hZzy's picture
Model save
eb09bec verified
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 2.992914501653283,
"eval_steps": 50,
"global_step": 1056,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"dpo_loss": 0.6931471824645996,
"epoch": 0.002834199338686821,
"grad_norm": 22.07642892977258,
"learning_rate": 4.716981132075472e-08,
"logits": -1.2867579460144043,
"logps": -84.34933471679688,
"loss": 0.7323,
"objective": 0.7293278574943542,
"ranking_simple": 0.5833333134651184,
"step": 1
},
{
"dpo_loss": 0.6926142573356628,
"epoch": 0.014170996693434105,
"grad_norm": 21.921066098011842,
"learning_rate": 2.358490566037736e-07,
"logits": -1.4302740097045898,
"logps": -83.70521545410156,
"loss": 0.7347,
"objective": 0.7329134345054626,
"ranking_simple": 0.4895833432674408,
"step": 5
},
{
"dpo_loss": 0.6933749318122864,
"epoch": 0.02834199338686821,
"grad_norm": 23.783506033511284,
"learning_rate": 4.716981132075472e-07,
"logits": -1.4030728340148926,
"logps": -83.88999938964844,
"loss": 0.7338,
"objective": 0.733830451965332,
"ranking_simple": 0.5833333134651184,
"step": 10
},
{
"dpo_loss": 0.687871515750885,
"epoch": 0.042512990080302314,
"grad_norm": 22.035924243469346,
"learning_rate": 7.075471698113208e-07,
"logits": -1.5375704765319824,
"logps": -82.89444732666016,
"loss": 0.7315,
"objective": 0.7324588894844055,
"ranking_simple": 0.574999988079071,
"step": 15
},
{
"dpo_loss": 0.6827042102813721,
"epoch": 0.05668398677373642,
"grad_norm": 20.603492069244336,
"learning_rate": 9.433962264150944e-07,
"logits": -1.377241611480713,
"logps": -81.56272888183594,
"loss": 0.726,
"objective": 0.7265715003013611,
"ranking_simple": 0.48750001192092896,
"step": 20
},
{
"dpo_loss": 0.6807191967964172,
"epoch": 0.07085498346717052,
"grad_norm": 19.555842775301652,
"learning_rate": 1.179245283018868e-06,
"logits": -1.4146537780761719,
"logps": -81.54867553710938,
"loss": 0.7175,
"objective": 0.7202857732772827,
"ranking_simple": 0.5708333253860474,
"step": 25
},
{
"dpo_loss": 0.6734613180160522,
"epoch": 0.08502598016060463,
"grad_norm": 19.538210305976264,
"learning_rate": 1.4150943396226415e-06,
"logits": -1.5176372528076172,
"logps": -83.47260284423828,
"loss": 0.7164,
"objective": 0.7164552807807922,
"ranking_simple": 0.48750001192092896,
"step": 30
},
{
"dpo_loss": 0.669752836227417,
"epoch": 0.09919697685403873,
"grad_norm": 21.78747337027822,
"learning_rate": 1.650943396226415e-06,
"logits": -1.5162482261657715,
"logps": -83.25272369384766,
"loss": 0.7048,
"objective": 0.70932936668396,
"ranking_simple": 0.5375000238418579,
"step": 35
},
{
"dpo_loss": 0.6691684722900391,
"epoch": 0.11336797354747284,
"grad_norm": 22.101328341871838,
"learning_rate": 1.8867924528301889e-06,
"logits": -1.522802710533142,
"logps": -81.73042297363281,
"loss": 0.6976,
"objective": 0.7093203663825989,
"ranking_simple": 0.5041666626930237,
"step": 40
},
{
"dpo_loss": 0.6488091945648193,
"epoch": 0.12753897024090693,
"grad_norm": 18.71576103393132,
"learning_rate": 2.1226415094339624e-06,
"logits": -1.530007004737854,
"logps": -80.75597381591797,
"loss": 0.6962,
"objective": 0.6907859444618225,
"ranking_simple": 0.5208333134651184,
"step": 45
},
{
"dpo_loss": 0.6503397822380066,
"epoch": 0.14170996693434104,
"grad_norm": 19.644583345415946,
"learning_rate": 2.358490566037736e-06,
"logits": -1.5319806337356567,
"logps": -83.5707015991211,
"loss": 0.6826,
"objective": 0.6971887946128845,
"ranking_simple": 0.5166666507720947,
"step": 50
},
{
"epoch": 0.14170996693434104,
"eval_dpo_loss": 0.6827681660652161,
"eval_logits": -1.5367929935455322,
"eval_logps": -92.3091812133789,
"eval_loss": 0.7254260182380676,
"eval_objective": 0.7272183299064636,
"eval_ranking_simple": 0.5279502868652344,
"eval_runtime": 368.8841,
"eval_samples_per_second": 15.696,
"eval_steps_per_second": 1.309,
"step": 50
},
{
"dpo_loss": 0.6309434771537781,
"epoch": 0.15588096362777515,
"grad_norm": 20.382972919722654,
"learning_rate": 2.5943396226415095e-06,
"logits": -1.5233420133590698,
"logps": -86.92516326904297,
"loss": 0.6735,
"objective": 0.6735023260116577,
"ranking_simple": 0.5291666388511658,
"step": 55
},
{
"dpo_loss": 0.6313819885253906,
"epoch": 0.17005196032120926,
"grad_norm": 18.13044483387993,
"learning_rate": 2.830188679245283e-06,
"logits": -1.4574053287506104,
"logps": -87.16276550292969,
"loss": 0.6756,
"objective": 0.680309534072876,
"ranking_simple": 0.5666666626930237,
"step": 60
},
{
"dpo_loss": 0.5997079610824585,
"epoch": 0.18422295701464336,
"grad_norm": 18.96425221102233,
"learning_rate": 3.0660377358490567e-06,
"logits": -1.554844617843628,
"logps": -83.63428497314453,
"loss": 0.6591,
"objective": 0.6451537609100342,
"ranking_simple": 0.6000000238418579,
"step": 65
},
{
"dpo_loss": 0.585650622844696,
"epoch": 0.19839395370807747,
"grad_norm": 17.70445379457848,
"learning_rate": 3.30188679245283e-06,
"logits": -1.6170697212219238,
"logps": -86.24830627441406,
"loss": 0.6426,
"objective": 0.6370573043823242,
"ranking_simple": 0.574999988079071,
"step": 70
},
{
"dpo_loss": 0.61463463306427,
"epoch": 0.21256495040151158,
"grad_norm": 21.475097381950583,
"learning_rate": 3.5377358490566038e-06,
"logits": -1.5737711191177368,
"logps": -88.68922424316406,
"loss": 0.6607,
"objective": 0.6711177229881287,
"ranking_simple": 0.5833333134651184,
"step": 75
},
{
"dpo_loss": 0.5694079995155334,
"epoch": 0.22673594709494568,
"grad_norm": 19.818901321925456,
"learning_rate": 3.7735849056603777e-06,
"logits": -1.6495484113693237,
"logps": -86.88624572753906,
"loss": 0.6294,
"objective": 0.6182950139045715,
"ranking_simple": 0.6291666626930237,
"step": 80
},
{
"dpo_loss": 0.5784618854522705,
"epoch": 0.2409069437883798,
"grad_norm": 15.843708271810604,
"learning_rate": 4.009433962264152e-06,
"logits": -1.6493085622787476,
"logps": -86.69010162353516,
"loss": 0.637,
"objective": 0.6347183585166931,
"ranking_simple": 0.5958333611488342,
"step": 85
},
{
"dpo_loss": 0.5171253085136414,
"epoch": 0.25507794048181387,
"grad_norm": 15.928618564308922,
"learning_rate": 4.245283018867925e-06,
"logits": -1.7159606218338013,
"logps": -87.39143371582031,
"loss": 0.624,
"objective": 0.5702006816864014,
"ranking_simple": 0.625,
"step": 90
},
{
"dpo_loss": 0.5523179173469543,
"epoch": 0.269248937175248,
"grad_norm": 17.73357042390042,
"learning_rate": 4.481132075471699e-06,
"logits": -1.6336411237716675,
"logps": -92.6693344116211,
"loss": 0.6238,
"objective": 0.6184098720550537,
"ranking_simple": 0.5874999761581421,
"step": 95
},
{
"dpo_loss": 0.5685542821884155,
"epoch": 0.2834199338686821,
"grad_norm": 19.615233623313884,
"learning_rate": 4.716981132075472e-06,
"logits": -1.5585684776306152,
"logps": -92.34257507324219,
"loss": 0.6518,
"objective": 0.6242761015892029,
"ranking_simple": 0.5916666388511658,
"step": 100
},
{
"epoch": 0.2834199338686821,
"eval_dpo_loss": 0.6727089285850525,
"eval_logits": -1.6149238348007202,
"eval_logps": -99.43962097167969,
"eval_loss": 0.7250885367393494,
"eval_objective": 0.7208768725395203,
"eval_ranking_simple": 0.5383023023605347,
"eval_runtime": 370.4436,
"eval_samples_per_second": 15.63,
"eval_steps_per_second": 1.304,
"step": 100
},
{
"dpo_loss": 0.5984646677970886,
"epoch": 0.2975909305621162,
"grad_norm": 17.951005909363573,
"learning_rate": 4.952830188679246e-06,
"logits": -1.680674433708191,
"logps": -93.13371276855469,
"loss": 0.6355,
"objective": 0.6607668399810791,
"ranking_simple": 0.625,
"step": 105
},
{
"dpo_loss": 0.5449077486991882,
"epoch": 0.3117619272555503,
"grad_norm": 15.66083104197914,
"learning_rate": 4.999781286194085e-06,
"logits": -1.5491271018981934,
"logps": -89.25350189208984,
"loss": 0.618,
"objective": 0.6060856580734253,
"ranking_simple": 0.5916666388511658,
"step": 110
},
{
"dpo_loss": 0.5834077596664429,
"epoch": 0.32593292394898443,
"grad_norm": 16.011885176044043,
"learning_rate": 4.998892826944418e-06,
"logits": -1.4953746795654297,
"logps": -86.22123718261719,
"loss": 0.6324,
"objective": 0.6475391387939453,
"ranking_simple": 0.5916666388511658,
"step": 115
},
{
"dpo_loss": 0.5353394746780396,
"epoch": 0.3401039206424185,
"grad_norm": 15.412706668770728,
"learning_rate": 4.997321195347154e-06,
"logits": -1.4954169988632202,
"logps": -86.4836654663086,
"loss": 0.6279,
"objective": 0.5993959903717041,
"ranking_simple": 0.6208333373069763,
"step": 120
},
{
"dpo_loss": 0.5647523999214172,
"epoch": 0.35427491733585265,
"grad_norm": 15.94277628603713,
"learning_rate": 4.9950668210706795e-06,
"logits": -1.5816553831100464,
"logps": -87.99699401855469,
"loss": 0.6255,
"objective": 0.6341190934181213,
"ranking_simple": 0.5874999761581421,
"step": 125
},
{
"dpo_loss": 0.537562906742096,
"epoch": 0.3684459140292867,
"grad_norm": 13.361047111116015,
"learning_rate": 4.992130320438411e-06,
"logits": -1.486953854560852,
"logps": -82.83094787597656,
"loss": 0.626,
"objective": 0.6004360318183899,
"ranking_simple": 0.6291666626930237,
