just1nseo's picture
Model save
15f0f76 verified
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 1.0,
"eval_steps": 100,
"global_step": 684,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"dpo_losses": 0.6931471824645996,
"epoch": 0.0,
"grad_norm": 1.9721656718078973,
"learning_rate": 7.246376811594203e-09,
"logits/chosen": -2.8746490478515625,
"logits/rejected": -2.840811252593994,
"logps/chosen": -227.73272705078125,
"logps/rejected": -174.71890258789062,
"loss": 0.6931,
"positive_losses": 0.0,
"rewards/accuracies": 0.0,
"rewards/chosen": 0.0,
"rewards/margins": 0.0,
"rewards/margins_max": 0.0,
"rewards/margins_min": 0.0,
"rewards/margins_std": 0.0,
"rewards/rejected": 0.0,
"step": 1
},
{
"dpo_losses": 0.6933290958404541,
"epoch": 0.01,
"grad_norm": 25.64726309944567,
"learning_rate": 7.246376811594203e-08,
"logits/chosen": -2.7698583602905273,
"logits/rejected": -2.726212501525879,
"logps/chosen": -244.640380859375,
"logps/rejected": -240.38246154785156,
"loss": 0.6981,
"positive_losses": 0.04526931792497635,
"rewards/accuracies": 0.2638888955116272,
"rewards/chosen": -7.686785829719156e-05,
"rewards/margins": -0.0003627391706686467,
"rewards/margins_max": 0.0012993874261155725,
"rewards/margins_min": -0.0022027057129889727,
"rewards/margins_std": 0.0015639414777979255,
"rewards/rejected": 0.00028587126871570945,
"step": 10
},
{
"dpo_losses": 0.6930510401725769,
"epoch": 0.03,
"grad_norm": 20.592262133277757,
"learning_rate": 1.4492753623188405e-07,
"logits/chosen": -2.897423028945923,
"logits/rejected": -2.8587498664855957,
"logps/chosen": -317.21636962890625,
"logps/rejected": -239.84121704101562,
"loss": 0.6999,
"positive_losses": 0.07797051966190338,
"rewards/accuracies": 0.5249999761581421,
"rewards/chosen": 0.0003467574715614319,
"rewards/margins": 0.00019531897851265967,
"rewards/margins_max": 0.0034703321289271116,
"rewards/margins_min": -0.0027894906233996153,
"rewards/margins_std": 0.002780328970402479,
"rewards/rejected": 0.000151438478496857,
"step": 20
},
{
"dpo_losses": 0.6931239366531372,
"epoch": 0.04,
"grad_norm": 16.747192147468148,
"learning_rate": 2.1739130434782607e-07,
"logits/chosen": -2.811279535293579,
"logits/rejected": -2.790393114089966,
"logps/chosen": -264.90753173828125,
"logps/rejected": -223.55331420898438,
"loss": 0.6988,
"positive_losses": 0.052257489413022995,
"rewards/accuracies": 0.4625000059604645,
"rewards/chosen": 0.0009221597574651241,
"rewards/margins": 4.864180664299056e-05,
"rewards/margins_max": 0.002964555751532316,
"rewards/margins_min": -0.0025322353467345238,
"rewards/margins_std": 0.0024663216900080442,
"rewards/rejected": 0.0008735180017538369,
"step": 30
},
{
"dpo_losses": 0.6930278539657593,
"epoch": 0.06,
"grad_norm": 12.398083667958339,
"learning_rate": 2.898550724637681e-07,
"logits/chosen": -2.864264965057373,
"logits/rejected": -2.834386110305786,
"logps/chosen": -304.97265625,
"logps/rejected": -266.8782043457031,
"loss": 0.6961,
"positive_losses": 0.031120896339416504,
"rewards/accuracies": 0.4625000059604645,
"rewards/chosen": 0.0022245876025408506,
"rewards/margins": 0.0002419196825940162,
"rewards/margins_max": 0.004220693372189999,
"rewards/margins_min": -0.0033918179105967283,
"rewards/margins_std": 0.0034401887096464634,
"rewards/rejected": 0.001982667949050665,
"step": 40
},
{
"dpo_losses": 0.6931858658790588,
"epoch": 0.07,
"grad_norm": 8.878267358365656,
"learning_rate": 3.6231884057971015e-07,
"logits/chosen": -2.870177984237671,
"logits/rejected": -2.8384270668029785,
"logps/chosen": -260.7277526855469,
"logps/rejected": -234.89218139648438,
"loss": 0.6942,
"positive_losses": 0.022727489471435547,
"rewards/accuracies": 0.4124999940395355,
"rewards/chosen": 0.003678619861602783,
"rewards/margins": -7.445671508321539e-05,
"rewards/margins_max": 0.0031701656989753246,
"rewards/margins_min": -0.0034985982347279787,
"rewards/margins_std": 0.0029860776849091053,
"rewards/rejected": 0.003753076074644923,
"step": 50
},
{
"dpo_losses": 0.6925040483474731,
"epoch": 0.09,
"grad_norm": 1.6719366044432857,
"learning_rate": 4.3478260869565214e-07,
"logits/chosen": -2.891150951385498,
"logits/rejected": -2.843611478805542,
"logps/chosen": -280.95947265625,
"logps/rejected": -248.34915161132812,
"loss": 0.6931,
"positive_losses": 0.0025458037853240967,
"rewards/accuracies": 0.6625000238418579,
"rewards/chosen": 0.0058168042451143265,
"rewards/margins": 0.0012899416033178568,
"rewards/margins_max": 0.005339757539331913,
"rewards/margins_min": -0.002916422439739108,
"rewards/margins_std": 0.003642101539298892,
"rewards/rejected": 0.004526862408965826,
"step": 60
},
{
"dpo_losses": 0.6921008229255676,
"epoch": 0.1,
"grad_norm": 1.9262001556938266,
"learning_rate": 4.999967381905813e-07,
"logits/chosen": -2.7916743755340576,
"logits/rejected": -2.7574045658111572,
"logps/chosen": -290.14361572265625,
"logps/rejected": -212.0419158935547,
"loss": 0.6926,
"positive_losses": 0.004826736636459827,
"rewards/accuracies": 0.675000011920929,
"rewards/chosen": 0.006598347332328558,
"rewards/margins": 0.002098249038681388,
"rewards/margins_max": 0.006628250237554312,
"rewards/margins_min": -0.0018369384342804551,
"rewards/margins_std": 0.00378109747543931,
"rewards/rejected": 0.004500098060816526,
"step": 70
},
{
"dpo_losses": 0.6920569539070129,
"epoch": 0.12,
"grad_norm": 13.284239668101833,
"learning_rate": 4.996054240392509e-07,
"logits/chosen": -2.7785937786102295,
"logits/rejected": -2.759201765060425,
"logps/chosen": -264.8086242675781,
"logps/rejected": -243.39212036132812,
"loss": 0.6931,
"positive_losses": 0.014845085330307484,
"rewards/accuracies": 0.6499999761581421,
"rewards/chosen": 0.007849474437534809,
"rewards/margins": 0.0021913917735219,
"rewards/margins_max": 0.008834488689899445,
"rewards/margins_min": -0.00322969863191247,
"rewards/margins_std": 0.005265987012535334,
"rewards/rejected": 0.0056580835953354836,
"step": 80
},
{
"dpo_losses": 0.6912720203399658,
"epoch": 0.13,
"grad_norm": 2.317442586572389,
"learning_rate": 4.985629178361649e-07,
"logits/chosen": -2.8908092975616455,
"logits/rejected": -2.8579392433166504,
"logps/chosen": -272.96112060546875,
