Files
mistral-7b-inst-dpo-on-p-tw…/checkpoint-1867/trainer_state.json
ModelHub XC b18324889e 初始化项目,由ModelHub XC社区提供模型
Model: HCY123902/mistral-7b-inst-dpo-on-p-tw31-beta-1e-0
Source: Original Platform
2026-04-28 03:31:55 +08:00

4028 lines
156 KiB
JSON

{
"best_global_step": null,
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 1.0,
"eval_steps": 500,
"global_step": 1867,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.0005358338914936369,
"grad_norm": 31.966846466064453,
"learning_rate": 0.0,
"logits/chosen": -2.52783203125,
"logits/rejected": -2.498046875,
"logps/chosen": -277.71875,
"logps/rejected": -223.125,
"logps/weighted_chosen": -0.533447265625,
"logps/weighted_rejected": -0.5252685546875,
"loss": 0.6914,
"rewards/accuracies": 0.0,
"rewards/chosen": 0.0,
"rewards/margins": 0.0,
"rewards/rejected": 0.0,
"rewards/weighted_accuracies": 0.0,
"rewards/weighted_chosen": 0.0,
"rewards/weighted_margins": 0.0,
"rewards/weighted_rejected": 0.0,
"step": 1
},
{
"epoch": 0.0053583389149363695,
"grad_norm": 31.4305419921875,
"learning_rate": 4.81283422459893e-08,
"logits/chosen": -2.42626953125,
"logits/rejected": -2.3970539569854736,
"logps/chosen": -195.6822967529297,
"logps/rejected": -195.890625,
"logps/weighted_chosen": -0.5433688759803772,
"logps/weighted_rejected": -0.5479600429534912,
"loss": 0.6923,
"rewards/accuracies": 0.2951388955116272,
"rewards/chosen": -0.0735677108168602,
"rewards/margins": 0.00434027798473835,
"rewards/rejected": -0.0779079869389534,
"rewards/weighted_accuracies": 0.34375,
"rewards/weighted_chosen": 0.00025431314134038985,
"rewards/weighted_margins": 0.0005137125845067203,
"rewards/weighted_rejected": -0.0002593994140625,
"step": 10
},
{
"epoch": 0.010716677829872739,
"grad_norm": 17.30971908569336,
"learning_rate": 1.0160427807486631e-07,
"logits/chosen": -2.400830030441284,
"logits/rejected": -2.399658203125,
"logps/chosen": -199.91366577148438,
"logps/rejected": -202.21328735351562,
"logps/weighted_chosen": -0.537811279296875,
"logps/weighted_rejected": -0.548992931842804,
"loss": 0.6926,
"rewards/accuracies": 0.359375,
"rewards/chosen": -0.04570312425494194,
"rewards/margins": 0.02001953125,
"rewards/rejected": -0.06572265923023224,
"rewards/weighted_accuracies": 0.3656249940395355,
"rewards/weighted_chosen": -0.00011291504051769152,
"rewards/weighted_margins": 0.00011787414405262098,
"rewards/weighted_rejected": -0.00023076534853316844,
"step": 20
},
{
"epoch": 0.016075016744809108,
"grad_norm": 30.744884490966797,
"learning_rate": 1.5508021390374333e-07,
"logits/chosen": -2.3249268531799316,
"logits/rejected": -2.299609422683716,
"logps/chosen": -198.44375610351562,
"logps/rejected": -212.16641235351562,
"logps/weighted_chosen": -0.555804431438446,
"logps/weighted_rejected": -0.547528088092804,
"loss": 0.6927,
"rewards/accuracies": 0.4000000059604645,
"rewards/chosen": 0.0771484375,
"rewards/margins": 0.18007811903953552,
"rewards/rejected": -0.10292968899011612,
"rewards/weighted_accuracies": 0.3656249940395355,
"rewards/weighted_chosen": 0.0004203796270303428,
"rewards/weighted_margins": -0.0001464843808207661,
"rewards/weighted_rejected": 0.0005667686345987022,
"step": 30
},
{
"epoch": 0.021433355659745478,
"grad_norm": 51.52167510986328,
"learning_rate": 2.085561497326203e-07,
"logits/chosen": -2.3595948219299316,
"logits/rejected": -2.375,
"logps/chosen": -192.6171875,
"logps/rejected": -207.1945343017578,
"logps/weighted_chosen": -0.544903576374054,
"logps/weighted_rejected": -0.588610827922821,
"loss": 0.6926,
"rewards/accuracies": 0.31562501192092896,
"rewards/chosen": -0.08066406100988388,
"rewards/margins": -0.11210937798023224,
"rewards/rejected": 0.03144531324505806,
"rewards/weighted_accuracies": 0.40937501192092896,
"rewards/weighted_chosen": 0.0021041869185864925,
"rewards/weighted_margins": 0.0001922607480082661,
"rewards/weighted_rejected": 0.0019119263160973787,
"step": 40
},
{
"epoch": 0.02679169457468185,
"grad_norm": 29.50680160522461,
"learning_rate": 2.620320855614973e-07,
"logits/chosen": -2.3987059593200684,
"logits/rejected": -2.3811278343200684,
"logps/chosen": -213.0945281982422,
"logps/rejected": -228.8328094482422,
"logps/weighted_chosen": -0.5421508550643921,
"logps/weighted_rejected": -0.557452380657196,
"loss": 0.6925,
"rewards/accuracies": 0.390625,
"rewards/chosen": 0.06367187201976776,
"rewards/margins": 0.06484375149011612,
"rewards/rejected": -0.0011718750465661287,
"rewards/weighted_accuracies": 0.45625001192092896,
"rewards/weighted_chosen": 0.0056938170455396175,
"rewards/weighted_margins": 0.000339508056640625,
"rewards/weighted_rejected": 0.005351257510483265,
"step": 50
},
{
"epoch": 0.032150033489618215,
"grad_norm": 30.643386840820312,
"learning_rate": 3.155080213903743e-07,
"logits/chosen": -2.347216844558716,
"logits/rejected": -2.3227295875549316,
"logps/chosen": -211.7421875,
"logps/rejected": -218.0234375,
"logps/weighted_chosen": -0.536822497844696,
"logps/weighted_rejected": -0.568676769733429,
"loss": 0.6923,
"rewards/accuracies": 0.421875,
"rewards/chosen": 0.09257812798023224,
"rewards/margins": 0.10820312798023224,
"rewards/rejected": -0.015625,
"rewards/weighted_accuracies": 0.45625001192092896,
"rewards/weighted_chosen": 0.009661102667450905,
"rewards/weighted_margins": 0.000804901123046875,
"rewards/weighted_rejected": 0.00885620154440403,
"step": 60
},
{
"epoch": 0.03750837240455459,
"grad_norm": 27.744924545288086,
"learning_rate": 3.689839572192513e-07,
"logits/chosen": -2.442822217941284,
"logits/rejected": -2.4500975608825684,
"logps/chosen": -194.47891235351562,
"logps/rejected": -200.30624389648438,
"logps/weighted_chosen": -0.5030456781387329,
"logps/weighted_rejected": -0.5135132074356079,
"loss": 0.6906,
"rewards/accuracies": 0.4312500059604645,
"rewards/chosen": 0.03203124925494194,
"rewards/margins": 0.20937499403953552,
"rewards/rejected": -0.17734375596046448,
"rewards/weighted_accuracies": 0.5218750238418579,
"rewards/weighted_chosen": 0.01783904992043972,
"rewards/weighted_margins": 0.0045639039017260075,
"rewards/weighted_rejected": 0.013275146484375,
"step": 70
},
{
"epoch": 0.042866711319490956,
"grad_norm": 38.061683654785156,
"learning_rate": 4.2245989304812833e-07,
"logits/chosen": -2.403271436691284,
"logits/rejected": -2.430908203125,
"logps/chosen": -219.08438110351562,
"logps/rejected": -239.91561889648438,
"logps/weighted_chosen": -0.529132068157196,
"logps/weighted_rejected": -0.5560547113418579,
"loss": 0.6903,
"rewards/accuracies": 0.43437498807907104,
"rewards/chosen": 0.29023438692092896,
"rewards/margins": 0.27910155057907104,
"rewards/rejected": 0.01113281212747097,
"rewards/weighted_accuracies": 0.518750011920929,
"rewards/weighted_chosen": 0.022306060418486595,
"rewards/weighted_margins": 0.005640792660415173,
"rewards/weighted_rejected": 0.016665268689393997,
"step": 80
},
{
"epoch": 0.04822505023442733,
"grad_norm": 28.250593185424805,
"learning_rate": 4.7593582887700533e-07,
"logits/chosen": -2.392260789871216,
"logits/rejected": -2.4065184593200684,
"logps/chosen": -195.1953125,
"logps/rejected": -207.7804718017578,
"logps/weighted_chosen": -0.51409912109375,
"logps/weighted_rejected": -0.526226818561554,
"loss": 0.6906,
"rewards/accuracies": 0.5249999761581421,
"rewards/chosen": 0.10097656399011612,
"rewards/margins": 0.724414050579071,
"rewards/rejected": -0.6234375238418579,
"rewards/weighted_accuracies": 0.559374988079071,
"rewards/weighted_chosen": 0.025403594598174095,
"rewards/weighted_margins": 0.006229400634765625,
"rewards/weighted_rejected": 0.01916809007525444,
"step": 90
},
{
"epoch": 0.0535833891493637,
"grad_norm": 18.92244529724121,
"learning_rate": 5.294117647058823e-07,
"logits/chosen": -2.3767333030700684,
"logits/rejected": -2.3611817359924316,
"logps/chosen": -224.2859344482422,
"logps/rejected": -229.6570281982422,
"logps/weighted_chosen": -0.534497082233429,
"logps/weighted_rejected": -0.5674392580986023,
"loss": 0.6903,
"rewards/accuracies": 0.5093749761581421,
"rewards/chosen": -0.8511718511581421,
"rewards/margins": 1.123437523841858,
"rewards/rejected": -1.974609375,
"rewards/weighted_accuracies": 0.53125,
"rewards/weighted_chosen": 0.023969268426299095,
"rewards/weighted_margins": 0.00682754535228014,
"rewards/weighted_rejected": 0.01714172400534153,
"step": 100
},
{
"epoch": 0.058941728064300064,
"grad_norm": 50.693965911865234,
"learning_rate": 5.828877005347593e-07,
"logits/chosen": -2.3811888694763184,
"logits/rejected": -2.382458448410034,
"logps/chosen": -217.30703735351562,
"logps/rejected": -225.4343719482422,
"logps/weighted_chosen": -0.54705810546875,
"logps/weighted_rejected": -0.5594238042831421,
"loss": 0.6951,
"rewards/accuracies": 0.46562498807907104,
"rewards/chosen": -2.3193359375,
"rewards/margins": 0.624218761920929,
"rewards/rejected": -2.942578077316284,
"rewards/weighted_accuracies": 0.4781250059604645,
"rewards/weighted_chosen": 0.022613525390625,
"rewards/weighted_margins": -0.0009094238048419356,
"rewards/weighted_rejected": 0.02351989783346653,
"step": 110
},
{
"epoch": 0.06430006697923643,
"grad_norm": 27.526296615600586,
"learning_rate": 6.363636363636363e-07,
"logits/chosen": -2.459423780441284,
"logits/rejected": -2.4720702171325684,
"logps/chosen": -187.80313110351562,
"logps/rejected": -226.0906219482422,
"logps/weighted_chosen": -0.533001720905304,
"logps/weighted_rejected": -0.605639636516571,
"loss": 0.6903,
"rewards/accuracies": 0.49687498807907104,
"rewards/chosen": -3.5208983421325684,
"rewards/margins": 1.0193359851837158,
"rewards/rejected": -4.540820121765137,
"rewards/weighted_accuracies": 0.5062500238418579,
"rewards/weighted_chosen": 0.005961227230727673,
"rewards/weighted_margins": 0.009808349423110485,
"rewards/weighted_rejected": -0.0038513182662427425,
"step": 120
},
{
"epoch": 0.06965840589417281,
"grad_norm": 26.822662353515625,
"learning_rate": 6.898395721925134e-07,
"logits/chosen": -2.565673828125,
"logits/rejected": -2.564648389816284,
"logps/chosen": -231.10000610351562,
"logps/rejected": -240.86093139648438,
"logps/weighted_chosen": -0.537158191204071,
"logps/weighted_rejected": -0.5712524652481079,
"loss": 0.6953,
"rewards/accuracies": 0.53125,
"rewards/chosen": -2.906054735183716,
"rewards/margins": 1.6062500476837158,
"rewards/rejected": -4.512109279632568,
"rewards/weighted_accuracies": 0.49687498807907104,
"rewards/weighted_chosen": 0.00572967529296875,
"rewards/weighted_margins": -0.00041961669921875,
"rewards/weighted_rejected": 0.0061210631392896175,
"step": 130
},
{
"epoch": 0.07501674480910918,
"grad_norm": 19.93072509765625,
"learning_rate": 7.433155080213903e-07,
"logits/chosen": -2.53515625,
"logits/rejected": -2.5311522483825684,
"logps/chosen": -198.80313110351562,
"logps/rejected": -211.29061889648438,
"logps/weighted_chosen": -0.5693206787109375,
"logps/weighted_rejected": -0.592010498046875,
"loss": 0.694,
"rewards/accuracies": 0.543749988079071,
"rewards/chosen": -4.0048828125,
"rewards/margins": 2.3238282203674316,
"rewards/rejected": -6.327929496765137,
"rewards/weighted_accuracies": 0.518750011920929,
"rewards/weighted_chosen": -0.017972564324736595,
"rewards/weighted_margins": 0.005100250244140625,
"rewards/weighted_rejected": -0.02306365966796875,
"step": 140
},
{
"epoch": 0.08037508372404555,
"grad_norm": 26.39690399169922,
"learning_rate": 7.967914438502673e-07,
"logits/chosen": -2.509326219558716,
"logits/rejected": -2.508056640625,
"logps/chosen": -214.25,
"logps/rejected": -237.84530639648438,
"logps/weighted_chosen": -0.5889526605606079,
"logps/weighted_rejected": -0.585980236530304,
"loss": 0.7007,
"rewards/accuracies": 0.546875,
"rewards/chosen": -3.867382764816284,
"rewards/margins": 2.546581983566284,
"rewards/rejected": -6.4140625,
"rewards/weighted_accuracies": 0.550000011920929,
"rewards/weighted_chosen": -0.041257478296756744,
"rewards/weighted_margins": -0.005215453915297985,
"rewards/weighted_rejected": -0.03600044175982475,
"step": 150
},
{
"epoch": 0.08573342263898191,
"grad_norm": 25.74460792541504,
"learning_rate": 8.502673796791443e-07,
"logits/chosen": -2.433154344558716,
"logits/rejected": -2.414794921875,
"logps/chosen": -200.64999389648438,
"logps/rejected": -229.2624969482422,
"logps/weighted_chosen": -0.589752197265625,
"logps/weighted_rejected": -0.612139880657196,
"loss": 0.6933,
"rewards/accuracies": 0.5406249761581421,
"rewards/chosen": -6.800390720367432,
"rewards/margins": 0.5577148199081421,
"rewards/rejected": -7.36279296875,
"rewards/weighted_accuracies": 0.5218750238418579,
"rewards/weighted_chosen": -0.05205688625574112,
"rewards/weighted_margins": 0.006626891903579235,
"rewards/weighted_rejected": -0.05869255214929581,
"step": 160
},
{
"epoch": 0.09109176155391828,
"grad_norm": 24.530479431152344,
"learning_rate": 9.037433155080213e-07,
"logits/chosen": -2.5284180641174316,
"logits/rejected": -2.5411133766174316,
"logps/chosen": -204.92031860351562,
"logps/rejected": -217.046875,
"logps/weighted_chosen": -0.600720226764679,
"logps/weighted_rejected": -0.6278625726699829,
"loss": 0.6787,
"rewards/accuracies": 0.6031249761581421,
"rewards/chosen": -7.408789157867432,
"rewards/margins": 4.765625,
"rewards/rejected": -12.178515434265137,
"rewards/weighted_accuracies": 0.612500011920929,
"rewards/weighted_chosen": -0.05335540696978569,
"rewards/weighted_margins": 0.03420104831457138,
"rewards/weighted_rejected": -0.08757324516773224,
"step": 170
},
{
"epoch": 0.09645010046885466,
"grad_norm": 17.329570770263672,
"learning_rate": 9.572192513368984e-07,
"logits/chosen": -2.638134717941284,
"logits/rejected": -2.6302247047424316,
"logps/chosen": -219.87655639648438,
"logps/rejected": -240.1531219482422,
"logps/weighted_chosen": -0.637438952922821,
"logps/weighted_rejected": -0.700207531452179,
"loss": 0.682,
"rewards/accuracies": 0.5843750238418579,
"rewards/chosen": -10.272656440734863,
"rewards/margins": 4.742578029632568,
"rewards/rejected": -15.017578125,
"rewards/weighted_accuracies": 0.581250011920929,
"rewards/weighted_chosen": -0.09931640326976776,
"rewards/weighted_margins": 0.0342864990234375,
"rewards/weighted_rejected": -0.13359375298023224,
"step": 180
},
{
"epoch": 0.10180843938379103,
"grad_norm": 40.06588363647461,
"learning_rate": 9.999965031204306e-07,
"logits/chosen": -2.677978515625,
"logits/rejected": -2.667724609375,
"logps/chosen": -217.265625,
"logps/rejected": -217.9968719482422,
"logps/weighted_chosen": -0.713732898235321,
"logps/weighted_rejected": -0.745983898639679,
"loss": 0.6972,
"rewards/accuracies": 0.559374988079071,
"rewards/chosen": -16.422557830810547,
"rewards/margins": 3.5362305641174316,
"rewards/rejected": -19.955078125,
"rewards/weighted_accuracies": 0.578125,
"rewards/weighted_chosen": -0.1743011474609375,
"rewards/weighted_margins": 0.01756439171731472,
"rewards/weighted_rejected": -0.19183501601219177,
"step": 190
},
{
"epoch": 0.1071667782987274,
"grad_norm": 26.449254989624023,
"learning_rate": 9.998741174712533e-07,
"logits/chosen": -2.758007764816284,
"logits/rejected": -2.752197265625,
"logps/chosen": -251.4640655517578,
"logps/rejected": -266.7734375,
"logps/weighted_chosen": -0.721697986125946,
"logps/weighted_rejected": -0.7427002191543579,
"loss": 0.6956,
"rewards/accuracies": 0.543749988079071,
"rewards/chosen": -23.8759765625,
"rewards/margins": 3.2962889671325684,
"rewards/rejected": -27.181835174560547,
"rewards/weighted_accuracies": 0.5,
"rewards/weighted_chosen": -0.14485931396484375,
"rewards/weighted_margins": 0.00913848914206028,
"rewards/weighted_rejected": -0.153900146484375,
"step": 200
},
{
"epoch": 0.11252511721366376,
"grad_norm": 25.405826568603516,
"learning_rate": 9.995769367531952e-07,
"logits/chosen": -2.8843750953674316,
"logits/rejected": -2.8621582984924316,
"logps/chosen": -249.39999389648438,
"logps/rejected": -249.5625,
"logps/weighted_chosen": -0.6254638433456421,
"logps/weighted_rejected": -0.677978515625,
"loss": 0.6845,
"rewards/accuracies": 0.5531250238418579,
"rewards/chosen": -31.167186737060547,
"rewards/margins": 2.1792969703674316,
"rewards/rejected": -33.351173400878906,
"rewards/weighted_accuracies": 0.528124988079071,
"rewards/weighted_chosen": -0.11136627197265625,
"rewards/weighted_margins": 0.02726135216653347,
"rewards/weighted_rejected": -0.1386566162109375,
"step": 210
},
{
"epoch": 0.11788345612860013,
"grad_norm": 22.92214012145996,
"learning_rate": 9.991050648838675e-07,
"logits/chosen": -3.027539014816284,
"logits/rejected": -3.0234375,
"logps/chosen": -249.0265655517578,
"logps/rejected": -264.1703186035156,
"logps/weighted_chosen": -0.72125244140625,
"logps/weighted_rejected": -0.7577880620956421,
"loss": 0.6887,
"rewards/accuracies": 0.5687500238418579,
"rewards/chosen": -44.36796951293945,
"rewards/margins": 2.092968702316284,
"rewards/rejected": -46.47265625,
"rewards/weighted_accuracies": 0.5218750238418579,
"rewards/weighted_chosen": -0.15605469048023224,
"rewards/weighted_margins": 0.022979736328125,
"rewards/weighted_rejected": -0.17911987006664276,
"step": 220
},
{
"epoch": 0.12324179504353651,
"grad_norm": 18.913394927978516,
"learning_rate": 9.98458666866564e-07,
