silviasapora's picture
Model save
6b37744 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": 1199.2445068359375,
"learning_rate": 7.8125e-06,
"log_odds_chosen": -1.0599775314331055,
"log_odds_ratio": -11.002888679504395,
"logps/chosen": -22.7018985748291,
"logps/rejected": -21.64215660095215,
"loss": 482.2551,
"nll_loss": 9.837621688842773,
"po_loss": 5.232847690582275,
"rewards/accuracies": 0.4781250059604645,
"rewards/chosen": -11.35094928741455,
"rewards/margins": -0.5298711061477661,
"rewards/rejected": -10.821078300476074,
"step": 5
},
{
"epoch": 0.09481481481481481,
"grad_norm": 597.969970703125,
"learning_rate": 1.5625e-05,
"log_odds_chosen": -3.1238608360290527,
"log_odds_ratio": -11.037403106689453,
"logps/chosen": -21.978214263916016,
"logps/rejected": -18.854707717895508,
"loss": 448.5505,
"nll_loss": 8.742585182189941,
"po_loss": 5.274618148803711,
"rewards/accuracies": 0.4468750059604645,
"rewards/chosen": -10.989107131958008,
"rewards/margins": -1.5617557764053345,
"rewards/rejected": -9.427353858947754,
"step": 10
},
{
"epoch": 0.14222222222222222,
"grad_norm": 562.1139526367188,
"learning_rate": 2.34375e-05,
"log_odds_chosen": -1.3050577640533447,
"log_odds_ratio": -11.273977279663086,
"logps/chosen": -21.88943862915039,
"logps/rejected": -20.582889556884766,
"loss": 423.8629,
"nll_loss": 7.963566780090332,
"po_loss": 5.2821502685546875,
"rewards/accuracies": 0.515625,
"rewards/chosen": -10.944719314575195,
"rewards/margins": -0.6532740592956543,
"rewards/rejected": -10.291444778442383,
"step": 15
},
{
"epoch": 0.18962962962962962,
"grad_norm": 3476.161865234375,
"learning_rate": 3.125e-05,
"log_odds_chosen": -3.9486403465270996,
"log_odds_ratio": -8.879361152648926,
"logps/chosen": -16.218509674072266,
"logps/rejected": -12.272605895996094,
"loss": 321.7442,
"nll_loss": 5.964278221130371,
"po_loss": 4.09022855758667,
"rewards/accuracies": 0.4312500059604645,
"rewards/chosen": -8.109254837036133,
"rewards/margins": -1.972952127456665,
"rewards/rejected": -6.136302947998047,
"step": 20
},
{
"epoch": 0.23703703703703705,
"grad_norm": 253.1682586669922,
"learning_rate": 3.90625e-05,
"log_odds_chosen": -0.06654863059520721,
"log_odds_ratio": -1.5640277862548828,
"logps/chosen": -3.363724946975708,
"logps/rejected": -3.2796008586883545,
"loss": 89.1255,
"nll_loss": 2.5614538192749023,
"po_loss": 0.223719522356987,
"rewards/accuracies": 0.5531250238418579,
"rewards/chosen": -1.681862473487854,
"rewards/margins": -0.042061954736709595,
"rewards/rejected": -1.6398004293441772,
"step": 25
},
{
"epoch": 0.28444444444444444,
"grad_norm": 201.95008850097656,
"learning_rate": 4.6875e-05,
"log_odds_chosen": 0.09966959804296494,
"log_odds_ratio": -0.8472908735275269,
"logps/chosen": -1.7985613346099854,
"logps/rejected": -1.8834549188613892,
"loss": 62.7189,
"nll_loss": 2.001840591430664,
"po_loss": -0.04187442362308502,
"rewards/accuracies": 0.543749988079071,
"rewards/chosen": -0.8992806673049927,
"rewards/margins": 0.04244700074195862,
"rewards/rejected": -0.9417274594306946,
"step": 30
},
{
"epoch": 0.33185185185185184,
"grad_norm": 226.0881805419922,
"learning_rate": 4.998613757348784e-05,
"log_odds_chosen": 0.3407457172870636,
"log_odds_ratio": -0.7940031886100769,
"logps/chosen": -1.766862154006958,
"logps/rejected": -2.0710058212280273,
"loss": 56.0033,
"nll_loss": 1.9001868963241577,
"po_loss": -0.15008527040481567,
"rewards/accuracies": 0.550000011920929,
"rewards/chosen": -0.883431077003479,
"rewards/margins": 0.15207180380821228,
"rewards/rejected": -1.0355029106140137,
"step": 35
},
{
"epoch": 0.37925925925925924,
"grad_norm": 266.1627197265625,
"learning_rate": 4.990147841143462e-05,
"log_odds_chosen": 0.5402604341506958,
"log_odds_ratio": -0.7131898999214172,
"logps/chosen": -1.7148605585098267,
"logps/rejected": -2.1804628372192383,
"loss": 53.4684,
"nll_loss": 1.883787751197815,
"po_loss": -0.21290016174316406,
"rewards/accuracies": 0.6625000238418579,
"rewards/chosen": -0.8574302792549133,
"rewards/margins": 0.2328011691570282,