"logps/rejected": -245.47073364257812,
"loss": 0.6918,
"positive_losses": 0.008871316909790039,
"rewards/accuracies": 0.737500011920929,
"rewards/chosen": 0.009552950970828533,
"rewards/margins": 0.0037618838250637054,
"rewards/margins_max": 0.010290957987308502,
"rewards/margins_min": -0.002331961877644062,
"rewards/margins_std": 0.005662451032549143,
"rewards/rejected": 0.00579106854274869,
"step": 90
},
{
"dpo_losses": 0.6900182962417603,
"epoch": 0.15,
"grad_norm": 12.043613650426611,
"learning_rate": 4.968719393609756e-07,
"logits/chosen": -2.8450093269348145,
"logits/rejected": -2.7789740562438965,
"logps/chosen": -337.09600830078125,
"logps/rejected": -243.02517700195312,
"loss": 0.6918,
"positive_losses": 0.017842482775449753,
"rewards/accuracies": 0.7250000238418579,
"rewards/chosen": 0.012230691500008106,
"rewards/margins": 0.0062872739508748055,
"rewards/margins_max": 0.01721397042274475,
"rewards/margins_min": -0.002463764976710081,
"rewards/margins_std": 0.008927479386329651,
"rewards/rejected": 0.0059434170834720135,
"step": 100
},
{
"epoch": 0.15,
"eval_dpo_losses": 0.6922104358673096,
"eval_logits/chosen": -2.8104522228240967,
"eval_logits/rejected": -2.7718498706817627,
"eval_logps/chosen": -283.23919677734375,
"eval_logps/rejected": -257.4134826660156,
"eval_loss": 0.6930333375930786,
"eval_positive_losses": 0.004661041311919689,
"eval_rewards/accuracies": 0.6000000238418579,
"eval_rewards/chosen": 0.013542202301323414,
"eval_rewards/margins": 0.001887724851258099,
"eval_rewards/margins_max": 0.013032433576881886,
"eval_rewards/margins_min": -0.008088048547506332,
"eval_rewards/margins_std": 0.006959096062928438,
"eval_rewards/rejected": 0.011654475703835487,
"eval_runtime": 439.0202,
"eval_samples_per_second": 4.556,
"eval_steps_per_second": 0.285,
"step": 100
},
{
"dpo_losses": 0.6904088258743286,
"epoch": 0.16,
"grad_norm": 5.183696409472315,
"learning_rate": 4.945369001834514e-07,
"logits/chosen": -2.8782482147216797,
"logits/rejected": -2.8235890865325928,
"logps/chosen": -287.38250732421875,
"logps/rejected": -222.830078125,
"loss": 0.6908,
"positive_losses": 0.0031826018821448088,
"rewards/accuracies": 0.7749999761581421,
"rewards/chosen": 0.01270085759460926,
"rewards/margins": 0.00550065329298377,
"rewards/margins_max": 0.014992681331932545,
"rewards/margins_min": -0.0012518208241090178,
"rewards/margins_std": 0.007309773005545139,
"rewards/rejected": 0.0072002047672867775,
"step": 110
},
{
"dpo_losses": 0.6896742582321167,
"epoch": 0.18,
"grad_norm": 1.9553746809300425,
"learning_rate": 4.915638921541951e-07,
"logits/chosen": -2.8301899433135986,
"logits/rejected": -2.8140978813171387,
"logps/chosen": -256.82879638671875,
"logps/rejected": -236.7211456298828,
"loss": 0.6898,
"positive_losses": 0.0022869110107421875,
"rewards/accuracies": 0.7875000238418579,
"rewards/chosen": 0.013588182628154755,
"rewards/margins": 0.00698423758149147,
"rewards/margins_max": 0.018451116979122162,
"rewards/margins_min": -0.001020266441628337,
"rewards/margins_std": 0.008837779983878136,
"rewards/rejected": 0.006603945046663284,
"step": 120
},
{
"dpo_losses": 0.6898840665817261,
"epoch": 0.19,
"grad_norm": 1.6022519000216686,
"learning_rate": 4.879606715117018e-07,
"logits/chosen": -2.881978988647461,
"logits/rejected": -2.82822847366333,
"logps/chosen": -274.6767578125,
"logps/rejected": -234.8880615234375,
"loss": 0.6897,
"positive_losses": 0.007077217102050781,
"rewards/accuracies": 0.6625000238418579,
"rewards/chosen": 0.014239470474421978,
"rewards/margins": 0.006568121258169413,
"rewards/margins_max": 0.017170870676636696,
"rewards/margins_min": -0.0028213211335241795,
"rewards/margins_std": 0.00886800978332758,
"rewards/rejected": 0.00767135014757514,
"step": 130
},
{
"dpo_losses": 0.6880620718002319,
"epoch": 0.2,
"grad_norm": 3.638881806293095,
"learning_rate": 4.837366386472174e-07,
"logits/chosen": -2.8916850090026855,
"logits/rejected": -2.8278439044952393,
"logps/chosen": -296.95111083984375,
"logps/rejected": -253.7423553466797,
"loss": 0.6901,
"positive_losses": 0.002955913543701172,
"rewards/accuracies": 0.8125,
"rewards/chosen": 0.01859217882156372,
"rewards/margins": 0.010244466364383698,
"rewards/margins_max": 0.024400196969509125,
"rewards/margins_min": -0.0004118576762266457,
"rewards/margins_std": 0.011427570134401321,
"rewards/rejected": 0.008347713388502598,
"step": 140
},
{
"dpo_losses": 0.6885315179824829,
"epoch": 0.22,
"grad_norm": 6.527916454528557,
"learning_rate": 4.789028135801918e-07,
"logits/chosen": -2.852397918701172,
"logits/rejected": -2.8406243324279785,
"logps/chosen": -297.78564453125,
"logps/rejected": -271.2729187011719,
"loss": 0.6914,
"positive_losses": 0.054250240325927734,
"rewards/accuracies": 0.6499999761581421,
"rewards/chosen": 0.01786804385483265,
"rewards/margins": 0.009341379627585411,
"rewards/margins_max": 0.03060457669198513,
"rewards/margins_min": -0.008363897912204266,
"rewards/margins_std": 0.01767081953585148,
"rewards/rejected": 0.008526663295924664,
"step": 150
},
{
"dpo_losses": 0.6880606412887573,
"epoch": 0.23,
"grad_norm": 8.44151922500754,
"learning_rate": 4.7347180720830627e-07,
"logits/chosen": -2.8739285469055176,
"logits/rejected": -2.7969882488250732,
"logps/chosen": -300.0534973144531,
"logps/rejected": -260.5088195800781,
"loss": 0.6899,
"positive_losses": 0.006076812744140625,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": 0.01983806863427162,
"rewards/margins": 0.01025958452373743,
"rewards/margins_max": 0.028022507205605507,
"rewards/margins_min": -0.0024425422307103872,
"rewards/margins_std": 0.013813845813274384,
"rewards/rejected": 0.009578483179211617,
"step": 160
},
{
"dpo_losses": 0.684917151927948,
"epoch": 0.25,
"grad_norm": 1.8673882157090276,
"learning_rate": 4.6745778840708107e-07,
"logits/chosen": -2.8020739555358887,
"logits/rejected": -2.7900993824005127,
"logps/chosen": -295.7058410644531,
"logps/rejected": -225.8823699951172,
"loss": 0.6858,
"positive_losses": 0.0014563084114342928,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": 0.026786331087350845,
"rewards/margins": 0.01665376126766205,
"rewards/margins_max": 0.03810211643576622,