"step": 130
},
{
"dpo_loss": 0.5634739398956299,
"epoch": 0.3826169107227208,
"grad_norm": 14.998158738648598,
"learning_rate": 4.988512496260302e-06,
"logits": -1.3754876852035522,
"logps": -82.95647430419922,
"loss": 0.617,
"objective": 0.6264640092849731,
"ranking_simple": 0.5958333611488342,
"step": 135
},
{
"dpo_loss": 0.5888128876686096,
"epoch": 0.39678790741615494,
"grad_norm": 17.662529862358365,
"learning_rate": 4.984214337613357e-06,
"logits": -1.298575520515442,
"logps": -82.97203826904297,
"loss": 0.6332,
"objective": 0.6609014868736267,
"ranking_simple": 0.612500011920929,
"step": 140
},
{
"dpo_loss": 0.5268335342407227,
"epoch": 0.410958904109589,
"grad_norm": 15.28899400178431,
"learning_rate": 4.979237019571235e-06,
"logits": -1.3319251537322998,
"logps": -83.20392608642578,
"loss": 0.6108,
"objective": 0.5984794497489929,
"ranking_simple": 0.699999988079071,
"step": 145
},
{
"dpo_loss": 0.5154529213905334,
"epoch": 0.42512990080302315,
"grad_norm": 14.256755917631622,
"learning_rate": 4.97358190288299e-06,
"logits": -1.389101505279541,
"logps": -84.82242584228516,
"loss": 0.5964,
"objective": 0.5867258310317993,
"ranking_simple": 0.5874999761581421,
"step": 150
},
{
"epoch": 0.42512990080302315,
"eval_dpo_loss": 0.6714372634887695,
"eval_logits": -1.4408866167068481,
"eval_logps": -90.21656036376953,
"eval_loss": 0.7329566478729248,
"eval_objective": 0.7246823310852051,
"eval_ranking_simple": 0.5408902764320374,
"eval_runtime": 368.6547,
"eval_samples_per_second": 15.706,
"eval_steps_per_second": 1.31,
"step": 150
},
{
"dpo_loss": 0.509684681892395,
"epoch": 0.43930089749645723,
"grad_norm": 15.222485168209062,
"learning_rate": 4.967250533601059e-06,
"logits": -1.5175641775131226,
"logps": -85.33422088623047,
"loss": 0.6038,
"objective": 0.5891249179840088,
"ranking_simple": 0.6208333373069763,
"step": 155
},
{
"dpo_loss": 0.5627656579017639,
"epoch": 0.45347189418989137,
"grad_norm": 14.37066166651516,
"learning_rate": 4.9602446426585845e-06,
"logits": -1.444894552230835,
"logps": -85.82301330566406,
"loss": 0.6235,
"objective": 0.6422544121742249,
"ranking_simple": 0.6208333373069763,
"step": 160
},
{
"dpo_loss": 0.5278021693229675,
"epoch": 0.46764289088332545,
"grad_norm": 13.248877681668944,
"learning_rate": 4.952566145396197e-06,
"logits": -1.513926386833191,
"logps": -86.929443359375,
"loss": 0.6091,
"objective": 0.6052799820899963,
"ranking_simple": 0.612500011920929,
"step": 165
},
{
"dpo_loss": 0.5260218977928162,
"epoch": 0.4818138875767596,
"grad_norm": 11.286148523723739,
"learning_rate": 4.944217141038379e-06,
"logits": -1.5067484378814697,
"logps": -87.19461059570312,
"loss": 0.5865,
"objective": 0.5946651697158813,
"ranking_simple": 0.5791666507720947,
"step": 170
},
{
"dpo_loss": 0.532828688621521,
"epoch": 0.49598488427019366,
"grad_norm": 13.880514840470576,
"learning_rate": 4.935199912119558e-06,
"logits": -1.3940012454986572,
"logps": -87.2088851928711,
"loss": 0.5986,
"objective": 0.6150615811347961,
"ranking_simple": 0.6416666507720947,
"step": 175
},
{
"dpo_loss": 0.5224838852882385,
"epoch": 0.5101558809636277,
"grad_norm": 13.177690520129488,
"learning_rate": 4.925516923860083e-06,
"logits": -1.4137970209121704,
"logps": -87.7735366821289,
"loss": 0.572,
"objective": 0.609702467918396,
"ranking_simple": 0.6000000238418579,
"step": 180
},
{
"dpo_loss": 0.485040545463562,
"epoch": 0.5243268776570619,
"grad_norm": 12.914253483146682,
"learning_rate": 4.9151708234922605e-06,
"logits": -1.4367740154266357,
"logps": -85.89669036865234,
"loss": 0.566,
"objective": 0.5699875950813293,
"ranking_simple": 0.6541666388511658,
"step": 185
},
{
"dpo_loss": 0.5013444423675537,
"epoch": 0.538497874350496,
"grad_norm": 11.870478509302453,
"learning_rate": 4.904164439536626e-06,
"logits": -1.4781168699264526,
"logps": -84.88081359863281,
"loss": 0.5937,
"objective": 0.5846849083900452,
"ranking_simple": 0.675000011920929,
"step": 190
},
{
"dpo_loss": 0.47530683875083923,
"epoch": 0.5526688710439301,
"grad_norm": 13.38945439253625,
"learning_rate": 4.8925007810286555e-06,
"logits": -1.5154672861099243,
"logps": -84.33358001708984,
"loss": 0.6047,
"objective": 0.5678978562355042,
"ranking_simple": 0.6583333611488342,
"step": 195
},
{
"dpo_loss": 0.49561557173728943,
"epoch": 0.5668398677373642,
"grad_norm": 11.74052506226232,
"learning_rate": 4.880183036696123e-06,
"logits": -1.5464282035827637,
"logps": -85.40785217285156,
"loss": 0.5794,
"objective": 0.5721699595451355,
"ranking_simple": 0.6333333253860474,
"step": 200
},
{
"epoch": 0.5668398677373642,
"eval_dpo_loss": 0.6857984662055969,
"eval_logits": -1.5459333658218384,
"eval_logps": -90.88076782226562,
"eval_loss": 0.7543493509292603,
"eval_objective": 0.743674099445343,
"eval_ranking_simple": 0.5377846956253052,
"eval_runtime": 368.3702,
"eval_samples_per_second": 15.718,
"eval_steps_per_second": 1.311,
"step": 200
},
{
"dpo_loss": 0.49015435576438904,
"epoch": 0.5810108644307983,
"grad_norm": 13.14364469960144,
"learning_rate": 4.867214574087338e-06,
"logits": -1.4027189016342163,
"logps": -88.2087631225586,
"loss": 0.5739,
"objective": 0.5810206532478333,
"ranking_simple": 0.625,
"step": 205
},
{
"dpo_loss": 0.4410454034805298,
"epoch": 0.5951818611242324,
"grad_norm": 13.072326727623208,
"learning_rate": 4.853598938650487e-06,
"logits": -1.4029525518417358,
"logps": -86.05158996582031,
"loss": 0.5515,
"objective": 0.52931809425354,
"ranking_simple": 0.6708333492279053,
"step": 210
},
{
"dpo_loss": 0.47003600001335144,
"epoch": 0.6093528578176665,
"grad_norm": 12.681896576326764,
"learning_rate": 4.8393398527643495e-06,
"logits": -1.4985297918319702,
"logps": -84.5003433227539,
"loss": 0.5691,
"objective": 0.5656867623329163,
"ranking_simple": 0.6416666507720947,
"step": 215
},
{
"dpo_loss": 0.4842369854450226,
"epoch": 0.6235238545111006,
"grad_norm": 12.185860749706057,
"learning_rate": 4.824441214720629e-06,
"logits": -1.5714740753173828,
"logps": -84.36730194091797,
"loss": 0.5581,
"objective": 0.5698367953300476,
"ranking_simple": 0.6291666626930237,
"step": 220
},
{
"dpo_loss": 0.499014288187027,
"epoch": 0.6376948512045347,
"grad_norm": 11.86860643002718,
"learning_rate": 4.808907097658205e-06,
"logits": -1.5924923419952393,
"logps": -84.66715240478516,
"loss": 0.5884,
"objective": 0.5895112752914429,
"ranking_simple": 0.6083333492279053,
"step": 225
},
{
"dpo_loss": 0.45395275950431824,
"epoch": 0.6518658478979689,
"grad_norm": 11.069924512206331,
"learning_rate": 4.7927417484495756e-06,
"logits": -1.4241927862167358,
"logps": -83.54296112060547,
"loss": 0.5726,
"objective": 0.5383394956588745,
"ranking_simple": 0.6791666746139526,
"step": 230
},
{
"dpo_loss": 0.491834819316864,
"epoch": 0.6660368445914029,
"grad_norm": 10.918136207544526,
"learning_rate": 4.7759495865398035e-06,
"logits": -1.3152087926864624,
"logps": -84.92122650146484,
"loss": 0.5635,
"objective": 0.575198233127594,
"ranking_simple": 0.6083333492279053,
"step": 235
},
{
"dpo_loss": 0.4870225191116333,
"epoch": 0.680207841284837,
"grad_norm": 11.379260144258156,
"learning_rate": 4.758535202738287e-06,
"logits": -1.5095748901367188,
"logps": -87.46720123291016,
"loss": 0.5672,
"objective": 0.5821055173873901,
"ranking_simple": 0.6708333492279053,
"step": 240
},
{
"dpo_loss": 0.541431725025177,
"epoch": 0.6943788379782712,
"grad_norm": 12.918426431601308,
"learning_rate": 4.740503357963676e-06,
"logits": -1.5333224534988403,
"logps": -86.5069808959961,
"loss": 0.5804,
"objective": 0.6269174218177795,
"ranking_simple": 0.6666666865348816,
"step": 245
},
{
"dpo_loss": 0.46632593870162964,
"epoch": 0.7085498346717053,
"grad_norm": 11.479829992169247,
"learning_rate": 4.721858981942284e-06,
"logits": -1.311442494392395,
"logps": -82.5728530883789,
"loss": 0.5802,
"objective": 0.559305727481842,
"ranking_simple": 0.6791666746139526,
"step": 250
},
{
"epoch": 0.7085498346717053,
"eval_dpo_loss": 0.6874057650566101,
"eval_logits": -1.5326449871063232,
"eval_logps": -86.87519073486328,
"eval_loss": 0.7558908462524414,
"eval_objective": 0.7459491491317749,
"eval_ranking_simple": 0.5403726696968079,
"eval_runtime": 368.3582,
"eval_samples_per_second": 15.718,
"eval_steps_per_second": 1.311,
"step": 250
},
{
"dpo_loss": 0.4774031341075897,
"epoch": 0.7227208313651393,
"grad_norm": 9.487597580618754,
"learning_rate": 4.702607171860354e-06,
"logits": -1.5363638401031494,
"logps": -80.10908508300781,
"loss": 0.5586,
"objective": 0.5592390894889832,
"ranking_simple": 0.6291666626930237,
"step": 255
},
{
"dpo_loss": 0.4512685537338257,
"epoch": 0.7368918280585735,
"grad_norm": 10.042345370580236,
"learning_rate": 4.682753190970533e-06,
"logits": -1.5129272937774658,
"logps": -81.99565887451172,
"loss": 0.5574,
"objective": 0.5405219793319702,
"ranking_simple": 0.6333333253860474,
"step": 260
},
{
"dpo_loss": 0.4361381232738495,
"epoch": 0.7510628247520076,
