JW17's picture
Model save
7ba716b verified
raw
history blame
45.3 kB
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 3.0,
"eval_steps": 500,
"global_step": 702,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.04,
"grad_norm": 3.8125,
"learning_rate": 4.997496987311431e-06,
"log_odds_chosen": 0.16457681357860565,
"log_odds_ratio": -0.6622492671012878,
"logits/chosen": -2.9562315940856934,
"logits/rejected": -2.943472385406494,
"logps/chosen": -0.6016947031021118,
"logps/rejected": -0.675441563129425,
"loss": 0.9961,
"nll_loss": 0.7968690991401672,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.12033893167972565,
"rewards/margins": 0.014749372377991676,
"rewards/rejected": -0.13508832454681396,
"step": 10
},
{
"epoch": 0.09,
"grad_norm": 3.5,
"learning_rate": 4.989992961303738e-06,
"log_odds_chosen": 0.09340940415859222,
"log_odds_ratio": -0.6970354914665222,
"logits/chosen": -3.0346553325653076,
"logits/rejected": -3.047511577606201,
"logps/chosen": -0.5685297846794128,
"logps/rejected": -0.6139942407608032,
"loss": 0.4901,
"nll_loss": 0.35993438959121704,
"rewards/accuracies": 0.5249999761581421,
"rewards/chosen": -0.11370595544576645,
"rewards/margins": 0.00909288041293621,
"rewards/rejected": -0.1227988451719284,
"step": 20
},
{
"epoch": 0.13,
"grad_norm": 3.703125,
"learning_rate": 4.977502948114772e-06,
"log_odds_chosen": 0.06927172094583511,
"log_odds_ratio": -0.7004188895225525,
"logits/chosen": -3.142512798309326,
"logits/rejected": -3.1354269981384277,
"logps/chosen": -0.5674165487289429,
"logps/rejected": -0.590361475944519,
"loss": 0.4797,
"nll_loss": 0.3510151505470276,
"rewards/accuracies": 0.581250011920929,
"rewards/chosen": -0.11348330974578857,
"rewards/margins": 0.00458899512887001,
"rewards/rejected": -0.11807229369878769,
"step": 30
},
{
"epoch": 0.17,
"grad_norm": 3.484375,
"learning_rate": 4.960051957873726e-06,
"log_odds_chosen": 0.049079131335020065,
"log_odds_ratio": -0.709546685218811,
"logits/chosen": -3.2224674224853516,
"logits/rejected": -3.239351749420166,
"logps/chosen": -0.564799427986145,
"logps/rejected": -0.5826238393783569,
"loss": 0.4883,
"nll_loss": 0.3356490135192871,
"rewards/accuracies": 0.53125,
"rewards/chosen": -0.11295988410711288,
"rewards/margins": 0.003564885351806879,
"rewards/rejected": -0.11652477085590363,
"step": 40
},
{
"epoch": 0.21,
"grad_norm": 3.59375,
"learning_rate": 4.937674934620601e-06,
"log_odds_chosen": 0.1465633362531662,
"log_odds_ratio": -0.6634026169776917,
"logits/chosen": -3.2599196434020996,
"logits/rejected": -3.257382869720459,
"logps/chosen": -0.5410924553871155,
"logps/rejected": -0.6060024499893188,
"loss": 0.4577,
"nll_loss": 0.3106100261211395,
"rewards/accuracies": 0.6187499761581421,
"rewards/chosen": -0.10821850597858429,
"rewards/margins": 0.012981968931853771,
"rewards/rejected": -0.12120047956705093,
"step": 50
},
{
"epoch": 0.26,
"grad_norm": 3.234375,
"learning_rate": 4.9104166863339065e-06,
"log_odds_chosen": 0.061258088797330856,
"log_odds_ratio": -0.7155165076255798,
"logits/chosen": -3.3071017265319824,
"logits/rejected": -3.3195369243621826,
"logps/chosen": -0.6092182397842407,
"logps/rejected": -0.6360430121421814,
"loss": 0.494,
"nll_loss": 0.3727017939090729,
"rewards/accuracies": 0.5687500238418579,
"rewards/chosen": -0.12184363603591919,
"rewards/margins": 0.005364974495023489,
"rewards/rejected": -0.1272086203098297,
"step": 60
},
{
"epoch": 0.3,
"grad_norm": 3.28125,
"learning_rate": 4.878331795206705e-06,
"log_odds_chosen": 0.09167635440826416,
"log_odds_ratio": -0.7004056572914124,
"logits/chosen": -3.2954325675964355,
"logits/rejected": -3.311220169067383,
"logps/chosen": -0.5609530210494995,
"logps/rejected": -0.6090016961097717,
"loss": 0.4762,
"nll_loss": 0.31441202759742737,
"rewards/accuracies": 0.5562499761581421,
"rewards/chosen": -0.1121906191110611,
"rewards/margins": 0.009609726257622242,
"rewards/rejected": -0.12180032581090927,
"step": 70
},
{
"epoch": 0.34,
"grad_norm": 3.40625,
"learning_rate": 4.841484508350679e-06,
"log_odds_chosen": 0.1308586597442627,
"log_odds_ratio": -0.6873519420623779,
"logits/chosen": -3.2609565258026123,
"logits/rejected": -3.302786350250244,
"logps/chosen": -0.5968475937843323,
"logps/rejected": -0.6581549644470215,
"loss": 0.4895,
"nll_loss": 0.36970850825309753,
"rewards/accuracies": 0.5874999761581421,
"rewards/chosen": -0.1193695068359375,
"rewards/margins": 0.012261486612260342,
"rewards/rejected": -0.13163098692893982,
"step": 80
},
{
"epoch": 0.38,
"grad_norm": 3.78125,
"learning_rate": 4.799948609147061e-06,
"log_odds_chosen": 0.06979366391897202,
"log_odds_ratio": -0.7083112001419067,
"logits/chosen": -3.289924144744873,
"logits/rejected": -3.326308012008667,
"logps/chosen": -0.6024026274681091,
