7b-kto-10-40-i1 / trainer_state.json
BraylonDash's picture
Model save
1c4a63e verified
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 0.997867803837953,
"eval_steps": 500,
"global_step": 234,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.0,
"learning_rate": 2.0833333333333333e-07,
"logits/chosen": -2.024216890335083,
"logits/rejected": -1.8819010257720947,
"logps/chosen": -1305.559326171875,
"logps/rejected": -3790.57275390625,
"loss": 0.5,
"rewards/accuracies": 0.0,
"rewards/chosen": 0.0,
"rewards/margins": 0.0,
"rewards/rejected": 0.0,
"step": 1
},
{
"epoch": 0.04,
"learning_rate": 2.0833333333333334e-06,
"logits/chosen": -2.2936673164367676,
"logits/rejected": -2.1570396423339844,
"logps/chosen": -1376.771728515625,
"logps/rejected": -3041.7578125,
"loss": 0.4966,
"rewards/accuracies": 0.4305555522441864,
"rewards/chosen": -0.037909653037786484,
"rewards/margins": 0.009223480708897114,
"rewards/rejected": -0.04713314026594162,
"step": 10
},
{
"epoch": 0.09,
"learning_rate": 4.166666666666667e-06,
"logits/chosen": -2.1768686771392822,
"logits/rejected": -2.084563732147217,
"logps/chosen": -2079.7646484375,
"logps/rejected": -3319.23583984375,
"loss": 0.4897,
"rewards/accuracies": 0.5062500238418579,
"rewards/chosen": -0.7143774628639221,
"rewards/margins": 0.04834098741412163,
"rewards/rejected": -0.7627183794975281,
"step": 20
},
{
"epoch": 0.13,
"learning_rate": 4.989935734988098e-06,
"logits/chosen": -2.145012617111206,
"logits/rejected": -2.027907609939575,
"logps/chosen": -2782.030517578125,
"logps/rejected": -4380.7275390625,
"loss": 0.4871,
"rewards/accuracies": 0.4437499940395355,
"rewards/chosen": -1.3868815898895264,
"rewards/margins": 0.28924745321273804,
"rewards/rejected": -1.6761291027069092,
"step": 30
},
{
"epoch": 0.17,
"learning_rate": 4.9287250957321685e-06,
"logits/chosen": -2.2485415935516357,
"logits/rejected": -2.1334125995635986,
"logps/chosen": -2663.207763671875,
"logps/rejected": -3975.64892578125,
"loss": 0.4821,
"rewards/accuracies": 0.5562499761581421,
"rewards/chosen": -1.1883825063705444,
"rewards/margins": 0.22184336185455322,
"rewards/rejected": -1.4102258682250977,
"step": 40
},
{
"epoch": 0.21,
"learning_rate": 4.813260751184992e-06,
"logits/chosen": -2.2985146045684814,
"logits/rejected": -2.210111141204834,
"logps/chosen": -1755.809326171875,
"logps/rejected": -3258.091796875,
"loss": 0.4806,
"rewards/accuracies": 0.5625,
"rewards/chosen": -0.6678739190101624,
"rewards/margins": 0.2041586935520172,
"rewards/rejected": -0.872032642364502,
"step": 50
},
{
"epoch": 0.26,
"learning_rate": 4.646121984004666e-06,
"logits/chosen": -2.2362232208251953,
"logits/rejected": -2.1845810413360596,
"logps/chosen": -2074.88134765625,
"logps/rejected": -3796.244873046875,
"loss": 0.473,
"rewards/accuracies": 0.4937500059604645,
"rewards/chosen": -0.5578988790512085,
"rewards/margins": 0.24681946635246277,
"rewards/rejected": -0.8047183752059937,
"step": 60
},
{
"epoch": 0.3,
"learning_rate": 4.431042398061499e-06,
"logits/chosen": -2.382171392440796,
"logits/rejected": -2.307375431060791,
"logps/chosen": -1571.745361328125,
"logps/rejected": -3834.587158203125,
"loss": 0.4815,
"rewards/accuracies": 0.606249988079071,
"rewards/chosen": -0.559153139591217,
"rewards/margins": 0.5565303564071655,
"rewards/rejected": -1.1156834363937378,
"step": 70
},
{