"grad_norm": 10.77799422518219,
"learning_rate": 4.6623024671529555e-06,
"logits": -1.470523715019226,
"logps": -81.5511703491211,
"loss": 0.5503,
"objective": 0.5269137620925903,
"ranking_simple": 0.7124999761581421,
"step": 265
},
{
"dpo_loss": 0.48189878463745117,
"epoch": 0.7652338214454416,
"grad_norm": 12.181902084596548,
"learning_rate": 4.641260591431315e-06,
"logits": -1.5885664224624634,
"logps": -82.85069274902344,
"loss": 0.5554,
"objective": 0.5691978931427002,
"ranking_simple": 0.6416666507720947,
"step": 270
},
{
"dpo_loss": 0.47639578580856323,
"epoch": 0.7794048181388757,
"grad_norm": 10.23476141708296,
"learning_rate": 4.61963331644433e-06,
"logits": -1.4604405164718628,
"logps": -82.8023681640625,
"loss": 0.5548,
"objective": 0.55946284532547,
"ranking_simple": 0.637499988079071,
"step": 275
},
{
"dpo_loss": 0.43256130814552307,
"epoch": 0.7935758148323099,
"grad_norm": 10.705607992682587,
"learning_rate": 4.597426554873037e-06,
"logits": -1.5027648210525513,
"logps": -83.16519927978516,
"loss": 0.5354,
"objective": 0.5194212794303894,
"ranking_simple": 0.6708333492279053,
"step": 280
},
{
"dpo_loss": 0.45488685369491577,
"epoch": 0.807746811525744,
"grad_norm": 11.614191556355545,
"learning_rate": 4.574646377824316e-06,
"logits": -1.4621371030807495,
"logps": -84.6886215209961,
"loss": 0.5375,
"objective": 0.539017379283905,
"ranking_simple": 0.6000000238418579,
"step": 285
},
{
"dpo_loss": 0.4303934872150421,
"epoch": 0.821917808219178,
"grad_norm": 11.206240309054008,
"learning_rate": 4.551299013171111e-06,
"logits": -1.4472154378890991,
"logps": -83.35670471191406,
"loss": 0.5305,
"objective": 0.5253880620002747,
"ranking_simple": 0.6958333253860474,
"step": 290
},
{
"dpo_loss": 0.43744438886642456,
"epoch": 0.8360888049126122,
"grad_norm": 9.074371483374543,
"learning_rate": 4.5273908438498e-06,
"logits": -1.506995677947998,
"logps": -84.82511138916016,
"loss": 0.5517,
"objective": 0.5335044860839844,
"ranking_simple": 0.637499988079071,
"step": 295
},
{
"dpo_loss": 0.4651775360107422,
"epoch": 0.8502598016060463,
"grad_norm": 9.01878047466114,
"learning_rate": 4.502928406115152e-06,
"logits": -1.4889142513275146,
"logps": -84.10907745361328,
"loss": 0.5473,
"objective": 0.5568612217903137,
"ranking_simple": 0.6666666865348816,
"step": 300
},
{
"epoch": 0.8502598016060463,
"eval_dpo_loss": 0.6780434846878052,
"eval_logits": -1.5369890928268433,
"eval_logps": -92.34797668457031,
"eval_loss": 0.7457044720649719,
"eval_objective": 0.7388522028923035,
"eval_ranking_simple": 0.5486542582511902,
"eval_runtime": 369.8452,
"eval_samples_per_second": 15.655,
"eval_steps_per_second": 1.306,
"step": 300
},
{
"dpo_loss": 0.47844478487968445,
"epoch": 0.8644307982994804,
"grad_norm": 10.005579661781612,
"learning_rate": 4.477918387753388e-06,
"logits": -1.4307849407196045,
"logps": -86.45662689208984,
"loss": 0.5252,
"objective": 0.5761564373970032,
"ranking_simple": 0.7083333134651184,
"step": 305
},
{
"dpo_loss": 0.40813323855400085,
"epoch": 0.8786017949929145,
"grad_norm": 10.307837911346382,
"learning_rate": 4.452367626253805e-06,
"logits": -1.3825361728668213,
"logps": -86.61471557617188,
"loss": 0.5225,
"objective": 0.5045412182807922,
"ranking_simple": 0.6916666626930237,
"step": 310
},
{
"dpo_loss": 0.42051681876182556,
"epoch": 0.8927727916863486,
"grad_norm": 10.459183690361446,
"learning_rate": 4.426283106939474e-06,
"logits": -1.3151206970214844,
"logps": -85.77119445800781,
"loss": 0.5426,
"objective": 0.513097882270813,
"ranking_simple": 0.637499988079071,
"step": 315
},
{
"dpo_loss": 0.41688165068626404,
"epoch": 0.9069437883797827,
"grad_norm": 11.302451754279112,
"learning_rate": 4.399671961057523e-06,
"logits": -1.2146058082580566,
"logps": -87.50178527832031,
"loss": 0.5607,
"objective": 0.5194131731987,
"ranking_simple": 0.7166666388511658,
"step": 320
},
{
"dpo_loss": 0.4247673749923706,
"epoch": 0.9211147850732169,
"grad_norm": 9.112687814178505,
"learning_rate": 4.372541463829524e-06,
"logits": -1.2642977237701416,
"logps": -86.85433959960938,
"loss": 0.5281,
"objective": 0.5294176340103149,
"ranking_simple": 0.6875,
"step": 325
},
{
"dpo_loss": 0.3975774645805359,
"epoch": 0.9352857817666509,
"grad_norm": 10.380051826599395,
"learning_rate": 4.3448990324625244e-06,
"logits": -1.1297107934951782,
"logps": -85.07998657226562,
"loss": 0.5304,
"objective": 0.5025947690010071,
"ranking_simple": 0.7041666507720947,
"step": 330
},
{
"dpo_loss": 0.4059672951698303,
"epoch": 0.949456778460085,
"grad_norm": 9.237583761740382,
"learning_rate": 4.316752224121252e-06,
"logits": -1.118418574333191,
"logps": -84.82205200195312,
"loss": 0.5208,
"objective": 0.499520480632782,
"ranking_simple": 0.7041666507720947,
"step": 335
},
{
"dpo_loss": 0.4579353332519531,
"epoch": 0.9636277751535192,
"grad_norm": 9.19069599968856,
"learning_rate": 4.288108733862064e-06,
"logits": -1.200610876083374,
"logps": -83.12140655517578,
"loss": 0.5327,
"objective": 0.5504066348075867,
"ranking_simple": 0.6583333611488342,
"step": 340
},
{
"dpo_loss": 0.4034007489681244,
"epoch": 0.9777987718469532,
"grad_norm": 9.024982300418905,
"learning_rate": 4.2589763925291924e-06,
"logits": -1.2796326875686646,
"logps": -81.1929702758789,
"loss": 0.5202,
"objective": 0.4997580945491791,
"ranking_simple": 0.699999988079071,
"step": 345
},
{
"dpo_loss": 0.3821704089641571,
"epoch": 0.9919697685403873,
"grad_norm": 10.161439402885936,
"learning_rate": 4.229363164613874e-06,
"logits": -1.2219146490097046,
"logps": -84.29117584228516,
"loss": 0.5104,
"objective": 0.4823973476886749,
"ranking_simple": 0.7708333134651184,
"step": 350
},
{
"epoch": 0.9919697685403873,
"eval_dpo_loss": 0.6766201257705688,
"eval_logits": -1.3364328145980835,
"eval_logps": -88.19395446777344,
"eval_loss": 0.7516013979911804,
"eval_objective": 0.7371890544891357,
"eval_ranking_simple": 0.5429606437683105,
"eval_runtime": 368.1018,
"eval_samples_per_second": 15.729,
"eval_steps_per_second": 1.312,
"step": 350
},
{
"dpo_loss": 0.34969767928123474,
"epoch": 1.0061407652338215,
"grad_norm": 6.289440966432375,
"learning_rate": 4.199277146076933e-06,
"logits": -1.3432971239089966,
"logps": -82.10868072509766,
"loss": 0.4782,
"objective": 0.4609685242176056,
"ranking_simple": 0.7458333373069763,
"step": 355
},
{
"dpo_loss": 0.33886921405792236,
"epoch": 1.0203117619272555,
"grad_norm": 8.403207409711419,
"learning_rate": 4.168726562135432e-06,
"logits": -1.2731363773345947,
"logps": -83.29503631591797,
"loss": 0.4406,
"objective": 0.4606216251850128,
"ranking_simple": 0.7208333611488342,
"step": 360
},
{
"dpo_loss": 0.2881692051887512,
"epoch": 1.0344827586206897,
"grad_norm": 6.468809468648868,
"learning_rate": 4.137719765013974e-06,
"logits": -1.298916220664978,
"logps": -84.39533233642578,
"loss": 0.4218,
"objective": 0.4092879295349121,
"ranking_simple": 0.7416666746139526,
"step": 365
},
{
"dpo_loss": 0.31373023986816406,
"epoch": 1.0486537553141237,
"grad_norm": 10.120074264563044,
"learning_rate": 4.106265231661292e-06,
"logits": -1.257562518119812,
"logps": -81.22982788085938,
"loss": 0.4476,
"objective": 0.4375106990337372,
"ranking_simple": 0.7333333492279053,
"step": 370
},
{
"dpo_loss": 0.3089008033275604,
"epoch": 1.0628247520075578,
"grad_norm": 7.557866547749679,
"learning_rate": 4.074371561432731e-06,
"logits": -1.3321465253829956,
"logps": -83.06178283691406,
"loss": 0.4494,
"objective": 0.4439465403556824,
"ranking_simple": 0.762499988079071,
"step": 375
},
{
"dpo_loss": 0.3182891309261322,
"epoch": 1.076995748700992,
"grad_norm": 7.153131025534419,
"learning_rate": 4.042047473739278e-06,
"logits": -1.2969304323196411,
"logps": -84.80577087402344,
"loss": 0.4362,
"objective": 0.4368627667427063,
"ranking_simple": 0.7833333611488342,
"step": 380
},
{
"dpo_loss": 0.33301401138305664,
"epoch": 1.091166745394426,
"grad_norm": 7.370055629692345,
"learning_rate": 4.009301805663752e-06,
"logits": -1.1952742338180542,
"logps": -83.80043029785156,
"loss": 0.4555,
"objective": 0.4633476436138153,
"ranking_simple": 0.7291666865348816,
"step": 385
},
{
"dpo_loss": 0.3350224494934082,
"epoch": 1.10533774208786,
"grad_norm": 6.621937340714611,
"learning_rate": 3.976143509544843e-06,
"logits": -1.144229531288147,
"logps": -85.45323944091797,
"loss": 0.4405,
"objective": 0.4582778811454773,
"ranking_simple": 0.762499988079071,
"step": 390
},
{
"dpo_loss": 0.2885909676551819,
"epoch": 1.1195087387812943,
"grad_norm": 7.285249021586949,
"learning_rate": 3.9425816505296254e-06,
"logits": -1.1913460493087769,
"logps": -84.71748352050781,
"loss": 0.4457,
"objective": 0.40855225920677185,
"ranking_simple": 0.7708333134651184,
"step": 395
},
{
"dpo_loss": 0.3148750960826874,
"epoch": 1.1336797354747283,
"grad_norm": 7.611492036237173,
"learning_rate": 3.908625404095242e-06,
"logits": -1.2308555841445923,
"logps": -82.81310272216797,
"loss": 0.4425,
"objective": 0.4325387179851532,