"rewards/margins_min": -0.0024380278773605824,
"rewards/margins_std": 0.017996035516262054,
"rewards/rejected": 0.010132570751011372,
"step": 170
},
{
"dpo_losses": 0.6871532201766968,
"epoch": 0.26,
"grad_norm": 2.603750597458806,
"learning_rate": 4.6087644706489703e-07,
"logits/chosen": -2.833789587020874,
"logits/rejected": -2.7939248085021973,
"logps/chosen": -294.24456787109375,
"logps/rejected": -254.28988647460938,
"loss": 0.6889,
"positive_losses": 0.024321984499692917,
"rewards/accuracies": 0.75,
"rewards/chosen": 0.022618165239691734,
"rewards/margins": 0.01210461463779211,
"rewards/margins_max": 0.030625376850366592,
"rewards/margins_min": -0.004277624189853668,
"rewards/margins_std": 0.015746023505926132,
"rewards/rejected": 0.010513553395867348,
"step": 180
},
{
"dpo_losses": 0.686254620552063,
"epoch": 0.28,
"grad_norm": 1.7335945641686348,
"learning_rate": 4.537449531498687e-07,
"logits/chosen": -2.7748355865478516,
"logits/rejected": -2.725888729095459,
"logps/chosen": -280.95428466796875,
"logps/rejected": -228.15097045898438,
"loss": 0.6878,
"positive_losses": 0.012805747799575329,
"rewards/accuracies": 0.7875000238418579,
"rewards/chosen": 0.028492039069533348,
"rewards/margins": 0.013932852074503899,
"rewards/margins_max": 0.03530807048082352,
"rewards/margins_min": -0.002935736905783415,
"rewards/margins_std": 0.017064960673451424,
"rewards/rejected": 0.0145591851323843,
"step": 190
},
{
"dpo_losses": 0.6861851811408997,
"epoch": 0.29,
"grad_norm": 1.5625317379589145,
"learning_rate": 4.4608191191535736e-07,
"logits/chosen": -2.8794567584991455,
"logits/rejected": -2.835148334503174,
"logps/chosen": -274.84735107421875,
"logps/rejected": -242.3157958984375,
"loss": 0.6867,
"positive_losses": 0.004458713345229626,
"rewards/accuracies": 0.6875,
"rewards/chosen": 0.02746543660759926,
"rewards/margins": 0.014083576388657093,
"rewards/margins_max": 0.03864169120788574,
"rewards/margins_min": -0.005826378706842661,
"rewards/margins_std": 0.019664574414491653,
"rewards/rejected": 0.01338185928761959,
"step": 200
},
{
"epoch": 0.29,
"eval_dpo_losses": 0.6900327801704407,
"eval_logits/chosen": -2.8075411319732666,
"eval_logits/rejected": -2.7689805030822754,
"eval_logps/chosen": -281.6053161621094,
"eval_logps/rejected": -256.2250061035156,
"eval_loss": 0.6917006373405457,
"eval_positive_losses": 0.012367370538413525,
"eval_rewards/accuracies": 0.6100000143051147,
"eval_rewards/chosen": 0.02988099865615368,
"eval_rewards/margins": 0.006341494619846344,
"eval_rewards/margins_max": 0.03729286044836044,
"eval_rewards/margins_min": -0.02144700102508068,
"eval_rewards/margins_std": 0.01944614388048649,
"eval_rewards/rejected": 0.023539502173662186,
"eval_runtime": 457.2097,
"eval_samples_per_second": 4.374,
"eval_steps_per_second": 0.273,
"step": 200
},
{
"dpo_losses": 0.6849207878112793,
"epoch": 0.31,
"grad_norm": 6.776479704467401,
"learning_rate": 4.379073153609896e-07,
"logits/chosen": -2.85437273979187,
"logits/rejected": -2.8159098625183105,
"logps/chosen": -298.1737365722656,
"logps/rejected": -258.9627685546875,
"loss": 0.6892,
"positive_losses": 0.024309396743774414,
"rewards/accuracies": 0.737500011920929,
"rewards/chosen": 0.031262680888175964,
"rewards/margins": 0.016705606132745743,
"rewards/margins_max": 0.0462987944483757,
"rewards/margins_min": -0.009217771701514721,
"rewards/margins_std": 0.024988915771245956,
"rewards/rejected": 0.014557072892785072,
"step": 210
},
{
"dpo_losses": 0.6835730075836182,
"epoch": 0.32,
"grad_norm": 1.7559811016317117,
"learning_rate": 4.292424900758128e-07,
"logits/chosen": -2.8128745555877686,
"logits/rejected": -2.7835748195648193,
"logps/chosen": -278.7519836425781,
"logps/rejected": -221.44998168945312,
"loss": 0.687,
"positive_losses": 0.01152725238353014,
"rewards/accuracies": 0.7875000238418579,
"rewards/chosen": 0.032679036259651184,
"rewards/margins": 0.019430814310908318,
"rewards/margins_max": 0.04703899472951889,
"rewards/margins_min": -0.004503333941102028,
"rewards/margins_std": 0.02318265102803707,
"rewards/rejected": 0.013248221948742867,
"step": 220
},
{
"dpo_losses": 0.6845378279685974,
"epoch": 0.34,
"grad_norm": 5.2855521291360335,
"learning_rate": 4.201100415996597e-07,
"logits/chosen": -2.7867586612701416,
"logits/rejected": -2.7540698051452637,
"logps/chosen": -262.8429260253906,
"logps/rejected": -259.90594482421875,
"loss": 0.6873,
"positive_losses": 0.010059595108032227,
"rewards/accuracies": 0.737500011920929,
"rewards/chosen": 0.033603329211473465,
"rewards/margins": 0.017468122765421867,
"rewards/margins_max": 0.05101371556520462,
"rewards/margins_min": -0.005049402825534344,
"rewards/margins_std": 0.025209274142980576,
"rewards/rejected": 0.016135206446051598,
"step": 230
},
{
"dpo_losses": 0.6822143793106079,
"epoch": 0.35,
"grad_norm": 7.1123081067417635,
"learning_rate": 4.1053379544787557e-07,
"logits/chosen": -2.8566319942474365,
"logits/rejected": -2.8126063346862793,
"logps/chosen": -285.74066162109375,
"logps/rejected": -246.8583221435547,
"loss": 0.6875,
"positive_losses": 0.0023063658736646175,
"rewards/accuracies": 0.8125,
"rewards/chosen": 0.03864200785756111,
"rewards/margins": 0.02219047024846077,
"rewards/margins_max": 0.04993893578648567,
"rewards/margins_min": -0.002027359325438738,
"rewards/margins_std": 0.023801075294613838,
"rewards/rejected": 0.01645154133439064,
"step": 240
},
{
"dpo_losses": 0.6844903230667114,
"epoch": 0.37,
"grad_norm": 2.225597078597608,
"learning_rate": 4.0053873495326964e-07,
"logits/chosen": -2.8150901794433594,
"logits/rejected": -2.798675537109375,
"logps/chosen": -243.178955078125,
"logps/rejected": -226.9827117919922,
"loss": 0.6875,
"positive_losses": 0.0484701506793499,
"rewards/accuracies": 0.7250000238418579,
"rewards/chosen": 0.033011239022016525,
"rewards/margins": 0.017598429694771767,
"rewards/margins_max": 0.047145940363407135,
"rewards/margins_min": -0.0065514100715518,
"rewards/margins_std": 0.02364528551697731,
"rewards/rejected": 0.015412810258567333,
"step": 250
},
{
"dpo_losses": 0.6823347806930542,
"epoch": 0.38,
"grad_norm": 1.8954330972040565,
"learning_rate": 3.9015093608745143e-07,
