards/chosen": -0.04224226251244545,
"rewards/margins": 0.049028728157281876,
"rewards/rejected": -0.09127099066972733,
"step": 265
},
{
"epoch": 2.549925925925926,
"grad_norm": 27.086530685424805,
"learning_rate": 3.055003141378948e-06,
"log_odds_chosen": 1.311583399772644,
"log_odds_ratio": -0.3499976694583893,
"logits/chosen": -2.8330342769622803,
"logits/rejected": -2.7891921997070312,
"logps/chosen": -0.43048733472824097,
"logps/rejected": -0.9034746885299683,
"loss": 19.0972,
"nll_loss": 0.6235347986221313,
"rewards/accuracies": 0.8812500238418579,
"rewards/chosen": -0.043048735707998276,
"rewards/margins": 0.04729873687028885,
"rewards/rejected": -0.09034748375415802,
"step": 270
},
{
"epoch": 2.5973333333333333,
"grad_norm": 26.073270797729492,
"learning_rate": 2.424427429704365e-06,
"log_odds_chosen": 1.382707118988037,
"log_odds_ratio": -0.3385299742221832,
"logits/chosen": -2.8038322925567627,
"logits/rejected": -2.691887378692627,
"logps/chosen": -0.43798762559890747,
"logps/rejected": -0.9750789403915405,
"loss": 20.113,
"nll_loss": 0.6334654688835144,
"rewards/accuracies": 0.8687499761581421,
"rewards/chosen": -0.04379875957965851,
"rewards/margins": 0.05370912700891495,
"rewards/rejected": -0.09750787913799286,
"step": 275
},
{
"epoch": 2.644740740740741,
"grad_norm": 28.34626007080078,
"learning_rate": 1.8633852284264508e-06,
"log_odds_chosen": 1.2504308223724365,
"log_odds_ratio": -0.367613822221756,
"logits/chosen": -2.7782769203186035,
"logits/rejected": -2.7843925952911377,
"logps/chosen": -0.4142681062221527,
"logps/rejected": -0.8895042538642883,
"loss": 19.2085,
"nll_loss": 0.591576874256134,
"rewards/accuracies": 0.887499988079071,
"rewards/chosen": -0.04142681136727333,
"rewards/margins": 0.047523610293865204,
"rewards/rejected": -0.08895042538642883,
"step": 280
},
{
"epoch": 2.6921481481481484,
"grad_norm": 30.690038681030273,
"learning_rate": 1.3736045660864034e-06,
"log_odds_chosen": 1.2426550388336182,
"log_odds_ratio": -0.36959585547447205,
"logits/chosen": -2.8215603828430176,
"logits/rejected": -2.7462406158447266,
"logps/chosen": -0.4342527389526367,
"logps/rejected": -0.8976672887802124,
"loss": 19.7584,
"nll_loss": 0.6197787523269653,
"rewards/accuracies": 0.8374999761581421,
"rewards/chosen": -0.04342527687549591,
"rewards/margins": 0.04634145647287369,
"rewards/rejected": -0.0897667333483696,
"step": 285
},
{
"epoch": 2.7395555555555555,
"grad_norm": 28.228126525878906,
"learning_rate": 9.565939833279192e-07,
"log_odds_chosen": 1.3652734756469727,
"log_odds_ratio": -0.3423106074333191,
"logits/chosen": -2.7341599464416504,
"logits/rejected": -2.7509799003601074,
"logps/chosen": -0.38983267545700073,
"logps/rejected": -0.8893811106681824,
"loss": 19.743,
"nll_loss": 0.6161855459213257,
"rewards/accuracies": 0.8687499761581421,
"rewards/chosen": -0.038983266800642014,
"rewards/margins": 0.0499548502266407,
"rewards/rejected": -0.08893812447786331,
"step": 290
},
{
"epoch": 2.786962962962963,
"grad_norm": 27.338224411010742,
"learning_rate": 6.136378865420872e-07,