"rewards/rejected": -1.0902314186096191,
"step": 40
},
{
"epoch": 0.4266666666666667,
"grad_norm": 335.4718322753906,
"learning_rate": 4.97401218720448e-05,
"log_odds_chosen": 1.1245437860488892,
"log_odds_ratio": -0.6130434274673462,
"logps/chosen": -1.6581999063491821,
"logps/rejected": -2.69628643989563,
"loss": 45.3042,
"nll_loss": 1.8416335582733154,
"po_loss": -0.4258776307106018,
"rewards/accuracies": 0.715624988079071,
"rewards/chosen": -0.8290999531745911,
"rewards/margins": 0.5190432667732239,
"rewards/rejected": -1.348143219947815,
"step": 45
},
{
"epoch": 0.4740740740740741,
"grad_norm": 170.5546875,
"learning_rate": 4.9502564938797946e-05,
"log_odds_chosen": 0.9557003974914551,
"log_odds_ratio": -0.6759911775588989,
"logps/chosen": -1.6203289031982422,
"logps/rejected": -2.4938411712646484,
"loss": 47.41,
"nll_loss": 1.8312504291534424,
"po_loss": -0.34968656301498413,
"rewards/accuracies": 0.65625,
"rewards/chosen": -0.8101644515991211,
"rewards/margins": 0.4367561340332031,
"rewards/rejected": -1.2469205856323242,
"step": 50
},
{
"epoch": 0.5214814814814814,
"grad_norm": 225.52430725097656,
"learning_rate": 4.918953929490768e-05,
"log_odds_chosen": 1.1236488819122314,
"log_odds_ratio": -0.6741882562637329,
"logps/chosen": -1.6877021789550781,
"logps/rejected": -2.727043390274048,
"loss": 45.2223,
"nll_loss": 1.8263689279556274,
"po_loss": -0.41317272186279297,
"rewards/accuracies": 0.6781250238418579,
"rewards/chosen": -0.8438510894775391,
"rewards/margins": 0.5196704864501953,
"rewards/rejected": -1.363521695137024,
"step": 55
},
{
"epoch": 0.5688888888888889,
"grad_norm": 355.39068603515625,
"learning_rate": 4.88020090697132e-05,
"log_odds_chosen": 1.1692771911621094,
"log_odds_ratio": -0.6426931619644165,
"logps/chosen": -1.5684707164764404,
"logps/rejected": -2.65775728225708,
"loss": 44.6871,
"nll_loss": 1.8004539012908936,
"po_loss": -0.40398064255714417,
"rewards/accuracies": 0.653124988079071,
"rewards/chosen": -0.7842353582382202,
"rewards/margins": 0.544643223285675,
"rewards/rejected": -1.32887864112854,
"step": 60
},
{
"epoch": 0.6162962962962963,
"grad_norm": 162.62863159179688,
"learning_rate": 4.834116786912897e-05,
"log_odds_chosen": 1.382725715637207,
"log_odds_ratio": -0.705934464931488,
"logps/chosen": -1.7009910345077515,
"logps/rejected": -2.989063024520874,
"loss": 43.7142,
"nll_loss": 1.8565829992294312,
"po_loss": -0.49051398038864136,
"rewards/accuracies": 0.6937500238418579,
"rewards/chosen": -0.8504955172538757,
"rewards/margins": 0.6440361142158508,
"rewards/rejected": -1.494531512260437,
"step": 65
},
{
"epoch": 0.6637037037037037,
"grad_norm": 889.6281127929688,
"learning_rate": 4.7808435099299045e-05,
"log_odds_chosen": 3.606097459793091,
"log_odds_ratio": -0.7058836221694946,
"logps/chosen": -2.0993778705596924,
"logps/rejected": -5.576085090637207,
"loss": 33.5164,
"nll_loss": 2.1465699672698975,
"po_loss": -1.0991829633712769,
"rewards/accuracies": 0.7593749761581421,
"rewards/chosen": -1.0496889352798462,
"rewards/margins": 1.7383534908294678,
"rewards/rejected": -2.7880425453186035,
"step": 70
},
{
"epoch": 0.7111111111111111,
"grad_norm": 218.24794006347656,
"learning_rate": 4.720545159477922e-05,
"log_odds_chosen": 2.234952926635742,
"log_odds_ratio": -0.8444352149963379,
"logps/chosen": -1.784437894821167,
"logps/rejected": -3.926750659942627,
"loss": 35.6981,
"nll_loss": 1.8649476766586304,
"po_loss": -0.7493817806243896,
"rewards/accuracies": 0.690625011920929,
"rewards/chosen": -0.8922189474105835,
"rewards/margins": 1.071156620979309,
"rewards/rejected": -1.9633753299713135,
"step": 75
},
{
"epoch": 0.7585185185185185,
"grad_norm": 1788.55126953125,
"learning_rate": 4.653407456471222e-05,
"log_odds_chosen": 1.6183449029922485,
"log_odds_ratio": -0.846682071685791,
"logps/chosen": -1.728458046913147,
"logps/rejected": -3.26391339302063,
"loss": 42.0319,
"nll_loss": 1.8288921117782593,
"po_loss": -0.5153962969779968,
"rewards/accuracies": 0.621874988079071,
"rewards/chosen": -0.8642290234565735,