"logits/chosen": -2.87410306930542,
"logits/rejected": -2.808983087539673,
"logps/chosen": -283.3040771484375,
"logps/rejected": -240.4110565185547,
"loss": 0.6824,
"positive_losses": 0.011256242170929909,
"rewards/accuracies": 0.824999988079071,
"rewards/chosen": 0.03804592788219452,
"rewards/margins": 0.021955588832497597,
"rewards/margins_max": 0.05776814743876457,
"rewards/margins_min": -0.004212898667901754,
"rewards/margins_std": 0.028291797265410423,
"rewards/rejected": 0.016090339049696922,
"step": 260
},
{
"dpo_losses": 0.681152880191803,
"epoch": 0.39,
"grad_norm": 2.3462040578320438,
"learning_rate": 3.79397499431599e-07,
"logits/chosen": -2.838275909423828,
"logits/rejected": -2.7828450202941895,
"logps/chosen": -291.1107177734375,
"logps/rejected": -261.1417236328125,
"loss": 0.6844,
"positive_losses": 0.031092902645468712,
"rewards/accuracies": 0.762499988079071,
"rewards/chosen": 0.04311501979827881,
"rewards/margins": 0.024433860555291176,
"rewards/margins_max": 0.05960015207529068,
"rewards/margins_min": -0.005230102688074112,
"rewards/margins_std": 0.028690297156572342,
"rewards/rejected": 0.018681161105632782,
"step": 270
},
{
"dpo_losses": 0.6834903955459595,
"epoch": 0.41,
"grad_norm": 1.678928576392419,
"learning_rate": 3.6830647947413694e-07,
"logits/chosen": -2.906564235687256,
"logits/rejected": -2.8589329719543457,
"logps/chosen": -273.8109436035156,
"logps/rejected": -246.07925415039062,
"loss": 0.6848,
"positive_losses": 0.03764519840478897,
"rewards/accuracies": 0.762499988079071,
"rewards/chosen": 0.03491830825805664,
"rewards/margins": 0.019722772762179375,
"rewards/margins_max": 0.05492109805345535,
"rewards/margins_min": -0.009231898002326488,
"rewards/margins_std": 0.029070580378174782,
"rewards/rejected": 0.015195533633232117,
"step": 280
},
{
"dpo_losses": 0.6819528937339783,
"epoch": 0.42,
"grad_norm": 1.5933364630845321,
"learning_rate": 3.5690681141977837e-07,
"logits/chosen": -2.8415896892547607,
"logits/rejected": -2.795182704925537,
"logps/chosen": -252.76455688476562,
"logps/rejected": -203.42364501953125,
"loss": 0.6836,
"positive_losses": 0.024924278259277344,
"rewards/accuracies": 0.762499988079071,
"rewards/chosen": 0.039224088191986084,
"rewards/margins": 0.022878160700201988,
"rewards/margins_max": 0.06510117650032043,
"rewards/margins_min": -0.010645577684044838,
"rewards/margins_std": 0.034357957541942596,
"rewards/rejected": 0.016345927491784096,
"step": 290
},
{
"dpo_losses": 0.6809358596801758,
"epoch": 0.44,
"grad_norm": 10.347091457783785,
"learning_rate": 3.4522823570088067e-07,
"logits/chosen": -2.840087413787842,
"logits/rejected": -2.8272616863250732,
"logps/chosen": -259.23687744140625,
"logps/rejected": -251.6184844970703,
"loss": 0.6825,
"positive_losses": 0.014734840020537376,
"rewards/accuracies": 0.675000011920929,
"rewards/chosen": 0.03966939076781273,
"rewards/margins": 0.02503836713731289,
"rewards/margins_max": 0.07522507011890411,
"rewards/margins_min": -0.010506678372621536,
"rewards/margins_std": 0.03809111937880516,
"rewards/rejected": 0.01463102363049984,
"step": 300
},
{
"epoch": 0.44,
"eval_dpo_losses": 0.6878849864006042,
"eval_logits/chosen": -2.8020598888397217,
"eval_logits/rejected": -2.763356924057007,
"eval_logps/chosen": -280.1855163574219,
"eval_logps/rejected": -255.25390625,
"eval_loss": 0.69089674949646,
"eval_positive_losses": 0.022481659427285194,
"eval_rewards/accuracies": 0.625,
"eval_rewards/chosen": 0.04407881200313568,
"eval_rewards/margins": 0.010828724130988121,
"eval_rewards/margins_max": 0.061860937625169754,
"eval_rewards/margins_min": -0.03410045802593231,
"eval_rewards/margins_std": 0.03193770721554756,
"eval_rewards/rejected": 0.03325008600950241,
"eval_runtime": 439.4476,
"eval_samples_per_second": 4.551,
"eval_steps_per_second": 0.284,
"step": 300
},
{
"dpo_losses": 0.6795910596847534,
"epoch": 0.45,
"grad_norm": 1.8194274060306528,
"learning_rate": 3.3330122038805277e-07,
"logits/chosen": -2.861931562423706,
"logits/rejected": -2.8135368824005127,
"logps/chosen": -271.8392639160156,
"logps/rejected": -225.01708984375,
"loss": 0.6837,
"positive_losses": 0.02801961824297905,
"rewards/accuracies": 0.8500000238418579,
"rewards/chosen": 0.04437270388007164,
"rewards/margins": 0.027612268924713135,
"rewards/margins_max": 0.07083487510681152,
"rewards/margins_min": -0.0025140452198684216,
"rewards/margins_std": 0.033803679049015045,
"rewards/rejected": 0.016760434955358505,
"step": 310
},
{
"dpo_losses": 0.6798790693283081,
"epoch": 0.47,
"grad_norm": 1.8110670147086625,
"learning_rate": 3.2115688170243734e-07,
"logits/chosen": -2.8523640632629395,
"logits/rejected": -2.82257080078125,
"logps/chosen": -270.28253173828125,
"logps/rejected": -248.6698455810547,
"loss": 0.6802,
"positive_losses": 0.009492230601608753,
"rewards/accuracies": 0.762499988079071,
"rewards/chosen": 0.04459889978170395,
"rewards/margins": 0.027110621333122253,
"rewards/margins_max": 0.06967566907405853,
"rewards/margins_min": -0.005822637118399143,
"rewards/margins_std": 0.03437542915344238,
"rewards/rejected": 0.017488282173871994,
"step": 320
},
{
"dpo_losses": 0.6773503422737122,
"epoch": 0.48,
"grad_norm": 9.150401672552029,
"learning_rate": 3.088269028370435e-07,
"logits/chosen": -2.875394821166992,
"logits/rejected": -2.8434793949127197,
"logps/chosen": -295.0433349609375,
"logps/rejected": -246.5893096923828,
"loss": 0.6835,
"positive_losses": 0.03743886947631836,
"rewards/accuracies": 0.7875000238418579,
"rewards/chosen": 0.04838422313332558,
"rewards/margins": 0.03234843909740448,
"rewards/margins_max": 0.08645691722631454,
"rewards/margins_min": -0.006109377834945917,
"rewards/margins_std": 0.04178870469331741,
"rewards/rejected": 0.016035789623856544,
"step": 330
},
{
"dpo_losses": 0.6774032115936279,
"epoch": 0.5,
"grad_norm": 4.795115194677247,
"learning_rate": 2.9634345129891294e-07,
"logits/chosen": -2.8238093852996826,
"logits/rejected": -2.785935163497925,
"logps/chosen": -287.4772033691406,
"logps/rejected": -254.5933074951172,
"loss": 0.6804,
"positive_losses": 0.03258161619305611,
"rewards/accuracies": 0.762499988079071,
