k-r-l's picture
Training in progress, step 62, checkpoint
5593a5a verified
raw
history blame
41.2 kB
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 0.6318471337579618,
"eval_steps": 500,
"global_step": 62,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.01019108280254777,
"grad_norm": 1.9706840515136719,
"learning_rate": 8e-05,
"log_odds_chosen": 0.043639570474624634,
"log_odds_ratio": -0.6778613328933716,
"logits/chosen": -0.5009905099868774,
"logits/rejected": -0.5356172919273376,
"logps/chosen": -2.954733371734619,
"logps/rejected": -2.996232509613037,
"loss": 3.6057,
"nll_loss": 3.5378708839416504,
"rewards/accuracies": 0.5625,
"rewards/chosen": -0.2954733371734619,
"rewards/margins": 0.00414988212287426,
"rewards/rejected": -0.2996232211589813,
"step": 1
},
{
"epoch": 0.02038216560509554,
"grad_norm": 1.9629229307174683,
"learning_rate": 6e-05,
"log_odds_chosen": 0.10935741662979126,
"log_odds_ratio": -0.6454683542251587,
"logits/chosen": -0.5241304636001587,
"logits/rejected": -0.5285190939903259,
"logps/chosen": -2.8890933990478516,
"logps/rejected": -2.9919257164001465,
"loss": 3.4851,
"nll_loss": 3.420555830001831,
"rewards/accuracies": 0.71875,
"rewards/chosen": -0.28890931606292725,
"rewards/margins": 0.010283238254487514,
"rewards/rejected": -0.2991925776004791,
"step": 2
},
{
"epoch": 0.030573248407643312,
"grad_norm": 2.020477533340454,
"learning_rate": 4e-05,
"log_odds_chosen": 0.069330133497715,
"log_odds_ratio": -0.6641735434532166,
"logits/chosen": -0.4971535801887512,
"logits/rejected": -0.5353123545646667,
"logps/chosen": -2.731393575668335,
"logps/rejected": -2.795962333679199,
"loss": 3.3144,
"nll_loss": 3.2479400634765625,
"rewards/accuracies": 0.5625,
"rewards/chosen": -0.2731393277645111,
"rewards/margins": 0.006456875707954168,
"rewards/rejected": -0.279596209526062,
"step": 3
},
{
"epoch": 0.04076433121019108,
"grad_norm": 2.3341853618621826,
"learning_rate": 2e-05,
"log_odds_chosen": 0.12229391932487488,
"log_odds_ratio": -0.6381683945655823,
"logits/chosen": -0.4815675914287567,
"logits/rejected": -0.5111523866653442,
"logps/chosen": -2.8219356536865234,
"logps/rejected": -2.9366464614868164,
"loss": 3.3456,
"nll_loss": 3.281827211380005,
"rewards/accuracies": 0.75,
"rewards/chosen": -0.28219354152679443,
"rewards/margins": 0.011471111327409744,
"rewards/rejected": -0.2936646640300751,
"step": 4
},
{
"epoch": 0.050955414012738856,
"grad_norm": 2.1892218589782715,
"learning_rate": 0.0,
"log_odds_chosen": 0.06512074917554855,
"log_odds_ratio": -0.6671069860458374,
"logits/chosen": -0.5229994654655457,
"logits/rejected": -0.5436381697654724,
"logps/chosen": -2.7552542686462402,
"logps/rejected": -2.816582679748535,
"loss": 3.2584,
"nll_loss": 3.191640615463257,
"rewards/accuracies": 0.625,
"rewards/chosen": -0.27552542090415955,
"rewards/margins": 0.006132860202342272,
"rewards/rejected": -0.2816582918167114,
"step": 5
},
{
"epoch": 0.061146496815286625,
"grad_norm": 1.755154013633728,
"learning_rate": 9.79591836734694e-05,
"log_odds_chosen": 0.09281320869922638,
"log_odds_ratio": -0.6549628973007202,
"logits/chosen": -0.5153884291648865,
"logits/rejected": -0.5407605767250061,
"logps/chosen": -2.8005495071411133,
"logps/rejected": -2.8862051963806152,
"loss": 3.2371,
"nll_loss": 3.171649217605591,
"rewards/accuracies": 0.625,
"rewards/chosen": -0.2800549566745758,
"rewards/margins": 0.008565555326640606,
"rewards/rejected": -0.2886205315589905,
"step": 6
},
{
"epoch": 0.07133757961783439,
"grad_norm": 1.8428237438201904,
"learning_rate": 9.761904761904762e-05,
"log_odds_chosen": 0.03281838819384575,
"log_odds_ratio": -0.6809508800506592,
"logits/chosen": -0.5399191975593567,
"logits/rejected": -0.5274120569229126,
"logps/chosen": -2.9202024936676025,
"logps/rejected": -2.9510116577148438,
"loss": 3.3829,
"nll_loss": 3.314777374267578,
"rewards/accuracies": 0.59375,
"rewards/chosen": -0.2920202612876892,
"rewards/margins": 0.003080902621150017,
"rewards/rejected": -0.2951011657714844,
"step": 7
},
{
"epoch": 0.08152866242038216,
"grad_norm": 1.8741809129714966,
"learning_rate": 9.727891156462585e-05,
"log_odds_chosen": 0.09863043576478958,
"log_odds_ratio": -0.6496734619140625,
"logits/chosen": -0.5901607275009155,
"logits/rejected": -0.5661267042160034,
"logps/chosen": -2.7296760082244873,
"logps/rejected": -2.8212223052978516,
