3b8f28089f
Model: jackf857/qwen3-8b-base-epsilon-dpo-hh-helpful-4xh200-batch-64-20260424-040306 Source: Original Platform
14465 lines
511 KiB
JSON
14465 lines
511 KiB
JSON
{
|
|
"best_global_step": null,
|
|
"best_metric": null,
|
|
"best_model_checkpoint": null,
|
|
"epoch": 1.0,
|
|
"eval_steps": 100,
|
|
"global_step": 681,
|
|
"is_hyper_param_search": false,
|
|
"is_local_process_zero": true,
|
|
"is_world_process_zero": true,
|
|
"log_history": [
|
|
{
|
|
"epoch": 0.0014684287812041115,
|
|
"grad_norm": 38.745460510253906,
|
|
"kl/avg_steps": 0.09375,
|
|
"kl/beta": 0.10000000149011612,
|
|
"kl/n_epsilon_steps": 0.453125,
|
|
"kl/p_epsilon_steps": 0.546875,
|
|
"learning_rate": 0.0,
|
|
"logits/chosen": -1.585817575454712,
|
|
"logits/rejected": -0.5333532691001892,
|
|
"logps/chosen": -85.37664031982422,
|
|
"logps/ref_chosen": -85.43083190917969,
|
|
"logps/ref_rejected": -79.90458679199219,
|
|
"logps/rejected": -79.91163635253906,
|
|
"loss": 1.381,
|
|
"rewards/accuracies": 0.5625,
|
|
"rewards/chosen": 0.005238114856183529,
|
|
"rewards/margins": 0.005787555128335953,
|
|
"rewards/rejected": -0.0005494409706443548,
|
|
"step": 1
|
|
},
|
|
{
|
|
"epoch": 0.002936857562408223,
|
|
"grad_norm": 29.78634262084961,
|
|
"kl/avg_steps": -0.3125,
|
|
"kl/beta": 0.09990634024143219,
|
|
"kl/n_epsilon_steps": 0.65625,
|
|
"kl/p_epsilon_steps": 0.34375,
|
|
"learning_rate": 7.246376811594203e-09,
|
|
"logits/chosen": -0.7526164054870605,
|
|
"logits/rejected": -0.3610996603965759,
|
|
"logps/chosen": -82.11383056640625,
|
|
"logps/ref_chosen": -82.06892395019531,
|
|
"logps/ref_rejected": -81.65457153320312,
|
|
"logps/rejected": -81.57505798339844,
|
|
"loss": 1.3995,
|
|
"rewards/accuracies": 0.375,
|
|
"rewards/chosen": -0.004650775343179703,
|
|
"rewards/margins": -0.012771460227668285,
|
|
"rewards/rejected": 0.008120683953166008,
|
|
"step": 2
|
|
},
|
|
{
|
|
"epoch": 0.004405286343612335,
|
|
"grad_norm": 26.100635528564453,
|
|
"kl/avg_steps": 0.09375,
|
|
"kl/beta": 0.10021952539682388,
|
|
"kl/n_epsilon_steps": 0.453125,
|
|
"kl/p_epsilon_steps": 0.546875,
|
|
"learning_rate": 1.4492753623188406e-08,
|
|
"logits/chosen": -1.002709150314331,
|
|
"logits/rejected": -0.5633082985877991,
|
|
"logps/chosen": -93.7535629272461,
|
|
"logps/ref_chosen": -93.81098937988281,
|
|
"logps/ref_rejected": -74.22950744628906,
|
|
"logps/rejected": -74.23006439208984,
|
|
"loss": 1.3813,
|
|
"rewards/accuracies": 0.578125,
|
|
"rewards/chosen": 0.005534623749554157,
|
|
"rewards/margins": 0.00546031491830945,
|
|
"rewards/rejected": 7.430883124470711e-05,
|
|
"step": 3
|
|
},
|
|
{
|
|
"epoch": 0.005873715124816446,
|
|
"grad_norm": 32.92469024658203,
|
|
"kl/avg_steps": 0.09375,
|
|
"kl/beta": 0.10012565553188324,
|
|
"kl/n_epsilon_steps": 0.453125,
|
|
"kl/p_epsilon_steps": 0.546875,
|
|
"learning_rate": 2.1739130434782606e-08,
|
|
"logits/chosen": -0.8497915267944336,
|
|
"logits/rejected": -0.17156964540481567,
|
|
"logps/chosen": -87.32073211669922,
|
|
"logps/ref_chosen": -87.29246520996094,
|
|
"logps/ref_rejected": -93.82425689697266,
|
|
"logps/rejected": -93.79373168945312,
|
|
"loss": 1.3929,
|
|
"rewards/accuracies": 0.5625,
|
|
"rewards/chosen": -0.003004699246957898,
|
|
"rewards/margins": -0.006174374371767044,
|
|
"rewards/rejected": 0.0031696748919785023,
|
|
"step": 4
|
|
},
|
|
{
|
|
"epoch": 0.007342143906020558,
|
|
"grad_norm": 35.190330505371094,
|
|
"kl/avg_steps": 0.03125,
|
|
"kl/beta": 0.10003187507390976,
|
|
"kl/n_epsilon_steps": 0.484375,
|
|
"kl/p_epsilon_steps": 0.515625,
|
|
"learning_rate": 2.898550724637681e-08,
|
|
"logits/chosen": -1.187368392944336,
|
|
"logits/rejected": -0.5294585227966309,
|
|
"logps/chosen": -89.35664367675781,
|
|
"logps/ref_chosen": -89.33675384521484,
|
|
"logps/ref_rejected": -88.74783325195312,
|
|
"logps/rejected": -88.785400390625,
|
|
"loss": 1.3853,
|
|
"rewards/accuracies": 0.5625,
|
|
"rewards/chosen": -0.0021656095050275326,
|
|
"rewards/margins": 0.0014310609549283981,
|
|
"rewards/rejected": -0.0035966699942946434,
|
|
"step": 5
|
|
},
|
|
{
|
|
"epoch": 0.00881057268722467,
|
|
"grad_norm": 36.109169006347656,
|
|
"kl/avg_steps": 0.109375,
|
|
"kl/beta": 0.10000062733888626,
|
|
"kl/n_epsilon_steps": 0.4375,
|
|
"kl/p_epsilon_steps": 0.546875,
|
|
"learning_rate": 3.6231884057971014e-08,
|
|
"logits/chosen": -1.2944458723068237,
|
|
"logits/rejected": -0.41752371191978455,
|
|
"logps/chosen": -97.32476043701172,
|
|
"logps/ref_chosen": -97.32147216796875,
|
|
"logps/ref_rejected": -97.88345336914062,
|
|
"logps/rejected": -97.89209747314453,
|
|
"loss": 1.3866,
|
|
"rewards/accuracies": 0.59375,
|
|
"rewards/chosen": -0.0005067111924290657,
|
|
"rewards/margins": 0.00019457843154668808,
|
|
"rewards/rejected": -0.0007012896239757538,
|
|
"step": 6
|
|
},
|
|
{
|
|
"epoch": 0.010279001468428781,
|
|
"grad_norm": 37.86967086791992,
|
|
"kl/avg_steps": 0.09375,
|
|
"kl/beta": 0.09989137202501297,
|
|
"kl/n_epsilon_steps": 0.453125,
|
|
"kl/p_epsilon_steps": 0.546875,
|
|
"learning_rate": 4.347826086956521e-08,
|
|
"logits/chosen": -0.6984870433807373,
|
|
"logits/rejected": -0.42031070590019226,
|
|
"logps/chosen": -86.60205078125,
|
|
"logps/ref_chosen": -86.64852905273438,
|
|
"logps/ref_rejected": -109.61618041992188,
|
|
"logps/rejected": -109.63433837890625,
|
|
"loss": 1.3805,
|
|
"rewards/accuracies": 0.53125,
|
|
"rewards/chosen": 0.004504315089434385,
|
|
"rewards/margins": 0.006173794623464346,
|
|
"rewards/rejected": -0.0016694795340299606,
|
|
"step": 7
|
|
},
|
|
{
|
|
"epoch": 0.011747430249632892,
|
|
"grad_norm": 32.797054290771484,
|
|
"kl/avg_steps": -0.125,
|
|
"kl/beta": 0.0997978076338768,
|
|
"kl/n_epsilon_steps": 0.5625,
|
|
"kl/p_epsilon_steps": 0.4375,
|
|
"learning_rate": 5.0724637681159424e-08,
|
|
"logits/chosen": -1.414137363433838,
|
|
"logits/rejected": -0.4588002562522888,
|
|
"logps/chosen": -89.94332885742188,
|
|
"logps/ref_chosen": -89.9236831665039,
|
|
"logps/ref_rejected": -86.22803497314453,
|
|
"logps/rejected": -86.1485366821289,
|
|
"loss": 1.3968,
|
|
"rewards/accuracies": 0.421875,
|
|
"rewards/chosen": -0.002112824469804764,
|
|
"rewards/margins": -0.010171930305659771,
|
|
"rewards/rejected": 0.008059106767177582,
|
|
"step": 8
|
|
},
|
|
{
|
|
"epoch": 0.013215859030837005,
|
|
"grad_norm": 35.72300338745117,
|
|
"kl/avg_steps": 0.15625,
|
|
"kl/beta": 0.09992270916700363,
|
|