"logits/chosen": -2.95361328125,
"logits/rejected": -2.9599609375,
"logps/chosen": -252.1765594482422,
"logps/rejected": -273.97186279296875,
"logps/weighted_chosen": -0.7530273199081421,
"logps/weighted_rejected": -0.7951415777206421,
"loss": 0.6805,
"rewards/accuracies": 0.5874999761581421,
"rewards/chosen": -55.478126525878906,
"rewards/margins": 9.685155868530273,
"rewards/rejected": -65.15702819824219,
"rewards/weighted_accuracies": 0.546875,
"rewards/weighted_chosen": -0.202239990234375,
"rewards/weighted_margins": 0.04668273776769638,
"rewards/weighted_rejected": -0.248951718211174,
"step": 230
},
{
"epoch": 0.12860013395847286,
"grad_norm": 18.727752685546875,
"learning_rate": 9.97637968732563e-07,
"logits/chosen": -2.9483885765075684,
"logits/rejected": -2.934033155441284,
"logps/chosen": -264.1937561035156,
"logps/rejected": -281.26873779296875,
"logps/weighted_chosen": -0.7586425542831421,
"logps/weighted_rejected": -0.777172863483429,
"loss": 0.7011,
"rewards/accuracies": 0.565625011920929,
"rewards/chosen": -60.15625,
"rewards/margins": 7.514062404632568,
"rewards/rejected": -67.6640625,
"rewards/weighted_accuracies": 0.528124988079071,
"rewards/weighted_chosen": -0.22589111328125,
"rewards/weighted_margins": 0.0014251709217205644,
"rewards/weighted_rejected": -0.22734984755516052,
"step": 240
},
{
"epoch": 0.13395847287340926,
"grad_norm": 17.17424774169922,
"learning_rate": 9.966432574620906e-07,
"logits/chosen": -2.994140625,
"logits/rejected": -2.995654344558716,
"logps/chosen": -250.4296875,
"logps/rejected": -282.0859375,
"logps/weighted_chosen": -0.707995593547821,
"logps/weighted_rejected": -0.76885986328125,
"loss": 0.6897,
"rewards/accuracies": 0.590624988079071,
"rewards/chosen": -55.573829650878906,
"rewards/margins": 7.142187595367432,
"rewards/rejected": -62.73515701293945,
"rewards/weighted_accuracies": 0.528124988079071,
"rewards/weighted_chosen": -0.18481139838695526,
"rewards/weighted_margins": 0.01998596265912056,
"rewards/weighted_rejected": -0.20475539565086365,
"step": 250
},
{
"epoch": 0.13931681178834562,
"grad_norm": 18.692733764648438,
"learning_rate": 9.954748808839674e-07,
"logits/chosen": -2.887255907058716,
"logits/rejected": -2.898193359375,
"logps/chosen": -280.46875,
"logps/rejected": -280.6640625,
"logps/weighted_chosen": -0.7401489019393921,
"logps/weighted_rejected": -0.776611328125,
"loss": 0.6869,
"rewards/accuracies": 0.574999988079071,
"rewards/chosen": -53.8828125,
"rewards/margins": 5.946484565734863,
"rewards/rejected": -59.81757736206055,
"rewards/weighted_accuracies": 0.53125,
"rewards/weighted_chosen": -0.18103942275047302,
"rewards/weighted_margins": 0.02960510179400444,
"rewards/weighted_rejected": -0.21054688096046448,
"step": 260
},
{
"epoch": 0.144675150703282,
"grad_norm": 16.58523941040039,
"learning_rate": 9.941332475539824e-07,
"logits/chosen": -2.8628907203674316,
"logits/rejected": -2.87939453125,
"logps/chosen": -267.26251220703125,
"logps/rejected": -306.49688720703125,
"logps/weighted_chosen": -0.7835937738418579,
"logps/weighted_rejected": -0.8957885503768921,
"loss": 0.6712,
"rewards/accuracies": 0.625,
"rewards/chosen": -57.08281326293945,
"rewards/margins": 12.824609756469727,
"rewards/rejected": -69.8984375,
"rewards/weighted_accuracies": 0.596875011920929,
"rewards/weighted_chosen": -0.2109939604997635,
"rewards/weighted_margins": 0.07429657131433487,
"rewards/weighted_rejected": -0.2853530943393707,
"step": 270
},
{
"epoch": 0.15003348961821836,
"grad_norm": 23.606122970581055,
"learning_rate": 9.926188266120295e-07,
"logits/chosen": -2.9261717796325684,
"logits/rejected": -2.9302000999450684,
"logps/chosen": -252.35781860351562,
"logps/rejected": -270.3921813964844,
"logps/weighted_chosen": -0.757946789264679,
"logps/weighted_rejected": -0.8466796875,
"loss": 0.6669,
"rewards/accuracies": 0.5874999761581421,
"rewards/chosen": -56.718360900878906,
"rewards/margins": 10.275781631469727,
"rewards/rejected": -66.990234375,
"rewards/weighted_accuracies": 0.606249988079071,
"rewards/weighted_chosen": -0.2161209136247635,
"rewards/weighted_margins": 0.08074493706226349,
"rewards/weighted_rejected": -0.2967941164970398,
"step": 280
},
{
"epoch": 0.15539182853315472,
"grad_norm": 25.32463836669922,
"learning_rate": 9.909321476180591e-07,
"logits/chosen": -2.9480957984924316,
"logits/rejected": -2.9312500953674316,
"logps/chosen": -281.33123779296875,
"logps/rejected": -317.3374938964844,
"logps/weighted_chosen": -0.870861828327179,
"logps/weighted_rejected": -0.9639892578125,
"loss": 0.6868,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -78.2476577758789,
"rewards/margins": 12.921093940734863,
"rewards/rejected": -91.15312194824219,
"rewards/weighted_accuracies": 0.534375011920929,
"rewards/weighted_chosen": -0.32458800077438354,
"rewards/weighted_margins": 0.04551086574792862,
"rewards/weighted_rejected": -0.37019044160842896,
"step": 290
},
{
"epoch": 0.1607501674480911,
"grad_norm": 21.94011116027832,
"learning_rate": 9.890738003669027e-07,
"logits/chosen": -3.082812547683716,
"logits/rejected": -3.089111328125,
"logps/chosen": -288.62811279296875,
"logps/rejected": -311.38751220703125,
"logps/weighted_chosen": -0.955078125,
"logps/weighted_rejected": -1.0396239757537842,
"loss": 0.687,
"rewards/accuracies": 0.5625,
"rewards/chosen": -93.6109390258789,
"rewards/margins": 9.185546875,
"rewards/rejected": -102.78202819824219,
"rewards/weighted_accuracies": 0.559374988079071,
"rewards/weighted_chosen": -0.40625,
"rewards/weighted_margins": 0.07210388034582138,
"rewards/weighted_rejected": -0.478240966796875,
"step": 300
},
{
"epoch": 0.16610850636302746,
"grad_norm": 25.38623046875,
"learning_rate": 9.870444346820348e-07,
"logits/chosen": -3.173828125,
"logits/rejected": -3.147998094558716,
"logps/chosen": -297.15936279296875,
"logps/rejected": -320.50311279296875,
"logps/weighted_chosen": -0.89190673828125,
"logps/weighted_rejected": -0.923266589641571,
"loss": 0.6987,
"rewards/accuracies": 0.6187499761581421,
"rewards/chosen": -83.2046890258789,
"rewards/margins": 12.489062309265137,
"rewards/rejected": -95.6937484741211,
"rewards/weighted_accuracies": 0.581250011920929,
"rewards/weighted_chosen": -0.33665162324905396,
"rewards/weighted_margins": 0.00975952111184597,
"rewards/weighted_rejected": -0.34637755155563354,
"step": 310
},
{
"epoch": 0.17146684527796383,
"grad_norm": 17.789676666259766,
"learning_rate": 9.848447601883433e-07,
"logits/chosen": -3.236083984375,
"logits/rejected": -3.2168946266174316,
"logps/chosen": -267.09844970703125,
"logps/rejected": -310.3031311035156,
"logps/weighted_chosen": -0.8338623046875,
"logps/weighted_rejected": -0.918957531452179,
"loss": 0.6805,
"rewards/accuracies": 0.612500011920929,
"rewards/chosen": -70.2816390991211,
"rewards/margins": 19.578907012939453,
"rewards/rejected": -89.81953430175781,
"rewards/weighted_accuracies": 0.590624988079071,
"rewards/weighted_chosen": -0.2871032655239105,
"rewards/weighted_margins": 0.06427612155675888,
"rewards/weighted_rejected": -0.3513946533203125,
"step": 320
},
{
"epoch": 0.1768251841929002,
"grad_norm": 24.225061416625977,
"learning_rate": 9.824755460639899e-07,
"logits/chosen": -3.1796875,
"logits/rejected": -3.179150342941284,
"logps/chosen": -265.12030029296875,
"logps/rejected": -308.19842529296875,
"logps/weighted_chosen": -0.833850085735321,
"logps/weighted_rejected": -0.8948730230331421,
"loss": 0.6836,
"rewards/accuracies": 0.6156250238418579,
"rewards/chosen": -73.27030944824219,
"rewards/margins": 16.651952743530273,
"rewards/rejected": -89.87968444824219,
"rewards/weighted_accuracies": 0.5531250238418579,
"rewards/weighted_chosen": -0.31196290254592896,
"rewards/weighted_margins": 0.04750366136431694,
"rewards/weighted_rejected": -0.35954588651657104,
"step": 330
},
{
"epoch": 0.18218352310783656,
"grad_norm": 19.119462966918945,
"learning_rate": 9.799376207714444e-07,
"logits/chosen": -3.1602540016174316,
"logits/rejected": -3.154492139816284,
"logps/chosen": -277.9624938964844,
"logps/rejected": -296.7093811035156,
"logps/weighted_chosen": -0.902087390422821,
"logps/weighted_rejected": -0.974658191204071,
"loss": 0.6727,
"rewards/accuracies": 0.5406249761581421,
"rewards/chosen": -77.0367202758789,
"rewards/margins": 9.926562309265137,
"rewards/rejected": -86.9476547241211,
"rewards/weighted_accuracies": 0.559374988079071,
"rewards/weighted_chosen": -0.334219366312027,
"rewards/weighted_margins": 0.06939239799976349,
"rewards/weighted_rejected": -0.4036361575126648,
"step": 340
},
{
"epoch": 0.18754186202277295,
"grad_norm": 22.155990600585938,
"learning_rate": 9.772318717677903e-07,
"logits/chosen": -3.130908250808716,
"logits/rejected": -3.132128953933716,
"logps/chosen": -288.41094970703125,
"logps/rejected": -313.0843811035156,
"logps/weighted_chosen": -0.9267333745956421,
"logps/weighted_rejected": -1.040771484375,
"loss": 0.6683,
"rewards/accuracies": 0.625,
"rewards/chosen": -88.24609375,
"rewards/margins": 18.800390243530273,
"rewards/rejected": -107.046875,
"rewards/weighted_accuracies": 0.6156250238418579,
"rewards/weighted_chosen": -0.39622801542282104,
"rewards/weighted_margins": 0.08176574856042862,
"rewards/weighted_rejected": -0.47796326875686646,
"step": 350
},
{
"epoch": 0.19290020093770932,
"grad_norm": 26.421598434448242,
"learning_rate": 9.743592451943998e-07,
"logits/chosen": -3.229687452316284,
"logits/rejected": -3.228222608566284,
"logps/chosen": -327.7328186035156,
"logps/rejected": -361.0015563964844,
"logps/weighted_chosen": -0.990966796875,
"logps/weighted_rejected": -1.057519555091858,
"loss": 0.6726,
"rewards/accuracies": 0.590624988079071,
"rewards/chosen": -116.12812805175781,
"rewards/margins": 18.506641387939453,
"rewards/rejected": -134.654296875,
"rewards/weighted_accuracies": 0.543749988079071,
"rewards/weighted_chosen": -0.4190429747104645,
"rewards/weighted_margins": 0.08066101372241974,
"rewards/weighted_rejected": -0.499612420797348,
"step": 360
},
{
"epoch": 0.1982585398526457,
"grad_norm": 19.50035858154297,
"learning_rate": 9.713207455460892e-07,
"logits/chosen": -3.184375047683716,
"logits/rejected": -3.209912061691284,
"logps/chosen": -326.4937438964844,
"logps/rejected": -343.4593811035156,
"logps/weighted_chosen": -1.00506591796875,
"logps/weighted_rejected": -1.092797875404358,
"loss": 0.6762,
"rewards/accuracies": 0.5687500238418579,
"rewards/chosen": -120.6937484741211,
"rewards/margins": 6.901953220367432,
"rewards/rejected": -127.60469055175781,
"rewards/weighted_accuracies": 0.5531250238418579,
"rewards/weighted_chosen": -0.4098266661167145,
"rewards/weighted_margins": 0.08511047065258026,
"rewards/weighted_rejected": -0.495004266500473,
"step": 370
},
{
"epoch": 0.20361687876758205,
"grad_norm": 17.131114959716797,
"learning_rate": 9.681174353198686e-07,
"logits/chosen": -3.2510743141174316,
"logits/rejected": -3.23828125,
"logps/chosen": -330.140625,
"logps/rejected": -368.10467529296875,
"logps/weighted_chosen": -0.9594482183456421,
"logps/weighted_rejected": -1.056616187095642,
"loss": 0.6719,
"rewards/accuracies": 0.6156250238418579,
"rewards/chosen": -119.54219055175781,
"rewards/margins": 20.350391387939453,
"rewards/rejected": -139.82461547851562,
"rewards/weighted_accuracies": 0.574999988079071,
"rewards/weighted_chosen": -0.415090948343277,
"rewards/weighted_margins": 0.086700439453125,
"rewards/weighted_rejected": -0.501788318157196,
"step": 380
},
{
"epoch": 0.20897521768251842,
"grad_norm": 38.318756103515625,
"learning_rate": 9.647504346434103e-07,
"logits/chosen": -3.445019483566284,
"logits/rejected": -3.4564452171325684,
"logps/chosen": -371.9156188964844,
"logps/rejected": -389.43438720703125,
"logps/weighted_chosen": -1.062597632408142,
"logps/weighted_rejected": -1.165747046470642,
"loss": 0.6656,
"rewards/accuracies": 0.578125,
"rewards/chosen": -150.7859344482422,
"rewards/margins": 19.3828125,
"rewards/rejected": -170.19686889648438,
"rewards/weighted_accuracies": 0.578125,
"rewards/weighted_chosen": -0.49369508028030396,
"rewards/weighted_margins": 0.09916381537914276,
"rewards/weighted_rejected": -0.5928589105606079,
"step": 390
},
{
"epoch": 0.2143335565974548,
"grad_norm": 37.75257873535156,
"learning_rate": 9.612209208833646e-07,
"logits/chosen": -3.4891600608825684,
"logits/rejected": -3.4913086891174316,
"logps/chosen": -373.2046813964844,
"logps/rejected": -387.3968811035156,
"logps/weighted_chosen": -1.159387230873108,
"logps/weighted_rejected": -1.2549560070037842,
"loss": 0.6802,
"rewards/accuracies": 0.6031249761581421,
"rewards/chosen": -173.24063110351562,
"rewards/margins": 12.66796875,
"rewards/rejected": -185.9562530517578,
"rewards/weighted_accuracies": 0.5531250238418579,
"rewards/weighted_chosen": -0.619030773639679,
"rewards/weighted_margins": 0.07501526176929474,
"rewards/weighted_rejected": -0.694287121295929,
"step": 400
},
{
"epoch": 0.21969189551239116,
"grad_norm": 62.04045486450195,
"learning_rate": 9.5753012823366e-07,
"logits/chosen": -3.5625,
"logits/rejected": -3.555224657058716,
"logps/chosen": -392.5843811035156,
"logps/rejected": -429.76873779296875,
"logps/weighted_chosen": -1.109643578529358,
"logps/weighted_rejected": -1.2004883289337158,
"loss": 0.6903,
"rewards/accuracies": 0.6156250238418579,
"rewards/chosen": -180.3562469482422,
"rewards/margins": 26.413280487060547,
"rewards/rejected": -206.75936889648438,
"rewards/weighted_accuracies": 0.5375000238418579,
"rewards/weighted_chosen": -0.5794036984443665,
"rewards/weighted_margins": 0.08088378608226776,
"rewards/weighted_rejected": -0.660137951374054,
"step": 410
},
{
"epoch": 0.22505023442732752,
"grad_norm": 29.432497024536133,
"learning_rate": 9.536793472839324e-07,
"logits/chosen": -3.7582030296325684,
"logits/rejected": -3.77783203125,
"logps/chosen": -413.109375,
"logps/rejected": -459.015625,
"logps/weighted_chosen": -1.0903198719024658,
"logps/weighted_rejected": -1.1765625476837158,
"loss": 0.6833,
"rewards/accuracies": 0.6031249761581421,
"rewards/chosen": -182.9140625,
"rewards/margins": 33.59453201293945,
"rewards/rejected": -216.5343780517578,
"rewards/weighted_accuracies": 0.590624988079071,
"rewards/weighted_chosen": -0.528045654296875,
"rewards/weighted_margins": 0.08157958835363388,
"rewards/weighted_rejected": -0.609661877155304,
"step": 420
},
{
"epoch": 0.2304085733422639,
"grad_norm": 34.61969757080078,
"learning_rate": 9.496699245682351e-07,
"logits/chosen": -3.876757860183716,
"logits/rejected": -3.873339891433716,
"logps/chosen": -411.23748779296875,
"logps/rejected": -474.3687438964844,
"logps/weighted_chosen": -1.104589819908142,
"logps/weighted_rejected": -1.18817138671875,
"loss": 0.6878,
"rewards/accuracies": 0.6156250238418579,
"rewards/chosen": -190.3468780517578,
"rewards/margins": 43.072265625,
"rewards/rejected": -233.43124389648438,
"rewards/weighted_accuracies": 0.5531250238418579,
"rewards/weighted_chosen": -0.543225109577179,
"rewards/weighted_margins": 0.05898132175207138,
"rewards/weighted_rejected": -0.6022491455078125,
"step": 430
},
{
"epoch": 0.23576691225720026,
"grad_norm": 25.689876556396484,
"learning_rate": 9.455032620941839e-07,
"logits/chosen": -3.893261671066284,
"logits/rejected": -3.866406202316284,
"logps/chosen": -390.3531188964844,
"logps/rejected": -474.80780029296875,
"logps/weighted_chosen": -1.0908081531524658,
"logps/weighted_rejected": -1.19012451171875,
"loss": 0.6673,
"rewards/accuracies": 0.621874988079071,
"rewards/chosen": -196.52188110351562,
"rewards/margins": 52.310546875,
"rewards/rejected": -248.7218780517578,
"rewards/weighted_accuracies": 0.5718749761581421,
"rewards/weighted_chosen": -0.555419921875,
"rewards/weighted_margins": 0.09650878608226776,
"rewards/weighted_rejected": -0.651898205280304,
"step": 440
},
{
"epoch": 0.24112525117213665,
"grad_norm": 54.604705810546875,
"learning_rate": 9.411808168527066e-07,
"logits/chosen": -3.88916015625,
"logits/rejected": -3.8941407203674316,
"logps/chosen": -411.3031311035156,
"logps/rejected": -472.203125,
"logps/weighted_chosen": -1.082067847251892,
"logps/weighted_rejected": -1.1839110851287842,
"loss": 0.6816,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -205.703125,
"rewards/margins": 46.81132888793945,
"rewards/rejected": -252.5343780517578,
"rewards/weighted_accuracies": 0.590624988079071,
"rewards/weighted_chosen": -0.554046630859375,
"rewards/weighted_margins": 0.07362671196460724,
"rewards/weighted_rejected": -0.6277831792831421,
"step": 450
},
{
"epoch": 0.24648359008707302,
"grad_norm": 26.44339942932129,
"learning_rate": 9.367041003085648e-07,
"logits/chosen": -3.8785157203674316,
"logits/rejected": -3.866894483566284,
"logps/chosen": -523.2593994140625,
"logps/rejected": -523.2750244140625,
"logps/weighted_chosen": -1.2444579601287842,
"logps/weighted_rejected": -1.3485107421875,
"loss": 0.6775,
"rewards/accuracies": 0.5625,
"rewards/chosen": -304.52813720703125,
"rewards/margins": 4.764062404632568,
"rewards/rejected": -309.3812561035156,
"rewards/weighted_accuracies": 0.590624988079071,
"rewards/weighted_chosen": -0.6716858148574829,
"rewards/weighted_margins": 0.08544921875,