"logps/rejected": -0.6323495507240295,
"loss": 0.4868,
"nll_loss": 0.376103937625885,
"rewards/accuracies": 0.581250011920929,
"rewards/chosen": -0.12048052251338959,
"rewards/margins": 0.005989375524222851,
"rewards/rejected": -0.12646988034248352,
"step": 90
},
{
"epoch": 0.43,
"grad_norm": 3.4375,
"learning_rate": 4.753807269502041e-06,
"log_odds_chosen": 0.12073998153209686,
"log_odds_ratio": -0.6814596652984619,
"logits/chosen": -3.2658774852752686,
"logits/rejected": -3.276677370071411,
"logps/chosen": -0.5358774662017822,
"logps/rejected": -0.5820270776748657,
"loss": 0.4686,
"nll_loss": 0.3315598964691162,
"rewards/accuracies": 0.550000011920929,
"rewards/chosen": -0.10717550665140152,
"rewards/margins": 0.009229922667145729,
"rewards/rejected": -0.1164054125547409,
"step": 100
},
{
"epoch": 0.47,
"grad_norm": 3.140625,
"learning_rate": 4.703152883302498e-06,
"log_odds_chosen": 0.12317148596048355,
"log_odds_ratio": -0.6822798848152161,
"logits/chosen": -3.279986619949341,
"logits/rejected": -3.3097903728485107,
"logps/chosen": -0.5460432171821594,
"logps/rejected": -0.6105918884277344,
"loss": 0.4745,
"nll_loss": 0.3556433618068695,
"rewards/accuracies": 0.5625,
"rewards/chosen": -0.10920864343643188,
"rewards/margins": 0.01290974486619234,
"rewards/rejected": -0.12211839854717255,
"step": 110
},
{
"epoch": 0.51,
"grad_norm": 3.03125,
"learning_rate": 4.648086881405542e-06,
"log_odds_chosen": 0.11489371955394745,
"log_odds_ratio": -0.689535915851593,
"logits/chosen": -3.255175828933716,
"logits/rejected": -3.275517225265503,
"logps/chosen": -0.5755141377449036,
"logps/rejected": -0.6235055923461914,
"loss": 0.475,
"nll_loss": 0.32820188999176025,
"rewards/accuracies": 0.581250011920929,
"rewards/chosen": -0.11510282754898071,
"rewards/margins": 0.00959828682243824,
"rewards/rejected": -0.1247011050581932,
"step": 120
},
{
"epoch": 0.56,
"grad_norm": 2.890625,
"learning_rate": 4.588719528532342e-06,
"log_odds_chosen": 0.09126593172550201,
"log_odds_ratio": -0.697300910949707,
"logits/chosen": -3.2629482746124268,
"logits/rejected": -3.2472147941589355,
"logps/chosen": -0.5701687335968018,
"logps/rejected": -0.618691623210907,
"loss": 0.4766,
"nll_loss": 0.3278309404850006,
"rewards/accuracies": 0.5249999761581421,
"rewards/chosen": -0.1140337586402893,
"rewards/margins": 0.00970458984375,
"rewards/rejected": -0.12373832613229752,
"step": 130
},
{
"epoch": 0.6,
"grad_norm": 3.234375,
"learning_rate": 4.525169702472917e-06,
"log_odds_chosen": 0.055292803794145584,
"log_odds_ratio": -0.7131302356719971,
"logits/chosen": -3.1946372985839844,
"logits/rejected": -3.208055019378662,
"logps/chosen": -0.5659917593002319,
"logps/rejected": -0.5857821702957153,
"loss": 0.4741,
"nll_loss": 0.32717862725257874,
"rewards/accuracies": 0.512499988079071,
"rewards/chosen": -0.11319835484027863,
"rewards/margins": 0.003958091139793396,
"rewards/rejected": -0.11715643107891083,
"step": 140
},
{
"epoch": 0.64,
"grad_norm": 3.28125,
"learning_rate": 4.457564656044056e-06,
"log_odds_chosen": 0.10363437980413437,
"log_odds_ratio": -0.6924600005149841,
"logits/chosen": -3.227095127105713,
"logits/rejected": -3.269847869873047,
"logps/chosen": -0.540108859539032,
"logps/rejected": -0.5899955034255981,
"loss": 0.4874,
"nll_loss": 0.3099691569805145,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.10802175849676132,
"rewards/margins": 0.009977328591048717,
"rewards/rejected": -0.11799909919500351,
"step": 150
},
{
"epoch": 0.68,
"grad_norm": 3.25,
"learning_rate": 4.386039762276976e-06,
"log_odds_chosen": 0.12246614694595337,
"log_odds_ratio": -0.6819506883621216,
"logits/chosen": -3.1787781715393066,
"logits/rejected": -3.1843771934509277,
"logps/chosen": -0.556765615940094,
"logps/rejected": -0.6073147058486938,
"loss": 0.4718,
"nll_loss": 0.34596166014671326,
"rewards/accuracies": 0.59375,
"rewards/chosen": -0.11135313659906387,
"rewards/margins": 0.010109812021255493,
"rewards/rejected": -0.12146292626857758,
"step": 160
},
{
"epoch": 0.73,
"grad_norm": 3.609375,
"learning_rate": 4.310738243344996e-06,
"log_odds_chosen": 0.09136945009231567,
"log_odds_ratio": -0.6929733753204346,
"logits/chosen": -3.207054853439331,
"logits/rejected": -3.225612163543701,
"logps/chosen": -0.5217511653900146,
"logps/rejected": -0.5743865966796875,
"loss": 0.4775,
"nll_loss": 0.34531891345977783,
"rewards/accuracies": 0.550000011920929,
"rewards/chosen": -0.10435023158788681,
"rewards/margins": 0.010527082718908787,
"rewards/rejected": -0.11487732082605362,
"step": 170
},
{
"epoch": 0.77,
"grad_norm": 3.21875,
"learning_rate": 4.231810883773999e-06,
"log_odds_chosen": 0.13569065928459167,
"log_odds_ratio": -0.6776682138442993,
"logits/chosen": -3.176665782928467,