"rewards/chosen": 0.04798700287938118,
"rewards/margins": 0.032188571989536285,
"rewards/margins_max": 0.0826764851808548,
"rewards/margins_min": -0.006551785860210657,
"rewards/margins_std": 0.04137192294001579,
"rewards/rejected": 0.015798425301909447,
"step": 340
},
{
"dpo_losses": 0.6810469627380371,
"epoch": 0.51,
"grad_norm": 11.724114006745417,
"learning_rate": 2.8373909498776744e-07,
"logits/chosen": -2.8722732067108154,
"logits/rejected": -2.854133129119873,
"logps/chosen": -279.33917236328125,
"logps/rejected": -276.32232666015625,
"loss": 0.6845,
"positive_losses": 0.04124107211828232,
"rewards/accuracies": 0.762499988079071,
"rewards/chosen": 0.04524795338511467,
"rewards/margins": 0.024712715297937393,
"rewards/margins_max": 0.06821637600660324,
"rewards/margins_min": -0.0043562790378928185,
"rewards/margins_std": 0.03368216007947922,
"rewards/rejected": 0.020535241812467575,
"step": 350
},
{
"dpo_losses": 0.6761634945869446,
"epoch": 0.53,
"grad_norm": 5.401380341327095,
"learning_rate": 2.710467172300768e-07,
"logits/chosen": -2.7806248664855957,
"logits/rejected": -2.74116587638855,
"logps/chosen": -318.1079406738281,
"logps/rejected": -248.9733428955078,
"loss": 0.6774,
"positive_losses": 0.012074279598891735,
"rewards/accuracies": 0.7875000238418579,
"rewards/chosen": 0.05469100549817085,
"rewards/margins": 0.03459729254245758,
"rewards/margins_max": 0.0798027291893959,
"rewards/margins_min": -0.0013110924046486616,
"rewards/margins_std": 0.03574884682893753,
"rewards/rejected": 0.020093712955713272,
"step": 360
},
{
"dpo_losses": 0.6761563420295715,
"epoch": 0.54,
"grad_norm": 8.370611151073765,
"learning_rate": 2.582994309902146e-07,
"logits/chosen": -2.8219902515411377,
"logits/rejected": -2.7716636657714844,
"logps/chosen": -304.6541748046875,
"logps/rejected": -254.8093719482422,
"loss": 0.6842,
"positive_losses": 0.04906577989459038,
"rewards/accuracies": 0.737500011920929,
"rewards/chosen": 0.054839521646499634,
"rewards/margins": 0.03491091728210449,
"rewards/margins_max": 0.09758211672306061,
"rewards/margins_min": -0.014001704752445221,
"rewards/margins_std": 0.05021950602531433,
"rewards/rejected": 0.019928600639104843,
"step": 370
},
{
"dpo_losses": 0.6737654805183411,
"epoch": 0.56,
"grad_norm": 2.3064692447314554,
"learning_rate": 2.455304924825151e-07,
"logits/chosen": -2.837787389755249,
"logits/rejected": -2.831657886505127,
"logps/chosen": -275.2360534667969,
"logps/rejected": -255.569580078125,
"loss": 0.6738,
"positive_losses": 0.01725444756448269,
"rewards/accuracies": 0.8125,
"rewards/chosen": 0.05640103667974472,
"rewards/margins": 0.0399165078997612,
"rewards/margins_max": 0.09945692121982574,
"rewards/margins_min": -0.001225404441356659,
"rewards/margins_std": 0.045562874525785446,
"rewards/rejected": 0.016484536230564117,
"step": 380
},
{
"dpo_losses": 0.6766735911369324,
"epoch": 0.57,
"grad_norm": 1.805419113024878,
"learning_rate": 2.3277321440960732e-07,
"logits/chosen": -2.838366985321045,
"logits/rejected": -2.793659210205078,
"logps/chosen": -266.8939208984375,
"logps/rejected": -261.4732666015625,
"loss": 0.6807,
"positive_losses": 0.028429364785552025,
"rewards/accuracies": 0.7250000238418579,
"rewards/chosen": 0.0527314729988575,
"rewards/margins": 0.034005872905254364,
"rewards/margins_max": 0.10273897647857666,
"rewards/margins_min": -0.011970164254307747,
"rewards/margins_std": 0.05079951137304306,
"rewards/rejected": 0.018725596368312836,
"step": 390
},
{
"dpo_losses": 0.6763077974319458,
"epoch": 0.58,
"grad_norm": 6.733018778387388,
"learning_rate": 2.2006087905337698e-07,
"logits/chosen": -2.868297815322876,
"logits/rejected": -2.8423171043395996,
"logps/chosen": -238.1276092529297,
"logps/rejected": -231.619384765625,
"loss": 0.6808,
"positive_losses": 0.045419443398714066,
"rewards/accuracies": 0.7749999761581421,
"rewards/chosen": 0.05248277261853218,
"rewards/margins": 0.03450237587094307,
"rewards/margins_max": 0.09071175009012222,
"rewards/margins_min": -0.006746978498995304,
"rewards/margins_std": 0.04475879669189453,
"rewards/rejected": 0.01798040233552456,
"step": 400
},
{
"epoch": 0.58,
"eval_dpo_losses": 0.6860782504081726,
"eval_logits/chosen": -2.7994227409362793,
"eval_logits/rejected": -2.760826826095581,
"eval_logps/chosen": -279.1606140136719,
"eval_logps/rejected": -254.61215209960938,
"eval_loss": 0.6905239224433899,
"eval_positive_losses": 0.0314272977411747,
"eval_rewards/accuracies": 0.6169999837875366,
"eval_rewards/chosen": 0.05432803928852081,
"eval_rewards/margins": 0.014660396613180637,
"eval_rewards/margins_max": 0.08083328604698181,
"eval_rewards/margins_min": -0.043623533099889755,
"eval_rewards/margins_std": 0.04150667414069176,
"eval_rewards/rejected": 0.03966763988137245,
"eval_runtime": 428.0845,
"eval_samples_per_second": 4.672,
"eval_steps_per_second": 0.292,
"step": 400
},
{
"dpo_losses": 0.6729803085327148,
"epoch": 0.6,
"grad_norm": 6.255311463771333,
"learning_rate": 2.0742665144529372e-07,
"logits/chosen": -2.852710723876953,
"logits/rejected": -2.812171459197998,
"logps/chosen": -307.93133544921875,
"logps/rejected": -281.7948303222656,
"loss": 0.6784,
"positive_losses": 0.0057888031005859375,
"rewards/accuracies": 0.7250000238418579,
"rewards/chosen": 0.06238215044140816,
"rewards/margins": 0.04161245375871658,
"rewards/margins_max": 0.1156817227602005,
"rewards/margins_min": -0.008250080049037933,
"rewards/margins_std": 0.05661824345588684,
"rewards/rejected": 0.020769696682691574,
"step": 410
},
{
"dpo_losses": 0.6805132627487183,
"epoch": 0.61,
"grad_norm": 2.3659410877777494,
"learning_rate": 1.9490349284263033e-07,
"logits/chosen": -2.788618564605713,
"logits/rejected": -2.7586796283721924,
"logps/chosen": -255.90536499023438,
"logps/rejected": -242.9296417236328,
"loss": 0.6804,
"positive_losses": 0.0345400795340538,
"rewards/accuracies": 0.7250000238418579,
"rewards/chosen": 0.051929228007793427,
"rewards/margins": 0.026662463322281837,
"rewards/margins_max": 0.08727220445871353,
"rewards/margins_min": -0.02731608971953392,
"rewards/margins_std": 0.053510747849941254,
"rewards/rejected": 0.02526676654815674,
"step": 420
},
{