"rewards/weighted_rejected": -0.757153332233429,
"step": 460
},
{
"epoch": 0.25184192900200936,
"grad_norm": 28.781457901000977,
"learning_rate": 9.320746778718274e-07,
"logits/chosen": -4.042578220367432,
"logits/rejected": -4.053515434265137,
"logps/chosen": -444.046875,
"logps/rejected": -514.0531005859375,
"logps/weighted_chosen": -1.281274437904358,
"logps/weighted_rejected": -1.4283936023712158,
"loss": 0.6526,
"rewards/accuracies": 0.6187499761581421,
"rewards/chosen": -256.4828186035156,
"rewards/margins": 53.646095275878906,
"rewards/rejected": -310.15313720703125,
"rewards/weighted_accuracies": 0.596875011920929,
"rewards/weighted_chosen": -0.714306652545929,
"rewards/weighted_margins": 0.13612060248851776,
"rewards/weighted_rejected": -0.8504638671875,
"step": 470
},
{
"epoch": 0.2572002679169457,
"grad_norm": 58.31265640258789,
"learning_rate": 9.272941683504808e-07,
"logits/chosen": -4.060449123382568,
"logits/rejected": -4.083691596984863,
"logps/chosen": -564.7312622070312,
"logps/rejected": -582.9124755859375,
"logps/weighted_chosen": -1.305688500404358,
"logps/weighted_rejected": -1.470678687095642,
"loss": 0.6627,
"rewards/accuracies": 0.5687500238418579,
"rewards/chosen": -346.97344970703125,
"rewards/margins": 15.486719131469727,
"rewards/rejected": -362.28594970703125,
"rewards/weighted_accuracies": 0.596875011920929,
"rewards/weighted_chosen": -0.7694152593612671,
"rewards/weighted_margins": 0.14488831162452698,
"rewards/weighted_rejected": -0.9143127202987671,
"step": 480
},
{
"epoch": 0.2625586068318821,
"grad_norm": 28.284282684326172,
"learning_rate": 9.223642433843679e-07,
"logits/chosen": -3.8697266578674316,
"logits/rejected": -3.857128858566284,
"logps/chosen": -500.0718688964844,
"logps/rejected": -534.6281127929688,
"logps/weighted_chosen": -1.1939697265625,
"logps/weighted_rejected": -1.323388695716858,
"loss": 0.6567,
"rewards/accuracies": 0.621874988079071,
"rewards/chosen": -298.09063720703125,
"rewards/margins": 29.775781631469727,
"rewards/rejected": -328.0640563964844,
"rewards/weighted_accuracies": 0.59375,
"rewards/weighted_chosen": -0.631103515625,
"rewards/weighted_margins": 0.13239136338233948,
"rewards/weighted_rejected": -0.763378918170929,
"step": 490
},
{
"epoch": 0.2679169457468185,
"grad_norm": 32.619327545166016,
"learning_rate": 9.172866268606513e-07,
"logits/chosen": -3.828320264816284,
"logits/rejected": -3.82373046875,
"logps/chosen": -527.2687377929688,
"logps/rejected": -588.8343505859375,
"logps/weighted_chosen": -1.291723608970642,
"logps/weighted_rejected": -1.4711182117462158,
"loss": 0.6559,
"rewards/accuracies": 0.625,
"rewards/chosen": -312.36248779296875,
"rewards/margins": 48.47734451293945,
"rewards/rejected": -360.8890686035156,
"rewards/weighted_accuracies": 0.59375,
"rewards/weighted_chosen": -0.727709949016571,
"rewards/weighted_margins": 0.145904541015625,
"rewards/weighted_rejected": -0.8731445074081421,
"step": 500
},
{
"epoch": 0.2679169457468185,
"eval_logits/chosen": -3.994471788406372,
"eval_logits/rejected": -4.005757808685303,
"eval_logps/chosen": -530.9920043945312,
"eval_logps/rejected": -592.9425659179688,
"eval_logps/weighted_chosen": -1.3731945753097534,
"eval_logps/weighted_rejected": -1.5277278423309326,
"eval_loss": 0.6660141348838806,
"eval_rewards/accuracies": 0.5874499082565308,
"eval_rewards/chosen": -328.0513916015625,
"eval_rewards/margins": 48.42765426635742,
"eval_rewards/rejected": -376.4005432128906,
"eval_rewards/weighted_accuracies": 0.5901201367378235,
"eval_rewards/weighted_chosen": -0.8177139759063721,
"eval_rewards/weighted_margins": 0.13399113714694977,
"eval_rewards/weighted_rejected": -0.9517050385475159,
"eval_runtime": 1889.7494,
"eval_samples_per_second": 1.584,
"eval_steps_per_second": 0.396,
"step": 500
},
{
"epoch": 0.2732752846617549,
"grad_norm": 36.31141662597656,
"learning_rate": 9.120630943110077e-07,
"logits/chosen": -4.06494140625,
"logits/rejected": -4.080273628234863,
"logps/chosen": -546.140625,
"logps/rejected": -658.2093505859375,
"logps/weighted_chosen": -1.464135766029358,
"logps/weighted_rejected": -1.644873023033142,
"loss": 0.6696,
"rewards/accuracies": 0.6187499761581421,
"rewards/chosen": -355.78594970703125,
"rewards/margins": 91.36328125,
"rewards/rejected": -447.06561279296875,
"rewards/weighted_accuracies": 0.6187499761581421,
"rewards/weighted_chosen": -0.9257446527481079,
"rewards/weighted_margins": 0.15970154106616974,
"rewards/weighted_rejected": -1.085382103919983,
"step": 510
},
{
"epoch": 0.27863362357669125,
"grad_norm": 35.842044830322266,
"learning_rate": 9.066954722907638e-07,
"logits/chosen": -4.306347846984863,
"logits/rejected": -4.325781345367432,
"logps/chosen": -809.96875,
"logps/rejected": -865.5875244140625,
"logps/weighted_chosen": -1.7161865234375,
"logps/weighted_rejected": -1.847753882408142,
"loss": 0.6806,
"rewards/accuracies": 0.609375,
"rewards/chosen": -589.8624877929688,
"rewards/margins": 49.43046951293945,
"rewards/rejected": -639.21875,
"rewards/weighted_accuracies": 0.581250011920929,
"rewards/weighted_chosen": -1.1625487804412842,
"rewards/weighted_margins": 0.126800537109375,
"rewards/weighted_rejected": -1.289453148841858,
"step": 520
},
{
"epoch": 0.2839919624916276,
"grad_norm": 35.471004486083984,
"learning_rate": 9.01185637740189e-07,
"logits/chosen": -4.103613376617432,
"logits/rejected": -4.137109279632568,
"logps/chosen": -668.6375122070312,
"logps/rejected": -714.5843505859375,
"logps/weighted_chosen": -1.6459472179412842,
"logps/weighted_rejected": -1.787329077720642,
"loss": 0.6869,
"rewards/accuracies": 0.5687500238418579,
"rewards/chosen": -462.6499938964844,
"rewards/margins": 39.26484298706055,
"rewards/rejected": -501.71875,
"rewards/weighted_accuracies": 0.565625011920929,
"rewards/weighted_chosen": -1.0964477062225342,
"rewards/weighted_margins": 0.13413086533546448,
"rewards/weighted_rejected": -1.2302672863006592,
"step": 530
},
{
"epoch": 0.289350301406564,
"grad_norm": 22.4523868560791,
"learning_rate": 8.955355173281707e-07,
"logits/chosen": -3.6913084983825684,
"logits/rejected": -3.6981444358825684,
"logps/chosen": -521.2437744140625,
"logps/rejected": -594.4593505859375,
"logps/weighted_chosen": -1.2616455554962158,
"logps/weighted_rejected": -1.4534423351287842,
"loss": 0.6462,
"rewards/accuracies": 0.625,
"rewards/chosen": -323.74688720703125,
"rewards/margins": 58.17578125,
"rewards/rejected": -381.83282470703125,
"rewards/weighted_accuracies": 0.612500011920929,
"rewards/weighted_chosen": -0.7408813238143921,
"rewards/weighted_margins": 0.15880127251148224,
"rewards/weighted_rejected": -0.89971923828125,
"step": 540
},
{
"epoch": 0.29470864032150035,
"grad_norm": 53.5711784362793,
"learning_rate": 8.897470867785002e-07,
"logits/chosen": -3.683789014816284,
"logits/rejected": -3.6670899391174316,
"logps/chosen": -669.2562255859375,
"logps/rejected": -769.59375,
"logps/weighted_chosen": -1.6165649890899658,
"logps/weighted_rejected": -1.8085448741912842,
"loss": 0.6561,
"rewards/accuracies": 0.653124988079071,
"rewards/chosen": -457.6937561035156,
"rewards/margins": 77.93281555175781,
"rewards/rejected": -535.5343627929688,
"rewards/weighted_accuracies": 0.606249988079071,
"rewards/weighted_chosen": -1.055206298828125,
"rewards/weighted_margins": 0.17585143446922302,
"rewards/weighted_rejected": -1.230871558189392,
"step": 550
},
{
"epoch": 0.3000669792364367,
"grad_norm": 49.29216384887695,
"learning_rate": 8.838223701790055e-07,
"logits/chosen": -3.755566358566284,
"logits/rejected": -3.782177686691284,
"logps/chosen": -865.2218627929688,
"logps/rejected": -996.9562377929688,
"logps/weighted_chosen": -1.795434594154358,
"logps/weighted_rejected": -2.018505811691284,
"loss": 0.648,
"rewards/accuracies": 0.653124988079071,
"rewards/chosen": -615.8656005859375,
"rewards/margins": 98.30000305175781,
"rewards/rejected": -714.1875,
"rewards/weighted_accuracies": 0.606249988079071,
"rewards/weighted_chosen": -1.202539086341858,
"rewards/weighted_margins": 0.18366089463233948,
"rewards/weighted_rejected": -1.3866698741912842,
"step": 560
},
{
"epoch": 0.3054253181513731,
"grad_norm": 50.555702209472656,
"learning_rate": 8.777634392737718e-07,
"logits/chosen": -3.892578125,
"logits/rejected": -3.912402391433716,
"logps/chosen": -780.0984497070312,
"logps/rejected": -869.4921875,
"logps/weighted_chosen": -1.9073486328125,
"logps/weighted_rejected": -2.052197217941284,
"loss": 0.6896,
"rewards/accuracies": 0.6656249761581421,
"rewards/chosen": -588.2203369140625,
"rewards/margins": 76.50312805175781,
"rewards/rejected": -664.6828002929688,
"rewards/weighted_accuracies": 0.5687500238418579,
"rewards/weighted_chosen": -1.3397338390350342,
"rewards/weighted_margins": 0.13177490234375,
"rewards/weighted_rejected": -1.472259521484375,
"step": 570
},
{
"epoch": 0.31078365706630945,
"grad_norm": 29.3996524810791,
"learning_rate": 8.71572412738697e-07,
"logits/chosen": -3.8792967796325684,
"logits/rejected": -3.884960889816284,
"logps/chosen": -791.7625122070312,
"logps/rejected": -904.4625244140625,
"logps/weighted_chosen": -1.7471435070037842,
"logps/weighted_rejected": -1.941796898841858,
"loss": 0.662,
"rewards/accuracies": 0.653124988079071,
"rewards/chosen": -584.3125,
"rewards/margins": 103.31172180175781,
"rewards/rejected": -687.5281372070312,
"rewards/weighted_accuracies": 0.612500011920929,
"rewards/weighted_chosen": -1.219970703125,
"rewards/weighted_margins": 0.17452391982078552,
"rewards/weighted_rejected": -1.3947632312774658,
"step": 580
},
{
"epoch": 0.3161419959812458,
"grad_norm": 62.64840316772461,
"learning_rate": 8.652514554406387e-07,
"logits/chosen": -3.75537109375,
"logits/rejected": -3.780566453933716,
"logps/chosen": -766.2156372070312,
"logps/rejected": -876.4124755859375,
"logps/weighted_chosen": -1.5784912109375,
"logps/weighted_rejected": -1.806738257408142,
"loss": 0.6446,
"rewards/accuracies": 0.628125011920929,
"rewards/chosen": -561.0531005859375,
"rewards/margins": 91.5546875,
"rewards/rejected": -652.6312255859375,
"rewards/weighted_accuracies": 0.590624988079071,
"rewards/weighted_chosen": -1.047125220298767,
"rewards/weighted_margins": 0.18712158501148224,
"rewards/weighted_rejected": -1.234130859375,
"step": 590
},
{
"epoch": 0.3215003348961822,
"grad_norm": 37.892852783203125,
"learning_rate": 8.588027776804058e-07,
"logits/chosen": -4.111035346984863,
"logits/rejected": -4.132616996765137,
"logps/chosen": -954.3093872070312,
"logps/rejected": -1287.2874755859375,
"logps/weighted_chosen": -1.9705321788787842,
"logps/weighted_rejected": -2.290820360183716,
"loss": 0.6267,
"rewards/accuracies": 0.65625,
"rewards/chosen": -767.3687744140625,
"rewards/margins": 301.2730407714844,
"rewards/rejected": -1068.2515869140625,
"rewards/weighted_accuracies": 0.6468750238418579,
"rewards/weighted_chosen": -1.4273865222930908,
"rewards/weighted_margins": 0.2945709228515625,
"rewards/weighted_rejected": -1.72186279296875,
"step": 600
},
{
"epoch": 0.32685867381111855,
"grad_norm": 65.1230239868164,
"learning_rate": 8.522286344198657e-07,
"logits/chosen": -4.340234279632568,
"logits/rejected": -4.395800590515137,
"logps/chosen": -1246.2874755859375,
"logps/rejected": -1507.5562744140625,
"logps/weighted_chosen": -2.40673828125,
"logps/weighted_rejected": -2.769335985183716,
"loss": 0.6572,
"rewards/accuracies": 0.606249988079071,
"rewards/chosen": -1049.7093505859375,
"rewards/margins": 244.8625030517578,
"rewards/rejected": -1294.578125,
"rewards/weighted_accuracies": 0.628125011920929,
"rewards/weighted_chosen": -1.853796362876892,
"rewards/weighted_margins": 0.33940428495407104,
"rewards/weighted_rejected": -2.1933226585388184,
"step": 610
},
{
"epoch": 0.3322170127260549,
"grad_norm": 29.82529640197754,
"learning_rate": 8.455313244934324e-07,
"logits/chosen": -4.122754096984863,
"logits/rejected": -4.150390625,
"logps/chosen": -1042.171875,
"logps/rejected": -1219.421875,
"logps/weighted_chosen": -2.124218702316284,
"logps/weighted_rejected": -2.362744092941284,
"loss": 0.657,
"rewards/accuracies": 0.6156250238418579,
"rewards/chosen": -852.1281127929688,
"rewards/margins": 169.6007843017578,
"rewards/rejected": -1021.8499755859375,
"rewards/weighted_accuracies": 0.659375011920929,
"rewards/weighted_chosen": -1.55322265625,
"rewards/weighted_margins": 0.25230103731155396,
"rewards/weighted_rejected": -1.8051025867462158,
"step": 620
},
{
"epoch": 0.3375753516409913,
"grad_norm": 37.24652862548828,
"learning_rate": 8.38713189804215e-07,
"logits/chosen": -3.853710889816284,
"logits/rejected": -3.8750977516174316,
"logps/chosen": -865.640625,
"logps/rejected": -959.234375,
"logps/weighted_chosen": -1.747460961341858,
"logps/weighted_rejected": -1.9114258289337158,
"loss": 0.6618,
"rewards/accuracies": 0.5874999761581421,
"rewards/chosen": -657.8468627929688,
"rewards/margins": 90.248046875,
"rewards/rejected": -747.7999877929688,
"rewards/weighted_accuracies": 0.606249988079071,
"rewards/weighted_chosen": -1.1962401866912842,
"rewards/weighted_margins": 0.14417724311351776,
"rewards/weighted_rejected": -1.3405029773712158,
"step": 630
},
{
"epoch": 0.34293369055592765,
"grad_norm": 121.3833999633789,
"learning_rate": 8.317766145051057e-07,
"logits/chosen": -4.014550685882568,
"logits/rejected": -4.060400485992432,
"logps/chosen": -1148.1968994140625,
"logps/rejected": -1330.6624755859375,
"logps/weighted_chosen": -2.155810594558716,
"logps/weighted_rejected": -2.436279296875,
"loss": 0.6494,
"rewards/accuracies": 0.621874988079071,
"rewards/chosen": -948.5125122070312,
"rewards/margins": 166.63320922851562,
"rewards/rejected": -1115.5250244140625,
"rewards/weighted_accuracies": 0.653124988079071,
"rewards/weighted_chosen": -1.6124267578125,
"rewards/weighted_margins": 0.25153809785842896,
"rewards/weighted_rejected": -1.8647949695587158,
"step": 640
},
{
"epoch": 0.348292029470864,
"grad_norm": 80.68321990966797,
"learning_rate": 8.247240241650917e-07,
"logits/chosen": -4.319043159484863,
"logits/rejected": -4.347851753234863,
"logps/chosen": -1620.6312255859375,
"logps/rejected": -1847.887451171875,
"logps/weighted_chosen": -2.8577637672424316,
"logps/weighted_rejected": -3.252002000808716,
"loss": 0.6113,
"rewards/accuracies": 0.6187499761581421,
"rewards/chosen": -1411.9906005859375,
"rewards/margins": 225.7062530517578,
"rewards/rejected": -1637.65625,
"rewards/weighted_accuracies": 0.675000011920929,
"rewards/weighted_chosen": -2.3053221702575684,
"rewards/weighted_margins": 0.40415650606155396,
"rewards/weighted_rejected": -2.710131883621216,
"step": 650
},
{
"epoch": 0.3536503683858004,
"grad_norm": 84.14534759521484,
"learning_rate": 8.175578849210894e-07,
"logits/chosen": -4.389941215515137,
"logits/rejected": -4.449902534484863,
"logps/chosen": -1601.6624755859375,
"logps/rejected": -1866.512451171875,
"logps/weighted_chosen": -2.7519774436950684,
"logps/weighted_rejected": -3.2392578125,
"loss": 0.5898,
"rewards/accuracies": 0.6343749761581421,
"rewards/chosen": -1391.625,
"rewards/margins": 258.7890625,
"rewards/rejected": -1651.15625,
"rewards/weighted_accuracies": 0.6656249761581421,
"rewards/weighted_chosen": -2.201342821121216,
"rewards/weighted_margins": 0.4779724180698395,
"rewards/weighted_rejected": -2.679443359375,
"step": 660
},
{
"epoch": 0.35900870730073675,
"grad_norm": 39.30218505859375,
"learning_rate": 8.102807026155873e-07,
"logits/chosen": -4.323534965515137,
"logits/rejected": -4.3671875,
"logps/chosen": -1877.043701171875,
"logps/rejected": -2115.753173828125,
"logps/weighted_chosen": -3.111035108566284,
"logps/weighted_rejected": -3.53955078125,
"loss": 0.6449,
"rewards/accuracies": 0.609375,
"rewards/chosen": -1668.4375,
"rewards/margins": 238.2890625,
"rewards/rejected": -1906.921875,
"rewards/weighted_accuracies": 0.659375011920929,
"rewards/weighted_chosen": -2.5547852516174316,
"rewards/weighted_margins": 0.4420410096645355,
"rewards/weighted_rejected": -2.9969725608825684,
"step": 670
},
{
"epoch": 0.3643670462156731,
"grad_norm": 49.729007720947266,
"learning_rate": 8.028950219204099e-07,
"logits/chosen": -3.9090819358825684,
"logits/rejected": -3.9154295921325684,
"logps/chosen": -1154.637451171875,
"logps/rejected": -1364.612548828125,
"logps/weighted_chosen": -2.158496141433716,
"logps/weighted_rejected": -2.4698243141174316,
"loss": 0.6288,
"rewards/accuracies": 0.6156250238418579,
"rewards/chosen": -945.5218505859375,
"rewards/margins": 190.38906860351562,
"rewards/rejected": -1135.9156494140625,
"rewards/weighted_accuracies": 0.6468750238418579,
"rewards/weighted_chosen": -1.6334960460662842,
"rewards/weighted_margins": 0.27833253145217896,
"rewards/weighted_rejected": -1.9119141101837158,
"step": 680
},
{
"epoch": 0.3697253851306095,
"grad_norm": 51.215213775634766,
"learning_rate": 7.954034254469e-07,
"logits/chosen": -3.8226561546325684,
"logits/rejected": -3.847851514816284,
"logps/chosen": -1110.840576171875,
"logps/rejected": -1422.15625,
"logps/weighted_chosen": -2.027099609375,
"logps/weighted_rejected": -2.3323731422424316,
"loss": 0.6364,
"rewards/accuracies": 0.643750011920929,