"loss": 3.1928,
"nll_loss": 3.1278655529022217,
"rewards/accuracies": 0.65625,
"rewards/chosen": -0.2729676365852356,
"rewards/margins": 0.009154656901955605,
"rewards/rejected": -0.28212225437164307,
"step": 8
},
{
"epoch": 0.09171974522292993,
"grad_norm": 1.9783447980880737,
"learning_rate": 9.693877551020408e-05,
"log_odds_chosen": 0.05389028042554855,
"log_odds_ratio": -0.6711492538452148,
"logits/chosen": -0.5444136261940002,
"logits/rejected": -0.564105212688446,
"logps/chosen": -2.688671112060547,
"logps/rejected": -2.7393460273742676,
"loss": 3.1182,
"nll_loss": 3.051093101501465,
"rewards/accuracies": 0.59375,
"rewards/chosen": -0.2688670754432678,
"rewards/margins": 0.005067505408078432,
"rewards/rejected": -0.27393460273742676,
"step": 9
},
{
"epoch": 0.10191082802547771,
"grad_norm": 3.7560782432556152,
"learning_rate": 9.659863945578231e-05,
"log_odds_chosen": 0.05389935150742531,
"log_odds_ratio": -0.6712656617164612,
"logits/chosen": -0.5685052871704102,
"logits/rejected": -0.5835959911346436,
"logps/chosen": -2.7232255935668945,
"logps/rejected": -2.772873878479004,
"loss": 3.1185,
"nll_loss": 3.0513598918914795,
"rewards/accuracies": 0.65625,
"rewards/chosen": -0.27232253551483154,
"rewards/margins": 0.004964826162904501,
"rewards/rejected": -0.27728739380836487,
"step": 10
},
{
"epoch": 0.11210191082802548,
"grad_norm": 2.745220422744751,
"learning_rate": 9.625850340136054e-05,
"log_odds_chosen": 0.0562276728451252,
"log_odds_ratio": -0.6704986095428467,
"logits/chosen": -0.5921126008033752,
"logits/rejected": -0.5893128514289856,
"logps/chosen": -2.645206928253174,
"logps/rejected": -2.696500539779663,
"loss": 3.0034,
"nll_loss": 2.936365842819214,
"rewards/accuracies": 0.65625,
"rewards/chosen": -0.26452070474624634,
"rewards/margins": 0.005129385739564896,
"rewards/rejected": -0.26965007185935974,
"step": 11
},
{
"epoch": 0.12229299363057325,
"grad_norm": 1.8458824157714844,
"learning_rate": 9.591836734693878e-05,
"log_odds_chosen": -0.026220712810754776,
"log_odds_ratio": -0.7125887870788574,
"logits/chosen": -0.5802856087684631,
"logits/rejected": -0.6202648878097534,
"logps/chosen": -2.6887969970703125,
"logps/rejected": -2.6625289916992188,
"loss": 3.0166,
"nll_loss": 2.945356845855713,
"rewards/accuracies": 0.40625,
"rewards/chosen": -0.2688797116279602,
"rewards/margins": -0.0026267748326063156,
"rewards/rejected": -0.26625293493270874,
"step": 12
},
{
"epoch": 0.13248407643312102,
"grad_norm": 1.8597255945205688,
"learning_rate": 9.557823129251701e-05,
"log_odds_chosen": 0.0966324731707573,
"log_odds_ratio": -0.6508103013038635,
"logits/chosen": -0.5836719274520874,
"logits/rejected": -0.6042653322219849,
"logps/chosen": -2.5945496559143066,
"logps/rejected": -2.6839816570281982,
"loss": 2.891,
"nll_loss": 2.8259291648864746,
"rewards/accuracies": 0.6875,
"rewards/chosen": -0.25945499539375305,
"rewards/margins": 0.008943157270550728,
"rewards/rejected": -0.26839813590049744,
"step": 13
},
{
"epoch": 0.14267515923566879,
"grad_norm": 1.9950586557388306,
"learning_rate": 9.523809523809524e-05,
"log_odds_chosen": 0.07333967834711075,
"log_odds_ratio": -0.6641579270362854,
"logits/chosen": -0.5741627812385559,
"logits/rejected": -0.6016291379928589,
"logps/chosen": -2.556469678878784,
"logps/rejected": -2.6230132579803467,
"loss": 2.8169,
"nll_loss": 2.7504758834838867,
"rewards/accuracies": 0.625,
"rewards/chosen": -0.2556469738483429,
"rewards/margins": 0.006654377095401287,
"rewards/rejected": -0.26230132579803467,
"step": 14
},
{
"epoch": 0.15286624203821655,
"grad_norm": 2.0696280002593994,
"learning_rate": 9.489795918367348e-05,
"log_odds_chosen": 0.12293906509876251,
"log_odds_ratio": -0.6370925903320312,
"logits/chosen": -0.5484431385993958,
"logits/rejected": -0.5725057721138,
"logps/chosen": -2.460176944732666,
"logps/rejected": -2.5726397037506104,
"loss": 2.7249,
"nll_loss": 2.661222219467163,
"rewards/accuracies": 0.78125,
"rewards/chosen": -0.2460177093744278,
"rewards/margins": 0.011246275156736374,
"rewards/rejected": -0.25726398825645447,
"step": 15
},
{
"epoch": 0.16305732484076432,
"grad_norm": 3.213852882385254,
"learning_rate": 9.455782312925171e-05,
"log_odds_chosen": 0.0818968415260315,
"log_odds_ratio": -0.6613866090774536,
"logits/chosen": -0.5976958274841309,
"logits/rejected": -0.6143887639045715,