"logits/rejected": -3.197835922241211,
"logps/chosen": -0.5234608054161072,
"logps/rejected": -0.5825861692428589,
"loss": 0.4797,
"nll_loss": 0.358028769493103,
"rewards/accuracies": 0.5687500238418579,
"rewards/chosen": -0.10469217598438263,
"rewards/margins": 0.011825057677924633,
"rewards/rejected": -0.11651722341775894,
"step": 180
},
{
"epoch": 0.81,
"grad_norm": 2.96875,
"learning_rate": 4.149415728509971e-06,
"log_odds_chosen": 0.11225181818008423,
"log_odds_ratio": -0.6859180331230164,
"logits/chosen": -3.2352004051208496,
"logits/rejected": -3.2568259239196777,
"logps/chosen": -0.5570780634880066,
"logps/rejected": -0.6075831651687622,
"loss": 0.4563,
"nll_loss": 0.33723753690719604,
"rewards/accuracies": 0.5687500238418579,
"rewards/chosen": -0.11141560971736908,
"rewards/margins": 0.010101011954247952,
"rewards/rejected": -0.1215166226029396,
"step": 190
},
{
"epoch": 0.85,
"grad_norm": 3.09375,
"learning_rate": 4.063717766448194e-06,
"log_odds_chosen": 0.09286753088235855,
"log_odds_ratio": -0.6982024908065796,
"logits/chosen": -3.177328109741211,
"logits/rejected": -3.204103946685791,
"logps/chosen": -0.5489121675491333,
"logps/rejected": -0.5975584387779236,
"loss": 0.4871,
"nll_loss": 0.36433055996894836,
"rewards/accuracies": 0.543749988079071,
"rewards/chosen": -0.10978243499994278,
"rewards/margins": 0.009729253128170967,
"rewards/rejected": -0.1195116862654686,
"step": 200
},
{
"epoch": 0.9,
"grad_norm": 3.234375,
"learning_rate": 3.974888600057808e-06,
"log_odds_chosen": 0.06380079686641693,
"log_odds_ratio": -0.7078055143356323,
"logits/chosen": -3.179089307785034,
"logits/rejected": -3.1974921226501465,
"logps/chosen": -0.5259910821914673,
"logps/rejected": -0.5546124577522278,
"loss": 0.4842,
"nll_loss": 0.3254118859767914,
"rewards/accuracies": 0.53125,
"rewards/chosen": -0.1051982045173645,
"rewards/margins": 0.0057242861948907375,
"rewards/rejected": -0.11092247813940048,
"step": 210
},
{
"epoch": 0.94,
"grad_norm": 3.1875,
"learning_rate": 3.883106101763285e-06,
"log_odds_chosen": 0.07409624010324478,
"log_odds_ratio": -0.7109658122062683,
"logits/chosen": -3.2040557861328125,
"logits/rejected": -3.232853651046753,
"logps/chosen": -0.5239503979682922,
"logps/rejected": -0.5682771801948547,
"loss": 0.4671,
"nll_loss": 0.32592883706092834,
"rewards/accuracies": 0.518750011920929,
"rewards/chosen": -0.10479007661342621,
"rewards/margins": 0.008865350857377052,
"rewards/rejected": -0.1136554479598999,
"step": 220
},
{
"epoch": 0.98,
"grad_norm": 3.375,
"learning_rate": 3.7885540577708806e-06,
"log_odds_chosen": 0.17323985695838928,
"log_odds_ratio": -0.6644941568374634,
"logits/chosen": -3.242084503173828,
"logits/rejected": -3.256577730178833,
"logps/chosen": -0.5062168836593628,
"logps/rejected": -0.5952838063240051,
"loss": 0.4719,
"nll_loss": 0.31021052598953247,
"rewards/accuracies": 0.606249988079071,
"rewards/chosen": -0.10124337673187256,
"rewards/margins": 0.01781339757144451,
"rewards/rejected": -0.11905677616596222,
"step": 230
},
{
"epoch": 1.03,
"grad_norm": 3.84375,
"learning_rate": 3.6914218000532697e-06,
"log_odds_chosen": 0.369870126247406,
"log_odds_ratio": -0.5746415257453918,
"logits/chosen": -3.2143750190734863,
"logits/rejected": -3.2405693531036377,
"logps/chosen": -0.4556041657924652,
"logps/rejected": -0.6035604476928711,
"loss": 0.4236,
"nll_loss": 0.2927762269973755,
"rewards/accuracies": 0.6937500238418579,
"rewards/chosen": -0.09112082421779633,
"rewards/margins": 0.029591253027319908,
"rewards/rejected": -0.12071208655834198,
"step": 240
},
{
"epoch": 1.07,
"grad_norm": 3.8125,
"learning_rate": 3.5919038272292824e-06,
"log_odds_chosen": 0.6750356554985046,
"log_odds_ratio": -0.4590001702308655,
"logits/chosen": -3.1941792964935303,
"logits/rejected": -3.229973554611206,
"logps/chosen": -0.41714033484458923,
"logps/rejected": -0.676815390586853,
"loss": 0.3754,
"nll_loss": 0.2869933545589447,
"rewards/accuracies": 0.8374999761581421,
"rewards/chosen": -0.08342806994915009,
"rewards/margins": 0.05193500965833664,
"rewards/rejected": -0.13536307215690613,
"step": 250
},
{
"epoch": 1.11,
"grad_norm": 4.0,
"learning_rate": 3.4901994150978926e-06,
"log_odds_chosen": 0.7411336302757263,
"log_odds_ratio": -0.43500590324401855,
"logits/chosen": -3.248504161834717,
"logits/rejected": -3.2485687732696533,
"logps/chosen": -0.45796918869018555,
"logps/rejected": -0.7692248225212097,
"loss": 0.3752,
"nll_loss": 0.29675063490867615,
"rewards/accuracies": 0.8812500238418579,
"rewards/chosen": -0.09159384667873383,
"rewards/margins": 0.06225113198161125,
"rewards/rejected": -0.15384498238563538,
"step": 260
},
{
"epoch": 1.15,
"grad_norm": 4.125,
"learning_rate": 3.386512217606339e-06,
"log_odds_chosen": 0.672085702419281,