"kl/n_epsilon_steps": 0.421875,
|
|
"kl/p_epsilon_steps": 0.578125,
|
|
"learning_rate": 5.797101449275362e-08,
|
|
"logits/chosen": -0.9759007692337036,
|
|
"logits/rejected": -0.3913915753364563,
|
|
"logps/chosen": -103.79417419433594,
|
|
"logps/ref_chosen": -103.85713195800781,
|
|
"logps/ref_rejected": -104.31932067871094,
|
|
"logps/rejected": -104.34971618652344,
|
|
"loss": 1.378,
|
|
"rewards/accuracies": 0.59375,
|
|
"rewards/chosen": 0.006141543388366699,
|
|
"rewards/margins": 0.008944813162088394,
|
|
"rewards/rejected": -0.002803270472213626,
|
|
"step": 9
|
|
},
|
|
{
|
|
"epoch": 0.014684287812041116,
|
|
"grad_norm": 33.81359100341797,
|
|
"kl/avg_steps": -0.25,
|
|
"kl/beta": 0.09976682811975479,
|
|
"kl/n_epsilon_steps": 0.625,
|
|
"kl/p_epsilon_steps": 0.375,
|
|
"learning_rate": 6.521739130434782e-08,
|
|
"logits/chosen": -1.2568838596343994,
|
|
"logits/rejected": -0.34031057357788086,
|
|
"logps/chosen": -76.20588684082031,
|
|
"logps/ref_chosen": -76.20436096191406,
|
|
"logps/ref_rejected": -87.15210723876953,
|
|
"logps/rejected": -87.01283264160156,
|
|
"loss": 1.401,
|
|
"rewards/accuracies": 0.359375,
|
|
"rewards/chosen": -0.00028783950256183743,
|
|
"rewards/margins": -0.014337141066789627,
|
|
"rewards/rejected": 0.014049299992620945,
|
|
"step": 10
|
|
},
|
|
{
|
|
"epoch": 0.016152716593245228,
|
|
"grad_norm": 37.09693908691406,
|
|
"kl/avg_steps": -0.015625,
|
|
"kl/beta": 0.10001686960458755,
|
|
"kl/n_epsilon_steps": 0.5,
|
|
"kl/p_epsilon_steps": 0.484375,
|
|
"learning_rate": 7.246376811594203e-08,
|
|
"logits/chosen": -1.4052127599716187,
|
|
"logits/rejected": -0.392024964094162,
|
|
"logps/chosen": -82.30293273925781,
|
|
"logps/ref_chosen": -82.36649322509766,
|
|
"logps/ref_rejected": -94.26461791992188,
|
|
"logps/rejected": -94.2509536743164,
|
|
"loss": 1.3819,
|
|
"rewards/accuracies": 0.515625,
|
|
"rewards/chosen": 0.006203308701515198,
|
|
"rewards/margins": 0.004709784872829914,
|
|
"rewards/rejected": 0.0014935237122699618,
|
|
"step": 11
|
|
},
|
|
{
|
|
"epoch": 0.01762114537444934,
|
|
"grad_norm": 39.88624572753906,
|
|
"kl/avg_steps": 0.125,
|
|
"kl/beta": 0.1000325009226799,
|
|
"kl/n_epsilon_steps": 0.4375,
|
|
"kl/p_epsilon_steps": 0.5625,
|
|
"learning_rate": 7.971014492753623e-08,
|
|
"logits/chosen": -0.7849316596984863,
|
|
"logits/rejected": -0.267448753118515,
|
|
"logps/chosen": -99.06604766845703,
|
|
"logps/ref_chosen": -99.10549926757812,
|
|
"logps/ref_rejected": -110.27140808105469,
|
|
"logps/rejected": -110.31909942626953,
|
|
"loss": 1.3785,
|
|
"rewards/accuracies": 0.578125,
|
|
"rewards/chosen": 0.003762049600481987,
|
|
"rewards/margins": 0.008365876972675323,
|
|
"rewards/rejected": -0.0046038273721933365,
|
|
"step": 12
|
|
},
|
|
{
|
|
"epoch": 0.01908957415565345,
|
|
"grad_norm": 41.430633544921875,
|
|
"kl/avg_steps": 0.15625,
|
|
"kl/beta": 0.09990761429071426,
|
|
"kl/n_epsilon_steps": 0.421875,
|
|
"kl/p_epsilon_steps": 0.578125,
|
|
"learning_rate": 8.695652173913042e-08,
|
|
"logits/chosen": -1.5717546939849854,
|
|
"logits/rejected": -0.7266464829444885,
|
|
"logps/chosen": -90.52992248535156,
|
|
"logps/ref_chosen": -90.55973052978516,
|
|
"logps/ref_rejected": -93.69110107421875,
|
|
"logps/rejected": -93.72262573242188,
|
|
"loss": 1.3812,
|
|
"rewards/accuracies": 0.578125,
|
|
"rewards/chosen": 0.002787390723824501,
|
|
"rewards/margins": 0.00573158822953701,
|
|
"rewards/rejected": -0.0029441972728818655,
|
|
"step": 13
|
|
},
|
|
{
|
|
"epoch": 0.020558002936857563,
|
|
"grad_norm": 35.816402435302734,
|
|
"kl/avg_steps": 0.03125,
|
|
"kl/beta": 0.09975175559520721,
|
|
"kl/n_epsilon_steps": 0.484375,
|
|
"kl/p_epsilon_steps": 0.515625,
|
|
"learning_rate": 9.420289855072464e-08,
|
|
"logits/chosen": -0.5540125370025635,
|
|
"logits/rejected": -0.28503644466400146,
|
|
"logps/chosen": -99.85889434814453,
|
|
"logps/ref_chosen": -99.82717895507812,
|
|
"logps/ref_rejected": -108.94200134277344,
|
|
"logps/rejected": -108.9466552734375,
|
|
"loss": 1.3898,
|
|
"rewards/accuracies": 0.515625,
|
|
"rewards/chosen": -0.003366068471223116,
|
|
"rewards/margins": -0.0030267564579844475,
|
|
"rewards/rejected": -0.00033931387588381767,
|
|
"step": 14
|
|
},
|
|
{
|
|
"epoch": 0.022026431718061675,
|
|
"grad_norm": 31.496644973754883,
|
|
"kl/avg_steps": -0.1875,
|
|
"kl/beta": 0.09972058981657028,
|
|
"kl/n_epsilon_steps": 0.59375,
|
|
"kl/p_epsilon_steps": 0.40625,
|
|
"learning_rate": 1.0144927536231885e-07,
|
|
"logits/chosen": -0.8750624060630798,
|
|
"logits/rejected": -0.2879735827445984,
|
|
"logps/chosen": -78.86597442626953,
|
|
"logps/ref_chosen": -78.90997314453125,
|
|
"logps/ref_rejected": -90.06234741210938,
|
|
"logps/rejected": -90.09466552734375,
|
|
"loss": 1.3795,
|
|
"rewards/accuracies": 0.4375,
|
|
"rewards/chosen": 0.004188378341495991,
|
|
"rewards/margins": 0.007276091258972883,
|
|
"rewards/rejected": -0.0030877136159688234,
|
|
"step": 15
|
|
},
|
|
{
|
|
"epoch": 0.023494860499265784,
|
|
"grad_norm": 34.249393463134766,
|
|
"kl/avg_steps": 0.15625,
|
|
"kl/beta": 0.09990791976451874,
|
|
"kl/n_epsilon_steps": 0.421875,
|
|
"kl/p_epsilon_steps": 0.578125,
|
|
"learning_rate": 1.0869565217391303e-07,
|
|
"logits/chosen": -0.9704724550247192,
|
|
"logits/rejected": -0.0901266559958458,
|
|
"logps/chosen": -97.3776626586914,
|
|
"logps/ref_chosen": -97.42327880859375,
|
|
"logps/ref_rejected": -90.59945678710938,
|
|
"logps/rejected": -90.60440826416016,
|
|
"loss": 1.3824,
|
|
"rewards/accuracies": 0.578125,
|
|
"rewards/chosen": 0.0043681650422513485,
|
|
"rewards/margins": 0.004663495346903801,
|
|
"rewards/rejected": -0.0002953286748379469,
|
|
"step": 16
|
|
},
|
|
{
|
|
"epoch": 0.024963289280469897,
|
|
"grad_norm": 35.82853698730469,
|
|
"kl/avg_steps": 0.03125,
|
|
"kl/beta": 0.09975205361843109,
|
|
"kl/n_epsilon_steps": 0.484375,
|
|
"kl/p_epsilon_steps": 0.515625,
|
|
"learning_rate": 1.1594202898550725e-07,
|
|
"logits/chosen": -1.0136702060699463,
|
|
"logits/rejected": -0.5759009122848511,
|
|
"logps/chosen": -104.36908721923828,
|
|
"logps/ref_chosen": -104.36431121826172,
|
|
"logps/ref_rejected": -90.46772766113281,
|
|
"logps/rejected": -90.47051239013672,
|
|
"loss": 1.3875,
|
|
"rewards/accuracies": 0.515625,
|
|
"rewards/chosen": -0.0006631199503317475,
|
|
"rewards/margins": -0.0005828813882544637,
|
|
"rewards/rejected": -8.023856207728386e-05,
|
|
"step": 17
|
|
},
|
|
{
|