"epoch": 0.34,
"learning_rate": 4.172826515897146e-06,
"logits/chosen": -2.408031940460205,
"logits/rejected": -2.3543269634246826,
"logps/chosen": -2339.724609375,
"logps/rejected": -3922.291748046875,
"loss": 0.4755,
"rewards/accuracies": 0.53125,
"rewards/chosen": -1.0054388046264648,
"rewards/margins": 0.39337557554244995,
"rewards/rejected": -1.39881432056427,
"step": 80
},
{
"epoch": 0.38,
"learning_rate": 3.8772424536302565e-06,
"logits/chosen": -2.4260120391845703,
"logits/rejected": -2.373760461807251,
"logps/chosen": -2360.41552734375,
"logps/rejected": -3782.758544921875,
"loss": 0.4727,
"rewards/accuracies": 0.518750011920929,
"rewards/chosen": -0.8136354684829712,
"rewards/margins": 0.4064968228340149,
"rewards/rejected": -1.2201323509216309,
"step": 90
},
{
"epoch": 0.43,
"learning_rate": 3.5508930707739143e-06,
"logits/chosen": -2.4029459953308105,
"logits/rejected": -2.3526365756988525,
"logps/chosen": -2014.8980712890625,
"logps/rejected": -3683.920654296875,
"loss": 0.4696,
"rewards/accuracies": 0.5375000238418579,
"rewards/chosen": -0.5975068211555481,
"rewards/margins": 0.31760281324386597,
"rewards/rejected": -0.9151096343994141,
"step": 100
},
{
"epoch": 0.47,
"learning_rate": 3.201068473265007e-06,
"logits/chosen": -2.452263355255127,
"logits/rejected": -2.4388153553009033,
"logps/chosen": -2102.702392578125,
"logps/rejected": -4107.12255859375,
"loss": 0.4725,
"rewards/accuracies": 0.53125,
"rewards/chosen": -0.7340173125267029,
"rewards/margins": 0.48010140657424927,
"rewards/rejected": -1.2141185998916626,
"step": 110
},
{
"epoch": 0.51,
"learning_rate": 2.835583164544139e-06,
"logits/chosen": -2.4622802734375,
"logits/rejected": -2.453519344329834,
"logps/chosen": -2598.88037109375,
"logps/rejected": -4125.0986328125,
"loss": 0.4738,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.9378460049629211,
"rewards/margins": 0.3462037444114685,
"rewards/rejected": -1.2840497493743896,
"step": 120
},
{
"epoch": 0.55,
"learning_rate": 2.4626014824618418e-06,
"logits/chosen": -2.5731091499328613,
"logits/rejected": -2.559770107269287,
"logps/chosen": -2807.160400390625,
"logps/rejected": -4325.0205078125,
"loss": 0.4802,
"rewards/accuracies": 0.550000011920929,
"rewards/chosen": -1.1455607414245605,
"rewards/margins": 0.6021825075149536,
"rewards/rejected": -1.7477432489395142,
"step": 130
},
{
"epoch": 0.6,
"learning_rate": 2.090455221462156e-06,
"logits/chosen": -2.468670606613159,
"logits/rejected": -2.4780094623565674,
"logps/chosen": -2137.4091796875,
"logps/rejected": -4046.2734375,
"loss": 0.4714,
"rewards/accuracies": 0.574999988079071,
"rewards/chosen": -0.8530462980270386,
"rewards/margins": 0.3694917857646942,
"rewards/rejected": -1.2225382328033447,
"step": 140
},
{
"epoch": 0.64,
"learning_rate": 1.7274575140626318e-06,
"logits/chosen": -2.4550023078918457,
"logits/rejected": -2.4524893760681152,
"logps/chosen": -1734.0328369140625,
"logps/rejected": -3453.727783203125,
"loss": 0.4732,
"rewards/accuracies": 0.550000011920929,
"rewards/chosen": -0.5440791845321655,
"rewards/margins": 0.4087960124015808,
"rewards/rejected": -0.9528751373291016,
"step": 150
},
{
"epoch": 0.68,
"learning_rate": 1.3817171292109182e-06,
"logits/chosen": -2.520023822784424,
"logits/rejected": -2.549712657928467,
"logps/chosen": -2198.69189453125,
"logps/rejected": -3628.389892578125,