"log_odds_ratio": -0.4590581953525543,
"logits/chosen": -3.210575580596924,
"logits/rejected": -3.2388033866882324,
"logps/chosen": -0.4334833025932312,
"logps/rejected": -0.7109585404396057,
"loss": 0.3831,
"nll_loss": 0.29120856523513794,
"rewards/accuracies": 0.8374999761581421,
"rewards/chosen": -0.08669666945934296,
"rewards/margins": 0.055495042353868484,
"rewards/rejected": -0.14219172298908234,
"step": 270
},
{
"epoch": 1.2,
"grad_norm": 4.1875,
"learning_rate": 3.281049859051394e-06,
"log_odds_chosen": 0.6232529282569885,
"log_odds_ratio": -0.4691483974456787,
"logits/chosen": -3.1913466453552246,
"logits/rejected": -3.213522434234619,
"logps/chosen": -0.457376629114151,
"logps/rejected": -0.7198972702026367,
"loss": 0.3748,
"nll_loss": 0.2882610559463501,
"rewards/accuracies": 0.8374999761581421,
"rewards/chosen": -0.09147532284259796,
"rewards/margins": 0.05250414088368416,
"rewards/rejected": -0.14397946000099182,
"step": 280
},
{
"epoch": 1.24,
"grad_norm": 4.125,
"learning_rate": 3.17402351833036e-06,
"log_odds_chosen": 0.710340678691864,
"log_odds_ratio": -0.44652828574180603,
"logits/chosen": -3.149742841720581,
"logits/rejected": -3.189366579055786,
"logps/chosen": -0.43936777114868164,
"logps/rejected": -0.72941654920578,
"loss": 0.3629,
"nll_loss": 0.29283395409584045,
"rewards/accuracies": 0.8687499761581421,
"rewards/chosen": -0.08787355571985245,
"rewards/margins": 0.05800975114107132,
"rewards/rejected": -0.14588329195976257,
"step": 290
},
{
"epoch": 1.28,
"grad_norm": 4.09375,
"learning_rate": 3.0656475060743065e-06,
"log_odds_chosen": 0.7323763966560364,
"log_odds_ratio": -0.4345877766609192,
"logits/chosen": -3.1514878273010254,
"logits/rejected": -3.199571132659912,
"logps/chosen": -0.44486841559410095,
"logps/rejected": -0.7583059072494507,
"loss": 0.3627,
"nll_loss": 0.29908788204193115,
"rewards/accuracies": 0.887499988079071,
"rewards/chosen": -0.08897367864847183,
"rewards/margins": 0.06268750131130219,
"rewards/rejected": -0.15166118741035461,
"step": 300
},
{
"epoch": 1.32,
"grad_norm": 4.4375,
"learning_rate": 2.956138835510282e-06,
"log_odds_chosen": 0.7746503353118896,
"log_odds_ratio": -0.42378655076026917,
"logits/chosen": -3.1741645336151123,
"logits/rejected": -3.2151100635528564,
"logps/chosen": -0.4183032512664795,
"logps/rejected": -0.7279097437858582,
"loss": 0.3604,
"nll_loss": 0.26949578523635864,
"rewards/accuracies": 0.875,
"rewards/chosen": -0.08366064727306366,
"rewards/margins": 0.06192130595445633,
"rewards/rejected": -0.14558196067810059,
"step": 310
},
{
"epoch": 1.37,
"grad_norm": 4.28125,
"learning_rate": 2.8457167879118332e-06,
"log_odds_chosen": 0.7689862251281738,
"log_odds_ratio": -0.43063870072364807,
"logits/chosen": -3.159554958343506,
"logits/rejected": -3.188426971435547,
"logps/chosen": -0.3951800465583801,
"logps/rejected": -0.7116678953170776,
"loss": 0.3763,
"nll_loss": 0.27707797288894653,
"rewards/accuracies": 0.8687499761581421,
"rewards/chosen": -0.07903601229190826,
"rewards/margins": 0.0632975846529007,
"rewards/rejected": -0.14233359694480896,
"step": 320
},
{
"epoch": 1.41,
"grad_norm": 3.921875,
"learning_rate": 2.7346024735079483e-06,
"log_odds_chosen": 0.7219299674034119,
"log_odds_ratio": -0.44966334104537964,
"logits/chosen": -3.127485752105713,
"logits/rejected": -3.198312997817993,
"logps/chosen": -0.41490626335144043,
"logps/rejected": -0.7052132487297058,
"loss": 0.3787,
"nll_loss": 0.2818581163883209,
"rewards/accuracies": 0.831250011920929,
"rewards/chosen": -0.08298125118017197,
"rewards/margins": 0.058061398565769196,
"rewards/rejected": -0.14104264974594116,
"step": 330
},
{
"epoch": 1.45,
"grad_norm": 4.34375,
"learning_rate": 2.6230183887296955e-06,
"log_odds_chosen": 0.7794576287269592,
"log_odds_ratio": -0.42089906334877014,
"logits/chosen": -3.129941463470459,
"logits/rejected": -3.1785807609558105,
"logps/chosen": -0.4315931797027588,
"logps/rejected": -0.7535022497177124,
"loss": 0.3603,
"nll_loss": 0.30001071095466614,
"rewards/accuracies": 0.8687499761581421,
"rewards/chosen": -0.08631863445043564,
"rewards/margins": 0.06438182294368744,
"rewards/rejected": -0.15070047974586487,
"step": 340
},
{
"epoch": 1.5,
"grad_norm": 4.53125,
"learning_rate": 2.511187970681109e-06,
"log_odds_chosen": 0.6614913940429688,
"log_odds_ratio": -0.4707488417625427,
"logits/chosen": -3.1603000164031982,
"logits/rejected": -3.1471877098083496,
"logps/chosen": -0.4360755980014801,
"logps/rejected": -0.6958964467048645,
"loss": 0.3618,
"nll_loss": 0.2804470360279083,
"rewards/accuracies": 0.824999988079071,
"rewards/chosen": -0.0872151255607605,
"rewards/margins": 0.0519641637802124,
"rewards/rejected": -0.1391792744398117,
"step": 350
},
{
"epoch": 1.54,
"grad_norm": 3.828125,