|
"epoch": 0.02643171806167401,
|
|
"grad_norm": 41.663047790527344,
|
|
"kl/avg_steps": 0.0625,
|
|
"kl/beta": 0.09972089529037476,
|
|
"kl/n_epsilon_steps": 0.46875,
|
|
"kl/p_epsilon_steps": 0.53125,
|
|
"learning_rate": 1.2318840579710146e-07,
|
|
"logits/chosen": -1.795450210571289,
|
|
"logits/rejected": -0.8455245494842529,
|
|
"logps/chosen": -87.06967163085938,
|
|
"logps/ref_chosen": -87.09195709228516,
|
|
"logps/ref_rejected": -81.85072326660156,
|
|
"logps/rejected": -81.87223052978516,
|
|
"loss": 1.383,
|
|
"rewards/accuracies": 0.5625,
|
|
"rewards/chosen": 0.0020006708800792694,
|
|
"rewards/margins": 0.003985004499554634,
|
|
"rewards/rejected": -0.001984333386644721,
|
|
"step": 18
|
|
},
|
|
{
|
|
"epoch": 0.027900146842878122,
|
|
"grad_norm": 31.180570602416992,
|
|
"kl/avg_steps": 0.0625,
|
|
"kl/beta": 0.09965860843658447,
|
|
"kl/n_epsilon_steps": 0.46875,
|
|
"kl/p_epsilon_steps": 0.53125,
|
|
"learning_rate": 1.3043478260869563e-07,
|
|
"logits/chosen": -1.0867879390716553,
|
|
"logits/rejected": -0.03352098539471626,
|
|
"logps/chosen": -105.85330963134766,
|
|
"logps/ref_chosen": -105.87354278564453,
|
|
"logps/ref_rejected": -96.93023681640625,
|
|
"logps/rejected": -96.95452880859375,
|
|
"loss": 1.3826,
|
|
"rewards/accuracies": 0.546875,
|
|
"rewards/chosen": 0.0018369832541793585,
|
|
"rewards/margins": 0.004121738485991955,
|
|
"rewards/rejected": -0.0022847556974738836,
|
|
"step": 19
|
|
},
|
|
{
|
|
"epoch": 0.02936857562408223,
|
|
"grad_norm": 32.30035400390625,
|
|
"kl/avg_steps": 0.09375,
|
|
"kl/beta": 0.09959635883569717,
|
|
"kl/n_epsilon_steps": 0.453125,
|
|
"kl/p_epsilon_steps": 0.546875,
|
|
"learning_rate": 1.3768115942028986e-07,
|
|
"logits/chosen": -1.2128328084945679,
|
|
"logits/rejected": -0.32826870679855347,
|
|
"logps/chosen": -90.72392272949219,
|
|
"logps/ref_chosen": -90.75811767578125,
|
|
"logps/ref_rejected": -85.91232299804688,
|
|
"logps/rejected": -85.96060180664062,
|
|
"loss": 1.3791,
|
|
"rewards/accuracies": 0.546875,
|
|
"rewards/chosen": 0.003147183684632182,
|
|
"rewards/margins": 0.007826998829841614,
|
|
"rewards/rejected": -0.004679815378040075,
|
|
"step": 20
|
|
},
|
|
{
|
|
"epoch": 0.030837004405286344,
|
|
"grad_norm": 31.352293014526367,
|
|
"kl/avg_steps": 0.125,
|
|
"kl/beta": 0.09950307011604309,
|
|
"kl/n_epsilon_steps": 0.4375,
|
|
"kl/p_epsilon_steps": 0.5625,
|
|
"learning_rate": 1.4492753623188405e-07,
|
|
"logits/chosen": -0.9726300239562988,
|
|
"logits/rejected": -0.411948561668396,
|
|
"logps/chosen": -80.2506332397461,
|
|
"logps/ref_chosen": -80.33346557617188,
|
|
"logps/ref_rejected": -83.9337387084961,
|
|
"logps/rejected": -83.91175842285156,
|
|
"loss": 1.3808,
|
|
"rewards/accuracies": 0.59375,
|
|
"rewards/chosen": 0.008116335608065128,
|
|
"rewards/margins": 0.005770478397607803,
|
|
"rewards/rejected": 0.002345857210457325,
|
|
"step": 21
|
|
},
|
|
{
|
|
"epoch": 0.032305433186490456,
|
|
"grad_norm": 42.61378860473633,
|
|
"kl/avg_steps": 0.09375,
|
|
"kl/beta": 0.09937884658575058,
|
|
"kl/n_epsilon_steps": 0.453125,
|
|
"kl/p_epsilon_steps": 0.546875,
|
|
"learning_rate": 1.5217391304347825e-07,
|
|
"logits/chosen": -0.9675413370132446,
|
|
"logits/rejected": -0.6590346693992615,
|
|
"logps/chosen": -95.35507202148438,
|
|
"logps/ref_chosen": -95.39530181884766,
|
|
"logps/ref_rejected": -103.47351837158203,
|
|
"logps/rejected": -103.48601531982422,
|
|
"loss": 1.3819,
|
|
"rewards/accuracies": 0.546875,
|
|
"rewards/chosen": 0.0037977853789925575,
|
|
"rewards/margins": 0.004873580764979124,
|
|
"rewards/rejected": -0.0010757955024018884,
|
|
"step": 22
|
|
},
|
|
{
|
|
"epoch": 0.033773861967694566,
|
|
"grad_norm": 31.29548454284668,
|
|
"kl/avg_steps": -0.09375,
|
|
"kl/beta": 0.09928576648235321,
|
|
"kl/n_epsilon_steps": 0.546875,
|
|
"kl/p_epsilon_steps": 0.453125,
|
|
"learning_rate": 1.5942028985507245e-07,
|
|
"logits/chosen": -1.156263828277588,
|
|
"logits/rejected": -0.3607323169708252,
|
|
"logps/chosen": -90.63298034667969,
|
|
"logps/ref_chosen": -90.63751220703125,
|
|
"logps/ref_rejected": -86.59425354003906,
|
|
"logps/rejected": -86.54367065429688,
|
|
"loss": 1.3916,
|
|
"rewards/accuracies": 0.4375,
|
|
"rewards/chosen": 0.00023264711489900947,
|
|
"rewards/margins": -0.004911544732749462,
|
|
"rewards/rejected": 0.005144191440194845,
|
|
"step": 23
|
|
},
|
|
{
|
|
"epoch": 0.03524229074889868,
|
|
"grad_norm": 44.2933464050293,
|
|
"kl/avg_steps": -0.03125,
|
|
"kl/beta": 0.09937893599271774,
|
|
"kl/n_epsilon_steps": 0.515625,
|
|
"kl/p_epsilon_steps": 0.484375,
|
|
"learning_rate": 1.6666666666666665e-07,
|
|
"logits/chosen": -0.8095067143440247,
|
|
"logits/rejected": -0.5244461297988892,
|
|
"logps/chosen": -69.98039245605469,
|
|
"logps/ref_chosen": -69.91728973388672,
|
|
"logps/ref_rejected": -106.60990142822266,
|
|
"logps/rejected": -106.59368133544922,
|
|
"loss": 1.395,
|
|
"rewards/accuracies": 0.453125,
|
|
"rewards/chosen": -0.006438862532377243,
|
|
"rewards/margins": -0.008207438513636589,
|
|
"rewards/rejected": 0.0017685755155980587,
|
|
"step": 24
|
|
},
|
|
{
|
|
"epoch": 0.03671071953010279,
|
|
"grad_norm": 36.52031326293945,
|
|
"kl/avg_steps": 0.25,
|
|
"kl/beta": 0.09941000491380692,
|
|
"kl/n_epsilon_steps": 0.375,
|
|
"kl/p_epsilon_steps": 0.625,
|
|
"learning_rate": 1.7391304347826085e-07,
|
|
"logits/chosen": -1.2055686712265015,
|
|
"logits/rejected": -0.4221525490283966,
|
|
"logps/chosen": -80.82945251464844,
|
|
"logps/ref_chosen": -80.82548522949219,
|
|
"logps/ref_rejected": -95.95710754394531,
|
|
"logps/rejected": -96.06101989746094,
|
|
"loss": 1.3773,
|
|
"rewards/accuracies": 0.65625,
|
|
"rewards/chosen": -0.0005520773120224476,
|
|
"rewards/margins": 0.009578470140695572,
|
|
"rewards/rejected": -0.010130547918379307,
|
|
"step": 25
|
|
},
|
|
{
|
|
"epoch": 0.0381791483113069,
|
|
"grad_norm": 40.3887939453125,
|
|
"kl/avg_steps": 0.15625,
|
|
"kl/beta": 0.09916209429502487,
|
|
"kl/n_epsilon_steps": 0.421875,
|
|
"kl/p_epsilon_steps": 0.578125,
|
|
"learning_rate": 1.8115942028985507e-07,
|
|
"logits/chosen": -1.4493508338928223,
|
|
"logits/rejected": -0.5392433404922485,
|
|
"logps/chosen": -88.89604949951172,
|
|
"logps/ref_chosen": -88.90116882324219,
|
|
"logps/ref_rejected": -109.82818603515625,
|
|
"logps/rejected": -109.86863708496094,
|
|
"loss": 1.3825,
|
|
"rewards/accuracies": 0.53125,
|
|
"rewards/chosen": 0.00035169871989637613,