"rewards/margins": 0.7677275538444519,
"rewards/rejected": -1.631956696510315,
"step": 80
},
{
"epoch": 0.8059259259259259,
"grad_norm": 608.4232788085938,
"learning_rate": 4.579637187256222e-05,
"log_odds_chosen": 2.1726527214050293,
"log_odds_ratio": -0.8868170976638794,
"logps/chosen": -2.386735200881958,
"logps/rejected": -4.454265594482422,
"loss": 40.3963,
"nll_loss": 2.1039323806762695,
"po_loss": -0.8415490388870239,
"rewards/accuracies": 0.671875,
"rewards/chosen": -1.193367600440979,
"rewards/margins": 1.0337653160095215,
"rewards/rejected": -2.227132797241211,
"step": 85
},
{
"epoch": 0.8533333333333334,
"grad_norm": 2436.2255859375,
"learning_rate": 4.499461566702685e-05,
"log_odds_chosen": 3.146548271179199,
"log_odds_ratio": -0.9941431283950806,
"logps/chosen": -2.020655393600464,
"logps/rejected": -5.0438432693481445,
"loss": 36.2214,
"nll_loss": 2.022157669067383,
"po_loss": -0.890238881111145,
"rewards/accuracies": 0.715624988079071,
"rewards/chosen": -1.010327696800232,
"rewards/margins": 1.5115936994552612,
"rewards/rejected": -2.5219216346740723,
"step": 90
},
{
"epoch": 0.9007407407407407,
"grad_norm": 1364.458251953125,
"learning_rate": 4.413127538374411e-05,
"log_odds_chosen": 2.372877597808838,
"log_odds_ratio": -0.8819985389709473,
"logps/chosen": -1.8147878646850586,
"logps/rejected": -4.103789806365967,
"loss": 35.7507,
"nll_loss": 1.7868406772613525,
"po_loss": -0.6696311235427856,
"rewards/accuracies": 0.640625,
"rewards/chosen": -0.9073939323425293,
"rewards/margins": 1.144500970840454,
"rewards/rejected": -2.0518949031829834,
"step": 95
},
{
"epoch": 0.9481481481481482,
"grad_norm": 1274.0474853515625,
"learning_rate": 4.320901013934887e-05,
"log_odds_chosen": 2.14143705368042,
"log_odds_ratio": -0.786421000957489,
"logps/chosen": -1.7358782291412354,
"logps/rejected": -3.7556350231170654,
"loss": 33.5542,
"nll_loss": 1.854790449142456,
"po_loss": -0.8062206506729126,
"rewards/accuracies": 0.6968749761581421,
"rewards/chosen": -0.8679391145706177,
"rewards/margins": 1.0098785161972046,
"rewards/rejected": -1.8778175115585327,
"step": 100
},
{
"epoch": 0.9955555555555555,
"grad_norm": 568.3045043945312,
"learning_rate": 4.223066054130568e-05,
"log_odds_chosen": 2.152684450149536,
"log_odds_ratio": -0.7120019197463989,
"logps/chosen": -1.5902575254440308,
"logps/rejected": -3.6419010162353516,
"loss": 31.711,
"nll_loss": 1.7339718341827393,
"po_loss": -0.7430030107498169,
"rewards/accuracies": 0.6812499761581421,
"rewards/chosen": -0.7951287627220154,
"rewards/margins": 1.0258219242095947,
"rewards/rejected": -1.8209505081176758,
"step": 105
},
{
"epoch": 1.037925925925926,
"grad_norm": 628.1900634765625,
"learning_rate": 4.1199239938743797e-05,
"log_odds_chosen": 3.4075300693511963,
"log_odds_ratio": -0.5813586115837097,
"logps/chosen": -1.8704923391342163,
"logps/rejected": -5.093435287475586,
"loss": 16.708,
"nll_loss": 1.6286418437957764,
"po_loss": -1.0444445610046387,
"rewards/accuracies": 0.7552447319030762,
"rewards/chosen": -0.9352461695671082,
"rewards/margins": 1.6114716529846191,
"rewards/rejected": -2.546717643737793,
"step": 110
},
{
"epoch": 1.0853333333333333,
"grad_norm": 1226.1982421875,
"learning_rate": 4.0117925141242174e-05,
"log_odds_chosen": 3.840623378753662,
"log_odds_ratio": -0.8993012309074402,
"logps/chosen": -1.9903583526611328,
"logps/rejected": -5.609218597412109,
"loss": 14.4574,
"nll_loss": 1.720370888710022,
"po_loss": -1.2685784101486206,
"rewards/accuracies": 0.7749999761581421,
"rewards/chosen": -0.9951791763305664,
"rewards/margins": 1.8094301223754883,
"rewards/rejected": -2.8046092987060547,
"step": 115
},
{
"epoch": 1.1327407407407408,
"grad_norm": 1661.664794921875,
"learning_rate": 3.899004663415084e-05,
"log_odds_chosen": 3.6831917762756348,
"log_odds_ratio": -0.5217957496643066,
"logps/chosen": -1.4346517324447632,
"logps/rejected": -4.917321681976318,
"loss": 14.9163,
"nll_loss": 1.5880098342895508,
"po_loss": -1.1218748092651367,
"rewards/accuracies": 0.7593749761581421,