"learning_rate": 2.399335149726463e-06,
"log_odds_chosen": 0.6727465391159058,
"log_odds_ratio": -0.4600442945957184,
"logits/chosen": -3.1209638118743896,
"logits/rejected": -3.1621603965759277,
"logps/chosen": -0.4566575884819031,
"logps/rejected": -0.7309259176254272,
"loss": 0.3719,
"nll_loss": 0.3025999665260315,
"rewards/accuracies": 0.824999988079071,
"rewards/chosen": -0.09133151918649673,
"rewards/margins": 0.0548536591231823,
"rewards/rejected": -0.14618518948554993,
"step": 360
},
{
"epoch": 1.58,
"grad_norm": 4.46875,
"learning_rate": 2.2876839010898377e-06,
"log_odds_chosen": 0.7592717409133911,
"log_odds_ratio": -0.43831682205200195,
"logits/chosen": -3.1502082347869873,
"logits/rejected": -3.207881212234497,
"logps/chosen": -0.42622488737106323,
"logps/rejected": -0.7278318405151367,
"loss": 0.3498,
"nll_loss": 0.2541278004646301,
"rewards/accuracies": 0.8812500238418579,
"rewards/chosen": -0.08524497598409653,
"rewards/margins": 0.0603213906288147,
"rewards/rejected": -0.14556637406349182,
"step": 370
},
{
"epoch": 1.62,
"grad_norm": 4.90625,
"learning_rate": 2.1764577963648616e-06,
"log_odds_chosen": 0.6783360838890076,
"log_odds_ratio": -0.46248960494995117,
"logits/chosen": -3.146242141723633,
"logits/rejected": -3.164093017578125,
"logps/chosen": -0.42164331674575806,
"logps/rejected": -0.6987554430961609,
"loss": 0.3636,
"nll_loss": 0.30008167028427124,
"rewards/accuracies": 0.84375,
"rewards/chosen": -0.08432865887880325,
"rewards/margins": 0.05542243644595146,
"rewards/rejected": -0.1397511065006256,
"step": 380
},
{
"epoch": 1.67,
"grad_norm": 4.4375,
"learning_rate": 2.0658795558326745e-06,
"log_odds_chosen": 0.6860665082931519,
"log_odds_ratio": -0.44539427757263184,
"logits/chosen": -3.1428816318511963,
"logits/rejected": -3.202255964279175,
"logps/chosen": -0.4127088487148285,
"logps/rejected": -0.6815465688705444,
"loss": 0.3548,
"nll_loss": 0.25606799125671387,
"rewards/accuracies": 0.8812500238418579,
"rewards/chosen": -0.08254177868366241,
"rewards/margins": 0.05376753211021423,
"rewards/rejected": -0.13630932569503784,
"step": 390
},
{
"epoch": 1.71,
"grad_norm": 4.15625,
"learning_rate": 1.956170602484582e-06,
"log_odds_chosen": 0.8614572286605835,
"log_odds_ratio": -0.3955555558204651,
"logits/chosen": -3.1366238594055176,
"logits/rejected": -3.1604573726654053,
"logps/chosen": -0.3913766145706177,
"logps/rejected": -0.7423630952835083,
"loss": 0.3487,
"nll_loss": 0.24887590110301971,
"rewards/accuracies": 0.887499988079071,
"rewards/chosen": -0.07827533036470413,
"rewards/margins": 0.07019731402397156,
"rewards/rejected": -0.1484726369380951,
"step": 400
},
{
"epoch": 1.75,
"grad_norm": 4.625,
"learning_rate": 1.8475506186424075e-06,
"log_odds_chosen": 0.7284751534461975,
"log_odds_ratio": -0.4521869122982025,
"logits/chosen": -3.1315054893493652,
"logits/rejected": -3.1535370349884033,
"logps/chosen": -0.4371741712093353,
"logps/rejected": -0.7353378534317017,
"loss": 0.3711,
"nll_loss": 0.2623272240161896,
"rewards/accuracies": 0.8500000238418579,
"rewards/chosen": -0.08743483573198318,
"rewards/margins": 0.059632737189531326,
"rewards/rejected": -0.1470675766468048,
"step": 410
},
{
"epoch": 1.79,
"grad_norm": 3.796875,
"learning_rate": 1.740237106064383e-06,
"log_odds_chosen": 0.7518190145492554,
"log_odds_ratio": -0.43122825026512146,
"logits/chosen": -3.155817747116089,
"logits/rejected": -3.1722676753997803,
"logps/chosen": -0.44021105766296387,
"logps/rejected": -0.752731442451477,
"loss": 0.3521,
"nll_loss": 0.27316415309906006,
"rewards/accuracies": 0.8687499761581421,
"rewards/chosen": -0.08804221451282501,
"rewards/margins": 0.06250409036874771,
"rewards/rejected": -0.15054629743099213,
"step": 420
},
{
"epoch": 1.84,
"grad_norm": 4.0625,
"learning_rate": 1.6344449504174193e-06,
"log_odds_chosen": 0.7577536106109619,
"log_odds_ratio": -0.4262828230857849,
"logits/chosen": -3.112555980682373,
"logits/rejected": -3.175693988800049,
"logps/chosen": -0.4249524176120758,
"logps/rejected": -0.7365177273750305,
"loss": 0.3772,
"nll_loss": 0.29327088594436646,
"rewards/accuracies": 0.875,
"rewards/chosen": -0.084990493953228,
"rewards/margins": 0.06231304258108139,
"rewards/rejected": -0.14730355143547058,
"step": 430
},
{
"epoch": 1.88,
"grad_norm": 4.625,
"learning_rate": 1.5303859909878632e-06,
"log_odds_chosen": 0.6891047954559326,
"log_odds_ratio": -0.45858821272850037,
"logits/chosen": -3.18835711479187,
"logits/rejected": -3.198822021484375,
"logps/chosen": -0.43595075607299805,
"logps/rejected": -0.707006573677063,
"loss": 0.3569,
"nll_loss": 0.2663270831108093,
"rewards/accuracies": 0.84375,
"rewards/chosen": -0.08719015121459961,
"rewards/margins": 0.054211169481277466,
"rewards/rejected": -0.14140132069587708,