"ranking_simple": 0.7583333253860474,
"step": 400
},
{
"epoch": 1.1336797354747283,
"eval_dpo_loss": 0.6866451501846313,
"eval_logits": -1.2226030826568604,
"eval_logps": -88.75952911376953,
"eval_loss": 0.7568410634994507,
"eval_objective": 0.7489107251167297,
"eval_ranking_simple": 0.5439958572387695,
"eval_runtime": 368.4645,
"eval_samples_per_second": 15.714,
"eval_steps_per_second": 1.311,
"step": 400
},
{
"dpo_loss": 0.35161828994750977,
"epoch": 1.1478507321681626,
"grad_norm": 7.5393253270016265,
"learning_rate": 3.8742840535404155e-06,
"logits": -1.1856341361999512,
"logps": -84.8878173828125,
"loss": 0.4461,
"objective": 0.47060316801071167,
"ranking_simple": 0.7166666388511658,
"step": 405
},
{
"dpo_loss": 0.338571161031723,
"epoch": 1.1620217288615966,
"grad_norm": 5.801736886678098,
"learning_rate": 3.839566987447492e-06,
"logits": -1.2132126092910767,
"logps": -85.20359802246094,
"loss": 0.4339,
"objective": 0.46073082089424133,
"ranking_simple": 0.7333333492279053,
"step": 410
},
{
"dpo_loss": 0.3313528895378113,
"epoch": 1.1761927255550306,
"grad_norm": 7.544361991295727,
"learning_rate": 3.8044836971156935e-06,
"logits": -1.1481475830078125,
"logps": -84.14738464355469,
"loss": 0.4504,
"objective": 0.45183154940605164,
"ranking_simple": 0.6833333373069763,
"step": 415
},
{
"dpo_loss": 0.31563931703567505,
"epoch": 1.1903637222484649,
"grad_norm": 6.99992341058847,
"learning_rate": 3.7690437739662928e-06,
"logits": -1.1214243173599243,
"logps": -84.69954681396484,
"loss": 0.4436,
"objective": 0.43789881467819214,
"ranking_simple": 0.7166666388511658,
"step": 420
},
{
"dpo_loss": 0.3403506278991699,
"epoch": 1.204534718941899,
"grad_norm": 7.423010911616055,
"learning_rate": 3.7332569069204127e-06,
"logits": -1.179802417755127,
"logps": -83.71639251708984,
"loss": 0.4526,
"objective": 0.46505507826805115,
"ranking_simple": 0.7250000238418579,
"step": 425
},
{
"dpo_loss": 0.3509117066860199,
"epoch": 1.2187057156353331,
"grad_norm": 7.343865460650952,
"learning_rate": 3.697132879750174e-06,
"logits": -1.190435528755188,
"logps": -86.54014587402344,
"loss": 0.4481,
"objective": 0.4735199213027954,
"ranking_simple": 0.6958333253860474,
"step": 430
},
{
"dpo_loss": 0.3211102783679962,
"epoch": 1.2328767123287672,
"grad_norm": 8.246553327807998,
"learning_rate": 3.6606815684039098e-06,
"logits": -1.1106759309768677,
"logps": -84.52306365966797,
"loss": 0.4591,
"objective": 0.4374840557575226,
"ranking_simple": 0.7791666388511658,
"step": 435
},
{
"dpo_loss": 0.3193610608577728,
"epoch": 1.2470477090222012,
"grad_norm": 8.720099662621916,
"learning_rate": 3.6239129383061764e-06,
"logits": -1.1405307054519653,
"logps": -85.81301879882812,
"loss": 0.4554,
"objective": 0.44526031613349915,
"ranking_simple": 0.7958333492279053,
"step": 440
},
{
"dpo_loss": 0.3264698088169098,
"epoch": 1.2612187057156352,
"grad_norm": 7.381389964368484,
"learning_rate": 3.586837041633312e-06,
"logits": -1.1591824293136597,
"logps": -86.23896026611328,
"loss": 0.4648,
"objective": 0.45790377259254456,
"ranking_simple": 0.7875000238418579,
"step": 445
},
{
"dpo_loss": 0.32597142457962036,
"epoch": 1.2753897024090695,
"grad_norm": 7.58156282251525,
"learning_rate": 3.5494640145652647e-06,
"logits": -1.3325276374816895,
"logps": -85.49541473388672,
"loss": 0.4544,
"objective": 0.456112265586853,
"ranking_simple": 0.7791666388511658,
"step": 450
},
{
"epoch": 1.2753897024090695,
"eval_dpo_loss": 0.6750319004058838,
"eval_logits": -1.3088560104370117,
"eval_logps": -90.05509185791016,
"eval_loss": 0.7455039620399475,
"eval_objective": 0.736483633518219,
"eval_ranking_simple": 0.5481366515159607,
"eval_runtime": 373.8032,
"eval_samples_per_second": 15.489,
"eval_steps_per_second": 1.292,
"step": 450
},
{
"dpo_loss": 0.3323071599006653,
"epoch": 1.2895606991025035,
"grad_norm": 7.576847549255986,
"learning_rate": 3.511804074514468e-06,
"logits": -1.1815721988677979,
"logps": -86.30281829833984,
"loss": 0.4633,
"objective": 0.45755326747894287,
"ranking_simple": 0.7291666865348816,
"step": 455
},
{
"dpo_loss": 0.32099449634552,
"epoch": 1.3037316957959377,
"grad_norm": 7.0611969685720934,
"learning_rate": 3.4738675173325008e-06,
"logits": -1.3054893016815186,
"logps": -85.12818145751953,
"loss": 0.449,
"objective": 0.4495506286621094,
"ranking_simple": 0.7166666388511658,
"step": 460
},
{
"dpo_loss": 0.3063493072986603,
"epoch": 1.3179026924893718,
"grad_norm": 7.25813572886754,
"learning_rate": 3.435664714495301e-06,
"logits": -1.3341224193572998,
"logps": -82.83889770507812,
"loss": 0.4533,
"objective": 0.4265134036540985,
"ranking_simple": 0.737500011920929,
"step": 465
},
{
"dpo_loss": 0.3599289655685425,
"epoch": 1.3320736891828058,
"grad_norm": 7.324136192857674,
"learning_rate": 3.397206110267713e-06,
"logits": -1.409017562866211,
"logps": -81.51747131347656,
"loss": 0.4597,
"objective": 0.47407808899879456,
"ranking_simple": 0.7458333373069763,
"step": 470
},
{
"dpo_loss": 0.32313933968544006,
"epoch": 1.34624468587624,
"grad_norm": 7.632057920114339,
"learning_rate": 3.3585022188481247e-06,
"logits": -1.3106532096862793,
"logps": -84.2204360961914,
"loss": 0.4525,
"objective": 0.44208064675331116,
"ranking_simple": 0.7749999761581421,
"step": 475
},
{
"dpo_loss": 0.3283863365650177,
"epoch": 1.360415682569674,
"grad_norm": 8.810666497005016,
"learning_rate": 3.3195636214939943e-06,
"logits": -1.350165843963623,
"logps": -84.08671569824219,
"loss": 0.4602,
"objective": 0.45167192816734314,
"ranking_simple": 0.7166666388511658,
"step": 480
},
{
"dpo_loss": 0.3586553633213043,
"epoch": 1.3745866792631083,
"grad_norm": 9.1074396847746,
"learning_rate": 3.2804009636290403e-06,
"logits": -1.3092820644378662,
"logps": -82.56153106689453,
"loss": 0.4755,
"objective": 0.4738306701183319,
"ranking_simple": 0.737500011920929,
"step": 485
},
{
"dpo_loss": 0.3484760820865631,
"epoch": 1.3887576759565423,
"grad_norm": 8.495980198743673,
"learning_rate": 3.2410249519328848e-06,
"logits": -1.2295089960098267,
"logps": -85.73413848876953,
"loss": 0.4478,
"objective": 0.46845272183418274,
"ranking_simple": 0.7458333373069763,
"step": 490
},
{
"dpo_loss": 0.3095528483390808,
"epoch": 1.4029286726499763,
"grad_norm": 8.616754835251825,
"learning_rate": 3.201446351413958e-06,
"logits": -1.2363132238388062,
"logps": -86.17890930175781,
"loss": 0.4491,
"objective": 0.4346270263195038,
"ranking_simple": 0.7583333253860474,
"step": 495
},
{
"dpo_loss": 0.33423519134521484,
"epoch": 1.4170996693434104,
"grad_norm": 9.483021181746007,
"learning_rate": 3.1616759824664543e-06,
"logits": -1.1377207040786743,
"logps": -84.53422546386719,
"loss": 0.4624,
"objective": 0.46367743611335754,
"ranking_simple": 0.7458333373069763,
"step": 500
},
{
"epoch": 1.4170996693434104,
"eval_dpo_loss": 0.67823725938797,
"eval_logits": -1.2444933652877808,
"eval_logps": -89.6255874633789,
"eval_loss": 0.7469730973243713,
"eval_objective": 0.7386783957481384,
"eval_ranking_simple": 0.5533125996589661,
"eval_runtime": 371.626,
"eval_samples_per_second": 15.58,
"eval_steps_per_second": 1.3,
"step": 500
},
{
"dpo_loss": 0.3274858891963959,
"epoch": 1.4312706660368446,
"grad_norm": 7.65492855456284,
"learning_rate": 3.121724717912138e-06,
"logits": -1.1723324060440063,
"logps": -85.37480163574219,
"loss": 0.4519,
"objective": 0.45082953572273254,
"ranking_simple": 0.7666666507720947,
"step": 505
},
{
"dpo_loss": 0.3435133397579193,
"epoch": 1.4454416627302786,
"grad_norm": 7.695387713966003,
"learning_rate": 3.081603480027826e-06,
"logits": -1.1867634057998657,
"logps": -85.25582885742188,
"loss": 0.4577,
"objective": 0.4627363979816437,
"ranking_simple": 0.7416666746139526,
"step": 510
},
{
"dpo_loss": 0.33665701746940613,
"epoch": 1.4596126594237129,
"grad_norm": 6.852553951419641,
"learning_rate": 3.04132323755935e-06,
"logits": -1.2323737144470215,
"logps": -85.88379669189453,
"loss": 0.4456,
"objective": 0.4514175057411194,
"ranking_simple": 0.6875,
"step": 515
},
{
"dpo_loss": 0.29371899366378784,
"epoch": 1.473783656117147,
"grad_norm": 6.810862817314035,
"learning_rate": 3.0008950027228035e-06,
"logits": -1.1436675786972046,
"logps": -84.27217102050781,
"loss": 0.4264,
"objective": 0.42438414692878723,
"ranking_simple": 0.7749999761581421,
"step": 520
},
{
"dpo_loss": 0.3257172703742981,
"epoch": 1.487954652810581,
"grad_norm": 8.583092428586104,
"learning_rate": 2.960329828193918e-06,
"logits": -1.013688325881958,
"logps": -83.8447265625,
"loss": 0.462,
"objective": 0.45160239934921265,
"ranking_simple": 0.7416666746139526,
"step": 525
},
{
"dpo_loss": 0.32752183079719543,
"epoch": 1.5021256495040152,
"grad_norm": 7.949241972971005,
"learning_rate": 2.9196388040863695e-06,
"logits": -1.2097468376159668,
"logps": -87.19517517089844,
"loss": 0.4528,
"objective": 0.4474882483482361,
"ranking_simple": 0.7333333492279053,
"step": 530
},
{
"dpo_loss": 0.2954551577568054,
"epoch": 1.5162966461974492,