"rewards/chosen": -0.7173258662223816,
"rewards/margins": 1.7413349151611328,
"rewards/rejected": -2.458660840988159,
"step": 120
},
{
"epoch": 1.1801481481481482,
"grad_norm": 571.31982421875,
"learning_rate": 3.781907832058587e-05,
"log_odds_chosen": 6.352164268493652,
"log_odds_ratio": -0.6368833780288696,
"logps/chosen": -1.6029341220855713,
"logps/rejected": -7.754278659820557,
"loss": 11.4892,
"nll_loss": 1.7506815195083618,
"po_loss": -1.3916442394256592,
"rewards/accuracies": 0.765625,
"rewards/chosen": -0.8014670610427856,
"rewards/margins": 3.075672149658203,
"rewards/rejected": -3.8771393299102783,
"step": 125
},
{
"epoch": 1.2275555555555555,
"grad_norm": 831.4838256835938,
"learning_rate": 3.660862682169282e-05,
"log_odds_chosen": 5.919846534729004,
"log_odds_ratio": -0.5129478573799133,
"logps/chosen": -1.4314855337142944,
"logps/rejected": -7.115793704986572,
"loss": 9.806,
"nll_loss": 1.6523067951202393,
"po_loss": -1.3458693027496338,
"rewards/accuracies": 0.7875000238418579,
"rewards/chosen": -0.7157427668571472,
"rewards/margins": 2.842153549194336,
"rewards/rejected": -3.557896852493286,
"step": 130
},
{
"epoch": 1.274962962962963,
"grad_norm": 586.3059692382812,
"learning_rate": 3.5362420368134356e-05,
"log_odds_chosen": 5.599277019500732,
"log_odds_ratio": -0.624258279800415,
"logps/chosen": -1.7905899286270142,
"logps/rejected": -7.161185264587402,
"loss": 12.6168,
"nll_loss": 1.760801076889038,
"po_loss": -1.3665275573730469,
"rewards/accuracies": 0.778124988079071,
"rewards/chosen": -0.8952949643135071,
"rewards/margins": 2.6852974891662598,
"rewards/rejected": -3.580592632293701,
"step": 135
},
{
"epoch": 1.3223703703703704,
"grad_norm": 880.5230102539062,
"learning_rate": 3.408429731701635e-05,
"log_odds_chosen": 3.740973711013794,
"log_odds_ratio": -0.5780373215675354,
"logps/chosen": -1.4464441537857056,
"logps/rejected": -4.967782020568848,
"loss": 15.2271,
"nll_loss": 1.6292192935943604,
"po_loss": -1.1533715724945068,
"rewards/accuracies": 0.762499988079071,
"rewards/chosen": -0.7232220768928528,
"rewards/margins": 1.7606691122055054,
"rewards/rejected": -2.483891010284424,
"step": 140
},
{
"epoch": 1.3697777777777778,
"grad_norm": 663.084228515625,
"learning_rate": 3.2778194329621104e-05,
"log_odds_chosen": 6.911275386810303,
"log_odds_ratio": -0.8403680920600891,
"logps/chosen": -1.9615440368652344,
"logps/rejected": -8.633264541625977,
"loss": 8.7443,
"nll_loss": 1.8974125385284424,
"po_loss": -1.624152421951294,
"rewards/accuracies": 0.796875,
"rewards/chosen": -0.9807720184326172,
"rewards/margins": 3.335860013961792,
"rewards/rejected": -4.316632270812988,
"step": 145
},
{
"epoch": 1.417185185185185,
"grad_norm": 902.1053466796875,
"learning_rate": 3.144813424636031e-05,
"log_odds_chosen": 7.530759334564209,
"log_odds_ratio": -0.5000367164611816,
"logps/chosen": -1.5284488201141357,
"logps/rejected": -8.809328079223633,
"loss": -5.3961,
"nll_loss": 1.633239984512329,
"po_loss": -1.8018696308135986,
"rewards/accuracies": 0.809374988079071,
"rewards/chosen": -0.7642244100570679,
"rewards/margins": 3.640439510345459,
"rewards/rejected": -4.404664039611816,
"step": 150
},
{
"epoch": 1.4645925925925927,
"grad_norm": 641.8519897460938,
"learning_rate": 3.0098213696293542e-05,
"log_odds_chosen": 6.644770622253418,
"log_odds_ratio": -1.1336581707000732,
"logps/chosen": -2.2592227458953857,
"logps/rejected": -8.666966438293457,
"loss": 19.4596,
"nll_loss": 2.1993162631988525,
"po_loss": -1.591204047203064,
"rewards/accuracies": 0.765625,
"rewards/chosen": -1.1296113729476929,
"rewards/margins": 3.203871488571167,
"rewards/rejected": -4.3334832191467285,
"step": 155
},
{
"epoch": 1.512,
"grad_norm": 991.56591796875,
"learning_rate": 2.8732590479375165e-05,
"log_odds_chosen": 7.27960205078125,
"log_odds_ratio": -0.6811318397521973,
"logps/chosen": -1.9771201610565186,
"logps/rejected": -8.984837532043457,
"loss": -4.91,
"nll_loss": 1.8900007009506226,
"po_loss": -2.043437957763672,
"rewards/accuracies": 0.768750011920929,