|
|
"rewards/margins": 0.004207686521112919,
|
|
"rewards/rejected": -0.0038559872191399336,
|
|
"step": 26
|
|
},
|
|
{
|
|
"epoch": 0.039647577092511016,
|
|
"grad_norm": 43.48580551147461,
|
|
"kl/avg_steps": 0.28125,
|
|
"kl/beta": 0.09900739789009094,
|
|
"kl/n_epsilon_steps": 0.359375,
|
|
"kl/p_epsilon_steps": 0.640625,
|
|
"learning_rate": 1.8840579710144927e-07,
|
|
"logits/chosen": -1.6640090942382812,
|
|
"logits/rejected": -0.6294593811035156,
|
|
"logps/chosen": -77.5752182006836,
|
|
"logps/ref_chosen": -77.59600830078125,
|
|
"logps/ref_rejected": -103.93850708007812,
|
|
"logps/rejected": -104.07573699951172,
|
|
"loss": 1.3714,
|
|
"rewards/accuracies": 0.65625,
|
|
"rewards/chosen": 0.0019225336145609617,
|
|
"rewards/margins": 0.01532358955591917,
|
|
"rewards/rejected": -0.013401055708527565,
|
|
"step": 27
|
|
},
|
|
{
|
|
"epoch": 0.041116005873715125,
|
|
"grad_norm": 35.73523712158203,
|
|
"kl/avg_steps": 0.4375,
|
|
"kl/beta": 0.09872972220182419,
|
|
"kl/n_epsilon_steps": 0.28125,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 1.9565217391304347e-07,
|
|
"logits/chosen": -0.956201434135437,
|
|
"logits/rejected": -0.39606085419654846,
|
|
"logps/chosen": -102.17597961425781,
|
|
"logps/ref_chosen": -102.22856140136719,
|
|
"logps/ref_rejected": -96.9594955444336,
|
|
"logps/rejected": -97.06103515625,
|
|
"loss": 1.3723,
|
|
"rewards/accuracies": 0.71875,
|
|
"rewards/chosen": 0.004990905057638884,
|
|
"rewards/margins": 0.014799000695347786,
|
|
"rewards/rejected": -0.00980809610337019,
|
|
"step": 28
|
|
},
|
|
{
|
|
"epoch": 0.042584434654919234,
|
|
"grad_norm": 41.57979965209961,
|
|
"kl/avg_steps": 0.15625,
|
|
"kl/beta": 0.09829965978860855,
|
|
"kl/n_epsilon_steps": 0.421875,
|
|
"kl/p_epsilon_steps": 0.578125,
|
|
"learning_rate": 2.028985507246377e-07,
|
|
"logits/chosen": -1.4259010553359985,
|
|
"logits/rejected": -0.6437522172927856,
|
|
"logps/chosen": -88.62876892089844,
|
|
"logps/ref_chosen": -88.64704895019531,
|
|
"logps/ref_rejected": -102.96011352539062,
|
|
"logps/rejected": -103.11316680908203,
|
|
"loss": 1.3704,
|
|
"rewards/accuracies": 0.609375,
|
|
"rewards/chosen": 0.0016324977623298764,
|
|
"rewards/margins": 0.01648685149848461,
|
|
"rewards/rejected": -0.014854353852570057,
|
|
"step": 29
|
|
},
|
|
{
|
|
"epoch": 0.04405286343612335,
|
|
"grad_norm": 38.55412292480469,
|
|
"kl/avg_steps": 0.0625,
|
|
"kl/beta": 0.09814630448818207,
|
|
"kl/n_epsilon_steps": 0.46875,
|
|
"kl/p_epsilon_steps": 0.53125,
|
|
"learning_rate": 2.1014492753623187e-07,
|
|
"logits/chosen": -0.9132494926452637,
|
|
"logits/rejected": -0.49612629413604736,
|
|
"logps/chosen": -88.3867416381836,
|
|
"logps/ref_chosen": -88.38838958740234,
|
|
"logps/ref_rejected": -102.31889343261719,
|
|
"logps/rejected": -102.37272644042969,
|
|
"loss": 1.3817,
|
|
"rewards/accuracies": 0.578125,
|
|
"rewards/chosen": 6.297486834228039e-06,
|
|
"rewards/margins": 0.0051061781123280525,
|
|
"rewards/rejected": -0.005099880509078503,
|
|
"step": 30
|
|
},
|
|
{
|
|
"epoch": 0.04552129221732746,
|
|
"grad_norm": 30.63753890991211,
|
|
"kl/avg_steps": 0.09375,
|
|
"kl/beta": 0.09808500111103058,
|
|
"kl/n_epsilon_steps": 0.453125,
|
|
"kl/p_epsilon_steps": 0.546875,
|
|
"learning_rate": 2.1739130434782607e-07,
|
|
"logits/chosen": -0.8628894090652466,
|
|
"logits/rejected": -0.3569292426109314,
|
|
"logps/chosen": -101.13359832763672,
|
|
"logps/ref_chosen": -101.12565612792969,
|
|
"logps/ref_rejected": -79.85842895507812,
|
|
"logps/rejected": -79.95623779296875,
|
|
"loss": 1.3782,
|
|
"rewards/accuracies": 0.578125,
|
|
"rewards/chosen": -0.0009215597528964281,
|
|
"rewards/margins": 0.008493051864206791,
|
|
"rewards/rejected": -0.009414611384272575,
|
|
"step": 31
|
|
},
|
|
{
|
|
"epoch": 0.04698972099853157,
|
|
"grad_norm": 34.43489456176758,
|
|
"kl/avg_steps": 0.0,
|
|
"kl/beta": 0.09799313545227051,
|
|
"kl/n_epsilon_steps": 0.5,
|
|
"kl/p_epsilon_steps": 0.5,
|
|
"learning_rate": 2.2463768115942027e-07,
|
|
"logits/chosen": -0.9892777800559998,
|
|
"logits/rejected": -0.34080418944358826,
|
|
"logps/chosen": -96.68499755859375,
|
|
"logps/ref_chosen": -96.61703491210938,
|
|
"logps/ref_rejected": -96.224365234375,
|
|
"logps/rejected": -96.38025665283203,
|
|
"loss": 1.3789,
|
|
"rewards/accuracies": 0.5625,
|
|
"rewards/chosen": -0.006817132234573364,
|
|
"rewards/margins": 0.008191294968128204,
|
|
"rewards/rejected": -0.015008427202701569,
|
|
"step": 32
|
|
},
|
|
{
|
|
"epoch": 0.048458149779735685,
|
|
"grad_norm": 32.965362548828125,
|
|
"kl/avg_steps": 0.140625,
|
|
"kl/beta": 0.09799313545227051,
|
|
"kl/n_epsilon_steps": 0.421875,
|
|
"kl/p_epsilon_steps": 0.5625,
|
|
"learning_rate": 2.318840579710145e-07,
|
|
"logits/chosen": -1.2965284585952759,
|
|
"logits/rejected": -0.3348070979118347,
|
|
"logps/chosen": -81.51568603515625,
|
|
"logps/ref_chosen": -81.5210189819336,
|
|
"logps/ref_rejected": -93.80595397949219,
|
|
"logps/rejected": -93.91485595703125,
|
|
"loss": 1.376,
|
|
"rewards/accuracies": 0.578125,
|
|
"rewards/chosen": 0.0003811400383710861,
|
|
"rewards/margins": 0.010841380804777145,
|
|
"rewards/rejected": -0.01046024076640606,
|
|
"step": 33
|
|
},
|
|
{
|
|
"epoch": 0.049926578560939794,
|
|
"grad_norm": 39.99357604980469,
|
|
"kl/avg_steps": 0.234375,
|
|
"kl/beta": 0.09785552322864532,
|
|
"kl/n_epsilon_steps": 0.375,
|
|
"kl/p_epsilon_steps": 0.609375,
|
|
"learning_rate": 2.391304347826087e-07,
|
|
"logits/chosen": -1.2087818384170532,
|
|
"logits/rejected": -0.19416889548301697,
|
|
"logps/chosen": -77.15280151367188,
|
|
"logps/ref_chosen": -77.20204162597656,
|
|
"logps/ref_rejected": -106.71875762939453,
|
|
"logps/rejected": -106.84297943115234,
|
|
"loss": 1.3705,
|
|
"rewards/accuracies": 0.65625,
|
|
"rewards/chosen": 0.004645414184778929,
|
|
"rewards/margins": 0.016573915258049965,
|
|
"rewards/rejected": -0.011928501538932323,
|
|
"step": 34
|
|
},
|
|
{
|
|
"epoch": 0.0513950073421439,
|
|
"grad_norm": 41.544822692871094,
|
|
"kl/avg_steps": 0.34375,
|
|
"kl/beta": 0.0976267158985138,
|
|
"kl/n_epsilon_steps": 0.328125,
|
|
"kl/p_epsilon_steps": 0.671875,
|
|
"learning_rate": 2.463768115942029e-07,
|
|
"logits/chosen": -1.575798749923706,
|
|
"logits/rejected": -0.5509282350540161,
|
|
"logps/chosen": -77.5616683959961,
|
|
"logps/ref_chosen": -77.57035827636719,
|
|
"logps/ref_rejected": -112.18855285644531,