"grad_norm": 6.728845085484624,
"learning_rate": 2.8788330549198512e-06,
"logits": -1.2219938039779663,
"logps": -86.1001205444336,
"loss": 0.4194,
"objective": 0.4230054020881653,
"ranking_simple": 0.7541666626930237,
"step": 535
},
{
"dpo_loss": 0.3660537302494049,
"epoch": 1.5304676428908834,
"grad_norm": 8.075282408782796,
"learning_rate": 2.8379237365787426e-06,
"logits": -1.1332299709320068,
"logps": -86.49842834472656,
"loss": 0.4601,
"objective": 0.48352059721946716,
"ranking_simple": 0.7666666507720947,
"step": 540
},
{
"dpo_loss": 0.3429103195667267,
"epoch": 1.5446386395843175,
"grad_norm": 8.419322188891892,
"learning_rate": 2.7969220332622004e-06,
"logits": -1.2047460079193115,
"logps": -87.4766616821289,
"loss": 0.4602,
"objective": 0.4581466615200043,
"ranking_simple": 0.7749999761581421,
"step": 545
},
{
"dpo_loss": 0.2875075042247772,
"epoch": 1.5588096362777515,
"grad_norm": 7.159416578046968,
"learning_rate": 2.7558391544265127e-06,
"logits": -1.184915542602539,
"logps": -88.01736450195312,
"loss": 0.4391,
"objective": 0.41795456409454346,
"ranking_simple": 0.762499988079071,
"step": 550
},
{
"epoch": 1.5588096362777515,
"eval_dpo_loss": 0.6695489287376404,
"eval_logits": -1.1983243227005005,
"eval_logps": -91.99543762207031,
"eval_loss": 0.7384570837020874,
"eval_objective": 0.7304210066795349,
"eval_ranking_simple": 0.5486542582511902,
"eval_runtime": 368.2566,
"eval_samples_per_second": 15.723,
"eval_steps_per_second": 1.312,
"step": 550
},
{
"dpo_loss": 0.33032724261283875,
"epoch": 1.5729806329711855,
"grad_norm": 7.5647449089794305,
"learning_rate": 2.714686331720543e-06,
"logits": -0.9676509499549866,
"logps": -88.15595245361328,
"loss": 0.4441,
"objective": 0.44403356313705444,
"ranking_simple": 0.737500011920929,
"step": 555
},
{
"dpo_loss": 0.3146314024925232,
"epoch": 1.5871516296646198,
"grad_norm": 7.8310551029619555,
"learning_rate": 2.6734748159151104e-06,
"logits": -1.0463515520095825,
"logps": -85.61664581298828,
"loss": 0.4414,
"objective": 0.42937740683555603,
"ranking_simple": 0.7458333373069763,
"step": 560
},
{
"dpo_loss": 0.31714820861816406,
"epoch": 1.601322626358054,
"grad_norm": 7.188369880783801,
"learning_rate": 2.632215873827142e-06,
"logits": -1.049370527267456,
"logps": -86.86510467529297,
"loss": 0.4496,
"objective": 0.4433501362800598,
"ranking_simple": 0.7333333492279053,
"step": 565
},
{
"dpo_loss": 0.31138813495635986,
"epoch": 1.615493623051488,
"grad_norm": 7.613684765151376,
"learning_rate": 2.5909207852394363e-06,
"logits": -1.1900863647460938,
"logps": -88.4011001586914,
"loss": 0.4325,
"objective": 0.4364365339279175,
"ranking_simple": 0.7583333253860474,
"step": 570
},
{
"dpo_loss": 0.3405741751194,
"epoch": 1.629664619744922,
"grad_norm": 7.994470235278288,
"learning_rate": 2.5496008398168844e-06,
"logits": -1.1202160120010376,
"logps": -85.93099975585938,
"loss": 0.4543,
"objective": 0.46565988659858704,
"ranking_simple": 0.7666666507720947,
"step": 575
},
{
"dpo_loss": 0.32562312483787537,
"epoch": 1.643835616438356,
"grad_norm": 7.561927544761709,
"learning_rate": 2.508267334019988e-06,
"logits": -1.1339008808135986,
"logps": -85.58094787597656,
"loss": 0.4384,
"objective": 0.4460238814353943,
"ranking_simple": 0.7666666507720947,
"step": 580
},
{
"dpo_loss": 0.32099297642707825,
"epoch": 1.6580066131317903,
"grad_norm": 7.262395386385144,
"learning_rate": 2.46693156801652e-06,
"logits": -0.9802089333534241,
"logps": -84.47058868408203,
"loss": 0.4483,
"objective": 0.43592384457588196,
"ranking_simple": 0.7208333611488342,
"step": 585
},
{
"dpo_loss": 0.31580641865730286,
"epoch": 1.6721776098252243,
"grad_norm": 6.800312627956946,
"learning_rate": 2.4256048425921693e-06,
"logits": -1.01039457321167,
"logps": -86.4159164428711,
"loss": 0.444,
"objective": 0.44133689999580383,
"ranking_simple": 0.7541666626930237,
"step": 590
},
{
"dpo_loss": 0.32943814992904663,
"epoch": 1.6863486065186586,
"grad_norm": 6.937730325378872,
"learning_rate": 2.384298456061023e-06,
"logits": -0.9875913858413696,
"logps": -86.71006774902344,
"loss": 0.4342,
"objective": 0.44936403632164,
"ranking_simple": 0.75,
"step": 595
},
{
"dpo_loss": 0.28611433506011963,
"epoch": 1.7005196032120926,
"grad_norm": 7.2292134497805245,
"learning_rate": 2.3430237011767166e-06,
"logits": -1.0400419235229492,
"logps": -87.26065063476562,
"loss": 0.4285,
"objective": 0.4172123074531555,
"ranking_simple": 0.75,
"step": 600
},
{
"epoch": 1.7005196032120926,
"eval_dpo_loss": 0.6726422905921936,
"eval_logits": -1.1181021928787231,
"eval_logps": -91.40369415283203,
"eval_loss": 0.7407769560813904,
"eval_objective": 0.7317408919334412,
"eval_ranking_simple": 0.5502070188522339,
"eval_runtime": 368.21,
"eval_samples_per_second": 15.725,
"eval_steps_per_second": 1.312,
"step": 600
},
{
"dpo_loss": 0.31679973006248474,
"epoch": 1.7146905999055266,
"grad_norm": 8.641164554516212,
"learning_rate": 2.30179186204511e-06,
"logits": -0.9793424606323242,
"logps": -87.53156280517578,
"loss": 0.4514,
"objective": 0.439665287733078,
"ranking_simple": 0.7291666865348816,
"step": 605
},
{
"dpo_loss": 0.31452521681785583,
"epoch": 1.7288615965989607,
"grad_norm": 8.523948659267386,
"learning_rate": 2.2606142110393248e-06,
"logits": -1.05043625831604,
"logps": -86.67326354980469,
"loss": 0.4321,
"objective": 0.4331686198711395,
"ranking_simple": 0.7833333611488342,
"step": 610
},
{
"dpo_loss": 0.3183232843875885,
"epoch": 1.743032593292395,
"grad_norm": 7.54765334989791,
"learning_rate": 2.2195020057179897e-06,
"logits": -1.0829071998596191,
"logps": -87.68238830566406,
"loss": 0.4403,
"objective": 0.4391016960144043,
"ranking_simple": 0.7166666388511658,
"step": 615
},
{
"dpo_loss": 0.31284984946250916,
"epoch": 1.7572035899858292,
"grad_norm": 8.306713660930198,
"learning_rate": 2.1784664857475356e-06,
"logits": -1.133055567741394,
"logps": -88.2243881225586,
"loss": 0.4481,
"objective": 0.4363309144973755,
"ranking_simple": 0.7583333253860474,
"step": 620
},
{
"dpo_loss": 0.3389653265476227,
"epoch": 1.7713745866792632,
"grad_norm": 8.27642622610812,
"learning_rate": 2.1375188698293855e-06,
"logits": -1.0864481925964355,
"logps": -87.53520202636719,
"loss": 0.4464,
"objective": 0.4615752100944519,
"ranking_simple": 0.7458333373069763,
"step": 625
},
{
"dpo_loss": 0.3449970781803131,
"epoch": 1.7855455833726972,
"grad_norm": 7.582885342390072,
"learning_rate": 2.096670352632873e-06,
"logits": -1.17921781539917,
"logps": -88.3585205078125,
"loss": 0.4521,
"objective": 0.4663265645503998,
"ranking_simple": 0.7124999761581421,
"step": 630
},
{
"dpo_loss": 0.2966757118701935,
"epoch": 1.7997165800661312,
"grad_norm": 7.391390201817734,
"learning_rate": 2.0559321017347286e-06,
"logits": -1.3011022806167603,
"logps": -87.97920989990234,
"loss": 0.4423,
"objective": 0.42836812138557434,
"ranking_simple": 0.7416666746139526,
"step": 635
},
{
"dpo_loss": 0.2997787892818451,
"epoch": 1.8138875767595655,
"grad_norm": 7.910229224489209,
"learning_rate": 2.01531525456598e-06,
"logits": -1.2456448078155518,
"logps": -88.25955963134766,
"loss": 0.4321,
"objective": 0.43085435032844543,
"ranking_simple": 0.7291666865348816,
"step": 640
},
{
"dpo_loss": 0.3190990388393402,
"epoch": 1.8280585734529995,
"grad_norm": 7.336498748426566,
"learning_rate": 1.974830915367086e-06,
"logits": -1.2663050889968872,
"logps": -88.02909851074219,
"loss": 0.448,
"objective": 0.44558364152908325,
"ranking_simple": 0.7416666746139526,
"step": 645
},
{
"dpo_loss": 0.31604960560798645,
"epoch": 1.8422295701464337,
"grad_norm": 8.01809767773599,
"learning_rate": 1.93449015215215e-06,
"logits": -1.3034510612487793,
"logps": -87.49658966064453,
"loss": 0.4553,
"objective": 0.4388020634651184,
"ranking_simple": 0.7583333253860474,
"step": 650
},
{
"epoch": 1.8422295701464337,
"eval_dpo_loss": 0.6739967465400696,
"eval_logits": -1.272489070892334,
"eval_logps": -90.41595458984375,
"eval_loss": 0.7425940036773682,
"eval_objective": 0.7334864139556885,
"eval_ranking_simple": 0.5559006333351135,
"eval_runtime": 368.2584,
"eval_samples_per_second": 15.723,
"eval_steps_per_second": 1.312,
"step": 650
},
{
"dpo_loss": 0.3332468867301941,
"epoch": 1.8564005668398678,
"grad_norm": 6.515064472181348,
"learning_rate": 1.8943039936830347e-06,
"logits": -1.153478980064392,
"logps": -83.99712371826172,
"loss": 0.4429,
"objective": 0.441272109746933,
"ranking_simple": 0.7416666746139526,
"step": 655
},
{
"dpo_loss": 0.3361124098300934,
"epoch": 1.8705715635333018,
"grad_norm": 7.186294756150463,
"learning_rate": 1.8542834264542091e-06,
"logits": -1.1785982847213745,
"logps": -87.21165466308594,
"loss": 0.4408,
"objective": 0.4524908661842346,
"ranking_simple": 0.737500011920929,
"step": 660
},
{
"dpo_loss": 0.30988797545433044,
"epoch": 1.8847425602267358,
"grad_norm": 7.223583667691921,
"learning_rate": 1.814439391689151e-06,
"logits": -1.1908276081085205,