"log_odds_chosen": 1.2006913423538208,
"log_odds_ratio": -0.3960389494895935,
"logits/chosen": -2.686972141265869,
"logits/rejected": -2.7599165439605713,
"logps/chosen": -0.45548802614212036,
"logps/rejected": -0.915076732635498,
"loss": 19.6585,
"nll_loss": 0.6340048909187317,
"rewards/accuracies": 0.8500000238418579,
"rewards/chosen": -0.04554879665374756,
"rewards/margins": 0.04595886915922165,
"rewards/rejected": -0.09150765836238861,
"step": 295
},
{
"epoch": 2.83437037037037,
"grad_norm": 29.504423141479492,
"learning_rate": 3.45792591853214e-07,
"log_odds_chosen": 1.189578890800476,
"log_odds_ratio": -0.3797909617424011,
"logits/chosen": -2.7965073585510254,
"logits/rejected": -2.6686172485351562,
"logps/chosen": -0.4425443112850189,
"logps/rejected": -0.9239686131477356,
"loss": 19.6148,
"nll_loss": 0.6315287947654724,
"rewards/accuracies": 0.8687499761581421,
"rewards/chosen": -0.04425443336367607,
"rewards/margins": 0.04814242199063301,
"rewards/rejected": -0.09239685535430908,
"step": 300
},
{
"epoch": 2.8817777777777778,
"grad_norm": 26.830537796020508,
"learning_rate": 1.538830716302092e-07,
"log_odds_chosen": 1.2787503004074097,
"log_odds_ratio": -0.34538131952285767,
"logits/chosen": -2.859900951385498,
"logits/rejected": -2.7875828742980957,
"logps/chosen": -0.458304226398468,
"logps/rejected": -0.9439491033554077,
"loss": 19.4322,
"nll_loss": 0.6039472222328186,
"rewards/accuracies": 0.856249988079071,
"rewards/chosen": -0.04583042114973068,
"rewards/margins": 0.04856449365615845,
"rewards/rejected": -0.09439491480588913,
"step": 305
},
{
"epoch": 2.9291851851851853,
"grad_norm": 28.568008422851562,
"learning_rate": 3.8500413544415025e-08,
"log_odds_chosen": 1.291078805923462,
"log_odds_ratio": -0.36775314807891846,
"logits/chosen": -2.760559558868408,
"logits/rejected": -2.6593780517578125,
"logps/chosen": -0.41886311769485474,
"logps/rejected": -0.8871288299560547,
"loss": 19.394,
"nll_loss": 0.6148114204406738,
"rewards/accuracies": 0.8374999761581421,
"rewards/chosen": -0.041886307299137115,
"rewards/margins": 0.04682657867670059,
"rewards/rejected": -0.08871288597583771,
"step": 310
},
{
"epoch": 2.9765925925925925,
"grad_norm": 26.362131118774414,
"learning_rate": 0.0,
"log_odds_chosen": 1.38863205909729,
"log_odds_ratio": -0.3244430124759674,
"logits/chosen": -2.7366228103637695,
"logits/rejected": -2.6630187034606934,
"logps/chosen": -0.4075457453727722,
"logps/rejected": -0.9129207730293274,
"loss": 18.6997,
"nll_loss": 0.5973876118659973,
"rewards/accuracies": 0.90625,
"rewards/chosen": -0.04075457900762558,
"rewards/margins": 0.05053749680519104,
"rewards/rejected": -0.09129206836223602,
"step": 315
},
{
"epoch": 2.9765925925925925,
"step": 315,
"total_flos": 0.0,
"train_loss": 26.468634469168528,
"train_runtime": 6925.9574,
"train_samples_per_second": 2.924,
"train_steps_per_second": 0.045
}
],
"logging_steps": 5,
"max_steps": 315,
"num_input_tokens_seen": 0,
"num_train_epochs": 3,
"save_steps": 100000,
"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
}