"logps/chosen": -2.413341522216797,
"logps/rejected": -2.4865636825561523,
"loss": 2.6428,
"nll_loss": 2.5766918659210205,
"rewards/accuracies": 0.5,
"rewards/chosen": -0.24133417010307312,
"rewards/margins": 0.007322215940803289,
"rewards/rejected": -0.24865636229515076,
"step": 16
},
{
"epoch": 0.1732484076433121,
"grad_norm": 3.263835906982422,
"learning_rate": 9.421768707482994e-05,
"log_odds_chosen": 0.12279009073972702,
"log_odds_ratio": -0.6375764608383179,
"logits/chosen": -0.5939255952835083,
"logits/rejected": -0.5961549878120422,
"logps/chosen": -2.3610243797302246,
"logps/rejected": -2.47149395942688,
"loss": 2.568,
"nll_loss": 2.5042312145233154,
"rewards/accuracies": 0.71875,
"rewards/chosen": -0.23610244691371918,
"rewards/margins": 0.011046944186091423,
"rewards/rejected": -0.24714939296245575,
"step": 17
},
{
"epoch": 0.18343949044585986,
"grad_norm": 2.0416226387023926,
"learning_rate": 9.387755102040817e-05,
"log_odds_chosen": 0.1555405557155609,
"log_odds_ratio": -0.6227366328239441,
"logits/chosen": -0.586774468421936,
"logits/rejected": -0.6310256123542786,
"logps/chosen": -2.2520110607147217,
"logps/rejected": -2.390634775161743,
"loss": 2.4749,
"nll_loss": 2.412618637084961,
"rewards/accuracies": 0.84375,
"rewards/chosen": -0.22520111501216888,
"rewards/margins": 0.013862377032637596,
"rewards/rejected": -0.23906347155570984,
"step": 18
},
{
"epoch": 0.19363057324840766,
"grad_norm": 1.8552919626235962,
"learning_rate": 9.35374149659864e-05,
"log_odds_chosen": 0.1357956975698471,
"log_odds_ratio": -0.6316878199577332,
"logits/chosen": -0.6498819589614868,
"logits/rejected": -0.6668039560317993,
"logps/chosen": -2.22139835357666,
"logps/rejected": -2.3428542613983154,
"loss": 2.3994,
"nll_loss": 2.336277723312378,
"rewards/accuracies": 0.75,
"rewards/chosen": -0.22213983535766602,
"rewards/margins": 0.012145599350333214,
"rewards/rejected": -0.23428542912006378,
"step": 19
},
{
"epoch": 0.20382165605095542,
"grad_norm": 1.9783782958984375,
"learning_rate": 9.319727891156463e-05,
"log_odds_chosen": 0.09560593217611313,
"log_odds_ratio": -0.6527409553527832,
"logits/chosen": -0.6574750542640686,
"logits/rejected": -0.6695252656936646,
"logps/chosen": -2.228286027908325,
"logps/rejected": -2.3149209022521973,
"loss": 2.4197,
"nll_loss": 2.3543879985809326,
"rewards/accuracies": 0.6875,
"rewards/chosen": -0.22282862663269043,
"rewards/margins": 0.008663482964038849,
"rewards/rejected": -0.23149211704730988,
"step": 20
},
{
"epoch": 0.2140127388535032,
"grad_norm": 2.270171880722046,
"learning_rate": 9.285714285714286e-05,
"log_odds_chosen": 0.13292132318019867,
"log_odds_ratio": -0.6331073641777039,
"logits/chosen": -0.6325170993804932,
"logits/rejected": -0.6448156237602234,
"logps/chosen": -2.0987019538879395,
"logps/rejected": -2.2161264419555664,
"loss": 2.2766,
"nll_loss": 2.2132651805877686,
"rewards/accuracies": 0.84375,
"rewards/chosen": -0.20987018942832947,
"rewards/margins": 0.01174245961010456,
"rewards/rejected": -0.22161263227462769,
"step": 21
},
{
"epoch": 0.22420382165605096,
"grad_norm": 2.6400585174560547,
"learning_rate": 9.25170068027211e-05,
"log_odds_chosen": 0.0851236879825592,
"log_odds_ratio": -0.6567416191101074,
"logits/chosen": -0.6978996992111206,
"logits/rejected": -0.7123661041259766,
"logps/chosen": -2.0720839500427246,
"logps/rejected": -2.1482036113739014,
"loss": 2.2512,
"nll_loss": 2.1855506896972656,
"rewards/accuracies": 0.6875,
"rewards/chosen": -0.20720840990543365,
"rewards/margins": 0.007611965760588646,
"rewards/rejected": -0.21482038497924805,
"step": 22
},
{
"epoch": 0.23439490445859873,
"grad_norm": 3.0456278324127197,
"learning_rate": 9.217687074829933e-05,
"log_odds_chosen": 0.12218278646469116,
"log_odds_ratio": -0.639614462852478,
"logits/chosen": -0.6912589073181152,
"logits/rejected": -0.7216678261756897,
"logps/chosen": -1.9978296756744385,
"logps/rejected": -2.103085517883301,
"loss": 2.1711,
"nll_loss": 2.1071152687072754,
"rewards/accuracies": 0.6875,
"rewards/chosen": -0.19978296756744385,
"rewards/margins": 0.010525593534111977,
"rewards/rejected": -0.21030855178833008,
"step": 23
},
{
"epoch": 0.2445859872611465,
"grad_norm": 2.3382904529571533,
"learning_rate": 9.183673469387756e-05,
"log_odds_chosen": 0.043098174035549164,
"log_odds_ratio": -0.675950825214386,