|
|
"logps/rejected": -112.379638671875,
|
|
"loss": 1.3678,
|
|
"rewards/accuracies": 0.65625,
|
|
"rewards/chosen": 0.0007331545930355787,
|
|
"rewards/margins": 0.019134394824504852,
|
|
"rewards/rejected": -0.018401240929961205,
|
|
"step": 35
|
|
},
|
|
{
|
|
"epoch": 0.05286343612334802,
|
|
"grad_norm": 31.387723922729492,
|
|
"kl/avg_steps": 0.28125,
|
|
"kl/beta": 0.09729227423667908,
|
|
"kl/n_epsilon_steps": 0.359375,
|
|
"kl/p_epsilon_steps": 0.640625,
|
|
"learning_rate": 2.536231884057971e-07,
|
|
"logits/chosen": -1.432613730430603,
|
|
"logits/rejected": -0.642814040184021,
|
|
"logps/chosen": -83.33061218261719,
|
|
"logps/ref_chosen": -83.28824615478516,
|
|
"logps/ref_rejected": -91.75741577148438,
|
|
"logps/rejected": -92.01959228515625,
|
|
"loss": 1.3662,
|
|
"rewards/accuracies": 0.625,
|
|
"rewards/chosen": -0.0043023210018873215,
|
|
"rewards/margins": 0.02096126601099968,
|
|
"rewards/rejected": -0.025263587012887,
|
|
"step": 36
|
|
},
|
|
{
|
|
"epoch": 0.05433186490455213,
|
|
"grad_norm": 34.76736831665039,
|
|
"kl/avg_steps": -0.03125,
|
|
"kl/beta": 0.09701940417289734,
|
|
"kl/n_epsilon_steps": 0.515625,
|
|
"kl/p_epsilon_steps": 0.484375,
|
|
"learning_rate": 2.6086956521739126e-07,
|
|
"logits/chosen": -0.9882210493087769,
|
|
"logits/rejected": -0.7192566990852356,
|
|
"logps/chosen": -94.82408142089844,
|
|
"logps/ref_chosen": -94.77108764648438,
|
|
"logps/ref_rejected": -85.7172622680664,
|
|
"logps/rejected": -85.97396087646484,
|
|
"loss": 1.368,
|
|
"rewards/accuracies": 0.53125,
|
|
"rewards/chosen": -0.005295893643051386,
|
|
"rewards/margins": 0.01935383304953575,
|
|
"rewards/rejected": -0.0246497243642807,
|
|
"step": 37
|
|
},
|
|
{
|
|
"epoch": 0.055800293685756244,
|
|
"grad_norm": 35.520565032958984,
|
|
"kl/avg_steps": 0.3125,
|
|
"kl/beta": 0.09704973548650742,
|
|
"kl/n_epsilon_steps": 0.34375,
|
|
"kl/p_epsilon_steps": 0.65625,
|
|
"learning_rate": 2.681159420289855e-07,
|
|
"logits/chosen": -0.9708235263824463,
|
|
"logits/rejected": -0.36027270555496216,
|
|
"logps/chosen": -75.92189025878906,
|
|
"logps/ref_chosen": -75.97850799560547,
|
|
"logps/ref_rejected": -104.10401916503906,
|
|
"logps/rejected": -104.33356475830078,
|
|
"loss": 1.3601,
|
|
"rewards/accuracies": 0.65625,
|
|
"rewards/chosen": 0.0053534312173724174,
|
|
"rewards/margins": 0.027343016117811203,
|
|
"rewards/rejected": -0.02198958396911621,
|
|
"step": 38
|
|
},
|
|
{
|
|
"epoch": 0.05726872246696035,
|
|
"grad_norm": 34.98324966430664,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.09674739837646484,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 2.753623188405797e-07,
|
|
"logits/chosen": -1.2482174634933472,
|
|
"logits/rejected": -0.5844467878341675,
|
|
"logps/chosen": -81.19099426269531,
|
|
"logps/ref_chosen": -81.18577575683594,
|
|
"logps/ref_rejected": -84.1959228515625,
|
|
"logps/rejected": -84.48798370361328,
|
|
"loss": 1.3596,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -0.0006124734645709395,
|
|
"rewards/margins": 0.02736220322549343,
|
|
"rewards/rejected": -0.027974674478173256,
|
|
"step": 39
|
|
},
|
|
{
|
|
"epoch": 0.05873715124816446,
|
|
"grad_norm": 27.852684020996094,
|
|
"kl/avg_steps": 0.40625,
|
|
"kl/beta": 0.09629600495100021,
|
|
"kl/n_epsilon_steps": 0.296875,
|
|
"kl/p_epsilon_steps": 0.703125,
|
|
"learning_rate": 2.8260869565217386e-07,
|
|
"logits/chosen": -0.8650610446929932,
|
|
"logits/rejected": -0.20084291696548462,
|
|
"logps/chosen": -83.25595092773438,
|
|
"logps/ref_chosen": -83.33256530761719,
|
|
"logps/ref_rejected": -80.25591278076172,
|
|
"logps/rejected": -80.43058776855469,
|
|
"loss": 1.363,
|
|
"rewards/accuracies": 0.6875,
|
|
"rewards/chosen": 0.007215453311800957,
|
|
"rewards/margins": 0.023813467472791672,
|
|
"rewards/rejected": -0.016598014160990715,
|
|
"step": 40
|
|
},
|
|
{
|
|
"epoch": 0.06020558002936858,
|
|
"grad_norm": 31.773061752319336,
|
|
"kl/avg_steps": 0.34375,
|
|
"kl/beta": 0.09590639173984528,
|
|
"kl/n_epsilon_steps": 0.328125,
|
|
"kl/p_epsilon_steps": 0.671875,
|
|
"learning_rate": 2.898550724637681e-07,
|
|
"logits/chosen": -1.026604413986206,
|
|
"logits/rejected": -0.37627846002578735,
|
|
"logps/chosen": -93.19059753417969,
|
|
"logps/ref_chosen": -93.14866638183594,
|
|
"logps/ref_rejected": -102.07920837402344,
|
|
"logps/rejected": -102.35368347167969,
|
|
"loss": 1.3656,
|
|
"rewards/accuracies": 0.671875,
|
|
"rewards/chosen": -0.004216345027089119,
|
|
"rewards/margins": 0.021797288209199905,
|
|
"rewards/rejected": -0.026013631373643875,
|
|
"step": 41
|
|
},
|
|
{
|
|
"epoch": 0.06167400881057269,
|
|
"grad_norm": 38.71194839477539,
|
|
"kl/avg_steps": 0.40625,
|
|
"kl/beta": 0.09557784348726273,
|
|
"kl/n_epsilon_steps": 0.296875,
|
|
"kl/p_epsilon_steps": 0.703125,
|
|
"learning_rate": 2.971014492753623e-07,
|
|
"logits/chosen": -1.2612457275390625,
|
|
"logits/rejected": -0.5760804414749146,
|
|
"logps/chosen": -90.67225646972656,
|
|
"logps/ref_chosen": -90.70162200927734,
|
|
"logps/ref_rejected": -113.8646469116211,
|
|
"logps/rejected": -114.30716705322266,
|
|
"loss": 1.3432,
|
|
"rewards/accuracies": 0.734375,
|
|
"rewards/chosen": 0.002619321458041668,
|
|
"rewards/margins": 0.044551268219947815,
|
|
"rewards/rejected": -0.041931942105293274,
|
|
"step": 42
|
|
},
|
|
{
|
|
"epoch": 0.0631424375917768,
|
|
"grad_norm": 35.74715805053711,
|
|
"kl/avg_steps": 0.4375,
|
|
"kl/beta": 0.09519112855195999,
|
|
"kl/n_epsilon_steps": 0.28125,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 3.043478260869565e-07,
|
|
"logits/chosen": -1.2501815557479858,
|
|
"logits/rejected": -0.3675554692745209,
|
|
"logps/chosen": -89.61614990234375,
|
|
"logps/ref_chosen": -89.64402770996094,
|
|
"logps/ref_rejected": -103.95185852050781,
|
|
"logps/rejected": -104.35449981689453,
|
|
"loss": 1.3476,
|
|
"rewards/accuracies": 0.703125,
|
|
"rewards/chosen": 0.002490551210939884,
|
|
"rewards/margins": 0.0404253825545311,
|
|
"rewards/rejected": -0.03793483227491379,
|
|
"step": 43
|
|
},
|
|
{
|
|
"epoch": 0.06461086637298091,
|
|
"grad_norm": 32.69614028930664,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.09477648138999939,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 3.115942028985507e-07,
|
|
"logits/chosen": -1.5350993871688843,
|
|
"logits/rejected": -0.39478617906570435,
|
|
"logps/chosen": -81.98731994628906,
|
|
"logps/ref_chosen": -81.8783187866211,