"rewards/chosen": -902.6968994140625,
"rewards/margins": 299.96954345703125,
"rewards/rejected": -1202.6640625,
"rewards/weighted_accuracies": 0.628125011920929,
"rewards/weighted_chosen": -1.477014183998108,
"rewards/weighted_margins": 0.2761291563510895,
"rewards/weighted_rejected": -1.753381371498108,
"step": 690
},
{
"epoch": 0.3750837240455459,
"grad_norm": 75.38801574707031,
"learning_rate": 7.878085328428368e-07,
"logits/chosen": -3.9351563453674316,
"logits/rejected": -3.950878858566284,
"logps/chosen": -1454.487548828125,
"logps/rejected": -1619.846923828125,
"logps/weighted_chosen": -2.549121141433716,
"logps/weighted_rejected": -2.860668897628784,
"loss": 0.6352,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -1233.6937255859375,
"rewards/margins": 161.18359375,
"rewards/rejected": -1394.9234619140625,
"rewards/weighted_accuracies": 0.6625000238418579,
"rewards/weighted_chosen": -1.9811279773712158,
"rewards/weighted_margins": 0.295562744140625,
"rewards/weighted_rejected": -2.277050733566284,
"step": 700
},
{
"epoch": 0.3804420629604823,
"grad_norm": 47.24467468261719,
"learning_rate": 7.801129998764014e-07,
"logits/chosen": -4.013671875,
"logits/rejected": -4.046777248382568,
"logps/chosen": -1444.081298828125,
"logps/rejected": -1578.3359375,
"logps/weighted_chosen": -2.5169920921325684,
"logps/weighted_rejected": -2.897216796875,
"loss": 0.6181,
"rewards/accuracies": 0.606249988079071,
"rewards/chosen": -1228.768798828125,
"rewards/margins": 132.5593719482422,
"rewards/rejected": -1361.543701171875,
"rewards/weighted_accuracies": 0.653124988079071,
"rewards/weighted_chosen": -1.987060546875,
"rewards/weighted_margins": 0.3536315858364105,
"rewards/weighted_rejected": -2.3398680686950684,
"step": 710
},
{
"epoch": 0.38580040187541864,
"grad_norm": 48.87159729003906,
"learning_rate": 7.723195175075135e-07,
"logits/chosen": -3.8736329078674316,
"logits/rejected": -3.909863233566284,
"logps/chosen": -1098.1187744140625,
"logps/rejected": -1359.378173828125,
"logps/weighted_chosen": -2.243359327316284,
"logps/weighted_rejected": -2.5437254905700684,
"loss": 0.6502,
"rewards/accuracies": 0.671875,
"rewards/chosen": -911.5625,
"rewards/margins": 230.42422485351562,
"rewards/rejected": -1141.9156494140625,
"rewards/weighted_accuracies": 0.6343749761581421,
"rewards/weighted_chosen": -1.6927001476287842,
"rewards/weighted_margins": 0.2898498475551605,
"rewards/weighted_rejected": -1.982519507408142,
"step": 720
},
{
"epoch": 0.391158740790355,
"grad_norm": 46.05818176269531,
"learning_rate": 7.644308109468609e-07,
"logits/chosen": -3.802734375,
"logits/rejected": -3.804003953933716,
"logps/chosen": -1067.456298828125,
"logps/rejected": -1283.940673828125,
"logps/weighted_chosen": -2.007983446121216,
"logps/weighted_rejected": -2.337109327316284,
"loss": 0.6166,
"rewards/accuracies": 0.625,
"rewards/chosen": -863.9593505859375,
"rewards/margins": 216.0382843017578,
"rewards/rejected": -1079.768798828125,
"rewards/weighted_accuracies": 0.6625000238418579,
"rewards/weighted_chosen": -1.4805786609649658,
"rewards/weighted_margins": 0.3025146424770355,
"rewards/weighted_rejected": -1.7832763195037842,
"step": 730
},
{
"epoch": 0.3965170797052914,
"grad_norm": 55.81437683105469,
"learning_rate": 7.564496387029531e-07,
"logits/chosen": -4.071093559265137,
"logits/rejected": -4.120214939117432,
"logps/chosen": -1393.003173828125,
"logps/rejected": -1683.3375244140625,
"logps/weighted_chosen": -2.53515625,
"logps/weighted_rejected": -2.929443359375,
"loss": 0.6251,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -1187.206298828125,
"rewards/margins": 281.1695251464844,
"rewards/rejected": -1468.278076171875,
"rewards/weighted_accuracies": 0.621874988079071,
"rewards/weighted_chosen": -1.9933350086212158,
"rewards/weighted_margins": 0.38404542207717896,
"rewards/weighted_rejected": -2.378063917160034,
"step": 740
},
{
"epoch": 0.40187541862022774,
"grad_norm": 70.06423950195312,
"learning_rate": 7.483787916175306e-07,
"logits/chosen": -4.136034965515137,
"logits/rejected": -4.169043064117432,
"logps/chosen": -1402.487548828125,
"logps/rejected": -1632.203125,
"logps/weighted_chosen": -2.555224657058716,
"logps/weighted_rejected": -2.929248094558716,
"loss": 0.6341,
"rewards/accuracies": 0.628125011920929,
"rewards/chosen": -1186.949951171875,
"rewards/margins": 219.8984375,
"rewards/rejected": -1406.706298828125,
"rewards/weighted_accuracies": 0.65625,
"rewards/weighted_chosen": -2.005810499191284,
"rewards/weighted_margins": 0.36415404081344604,
"rewards/weighted_rejected": -2.3702635765075684,
"step": 750
},
{
"epoch": 0.4072337575351641,
"grad_norm": 70.14170837402344,
"learning_rate": 7.402210918896689e-07,
"logits/chosen": -4.0087890625,
"logits/rejected": -4.039746284484863,
"logps/chosen": -1279.356201171875,
"logps/rejected": -1483.34375,
"logps/weighted_chosen": -2.5243163108825684,
"logps/weighted_rejected": -2.907763719558716,
"loss": 0.624,
"rewards/accuracies": 0.653124988079071,
"rewards/chosen": -1078.0062255859375,
"rewards/margins": 200.91171264648438,
"rewards/rejected": -1279.0718994140625,
"rewards/weighted_accuracies": 0.671875,
"rewards/weighted_chosen": -1.9707763195037842,
"rewards/weighted_margins": 0.36932373046875,
"rewards/weighted_rejected": -2.340039014816284,
"step": 760
},
{
"epoch": 0.4125920964501005,
"grad_norm": 42.04954147338867,
"learning_rate": 7.31979392088917e-07,
"logits/chosen": -4.008203029632568,
"logits/rejected": -4.041699409484863,
"logps/chosen": -1906.106201171875,
"logps/rejected": -2324.574951171875,
"logps/weighted_chosen": -3.2344727516174316,
"logps/weighted_rejected": -3.656445264816284,
"loss": 0.6464,
"rewards/accuracies": 0.6343749761581421,
"rewards/chosen": -1681.8343505859375,
"rewards/margins": 403.7398376464844,
"rewards/rejected": -2085.55615234375,
"rewards/weighted_accuracies": 0.653124988079071,
"rewards/weighted_chosen": -2.6626219749450684,
"rewards/weighted_margins": 0.41541749238967896,
"rewards/weighted_rejected": -3.079028367996216,
"step": 770
},
{
"epoch": 0.41795043536503684,
"grad_norm": 60.82102966308594,
"learning_rate": 7.236565741578162e-07,
"logits/chosen": -3.94140625,
"logits/rejected": -3.970410108566284,
"logps/chosen": -1391.1500244140625,
"logps/rejected": -1722.4625244140625,
"logps/weighted_chosen": -2.736621141433716,
"logps/weighted_rejected": -3.1324706077575684,
"loss": 0.6103,
"rewards/accuracies": 0.6187499761581421,
"rewards/chosen": -1210.75,
"rewards/margins": 304.0874938964844,
"rewards/rejected": -1515.206298828125,
"rewards/weighted_accuracies": 0.6781250238418579,
"rewards/weighted_chosen": -2.1943359375,
"rewards/weighted_margins": 0.36918336153030396,
"rewards/weighted_rejected": -2.5628418922424316,
"step": 780
},
{
"epoch": 0.4233087742799732,
"grad_norm": 57.50530242919922,
"learning_rate": 7.152555484041475e-07,
"logits/chosen": -3.9559569358825684,
"logits/rejected": -3.97998046875,
"logps/chosen": -2113.05615234375,
"logps/rejected": -2432.09375,
"logps/weighted_chosen": -3.4646973609924316,
"logps/weighted_rejected": -3.819775342941284,
"loss": 0.665,
"rewards/accuracies": 0.6156250238418579,
"rewards/chosen": -1889.3656005859375,
"rewards/margins": 320.93048095703125,
"rewards/rejected": -2209.949951171875,
"rewards/weighted_accuracies": 0.606249988079071,
"rewards/weighted_chosen": -2.8851561546325684,
"rewards/weighted_margins": 0.3280883729457855,
"rewards/weighted_rejected": -3.2130370140075684,
"step": 790
},
{
"epoch": 0.4286671131949096,
"grad_norm": 66.44673919677734,
"learning_rate": 7.067792524832603e-07,
"logits/chosen": -3.87353515625,
"logits/rejected": -3.876953125,
"logps/chosen": -1906.387451171875,
"logps/rejected": -2406.27490234375,
"logps/weighted_chosen": -3.2735352516174316,
"logps/weighted_rejected": -3.800585985183716,
"loss": 0.5965,
"rewards/accuracies": 0.659375011920929,
"rewards/chosen": -1719.875,
"rewards/margins": 478.44219970703125,
"rewards/rejected": -2197.93115234375,
"rewards/weighted_accuracies": 0.668749988079071,
"rewards/weighted_chosen": -2.7166504859924316,
"rewards/weighted_margins": 0.4966064393520355,
"rewards/weighted_rejected": -3.2127442359924316,
"step": 800
},
{
"epoch": 0.43402545210984594,
"grad_norm": 46.25076675415039,
"learning_rate": 6.982306503708387e-07,
"logits/chosen": -3.9263672828674316,
"logits/rejected": -3.9388670921325684,
"logps/chosen": -2048.074951171875,
"logps/rejected": -2567.831298828125,
"logps/weighted_chosen": -3.1552734375,
"logps/weighted_rejected": -3.6259765625,
"loss": 0.6162,
"rewards/accuracies": 0.6187499761581421,
"rewards/chosen": -1836.296875,
"rewards/margins": 497.3296813964844,
"rewards/rejected": -2333.86865234375,
"rewards/weighted_accuracies": 0.643750011920929,
"rewards/weighted_chosen": -2.6156249046325684,
"rewards/weighted_margins": 0.43950194120407104,
"rewards/weighted_rejected": -3.054370164871216,
"step": 810
},
{
"epoch": 0.4393837910247823,
"grad_norm": 49.19358825683594,
"learning_rate": 6.896127313264642e-07,
"logits/chosen": -3.8050780296325684,
"logits/rejected": -3.810253858566284,
"logps/chosen": -1704.9000244140625,
"logps/rejected": -2074.984375,
"logps/weighted_chosen": -2.6828856468200684,
"logps/weighted_rejected": -3.102343797683716,
"loss": 0.6091,
"rewards/accuracies": 0.612500011920929,
"rewards/chosen": -1484.278076171875,
"rewards/margins": 349.5960998535156,
"rewards/rejected": -1833.125,
"rewards/weighted_accuracies": 0.668749988079071,
"rewards/weighted_chosen": -2.126025438308716,
"rewards/weighted_margins": 0.3870300352573395,
"rewards/weighted_rejected": -2.5131592750549316,
"step": 820
},
{
"epoch": 0.4447421299397187,
"grad_norm": 60.69411849975586,
"learning_rate": 6.809285088483361e-07,
"logits/chosen": -3.877636671066284,
"logits/rejected": -3.895800828933716,
"logps/chosen": -2122.018798828125,
"logps/rejected": -2766.80615234375,
"logps/weighted_chosen": -3.4732422828674316,
"logps/weighted_rejected": -4.013476371765137,
"loss": 0.589,
"rewards/accuracies": 0.625,
"rewards/chosen": -1923.1968994140625,
"rewards/margins": 628.5843505859375,
"rewards/rejected": -2552.231201171875,
"rewards/weighted_accuracies": 0.6812499761581421,
"rewards/weighted_chosen": -2.9307618141174316,
"rewards/weighted_margins": 0.514050304889679,
"rewards/weighted_rejected": -3.4447021484375,
"step": 830
},
{
"epoch": 0.45010046885465504,
"grad_norm": 41.00968551635742,
"learning_rate": 6.721810196195174e-07,
"logits/chosen": -3.9095702171325684,
"logits/rejected": -3.932910203933716,
"logps/chosen": -2189.53125,
"logps/rejected": -2920.6875,
"logps/weighted_chosen": -3.7161622047424316,
"logps/weighted_rejected": -4.303027153015137,
"loss": 0.5847,
"rewards/accuracies": 0.653124988079071,
"rewards/chosen": -1995.074951171875,
"rewards/margins": 712.40625,
"rewards/rejected": -2706.63134765625,
"rewards/weighted_accuracies": 0.671875,
"rewards/weighted_chosen": -3.1697754859924316,
"rewards/weighted_margins": 0.5831848382949829,
"rewards/weighted_rejected": -3.75390625,
"step": 840
},
{
"epoch": 0.4554588077695914,
"grad_norm": 35.70737838745117,
"learning_rate": 6.633733224460737e-07,
"logits/chosen": -3.749706983566284,
"logits/rejected": -3.76806640625,
"logps/chosen": -1573.5374755859375,
"logps/rejected": -2201.737548828125,
"logps/weighted_chosen": -2.6988282203674316,
"logps/weighted_rejected": -3.203369140625,
"loss": 0.5615,
"rewards/accuracies": 0.675000011920929,
"rewards/chosen": -1371.199951171875,
"rewards/margins": 607.2484130859375,
"rewards/rejected": -1978.456298828125,
"rewards/weighted_accuracies": 0.706250011920929,
"rewards/weighted_chosen": -2.152587890625,
"rewards/weighted_margins": 0.48516845703125,
"rewards/weighted_rejected": -2.6385498046875,
"step": 850
},
{
"epoch": 0.4608171466845278,
"grad_norm": 73.64376831054688,
"learning_rate": 6.545084971874736e-07,
"logits/chosen": -3.9317383766174316,
"logits/rejected": -3.955859422683716,
"logps/chosen": -2716.58740234375,
"logps/rejected": -3423.94384765625,
"logps/weighted_chosen": -4.377148628234863,
"logps/weighted_rejected": -5.021288871765137,
"loss": 0.583,
"rewards/accuracies": 0.6468750238418579,
"rewards/chosen": -2526.012451171875,
"rewards/margins": 687.0515747070312,
"rewards/rejected": -3212.625,
"rewards/weighted_accuracies": 0.699999988079071,
"rewards/weighted_chosen": -3.8341307640075684,
"rewards/weighted_margins": 0.6270385980606079,
"rewards/weighted_rejected": -4.462500095367432,
"step": 860
},
{
"epoch": 0.46617548559946415,
"grad_norm": 54.467132568359375,
"learning_rate": 6.455896436796313e-07,
"logits/chosen": -3.9200196266174316,
"logits/rejected": -3.932910203933716,
"logps/chosen": -3902.112548828125,
"logps/rejected": -4818.7626953125,
"logps/weighted_chosen": -5.9072265625,
"logps/weighted_rejected": -6.536718845367432,
"loss": 0.6265,
"rewards/accuracies": 0.596875011920929,
"rewards/chosen": -3707.512451171875,
"rewards/margins": 898.1077880859375,
"rewards/rejected": -4605.3125,
"rewards/weighted_accuracies": 0.671875,
"rewards/weighted_chosen": -5.350195407867432,
"rewards/weighted_margins": 0.618518054485321,
"rewards/weighted_rejected": -5.966601371765137,
"step": 870
},
{
"epoch": 0.4715338245144005,
"grad_norm": 58.258487701416016,
"learning_rate": 6.3661988065096e-07,
"logits/chosen": -3.58642578125,
"logits/rejected": -3.57080078125,
"logps/chosen": -2729.300048828125,
"logps/rejected": -3327.96875,
"logps/weighted_chosen": -4.433691501617432,
"logps/weighted_rejected": -4.90625,
"loss": 0.5904,
"rewards/accuracies": 0.6312500238418579,
"rewards/chosen": -2531.981201171875,
"rewards/margins": 591.4375,
"rewards/rejected": -3122.64990234375,
"rewards/weighted_accuracies": 0.671875,
"rewards/weighted_chosen": -3.89013671875,
"rewards/weighted_margins": 0.4582275450229645,
"rewards/weighted_rejected": -4.347851753234863,
"step": 880
},
{
"epoch": 0.4768921634293369,
"grad_norm": 47.88425827026367,
"learning_rate": 6.276023446318213e-07,
"logits/chosen": -3.6739258766174316,
"logits/rejected": -3.68017578125,
"logps/chosen": -2387.86865234375,
"logps/rejected": -2776.918701171875,
"logps/weighted_chosen": -3.9383788108825684,
"logps/weighted_rejected": -4.536913871765137,
"loss": 0.5673,
"rewards/accuracies": 0.690625011920929,
"rewards/chosen": -2187.94677734375,
"rewards/margins": 379.22186279296875,
"rewards/rejected": -2567.375,
"rewards/weighted_accuracies": 0.746874988079071,
"rewards/weighted_chosen": -3.3819336891174316,
"rewards/weighted_margins": 0.5686889886856079,
"rewards/weighted_rejected": -3.9501953125,
"step": 890
},
{
"epoch": 0.4822505023442733,
"grad_norm": 66.42100524902344,
"learning_rate": 6.185401888577487e-07,
"logits/chosen": -3.93115234375,
"logits/rejected": -3.9623045921325684,
"logps/chosen": -2366.125,
"logps/rejected": -3888.5625,
"logps/weighted_chosen": -3.873828172683716,
"logps/weighted_rejected": -4.703711032867432,
"loss": 0.5527,
"rewards/accuracies": 0.6656249761581421,
"rewards/chosen": -2176.15625,
"rewards/margins": 1495.7890625,
"rewards/rejected": -3671.89990234375,
"rewards/weighted_accuracies": 0.6812499761581421,
"rewards/weighted_chosen": -3.337158203125,
"rewards/weighted_margins": 0.809399425983429,
"rewards/weighted_rejected": -4.147070407867432,
"step": 900
},
{
"epoch": 0.48760884125920967,
"grad_norm": 55.539608001708984,
"learning_rate": 6.094365821668307e-07,
"logits/chosen": -3.839550733566284,
"logits/rejected": -3.8837890625,
"logps/chosen": -1797.953125,
"logps/rejected": -2595.637451171875,
"logps/weighted_chosen": -2.9969725608825684,
"logps/weighted_rejected": -3.619189500808716,
"loss": 0.5987,
"rewards/accuracies": 0.6812499761581421,
"rewards/chosen": -1617.2125244140625,
"rewards/margins": 776.7586059570312,
"rewards/rejected": -2393.4765625,
"rewards/weighted_accuracies": 0.684374988079071,
"rewards/weighted_chosen": -2.473498582839966,
"rewards/weighted_margins": 0.5845702886581421,
"rewards/weighted_rejected": -3.0577635765075684,
"step": 910
},
{
"epoch": 0.49296718017414604,
"grad_norm": 45.468833923339844,
"learning_rate": 6.002947078916364e-07,
"logits/chosen": -3.766308546066284,
"logits/rejected": -3.795214891433716,
"logps/chosen": -1605.893798828125,
"logps/rejected": -2098.050048828125,
"logps/weighted_chosen": -2.6654295921325684,
"logps/weighted_rejected": -3.1903319358825684,
"loss": 0.5926,
"rewards/accuracies": 0.640625,
"rewards/chosen": -1392.9937744140625,
"rewards/margins": 466.5140686035156,
"rewards/rejected": -1860.2249755859375,
"rewards/weighted_accuracies": 0.6812499761581421,
"rewards/weighted_chosen": -2.1322021484375,
"rewards/weighted_margins": 0.5062255859375,
"rewards/weighted_rejected": -2.6379637718200684,
"step": 920
},
{
"epoch": 0.4983255190890824,
"grad_norm": 40.75309753417969,
"learning_rate": 5.911177627460738e-07,
"logits/chosen": -3.7451171875,
"logits/rejected": -3.78662109375,
"logps/chosen": -1467.300048828125,
"logps/rejected": -2011.8218994140625,
"logps/weighted_chosen": -2.669970750808716,
"logps/weighted_rejected": -3.193066358566284,
"loss": 0.6164,
"rewards/accuracies": 0.6468750238418579,