"logps": -85.18921661376953,
"loss": 0.4392,
"objective": 0.4265805184841156,
"ranking_simple": 0.7875000238418579,
"step": 665
},
{
"dpo_loss": 0.3152609169483185,
"epoch": 1.89891355692017,
"grad_norm": 7.236738362974331,
"learning_rate": 1.7747827823491253e-06,
"logits": -1.1136568784713745,
"logps": -85.27536010742188,
"loss": 0.437,
"objective": 0.43872639536857605,
"ranking_simple": 0.7166666388511658,
"step": 670
},
{
"dpo_loss": 0.3269253373146057,
"epoch": 1.9130845536136043,
"grad_norm": 7.561018148695705,
"learning_rate": 1.7353244401551566e-06,
"logits": -1.1110624074935913,
"logps": -86.15958404541016,
"loss": 0.4438,
"objective": 0.45002037286758423,
"ranking_simple": 0.7916666865348816,
"step": 675
},
{
"dpo_loss": 0.32910048961639404,
"epoch": 1.9272555503070383,
"grad_norm": 7.152885290887085,
"learning_rate": 1.6960751526240122e-06,
"logits": -1.1628243923187256,
"logps": -88.80421447753906,
"loss": 0.4433,
"objective": 0.4564170241355896,
"ranking_simple": 0.8041666746139526,
"step": 680
},
{
"dpo_loss": 0.2932046949863434,
"epoch": 1.9414265470004723,
"grad_norm": 6.233579110675445,
"learning_rate": 1.6570456501189996e-06,
"logits": -1.1053887605667114,
"logps": -87.36367797851562,
"loss": 0.4164,
"objective": 0.424954891204834,
"ranking_simple": 0.699999988079071,
"step": 685
},
{
"dpo_loss": 0.3040808141231537,
"epoch": 1.9555975436939064,
"grad_norm": 7.4587976233319795,
"learning_rate": 1.6182466029163974e-06,
"logits": -1.1335686445236206,
"logps": -87.29252624511719,
"loss": 0.439,
"objective": 0.4274420142173767,
"ranking_simple": 0.75,
"step": 690
},
{
"dpo_loss": 0.33969834446907043,
"epoch": 1.9697685403873406,
"grad_norm": 8.387180426204326,
"learning_rate": 1.5796886182883053e-06,
"logits": -1.048902988433838,
"logps": -87.7245101928711,
"loss": 0.4375,
"objective": 0.44796222448349,
"ranking_simple": 0.75,
"step": 695
},
{
"dpo_loss": 0.3006362318992615,
"epoch": 1.9839395370807746,
"grad_norm": 7.023996943402399,
"learning_rate": 1.541382237602721e-06,
"logits": -1.0404043197631836,
"logps": -88.86479187011719,
"loss": 0.4307,
"objective": 0.4227179288864136,
"ranking_simple": 0.7416666746139526,
"step": 700
},
{
"epoch": 1.9839395370807746,
"eval_dpo_loss": 0.6734952926635742,
"eval_logits": -1.2351235151290894,
"eval_logps": -91.78550720214844,
"eval_loss": 0.740387499332428,
"eval_objective": 0.7341719269752502,
"eval_ranking_simple": 0.5584886074066162,
"eval_runtime": 367.9084,
"eval_samples_per_second": 15.738,
"eval_steps_per_second": 1.313,
"step": 700
},
{
"dpo_loss": 0.3133445382118225,
"epoch": 1.9981105337742089,
"grad_norm": 7.983253403779177,
"learning_rate": 1.5033379334416376e-06,
"logits": -0.9815789461135864,
"logps": -87.82261657714844,
"loss": 0.448,
"objective": 0.4431394636631012,
"ranking_simple": 0.7208333611488342,
"step": 705
},
{
"dpo_loss": 0.26884225010871887,
"epoch": 2.012281530467643,
"grad_norm": 4.711433926223709,
"learning_rate": 1.465566106737942e-06,
"logits": -1.132247805595398,
"logps": -86.83318328857422,
"loss": 0.3883,
"objective": 0.4008704125881195,
"ranking_simple": 0.800000011920929,
"step": 710
},
{
"dpo_loss": 0.24094882607460022,
"epoch": 2.026452527161077,
"grad_norm": 4.45009467250422,
"learning_rate": 1.4280770839319073e-06,
"logits": -1.1357406377792358,
"logps": -85.59795379638672,
"loss": 0.3752,
"objective": 0.37576305866241455,
"ranking_simple": 0.8333333134651184,
"step": 715
},
{
"dpo_loss": 0.22363178431987762,
"epoch": 2.040623523854511,
"grad_norm": 4.462329734353584,
"learning_rate": 1.3908811141480408e-06,
"logits": -1.1281845569610596,
"logps": -86.89059448242188,
"loss": 0.371,
"objective": 0.3674115538597107,
"ranking_simple": 0.8166666626930237,
"step": 720
},
{
"dpo_loss": 0.22502318024635315,
"epoch": 2.0547945205479454,
"grad_norm": 4.243136918683414,
"learning_rate": 1.353988366393083e-06,
"logits": -1.1830660104751587,
"logps": -89.62688446044922,
"loss": 0.3719,
"objective": 0.37353357672691345,
"ranking_simple": 0.7958333492279053,
"step": 725
},
{
"dpo_loss": 0.24978047609329224,
"epoch": 2.0689655172413794,
"grad_norm": 4.567188933527537,
"learning_rate": 1.3174089267758983e-06,
"logits": -1.2705327272415161,
"logps": -87.48053741455078,
"loss": 0.3752,
"objective": 0.3849594295024872,
"ranking_simple": 0.7833333611488342,
"step": 730
},
{
"dpo_loss": 0.25259262323379517,
"epoch": 2.0831365139348135,
"grad_norm": 5.151937950415065,
"learning_rate": 1.2811527957500344e-06,
"logits": -1.1913000345230103,
"logps": -87.4656753540039,
"loss": 0.3812,
"objective": 0.3858603835105896,
"ranking_simple": 0.762499988079071,
"step": 735
},
{
"dpo_loss": 0.2209930121898651,
"epoch": 2.0973075106282475,
"grad_norm": 4.254397078002712,
"learning_rate": 1.245229885379699e-06,
"logits": -1.2872763872146606,
"logps": -86.62919616699219,
"loss": 0.3727,
"objective": 0.3619976043701172,
"ranking_simple": 0.8458333611488342,
"step": 740
},
{
"dpo_loss": 0.2264253944158554,
"epoch": 2.1114785073216815,
"grad_norm": 4.542965704767376,
"learning_rate": 1.2096500166298992e-06,
"logits": -1.2000243663787842,
"logps": -88.8626937866211,
"loss": 0.3762,
"objective": 0.3712898790836334,
"ranking_simple": 0.8125,
"step": 745
},
{
"dpo_loss": 0.23141829669475555,
"epoch": 2.1256495040151155,
"grad_norm": 4.288133779032122,
"learning_rate": 1.1744229166814889e-06,
"logits": -1.2524789571762085,
"logps": -87.17241668701172,
"loss": 0.3755,
"objective": 0.37335434556007385,
"ranking_simple": 0.7791666388511658,
"step": 750
},
{
"epoch": 2.1256495040151155,
"eval_dpo_loss": 0.6762245893478394,
"eval_logits": -1.3013161420822144,
"eval_logps": -93.2393569946289,
"eval_loss": 0.7430183291435242,
"eval_objective": 0.736893355846405,
"eval_ranking_simple": 0.5486542582511902,
"eval_runtime": 368.204,
"eval_samples_per_second": 15.725,
"eval_steps_per_second": 1.312,
"step": 750
},
{
"dpo_loss": 0.2273796647787094,
"epoch": 2.13982050070855,
"grad_norm": 4.5000688322821345,
"learning_rate": 1.1395582162718524e-06,
"logits": -1.2545628547668457,
"logps": -90.6506576538086,
"loss": 0.3758,
"objective": 0.3768024444580078,
"ranking_simple": 0.8083333373069763,
"step": 755
},
{
"dpo_loss": 0.2525111138820648,
"epoch": 2.153991497401984,
"grad_norm": 5.011749929420246,
"learning_rate": 1.1050654470619602e-06,
"logits": -1.1907628774642944,
"logps": -85.97274017333984,
"loss": 0.3783,
"objective": 0.39089202880859375,
"ranking_simple": 0.7875000238418579,
"step": 760
},
{
"dpo_loss": 0.23322957754135132,
"epoch": 2.168162494095418,
"grad_norm": 5.201612157262786,
"learning_rate": 1.0709540390305061e-06,
"logits": -1.2566605806350708,
"logps": -87.6942367553711,
"loss": 0.3849,
"objective": 0.3768391013145447,
"ranking_simple": 0.8125,
"step": 765
},
{
"dpo_loss": 0.22707949578762054,
"epoch": 2.182333490788852,
"grad_norm": 4.70373130482108,
"learning_rate": 1.0372333178958462e-06,
"logits": -1.2721644639968872,
"logps": -88.22640991210938,
"loss": 0.377,
"objective": 0.3728346526622772,
"ranking_simple": 0.7875000238418579,
"step": 770
},
{
"dpo_loss": 0.22152571380138397,
"epoch": 2.196504487482286,
"grad_norm": 5.587786749861536,
"learning_rate": 1.0039125025664392e-06,
"logits": -1.2534898519515991,
"logps": -89.17953491210938,
"loss": 0.3725,
"objective": 0.3679354786872864,
"ranking_simple": 0.8208333253860474,
"step": 775
},
{
"dpo_loss": 0.250379741191864,
"epoch": 2.21067548417572,
"grad_norm": 5.002867020526872,
"learning_rate": 9.710007026204896e-07,
"logits": -1.1392863988876343,
"logps": -89.51647186279297,
"loss": 0.3774,
"objective": 0.39055460691452026,
"ranking_simple": 0.737500011920929,
"step": 780
},
{
"dpo_loss": 0.22285352647304535,
"epoch": 2.2248464808691546,
"grad_norm": 5.843250372899321,
"learning_rate": 9.385069158154805e-07,
"logits": -1.1372294425964355,
"logps": -87.82280731201172,
"loss": 0.3756,
"objective": 0.37054911255836487,
"ranking_simple": 0.7916666865348816,
"step": 785
},
{
"dpo_loss": 0.2125234156847,
"epoch": 2.2390174775625886,
"grad_norm": 4.914317874167112,
"learning_rate": 9.064400256282757e-07,
"logits": -1.1291053295135498,
"logps": -88.7917709350586,
"loss": 0.3784,
"objective": 0.36886271834373474,
"ranking_simple": 0.8500000238418579,
"step": 790
},
{
"dpo_loss": 0.24073070287704468,
"epoch": 2.2531884742560226,
"grad_norm": 4.77090719779272,
"learning_rate": 8.74808798826467e-07,
"logits": -1.216113805770874,
"logps": -89.64153289794922,
"loss": 0.3768,
"objective": 0.37714874744415283,
"ranking_simple": 0.824999988079071,
"step": 795
},
{
"dpo_loss": 0.23036065697669983,
"epoch": 2.2673594709494567,
"grad_norm": 4.57839809888007,
"learning_rate": 8.436218830716259e-07,
"logits": -1.2197948694229126,
"logps": -90.15279388427734,
"loss": 0.3794,
"objective": 0.3776319622993469,
"ranking_simple": 0.8333333134651184,
"step": 800
},
{
"epoch": 2.2673594709494567,
"eval_dpo_loss": 0.6726187467575073,