"loss": 0.4758,
"rewards/accuracies": 0.5375000238418579,
"rewards/chosen": -0.6990865468978882,
"rewards/margins": 0.3725363612174988,
"rewards/rejected": -1.0716229677200317,
"step": 160
},
{
"epoch": 0.72,
"learning_rate": 1.0609573357858166e-06,
"logits/chosen": -2.5392794609069824,
"logits/rejected": -2.5447866916656494,
"logps/chosen": -1648.3802490234375,
"logps/rejected": -4229.2060546875,
"loss": 0.4682,
"rewards/accuracies": 0.574999988079071,
"rewards/chosen": -0.7170668840408325,
"rewards/margins": 0.6884183883666992,
"rewards/rejected": -1.4054853916168213,
"step": 170
},
{
"epoch": 0.77,
"learning_rate": 7.723433775328385e-07,
"logits/chosen": -2.524719476699829,
"logits/rejected": -2.5857484340667725,
"logps/chosen": -2204.722900390625,
"logps/rejected": -4786.68408203125,
"loss": 0.47,
"rewards/accuracies": 0.6000000238418579,
"rewards/chosen": -0.9566167593002319,
"rewards/margins": 0.7360025644302368,
"rewards/rejected": -1.6926193237304688,
"step": 180
},
{
"epoch": 0.81,
"learning_rate": 5.223224133591475e-07,
"logits/chosen": -2.5650665760040283,
"logits/rejected": -2.5723259449005127,
"logps/chosen": -2591.955322265625,
"logps/rejected": -4333.18701171875,
"loss": 0.4722,
"rewards/accuracies": 0.5375000238418579,
"rewards/chosen": -1.0086050033569336,
"rewards/margins": 0.689775824546814,
"rewards/rejected": -1.6983808279037476,
"step": 190
},
{
"epoch": 0.85,
"learning_rate": 3.164794984571759e-07,
"logits/chosen": -2.4675166606903076,
"logits/rejected": -2.536062717437744,
"logps/chosen": -2089.778564453125,
"logps/rejected": -4025.212158203125,
"loss": 0.474,
"rewards/accuracies": 0.48750001192092896,
"rewards/chosen": -0.8992789387702942,
"rewards/margins": 0.5448096990585327,
"rewards/rejected": -1.4440886974334717,
"step": 200
},
{
"epoch": 0.9,
"learning_rate": 1.59412823400657e-07,
"logits/chosen": -2.560384750366211,
"logits/rejected": -2.6164653301239014,
"logps/chosen": -2061.99658203125,
"logps/rejected": -3734.251953125,
"loss": 0.4744,
"rewards/accuracies": 0.48750001192092896,
"rewards/chosen": -0.9695846438407898,
"rewards/margins": 0.20665684342384338,
"rewards/rejected": -1.176241397857666,
"step": 210
},
{
"epoch": 0.94,
"learning_rate": 5.463099816548578e-08,
"logits/chosen": -2.4755778312683105,
"logits/rejected": -2.5548205375671387,
"logps/chosen": -1978.981689453125,
"logps/rejected": -4176.359375,
"loss": 0.4721,
"rewards/accuracies": 0.59375,
"rewards/chosen": -0.8206374049186707,
"rewards/margins": 0.5447834730148315,
"rewards/rejected": -1.365420937538147,
"step": 220
},
{
"epoch": 0.98,
"learning_rate": 4.474675580662113e-09,
"logits/chosen": -2.485661745071411,
"logits/rejected": -2.55751371383667,
"logps/chosen": -2132.883056640625,
"logps/rejected": -4433.28271484375,
"loss": 0.4707,
"rewards/accuracies": 0.606249988079071,
"rewards/chosen": -0.7833830714225769,
"rewards/margins": 0.7396507263183594,
"rewards/rejected": -1.523033857345581,
"step": 230
},
{
"epoch": 1.0,
"step": 234,
"total_flos": 0.0,
"train_loss": 0.47607828460187995,
"train_runtime": 5505.1583,
"train_samples_per_second": 2.725,
"train_steps_per_second": 0.043
}
],
"logging_steps": 10,
"max_steps": 234,
"num_input_tokens_seen": 0,
"num_train_epochs": 1,
"save_steps": 20,
"total_flos": 0.0,
"train_batch_size": 4,
"trial_name": null,
"trial_params": null
}