"rewards/chosen": -1248.856201171875,
"rewards/margins": 535.0531005859375,
"rewards/rejected": -1784.440673828125,
"rewards/weighted_accuracies": 0.6656249761581421,
"rewards/weighted_chosen": -2.093676805496216,
"rewards/weighted_margins": 0.520129382610321,
"rewards/weighted_rejected": -2.613964796066284,
"step": 930
},
{
"epoch": 0.5036838580040187,
"grad_norm": 56.69281768798828,
"learning_rate": 5.819089557075688e-07,
"logits/chosen": -3.661328077316284,
"logits/rejected": -3.725781202316284,
"logps/chosen": -1287.831298828125,
"logps/rejected": -1729.199951171875,
"logps/weighted_chosen": -2.5040040016174316,
"logps/weighted_rejected": -3.1222167015075684,
"loss": 0.5475,
"rewards/accuracies": 0.6781250238418579,
"rewards/chosen": -1086.8968505859375,
"rewards/margins": 426.1015625,
"rewards/rejected": -1512.9375,
"rewards/weighted_accuracies": 0.737500011920929,
"rewards/weighted_chosen": -1.9895751476287842,
"rewards/weighted_margins": 0.597973644733429,
"rewards/weighted_rejected": -2.5875487327575684,
"step": 940
},
{
"epoch": 0.5090421969189551,
"grad_norm": 198.49781799316406,
"learning_rate": 5.726715068949564e-07,
"logits/chosen": -3.756054639816284,
"logits/rejected": -3.787304639816284,
"logps/chosen": -2703.00634765625,
"logps/rejected": -2929.199951171875,
"logps/weighted_chosen": -4.483007907867432,
"logps/weighted_rejected": -5.202538967132568,
"loss": 0.6233,
"rewards/accuracies": 0.6468750238418579,
"rewards/chosen": -2486.63134765625,
"rewards/margins": 220.6328125,
"rewards/rejected": -2707.15625,
"rewards/weighted_accuracies": 0.675000011920929,
"rewards/weighted_chosen": -3.906005859375,
"rewards/weighted_margins": 0.69610595703125,
"rewards/weighted_rejected": -4.601171970367432,
"step": 950
},
{
"epoch": 0.5144005358338914,
"grad_norm": 94.89183044433594,
"learning_rate": 5.634086464424742e-07,
"logits/chosen": -3.745898485183716,
"logits/rejected": -3.765429735183716,
"logps/chosen": -2798.706298828125,
"logps/rejected": -3669.106201171875,
"logps/weighted_chosen": -5.272851467132568,
"logps/weighted_rejected": -6.047753810882568,
"loss": 0.5864,
"rewards/accuracies": 0.653124988079071,
"rewards/chosen": -2610.640625,
"rewards/margins": 852.515625,
"rewards/rejected": -3462.06884765625,
"rewards/weighted_accuracies": 0.715624988079071,
"rewards/weighted_chosen": -4.717529296875,
"rewards/weighted_margins": 0.750353991985321,
"rewards/weighted_rejected": -5.467114448547363,
"step": 960
},
{
"epoch": 0.5197588747488279,
"grad_norm": 46.3341178894043,
"learning_rate": 5.54123613370256e-07,
"logits/chosen": -3.57080078125,
"logits/rejected": -3.592968702316284,
"logps/chosen": -2180.831298828125,
"logps/rejected": -2691.08740234375,
"logps/weighted_chosen": -4.2587890625,
"logps/weighted_rejected": -4.9375,
"loss": 0.562,
"rewards/accuracies": 0.6499999761581421,
"rewards/chosen": -1985.668701171875,
"rewards/margins": 503.3203125,
"rewards/rejected": -2488.737548828125,
"rewards/weighted_accuracies": 0.690625011920929,
"rewards/weighted_chosen": -3.743603467941284,
"rewards/weighted_margins": 0.6626037359237671,
"rewards/weighted_rejected": -4.405859470367432,
"step": 970
},
{
"epoch": 0.5251172136637642,
"grad_norm": 44.62388610839844,
"learning_rate": 5.448196544517167e-07,
"logits/chosen": -3.4229493141174316,
"logits/rejected": -3.43896484375,
"logps/chosen": -1911.8687744140625,
"logps/rejected": -2449.68115234375,
"logps/weighted_chosen": -3.70849609375,
"logps/weighted_rejected": -4.321630954742432,
"loss": 0.5762,
"rewards/accuracies": 0.668749988079071,
"rewards/chosen": -1712.768798828125,
"rewards/margins": 526.8828125,
"rewards/rejected": -2239.606201171875,
"rewards/weighted_accuracies": 0.659375011920929,
"rewards/weighted_chosen": -3.1871094703674316,
"rewards/weighted_margins": 0.5860961675643921,
"rewards/weighted_rejected": -3.773242235183716,
"step": 980
},
{
"epoch": 0.5304755525787006,
"grad_norm": 49.71703338623047,
"learning_rate": 5.355000230782267e-07,
"logits/chosen": -3.386523485183716,
"logits/rejected": -3.404003858566284,
"logps/chosen": -1865.487548828125,
"logps/rejected": -2300.90625,
"logps/weighted_chosen": -3.5951170921325684,
"logps/weighted_rejected": -4.100878715515137,
"loss": 0.602,
"rewards/accuracies": 0.625,
"rewards/chosen": -1660.425048828125,
"rewards/margins": 418.15234375,
"rewards/rejected": -2078.49072265625,
"rewards/weighted_accuracies": 0.684374988079071,
"rewards/weighted_chosen": -3.0364990234375,
"rewards/weighted_margins": 0.4848266541957855,
"rewards/weighted_rejected": -3.5208983421325684,
"step": 990
},
{
"epoch": 0.535833891493637,
"grad_norm": 54.951995849609375,
"learning_rate": 5.26167978121472e-07,
"logits/chosen": -3.4642577171325684,
"logits/rejected": -3.4644532203674316,
"logps/chosen": -2232.06884765625,
"logps/rejected": -2963.16259765625,
"logps/weighted_chosen": -4.036523342132568,
"logps/weighted_rejected": -4.75390625,
"loss": 0.5535,
"rewards/accuracies": 0.637499988079071,
"rewards/chosen": -2023.6187744140625,
"rewards/margins": 708.0984497070312,
"rewards/rejected": -2731.19384765625,
"rewards/weighted_accuracies": 0.7250000238418579,
"rewards/weighted_chosen": -3.497631788253784,
"rewards/weighted_margins": 0.680041491985321,
"rewards/weighted_rejected": -4.178027153015137,
"step": 1000
},
{
"epoch": 0.535833891493637,
"eval_logits/chosen": -3.549294948577881,
"eval_logits/rejected": -3.559162139892578,
"eval_logps/chosen": -2485.90380859375,
"eval_logps/rejected": -3043.268310546875,
"eval_logps/weighted_chosen": -4.452103137969971,
"eval_logps/weighted_rejected": -5.134407043457031,
"eval_loss": 0.5839847326278687,
"eval_rewards/accuracies": 0.6418558359146118,
"eval_rewards/chosen": -2282.807861328125,
"eval_rewards/margins": 543.8858642578125,
"eval_rewards/rejected": -2826.91064453125,
"eval_rewards/weighted_accuracies": 0.6829105615615845,
"eval_rewards/weighted_chosen": -3.896622657775879,
"eval_rewards/weighted_margins": 0.6617620587348938,
"eval_rewards/weighted_rejected": -4.558384418487549,
"eval_runtime": 2155.1416,
"eval_samples_per_second": 1.389,
"eval_steps_per_second": 0.348,
"step": 1000
},
{
"epoch": 0.5411922304085733,
"grad_norm": 79.4778060913086,
"learning_rate": 5.16826782793897e-07,
"logits/chosen": -3.510449171066284,
"logits/rejected": -3.534863233566284,
"logps/chosen": -2491.237548828125,
"logps/rejected": -3156.91259765625,
"logps/weighted_chosen": -4.569384574890137,
"logps/weighted_rejected": -5.425585746765137,
"loss": 0.5358,
"rewards/accuracies": 0.65625,
"rewards/chosen": -2292.84375,
"rewards/margins": 644.8546752929688,
"rewards/rejected": -2937.61865234375,
"rewards/weighted_accuracies": 0.71875,
"rewards/weighted_chosen": -4.032910346984863,
"rewards/weighted_margins": 0.826770007610321,
"rewards/weighted_rejected": -4.859228610992432,
"step": 1010
},
{
"epoch": 0.5465505693235098,
"grad_norm": 61.079864501953125,
"learning_rate": 5.074797035076318e-07,
"logits/chosen": -3.5478515625,
"logits/rejected": -3.567187547683716,
"logps/chosen": -2653.21240234375,
"logps/rejected": -3437.8125,
"logps/weighted_chosen": -4.941064357757568,
"logps/weighted_rejected": -5.822851657867432,
"loss": 0.5441,
"rewards/accuracies": 0.690625011920929,
"rewards/chosen": -2446.54052734375,
"rewards/margins": 765.1218872070312,
"rewards/rejected": -3211.3125,
"rewards/weighted_accuracies": 0.746874988079071,
"rewards/weighted_chosen": -4.367761135101318,
"rewards/weighted_margins": 0.8629516363143921,
"rewards/weighted_rejected": -5.231005668640137,
"step": 1020
},
{
"epoch": 0.5519089082384461,
"grad_norm": 47.398704528808594,
"learning_rate": 4.981300087322984e-07,
"logits/chosen": -3.4212889671325684,
"logits/rejected": -3.440722703933716,
"logps/chosen": -2479.63134765625,
"logps/rejected": -3158.550048828125,
"logps/weighted_chosen": -4.55029296875,
"logps/weighted_rejected": -5.351855278015137,
"loss": 0.5583,
"rewards/accuracies": 0.6937500238418579,
"rewards/chosen": -2293.41259765625,
"rewards/margins": 655.875,
"rewards/rejected": -2951.34375,
"rewards/weighted_accuracies": 0.715624988079071,
"rewards/weighted_chosen": -3.9914307594299316,
"rewards/weighted_margins": 0.7806396484375,
"rewards/weighted_rejected": -4.77197265625,
"step": 1030
},
{
"epoch": 0.5572672471533825,
"grad_norm": 63.054168701171875,
"learning_rate": 4.887809678520975e-07,
"logits/chosen": -3.3487305641174316,
"logits/rejected": -3.3558592796325684,
"logps/chosen": -2364.0625,
"logps/rejected": -2929.425048828125,
"logps/weighted_chosen": -4.341748237609863,
"logps/weighted_rejected": -5.078515529632568,
"loss": 0.5706,
"rewards/accuracies": 0.640625,
"rewards/chosen": -2166.137451171875,
"rewards/margins": 556.125,
"rewards/rejected": -2720.5625,
"rewards/weighted_accuracies": 0.7281249761581421,
"rewards/weighted_chosen": -3.79736328125,
"rewards/weighted_margins": 0.707104504108429,
"rewards/weighted_rejected": -4.505615234375,
"step": 1040
},
{
"epoch": 0.5626255860683188,
"grad_norm": 66.39554595947266,
"learning_rate": 4.794358500225781e-07,
"logits/chosen": -3.335742235183716,
"logits/rejected": -3.342968702316284,
"logps/chosen": -2968.112548828125,
"logps/rejected": -3496.21240234375,
"logps/weighted_chosen": -4.970947265625,
"logps/weighted_rejected": -5.766699314117432,
"loss": 0.5347,
"rewards/accuracies": 0.6812499761581421,
"rewards/chosen": -2753.925048828125,
"rewards/margins": 525.3312377929688,
"rewards/rejected": -3278.762451171875,
"rewards/weighted_accuracies": 0.737500011920929,
"rewards/weighted_chosen": -4.405615329742432,
"rewards/weighted_margins": 0.7887328863143921,
"rewards/weighted_rejected": -5.193163871765137,
"step": 1050
},
{
"epoch": 0.5679839249832552,
"grad_norm": 57.971885681152344,
"learning_rate": 4.700979230274829e-07,
"logits/chosen": -3.3427734375,
"logits/rejected": -3.34375,
"logps/chosen": -3232.53759765625,
"logps/rejected": -3996.81884765625,
"logps/weighted_chosen": -5.7021484375,
"logps/weighted_rejected": -6.397656440734863,
"loss": 0.5907,
"rewards/accuracies": 0.6187499761581421,
"rewards/chosen": -3012.375,
"rewards/margins": 746.2734375,
"rewards/rejected": -3758.175048828125,
"rewards/weighted_accuracies": 0.6968749761581421,
"rewards/weighted_chosen": -5.131152153015137,
"rewards/weighted_margins": 0.684863269329071,
"rewards/weighted_rejected": -5.815527439117432,
"step": 1060
},
{
"epoch": 0.5733422638981915,
"grad_norm": 48.712318420410156,
"learning_rate": 4.6077045213607755e-07,
"logits/chosen": -3.2083983421325684,
"logits/rejected": -3.209277391433716,
"logps/chosen": -2662.112548828125,
"logps/rejected": -3147.550048828125,
"logps/weighted_chosen": -4.856835842132568,
"logps/weighted_rejected": -5.518847465515137,
"loss": 0.5756,
"rewards/accuracies": 0.6468750238418579,
"rewards/chosen": -2463.800048828125,
"rewards/margins": 479.9437561035156,
"rewards/rejected": -2944.550048828125,
"rewards/weighted_accuracies": 0.699999988079071,
"rewards/weighted_chosen": -4.350390434265137,
"rewards/weighted_margins": 0.623095691204071,
"rewards/weighted_rejected": -4.972216606140137,
"step": 1070
},
{
"epoch": 0.578700602813128,
"grad_norm": 48.57645797729492,
"learning_rate": 4.514566989613559e-07,
"logits/chosen": -3.207324266433716,
"logits/rejected": -3.2095704078674316,
"logps/chosen": -2496.25634765625,
"logps/rejected": -3165.72509765625,
"logps/weighted_chosen": -4.495214939117432,
"logps/weighted_rejected": -5.168164253234863,
"loss": 0.5612,
"rewards/accuracies": 0.621874988079071,
"rewards/chosen": -2264.86865234375,
"rewards/margins": 653.3562622070312,
"rewards/rejected": -2917.50634765625,
"rewards/weighted_accuracies": 0.706250011920929,
"rewards/weighted_chosen": -3.945605516433716,
"rewards/weighted_margins": 0.647961437702179,
"rewards/weighted_rejected": -4.594580173492432,
"step": 1080
},
{
"epoch": 0.5840589417280643,
"grad_norm": 65.56918334960938,
"learning_rate": 4.4215992031952614e-07,
"logits/chosen": -3.2230467796325684,
"logits/rejected": -3.2120118141174316,
"logps/chosen": -2563.293701171875,
"logps/rejected": -3421.875,
"logps/weighted_chosen": -5.21826171875,
"logps/weighted_rejected": -5.978906154632568,
"loss": 0.5596,
"rewards/accuracies": 0.643750011920929,
"rewards/chosen": -2379.03759765625,
"rewards/margins": 820.65625,
"rewards/rejected": -3198.41259765625,
"rewards/weighted_accuracies": 0.7093750238418579,
"rewards/weighted_chosen": -4.678418159484863,
"rewards/weighted_margins": 0.7149902582168579,
"rewards/weighted_rejected": -5.3935546875,
"step": 1090
},
{
"epoch": 0.5894172806430007,
"grad_norm": 68.06615447998047,
"learning_rate": 4.328833670911724e-07,
"logits/chosen": -3.170605421066284,
"logits/rejected": -3.166699171066284,
"logps/chosen": -3257.112548828125,
"logps/rejected": -4018.00634765625,
"logps/weighted_chosen": -5.70556640625,
"logps/weighted_rejected": -6.41455078125,
"loss": 0.5918,
"rewards/accuracies": 0.590624988079071,
"rewards/chosen": -3049.296875,
"rewards/margins": 746.9202880859375,
"rewards/rejected": -3795.425048828125,
"rewards/weighted_accuracies": 0.671875,
"rewards/weighted_chosen": -5.149609565734863,
"rewards/weighted_margins": 0.6976073980331421,
"rewards/weighted_rejected": -5.848535060882568,
"step": 1100
},
{
"epoch": 0.594775619557937,
"grad_norm": 47.62029266357422,
"learning_rate": 4.236302830844931e-07,
"logits/chosen": -3.06396484375,
"logits/rejected": -3.0479493141174316,
"logps/chosen": -2672.137451171875,
"logps/rejected": -3189.27490234375,
"logps/weighted_chosen": -4.959668159484863,
"logps/weighted_rejected": -5.608105659484863,
"loss": 0.6,
"rewards/accuracies": 0.6781250238418579,
"rewards/chosen": -2458.83740234375,
"rewards/margins": 508.86407470703125,
"rewards/rejected": -2967.31884765625,
"rewards/weighted_accuracies": 0.706250011920929,
"rewards/weighted_chosen": -4.393310546875,
"rewards/weighted_margins": 0.6424316167831421,
"rewards/weighted_rejected": -5.03564453125,
"step": 1110
},
{
"epoch": 0.6001339584728734,
"grad_norm": 59.24471664428711,
"learning_rate": 4.144039039010124e-07,
"logits/chosen": -2.9786133766174316,
"logits/rejected": -2.9564452171325684,
"logps/chosen": -2158.97509765625,
"logps/rejected": -2407.59375,
"logps/weighted_chosen": -4.17529296875,
"logps/weighted_rejected": -4.825976371765137,
"loss": 0.546,
"rewards/accuracies": 0.640625,
"rewards/chosen": -1953.484375,
"rewards/margins": 245.921875,
"rewards/rejected": -2199.78759765625,
"rewards/weighted_accuracies": 0.753125011920929,
"rewards/weighted_chosen": -3.6058592796325684,
"rewards/weighted_margins": 0.63897705078125,
"rewards/weighted_rejected": -4.246069431304932,
"step": 1120
},
{
"epoch": 0.6054922973878097,
"grad_norm": 63.221038818359375,
"learning_rate": 4.052074558041608e-07,
"logits/chosen": -3.0283203125,
"logits/rejected": -3.017285108566284,
"logps/chosen": -2134.97509765625,
"logps/rejected": -2573.737548828125,
"logps/weighted_chosen": -4.003759860992432,
"logps/weighted_rejected": -4.656933784484863,
"loss": 0.5513,
"rewards/accuracies": 0.628125011920929,
"rewards/chosen": -1908.206298828125,
"rewards/margins": 434.62811279296875,
"rewards/rejected": -2343.28125,
"rewards/weighted_accuracies": 0.753125011920929,
"rewards/weighted_chosen": -3.4220213890075684,
"rewards/weighted_margins": 0.6371825933456421,
"rewards/weighted_rejected": -4.058789253234863,
"step": 1130
},
{
"epoch": 0.6108506363027462,
"grad_norm": 52.066368103027344,
"learning_rate": 3.960441545911204e-07,
"logits/chosen": -3.1263670921325684,
"logits/rejected": -3.1119141578674316,
"logps/chosen": -1908.762451171875,
"logps/rejected": -2521.012451171875,
"logps/weighted_chosen": -4.003027439117432,
"logps/weighted_rejected": -4.765527248382568,
"loss": 0.5214,
"rewards/accuracies": 0.731249988079071,
"rewards/chosen": -1710.4625244140625,
"rewards/margins": 589.6312255859375,
"rewards/rejected": -2299.95947265625,
"rewards/weighted_accuracies": 0.7406250238418579,
"rewards/weighted_chosen": -3.4476075172424316,
"rewards/weighted_margins": 0.731738269329071,
"rewards/weighted_rejected": -4.180151462554932,
"step": 1140
},
{
"epoch": 0.6162089752176825,
"grad_norm": 70.68743896484375,
"learning_rate": 3.8691720446833187e-07,
"logits/chosen": -3.215527296066284,
"logits/rejected": -3.2021484375,
"logps/chosen": -2513.35009765625,
"logps/rejected": -3176.1875,
"logps/weighted_chosen": -5.13623046875,
"logps/weighted_rejected": -5.89794921875,
"loss": 0.5521,
"rewards/accuracies": 0.675000011920929,
"rewards/chosen": -2315.543701171875,
"rewards/margins": 643.4359130859375,
"rewards/rejected": -2958.637451171875,
"rewards/weighted_accuracies": 0.731249988079071,
"rewards/weighted_chosen": -4.585058689117432,
"rewards/weighted_margins": 0.741162121295929,
"rewards/weighted_rejected": -5.324120998382568,
"step": 1150
},
{
"epoch": 0.6215673141326189,
"grad_norm": 58.93326950073242,
"learning_rate": 3.778297969310529e-07,
"logits/chosen": -3.2992186546325684,
"logits/rejected": -3.297168016433716,
"logps/chosen": -2992.449951171875,
"logps/rejected": -3617.85009765625,
"logps/weighted_chosen": -5.724609375,