"rewards/chosen": -0.9885600805282593,
"rewards/margins": 3.5038580894470215,
"rewards/rejected": -4.4924187660217285,
"step": 160
},
{
"epoch": 1.5594074074074074,
"grad_norm": 593.0498657226562,
"learning_rate": 2.7355470760292956e-05,
"log_odds_chosen": 6.110047340393066,
"log_odds_ratio": -0.8749097585678101,
"logps/chosen": -2.053755283355713,
"logps/rejected": -7.891977787017822,
"loss": 7.4888,
"nll_loss": 2.0199038982391357,
"po_loss": -1.7858794927597046,
"rewards/accuracies": 0.8062499761581421,
"rewards/chosen": -1.0268776416778564,
"rewards/margins": 2.9191110134124756,
"rewards/rejected": -3.945988893508911,
"step": 165
},
{
"epoch": 1.6068148148148147,
"grad_norm": 883.1884765625,
"learning_rate": 2.597109611334169e-05,
"log_odds_chosen": 7.358190059661865,
"log_odds_ratio": -0.6941929459571838,
"logps/chosen": -1.9749581813812256,
"logps/rejected": -9.061105728149414,
"loss": -1.458,
"nll_loss": 1.907533049583435,
"po_loss": -1.9530951976776123,
"rewards/accuracies": 0.8125,
"rewards/chosen": -0.9874790906906128,
"rewards/margins": 3.5430731773376465,
"rewards/rejected": -4.530552864074707,
"step": 170
},
{
"epoch": 1.6542222222222223,
"grad_norm": 776.2947387695312,
"learning_rate": 2.458373045823404e-05,
"log_odds_chosen": 7.158226013183594,
"log_odds_ratio": -0.9282143712043762,
"logps/chosen": -2.6594738960266113,
"logps/rejected": -9.560735702514648,
"loss": 7.5769,
"nll_loss": 2.254758596420288,
"po_loss": -2.0179810523986816,
"rewards/accuracies": 0.784375011920929,
"rewards/chosen": -1.3297369480133057,
"rewards/margins": 3.4506313800811768,
"rewards/rejected": -4.780367851257324,
"step": 175
},
{
"epoch": 1.7016296296296296,
"grad_norm": 2007.3345947265625,
"learning_rate": 2.3197646927086697e-05,
"log_odds_chosen": 7.059978485107422,
"log_odds_ratio": -0.91358882188797,
"logps/chosen": -2.309866428375244,
"logps/rejected": -9.11004638671875,
"loss": -8.0333,
"nll_loss": 1.9352056980133057,
"po_loss": -2.1862475872039795,
"rewards/accuracies": 0.7718750238418579,
"rewards/chosen": -1.154933214187622,
"rewards/margins": 3.400089979171753,
"rewards/rejected": -4.555023193359375,
"step": 180
},
{
"epoch": 1.749037037037037,
"grad_norm": 846.7615356445312,
"learning_rate": 2.1817114703032176e-05,
"log_odds_chosen": 7.948319435119629,
"log_odds_ratio": -1.0959177017211914,
"logps/chosen": -2.901278495788574,
"logps/rejected": -10.634871482849121,
"loss": -11.2961,
"nll_loss": 2.0984649658203125,
"po_loss": -2.451467514038086,
"rewards/accuracies": 0.7718750238418579,
"rewards/chosen": -1.450639247894287,
"rewards/margins": 3.8667964935302734,
"rewards/rejected": -5.3174357414245605,
"step": 185
},
{
"epoch": 1.7964444444444445,
"grad_norm": 817.7583618164062,
"learning_rate": 2.0446385870993467e-05,
"log_odds_chosen": 6.724404335021973,
"log_odds_ratio": -0.5374380350112915,
"logps/chosen": -1.7088407278060913,
"logps/rejected": -8.153546333312988,
"loss": -27.5871,
"nll_loss": 1.6332111358642578,
"po_loss": -2.495307207107544,
"rewards/accuracies": 0.796875,
"rewards/chosen": -0.8544203639030457,
"rewards/margins": 3.2223522663116455,
"rewards/rejected": -4.076773166656494,
"step": 190
},
{
"epoch": 1.8438518518518519,
"grad_norm": 2679.26513671875,
"learning_rate": 1.9089682321121834e-05,
"log_odds_chosen": 9.504817008972168,
"log_odds_ratio": -1.4054466485977173,
"logps/chosen": -4.4236674308776855,
"logps/rejected": -13.693811416625977,
"loss": 20.7869,
"nll_loss": 3.1105668544769287,
"po_loss": -2.460975170135498,
"rewards/accuracies": 0.78125,
"rewards/chosen": -2.2118337154388428,
"rewards/margins": 4.635071754455566,
"rewards/rejected": -6.846905708312988,
"step": 195
},
{
"epoch": 1.8912592592592592,
"grad_norm": 896.3246459960938,
"learning_rate": 1.775118274523545e-05,
"log_odds_chosen": 6.952999114990234,
"log_odds_ratio": -1.0420355796813965,
"logps/chosen": -2.362539768218994,
"logps/rejected": -9.074200630187988,
"loss": -3.5533,
"nll_loss": 2.0847651958465576,
"po_loss": -2.1958060264587402,