|
|
"logps/ref_rejected": -113.421630859375,
|
|
"logps/rejected": -113.86365509033203,
|
|
"loss": 1.3562,
|
|
"rewards/accuracies": 0.734375,
|
|
"rewards/chosen": -0.010416124947369099,
|
|
"rewards/margins": 0.03108786605298519,
|
|
"rewards/rejected": -0.041503991931676865,
|
|
"step": 44
|
|
},
|
|
{
|
|
"epoch": 0.06607929515418502,
|
|
"grad_norm": 22.90825080871582,
|
|
"kl/avg_steps": 0.40625,
|
|
"kl/beta": 0.09433428943157196,
|
|
"kl/n_epsilon_steps": 0.296875,
|
|
"kl/p_epsilon_steps": 0.703125,
|
|
"learning_rate": 3.188405797101449e-07,
|
|
"logits/chosen": -0.9012535810470581,
|
|
"logits/rejected": -0.24215500056743622,
|
|
"logps/chosen": -77.44337463378906,
|
|
"logps/ref_chosen": -77.34459686279297,
|
|
"logps/ref_rejected": -84.24774169921875,
|
|
"logps/rejected": -84.58972930908203,
|
|
"loss": 1.3645,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -0.009415511973202229,
|
|
"rewards/margins": 0.022572031244635582,
|
|
"rewards/rejected": -0.031987544149160385,
|
|
"step": 45
|
|
},
|
|
{
|
|
"epoch": 0.06754772393538913,
|
|
"grad_norm": 31.163087844848633,
|
|
"kl/avg_steps": 0.546875,
|
|
"kl/beta": 0.09395260363817215,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 3.260869565217391e-07,
|
|
"logits/chosen": -1.383570909500122,
|
|
"logits/rejected": -0.5739269256591797,
|
|
"logps/chosen": -90.39727020263672,
|
|
"logps/ref_chosen": -90.3408203125,
|
|
"logps/ref_rejected": -100.676513671875,
|
|
"logps/rejected": -101.10578918457031,
|
|
"loss": 1.3528,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -0.005449555814266205,
|
|
"rewards/margins": 0.03452453762292862,
|
|
"rewards/rejected": -0.039974093437194824,
|
|
"step": 46
|
|
},
|
|
{
|
|
"epoch": 0.06901615271659324,
|
|
"grad_norm": 31.565263748168945,
|
|
"kl/avg_steps": 0.375,
|
|
"kl/beta": 0.09344159811735153,
|
|
"kl/n_epsilon_steps": 0.3125,
|
|
"kl/p_epsilon_steps": 0.6875,
|
|
"learning_rate": 3.333333333333333e-07,
|
|
"logits/chosen": -0.9797345995903015,
|
|
"logits/rejected": -0.36259299516677856,
|
|
"logps/chosen": -104.41678619384766,
|
|
"logps/ref_chosen": -104.41130065917969,
|
|
"logps/ref_rejected": -101.25489807128906,
|
|
"logps/rejected": -101.67523193359375,
|
|
"loss": 1.3497,
|
|
"rewards/accuracies": 0.671875,
|
|
"rewards/chosen": -0.0006828177720308304,
|
|
"rewards/margins": 0.03818219527602196,
|
|
"rewards/rejected": -0.03886501491069794,
|
|
"step": 47
|
|
},
|
|
{
|
|
"epoch": 0.07048458149779736,
|
|
"grad_norm": 35.98372268676758,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.09309250116348267,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 3.4057971014492755e-07,
|
|
"logits/chosen": -1.2902014255523682,
|
|
"logits/rejected": -0.7823787331581116,
|
|
"logps/chosen": -91.1424560546875,
|
|
"logps/ref_chosen": -91.10027313232422,
|
|
"logps/ref_rejected": -95.08057403564453,
|
|
"logps/rejected": -95.68135070800781,
|
|
"loss": 1.3375,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.004128246568143368,
|
|
"rewards/margins": 0.05126403272151947,
|
|
"rewards/rejected": -0.05539228022098541,
|
|
"step": 48
|
|
},
|
|
{
|
|
"epoch": 0.07195301027900147,
|
|
"grad_norm": 36.75701904296875,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.09260056167840958,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 3.478260869565217e-07,
|
|
"logits/chosen": -2.277641773223877,
|
|
"logits/rejected": -0.8717272281646729,
|
|
"logps/chosen": -92.91861724853516,
|
|
"logps/ref_chosen": -93.00367736816406,
|
|
"logps/ref_rejected": -91.74899291992188,
|
|
"logps/rejected": -92.42121887207031,
|
|
"loss": 1.3199,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": 0.007629199419170618,
|
|
"rewards/margins": 0.06930971145629883,
|
|
"rewards/rejected": -0.06168051436543465,
|
|
"step": 49
|
|
},
|
|
{
|
|
"epoch": 0.07342143906020558,
|
|
"grad_norm": 30.533233642578125,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.09216851741075516,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 3.5507246376811595e-07,
|
|
"logits/chosen": -1.142214059829712,
|
|
"logits/rejected": -0.34731245040893555,
|
|
"logps/chosen": -94.70893859863281,
|
|
"logps/ref_chosen": -94.62681579589844,
|
|
"logps/ref_rejected": -103.57435607910156,
|
|
"logps/rejected": -104.34310913085938,
|
|
"loss": 1.3264,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.007727333344519138,
|
|
"rewards/margins": 0.0625331848859787,
|
|
"rewards/rejected": -0.07026051729917526,
|
|
"step": 50
|
|
},
|
|
{
|
|
"epoch": 0.07488986784140969,
|
|
"grad_norm": 27.474159240722656,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.09162449836730957,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 3.6231884057971015e-07,
|
|
"logits/chosen": -1.5295517444610596,
|
|
"logits/rejected": -1.121992826461792,
|
|
"logps/chosen": -87.59027099609375,
|
|
"logps/ref_chosen": -87.50727844238281,
|
|
"logps/ref_rejected": -83.47235870361328,
|
|
"logps/rejected": -84.15083312988281,
|
|
"loss": 1.3355,
|
|
"rewards/accuracies": 0.734375,
|
|
"rewards/chosen": -0.007823294959962368,
|
|
"rewards/margins": 0.053790926933288574,
|
|
"rewards/rejected": -0.06161422282457352,
|
|
"step": 51
|
|
},
|
|
{
|
|
"epoch": 0.0763582966226138,
|
|
"grad_norm": 29.224390029907227,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.09114031493663788,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 3.695652173913043e-07,
|
|
"logits/chosen": -1.5085875988006592,
|
|
"logits/rejected": -0.8081971406936646,
|
|
"logps/chosen": -90.67510986328125,
|
|
"logps/ref_chosen": -90.63026428222656,
|
|
"logps/ref_rejected": -87.0390625,
|
|
"logps/rejected": -87.94956970214844,
|
|
"loss": 1.3116,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.004274226725101471,
|
|
"rewards/margins": 0.07802098244428635,
|
|
"rewards/rejected": -0.08229520916938782,
|
|
"step": 52
|
|
},
|
|
{
|
|
"epoch": 0.07782672540381791,
|
|
"grad_norm": 35.840335845947266,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.0905742272734642,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 3.7681159420289855e-07,
|
|
"logits/chosen": -1.548392415046692,
|
|
"logits/rejected": -1.1408261060714722,
|
|
"logps/chosen": -81.63391876220703,
|
|
"logps/ref_chosen": -81.58306884765625,
|
|
"logps/ref_rejected": -95.66152954101562,
|
|
"logps/rejected": -96.76312255859375,
|
|
"loss": 1.2987,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.004804985597729683,