"logps/weighted_rejected": -6.642480373382568,
"loss": 0.5462,
"rewards/accuracies": 0.596875011920929,
"rewards/chosen": -2794.61865234375,
"rewards/margins": 608.9249877929688,
"rewards/rejected": -3405.16259765625,
"rewards/weighted_accuracies": 0.71875,
"rewards/weighted_chosen": -5.183398246765137,
"rewards/weighted_margins": 0.877758800983429,
"rewards/weighted_rejected": -6.062597751617432,
"step": 1160
},
{
"epoch": 0.6269256530475552,
"grad_norm": 84.15338134765625,
"learning_rate": 3.687851096473624e-07,
"logits/chosen": -3.316210985183716,
"logits/rejected": -3.3095703125,
"logps/chosen": -2626.550048828125,
"logps/rejected": -3306.75,
"logps/weighted_chosen": -5.133008003234863,
"logps/weighted_rejected": -5.990136623382568,
"loss": 0.5551,
"rewards/accuracies": 0.6468750238418579,
"rewards/chosen": -2415.487548828125,
"rewards/margins": 686.6953125,
"rewards/rejected": -3102.28759765625,
"rewards/weighted_accuracies": 0.7124999761581421,
"rewards/weighted_chosen": -4.54345703125,
"rewards/weighted_margins": 0.861572265625,
"rewards/weighted_rejected": -5.404882907867432,
"step": 1170
},
{
"epoch": 0.6322839919624916,
"grad_norm": 71.16552734375,
"learning_rate": 3.5978630534699865e-07,
"logits/chosen": -3.3482422828674316,
"logits/rejected": -3.3451170921325684,
"logps/chosen": -2592.78759765625,
"logps/rejected": -3643.425048828125,
"logps/weighted_chosen": -5.055273532867432,
"logps/weighted_rejected": -5.8125,
"loss": 0.5426,
"rewards/accuracies": 0.6968749761581421,
"rewards/chosen": -2389.77490234375,
"rewards/margins": 1034.065673828125,
"rewards/rejected": -3422.65625,
"rewards/weighted_accuracies": 0.6875,
"rewards/weighted_chosen": -4.496972560882568,
"rewards/weighted_margins": 0.748461902141571,
"rewards/weighted_rejected": -5.245409965515137,
"step": 1180
},
{
"epoch": 0.6376423308774279,
"grad_norm": 67.54759979248047,
"learning_rate": 3.5083653071542197e-07,
"logits/chosen": -3.2964844703674316,
"logits/rejected": -3.294238328933716,
"logps/chosen": -2341.16259765625,
"logps/rejected": -2609.731201171875,
"logps/weighted_chosen": -4.109082221984863,
"logps/weighted_rejected": -4.714648246765137,
"loss": 0.6061,
"rewards/accuracies": 0.590624988079071,
"rewards/chosen": -2109.64697265625,
"rewards/margins": 264.6039123535156,
"rewards/rejected": -2374.91259765625,
"rewards/weighted_accuracies": 0.637499988079071,
"rewards/weighted_chosen": -3.56103515625,
"rewards/weighted_margins": 0.5668579339981079,
"rewards/weighted_rejected": -4.129150390625,
"step": 1190
},
{
"epoch": 0.6430006697923644,
"grad_norm": 78.90850067138672,
"learning_rate": 3.4193891529348795e-07,
"logits/chosen": -3.225878953933716,
"logits/rejected": -3.225878953933716,
"logps/chosen": -2252.49365234375,
"logps/rejected": -3008.043701171875,
"logps/weighted_chosen": -4.048047065734863,
"logps/weighted_rejected": -4.911328315734863,
"loss": 0.4958,
"rewards/accuracies": 0.6781250238418579,
"rewards/chosen": -2030.903076171875,
"rewards/margins": 770.2398681640625,
"rewards/rejected": -2799.190673828125,
"rewards/weighted_accuracies": 0.762499988079071,
"rewards/weighted_chosen": -3.5079102516174316,
"rewards/weighted_margins": 0.8606201410293579,
"rewards/weighted_rejected": -4.367968559265137,
"step": 1200
},
{
"epoch": 0.6483590087073008,
"grad_norm": 84.27739715576172,
"learning_rate": 3.330965703831146e-07,
"logits/chosen": -3.2769532203674316,
"logits/rejected": -3.26123046875,
"logps/chosen": -2761.58740234375,
"logps/rejected": -3984.36865234375,
"logps/weighted_chosen": -5.098242282867432,
"logps/weighted_rejected": -6.088086128234863,
"loss": 0.5283,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": -2558.112548828125,
"rewards/margins": 1182.4781494140625,
"rewards/rejected": -3739.737548828125,
"rewards/weighted_accuracies": 0.7406250238418579,
"rewards/weighted_chosen": -4.5693359375,
"rewards/weighted_margins": 0.9654296636581421,
"rewards/weighted_rejected": -5.533789157867432,
"step": 1210
},
{
"epoch": 0.6537173476222371,
"grad_norm": 90.22477722167969,
"learning_rate": 3.243125879593286e-07,
"logits/chosen": -3.3125977516174316,
"logits/rejected": -3.286816358566284,
"logps/chosen": -2985.77490234375,
"logps/rejected": -3794.83740234375,
"logps/weighted_chosen": -5.321044921875,
"logps/weighted_rejected": -6.415722846984863,
"loss": 0.5187,
"rewards/accuracies": 0.640625,
"rewards/chosen": -2769.6875,
"rewards/margins": 801.0875244140625,
"rewards/rejected": -3571.24365234375,
"rewards/weighted_accuracies": 0.737500011920929,
"rewards/weighted_chosen": -4.761767387390137,
"rewards/weighted_margins": 1.061132788658142,
"rewards/weighted_rejected": -5.822265625,
"step": 1220
},
{
"epoch": 0.6590756865371735,
"grad_norm": 69.05809783935547,
"learning_rate": 3.1559003958906903e-07,
"logits/chosen": -3.3125977516174316,
"logits/rejected": -3.315234422683716,
"logps/chosen": -2401.887451171875,
"logps/rejected": -3324.375,
"logps/weighted_chosen": -5.186327934265137,
"logps/weighted_rejected": -6.095703125,
"loss": 0.5257,
"rewards/accuracies": 0.6656249761581421,
"rewards/chosen": -2212.887451171875,
"rewards/margins": 904.328125,
"rewards/rejected": -3118.925048828125,
"rewards/weighted_accuracies": 0.75,
"rewards/weighted_chosen": -4.607031345367432,
"rewards/weighted_margins": 0.905322253704071,
"rewards/weighted_rejected": -5.51513671875,
"step": 1230
},
{
"epoch": 0.6644340254521098,
"grad_norm": 70.57830047607422,
"learning_rate": 3.069319753571269e-07,
"logits/chosen": -3.2847657203674316,
"logits/rejected": -3.280468702316284,
"logps/chosen": -2278.425048828125,
"logps/rejected": -3086.425048828125,
"logps/weighted_chosen": -4.578027248382568,
"logps/weighted_rejected": -5.426171779632568,
"loss": 0.5812,
"rewards/accuracies": 0.6656249761581421,
"rewards/chosen": -2084.93115234375,
"rewards/margins": 796.6468505859375,
"rewards/rejected": -2879.862548828125,
"rewards/weighted_accuracies": 0.6968749761581421,
"rewards/weighted_chosen": -4.033203125,
"rewards/weighted_margins": 0.810620129108429,
"rewards/weighted_rejected": -4.844140529632568,
"step": 1240
},
{
"epoch": 0.6697923643670463,
"grad_norm": 60.08575439453125,
"learning_rate": 2.983414227995975e-07,
"logits/chosen": -3.1739258766174316,
"logits/rejected": -3.1656250953674316,
"logps/chosen": -2231.875,
"logps/rejected": -2701.074951171875,
"logps/weighted_chosen": -4.2490234375,
"logps/weighted_rejected": -4.935449123382568,
"loss": 0.5667,
"rewards/accuracies": 0.653124988079071,
"rewards/chosen": -2014.362548828125,
"rewards/margins": 461.45001220703125,
"rewards/rejected": -2474.199951171875,
"rewards/weighted_accuracies": 0.703125,
"rewards/weighted_chosen": -3.686474561691284,
"rewards/weighted_margins": 0.6810302734375,
"rewards/weighted_rejected": -4.368066310882568,
"step": 1250
},
{
"epoch": 0.6751507032819826,
"grad_norm": 76.56852722167969,
"learning_rate": 2.898213858452173e-07,
"logits/chosen": -3.127246141433716,
"logits/rejected": -3.1114258766174316,
"logps/chosen": -2059.981201171875,
"logps/rejected": -2350.03759765625,
"logps/weighted_chosen": -4.0869140625,
"logps/weighted_rejected": -4.69140625,
"loss": 0.5527,
"rewards/accuracies": 0.675000011920929,
"rewards/chosen": -1861.9312744140625,
"rewards/margins": 285.4671936035156,
"rewards/rejected": -2147.18115234375,
"rewards/weighted_accuracies": 0.734375,
"rewards/weighted_chosen": -3.5528807640075684,
"rewards/weighted_margins": 0.589099109172821,
"rewards/weighted_rejected": -4.142333984375,
"step": 1260
},
{
"epoch": 0.680509042196919,
"grad_norm": 74.49163055419922,
"learning_rate": 2.8137484376495506e-07,
"logits/chosen": -3.12646484375,
"logits/rejected": -3.0918946266174316,
"logps/chosen": -2617.00634765625,
"logps/rejected": -3119.440673828125,
"logps/weighted_chosen": -4.663671970367432,
"logps/weighted_rejected": -5.35205078125,
"loss": 0.5693,
"rewards/accuracies": 0.6499999761581421,
"rewards/chosen": -2397.121826171875,
"rewards/margins": 509.5687561035156,
"rewards/rejected": -2906.49365234375,
"rewards/weighted_accuracies": 0.7250000238418579,
"rewards/weighted_chosen": -4.127978324890137,
"rewards/weighted_margins": 0.684863269329071,
"rewards/weighted_rejected": -4.812304496765137,
"step": 1270
},
{
"epoch": 0.6858673811118553,
"grad_norm": 93.80892181396484,
"learning_rate": 2.730047501302266e-07,
"logits/chosen": -3.093554735183716,
"logits/rejected": -3.0672850608825684,
"logps/chosen": -2769.96240234375,
"logps/rejected": -3371.99365234375,
"logps/weighted_chosen": -5.237597465515137,
"logps/weighted_rejected": -6.099804878234863,
"loss": 0.5304,
"rewards/accuracies": 0.675000011920929,
"rewards/chosen": -2569.33740234375,
"rewards/margins": 579.9187622070312,
"rewards/rejected": -3150.10009765625,
"rewards/weighted_accuracies": 0.737500011920929,
"rewards/weighted_chosen": -4.700976371765137,
"rewards/weighted_margins": 0.8484131097793579,
"rewards/weighted_rejected": -5.551074028015137,
"step": 1280
},
{
"epoch": 0.6912257200267917,
"grad_norm": 67.59648895263672,
"learning_rate": 2.647140317800944e-07,
"logits/chosen": -3.073046922683716,
"logits/rejected": -3.0606446266174316,
"logps/chosen": -2785.61865234375,
"logps/rejected": -3411.47509765625,
"logps/weighted_chosen": -5.348340034484863,
"logps/weighted_rejected": -6.2333984375,
"loss": 0.5383,
"rewards/accuracies": 0.690625011920929,
"rewards/chosen": -2587.71875,
"rewards/margins": 618.9249877929688,
"rewards/rejected": -3206.293701171875,
"rewards/weighted_accuracies": 0.71875,
"rewards/weighted_chosen": -4.78515625,
"rewards/weighted_margins": 0.86767578125,
"rewards/weighted_rejected": -5.651659965515137,
"step": 1290
},
{
"epoch": 0.696584058941728,
"grad_norm": 101.80194854736328,
"learning_rate": 2.5650558779781635e-07,
"logits/chosen": -3.100390672683716,
"logits/rejected": -3.0810546875,
"logps/chosen": -2814.846923828125,
"logps/rejected": -3689.050048828125,
"logps/weighted_chosen": -5.2373046875,
"logps/weighted_rejected": -6.142382621765137,
"loss": 0.534,
"rewards/accuracies": 0.675000011920929,
"rewards/chosen": -2618.82177734375,
"rewards/margins": 850.7999877929688,
"rewards/rejected": -3470.30615234375,
"rewards/weighted_accuracies": 0.71875,
"rewards/weighted_chosen": -4.687329292297363,
"rewards/weighted_margins": 0.886914074420929,
"rewards/weighted_rejected": -5.5751953125,
"step": 1300
},
{
"epoch": 0.7019423978566645,
"grad_norm": 54.778236389160156,
"learning_rate": 2.4838228849709997e-07,
"logits/chosen": -3.131542921066284,
"logits/rejected": -3.1089844703674316,
"logps/chosen": -2901.106201171875,
"logps/rejected": -3760.03759765625,
"logps/weighted_chosen": -5.477831840515137,
"logps/weighted_rejected": -6.385839939117432,
"loss": 0.5305,
"rewards/accuracies": 0.6937500238418579,
"rewards/chosen": -2706.14990234375,
"rewards/margins": 837.2843627929688,
"rewards/rejected": -3543.5625,
"rewards/weighted_accuracies": 0.715624988079071,
"rewards/weighted_chosen": -4.89501953125,
"rewards/weighted_margins": 0.8981689214706421,
"rewards/weighted_rejected": -5.79296875,
"step": 1310
},
{
"epoch": 0.7073007367716008,
"grad_norm": 71.93656921386719,
"learning_rate": 2.403469744184154e-07,
"logits/chosen": -3.1357421875,
"logits/rejected": -3.116015672683716,
"logps/chosen": -2888.69384765625,
"logps/rejected": -3361.96875,
"logps/weighted_chosen": -5.206933498382568,
"logps/weighted_rejected": -6.030077934265137,
"loss": 0.5677,
"rewards/accuracies": 0.6468750238418579,
"rewards/chosen": -2675.22802734375,
"rewards/margins": 466.41326904296875,
"rewards/rejected": -3142.762451171875,
"rewards/weighted_accuracies": 0.706250011920929,
"rewards/weighted_chosen": -4.642333984375,
"rewards/weighted_margins": 0.812817394733429,
"rewards/weighted_rejected": -5.456250190734863,
"step": 1320
},
{
"epoch": 0.7126590756865372,
"grad_norm": 59.53675842285156,
"learning_rate": 2.3240245533572167e-07,
"logits/chosen": -3.1578125953674316,
"logits/rejected": -3.1273436546325684,
"logps/chosen": -2736.11865234375,
"logps/rejected": -3383.52490234375,
"logps/weighted_chosen": -5.336230278015137,
"logps/weighted_rejected": -6.180273532867432,
"loss": 0.5341,
"rewards/accuracies": 0.675000011920929,
"rewards/chosen": -2544.64990234375,
"rewards/margins": 625.8937377929688,
"rewards/rejected": -3169.925048828125,
"rewards/weighted_accuracies": 0.762499988079071,
"rewards/weighted_chosen": -4.8037109375,
"rewards/weighted_margins": 0.7956176996231079,
"rewards/weighted_rejected": -5.6015625,
"step": 1330
},
{
"epoch": 0.7180174146014735,
"grad_norm": 63.56354522705078,
"learning_rate": 2.2455150927394878e-07,
"logits/chosen": -3.2085938453674316,
"logits/rejected": -3.174121141433716,
"logps/chosen": -2826.33740234375,
"logps/rejected": -3696.35009765625,
"logps/weighted_chosen": -5.438574314117432,
"logps/weighted_rejected": -6.519629001617432,
"loss": 0.4721,
"rewards/accuracies": 0.6625000238418579,
"rewards/chosen": -2625.762451171875,
"rewards/margins": 855.0718994140625,
"rewards/rejected": -3480.175048828125,
"rewards/weighted_accuracies": 0.778124988079071,
"rewards/weighted_chosen": -4.86962890625,
"rewards/weighted_margins": 1.0476806163787842,
"rewards/weighted_rejected": -5.919140815734863,
"step": 1340
},
{
"epoch": 0.7233757535164099,
"grad_norm": 65.85667419433594,
"learning_rate": 2.167968815375837e-07,
"logits/chosen": -3.3223633766174316,
"logits/rejected": -3.31201171875,
"logps/chosen": -2891.875,
"logps/rejected": -3532.35009765625,
"logps/weighted_chosen": -5.588574409484863,
"logps/weighted_rejected": -6.560546875,
"loss": 0.5111,
"rewards/accuracies": 0.7093750238418579,
"rewards/chosen": -2693.418701171875,
"rewards/margins": 642.3468627929688,
"rewards/rejected": -3337.02490234375,
"rewards/weighted_accuracies": 0.7406250238418579,
"rewards/weighted_chosen": -5.040429592132568,
"rewards/weighted_margins": 0.9643799066543579,
"rewards/weighted_rejected": -6.002245903015137,
"step": 1350
},
{
"epoch": 0.7287340924313462,
"grad_norm": 60.59130096435547,
"learning_rate": 2.0914128375069722e-07,
"logits/chosen": -3.3700194358825684,
"logits/rejected": -3.3509764671325684,
"logps/chosen": -2665.44384765625,
"logps/rejected": -3530.987548828125,
"logps/weighted_chosen": -5.399316310882568,
"logps/weighted_rejected": -6.340136528015137,
"loss": 0.5659,
"rewards/accuracies": 0.6812499761581421,
"rewards/chosen": -2463.637451171875,
"rewards/margins": 849.4015502929688,
"rewards/rejected": -3314.018798828125,
"rewards/weighted_accuracies": 0.7281249761581421,
"rewards/weighted_chosen": -4.841796875,
"rewards/weighted_margins": 0.9273926019668579,
"rewards/weighted_rejected": -5.770410060882568,
"step": 1360
},
{
"epoch": 0.7340924313462827,
"grad_norm": 104.74171447753906,
"learning_rate": 2.015873929087482e-07,
"logits/chosen": -3.359179735183716,
"logits/rejected": -3.3487305641174316,
"logps/chosen": -2628.96240234375,
"logps/rejected": -3298.78125,
"logps/weighted_chosen": -4.932714939117432,
"logps/weighted_rejected": -5.869042873382568,
"loss": 0.5271,
"rewards/accuracies": 0.675000011920929,
"rewards/chosen": -2423.85009765625,
"rewards/margins": 660.5452880859375,
"rewards/rejected": -3083.10009765625,
"rewards/weighted_accuracies": 0.734375,
"rewards/weighted_chosen": -4.387646675109863,
"rewards/weighted_margins": 0.9255615472793579,
"rewards/weighted_rejected": -5.314843654632568,
"step": 1370
},
{
"epoch": 0.739450770261219,
"grad_norm": 88.1015625,
"learning_rate": 1.9413785044249676e-07,
"logits/chosen": -3.3573241233825684,
"logits/rejected": -3.3525390625,
"logps/chosen": -2197.106201171875,
"logps/rejected": -2856.64990234375,
"logps/weighted_chosen": -4.386181831359863,
"logps/weighted_rejected": -5.436425685882568,
"loss": 0.486,
"rewards/accuracies": 0.7250000238418579,
"rewards/chosen": -1992.543701171875,
"rewards/margins": 648.078125,
"rewards/rejected": -2640.41259765625,
"rewards/weighted_accuracies": 0.762499988079071,
"rewards/weighted_chosen": -3.8707518577575684,
"rewards/weighted_margins": 1.014892578125,
"rewards/weighted_rejected": -4.8857421875,
"step": 1380
},
{
"epoch": 0.7448091091761554,
"grad_norm": 62.56149673461914,
"learning_rate": 1.8679526129435358e-07,
"logits/chosen": -3.3702149391174316,
"logits/rejected": -3.3526368141174316,
"logps/chosen": -2864.96875,
"logps/rejected": -3427.824951171875,
"logps/weighted_chosen": -5.074316501617432,
"logps/weighted_rejected": -6.08056640625,
"loss": 0.5087,
"rewards/accuracies": 0.640625,
"rewards/chosen": -2643.831298828125,
"rewards/margins": 562.1812744140625,
"rewards/rejected": -3208.293701171875,
"rewards/weighted_accuracies": 0.762499988079071,
"rewards/weighted_chosen": -4.50439453125,
"rewards/weighted_margins": 0.972485363483429,
"rewards/weighted_rejected": -5.479687690734863,
"step": 1390
},
{
"epoch": 0.7501674480910918,
"grad_norm": 89.40715026855469,
"learning_rate": 1.7956219300748792e-07,
"logits/chosen": -3.3399415016174316,
"logits/rejected": -3.330273389816284,
"logps/chosen": -3021.71240234375,
"logps/rejected": -3707.03759765625,
"logps/weighted_chosen": -5.5947265625,
"logps/weighted_rejected": -6.605370998382568,