"logits/chosen": -0.7050585746765137,
"logits/rejected": -0.6946679949760437,
"logps/chosen": -1.8801963329315186,
"logps/rejected": -1.9172497987747192,
"loss": 2.0647,
"nll_loss": 1.9970893859863281,
"rewards/accuracies": 0.53125,
"rewards/chosen": -0.18801963329315186,
"rewards/margins": 0.003705363254994154,
"rewards/rejected": -0.1917250007390976,
"step": 24
},
{
"epoch": 0.25477707006369427,
"grad_norm": 2.2423338890075684,
"learning_rate": 9.149659863945579e-05,
"log_odds_chosen": 0.051533281803131104,
"log_odds_ratio": -0.6742590665817261,
"logits/chosen": -0.7378531694412231,
"logits/rejected": -0.7483745217323303,
"logps/chosen": -1.7976934909820557,
"logps/rejected": -1.8392703533172607,
"loss": 1.9856,
"nll_loss": 1.9182195663452148,
"rewards/accuracies": 0.5625,
"rewards/chosen": -0.1797693371772766,
"rewards/margins": 0.004157697316259146,
"rewards/rejected": -0.1839270442724228,
"step": 25
},
{
"epoch": 0.26496815286624203,
"grad_norm": 2.2076170444488525,
"learning_rate": 9.115646258503402e-05,
"log_odds_chosen": 0.14921554923057556,
"log_odds_ratio": -0.6260839700698853,
"logits/chosen": -0.7678406834602356,
"logits/rejected": -0.7855570316314697,
"logps/chosen": -1.690258264541626,
"logps/rejected": -1.813413143157959,
"loss": 1.8961,
"nll_loss": 1.833521842956543,
"rewards/accuracies": 0.78125,
"rewards/chosen": -0.16902580857276917,
"rewards/margins": 0.012315504252910614,
"rewards/rejected": -0.18134130537509918,
"step": 26
},
{
"epoch": 0.2751592356687898,
"grad_norm": 2.486793279647827,
"learning_rate": 9.081632653061225e-05,
"log_odds_chosen": 0.15732049942016602,
"log_odds_ratio": -0.6271807551383972,
"logits/chosen": -0.7320041060447693,
"logits/rejected": -0.7608926296234131,
"logps/chosen": -1.6371859312057495,
"logps/rejected": -1.7642821073532104,
"loss": 1.8287,
"nll_loss": 1.7660211324691772,
"rewards/accuracies": 0.6875,
"rewards/chosen": -0.163718581199646,
"rewards/margins": 0.01270962879061699,
"rewards/rejected": -0.17642822861671448,
"step": 27
},
{
"epoch": 0.28535031847133757,
"grad_norm": 2.56262469291687,
"learning_rate": 9.047619047619048e-05,
"log_odds_chosen": 0.16445228457450867,
"log_odds_ratio": -0.6198588609695435,
"logits/chosen": -0.7562386989593506,
"logits/rejected": -0.7781034708023071,
"logps/chosen": -1.5056514739990234,
"logps/rejected": -1.634886622428894,
"loss": 1.7373,
"nll_loss": 1.6753149032592773,
"rewards/accuracies": 0.6875,
"rewards/chosen": -0.15056513249874115,
"rewards/margins": 0.012923511676490307,
"rewards/rejected": -0.16348864138126373,
"step": 28
},
{
"epoch": 0.29554140127388534,
"grad_norm": 2.2155702114105225,
"learning_rate": 9.01360544217687e-05,
"log_odds_chosen": 0.07792021334171295,
"log_odds_ratio": -0.6598246693611145,
"logits/chosen": -0.7097214460372925,
"logits/rejected": -0.7517991065979004,
"logps/chosen": -1.4955050945281982,
"logps/rejected": -1.5548988580703735,
"loss": 1.7355,
"nll_loss": 1.669482707977295,
"rewards/accuracies": 0.71875,
"rewards/chosen": -0.14955051243305206,
"rewards/margins": 0.005939368158578873,
"rewards/rejected": -0.15548989176750183,
"step": 29
},
{
"epoch": 0.3057324840764331,
"grad_norm": 2.1764378547668457,
"learning_rate": 8.979591836734695e-05,
"log_odds_chosen": 0.03990812599658966,
"log_odds_ratio": -0.6799441576004028,
"logits/chosen": -0.7283363938331604,
"logits/rejected": -0.7530458569526672,
"logps/chosen": -1.4264872074127197,
"logps/rejected": -1.4550918340682983,
"loss": 1.6459,
"nll_loss": 1.5778682231903076,
"rewards/accuracies": 0.625,
"rewards/chosen": -0.14264871180057526,
"rewards/margins": 0.0028604594990611076,
"rewards/rejected": -0.14550916850566864,
"step": 30
},
{
"epoch": 0.3159235668789809,
"grad_norm": 2.089293956756592,
"learning_rate": 8.945578231292518e-05,
"log_odds_chosen": 0.2132435142993927,
"log_odds_ratio": -0.5990588068962097,
"logits/chosen": -0.7453004121780396,
"logits/rejected": -0.7864771485328674,
"logps/chosen": -1.330241322517395,
"logps/rejected": -1.4900933504104614,
"loss": 1.5747,
"nll_loss": 1.5148346424102783,
"rewards/accuracies": 0.75,
"rewards/chosen": -0.1330241560935974,
"rewards/margins": 0.01598520018160343,
"rewards/rejected": -0.1490093469619751,
"step": 31
},
{
"epoch": 0.32611464968152865,
"grad_norm": 1.9524860382080078,
"learning_rate": 8.911564625850341e-05,
"log_odds_chosen": 0.07104514539241791,