"dpo_losses": 0.6737715005874634,
"epoch": 0.63,
"grad_norm": 2.271597415804245,
"learning_rate": 1.8252407473630605e-07,
"logits/chosen": -2.8645436763763428,
"logits/rejected": -2.8088572025299072,
"logps/chosen": -295.20916748046875,
"logps/rejected": -230.15567016601562,
"loss": 0.6795,
"positive_losses": 0.03842503949999809,
"rewards/accuracies": 0.8125,
"rewards/chosen": 0.058985937386751175,
"rewards/margins": 0.03972195088863373,
"rewards/margins_max": 0.0958792120218277,
"rewards/margins_min": -0.005095441360026598,
"rewards/margins_std": 0.044632021337747574,
"rewards/rejected": 0.01926398277282715,
"step": 430
},
{
"dpo_losses": 0.6767727136611938,
"epoch": 0.64,
"grad_norm": 10.115009795187461,
"learning_rate": 1.7032069361469764e-07,
"logits/chosen": -2.7652392387390137,
"logits/rejected": -2.7494540214538574,
"logps/chosen": -226.9882354736328,
"logps/rejected": -269.02301025390625,
"loss": 0.6792,
"positive_losses": 0.009693240746855736,
"rewards/accuracies": 0.7250000238418579,
"rewards/chosen": 0.05702495574951172,
"rewards/margins": 0.03372834250330925,
"rewards/margins_max": 0.09970168769359589,
"rewards/margins_min": -0.013438252732157707,
"rewards/margins_std": 0.051222048699855804,
"rewards/rejected": 0.02329660952091217,
"step": 440
},
{
"dpo_losses": 0.6781430244445801,
"epoch": 0.66,
"grad_norm": 1.5440271538041583,
"learning_rate": 1.58325186705788e-07,
"logits/chosen": -2.8817787170410156,
"logits/rejected": -2.8590903282165527,
"logps/chosen": -267.78729248046875,
"logps/rejected": -251.52542114257812,
"loss": 0.6868,
"positive_losses": 0.08534412086009979,
"rewards/accuracies": 0.7749999761581421,
"rewards/chosen": 0.05013534426689148,
"rewards/margins": 0.031011879444122314,
"rewards/margins_max": 0.09007111191749573,
"rewards/margins_min": -0.02042573131620884,
"rewards/margins_std": 0.04950300231575966,
"rewards/rejected": 0.019123462960124016,
"step": 450
},
{
"dpo_losses": 0.675542950630188,
"epoch": 0.67,
"grad_norm": 1.6924810012359601,
"learning_rate": 1.4656884891747395e-07,
"logits/chosen": -2.812002182006836,
"logits/rejected": -2.765144109725952,
"logps/chosen": -282.1943664550781,
"logps/rejected": -234.0194854736328,
"loss": 0.6808,
"positive_losses": 0.03008117713034153,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": 0.055878765881061554,
"rewards/margins": 0.03611396998167038,
"rewards/margins_max": 0.09502451121807098,
"rewards/margins_min": -0.008020835928618908,
"rewards/margins_std": 0.04623667150735855,
"rewards/rejected": 0.019764794036746025,
"step": 460
},
{
"dpo_losses": 0.6700197458267212,
"epoch": 0.69,
"grad_norm": 1.8078995560874291,
"learning_rate": 1.3508235119272466e-07,
"logits/chosen": -2.8690173625946045,
"logits/rejected": -2.823881149291992,
"logps/chosen": -322.7649230957031,
"logps/rejected": -289.1722106933594,
"loss": 0.6782,
"positive_losses": 0.07821483910083771,
"rewards/accuracies": 0.7749999761581421,
"rewards/chosen": 0.06522538512945175,
"rewards/margins": 0.04820011183619499,
"rewards/margins_max": 0.1358465701341629,
"rewards/margins_min": -0.011710671707987785,
"rewards/margins_std": 0.06825564056634903,
"rewards/rejected": 0.017025265842676163,
"step": 470
},
{
"dpo_losses": 0.6712939143180847,
"epoch": 0.7,
"grad_norm": 1.451487777642733,
"learning_rate": 1.2389566049259336e-07,
"logits/chosen": -2.811398983001709,
"logits/rejected": -2.7891058921813965,
"logps/chosen": -238.219970703125,
"logps/rejected": -230.36062622070312,
"loss": 0.6767,
"positive_losses": 0.054373931139707565,
"rewards/accuracies": 0.7875000238418579,
"rewards/chosen": 0.06273017823696136,
"rewards/margins": 0.0451803021132946,
"rewards/margins_max": 0.10969402641057968,
"rewards/margins_min": -0.0036571747623384,
"rewards/margins_std": 0.051760654896497726,
"rewards/rejected": 0.017549879848957062,
"step": 480
},
{
"dpo_losses": 0.6667095422744751,
"epoch": 0.72,
"grad_norm": 2.410585831948769,
"learning_rate": 1.1303796161583762e-07,
"logits/chosen": -2.9242846965789795,
"logits/rejected": -2.840134859085083,
"logps/chosen": -316.83929443359375,
"logps/rejected": -256.23675537109375,
"loss": 0.6785,
"positive_losses": 0.06970572471618652,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": 0.0768619254231453,
"rewards/margins": 0.05491393804550171,
"rewards/margins_max": 0.13214218616485596,
"rewards/margins_min": -0.004780332557857037,
"rewards/margins_std": 0.06225153058767319,
"rewards/rejected": 0.02194797992706299,
"step": 490
},
{
"dpo_losses": 0.6667904853820801,
"epoch": 0.73,
"grad_norm": 1.9094755680052469,
"learning_rate": 1.0253758105911167e-07,
"logits/chosen": -2.886962890625,
"logits/rejected": -2.840977430343628,
"logps/chosen": -331.50872802734375,
"logps/rejected": -285.4704895019531,
"loss": 0.6769,
"positive_losses": 0.03339347988367081,
"rewards/accuracies": 0.875,
"rewards/chosen": 0.07617944478988647,
"rewards/margins": 0.0544399730861187,
"rewards/margins_max": 0.13042126595973969,
"rewards/margins_min": -0.00018090903176926076,
"rewards/margins_std": 0.05866100266575813,
"rewards/rejected": 0.021739467978477478,
"step": 500
},
{
"epoch": 0.73,
"eval_dpo_losses": 0.6854846477508545,
"eval_logits/chosen": -2.797536611557007,
"eval_logits/rejected": -2.7586684226989746,
"eval_logps/chosen": -278.739990234375,
"eval_logps/rejected": -254.3216094970703,
"eval_loss": 0.6902235150337219,
"eval_positive_losses": 0.032978080213069916,
"eval_rewards/accuracies": 0.6119999885559082,
"eval_rewards/chosen": 0.058533985167741776,
"eval_rewards/margins": 0.015960821881890297,
"eval_rewards/margins_max": 0.08850042521953583,
"eval_rewards/margins_min": -0.048725374042987823,
"eval_rewards/margins_std": 0.045832596719264984,
"eval_rewards/rejected": 0.04257315769791603,
"eval_runtime": 448.6936,
"eval_samples_per_second": 4.457,
"eval_steps_per_second": 0.279,
"step": 500
},
{
"dpo_losses": 0.6742907166481018,
"epoch": 0.75,
"grad_norm": 2.2304663983815316,
"learning_rate": 9.242191311637049e-08,
"logits/chosen": -2.8333396911621094,
"logits/rejected": -2.8009631633758545,
"logps/chosen": -277.5825500488281,
"logps/rejected": -235.39108276367188,
"loss": 0.6774,