"step": 440
},
{
"epoch": 1.92,
"grad_norm": 4.84375,
"learning_rate": 1.4282685964923643e-06,
"log_odds_chosen": 0.8139818906784058,
"log_odds_ratio": -0.41608935594558716,
"logits/chosen": -3.1146390438079834,
"logits/rejected": -3.167144775390625,
"logps/chosen": -0.41870421171188354,
"logps/rejected": -0.7421842813491821,
"loss": 0.3699,
"nll_loss": 0.2976505160331726,
"rewards/accuracies": 0.8374999761581421,
"rewards/chosen": -0.08374085277318954,
"rewards/margins": 0.06469599902629852,
"rewards/rejected": -0.14843687415122986,
"step": 450
},
{
"epoch": 1.97,
"grad_norm": 4.53125,
"learning_rate": 1.328297247838241e-06,
"log_odds_chosen": 0.7360786199569702,
"log_odds_ratio": -0.4329991936683655,
"logits/chosen": -3.162118434906006,
"logits/rejected": -3.2087559700012207,
"logps/chosen": -0.4136020541191101,
"logps/rejected": -0.7124483585357666,
"loss": 0.3642,
"nll_loss": 0.2634555995464325,
"rewards/accuracies": 0.887499988079071,
"rewards/chosen": -0.08272042125463486,
"rewards/margins": 0.05976926162838936,
"rewards/rejected": -0.14248968660831451,
"step": 460
},
{
"epoch": 2.01,
"grad_norm": 5.3125,
"learning_rate": 1.2306721286688312e-06,
"log_odds_chosen": 0.8309246301651001,
"log_odds_ratio": -0.4013638496398926,
"logits/chosen": -3.150937795639038,
"logits/rejected": -3.205261707305908,
"logps/chosen": -0.4131666123867035,
"logps/rejected": -0.7659183740615845,
"loss": 0.3651,
"nll_loss": 0.2736993134021759,
"rewards/accuracies": 0.9375,
"rewards/chosen": -0.08263333141803741,
"rewards/margins": 0.07055036723613739,
"rewards/rejected": -0.1531836986541748,
"step": 470
},
{
"epoch": 2.05,
"grad_norm": 4.5625,
"learning_rate": 1.1355887245137383e-06,
"log_odds_chosen": 1.1850926876068115,
"log_odds_ratio": -0.30928006768226624,
"logits/chosen": -3.197258234024048,
"logits/rejected": -3.233253002166748,
"logps/chosen": -0.3570306599140167,
"logps/rejected": -0.8269041776657104,
"loss": 0.306,
"nll_loss": 0.21892933547496796,
"rewards/accuracies": 0.956250011920929,
"rewards/chosen": -0.07140612602233887,
"rewards/margins": 0.09397471696138382,
"rewards/rejected": -0.16538085043430328,
"step": 480
},
{
"epoch": 2.09,
"grad_norm": 4.3125,
"learning_rate": 1.043237431346622e-06,
"log_odds_chosen": 1.2696115970611572,
"log_odds_ratio": -0.29188066720962524,
"logits/chosen": -3.060274600982666,
"logits/rejected": -3.112060785293579,
"logps/chosen": -0.37511518597602844,
"logps/rejected": -0.8814083933830261,
"loss": 0.3133,
"nll_loss": 0.2801322340965271,
"rewards/accuracies": 0.96875,
"rewards/chosen": -0.07502304017543793,
"rewards/margins": 0.10125863552093506,
"rewards/rejected": -0.17628169059753418,
"step": 490
},
{
"epoch": 2.14,
"grad_norm": 4.40625,
"learning_rate": 9.538031743343628e-07,
"log_odds_chosen": 1.2528076171875,
"log_odds_ratio": -0.3035999834537506,
"logits/chosen": -3.0861003398895264,
"logits/rejected": -3.111483097076416,
"logps/chosen": -0.39375776052474976,
"logps/rejected": -0.9319614171981812,
"loss": 0.3176,
"nll_loss": 0.2627008259296417,
"rewards/accuracies": 0.956250011920929,
"rewards/chosen": -0.07875155657529831,
"rewards/margins": 0.10764074325561523,
"rewards/rejected": -0.18639230728149414,
"step": 500
},
{
"epoch": 2.18,
"grad_norm": 4.75,
"learning_rate": 8.674650375410379e-07,
"log_odds_chosen": 1.0996912717819214,
"log_odds_ratio": -0.34020066261291504,
"logits/chosen": -3.0855681896209717,
"logits/rejected": -3.1118760108947754,
"logps/chosen": -0.426668643951416,
"logps/rejected": -0.8785163760185242,
"loss": 0.3107,
"nll_loss": 0.2832065224647522,
"rewards/accuracies": 0.925000011920929,
"rewards/chosen": -0.08533372730016708,
"rewards/margins": 0.09036955237388611,
"rewards/rejected": -0.17570330202579498,
"step": 510
},
{
"epoch": 2.22,
"grad_norm": 4.3125,
"learning_rate": 7.843959053281663e-07,
"log_odds_chosen": 1.3144631385803223,
"log_odds_ratio": -0.2888423502445221,
"logits/chosen": -3.060899019241333,
"logits/rejected": -3.110773801803589,
"logps/chosen": -0.3530345559120178,
"logps/rejected": -0.8913728594779968,
"loss": 0.3117,
"nll_loss": 0.2591019570827484,
"rewards/accuracies": 0.956250011920929,
"rewards/chosen": -0.07060691714286804,
"rewards/margins": 0.10766766965389252,
"rewards/rejected": -0.17827460169792175,
"step": 520
},
{
"epoch": 2.26,
"grad_norm": 5.0,
"learning_rate": 7.047621161693152e-07,
"log_odds_chosen": 1.327051043510437,
"log_odds_ratio": -0.2728634774684906,
"logits/chosen": -3.0373597145080566,
"logits/rejected": -3.121220350265503,
"logps/chosen": -0.35444390773773193,
"logps/rejected": -0.9054730534553528,
"loss": 0.3201,
"nll_loss": 0.2551073729991913,
"rewards/accuracies": 0.981249988079071,
"rewards/chosen": -0.07088878005743027,
"rewards/margins": 0.11020584404468536,