"rewards/accuracies": 0.753125011920929,
"rewards/chosen": -1.181269884109497,
"rewards/margins": 3.355830430984497,
"rewards/rejected": -4.537100315093994,
"step": 200
},
{
"epoch": 1.9386666666666668,
"grad_norm": 602.3760986328125,
"learning_rate": 1.643500976631037e-05,
"log_odds_chosen": 4.505127429962158,
"log_odds_ratio": -0.7301878333091736,
"logps/chosen": -1.551640272140503,
"logps/rejected": -5.79341983795166,
"loss": 2.1605,
"nll_loss": 1.6449111700057983,
"po_loss": -1.5773966312408447,
"rewards/accuracies": 0.78125,
"rewards/chosen": -0.7758201360702515,
"rewards/margins": 2.120889663696289,
"rewards/rejected": -2.89670991897583,
"step": 205
},
{
"epoch": 1.986074074074074,
"grad_norm": 831.9571533203125,
"learning_rate": 1.514521724066537e-05,
"log_odds_chosen": 4.619518280029297,
"log_odds_ratio": -0.8430620431900024,
"logps/chosen": -1.8811943531036377,
"logps/rejected": -6.274863243103027,
"loss": 7.1653,
"nll_loss": 1.774757742881775,
"po_loss": -1.5508434772491455,
"rewards/accuracies": 0.746874988079071,
"rewards/chosen": -0.9405971765518188,
"rewards/margins": 2.1968345642089844,
"rewards/rejected": -3.1374316215515137,
"step": 210
},
{
"epoch": 2.0284444444444443,
"grad_norm": 1437.5135498046875,
"learning_rate": 1.3885777771950348e-05,
"log_odds_chosen": 6.645725727081299,
"log_odds_ratio": -0.6343217492103577,
"logps/chosen": -1.605512022972107,
"logps/rejected": -7.886458396911621,
"loss": -23.0718,
"nll_loss": 1.6514700651168823,
"po_loss": -2.4581761360168457,
"rewards/accuracies": 0.8216783404350281,
"rewards/chosen": -0.8027560114860535,
"rewards/margins": 3.1404736042022705,
"rewards/rejected": -3.9432291984558105,
"step": 215
},
{
"epoch": 2.075851851851852,
"grad_norm": 878.4225463867188,
"learning_rate": 1.2660570475395683e-05,
"log_odds_chosen": 7.114099025726318,
"log_odds_ratio": -0.7919400930404663,
"logps/chosen": -1.9719266891479492,
"logps/rejected": -8.785611152648926,
"loss": -21.0157,
"nll_loss": 1.8605375289916992,
"po_loss": -2.517279624938965,
"rewards/accuracies": 0.8031250238418579,
"rewards/chosen": -0.9859633445739746,
"rewards/margins": 3.4068424701690674,
"rewards/rejected": -4.392805576324463,
"step": 220
},
{
"epoch": 2.1232592592592594,
"grad_norm": 1384.3018798828125,
"learning_rate": 1.1473369030008974e-05,
"log_odds_chosen": 9.657078742980957,
"log_odds_ratio": -0.717284083366394,
"logps/chosen": -2.1501874923706055,
"logps/rejected": -11.44202995300293,
"loss": -49.1628,
"nll_loss": 1.9243361949920654,
"po_loss": -3.460672378540039,
"rewards/accuracies": 0.8500000238418579,
"rewards/chosen": -1.0750937461853027,
"rewards/margins": 4.645921230316162,
"rewards/rejected": -5.721014976501465,
"step": 225
},
{
"epoch": 2.1706666666666665,
"grad_norm": 1059.046142578125,
"learning_rate": 1.0327830055518842e-05,
"log_odds_chosen": 9.652680397033691,
"log_odds_ratio": -0.5430900454521179,
"logps/chosen": -1.8017524480819702,
"logps/rejected": -11.105250358581543,
"loss": -54.9112,
"nll_loss": 1.7806711196899414,
"po_loss": -3.496647596359253,
"rewards/accuracies": 0.8343750238418579,
"rewards/chosen": -0.9008762240409851,
"rewards/margins": 4.651749610900879,
"rewards/rejected": -5.5526251792907715,
"step": 230
},
{
"epoch": 2.218074074074074,
"grad_norm": 934.3763427734375,
"learning_rate": 9.227481849865235e-06,
"log_odds_chosen": 8.37989616394043,
"log_odds_ratio": -0.548344075679779,
"logps/chosen": -1.6562755107879639,
"logps/rejected": -9.679137229919434,
"loss": -44.0365,
"nll_loss": 1.69879949092865,
"po_loss": -3.074941635131836,
"rewards/accuracies": 0.831250011920929,
"rewards/chosen": -0.8281377553939819,
"rewards/margins": 4.011431694030762,
"rewards/rejected": -4.839568614959717,
"step": 235
},
{
"epoch": 2.2654814814814817,
"grad_norm": 1470.9927978515625,
"learning_rate": 8.175713521924978e-06,
"log_odds_chosen": 9.9420166015625,
"log_odds_ratio": -0.5666526556015015,
"logps/chosen": -1.586470365524292,
"logps/rejected": -11.144021987915039,
"loss": -54.6849,
"nll_loss": 1.6270078420639038,