|
|
"rewards/margins": 0.09401369839906693,
|
|
"rewards/rejected": -0.09881868213415146,
|
|
"step": 53
|
|
},
|
|
{
|
|
"epoch": 0.07929515418502203,
|
|
"grad_norm": 35.994388580322266,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.09003961831331253,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 3.8405797101449274e-07,
|
|
"logits/chosen": -1.3138582706451416,
|
|
"logits/rejected": -0.6569335460662842,
|
|
"logps/chosen": -88.93460083007812,
|
|
"logps/ref_chosen": -88.91016387939453,
|
|
"logps/ref_rejected": -99.1175537109375,
|
|
"logps/rejected": -100.24466705322266,
|
|
"loss": 1.2939,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -0.002394177485257387,
|
|
"rewards/margins": 0.09816494584083557,
|
|
"rewards/rejected": -0.10055913031101227,
|
|
"step": 54
|
|
},
|
|
{
|
|
"epoch": 0.08076358296622614,
|
|
"grad_norm": 31.913665771484375,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.08959165960550308,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 3.9130434782608694e-07,
|
|
"logits/chosen": -1.6863343715667725,
|
|
"logits/rejected": -0.6388131976127625,
|
|
"logps/chosen": -92.33493041992188,
|
|
"logps/ref_chosen": -92.45592498779297,
|
|
"logps/ref_rejected": -92.97093963623047,
|
|
"logps/rejected": -93.97943115234375,
|
|
"loss": 1.2933,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": 0.010466434992849827,
|
|
"rewards/margins": 0.10004328191280365,
|
|
"rewards/rejected": -0.0895768478512764,
|
|
"step": 55
|
|
},
|
|
{
|
|
"epoch": 0.08223201174743025,
|
|
"grad_norm": 29.783920288085938,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.08917365968227386,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 3.9855072463768114e-07,
|
|
"logits/chosen": -1.9710514545440674,
|
|
"logits/rejected": -0.6645182371139526,
|
|
"logps/chosen": -87.32936096191406,
|
|
"logps/ref_chosen": -87.23665618896484,
|
|
"logps/ref_rejected": -100.64553833007812,
|
|
"logps/rejected": -101.76547241210938,
|
|
"loss": 1.3033,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.008556234650313854,
|
|
"rewards/margins": 0.09043145924806595,
|
|
"rewards/rejected": -0.09898769855499268,
|
|
"step": 56
|
|
},
|
|
{
|
|
"epoch": 0.08370044052863436,
|
|
"grad_norm": 30.23967933654785,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.08867485821247101,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 4.057971014492754e-07,
|
|
"logits/chosen": -1.7121399641036987,
|
|
"logits/rejected": -0.9029750227928162,
|
|
"logps/chosen": -98.1844253540039,
|
|
"logps/ref_chosen": -98.15074157714844,
|
|
"logps/ref_rejected": -101.05284118652344,
|
|
"logps/rejected": -102.23455047607422,
|
|
"loss": 1.2927,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -0.003269542008638382,
|
|
"rewards/margins": 0.10057489573955536,
|
|
"rewards/rejected": -0.10384444147348404,
|
|
"step": 57
|
|
},
|
|
{
|
|
"epoch": 0.08516886930983847,
|
|
"grad_norm": 33.85087966918945,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.08823369443416595,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 4.1304347826086954e-07,
|
|
"logits/chosen": -1.996552586555481,
|
|
"logits/rejected": -1.058452844619751,
|
|
"logps/chosen": -99.30267333984375,
|
|
"logps/ref_chosen": -99.58097076416016,
|
|
"logps/ref_rejected": -91.22227478027344,
|
|
"logps/rejected": -92.50682067871094,
|
|
"loss": 1.2623,
|
|
"rewards/accuracies": 0.90625,
|
|
"rewards/chosen": 0.02414235845208168,
|
|
"rewards/margins": 0.13646608591079712,
|
|
"rewards/rejected": -0.11232372373342514,
|
|
"step": 58
|
|
},
|
|
{
|
|
"epoch": 0.08663729809104258,
|
|
"grad_norm": 30.440584182739258,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.08765843510627747,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 4.2028985507246374e-07,
|
|
"logits/chosen": -1.5328807830810547,
|
|
"logits/rejected": -0.9605180621147156,
|
|
"logps/chosen": -89.82688903808594,
|
|
"logps/ref_chosen": -89.80232238769531,
|
|
"logps/ref_rejected": -94.27667236328125,
|
|
"logps/rejected": -95.44361877441406,
|
|
"loss": 1.2934,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.0023305192589759827,
|
|
"rewards/margins": 0.09902888536453247,
|
|
"rewards/rejected": -0.10135940462350845,
|
|
"step": 59
|
|
},
|
|
{
|
|
"epoch": 0.0881057268722467,
|
|
"grad_norm": 25.22243309020996,
|
|
"kl/avg_steps": 0.40625,
|
|
"kl/beta": 0.08711396902799606,
|
|
"kl/n_epsilon_steps": 0.296875,
|
|
"kl/p_epsilon_steps": 0.703125,
|
|
"learning_rate": 4.2753623188405794e-07,
|
|
"logits/chosen": -1.692917823791504,
|
|
"logits/rejected": -0.8258851170539856,
|
|
"logps/chosen": -95.11837768554688,
|
|
"logps/ref_chosen": -95.15571594238281,
|
|
"logps/ref_rejected": -91.3724365234375,
|
|
"logps/rejected": -92.52310180664062,
|
|
"loss": 1.2904,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": 0.002980598248541355,
|
|
"rewards/margins": 0.10234124958515167,
|
|
"rewards/rejected": -0.0993606448173523,
|
|
"step": 60
|
|
},
|
|
{
|
|
"epoch": 0.08957415565345081,
|
|
"grad_norm": 25.07730484008789,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.08676150441169739,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 4.3478260869565214e-07,
|
|
"logits/chosen": -1.4720783233642578,
|
|
"logits/rejected": -0.7469910979270935,
|
|
"logps/chosen": -85.89845275878906,
|
|
"logps/ref_chosen": -85.73231506347656,
|
|
"logps/ref_rejected": -97.96575927734375,
|
|
"logps/rejected": -99.31712341308594,
|
|
"loss": 1.2933,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.014695134945213795,
|
|
"rewards/margins": 0.101577028632164,
|
|
"rewards/rejected": -0.11627216637134552,
|
|
"step": 61
|
|
},
|
|
{
|
|
"epoch": 0.09104258443465492,
|
|
"grad_norm": 24.972980499267578,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.08630301803350449,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 4.420289855072464e-07,
|
|
"logits/chosen": -1.7914103269577026,
|
|
"logits/rejected": -0.8692626953125,
|
|
"logps/chosen": -81.72990417480469,
|
|
"logps/ref_chosen": -81.63538360595703,
|
|
"logps/ref_rejected": -84.03831481933594,
|
|
"logps/rejected": -85.38259887695312,
|
|
"loss": 1.2896,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.008462773635983467,
|
|
"rewards/margins": 0.10654792189598083,
|
|
"rewards/rejected": -0.11501070111989975,
|
|
"step": 62
|
|
},
|
|
{
|
|
"epoch": 0.09251101321585903,