"rewards/rejected": -0.18109463155269623,
"step": 530
},
{
"epoch": 2.31,
"grad_norm": 4.8125,
"learning_rate": 6.28723129572247e-07,
"log_odds_chosen": 1.3342939615249634,
"log_odds_ratio": -0.27417218685150146,
"logits/chosen": -3.0797994136810303,
"logits/rejected": -3.1183524131774902,
"logps/chosen": -0.362510621547699,
"logps/rejected": -0.9240224957466125,
"loss": 0.314,
"nll_loss": 0.2499639093875885,
"rewards/accuracies": 0.9624999761581421,
"rewards/chosen": -0.07250212132930756,
"rewards/margins": 0.11230238527059555,
"rewards/rejected": -0.1848045140504837,
"step": 540
},
{
"epoch": 2.35,
"grad_norm": 4.5,
"learning_rate": 5.564312067755856e-07,
"log_odds_chosen": 1.2629302740097046,
"log_odds_ratio": -0.29583701491355896,
"logits/chosen": -3.106447696685791,
"logits/rejected": -3.153026580810547,
"logps/chosen": -0.36336833238601685,
"logps/rejected": -0.8715957403182983,
"loss": 0.3098,
"nll_loss": 0.2292967140674591,
"rewards/accuracies": 0.9750000238418579,
"rewards/chosen": -0.07267366349697113,
"rewards/margins": 0.10164550691843033,
"rewards/rejected": -0.17431916296482086,
"step": 550
},
{
"epoch": 2.39,
"grad_norm": 4.375,
"learning_rate": 4.880311058593617e-07,
"log_odds_chosen": 1.3260443210601807,
"log_odds_ratio": -0.2782818675041199,
"logits/chosen": -3.0871644020080566,
"logits/rejected": -3.1364452838897705,
"logps/chosen": -0.36632782220840454,
"logps/rejected": -0.924032986164093,
"loss": 0.2976,
"nll_loss": 0.22606725990772247,
"rewards/accuracies": 0.9750000238418579,
"rewards/chosen": -0.07326555997133255,
"rewards/margins": 0.11154104769229889,
"rewards/rejected": -0.18480661511421204,
"step": 560
},
{
"epoch": 2.44,
"grad_norm": 4.96875,
"learning_rate": 4.2365979187997094e-07,
"log_odds_chosen": 1.277055025100708,
"log_odds_ratio": -0.29783302545547485,
"logits/chosen": -3.0671963691711426,
"logits/rejected": -3.1140341758728027,
"logps/chosen": -0.381780207157135,
"logps/rejected": -0.9046441912651062,
"loss": 0.3148,
"nll_loss": 0.2656724154949188,
"rewards/accuracies": 0.9375,
"rewards/chosen": -0.07635603845119476,
"rewards/margins": 0.10457281768321991,
"rewards/rejected": -0.18092885613441467,
"step": 570
},
{
"epoch": 2.48,
"grad_norm": 4.9375,
"learning_rate": 3.634461626099495e-07,
"log_odds_chosen": 1.2928454875946045,
"log_odds_ratio": -0.28668928146362305,
"logits/chosen": -3.0869479179382324,
"logits/rejected": -3.1252989768981934,
"logps/chosen": -0.345980703830719,
"logps/rejected": -0.8593694567680359,
"loss": 0.3035,
"nll_loss": 0.22678887844085693,
"rewards/accuracies": 0.96875,
"rewards/chosen": -0.06919614970684052,
"rewards/margins": 0.10267776250839233,
"rewards/rejected": -0.17187389731407166,
"step": 580
},
{
"epoch": 2.52,
"grad_norm": 5.5,
"learning_rate": 3.075107904317667e-07,
"log_odds_chosen": 1.2825336456298828,
"log_odds_ratio": -0.29012447595596313,
"logits/chosen": -3.0542449951171875,
"logits/rejected": -3.106863021850586,
"logps/chosen": -0.36031287908554077,
"logps/rejected": -0.8658483624458313,
"loss": 0.3156,
"nll_loss": 0.2495698481798172,
"rewards/accuracies": 0.96875,
"rewards/chosen": -0.07206258177757263,
"rewards/margins": 0.10110710561275482,
"rewards/rejected": -0.17316967248916626,
"step": 590
},
{
"epoch": 2.56,
"grad_norm": 4.90625,
"learning_rate": 2.5596568090246546e-07,
"log_odds_chosen": 1.2836467027664185,
"log_odds_ratio": -0.28635507822036743,
"logits/chosen": -3.0764803886413574,
"logits/rejected": -3.1473774909973145,
"logps/chosen": -0.37658047676086426,
"logps/rejected": -0.9045608639717102,
"loss": 0.3141,
"nll_loss": 0.24526798725128174,
"rewards/accuracies": 0.9750000238418579,
"rewards/chosen": -0.07531610131263733,
"rewards/margins": 0.10559606552124023,
"rewards/rejected": -0.18091216683387756,
"step": 600
},
{
"epoch": 2.61,
"grad_norm": 4.28125,
"learning_rate": 2.0891404847259267e-07,
"log_odds_chosen": 1.2697139978408813,
"log_odds_ratio": -0.29423031210899353,
"logits/chosen": -3.090924024581909,
"logits/rejected": -3.146167039871216,
"logps/chosen": -0.38669395446777344,
"logps/rejected": -0.9116611480712891,
"loss": 0.313,
"nll_loss": 0.25895369052886963,
"rewards/accuracies": 0.96875,
"rewards/chosen": -0.0773387923836708,
"rewards/margins": 0.10499344766139984,
"rewards/rejected": -0.18233224749565125,
"step": 610
},
{
"epoch": 2.65,
"grad_norm": 4.96875,
"learning_rate": 1.6645010980854082e-07,
"log_odds_chosen": 1.3362077474594116,
"log_odds_ratio": -0.2740064263343811,
"logits/chosen": -3.075295925140381,
"logits/rejected": -3.110752582550049,
"logps/chosen": -0.3653913140296936,
"logps/rejected": -0.9164104461669922,
"loss": 0.3052,
"nll_loss": 0.26205652952194214,
"rewards/accuracies": 0.9750000238418579,