"loss": 0.5289,
"rewards/accuracies": 0.6656249761581421,
"rewards/chosen": -2813.96875,
"rewards/margins": 679.5452880859375,
"rewards/rejected": -3492.41259765625,
"rewards/weighted_accuracies": 0.731249988079071,
"rewards/weighted_chosen": -5.048632621765137,
"rewards/weighted_margins": 0.9895263910293579,
"rewards/weighted_rejected": -6.037695407867432,
"step": 1400
},
{
"epoch": 0.7555257870060281,
"grad_norm": 80.76519012451172,
"learning_rate": 1.7244117482801457e-07,
"logits/chosen": -3.2992186546325684,
"logits/rejected": -3.2899413108825684,
"logps/chosen": -3175.44384765625,
"logps/rejected": -4062.262451171875,
"logps/weighted_chosen": -5.7138671875,
"logps/weighted_rejected": -6.551953315734863,
"loss": 0.5516,
"rewards/accuracies": 0.668749988079071,
"rewards/chosen": -2974.456298828125,
"rewards/margins": 867.3734130859375,
"rewards/rejected": -3842.03759765625,
"rewards/weighted_accuracies": 0.7093750238418579,
"rewards/weighted_chosen": -5.172167778015137,
"rewards/weighted_margins": 0.8321533203125,
"rewards/weighted_rejected": -6.002831935882568,
"step": 1410
},
{
"epoch": 0.7608841259209645,
"grad_norm": 72.215087890625,
"learning_rate": 1.6543469682057104e-07,
"logits/chosen": -3.291308641433716,
"logits/rejected": -3.2666015625,
"logps/chosen": -2498.987548828125,
"logps/rejected": -3458.824951171875,
"logps/weighted_chosen": -4.884472846984863,
"logps/weighted_rejected": -5.822167873382568,
"loss": 0.5045,
"rewards/accuracies": 0.6781250238418579,
"rewards/chosen": -2304.59375,
"rewards/margins": 943.5999755859375,
"rewards/rejected": -3247.59375,
"rewards/weighted_accuracies": 0.737500011920929,
"rewards/weighted_chosen": -4.341552734375,
"rewards/weighted_margins": 0.907763659954071,
"rewards/weighted_rejected": -5.25,
"step": 1420
},
{
"epoch": 0.7662424648359009,
"grad_norm": 51.354888916015625,
"learning_rate": 1.5854520899759656e-07,
"logits/chosen": -3.2826170921325684,
"logits/rejected": -3.265820264816284,
"logps/chosen": -2747.356201171875,
"logps/rejected": -3650.862548828125,
"logps/weighted_chosen": -5.053515434265137,
"logps/weighted_rejected": -6.081445217132568,
"loss": 0.4876,
"rewards/accuracies": 0.6875,
"rewards/chosen": -2536.012451171875,
"rewards/margins": 885.2625122070312,
"rewards/rejected": -3421.18115234375,
"rewards/weighted_accuracies": 0.768750011920929,
"rewards/weighted_chosen": -4.487597465515137,
"rewards/weighted_margins": 1.0113036632537842,
"rewards/weighted_rejected": -5.4990234375,
"step": 1430
},
{
"epoch": 0.7716008037508373,
"grad_norm": 129.9513397216797,
"learning_rate": 1.5177512046261666e-07,
"logits/chosen": -3.2759766578674316,
"logits/rejected": -3.2642579078674316,
"logps/chosen": -2872.762451171875,
"logps/rejected": -3759.550048828125,
"logps/weighted_chosen": -5.417382717132568,
"logps/weighted_rejected": -6.450293064117432,
"loss": 0.5024,
"rewards/accuracies": 0.690625011920929,
"rewards/chosen": -2680.324951171875,
"rewards/margins": 868.4000244140625,
"rewards/rejected": -3547.94384765625,
"rewards/weighted_accuracies": 0.753125011920929,
"rewards/weighted_chosen": -4.894238471984863,
"rewards/weighted_margins": 1.013818383216858,
"rewards/weighted_rejected": -5.909081935882568,
"step": 1440
},
{
"epoch": 0.7769591426657736,
"grad_norm": 65.20556640625,
"learning_rate": 1.4512679856783123e-07,
"logits/chosen": -3.28271484375,
"logits/rejected": -3.2655272483825684,
"logps/chosen": -3131.39990234375,
"logps/rejected": -3905.887451171875,
"logps/weighted_chosen": -5.764355659484863,
"logps/weighted_rejected": -6.713574409484863,
"loss": 0.5299,
"rewards/accuracies": 0.671875,
"rewards/chosen": -2915.362548828125,
"rewards/margins": 759.78125,
"rewards/rejected": -3675.112548828125,
"rewards/weighted_accuracies": 0.734375,
"rewards/weighted_chosen": -5.2080078125,
"rewards/weighted_margins": 0.9625244140625,
"rewards/weighted_rejected": -6.171484470367432,
"step": 1450
},
{
"epoch": 0.78231748158071,
"grad_norm": 63.20075988769531,
"learning_rate": 1.3860256808630427e-07,
"logits/chosen": -3.2359375953674316,
"logits/rejected": -3.2249999046325684,
"logps/chosen": -2864.375,
"logps/rejected": -3565.1875,
"logps/weighted_chosen": -5.420361518859863,
"logps/weighted_rejected": -6.3603515625,
"loss": 0.5327,
"rewards/accuracies": 0.6625000238418579,
"rewards/chosen": -2649.88134765625,
"rewards/margins": 697.9484252929688,
"rewards/rejected": -3349.800048828125,
"rewards/weighted_accuracies": 0.7437499761581421,
"rewards/weighted_chosen": -4.828271389007568,
"rewards/weighted_margins": 0.9333130121231079,
"rewards/weighted_rejected": -5.76318359375,
"step": 1460
},
{
"epoch": 0.7876758204956463,
"grad_norm": 66.50836181640625,
"learning_rate": 1.3220471039904047e-07,
"logits/chosen": -3.240234375,
"logits/rejected": -3.222851514816284,
"logps/chosen": -2597.887451171875,
"logps/rejected": -3301.46240234375,
"logps/weighted_chosen": -5.174902439117432,
"logps/weighted_rejected": -6.165917873382568,
"loss": 0.504,
"rewards/accuracies": 0.690625011920929,
"rewards/chosen": -2409.58740234375,
"rewards/margins": 691.7374877929688,
"rewards/rejected": -3101.94384765625,
"rewards/weighted_accuracies": 0.746874988079071,
"rewards/weighted_chosen": -4.65380859375,
"rewards/weighted_margins": 0.966931164264679,
"rewards/weighted_rejected": -5.6201171875,
"step": 1470
},
{
"epoch": 0.7930341594105828,
"grad_norm": 73.51492309570312,
"learning_rate": 1.2593546269723647e-07,
"logits/chosen": -3.24755859375,
"logits/rejected": -3.22802734375,
"logps/chosen": -2874.012451171875,
"logps/rejected": -3520.91259765625,
"logps/weighted_chosen": -5.251416206359863,
"logps/weighted_rejected": -6.177832126617432,
"loss": 0.5262,
"rewards/accuracies": 0.6812499761581421,
"rewards/chosen": -2664.25634765625,
"rewards/margins": 644.0562744140625,
"rewards/rejected": -3306.612548828125,
"rewards/weighted_accuracies": 0.731249988079071,
"rewards/weighted_chosen": -4.698779106140137,
"rewards/weighted_margins": 0.9187988042831421,
"rewards/weighted_rejected": -5.615820407867432,
"step": 1480
},
{
"epoch": 0.7983924983255191,
"grad_norm": 44.982112884521484,
"learning_rate": 1.1979701719998454e-07,
"logits/chosen": -3.227734327316284,
"logits/rejected": -3.204296827316284,
"logps/chosen": -2508.987548828125,
"logps/rejected": -3531.949951171875,
"logps/weighted_chosen": -5.145898342132568,
"logps/weighted_rejected": -6.172753810882568,
"loss": 0.4814,
"rewards/accuracies": 0.706250011920929,
"rewards/chosen": -2321.043701171875,
"rewards/margins": 998.4359130859375,
"rewards/rejected": -3319.78759765625,
"rewards/weighted_accuracies": 0.7593749761581421,
"rewards/weighted_chosen": -4.614648342132568,
"rewards/weighted_margins": 0.997851550579071,
"rewards/weighted_rejected": -5.612011909484863,
"step": 1490
},
{
"epoch": 0.8037508372404555,
"grad_norm": 92.11717987060547,
"learning_rate": 1.1379152038770029e-07,
"logits/chosen": -3.220019578933716,
"logits/rejected": -3.2080078125,
"logps/chosen": -2920.581298828125,
"logps/rejected": -3633.59375,
"logps/weighted_chosen": -5.250292778015137,
"logps/weighted_rejected": -6.166601657867432,
"loss": 0.5298,
"rewards/accuracies": 0.671875,
"rewards/chosen": -2697.268798828125,
"rewards/margins": 713.0906372070312,
"rewards/rejected": -3409.893798828125,
"rewards/weighted_accuracies": 0.6968749761581421,
"rewards/weighted_chosen": -4.699120998382568,
"rewards/weighted_margins": 0.915234386920929,
"rewards/weighted_rejected": -5.615527153015137,
"step": 1500
},
{
"epoch": 0.8037508372404555,
"eval_logits/chosen": -3.2061080932617188,
"eval_logits/rejected": -3.1859562397003174,
"eval_logps/chosen": -2936.50732421875,
"eval_logps/rejected": -3657.35107421875,
"eval_logps/weighted_chosen": -5.540689468383789,
"eval_logps/weighted_rejected": -6.442659854888916,
"eval_loss": 0.5436545610427856,
"eval_rewards/accuracies": 0.6598798632621765,
"eval_rewards/chosen": -2733.882568359375,
"eval_rewards/margins": 707.40771484375,
"eval_rewards/rejected": -3441.0439453125,
"eval_rewards/weighted_accuracies": 0.7239652872085571,
"eval_rewards/weighted_chosen": -4.985209941864014,
"eval_rewards/weighted_margins": 0.8814277052879333,
"eval_rewards/weighted_rejected": -5.866636753082275,
"eval_runtime": 1716.6228,
"eval_samples_per_second": 1.744,
"eval_steps_per_second": 0.436,
"step": 1500
},
{
"epoch": 0.8091091761553918,
"grad_norm": 87.70408630371094,
"learning_rate": 1.0792107225154596e-07,
"logits/chosen": -3.2134766578674316,
"logits/rejected": -3.19921875,
"logps/chosen": -2812.77490234375,
"logps/rejected": -3478.71875,
"logps/weighted_chosen": -5.639452934265137,
"logps/weighted_rejected": -6.593945503234863,
"loss": 0.5414,
"rewards/accuracies": 0.671875,
"rewards/chosen": -2628.34375,
"rewards/margins": 655.0343627929688,
"rewards/rejected": -3283.39990234375,
"rewards/weighted_accuracies": 0.721875011920929,
"rewards/weighted_chosen": -5.105078220367432,
"rewards/weighted_margins": 0.9124511480331421,
"rewards/weighted_rejected": -6.016894340515137,
"step": 1510
},
{
"epoch": 0.8144675150703282,
"grad_norm": 75.14582061767578,
"learning_rate": 1.0218772555910954e-07,
"logits/chosen": -3.20361328125,
"logits/rejected": -3.1815428733825684,
"logps/chosen": -3134.0,
"logps/rejected": -3875.64990234375,
"logps/weighted_chosen": -5.807714939117432,
"logps/weighted_rejected": -6.793261528015137,
"loss": 0.5239,
"rewards/accuracies": 0.684374988079071,
"rewards/chosen": -2934.81884765625,
"rewards/margins": 720.140625,
"rewards/rejected": -3654.21240234375,
"rewards/weighted_accuracies": 0.715624988079071,
"rewards/weighted_chosen": -5.2802734375,
"rewards/weighted_margins": 0.9580078125,
"rewards/weighted_rejected": -6.235937595367432,
"step": 1520
},
{
"epoch": 0.8198258539852645,
"grad_norm": 65.42542266845703,
"learning_rate": 9.659348513659598e-08,
"logits/chosen": -3.2247071266174316,
"logits/rejected": -3.204882860183716,
"logps/chosen": -3170.949951171875,
"logps/rejected": -4248.7626953125,
"logps/weighted_chosen": -5.816210746765137,
"logps/weighted_rejected": -6.777734279632568,
"loss": 0.5135,
"rewards/accuracies": 0.6875,
"rewards/chosen": -2940.637451171875,
"rewards/margins": 1080.1109619140625,
"rewards/rejected": -4022.96240234375,
"rewards/weighted_accuracies": 0.734375,
"rewards/weighted_chosen": -5.244091987609863,
"rewards/weighted_margins": 0.9649413824081421,
"rewards/weighted_rejected": -6.210156440734863,
"step": 1530
},
{
"epoch": 0.825184192900201,
"grad_norm": 59.15487289428711,
"learning_rate": 9.114030716778432e-08,
"logits/chosen": -3.190722703933716,
"logits/rejected": -3.1700196266174316,
"logps/chosen": -3101.53759765625,
"logps/rejected": -4073.987548828125,
"logps/weighted_chosen": -5.633691310882568,
"logps/weighted_rejected": -6.672461032867432,
"loss": 0.5014,
"rewards/accuracies": 0.643750011920929,
"rewards/chosen": -2894.675048828125,
"rewards/margins": 959.1656494140625,
"rewards/rejected": -3854.237548828125,
"rewards/weighted_accuracies": 0.734375,
"rewards/weighted_chosen": -5.086767673492432,
"rewards/weighted_margins": 1.0181396007537842,
"rewards/weighted_rejected": -6.106005668640137,
"step": 1540
},
{
"epoch": 0.8305425318151373,
"grad_norm": 67.11317443847656,
"learning_rate": 8.58300985099918e-08,
"logits/chosen": -3.1815428733825684,
"logits/rejected": -3.159960985183716,
"logps/chosen": -2935.112548828125,
"logps/rejected": -3563.09375,
"logps/weighted_chosen": -5.665625095367432,
"logps/weighted_rejected": -6.660254001617432,
"loss": 0.5111,
"rewards/accuracies": 0.699999988079071,
"rewards/chosen": -2729.018798828125,
"rewards/margins": 614.7374877929688,
"rewards/rejected": -3343.293701171875,
"rewards/weighted_accuracies": 0.7250000238418579,
"rewards/weighted_chosen": -5.095117092132568,
"rewards/weighted_margins": 0.9812988042831421,
"rewards/weighted_rejected": -6.072949409484863,
"step": 1550
},
{
"epoch": 0.8359008707300737,
"grad_norm": 76.89031219482422,
"learning_rate": 8.066471602728803e-08,
"logits/chosen": -3.2030272483825684,
"logits/rejected": -3.1739258766174316,
"logps/chosen": -2904.71240234375,
"logps/rejected": -3688.875,
"logps/weighted_chosen": -5.528418064117432,
"logps/weighted_rejected": -6.6826171875,
"loss": 0.4935,
"rewards/accuracies": 0.6656249761581421,
"rewards/chosen": -2705.86865234375,
"rewards/margins": 773.3015747070312,
"rewards/rejected": -3478.66259765625,
"rewards/weighted_accuracies": 0.765625,
"rewards/weighted_chosen": -4.995409965515137,
"rewards/weighted_margins": 1.127539038658142,
"rewards/weighted_rejected": -6.12353515625,
"step": 1560
},
{
"epoch": 0.84125920964501,
"grad_norm": 99.03889465332031,
"learning_rate": 7.564596594119033e-08,
"logits/chosen": -3.19482421875,
"logits/rejected": -3.170214891433716,
"logps/chosen": -2978.71875,
"logps/rejected": -3497.5625,
"logps/weighted_chosen": -5.62548828125,
"logps/weighted_rejected": -6.575879096984863,
"loss": 0.5376,
"rewards/accuracies": 0.640625,
"rewards/chosen": -2773.112548828125,
"rewards/margins": 521.1875,
"rewards/rejected": -3295.28759765625,
"rewards/weighted_accuracies": 0.7437499761581421,
"rewards/weighted_chosen": -5.089892387390137,
"rewards/weighted_margins": 0.943676769733429,
"rewards/weighted_rejected": -6.034960746765137,
"step": 1570
},
{
"epoch": 0.8466175485599464,
"grad_norm": 69.2409896850586,
"learning_rate": 7.077560319906694e-08,
"logits/chosen": -3.185839891433716,
"logits/rejected": -3.168164014816284,
"logps/chosen": -2938.85009765625,
"logps/rejected": -3474.99365234375,
"logps/weighted_chosen": -5.257275581359863,
"logps/weighted_rejected": -6.167431831359863,
"loss": 0.5381,
"rewards/accuracies": 0.625,
"rewards/chosen": -2736.5,
"rewards/margins": 531.8015747070312,
"rewards/rejected": -3268.02490234375,
"rewards/weighted_accuracies": 0.7250000238418579,
"rewards/weighted_chosen": -4.705664157867432,
"rewards/weighted_margins": 0.9029175043106079,
"rewards/weighted_rejected": -5.607177734375,
"step": 1580
},
{
"epoch": 0.8519758874748827,
"grad_norm": 89.65556335449219,
"learning_rate": 6.605533086047088e-08,
"logits/chosen": -3.169726610183716,
"logits/rejected": -3.160937547683716,
"logps/chosen": -2645.53759765625,
"logps/rejected": -3500.5,
"logps/weighted_chosen": -5.13134765625,
"logps/weighted_rejected": -6.044335842132568,
"loss": 0.503,
"rewards/accuracies": 0.6812499761581421,
"rewards/chosen": -2438.22509765625,
"rewards/margins": 845.2281494140625,
"rewards/rejected": -3282.550048828125,
"rewards/weighted_accuracies": 0.7406250238418579,
"rewards/weighted_chosen": -4.585693359375,
"rewards/weighted_margins": 0.883007824420929,
"rewards/weighted_rejected": -5.467577934265137,
"step": 1590
},
{
"epoch": 0.8573342263898192,
"grad_norm": 71.87641906738281,
"learning_rate": 6.148679950161672e-08,
"logits/chosen": -3.196484327316284,
"logits/rejected": -3.173144578933716,
"logps/chosen": -3048.987548828125,
"logps/rejected": -3728.137451171875,
"logps/weighted_chosen": -5.3779296875,
"logps/weighted_rejected": -6.256543159484863,
"loss": 0.5426,
"rewards/accuracies": 0.65625,
"rewards/chosen": -2816.296875,
"rewards/margins": 676.265625,
"rewards/rejected": -3492.206298828125,
"rewards/weighted_accuracies": 0.737500011920929,
"rewards/weighted_chosen": -4.829297065734863,
"rewards/weighted_margins": 0.8684326410293579,
"rewards/weighted_rejected": -5.698828220367432,
"step": 1600
},
{
"epoch": 0.8626925653047556,
"grad_norm": 58.04913330078125,
"learning_rate": 5.707160663821009e-08,
"logits/chosen": -3.1783204078674316,
"logits/rejected": -3.151562452316284,
"logps/chosen": -2569.85009765625,
"logps/rejected": -3703.75,
"logps/weighted_chosen": -5.129004001617432,
"logps/weighted_rejected": -6.017480373382568,
"loss": 0.5375,
"rewards/accuracies": 0.706250011920929,
"rewards/chosen": -2379.08740234375,
"rewards/margins": 1106.7109375,
"rewards/rejected": -3485.64990234375,
"rewards/weighted_accuracies": 0.734375,
"rewards/weighted_chosen": -4.591894626617432,
"rewards/weighted_margins": 0.8587402105331421,
"rewards/weighted_rejected": -5.450879096984863,
"step": 1610
},
{
"epoch": 0.8680509042196919,
"grad_norm": 59.631710052490234,
"learning_rate": 5.2811296166831666e-08,
"logits/chosen": -3.1659178733825684,
"logits/rejected": -3.1470704078674316,
"logps/chosen": -2712.78759765625,
"logps/rejected": -3333.699951171875,
"logps/weighted_chosen": -5.164648532867432,
"logps/weighted_rejected": -6.154492378234863,
"loss": 0.4743,
"rewards/accuracies": 0.659375011920929,
"rewards/chosen": -2500.925048828125,
"rewards/margins": 610.2156372070312,
"rewards/rejected": -3110.875,
"rewards/weighted_accuracies": 0.765625,
"rewards/weighted_chosen": -4.611572265625,
"rewards/weighted_margins": 0.9793945550918579,
"rewards/weighted_rejected": -5.58984375,
"step": 1620
},
{
"epoch": 0.8734092431346283,
"grad_norm": 42.91954040527344,
"learning_rate": 4.870735782506979e-08,
"logits/chosen": -3.1717772483825684,
"logits/rejected": -3.153613328933716,
"logps/chosen": -2670.418701171875,
"logps/rejected": -3490.09375,
"logps/weighted_chosen": -5.15283203125,
"logps/weighted_rejected": -6.075976371765137,