"eval_logits": -1.2647498846054077,
"eval_logps": -93.31334686279297,
"eval_loss": 0.7400006651878357,
"eval_objective": 0.7334582805633545,
"eval_ranking_simple": 0.554347813129425,
"eval_runtime": 370.3208,
"eval_samples_per_second": 15.635,
"eval_steps_per_second": 1.304,
"step": 800
},
{
"dpo_loss": 0.24638353288173676,
"epoch": 2.2815304676428907,
"grad_norm": 4.775498293092394,
"learning_rate": 8.1288780455512e-07,
"logits": -1.2070590257644653,
"logps": -88.29240417480469,
"loss": 0.38,
"objective": 0.3882957696914673,
"ranking_simple": 0.7916666865348816,
"step": 805
},
{
"dpo_loss": 0.26376059651374817,
"epoch": 2.295701464336325,
"grad_norm": 4.107962497004369,
"learning_rate": 7.826149656671386e-07,
"logits": -1.3075222969055176,
"logps": -87.07465362548828,
"loss": 0.3775,
"objective": 0.38518133759498596,
"ranking_simple": 0.8208333253860474,
"step": 810
},
{
"dpo_loss": 0.23214052617549896,
"epoch": 2.309872461029759,
"grad_norm": 4.032286949010579,
"learning_rate": 7.528116426995605e-07,
"logits": -1.2393053770065308,
"logps": -89.05374145507812,
"loss": 0.3692,
"objective": 0.3722856342792511,
"ranking_simple": 0.7833333611488342,
"step": 815
},
{
"dpo_loss": 0.24578505754470825,
"epoch": 2.324043457723193,
"grad_norm": 4.0007768416623755,
"learning_rate": 7.234859835833022e-07,
"logits": -1.1707955598831177,
"logps": -88.1304702758789,
"loss": 0.3733,
"objective": 0.377387672662735,
"ranking_simple": 0.762499988079071,
"step": 820
},
{
"dpo_loss": 0.23438973724842072,
"epoch": 2.3382144544166272,
"grad_norm": 4.645717749878006,
"learning_rate": 6.94646005660749e-07,
"logits": -1.197273850440979,
"logps": -87.03529357910156,
"loss": 0.3747,
"objective": 0.3757517635822296,
"ranking_simple": 0.8083333373069763,
"step": 825
},
{
"dpo_loss": 0.22819119691848755,
"epoch": 2.3523854511100613,
"grad_norm": 5.87351004355499,
"learning_rate": 6.662995934939007e-07,
"logits": -1.2045276165008545,
"logps": -89.30978393554688,
"loss": 0.3712,
"objective": 0.3776562809944153,
"ranking_simple": 0.8125,
"step": 830
},
{
"dpo_loss": 0.24645249545574188,
"epoch": 2.3665564478034957,
"grad_norm": 5.364960343792073,
"learning_rate": 6.384544967088063e-07,
"logits": -1.177357792854309,
"logps": -88.5252914428711,
"loss": 0.3869,
"objective": 0.38127484917640686,
"ranking_simple": 0.8208333253860474,
"step": 835
},
{
"dpo_loss": 0.24418006837368011,
"epoch": 2.3807274444969297,
"grad_norm": 5.224294646577439,
"learning_rate": 6.111183278768956e-07,
"logits": -1.1215898990631104,
"logps": -88.96337890625,
"loss": 0.3837,
"objective": 0.38468509912490845,
"ranking_simple": 0.7583333253860474,
"step": 840
},
{
"dpo_loss": 0.2479609102010727,
"epoch": 2.3948984411903638,
"grad_norm": 5.589943019855736,
"learning_rate": 5.842985604337769e-07,
"logits": -1.1818034648895264,
"logps": -90.55766296386719,
"loss": 0.3851,
"objective": 0.3813749849796295,
"ranking_simple": 0.8041666746139526,
"step": 845
},
{
"dpo_loss": 0.22911126911640167,
"epoch": 2.409069437883798,
"grad_norm": 4.122848222733345,
"learning_rate": 5.580025266360764e-07,
"logits": -1.2163876295089722,
"logps": -88.67691040039062,
"loss": 0.373,
"objective": 0.37054920196533203,
"ranking_simple": 0.7749999761581421,
"step": 850
},
{
"epoch": 2.409069437883798,
"eval_dpo_loss": 0.6747334599494934,
"eval_logits": -1.2593425512313843,
"eval_logps": -92.93883514404297,
"eval_loss": 0.74095219373703,
"eval_objective": 0.7354456186294556,
"eval_ranking_simple": 0.5522774457931519,
"eval_runtime": 376.5034,
"eval_samples_per_second": 15.378,
"eval_steps_per_second": 1.283,
"step": 850
},
{
"dpo_loss": 0.24901343882083893,
"epoch": 2.423240434577232,
"grad_norm": 5.545917724699393,
"learning_rate": 5.322374155568688e-07,
"logits": -1.2266477346420288,
"logps": -87.11955261230469,
"loss": 0.3932,
"objective": 0.3926084637641907,
"ranking_simple": 0.800000011920929,
"step": 855
},
{
"dpo_loss": 0.24267633259296417,
"epoch": 2.4374114312706663,
"grad_norm": 4.734435718269958,
"learning_rate": 5.070102711202606e-07,
"logits": -1.1731507778167725,
"logps": -89.50730895996094,
"loss": 0.3789,
"objective": 0.38105666637420654,
"ranking_simple": 0.7916666865348816,
"step": 860
},
{
"dpo_loss": 0.23941467702388763,
"epoch": 2.4515824279641003,
"grad_norm": 5.519011513783408,
"learning_rate": 4.823279901756498e-07,
"logits": -1.204505443572998,
"logps": -88.21556854248047,
"loss": 0.3821,
"objective": 0.38131964206695557,
"ranking_simple": 0.762499988079071,
"step": 865
},
{
"dpo_loss": 0.2390858381986618,
"epoch": 2.4657534246575343,
"grad_norm": 4.890994580870049,
"learning_rate": 4.581973206121948e-07,
"logits": -1.2685866355895996,
"logps": -88.25513458251953,
"loss": 0.381,
"objective": 0.3830280900001526,
"ranking_simple": 0.7791666388511658,
"step": 870
},
{
"dpo_loss": 0.2303398698568344,
"epoch": 2.4799244213509684,
"grad_norm": 5.062132747484294,
"learning_rate": 4.3462485951401126e-07,
"logits": -1.204114556312561,
"logps": -88.9038314819336,
"loss": 0.3766,
"objective": 0.38060086965560913,
"ranking_simple": 0.8208333253860474,
"step": 875
},
{
"dpo_loss": 0.2502264678478241,
"epoch": 2.4940954180444024,
"grad_norm": 5.573699785814543,
"learning_rate": 4.116170513565942e-07,
"logits": -1.1215044260025024,
"logps": -87.70077514648438,
"loss": 0.3832,
"objective": 0.3906749486923218,
"ranking_simple": 0.7458333373069763,
"step": 880
},
{
"dpo_loss": 0.2577410042285919,
"epoch": 2.5082664147378364,
"grad_norm": 4.956979588584702,
"learning_rate": 3.891801862449629e-07,
"logits": -1.3180614709854126,
"logps": -86.54364776611328,
"loss": 0.3894,
"objective": 0.39261943101882935,
"ranking_simple": 0.800000011920929,
"step": 885
},
{
"dpo_loss": 0.21435794234275818,
"epoch": 2.5224374114312704,
"grad_norm": 5.7378961390152385,
"learning_rate": 3.6732039819400686e-07,
"logits": -1.2147475481033325,
"logps": -85.97087097167969,
"loss": 0.3823,
"objective": 0.3677336275577545,
"ranking_simple": 0.8291666507720947,
"step": 890
},
{
"dpo_loss": 0.24500012397766113,
"epoch": 2.536608408124705,
"grad_norm": 5.350118064676714,
"learning_rate": 3.46043663451511e-07,
"logits": -1.2147990465164185,
"logps": -88.52668762207031,
"loss": 0.3803,
"objective": 0.3881225883960724,
"ranking_simple": 0.762499988079071,
"step": 895
},
{
"dpo_loss": 0.27485665678977966,
"epoch": 2.550779404818139,
"grad_norm": 5.3127285866105405,
"learning_rate": 3.253557988643072e-07,
"logits": -1.1600372791290283,
"logps": -86.77176666259766,
"loss": 0.388,
"objective": 0.40204280614852905,
"ranking_simple": 0.7875000238418579,
"step": 900
},
{
"epoch": 2.550779404818139,
"eval_dpo_loss": 0.6756879687309265,
"eval_logits": -1.2939373254776,
"eval_logps": -92.89238739013672,
"eval_loss": 0.7418231964111328,
"eval_objective": 0.7363364696502686,
"eval_ranking_simple": 0.5502070188522339,
"eval_runtime": 369.0318,
"eval_samples_per_second": 15.69,
"eval_steps_per_second": 1.309,
"step": 900
},
{
"dpo_loss": 0.23769782483577728,
"epoch": 2.564950401511573,
"grad_norm": 4.493878506991506,
"learning_rate": 3.052624602880064e-07,
"logits": -1.2237998247146606,
"logps": -85.79032135009766,
"loss": 0.3781,
"objective": 0.3822983205318451,
"ranking_simple": 0.7875000238418579,
"step": 905
},
{
"dpo_loss": 0.23715509474277496,
"epoch": 2.579121398205007,
"grad_norm": 5.776134875432629,
"learning_rate": 2.8576914104074425e-07,
"logits": -1.2153693437576294,
"logps": -89.9877700805664,
"loss": 0.38,
"objective": 0.3822348713874817,
"ranking_simple": 0.7833333611488342,
"step": 910
},
{
"dpo_loss": 0.241657093167305,
"epoch": 2.593292394898441,
"grad_norm": 3.828702735136657,
"learning_rate": 2.6688117040136463e-07,
"logits": -1.3503960371017456,
"logps": -88.94493865966797,
"loss": 0.3732,
"objective": 0.3829546570777893,
"ranking_simple": 0.8083333373069763,
"step": 915
},
{
"dpo_loss": 0.2292027622461319,
"epoch": 2.6074633915918755,
"grad_norm": 4.721702099688327,
"learning_rate": 2.486037121524448e-07,
"logits": -1.1647106409072876,
"logps": -88.75472259521484,
"loss": 0.3765,
"objective": 0.3745788335800171,
"ranking_simple": 0.8125,
"step": 920
},
{
"dpo_loss": 0.22531941533088684,
"epoch": 2.6216343882853095,
"grad_norm": 4.89488363414913,
"learning_rate": 2.3094176316856982e-07,
"logits": -1.1882636547088623,
"logps": -87.45687866210938,
"loss": 0.3799,
"objective": 0.373484343290329,
"ranking_simple": 0.800000011920929,
"step": 925
},
{
"dpo_loss": 0.2518150508403778,
"epoch": 2.6358053849787435,
"grad_norm": 5.096542868734766,
"learning_rate": 2.13900152050239e-07,
"logits": -1.1838154792785645,
"logps": -87.14102935791016,
"loss": 0.3831,
"objective": 0.39408108592033386,
"ranking_simple": 0.7416666746139526,
"step": 930
},
{
"dpo_loss": 0.2513907849788666,
"epoch": 2.6499763816721775,
"grad_norm": 5.6389448500299935,
"learning_rate": 1.9748353780377234e-07,
"logits": -1.1840038299560547,
"logps": -86.4458999633789,
"loss": 0.3954,