"log_odds_ratio": -0.6634508371353149,
"logits/chosen": -0.7562890648841858,
"logits/rejected": -0.762592077255249,
"logps/chosen": -1.3261951208114624,
"logps/rejected": -1.3763710260391235,
"loss": 1.5646,
"nll_loss": 1.4982781410217285,
"rewards/accuracies": 0.625,
"rewards/chosen": -0.13261951506137848,
"rewards/margins": 0.005017599556595087,
"rewards/rejected": -0.13763710856437683,
"step": 32
},
{
"epoch": 0.3363057324840764,
"grad_norm": 1.859688401222229,
"learning_rate": 8.877551020408164e-05,
"log_odds_chosen": 0.2007061243057251,
"log_odds_ratio": -0.6077967882156372,
"logits/chosen": -0.7153605222702026,
"logits/rejected": -0.7660464644432068,
"logps/chosen": -1.1958274841308594,
"logps/rejected": -1.3376389741897583,
"loss": 1.473,
"nll_loss": 1.4121888875961304,
"rewards/accuracies": 0.71875,
"rewards/chosen": -0.11958275735378265,
"rewards/margins": 0.01418115571141243,
"rewards/rejected": -0.13376390933990479,
"step": 33
},
{
"epoch": 0.3464968152866242,
"grad_norm": 1.627384901046753,
"learning_rate": 8.843537414965987e-05,
"log_odds_chosen": 0.20318442583084106,
"log_odds_ratio": -0.60548335313797,
"logits/chosen": -0.6687170267105103,
"logits/rejected": -0.6979736089706421,
"logps/chosen": -1.1998878717422485,
"logps/rejected": -1.3429744243621826,
"loss": 1.4198,
"nll_loss": 1.3592194318771362,
"rewards/accuracies": 0.75,
"rewards/chosen": -0.11998879909515381,
"rewards/margins": 0.014308637008070946,
"rewards/rejected": -0.1342974454164505,
"step": 34
},
{
"epoch": 0.35668789808917195,
"grad_norm": 1.6395893096923828,
"learning_rate": 8.80952380952381e-05,
"log_odds_chosen": 0.1479332149028778,
"log_odds_ratio": -0.6287116408348083,
"logits/chosen": -0.6793037056922913,
"logits/rejected": -0.7141229510307312,
"logps/chosen": -1.1636970043182373,
"logps/rejected": -1.2673585414886475,
"loss": 1.3737,
"nll_loss": 1.310793161392212,
"rewards/accuracies": 0.6875,
"rewards/chosen": -0.11636970937252045,
"rewards/margins": 0.01036614365875721,
"rewards/rejected": -0.1267358511686325,
"step": 35
},
{
"epoch": 0.3668789808917197,
"grad_norm": 1.3266431093215942,
"learning_rate": 8.775510204081632e-05,
"log_odds_chosen": 0.16748954355716705,
"log_odds_ratio": -0.6182947754859924,
"logits/chosen": -0.6686299443244934,
"logits/rejected": -0.6554075479507446,
"logps/chosen": -1.0804071426391602,
"logps/rejected": -1.1927094459533691,
"loss": 1.328,
"nll_loss": 1.2661728858947754,
"rewards/accuracies": 0.75,
"rewards/chosen": -0.1080407053232193,
"rewards/margins": 0.011230221018195152,
"rewards/rejected": -0.1192709356546402,
"step": 36
},
{
"epoch": 0.37707006369426754,
"grad_norm": 1.6111111640930176,
"learning_rate": 8.741496598639455e-05,
"log_odds_chosen": 0.09395941346883774,
"log_odds_ratio": -0.6549113988876343,
"logits/chosen": -0.7055931091308594,
"logits/rejected": -0.6843895316123962,
"logps/chosen": -1.123265027999878,
"logps/rejected": -1.1842732429504395,
"loss": 1.3436,
"nll_loss": 1.2781226634979248,
"rewards/accuracies": 0.59375,
"rewards/chosen": -0.1123264953494072,
"rewards/margins": 0.006100829225033522,
"rewards/rejected": -0.1184273213148117,
"step": 37
},
{
"epoch": 0.3872611464968153,
"grad_norm": 1.3853479623794556,
"learning_rate": 8.70748299319728e-05,
"log_odds_chosen": 0.22841353714466095,
"log_odds_ratio": -0.5929134488105774,
"logits/chosen": -0.6214150786399841,
"logits/rejected": -0.6320711374282837,
"logps/chosen": -0.9967460632324219,
"logps/rejected": -1.1483112573623657,
"loss": 1.239,
"nll_loss": 1.179752230644226,
"rewards/accuracies": 0.84375,
"rewards/chosen": -0.09967461228370667,
"rewards/margins": 0.01515651773661375,
"rewards/rejected": -0.11483113467693329,
"step": 38
},
{
"epoch": 0.3974522292993631,
"grad_norm": 1.4109987020492554,
"learning_rate": 8.673469387755102e-05,
"log_odds_chosen": 0.15123127400875092,
"log_odds_ratio": -0.6313170194625854,
"logits/chosen": -0.5636093020439148,
"logits/rejected": -0.5879440903663635,
"logps/chosen": -0.9781452417373657,
"logps/rejected": -1.0714449882507324,
"loss": 1.2294,
"nll_loss": 1.1662321090698242,
"rewards/accuracies": 0.65625,
"rewards/chosen": -0.09781452268362045,
"rewards/margins": 0.00932997465133667,
"rewards/rejected": -0.10714450478553772,
"step": 39
},
{
"epoch": 0.40764331210191085,
"grad_norm": 1.505708932876587,