"loss": 0.5171,
"rewards/accuracies": 0.6812499761581421,
"rewards/chosen": -2461.41259765625,
"rewards/margins": 811.3484497070312,
"rewards/rejected": -3273.84375,
"rewards/weighted_accuracies": 0.734375,
"rewards/weighted_chosen": -4.607714653015137,
"rewards/weighted_margins": 0.912646472454071,
"rewards/weighted_rejected": -5.520215034484863,
"step": 1630
},
{
"epoch": 0.8787675820495646,
"grad_norm": 62.71977996826172,
"learning_rate": 4.4761226670592066e-08,
"logits/chosen": -3.1913084983825684,
"logits/rejected": -3.160839796066284,
"logps/chosen": -2645.762451171875,
"logps/rejected": -3426.86865234375,
"logps/weighted_chosen": -5.320898532867432,
"logps/weighted_rejected": -6.295214653015137,
"loss": 0.4987,
"rewards/accuracies": 0.684374988079071,
"rewards/chosen": -2440.668701171875,
"rewards/margins": 770.0343627929688,
"rewards/rejected": -3210.43115234375,
"rewards/weighted_accuracies": 0.7718750238418579,
"rewards/weighted_chosen": -4.76806640625,
"rewards/weighted_margins": 0.935864269733429,
"rewards/weighted_rejected": -5.705029487609863,
"step": 1640
},
{
"epoch": 0.884125920964501,
"grad_norm": 66.89678955078125,
"learning_rate": 4.097428257933716e-08,
"logits/chosen": -3.187695264816284,
"logits/rejected": -3.162890672683716,
"logps/chosen": -3017.77490234375,
"logps/rejected": -3876.0625,
"logps/weighted_chosen": -5.73974609375,
"logps/weighted_rejected": -6.728417873382568,
"loss": 0.5215,
"rewards/accuracies": 0.706250011920929,
"rewards/chosen": -2828.0,
"rewards/margins": 835.1031494140625,
"rewards/rejected": -3662.41552734375,
"rewards/weighted_accuracies": 0.7281249761581421,
"rewards/weighted_chosen": -5.231787204742432,
"rewards/weighted_margins": 0.9548095464706421,
"rewards/weighted_rejected": -6.185595512390137,
"step": 1650
},
{
"epoch": 0.8894842598794374,
"grad_norm": 82.63867950439453,
"learning_rate": 3.734784976300165e-08,
"logits/chosen": -3.175488233566284,
"logits/rejected": -3.150683641433716,
"logps/chosen": -2716.643798828125,
"logps/rejected": -3477.78759765625,
"logps/weighted_chosen": -5.629199028015137,
"logps/weighted_rejected": -6.379004001617432,
"loss": 0.612,
"rewards/accuracies": 0.628125011920929,
"rewards/chosen": -2517.60009765625,
"rewards/margins": 745.421875,
"rewards/rejected": -3263.012451171875,
"rewards/weighted_accuracies": 0.6468750238418579,
"rewards/weighted_chosen": -5.07275390625,
"rewards/weighted_margins": 0.7259765863418579,
"rewards/weighted_rejected": -5.798925876617432,
"step": 1660
},
{
"epoch": 0.8948425987943738,
"grad_norm": 65.32432556152344,
"learning_rate": 3.38831963059929e-08,
"logits/chosen": -3.1871094703674316,
"logits/rejected": -3.162109375,
"logps/chosen": -2769.24365234375,
"logps/rejected": -3995.64990234375,
"logps/weighted_chosen": -5.353320121765137,
"logps/weighted_rejected": -6.412793159484863,
"loss": 0.5078,
"rewards/accuracies": 0.715624988079071,
"rewards/chosen": -2571.46240234375,
"rewards/margins": 1197.0687255859375,
"rewards/rejected": -3769.637451171875,
"rewards/weighted_accuracies": 0.7437499761581421,
"rewards/weighted_chosen": -4.829003810882568,
"rewards/weighted_margins": 1.0150146484375,
"rewards/weighted_rejected": -5.844140529632568,
"step": 1670
},
{
"epoch": 0.9002009377093101,
"grad_norm": 57.24570083618164,
"learning_rate": 3.058153372200695e-08,
"logits/chosen": -3.1678709983825684,
"logits/rejected": -3.1615233421325684,
"logps/chosen": -2667.25,
"logps/rejected": -3383.625,
"logps/weighted_chosen": -5.261816501617432,
"logps/weighted_rejected": -6.082324028015137,
"loss": 0.55,
"rewards/accuracies": 0.65625,
"rewards/chosen": -2458.91259765625,
"rewards/margins": 713.3968505859375,
"rewards/rejected": -3172.75,
"rewards/weighted_accuracies": 0.706250011920929,
"rewards/weighted_chosen": -4.714550971984863,
"rewards/weighted_margins": 0.80712890625,
"rewards/weighted_rejected": -5.522851467132568,
"step": 1680
},
{
"epoch": 0.9055592766242465,
"grad_norm": 72.50177764892578,
"learning_rate": 2.7444016530389025e-08,
"logits/chosen": -3.1714844703674316,
"logits/rejected": -3.1485352516174316,
"logps/chosen": -2794.46240234375,
"logps/rejected": -3045.887451171875,
"logps/weighted_chosen": -5.3095703125,
"logps/weighted_rejected": -6.16455078125,
"loss": 0.5389,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -2587.04052734375,
"rewards/margins": 262.45233154296875,
"rewards/rejected": -2849.606201171875,
"rewards/weighted_accuracies": 0.71875,
"rewards/weighted_chosen": -4.750878810882568,
"rewards/weighted_margins": 0.848437488079071,
"rewards/weighted_rejected": -5.600341796875,
"step": 1690
},
{
"epoch": 0.9109176155391828,
"grad_norm": 87.17896270751953,
"learning_rate": 2.4471741852423233e-08,
"logits/chosen": -3.16796875,
"logits/rejected": -3.1455078125,
"logps/chosen": -3057.05615234375,
"logps/rejected": -3725.737548828125,
"logps/weighted_chosen": -5.580175876617432,
"logps/weighted_rejected": -6.583984375,
"loss": 0.5216,
"rewards/accuracies": 0.6625000238418579,
"rewards/chosen": -2826.356201171875,
"rewards/margins": 679.046875,
"rewards/rejected": -3504.35009765625,
"rewards/weighted_accuracies": 0.737500011920929,
"rewards/weighted_chosen": -5.033593654632568,
"rewards/weighted_margins": 0.997631847858429,
"rewards/weighted_rejected": -6.031445503234863,
"step": 1700
},
{
"epoch": 0.9162759544541192,
"grad_norm": 61.37962341308594,
"learning_rate": 2.1665749027693024e-08,
"logits/chosen": -3.1654295921325684,
"logits/rejected": -3.1329102516174316,
"logps/chosen": -2913.6875,
"logps/rejected": -3777.46240234375,
"logps/weighted_chosen": -5.392675876617432,
"logps/weighted_rejected": -6.394433498382568,
"loss": 0.5009,
"rewards/accuracies": 0.684374988079071,
"rewards/chosen": -2708.0,
"rewards/margins": 839.7125244140625,
"rewards/rejected": -3547.28759765625,
"rewards/weighted_accuracies": 0.75,
"rewards/weighted_chosen": -4.845507621765137,
"rewards/weighted_margins": 0.9913574457168579,
"rewards/weighted_rejected": -5.839062690734863,
"step": 1710
},
{
"epoch": 0.9216342933690556,
"grad_norm": 51.34814453125,
"learning_rate": 1.9027019250647036e-08,
"logits/chosen": -3.156933546066284,
"logits/rejected": -3.1332030296325684,
"logps/chosen": -2813.012451171875,
"logps/rejected": -3387.231201171875,
"logps/weighted_chosen": -5.587695121765137,
"logps/weighted_rejected": -6.439745903015137,
"loss": 0.5454,
"rewards/accuracies": 0.6625000238418579,
"rewards/chosen": -2612.96875,
"rewards/margins": 573.9093627929688,
"rewards/rejected": -3187.699951171875,
"rewards/weighted_accuracies": 0.7124999761581421,
"rewards/weighted_chosen": -5.025976657867432,
"rewards/weighted_margins": 0.864672839641571,
"rewards/weighted_rejected": -5.890527248382568,
"step": 1720
},
{
"epoch": 0.926992632283992,
"grad_norm": 73.26695251464844,
"learning_rate": 1.6556475227496812e-08,
"logits/chosen": -3.164355516433716,
"logits/rejected": -3.1456055641174316,
"logps/chosen": -2958.762451171875,
"logps/rejected": -3783.481201171875,
"logps/weighted_chosen": -5.6123046875,
"logps/weighted_rejected": -6.597070217132568,
"loss": 0.5266,
"rewards/accuracies": 0.6875,
"rewards/chosen": -2753.356201171875,
"rewards/margins": 819.8953247070312,
"rewards/rejected": -3573.69384765625,
"rewards/weighted_accuracies": 0.7437499761581421,
"rewards/weighted_chosen": -5.076123237609863,
"rewards/weighted_margins": 0.9881591796875,
"rewards/weighted_rejected": -6.064257621765137,
"step": 1730
},
{
"epoch": 0.9323509711989283,
"grad_norm": 93.54780578613281,
"learning_rate": 1.4254980853566246e-08,
"logits/chosen": -3.160449266433716,
"logits/rejected": -3.1363282203674316,
"logps/chosen": -2796.824951171875,
"logps/rejected": -3328.25,
"logps/weighted_chosen": -5.427636623382568,
"logps/weighted_rejected": -6.374413967132568,
"loss": 0.5045,
"rewards/accuracies": 0.671875,
"rewards/chosen": -2590.0625,
"rewards/margins": 533.0515747070312,
"rewards/rejected": -3125.112548828125,
"rewards/weighted_accuracies": 0.7562500238418579,
"rewards/weighted_chosen": -4.849609375,
"rewards/weighted_margins": 0.9527587890625,
"rewards/weighted_rejected": -5.801660060882568,
"step": 1740
},
{
"epoch": 0.9377093101138647,
"grad_norm": 66.60485076904297,
"learning_rate": 1.2123340911206815e-08,
"logits/chosen": -3.15869140625,
"logits/rejected": -3.131542921066284,
"logps/chosen": -2794.800048828125,
"logps/rejected": -3405.175048828125,
"logps/weighted_chosen": -5.47119140625,
"logps/weighted_rejected": -6.472265720367432,
"loss": 0.4748,
"rewards/accuracies": 0.71875,
"rewards/chosen": -2592.53125,
"rewards/margins": 597.4718627929688,
"rewards/rejected": -3189.85009765625,
"rewards/weighted_accuracies": 0.778124988079071,
"rewards/weighted_chosen": -4.9296875,
"rewards/weighted_margins": 0.9674072265625,
"rewards/weighted_rejected": -5.895898342132568,
"step": 1750
},
{
"epoch": 0.943067649028801,
"grad_norm": 49.572227478027344,
"learning_rate": 1.016230078838226e-08,
"logits/chosen": -3.1641602516174316,
"logits/rejected": -3.1470704078674316,
"logps/chosen": -2914.47509765625,
"logps/rejected": -3595.52490234375,
"logps/weighted_chosen": -5.606835842132568,
"logps/weighted_rejected": -6.583105564117432,
"loss": 0.487,
"rewards/accuracies": 0.690625011920929,
"rewards/chosen": -2710.293701171875,
"rewards/margins": 674.9749755859375,
"rewards/rejected": -3386.293701171875,
"rewards/weighted_accuracies": 0.753125011920929,
"rewards/weighted_chosen": -5.075976371765137,
"rewards/weighted_margins": 0.9598388671875,
"rewards/weighted_rejected": -6.036523342132568,
"step": 1760
},
{
"epoch": 0.9484259879437374,
"grad_norm": 62.872276306152344,
"learning_rate": 8.372546218022746e-09,
"logits/chosen": -3.1591796875,
"logits/rejected": -3.137500047683716,
"logps/chosen": -2905.28759765625,
"logps/rejected": -3870.824951171875,
"logps/weighted_chosen": -5.72802734375,
"logps/weighted_rejected": -6.716992378234863,
"loss": 0.4834,
"rewards/accuracies": 0.6781250238418579,
"rewards/chosen": -2698.89990234375,
"rewards/margins": 945.2531127929688,
"rewards/rejected": -3643.28759765625,
"rewards/weighted_accuracies": 0.7718750238418579,
"rewards/weighted_chosen": -5.190820217132568,
"rewards/weighted_margins": 0.977783203125,
"rewards/weighted_rejected": -6.167773246765137,
"step": 1770
},
{
"epoch": 0.9537843268586738,
"grad_norm": 75.22191619873047,
"learning_rate": 6.754703038239329e-09,
"logits/chosen": -3.1405272483825684,
"logits/rejected": -3.13525390625,
"logps/chosen": -2822.35009765625,
"logps/rejected": -3707.699951171875,
"logps/weighted_chosen": -5.446386814117432,
"logps/weighted_rejected": -6.650488376617432,
"loss": 0.486,
"rewards/accuracies": 0.6875,
"rewards/chosen": -2621.550048828125,
"rewards/margins": 878.3968505859375,
"rewards/rejected": -3499.91259765625,
"rewards/weighted_accuracies": 0.7718750238418579,
"rewards/weighted_chosen": -4.904931545257568,
"rewards/weighted_margins": 1.176538109779358,
"rewards/weighted_rejected": -6.079003810882568,
"step": 1780
},
{
"epoch": 0.9591426657736102,
"grad_norm": 67.21267700195312,
"learning_rate": 5.309336973481682e-09,
"logits/chosen": -3.1810545921325684,
"logits/rejected": -3.1556639671325684,
"logps/chosen": -3056.46240234375,
"logps/rejected": -3509.85009765625,
"logps/weighted_chosen": -5.707129001617432,
"logps/weighted_rejected": -6.433691501617432,
"loss": 0.5627,
"rewards/accuracies": 0.625,
"rewards/chosen": -2849.375,
"rewards/margins": 444.78436279296875,
"rewards/rejected": -3293.550048828125,
"rewards/weighted_accuracies": 0.7281249761581421,
"rewards/weighted_chosen": -5.164160251617432,
"rewards/weighted_margins": 0.703051745891571,
"rewards/weighted_rejected": -5.868066310882568,
"step": 1790
},
{
"epoch": 0.9645010046885466,
"grad_norm": 68.99085235595703,
"learning_rate": 4.036953436716895e-09,
"logits/chosen": -3.181640625,
"logits/rejected": -3.163769483566284,
"logps/chosen": -3049.543701171875,
"logps/rejected": -3517.0625,
"logps/weighted_chosen": -5.563622951507568,
"logps/weighted_rejected": -6.490624904632568,
"loss": 0.5335,
"rewards/accuracies": 0.6499999761581421,
"rewards/chosen": -2838.543701171875,
"rewards/margins": 461.46563720703125,
"rewards/rejected": -3300.699951171875,
"rewards/weighted_accuracies": 0.7250000238418579,
"rewards/weighted_chosen": -5.009814262390137,
"rewards/weighted_margins": 0.923999011516571,
"rewards/weighted_rejected": -5.932909965515137,
"step": 1800
},
{
"epoch": 0.9698593436034829,
"grad_norm": 84.05687713623047,
"learning_rate": 2.9379973526973967e-09,
"logits/chosen": -3.1620116233825684,
"logits/rejected": -3.146289110183716,
"logps/chosen": -2902.84375,
"logps/rejected": -3798.21240234375,
"logps/weighted_chosen": -5.657812595367432,
"logps/weighted_rejected": -6.6630859375,
"loss": 0.5073,
"rewards/accuracies": 0.668749988079071,
"rewards/chosen": -2701.234375,
"rewards/margins": 887.8859252929688,
"rewards/rejected": -3589.300048828125,
"rewards/weighted_accuracies": 0.7250000238418579,
"rewards/weighted_chosen": -5.122704982757568,
"rewards/weighted_margins": 0.9850708246231079,
"rewards/weighted_rejected": -6.108105659484863,
"step": 1810
},
{
"epoch": 0.9752176825184193,
"grad_norm": 69.16804504394531,
"learning_rate": 2.0128530023804656e-09,
"logits/chosen": -3.1546874046325684,
"logits/rejected": -3.133593797683716,
"logps/chosen": -3130.137451171875,
"logps/rejected": -3925.14990234375,
"logps/weighted_chosen": -5.845312595367432,
"logps/weighted_rejected": -6.761425971984863,
"loss": 0.5394,
"rewards/accuracies": 0.6499999761581421,
"rewards/chosen": -2926.543701171875,
"rewards/margins": 767.1781005859375,
"rewards/rejected": -3693.300048828125,
"rewards/weighted_accuracies": 0.721875011920929,
"rewards/weighted_chosen": -5.313769340515137,
"rewards/weighted_margins": 0.8848632574081421,
"rewards/weighted_rejected": -6.195996284484863,
"step": 1820
},
{
"epoch": 0.9805760214333556,
"grad_norm": 73.50269317626953,
"learning_rate": 1.2618438885537152e-09,
"logits/chosen": -3.161914110183716,
"logits/rejected": -3.138378858566284,
"logps/chosen": -3062.862548828125,
"logps/rejected": -3960.324951171875,
"logps/weighted_chosen": -5.643456935882568,
"logps/weighted_rejected": -6.707910060882568,
"loss": 0.4749,
"rewards/accuracies": 0.690625011920929,
"rewards/chosen": -2860.58740234375,
"rewards/margins": 884.25,
"rewards/rejected": -3744.0,
"rewards/weighted_accuracies": 0.7875000238418579,
"rewards/weighted_chosen": -5.090478420257568,
"rewards/weighted_margins": 1.050439476966858,
"rewards/weighted_rejected": -6.140429496765137,
"step": 1830
},
{
"epoch": 0.9859343603482921,
"grad_norm": 65.2632064819336,
"learning_rate": 6.852326227130833e-10,
"logits/chosen": -3.167773485183716,
"logits/rejected": -3.1527342796325684,
"logps/chosen": -2902.987548828125,
"logps/rejected": -3425.77490234375,
"logps/weighted_chosen": -5.632128715515137,
"logps/weighted_rejected": -6.458300590515137,
"loss": 0.5536,
"rewards/accuracies": 0.6499999761581421,
"rewards/chosen": -2691.75634765625,
"rewards/margins": 516.25,
"rewards/rejected": -3207.75634765625,
"rewards/weighted_accuracies": 0.734375,
"rewards/weighted_chosen": -5.106347560882568,
"rewards/weighted_margins": 0.8031982183456421,
"rewards/weighted_rejected": -5.910351753234863,
"step": 1840
},
{
"epoch": 0.9912926992632284,
"grad_norm": 66.69202423095703,
"learning_rate": 2.832208332333441e-10,
"logits/chosen": -3.16552734375,
"logits/rejected": -3.1412110328674316,
"logps/chosen": -2898.9375,
"logps/rejected": -3730.606201171875,
"logps/weighted_chosen": -5.487890720367432,
"logps/weighted_rejected": -6.541113376617432,
"loss": 0.4868,
"rewards/accuracies": 0.71875,
"rewards/chosen": -2677.90625,
"rewards/margins": 818.1031494140625,
"rewards/rejected": -3496.64990234375,
"rewards/weighted_accuracies": 0.7906249761581421,
"rewards/weighted_chosen": -4.917284965515137,
"rewards/weighted_margins": 1.014123558998108,
"rewards/weighted_rejected": -5.9306640625,
"step": 1850
},
{
"epoch": 0.9966510381781648,
"grad_norm": 50.651275634765625,
"learning_rate": 5.594909486328348e-11,
"logits/chosen": -3.1732420921325684,
"logits/rejected": -3.1553711891174316,
"logps/chosen": -3009.137451171875,
"logps/rejected": -3617.699951171875,
"logps/weighted_chosen": -5.572167873382568,
"logps/weighted_rejected": -6.584765434265137,
"loss": 0.5004,
"rewards/accuracies": 0.706250011920929,
"rewards/chosen": -2802.52490234375,
"rewards/margins": 580.9781494140625,
"rewards/rejected": -3383.074951171875,
"rewards/weighted_accuracies": 0.75,
"rewards/weighted_chosen": -5.014746189117432,
"rewards/weighted_margins": 0.967211902141571,
"rewards/weighted_rejected": -5.982421875,
"step": 1860
}
],
"logging_steps": 10,
"max_steps": 1867,
"num_input_tokens_seen": 0,
"num_train_epochs": 1,
"save_steps": 500,
"stateful_callbacks": {
"TrainerControl": {
"args": {
"should_epoch_stop": false,
"should_evaluate": false,
"should_log": false,
"should_save": true,
"should_training_stop": true
},
"attributes": {}
}
},
"total_flos": 0.0,
"train_batch_size": 1,
"trial_name": null,
"trial_params": null
}