"positive_losses": 0.007343196775764227,
"rewards/accuracies": 0.7749999761581421,
"rewards/chosen": 0.05940967798233032,
"rewards/margins": 0.038841020315885544,
"rewards/margins_max": 0.1059766411781311,
"rewards/margins_min": -0.00883492175489664,
"rewards/margins_std": 0.052775926887989044,
"rewards/rejected": 0.02056865207850933,
"step": 510
},
{
"dpo_losses": 0.6684087514877319,
"epoch": 0.76,
"grad_norm": 1.9114464768987025,
"learning_rate": 8.271734841028552e-08,
"logits/chosen": -2.8523058891296387,
"logits/rejected": -2.7778074741363525,
"logps/chosen": -335.83966064453125,
"logps/rejected": -264.36016845703125,
"loss": 0.6743,
"positive_losses": 0.0079803466796875,
"rewards/accuracies": 0.887499988079071,
"rewards/chosen": 0.06848612427711487,
"rewards/margins": 0.05102965235710144,
"rewards/margins_max": 0.12273881584405899,
"rewards/margins_min": -0.0025417390279471874,
"rewards/margins_std": 0.05730118602514267,
"rewards/rejected": 0.017456475645303726,
"step": 520
},
{
"dpo_losses": 0.6773437261581421,
"epoch": 0.77,
"grad_norm": 1.8982933837442546,
"learning_rate": 7.344920504212243e-08,
"logits/chosen": -2.839207172393799,
"logits/rejected": -2.7776246070861816,
"logps/chosen": -260.9156799316406,
"logps/rejected": -216.26376342773438,
"loss": 0.6801,
"positive_losses": 0.04813060909509659,
"rewards/accuracies": 0.7749999761581421,
"rewards/chosen": 0.054224640130996704,
"rewards/margins": 0.032528605312108994,
"rewards/margins_max": 0.09149853885173798,
"rewards/margins_min": -0.01699809357523918,
"rewards/margins_std": 0.04981226474046707,
"rewards/rejected": 0.02169603481888771,
"step": 530
},
{
"dpo_losses": 0.6705015301704407,
"epoch": 0.79,
"grad_norm": 1.723774449740064,
"learning_rate": 6.46416625397067e-08,
"logits/chosen": -2.834272861480713,
"logits/rejected": -2.7763218879699707,
"logps/chosen": -320.51885986328125,
"logps/rejected": -295.53875732421875,
"loss": 0.6751,
"positive_losses": 0.018300628289580345,
"rewards/accuracies": 0.8374999761581421,
"rewards/chosen": 0.06741219758987427,
"rewards/margins": 0.04666105657815933,
"rewards/margins_max": 0.10739126056432724,
"rewards/margins_min": -0.008476030081510544,
"rewards/margins_std": 0.05081728845834732,
"rewards/rejected": 0.020751142874360085,
"step": 540
},
{
"dpo_losses": 0.6752179861068726,
"epoch": 0.8,
"grad_norm": 1.6108967521538597,
"learning_rate": 5.6317698775795344e-08,
"logits/chosen": -2.8729920387268066,
"logits/rejected": -2.833665370941162,
"logps/chosen": -278.38275146484375,
"logps/rejected": -270.0639953613281,
"loss": 0.6723,
"positive_losses": 0.005997705273330212,
"rewards/accuracies": 0.75,
"rewards/chosen": 0.057433564215898514,
"rewards/margins": 0.037055402994155884,
"rewards/margins_max": 0.09951488673686981,
"rewards/margins_min": -0.01219971850514412,
"rewards/margins_std": 0.051766276359558105,
"rewards/rejected": 0.02037816308438778,
"step": 550
},
{
"dpo_losses": 0.6663364171981812,
"epoch": 0.82,
"grad_norm": 1.9255604336867507,
"learning_rate": 4.849903002143113e-08,
"logits/chosen": -2.9117777347564697,
"logits/rejected": -2.8610081672668457,
"logps/chosen": -328.8406066894531,
"logps/rejected": -272.1076965332031,
"loss": 0.6731,
"positive_losses": 0.038059137761592865,
"rewards/accuracies": 0.7749999761581421,
"rewards/chosen": 0.07473967969417572,
"rewards/margins": 0.055439990013837814,
"rewards/margins_max": 0.12707814574241638,
"rewards/margins_min": -0.002053080592304468,
"rewards/margins_std": 0.05929755046963692,
"rewards/rejected": 0.019299697130918503,
"step": 560
},
{
"dpo_losses": 0.6710665822029114,
"epoch": 0.83,
"grad_norm": 1.733507781542446,
"learning_rate": 4.1206054290670535e-08,
"logits/chosen": -2.8613619804382324,
"logits/rejected": -2.82637357711792,
"logps/chosen": -266.23773193359375,
"logps/rejected": -258.0359802246094,
"loss": 0.6756,
"positive_losses": 0.013837432488799095,
"rewards/accuracies": 0.7875000238418579,
"rewards/chosen": 0.06855504214763641,
"rewards/margins": 0.04551316797733307,
"rewards/margins_max": 0.1095261201262474,
"rewards/margins_min": -0.004014958627521992,
"rewards/margins_std": 0.0510232150554657,
"rewards/rejected": 0.02304188162088394,
"step": 570
},
{
"dpo_losses": 0.676629900932312,
"epoch": 0.85,
"grad_norm": 8.907703588890858,
"learning_rate": 3.44577981244944e-08,
"logits/chosen": -2.8228065967559814,
"logits/rejected": -2.809900999069214,
"logps/chosen": -257.9283752441406,
"logps/rejected": -256.25799560546875,
"loss": 0.6858,
"positive_losses": 0.03807668760418892,
"rewards/accuracies": 0.625,
"rewards/chosen": 0.05910138040781021,
"rewards/margins": 0.034723442047834396,
"rewards/margins_max": 0.12428979575634003,
"rewards/margins_min": -0.03289930149912834,
"rewards/margins_std": 0.07077684998512268,
"rewards/rejected": 0.024377940222620964,
"step": 580
},
{
"dpo_losses": 0.6756445169448853,
"epoch": 0.86,
"grad_norm": 8.494851741114475,
"learning_rate": 2.8271866952734814e-08,
"logits/chosen": -2.8694353103637695,
"logits/rejected": -2.844571113586426,
"logps/chosen": -302.86883544921875,
"logps/rejected": -284.4591369628906,
"loss": 0.6829,
"positive_losses": 0.07926483452320099,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": 0.05832204222679138,
"rewards/margins": 0.03624555841088295,
"rewards/margins_max": 0.10978803783655167,
"rewards/margins_min": -0.017319655045866966,
"rewards/margins_std": 0.05776820331811905,
"rewards/rejected": 0.022076481953263283,
"step": 590
},
{
"dpo_losses": 0.6685220003128052,
"epoch": 0.88,
"grad_norm": 2.0828943807385567,
"learning_rate": 2.2664399163518782e-08,
"logits/chosen": -2.7986605167388916,
"logits/rejected": -2.781226396560669,
"logps/chosen": -289.7279968261719,
"logps/rejected": -246.7861328125,
"loss": 0.67,
"positive_losses": 0.0,
"rewards/accuracies": 0.762499988079071,
"rewards/chosen": 0.07399223744869232,
"rewards/margins": 0.051160674542188644,
"rewards/margins_max": 0.12970498204231262,
"rewards/margins_min": -0.008301710709929466,
"rewards/margins_std": 0.06261205673217773,
"rewards/rejected": 0.02283155545592308,
"step": 600
},
{
"epoch": 0.88,
"eval_dpo_losses": 0.6851370930671692,