"learning_rate": 8.639455782312925e-05,
"log_odds_chosen": 0.15932558476924896,
"log_odds_ratio": -0.625272274017334,
"logits/chosen": -0.5791752934455872,
"logits/rejected": -0.6104503870010376,
"logps/chosen": -0.9371697902679443,
"logps/rejected": -1.0375370979309082,
"loss": 1.1666,
"nll_loss": 1.104070782661438,
"rewards/accuracies": 0.71875,
"rewards/chosen": -0.09371697902679443,
"rewards/margins": 0.010036729276180267,
"rewards/rejected": -0.1037537157535553,
"step": 40
},
{
"epoch": 0.4178343949044586,
"grad_norm": 1.3721827268600464,
"learning_rate": 8.605442176870748e-05,
"log_odds_chosen": 0.15735837817192078,
"log_odds_ratio": -0.6272850036621094,
"logits/chosen": -0.5897191762924194,
"logits/rejected": -0.5843300819396973,
"logps/chosen": -0.9254653453826904,
"logps/rejected": -1.0147194862365723,
"loss": 1.1898,
"nll_loss": 1.127084732055664,
"rewards/accuracies": 0.75,
"rewards/chosen": -0.09254653751850128,
"rewards/margins": 0.008925420232117176,
"rewards/rejected": -0.10147196054458618,
"step": 41
},
{
"epoch": 0.4280254777070064,
"grad_norm": 1.4123814105987549,
"learning_rate": 8.571428571428571e-05,
"log_odds_chosen": 0.1627001166343689,
"log_odds_ratio": -0.6306536197662354,
"logits/chosen": -0.5717113614082336,
"logits/rejected": -0.5668535232543945,
"logps/chosen": -0.877647876739502,
"logps/rejected": -0.9752596020698547,
"loss": 1.1148,
"nll_loss": 1.0516966581344604,
"rewards/accuracies": 0.71875,
"rewards/chosen": -0.08776478469371796,
"rewards/margins": 0.00976115744560957,
"rewards/rejected": -0.0975259393453598,
"step": 42
},
{
"epoch": 0.43821656050955415,
"grad_norm": 1.4072459936141968,
"learning_rate": 8.537414965986394e-05,
"log_odds_chosen": 0.26022908091545105,
"log_odds_ratio": -0.5852185487747192,
"logits/chosen": -0.5471714735031128,
"logits/rejected": -0.5991571545600891,
"logps/chosen": -0.8360999822616577,
"logps/rejected": -0.9906545281410217,
"loss": 1.0647,
"nll_loss": 1.0061821937561035,
"rewards/accuracies": 0.71875,
"rewards/chosen": -0.08360999822616577,
"rewards/margins": 0.015455447137355804,
"rewards/rejected": -0.09906545281410217,
"step": 43
},
{
"epoch": 0.4484076433121019,
"grad_norm": 1.263767957687378,
"learning_rate": 8.503401360544217e-05,
"log_odds_chosen": 0.15025389194488525,
"log_odds_ratio": -0.6313925981521606,
"logits/chosen": -0.5081976652145386,
"logits/rejected": -0.5151488184928894,
"logps/chosen": -0.860946774482727,
"logps/rejected": -0.9522109031677246,
"loss": 1.0628,
"nll_loss": 0.9996945858001709,
"rewards/accuracies": 0.6875,
"rewards/chosen": -0.0860946774482727,
"rewards/margins": 0.00912640430033207,
"rewards/rejected": -0.09522109478712082,
"step": 44
},
{
"epoch": 0.4585987261146497,
"grad_norm": 1.1564173698425293,
"learning_rate": 8.469387755102041e-05,
"log_odds_chosen": 0.25856250524520874,
"log_odds_ratio": -0.5882014632225037,
"logits/chosen": -0.47218459844589233,
"logits/rejected": -0.47766923904418945,
"logps/chosen": -0.786893904209137,
"logps/rejected": -0.9319732785224915,
"loss": 1.0045,
"nll_loss": 0.9457147121429443,
"rewards/accuracies": 0.78125,
"rewards/chosen": -0.07868938893079758,
"rewards/margins": 0.01450793445110321,
"rewards/rejected": -0.09319733083248138,
"step": 45
},
{
"epoch": 0.46878980891719746,
"grad_norm": 1.1476246118545532,
"learning_rate": 8.435374149659864e-05,
"log_odds_chosen": 0.1751195192337036,
"log_odds_ratio": -0.6188879013061523,
"logits/chosen": -0.4528760015964508,
"logits/rejected": -0.47072282433509827,
"logps/chosen": -0.7283643484115601,
"logps/rejected": -0.8211995363235474,
"loss": 0.9499,
"nll_loss": 0.8879653811454773,
"rewards/accuracies": 0.78125,
"rewards/chosen": -0.07283642888069153,
"rewards/margins": 0.009283517487347126,
"rewards/rejected": -0.08211994916200638,
"step": 46
},
{
"epoch": 0.4789808917197452,
"grad_norm": 1.082606554031372,
"learning_rate": 8.401360544217689e-05,
"log_odds_chosen": 0.239468514919281,
"log_odds_ratio": -0.5954229831695557,
"logits/chosen": -0.4875160753726959,
"logits/rejected": -0.48623812198638916,
"logps/chosen": -0.7523530125617981,
"logps/rejected": -0.877802848815918,
"loss": 0.954,
"nll_loss": 0.8944483399391174,
"rewards/accuracies": 0.78125,
"rewards/chosen": -0.07523529976606369,
"rewards/margins": 0.012544991448521614,
"rewards/rejected": -0.08778029680252075,
"step": 47
},