"po_loss": -3.335911989212036,
"rewards/accuracies": 0.8812500238418579,
"rewards/chosen": -0.793235182762146,
"rewards/margins": 4.778776168823242,
"rewards/rejected": -5.5720109939575195,
"step": 240
},
{
"epoch": 2.3128888888888888,
"grad_norm": 2696.906982421875,
"learning_rate": 7.1757645529443665e-06,
"log_odds_chosen": 11.02066421508789,
"log_odds_ratio": -0.523024320602417,
"logps/chosen": -1.5870704650878906,
"logps/rejected": -12.236452102661133,
"loss": -57.4956,
"nll_loss": 1.6896440982818604,
"po_loss": -3.4863803386688232,
"rewards/accuracies": 0.862500011920929,
"rewards/chosen": -0.7935352325439453,
"rewards/margins": 5.324690341949463,
"rewards/rejected": -6.118226051330566,
"step": 245
},
{
"epoch": 2.3602962962962963,
"grad_norm": 2218.82373046875,
"learning_rate": 6.230714818829733e-06,
"log_odds_chosen": 12.59239673614502,
"log_odds_ratio": -0.5458945035934448,
"logps/chosen": -2.0517947673797607,
"logps/rejected": -14.282119750976562,
"loss": -61.3005,
"nll_loss": 1.7380138635635376,
"po_loss": -3.6536529064178467,
"rewards/accuracies": 0.824999988079071,
"rewards/chosen": -1.0258973836898804,
"rewards/margins": 6.115162372589111,
"rewards/rejected": -7.141059875488281,
"step": 250
},
{
"epoch": 2.407703703703704,
"grad_norm": 3879.0703125,
"learning_rate": 5.343475104027743e-06,
"log_odds_chosen": 13.796175956726074,
"log_odds_ratio": -0.7049432992935181,
"logps/chosen": -2.48822283744812,
"logps/rejected": -15.938923835754395,
"loss": -54.8681,
"nll_loss": 2.094301700592041,
"po_loss": -3.8089308738708496,
"rewards/accuracies": 0.878125011920929,
"rewards/chosen": -1.24411141872406,
"rewards/margins": 6.725350379943848,
"rewards/rejected": -7.969461917877197,
"step": 255
},
{
"epoch": 2.455111111111111,
"grad_norm": 2860.3720703125,
"learning_rate": 4.516778136213037e-06,
"log_odds_chosen": 12.339117050170898,
"log_odds_ratio": -0.8312928080558777,
"logps/chosen": -2.22261381149292,
"logps/rejected": -14.213083267211914,
"loss": -51.4931,
"nll_loss": 2.0115973949432373,
"po_loss": -3.6207566261291504,
"rewards/accuracies": 0.859375,
"rewards/chosen": -1.11130690574646,
"rewards/margins": 5.995234489440918,
"rewards/rejected": -7.106541633605957,
"step": 260
},
{
"epoch": 2.5025185185185186,
"grad_norm": 1744.2803955078125,
"learning_rate": 3.7531701693965554e-06,
"log_odds_chosen": 10.812795639038086,
"log_odds_ratio": -0.5689637064933777,
"logps/chosen": -1.7130670547485352,
"logps/rejected": -12.172760009765625,
"loss": -62.0607,
"nll_loss": 1.6393451690673828,
"po_loss": -3.5787429809570312,
"rewards/accuracies": 0.840624988079071,
"rewards/chosen": -0.8565335273742676,
"rewards/margins": 5.229846000671387,
"rewards/rejected": -6.0863800048828125,
"step": 265
},
{
"epoch": 2.549925925925926,
"grad_norm": 1753.0069580078125,
"learning_rate": 3.055003141378948e-06,
"log_odds_chosen": 11.131937026977539,
"log_odds_ratio": -0.6443347930908203,
"logps/chosen": -1.8285239934921265,
"logps/rejected": -12.587448120117188,
"loss": -64.4668,
"nll_loss": 1.797912836074829,
"po_loss": -3.8124988079071045,
"rewards/accuracies": 0.8531249761581421,
"rewards/chosen": -0.9142619967460632,
"rewards/margins": 5.379462242126465,
"rewards/rejected": -6.293724060058594,
"step": 270
},
{
"epoch": 2.5973333333333333,
"grad_norm": 2948.94775390625,
"learning_rate": 2.424427429704365e-06,
"log_odds_chosen": 10.50131893157959,
"log_odds_ratio": -0.5303918123245239,
"logps/chosen": -1.9819530248641968,
"logps/rejected": -12.131158828735352,
"loss": -54.0842,
"nll_loss": 1.8271331787109375,
"po_loss": -3.5172641277313232,
"rewards/accuracies": 0.878125011920929,
"rewards/chosen": -0.9909765124320984,
"rewards/margins": 5.0746026039123535,
"rewards/rejected": -6.065579414367676,
"step": 275
},
{
"epoch": 2.644740740740741,
"grad_norm": 2592.339111328125,
"learning_rate": 1.8633852284264508e-06,
"log_odds_chosen": 12.00327205657959,
"log_odds_ratio": -0.676723837852478,
"logps/chosen": -1.9051532745361328,
"logps/rejected": -13.546533584594727,