"objective": 0.3891184628009796,
"ranking_simple": 0.8291666507720947,
"step": 935
},
{
"dpo_loss": 0.22740763425827026,
"epoch": 2.6641473783656116,
"grad_norm": 4.892959475477896,
"learning_rate": 1.8169640856758652e-07,
"logits": -1.2362346649169922,
"logps": -89.93408203125,
"loss": 0.3789,
"objective": 0.3699011206626892,
"ranking_simple": 0.8041666746139526,
"step": 940
},
{
"dpo_loss": 0.2212921679019928,
"epoch": 2.678318375059046,
"grad_norm": 4.26351950100694,
"learning_rate": 1.6654308038518057e-07,
"logits": -1.2807658910751343,
"logps": -88.23102569580078,
"loss": 0.3726,
"objective": 0.36465439200401306,
"ranking_simple": 0.800000011920929,
"step": 945
},
{
"dpo_loss": 0.24647028744220734,
"epoch": 2.69248937175248,
"grad_norm": 5.222304477193562,
"learning_rate": 1.5202769602517514e-07,
"logits": -1.1635686159133911,
"logps": -86.58324432373047,
"loss": 0.3866,
"objective": 0.3892979621887207,
"ranking_simple": 0.7958333492279053,
"step": 950
},
{
"epoch": 2.69248937175248,
"eval_dpo_loss": 0.6751933097839355,
"eval_logits": -1.2937095165252686,
"eval_logps": -92.32901763916016,
"eval_loss": 0.7418193221092224,
"eval_objective": 0.7358114719390869,
"eval_ranking_simple": 0.5507246255874634,
"eval_runtime": 369.9174,
"eval_samples_per_second": 15.652,
"eval_steps_per_second": 1.306,
"step": 950
},
{
"dpo_loss": 0.2517485022544861,
"epoch": 2.706660368445914,
"grad_norm": 4.388748582328918,
"learning_rate": 1.381542238487188e-07,
"logits": -1.243085265159607,
"logps": -88.0900650024414,
"loss": 0.3837,
"objective": 0.38995641469955444,
"ranking_simple": 0.8041666746139526,
"step": 955
},
{
"dpo_loss": 0.25272881984710693,
"epoch": 2.720831365139348,
"grad_norm": 5.315737788533154,
"learning_rate": 1.2492645672457838e-07,
"logits": -1.2224252223968506,
"logps": -87.57622528076172,
"loss": 0.3841,
"objective": 0.39599618315696716,
"ranking_simple": 0.7958333492279053,
"step": 960
},
{
"dpo_loss": 0.22980408370494843,
"epoch": 2.735002361832782,
"grad_norm": 5.693843860804482,
"learning_rate": 1.1234801099220787e-07,
"logits": -1.2591173648834229,
"logps": -87.08290100097656,
"loss": 0.3861,
"objective": 0.3821989893913269,
"ranking_simple": 0.8500000238418579,
"step": 965
},
{
"dpo_loss": 0.26472774147987366,
"epoch": 2.7491733585262166,
"grad_norm": 4.654019056163642,
"learning_rate": 1.004223254730749e-07,
"logits": -1.2617594003677368,
"logps": -88.49696350097656,
"loss": 0.3797,
"objective": 0.39959490299224854,
"ranking_simple": 0.8458333611488342,
"step": 970
},
{
"dpo_loss": 0.22332945466041565,
"epoch": 2.7633443552196506,
"grad_norm": 5.199760236696295,
"learning_rate": 8.915266053052374e-08,
"logits": -1.1656602621078491,
"logps": -86.72117614746094,
"loss": 0.3766,
"objective": 0.36580052971839905,
"ranking_simple": 0.8083333373069763,
"step": 975
},
{
"dpo_loss": 0.22700704634189606,
"epoch": 2.7775153519130846,
"grad_norm": 4.6165629384250435,
"learning_rate": 7.854209717842231e-08,
"logits": -1.3093292713165283,
"logps": -87.9115982055664,
"loss": 0.3728,
"objective": 0.37029343843460083,
"ranking_simple": 0.8291666507720947,
"step": 980
},
{
"dpo_loss": 0.23392988741397858,
"epoch": 2.7916863486065187,
"grad_norm": 4.595783043627286,
"learning_rate": 6.859353623884569e-08,
"logits": -1.1338623762130737,
"logps": -84.75395202636719,
"loss": 0.3729,
"objective": 0.3690672218799591,
"ranking_simple": 0.8083333373069763,
"step": 985
},
{
"dpo_loss": 0.24494783580303192,
"epoch": 2.8058573452999527,
"grad_norm": 4.754488601809905,
"learning_rate": 5.930969754901844e-08,
"logits": -1.1965256929397583,
"logps": -88.29127502441406,
"loss": 0.3801,
"objective": 0.38465526700019836,
"ranking_simple": 0.7749999761581421,
"step": 990
},
{
"dpo_loss": 0.24409013986587524,
"epoch": 2.820028341993387,
"grad_norm": 5.4353357294596245,
"learning_rate": 5.069311921774039e-08,
"logits": -1.2917786836624146,
"logps": -88.09101867675781,
"loss": 0.3817,
"objective": 0.3790687918663025,
"ranking_simple": 0.7916666865348816,
"step": 995
},
{
"dpo_loss": 0.24059538543224335,
"epoch": 2.8341993386868207,
"grad_norm": 4.725884973016709,
"learning_rate": 4.2746156931490756e-08,
"logits": -1.2047442197799683,
"logps": -85.76768493652344,
"loss": 0.3828,
"objective": 0.3797774612903595,
"ranking_simple": 0.8291666507720947,
"step": 1000
},
{
"epoch": 2.8341993386868207,
"eval_dpo_loss": 0.6749303340911865,
"eval_logits": -1.2945847511291504,
"eval_logps": -92.32601928710938,
"eval_loss": 0.7417021989822388,
"eval_objective": 0.7355546355247498,
"eval_ranking_simple": 0.5502070188522339,
"eval_runtime": 369.447,
"eval_samples_per_second": 15.672,
"eval_steps_per_second": 1.307,
"step": 1000
},
{
"dpo_loss": 0.23444737493991852,
"epoch": 2.848370335380255,
"grad_norm": 5.091471401481779,
"learning_rate": 3.547098331040916e-08,
"logits": -1.2024807929992676,
"logps": -86.74796295166016,
"loss": 0.382,
"objective": 0.3817732632160187,
"ranking_simple": 0.8041666746139526,
"step": 1005
},
{
"dpo_loss": 0.2292069047689438,
"epoch": 2.862541332073689,
"grad_norm": 4.7403703778744815,
"learning_rate": 2.8869587314321324e-08,
"logits": -1.2810461521148682,
"logps": -87.0550308227539,
"loss": 0.379,
"objective": 0.37782254815101624,
"ranking_simple": 0.7791666388511658,
"step": 1010
},
{
"dpo_loss": 0.2399337738752365,
"epoch": 2.8767123287671232,
"grad_norm": 5.329975170501087,
"learning_rate": 2.2943773698977935e-08,
"logits": -1.197946548461914,
"logps": -85.78963470458984,
"loss": 0.3813,
"objective": 0.3823747932910919,
"ranking_simple": 0.7916666865348816,
"step": 1015
},
{
"dpo_loss": 0.23225615918636322,
"epoch": 2.8908833254605573,
"grad_norm": 4.141247666962683,
"learning_rate": 1.7695162522652352e-08,
"logits": -1.1521192789077759,
"logps": -87.80615234375,
"loss": 0.383,
"objective": 0.3741118311882019,
"ranking_simple": 0.75,
"step": 1020
},
{
"dpo_loss": 0.24901318550109863,
"epoch": 2.9050543221539913,
"grad_norm": 4.809794853364304,
"learning_rate": 1.3125188703233815e-08,
"logits": -1.2677100896835327,
"logps": -87.76679229736328,
"loss": 0.3815,
"objective": 0.3834156394004822,
"ranking_simple": 0.8083333373069763,
"step": 1025
},
{
"dpo_loss": 0.2290908843278885,
"epoch": 2.9192253188474258,
"grad_norm": 3.635733503543428,
"learning_rate": 9.235101625932885e-09,
"logits": -1.3011952638626099,
"logps": -86.35026550292969,
"loss": 0.3798,
"objective": 0.3711520731449127,
"ranking_simple": 0.762499988079071,
"step": 1030
},
{
"dpo_loss": 0.23674419522285461,
"epoch": 2.9333963155408598,
"grad_norm": 4.9577484155223,
"learning_rate": 6.025964801714412e-09,
"logits": -1.2497870922088623,
"logps": -88.07068634033203,
"loss": 0.3866,
"objective": 0.3858194053173065,
"ranking_simple": 0.8083333373069763,
"step": 1035
},
{
"dpo_loss": 0.23205497860908508,
"epoch": 2.947567312234294,
"grad_norm": 4.605543722963177,
"learning_rate": 3.4986555765434415e-09,
"logits": -1.2593365907669067,
"logps": -88.1611099243164,
"loss": 0.3798,
"objective": 0.3837084472179413,
"ranking_simple": 0.7875000238418579,
"step": 1040
},
{
"dpo_loss": 0.24727579951286316,
"epoch": 2.961738308927728,
"grad_norm": 4.997984528026836,
"learning_rate": 1.6538648915270794e-09,
"logits": -1.2010886669158936,
"logps": -90.2540512084961,
"loss": 0.3835,
"objective": 0.3843297064304352,
"ranking_simple": 0.7833333611488342,
"step": 1045
},
{
"dpo_loss": 0.23686116933822632,
"epoch": 2.975909305621162,
"grad_norm": 4.330333752032827,
"learning_rate": 4.920970940180958e-10,
"logits": -1.2512930631637573,
"logps": -85.0908203125,
"loss": 0.3743,
"objective": 0.3781766891479492,
"ranking_simple": 0.8166666626930237,
"step": 1050
},
{
"epoch": 2.975909305621162,
"eval_dpo_loss": 0.6748924255371094,
"eval_logits": -1.2948576211929321,
"eval_logps": -92.3352279663086,
"eval_loss": 0.7416455745697021,
"eval_objective": 0.7355117201805115,
"eval_ranking_simple": 0.5502070188522339,
"eval_runtime": 370.411,
"eval_samples_per_second": 15.631,
"eval_steps_per_second": 1.304,
"step": 1050
},
{
"dpo_loss": 0.25150495767593384,
"epoch": 2.9900803023145963,
"grad_norm": 5.061308546886394,
"learning_rate": 1.3669799732163314e-11,
"logits": -1.2227600812911987,
"logps": -85.82769012451172,
"loss": 0.3854,
"objective": 0.38525307178497314,
"ranking_simple": 0.7833333611488342,
"step": 1055
},
{
"epoch": 2.992914501653283,
"step": 1056,
"total_flos": 0.0,
"train_loss": 0.4763158837220434,
"train_runtime": 34658.3644,
"train_samples_per_second": 4.397,
"train_steps_per_second": 0.03
}
],
"logging_steps": 5,
"max_steps": 1056,
"num_input_tokens_seen": 0,
"num_train_epochs": 3,
"save_steps": 500,
"stateful_callbacks": {
"TrainerControl": {
"args": {
"should_epoch_stop": false,
"should_evaluate": false,
"should_log": false,
"should_save": false,
"should_training_stop": false
},
"attributes": {}
}
},
"total_flos": 0.0,
"train_batch_size": 4,
"trial_name": null,
"trial_params": null
}