{
"epoch": 0.489171974522293,
"grad_norm": 1.1768563985824585,
"learning_rate": 8.367346938775511e-05,
"log_odds_chosen": 0.11137399077415466,
"log_odds_ratio": -0.6481903791427612,
"logits/chosen": -0.47852304577827454,
"logits/rejected": -0.49898433685302734,
"logps/chosen": -0.7479023933410645,
"logps/rejected": -0.8043348789215088,
"loss": 0.9558,
"nll_loss": 0.8910031914710999,
"rewards/accuracies": 0.59375,
"rewards/chosen": -0.07479023933410645,
"rewards/margins": 0.005643244832754135,
"rewards/rejected": -0.08043348044157028,
"step": 48
},
{
"epoch": 0.49936305732484076,
"grad_norm": 1.0716862678527832,
"learning_rate": 8.333333333333334e-05,
"log_odds_chosen": 0.20709852874279022,
"log_odds_ratio": -0.6083564758300781,
"logits/chosen": -0.4109807312488556,
"logits/rejected": -0.44194647669792175,
"logps/chosen": -0.7731195688247681,
"logps/rejected": -0.8942749500274658,
"loss": 0.9623,
"nll_loss": 0.9014848470687866,
"rewards/accuracies": 0.6875,
"rewards/chosen": -0.07731196284294128,
"rewards/margins": 0.012115536257624626,
"rewards/rejected": -0.08942750096321106,
"step": 49
},
{
"epoch": 0.5095541401273885,
"grad_norm": 0.9973980188369751,
"learning_rate": 8.299319727891157e-05,
"log_odds_chosen": 0.18566134572029114,
"log_odds_ratio": -0.61891108751297,
"logits/chosen": -0.4191117286682129,
"logits/rejected": -0.41958895325660706,
"logps/chosen": -0.702156662940979,
"logps/rejected": -0.7984734177589417,
"loss": 0.8786,
"nll_loss": 0.8166758418083191,
"rewards/accuracies": 0.71875,
"rewards/chosen": -0.07021567225456238,
"rewards/margins": 0.00963167380541563,
"rewards/rejected": -0.07984734326601028,
"step": 50
},
{
"epoch": 0.5197452229299363,
"grad_norm": 0.8124039769172668,
"learning_rate": 8.26530612244898e-05,
"log_odds_chosen": 0.18215714395046234,
"log_odds_ratio": -0.6213276386260986,
"logits/chosen": -0.4588465690612793,
"logits/rejected": -0.46019288897514343,
"logps/chosen": -0.7112851142883301,
"logps/rejected": -0.7996297478675842,
"loss": 0.8886,
"nll_loss": 0.8265119194984436,
"rewards/accuracies": 0.6875,
"rewards/chosen": -0.07112851738929749,
"rewards/margins": 0.008834462612867355,
"rewards/rejected": -0.07996297627687454,
"step": 51
},
{
"epoch": 0.5299363057324841,
"grad_norm": 0.777076780796051,
"learning_rate": 8.231292517006803e-05,
"log_odds_chosen": 0.2624254524707794,
"log_odds_ratio": -0.5843685269355774,
"logits/chosen": -0.3662329912185669,
"logits/rejected": -0.34878844022750854,
"logps/chosen": -0.6983358860015869,
"logps/rejected": -0.838935911655426,
"loss": 0.8633,
"nll_loss": 0.8049026131629944,
"rewards/accuracies": 0.75,
"rewards/chosen": -0.06983359158039093,
"rewards/margins": 0.014060002751648426,
"rewards/rejected": -0.08389358967542648,
"step": 52
},
{
"epoch": 0.5401273885350318,
"grad_norm": 0.5475600957870483,
"learning_rate": 8.197278911564626e-05,
"log_odds_chosen": 0.23267458379268646,
"log_odds_ratio": -0.6021015644073486,
"logits/chosen": -0.3862481415271759,
"logits/rejected": -0.38804885745048523,
"logps/chosen": -0.7031423449516296,
"logps/rejected": -0.8290390968322754,
"loss": 0.8748,
"nll_loss": 0.8145608901977539,
"rewards/accuracies": 0.71875,
"rewards/chosen": -0.07031424343585968,
"rewards/margins": 0.012589672580361366,
"rewards/rejected": -0.0829039141535759,
"step": 53
},
{
"epoch": 0.5503184713375796,
"grad_norm": 0.5666794776916504,
"learning_rate": 8.163265306122449e-05,
"log_odds_chosen": 0.25999322533607483,
"log_odds_ratio": -0.5887346863746643,
"logits/chosen": -0.347758412361145,
"logits/rejected": -0.3341512084007263,
"logps/chosen": -0.6920559406280518,
"logps/rejected": -0.8253049254417419,
"loss": 0.8235,
"nll_loss": 0.7646136283874512,
"rewards/accuracies": 0.78125,
"rewards/chosen": -0.06920559704303741,
"rewards/margins": 0.013324896804988384,
"rewards/rejected": -0.08253049850463867,
"step": 54
},
{
"epoch": 0.5605095541401274,
"grad_norm": 0.5203105211257935,
"learning_rate": 8.129251700680273e-05,
"log_odds_chosen": 0.18836164474487305,
"log_odds_ratio": -0.6200145483016968,
"logits/chosen": -0.3182459771633148,
"logits/rejected": -0.3383919596672058,
"logps/chosen": -0.6551133394241333,
"logps/rejected": -0.7530025839805603,
"loss": 0.8446,
"nll_loss": 0.7825708389282227,
"rewards/accuracies": 0.65625,
"rewards/chosen": -0.06551133096218109,