"loss": -61.6846,
"nll_loss": 1.8986514806747437,
"po_loss": -3.826294422149658,
"rewards/accuracies": 0.824999988079071,
"rewards/chosen": -0.9525766372680664,
"rewards/margins": 5.820690155029297,
"rewards/rejected": -6.773266792297363,
"step": 280
},
{
"epoch": 2.6921481481481484,
"grad_norm": 2699.58544921875,
"learning_rate": 1.3736045660864034e-06,
"log_odds_chosen": 11.029436111450195,
"log_odds_ratio": -0.5692117810249329,
"logps/chosen": -1.7028005123138428,
"logps/rejected": -12.390884399414062,
"loss": -66.3219,
"nll_loss": 1.6979774236679077,
"po_loss": -3.770538330078125,
"rewards/accuracies": 0.859375,
"rewards/chosen": -0.8514002561569214,
"rewards/margins": 5.344042778015137,
"rewards/rejected": -6.195442199707031,
"step": 285
},
{
"epoch": 2.7395555555555555,
"grad_norm": 2166.333984375,
"learning_rate": 9.565939833279192e-07,
"log_odds_chosen": 10.880716323852539,
"log_odds_ratio": -0.7237260341644287,
"logps/chosen": -2.316985607147217,
"logps/rejected": -12.853070259094238,
"loss": -52.5932,
"nll_loss": 1.9568628072738647,
"po_loss": -3.60040020942688,
"rewards/accuracies": 0.8374999761581421,
"rewards/chosen": -1.1584928035736084,
"rewards/margins": 5.26804256439209,
"rewards/rejected": -6.426535129547119,
"step": 290
},
{
"epoch": 2.786962962962963,
"grad_norm": 2485.855712890625,
"learning_rate": 6.136378865420872e-07,
"log_odds_chosen": 10.605603218078613,
"log_odds_ratio": -0.8477842211723328,
"logps/chosen": -2.2415826320648193,
"logps/rejected": -12.530155181884766,
"loss": -47.1903,
"nll_loss": 1.9998095035552979,
"po_loss": -3.474505662918091,
"rewards/accuracies": 0.7906249761581421,
"rewards/chosen": -1.1207913160324097,
"rewards/margins": 5.144285678863525,
"rewards/rejected": -6.265077590942383,
"step": 295
},
{
"epoch": 2.83437037037037,
"grad_norm": 2430.19970703125,
"learning_rate": 3.45792591853214e-07,
"log_odds_chosen": 11.173027992248535,
"log_odds_ratio": -0.5032259821891785,
"logps/chosen": -1.4856630563735962,
"logps/rejected": -12.28561019897461,
"loss": -71.7924,
"nll_loss": 1.597579002380371,
"po_loss": -3.8410916328430176,
"rewards/accuracies": 0.828125,
"rewards/chosen": -0.7428315281867981,
"rewards/margins": 5.399973392486572,
"rewards/rejected": -6.142805099487305,
"step": 300
},
{
"epoch": 2.8817777777777778,
"grad_norm": 1150.9241943359375,
"learning_rate": 1.538830716302092e-07,
"log_odds_chosen": 11.992634773254395,
"log_odds_ratio": -0.6470831632614136,
"logps/chosen": -1.8577916622161865,
"logps/rejected": -13.447807312011719,
"loss": -69.0062,
"nll_loss": 1.8080154657363892,
"po_loss": -3.9644579887390137,
"rewards/accuracies": 0.828125,
"rewards/chosen": -0.9288958311080933,
"rewards/margins": 5.795008182525635,
"rewards/rejected": -6.723903656005859,
"step": 305
},
{
"epoch": 2.9291851851851853,
"grad_norm": 1115.533447265625,
"learning_rate": 3.8500413544415025e-08,
"log_odds_chosen": 11.34982967376709,
"log_odds_ratio": -0.5919870734214783,
"logps/chosen": -1.6738865375518799,
"logps/rejected": -12.641853332519531,
"loss": -65.5454,
"nll_loss": 1.8010927438735962,
"po_loss": -3.849386692047119,
"rewards/accuracies": 0.8656250238418579,
"rewards/chosen": -0.8369432687759399,
"rewards/margins": 5.483983993530273,
"rewards/rejected": -6.320926666259766,
"step": 310
},
{
"epoch": 2.9765925925925925,
"grad_norm": 2356.891845703125,
"learning_rate": 0.0,
"log_odds_chosen": 11.529512405395508,
"log_odds_ratio": -0.520513653755188,
"logps/chosen": -1.613694429397583,
"logps/rejected": -12.730021476745605,
"loss": -73.7153,
"nll_loss": 1.6499922275543213,
"po_loss": -3.9535961151123047,
"rewards/accuracies": 0.840624988079071,
"rewards/chosen": -0.8068472146987915,
"rewards/margins": 5.558164119720459,
"rewards/rejected": -6.365010738372803,
"step": 315
},
{
"epoch": 2.9765925925925925,
"step": 315,
"total_flos": 0.0,
"train_loss": 22.20340297638424,
"train_runtime": 9567.4219,
"train_samples_per_second": 2.117,
"train_steps_per_second": 0.033
}
],
"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
}