"eval_logits/chosen": -2.798839569091797,
"eval_logits/rejected": -2.7602357864379883,
"eval_logps/chosen": -278.6050109863281,
"eval_logps/rejected": -254.26231384277344,
"eval_loss": 0.6902583241462708,
"eval_positive_losses": 0.03728185221552849,
"eval_rewards/accuracies": 0.6159999966621399,
"eval_rewards/chosen": 0.05988401547074318,
"eval_rewards/margins": 0.016717953607439995,
"eval_rewards/margins_max": 0.09297847747802734,
"eval_rewards/margins_min": -0.051056575030088425,
"eval_rewards/margins_std": 0.0480518564581871,
"eval_rewards/rejected": 0.043166060000658035,
"eval_runtime": 428.2792,
"eval_samples_per_second": 4.67,
"eval_steps_per_second": 0.292,
"step": 600
},
{
"dpo_losses": 0.6745679974555969,
"epoch": 0.89,
"grad_norm": 1.3564104662072647,
"learning_rate": 1.7650024000056414e-08,
"logits/chosen": -2.7833163738250732,
"logits/rejected": -2.7807528972625732,
"logps/chosen": -241.8382568359375,
"logps/rejected": -243.6163330078125,
"loss": 0.6742,
"positive_losses": 0.0,
"rewards/accuracies": 0.75,
"rewards/chosen": 0.06093451380729675,
"rewards/margins": 0.038374222815036774,
"rewards/margins_max": 0.10369594395160675,
"rewards/margins_min": -0.009305914863944054,
"rewards/margins_std": 0.05068794637918472,
"rewards/rejected": 0.022560294717550278,
"step": 610
},
{
"dpo_losses": 0.6677746772766113,
"epoch": 0.91,
"grad_norm": 7.979260503019614,
"learning_rate": 1.3241823394615437e-08,
"logits/chosen": -2.908095598220825,
"logits/rejected": -2.8685083389282227,
"logps/chosen": -304.2915954589844,
"logps/rejected": -286.26873779296875,
"loss": 0.675,
"positive_losses": 0.05257830768823624,
"rewards/accuracies": 0.8500000238418579,
"rewards/chosen": 0.07303957641124725,
"rewards/margins": 0.05244825407862663,
"rewards/margins_max": 0.11377520859241486,
"rewards/margins_min": -0.004015705548226833,
"rewards/margins_std": 0.0548991933465004,
"rewards/rejected": 0.020591318607330322,
"step": 620
},
{
"dpo_losses": 0.6700986623764038,
"epoch": 0.92,
"grad_norm": 8.477082267330537,
"learning_rate": 9.451297839253913e-09,
"logits/chosen": -2.7833220958709717,
"logits/rejected": -2.7590456008911133,
"logps/chosen": -276.22540283203125,
"logps/rejected": -278.2480163574219,
"loss": 0.6748,
"positive_losses": 0.00507011404260993,
"rewards/accuracies": 0.762499988079071,
"rewards/chosen": 0.06989948451519012,
"rewards/margins": 0.047721296548843384,
"rewards/margins_max": 0.11920352280139923,
"rewards/margins_min": -0.008280943147838116,
"rewards/margins_std": 0.05656985193490982,
"rewards/rejected": 0.02217818982899189,
"step": 630
},
{
"dpo_losses": 0.6701762080192566,
"epoch": 0.94,
"grad_norm": 1.4912730505697442,
"learning_rate": 6.288336382349463e-09,
"logits/chosen": -2.7986907958984375,
"logits/rejected": -2.7656140327453613,
"logps/chosen": -245.1352996826172,
"logps/rejected": -207.5049285888672,
"loss": 0.6724,
"positive_losses": 0.004144692327827215,
"rewards/accuracies": 0.800000011920929,
"rewards/chosen": 0.06695154309272766,
"rewards/margins": 0.04747752100229263,
"rewards/margins_max": 0.12407982349395752,
"rewards/margins_min": -0.003674466861411929,
"rewards/margins_std": 0.05860821157693863,
"rewards/rejected": 0.01947401836514473,
"step": 640
},
{
"dpo_losses": 0.6735118627548218,
"epoch": 0.95,
"grad_norm": 10.343240316228654,
"learning_rate": 3.7611908292010665e-09,
"logits/chosen": -2.859143018722534,
"logits/rejected": -2.8227005004882812,
"logps/chosen": -307.39508056640625,
"logps/rejected": -258.0931091308594,
"loss": 0.6774,
"positive_losses": 0.10746382176876068,
"rewards/accuracies": 0.7124999761581421,
"rewards/chosen": 0.05885057523846626,
"rewards/margins": 0.040867820382118225,
"rewards/margins_max": 0.11666282266378403,
"rewards/margins_min": -0.015247263014316559,
"rewards/margins_std": 0.05968106538057327,
"rewards/rejected": 0.017982754856348038,
"step": 650
},
{
"dpo_losses": 0.6789201498031616,
"epoch": 0.96,
"grad_norm": 12.216425135481341,
"learning_rate": 1.8764542140112527e-09,
"logits/chosen": -2.800410509109497,
"logits/rejected": -2.8051278591156006,
"logps/chosen": -219.88369750976562,
"logps/rejected": -224.5011749267578,
"loss": 0.6838,
"positive_losses": 0.09240980446338654,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": 0.0470733642578125,
"rewards/margins": 0.02947625145316124,
"rewards/margins_max": 0.0936986654996872,
"rewards/margins_min": -0.014550586231052876,
"rewards/margins_std": 0.048393599689006805,
"rewards/rejected": 0.01759711280465126,
"step": 660
},
{
"dpo_losses": 0.6718948483467102,
"epoch": 0.98,
"grad_norm": 7.347929588964824,
"learning_rate": 6.390435994127752e-10,
"logits/chosen": -2.8291993141174316,
"logits/rejected": -2.7946581840515137,
"logps/chosen": -279.2850646972656,
"logps/rejected": -295.1185607910156,
"loss": 0.6767,
"positive_losses": 0.03623781353235245,
"rewards/accuracies": 0.762499988079071,
"rewards/chosen": 0.06117644160985947,
"rewards/margins": 0.044048137962818146,
"rewards/margins_max": 0.11197330802679062,
"rewards/margins_min": -0.004027059301733971,
"rewards/margins_std": 0.05286269634962082,
"rewards/rejected": 0.01712830550968647,
"step": 670
},
{
"dpo_losses": 0.6698868870735168,
"epoch": 0.99,
"grad_norm": 9.194005207225429,
"learning_rate": 5.2187248413465555e-11,
"logits/chosen": -2.8184335231781006,
"logits/rejected": -2.7909679412841797,
"logps/chosen": -296.29473876953125,
"logps/rejected": -258.4402160644531,
"loss": 0.6757,
"positive_losses": 0.03602256625890732,
"rewards/accuracies": 0.75,
"rewards/chosen": 0.06865200400352478,
"rewards/margins": 0.04830945283174515,
"rewards/margins_max": 0.1308433711528778,
"rewards/margins_min": -0.012822682037949562,
"rewards/margins_std": 0.06532330065965652,
"rewards/rejected": 0.020342541858553886,
"step": 680
},
{
"epoch": 1.0,
"step": 684,
"total_flos": 0.0,
"train_loss": 0.6833745032026056,
"train_runtime": 9125.7542,
"train_samples_per_second": 1.199,
"train_steps_per_second": 0.075
}
],
"logging_steps": 10,
"max_steps": 684,
"num_input_tokens_seen": 0,
"num_train_epochs": 1,
"save_steps": 100,
"total_flos": 0.0,
"train_batch_size": 4,
"trial_name": null,
"trial_params": null
}