"rewards/chosen": -0.07307826727628708,
"rewards/margins": 0.1102038249373436,
"rewards/rejected": -0.18328210711479187,
"step": 620
},
{
"epoch": 2.69,
"grad_norm": 5.21875,
"learning_rate": 1.286588951321363e-07,
"log_odds_chosen": 1.2346899509429932,
"log_odds_ratio": -0.29600512981414795,
"logits/chosen": -3.094604015350342,
"logits/rejected": -3.125540256500244,
"logps/chosen": -0.393343448638916,
"logps/rejected": -0.9184632301330566,
"loss": 0.3236,
"nll_loss": 0.2790059447288513,
"rewards/accuracies": 0.96875,
"rewards/chosen": -0.07866868376731873,
"rewards/margins": 0.10502395778894424,
"rewards/rejected": -0.18369261920452118,
"step": 630
},
{
"epoch": 2.74,
"grad_norm": 4.6875,
"learning_rate": 9.561607795526007e-08,
"log_odds_chosen": 1.2945499420166016,
"log_odds_ratio": -0.2860717177391052,
"logits/chosen": -3.092331647872925,
"logits/rejected": -3.149021863937378,
"logps/chosen": -0.3724428415298462,
"logps/rejected": -0.9131635427474976,
"loss": 0.312,
"nll_loss": 0.2520996630191803,
"rewards/accuracies": 0.9750000238418579,
"rewards/chosen": -0.074488565325737,
"rewards/margins": 0.1081441268324852,
"rewards/rejected": -0.182632714509964,
"step": 640
},
{
"epoch": 2.78,
"grad_norm": 4.375,
"learning_rate": 6.738782355044048e-08,
"log_odds_chosen": 1.303377628326416,
"log_odds_ratio": -0.28758224844932556,
"logits/chosen": -3.0791025161743164,
"logits/rejected": -3.130547046661377,
"logps/chosen": -0.3604482114315033,
"logps/rejected": -0.8876269459724426,
"loss": 0.3161,
"nll_loss": 0.2479972094297409,
"rewards/accuracies": 0.9750000238418579,
"rewards/chosen": -0.07208964973688126,
"rewards/margins": 0.10543572902679443,
"rewards/rejected": -0.1775253713130951,
"step": 650
},
{
"epoch": 2.82,
"grad_norm": 4.8125,
"learning_rate": 4.40306564608381e-08,
"log_odds_chosen": 1.220916509628296,
"log_odds_ratio": -0.3061557412147522,
"logits/chosen": -3.0849318504333496,
"logits/rejected": -3.0918779373168945,
"logps/chosen": -0.38813266158103943,
"logps/rejected": -0.8825920224189758,
"loss": 0.3064,
"nll_loss": 0.26622262597084045,
"rewards/accuracies": 0.96875,
"rewards/chosen": -0.0776265487074852,
"rewards/margins": 0.09889186918735504,
"rewards/rejected": -0.17651841044425964,
"step": 660
},
{
"epoch": 2.86,
"grad_norm": 4.9375,
"learning_rate": 2.5591347314928572e-08,
"log_odds_chosen": 1.1736371517181396,
"log_odds_ratio": -0.3195890486240387,
"logits/chosen": -3.0899367332458496,
"logits/rejected": -3.1268470287323,
"logps/chosen": -0.3975283205509186,
"logps/rejected": -0.897696316242218,
"loss": 0.3181,
"nll_loss": 0.2676974833011627,
"rewards/accuracies": 0.9375,
"rewards/chosen": -0.07950566709041595,
"rewards/margins": 0.10003360360860825,
"rewards/rejected": -0.1795392781496048,
"step": 670
},
{
"epoch": 2.91,
"grad_norm": 3.921875,
"learning_rate": 1.2106819172520434e-08,
"log_odds_chosen": 1.2462742328643799,
"log_odds_ratio": -0.29843610525131226,
"logits/chosen": -3.0916683673858643,
"logits/rejected": -3.126143217086792,
"logps/chosen": -0.37134823203086853,
"logps/rejected": -0.8719061613082886,
"loss": 0.3068,
"nll_loss": 0.2421465665102005,
"rewards/accuracies": 0.9750000238418579,
"rewards/chosen": -0.07426965236663818,
"rewards/margins": 0.10011158138513565,
"rewards/rejected": -0.17438122630119324,
"step": 680
},
{
"epoch": 2.95,
"grad_norm": 4.0625,
"learning_rate": 3.6040735896455957e-09,
"log_odds_chosen": 1.2757344245910645,
"log_odds_ratio": -0.29123741388320923,
"logits/chosen": -3.0802972316741943,
"logits/rejected": -3.130801200866699,
"logps/chosen": -0.37011533975601196,
"logps/rejected": -0.8979229927062988,
"loss": 0.3024,
"nll_loss": 0.23298993706703186,
"rewards/accuracies": 0.9437500238418579,
"rewards/chosen": -0.07402306795120239,
"rewards/margins": 0.10556153953075409,
"rewards/rejected": -0.17958460748195648,
"step": 690
},
{
"epoch": 2.99,
"grad_norm": 5.0,
"learning_rate": 1.0013655036916758e-10,
"log_odds_chosen": 1.3622093200683594,
"log_odds_ratio": -0.2736757695674896,
"logits/chosen": -3.140601396560669,
"logits/rejected": -3.1706244945526123,
"logps/chosen": -0.34326601028442383,
"logps/rejected": -0.8912754058837891,
"loss": 0.3068,
"nll_loss": 0.21759584546089172,
"rewards/accuracies": 0.9750000238418579,
"rewards/chosen": -0.06865320354700089,
"rewards/margins": 0.10960187017917633,
"rewards/rejected": -0.1782551109790802,
"step": 700
},
{
"epoch": 3.0,
"step": 702,
"total_flos": 0.0,
"train_loss": 0.3926089354220279,
"train_runtime": 9824.6416,
"train_samples_per_second": 4.569,
"train_steps_per_second": 0.071
}
],
"logging_steps": 10,
"max_steps": 702,
"num_input_tokens_seen": 0,
"num_train_epochs": 3,
"save_steps": 500,
"total_flos": 0.0,
"train_batch_size": 8,
"trial_name": null,
"trial_params": null
}