"rewards/margins": 0.009788922034204006,
"rewards/rejected": -0.07530025392770767,
"step": 55
},
{
"epoch": 0.5707006369426751,
"grad_norm": 0.5612345933914185,
"learning_rate": 8.095238095238096e-05,
"log_odds_chosen": 0.1355578601360321,
"log_odds_ratio": -0.6451556086540222,
"logits/chosen": -0.29888439178466797,
"logits/rejected": -0.2906675338745117,
"logps/chosen": -0.6383944153785706,
"logps/rejected": -0.7033387422561646,
"loss": 0.8437,
"nll_loss": 0.7791876196861267,
"rewards/accuracies": 0.59375,
"rewards/chosen": -0.06383944302797318,
"rewards/margins": 0.006494428031146526,
"rewards/rejected": -0.07033387571573257,
"step": 56
},
{
"epoch": 0.5808917197452229,
"grad_norm": 0.8524444103240967,
"learning_rate": 8.061224489795919e-05,
"log_odds_chosen": 0.349676251411438,
"log_odds_ratio": -0.5498685836791992,
"logits/chosen": -0.2660248577594757,
"logits/rejected": -0.29482465982437134,
"logps/chosen": -0.5994170904159546,
"logps/rejected": -0.770711362361908,
"loss": 0.7653,
"nll_loss": 0.7103118300437927,
"rewards/accuracies": 0.84375,
"rewards/chosen": -0.05994170904159546,
"rewards/margins": 0.017129434272646904,
"rewards/rejected": -0.07707113772630692,
"step": 57
},
{
"epoch": 0.5910828025477707,
"grad_norm": 0.5966730713844299,
"learning_rate": 8.027210884353742e-05,
"log_odds_chosen": 0.4513949453830719,
"log_odds_ratio": -0.5082166194915771,
"logits/chosen": -0.35525086522102356,
"logits/rejected": -0.39397263526916504,
"logps/chosen": -0.6463137865066528,
"logps/rejected": -0.8910082578659058,
"loss": 0.7776,
"nll_loss": 0.7267977595329285,
"rewards/accuracies": 0.96875,
"rewards/chosen": -0.0646313801407814,
"rewards/margins": 0.02446945384144783,
"rewards/rejected": -0.08910083770751953,
"step": 58
},
{
"epoch": 0.6012738853503184,
"grad_norm": 0.6914296746253967,
"learning_rate": 7.993197278911565e-05,
"log_odds_chosen": 0.2607879042625427,
"log_odds_ratio": -0.5879091024398804,
"logits/chosen": -0.26178666949272156,
"logits/rejected": -0.2836032807826996,
"logps/chosen": -0.60906583070755,
"logps/rejected": -0.7373634576797485,
"loss": 0.7941,
"nll_loss": 0.7352855801582336,
"rewards/accuracies": 0.75,
"rewards/chosen": -0.060906581580638885,
"rewards/margins": 0.012829765677452087,
"rewards/rejected": -0.07373635470867157,
"step": 59
},
{
"epoch": 0.6114649681528662,
"grad_norm": 0.49855294823646545,
"learning_rate": 7.959183673469388e-05,
"log_odds_chosen": 0.17502906918525696,
"log_odds_ratio": -0.6250565648078918,
"logits/chosen": -0.27342450618743896,
"logits/rejected": -0.2793465256690979,
"logps/chosen": -0.643248975276947,
"logps/rejected": -0.7312005162239075,
"loss": 0.8101,
"nll_loss": 0.7475584149360657,
"rewards/accuracies": 0.625,
"rewards/chosen": -0.06432489305734634,
"rewards/margins": 0.008795149624347687,
"rewards/rejected": -0.07312005013227463,
"step": 60
},
{
"epoch": 0.621656050955414,
"grad_norm": 0.623975396156311,
"learning_rate": 7.925170068027211e-05,
"log_odds_chosen": 0.20475000143051147,
"log_odds_ratio": -0.6119533181190491,
"logits/chosen": -0.2819938361644745,
"logits/rejected": -0.2693704068660736,
"logps/chosen": -0.6357014179229736,
"logps/rejected": -0.7285926342010498,
"loss": 0.7994,
"nll_loss": 0.7381711602210999,
"rewards/accuracies": 0.625,
"rewards/chosen": -0.06357015669345856,
"rewards/margins": 0.009289107285439968,
"rewards/rejected": -0.0728592500090599,
"step": 61
},
{
"epoch": 0.6318471337579618,
"grad_norm": 0.63492351770401,
"learning_rate": 7.891156462585034e-05,
"log_odds_chosen": 0.22653746604919434,
"log_odds_ratio": -0.5998401045799255,
"logits/chosen": -0.2391633242368698,
"logits/rejected": -0.22683343291282654,
"logps/chosen": -0.6707414388656616,
"logps/rejected": -0.7814602255821228,
"loss": 0.8376,
"nll_loss": 0.7776267528533936,
"rewards/accuracies": 0.78125,
"rewards/chosen": -0.06707413494586945,
"rewards/margins": 0.011071886867284775,
"rewards/rejected": -0.07814602553844452,
"step": 62
}
],
"logging_steps": 1,
"max_steps": 294,
"num_input_tokens_seen": 0,
"num_train_epochs": 3,
"save_steps": 1,
"stateful_callbacks": {
"TrainerControl": {
"args": {
"should_epoch_stop": false,
"should_evaluate": false,
"should_log": false,
"should_save": true,
"should_training_stop": false
},
"attributes": {}
}
},
"total_flos": 0.0,
"train_batch_size": 4,
"trial_name": null,
"trial_params": null
}