|
|
"grad_norm": 30.612600326538086,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.0857403501868248,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 4.4927536231884053e-07,
|
|
"logits/chosen": -1.5954476594924927,
|
|
"logits/rejected": -0.8784996271133423,
|
|
"logps/chosen": -103.65047454833984,
|
|
"logps/ref_chosen": -103.62405395507812,
|
|
"logps/ref_rejected": -103.40303039550781,
|
|
"logps/rejected": -104.9129638671875,
|
|
"loss": 1.2686,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.002559835556894541,
|
|
"rewards/margins": 0.12582427263259888,
|
|
"rewards/rejected": -0.12838411331176758,
|
|
"step": 63
|
|
},
|
|
{
|
|
"epoch": 0.09397944199706314,
|
|
"grad_norm": 30.650426864624023,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.08520779758691788,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 4.5652173913043473e-07,
|
|
"logits/chosen": -1.671600580215454,
|
|
"logits/rejected": -0.9036776423454285,
|
|
"logps/chosen": -87.17109680175781,
|
|
"logps/ref_chosen": -87.0015869140625,
|
|
"logps/ref_rejected": -100.5854721069336,
|
|
"logps/rejected": -102.39601135253906,
|
|
"loss": 1.2612,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -0.014812503941357136,
|
|
"rewards/margins": 0.1382252722978592,
|
|
"rewards/rejected": -0.15303777158260345,
|
|
"step": 64
|
|
},
|
|
{
|
|
"epoch": 0.09544787077826726,
|
|
"grad_norm": 33.42892837524414,
|
|
"kl/avg_steps": 0.75,
|
|
"kl/beta": 0.0847838819026947,
|
|
"kl/n_epsilon_steps": 0.125,
|
|
"kl/p_epsilon_steps": 0.875,
|
|
"learning_rate": 4.63768115942029e-07,
|
|
"logits/chosen": -1.976373314857483,
|
|
"logits/rejected": -1.0799182653427124,
|
|
"logps/chosen": -91.29652404785156,
|
|
"logps/ref_chosen": -91.22191619873047,
|
|
"logps/ref_rejected": -115.33553314208984,
|
|
"logps/rejected": -117.1944351196289,
|
|
"loss": 1.2492,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -0.006513871252536774,
|
|
"rewards/margins": 0.14960609376430511,
|
|
"rewards/rejected": -0.15611997246742249,
|
|
"step": 65
|
|
},
|
|
{
|
|
"epoch": 0.09691629955947137,
|
|
"grad_norm": 22.01787567138672,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.08415273576974869,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 4.7101449275362313e-07,
|
|
"logits/chosen": -2.0008394718170166,
|
|
"logits/rejected": -1.0472722053527832,
|
|
"logps/chosen": -84.10005187988281,
|
|
"logps/ref_chosen": -83.78422546386719,
|
|
"logps/ref_rejected": -82.7520980834961,
|
|
"logps/rejected": -84.29136657714844,
|
|
"loss": 1.2937,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.026944037526845932,
|
|
"rewards/margins": 0.10165668278932571,
|
|
"rewards/rejected": -0.12860071659088135,
|
|
"step": 66
|
|
},
|
|
{
|
|
"epoch": 0.09838472834067548,
|
|
"grad_norm": 20.735403060913086,
|
|
"kl/avg_steps": 0.34375,
|
|
"kl/beta": 0.08373406529426575,
|
|
"kl/n_epsilon_steps": 0.328125,
|
|
"kl/p_epsilon_steps": 0.671875,
|
|
"learning_rate": 4.782608695652174e-07,
|
|
"logits/chosen": -2.0153298377990723,
|
|
"logits/rejected": -1.0666594505310059,
|
|
"logps/chosen": -88.51578521728516,
|
|
"logps/ref_chosen": -87.67295837402344,
|
|
"logps/ref_rejected": -79.08674621582031,
|
|
"logps/rejected": -81.0645751953125,
|
|
"loss": 1.3088,
|
|
"rewards/accuracies": 0.671875,
|
|
"rewards/chosen": -0.07106294482946396,
|
|
"rewards/margins": 0.09327295422554016,
|
|
"rewards/rejected": -0.16433589160442352,
|
|
"step": 67
|
|
},
|
|
{
|
|
"epoch": 0.09985315712187959,
|
|
"grad_norm": 24.614826202392578,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.08344721049070358,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.855072463768116e-07,
|
|
"logits/chosen": -1.7102564573287964,
|
|
"logits/rejected": -1.115820288658142,
|
|
"logps/chosen": -97.353759765625,
|
|
"logps/ref_chosen": -97.07884216308594,
|
|
"logps/ref_rejected": -84.11872863769531,
|
|
"logps/rejected": -86.50021362304688,
|
|
"loss": 1.2341,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.02329547144472599,
|
|
"rewards/margins": 0.1738092005252838,
|
|
"rewards/rejected": -0.19710469245910645,
|
|
"step": 68
|
|
},
|
|
{
|
|
"epoch": 0.1013215859030837,
|
|
"grad_norm": 26.68158531188965,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.0829804465174675,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.927536231884058e-07,
|
|
"logits/chosen": -2.055922269821167,
|
|
"logits/rejected": -1.1354026794433594,
|
|
"logps/chosen": -86.03010559082031,
|
|
"logps/ref_chosen": -85.71971130371094,
|
|
"logps/ref_rejected": -109.4802017211914,
|
|
"logps/rejected": -111.91256713867188,
|
|
"loss": 1.2348,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.025998366996645927,
|
|
"rewards/margins": 0.1740657240152359,
|
|
"rewards/rejected": -0.20006409287452698,
|
|
"step": 69
|
|
},
|
|
{
|
|
"epoch": 0.1027900146842878,
|
|
"grad_norm": 25.407556533813477,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.08251629024744034,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 5e-07,
|
|
"logits/chosen": -1.9641715288162231,
|
|
"logits/rejected": -1.39949631690979,
|
|
"logps/chosen": -95.62403869628906,
|
|
"logps/ref_chosen": -95.00994873046875,
|
|
"logps/ref_rejected": -96.21272277832031,
|
|
"logps/rejected": -99.49327087402344,
|
|
"loss": 1.201,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.05099921301007271,
|
|
"rewards/margins": 0.21753652393817902,
|
|
"rewards/rejected": -0.2685357332229614,
|
|
"step": 70
|
|
},
|
|
{
|
|
"epoch": 0.10425844346549193,
|
|
"grad_norm": 30.39597511291504,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.08208024501800537,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 4.999967061337492e-07,
|
|
"logits/chosen": -2.9308741092681885,
|
|
"logits/rejected": -1.5702245235443115,
|
|
"logps/chosen": -91.07780456542969,
|
|
"logps/ref_chosen": -90.97735595703125,
|
|
"logps/ref_rejected": -102.59103393554688,
|
|
"logps/rejected": -106.28335571289062,
|
|
"loss": 1.14,
|
|
"rewards/accuracies": 0.90625,
|
|
"rewards/chosen": -0.00880364328622818,
|
|
"rewards/margins": 0.291946142911911,
|
|
"rewards/rejected": -0.3007497787475586,
|
|
"step": 71
|
|
},
|
|
{
|
|
"epoch": 0.10572687224669604,
|
|
"grad_norm": 25.40204620361328,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.08149450272321701,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.999868246217933e-07,
|
|
"logits/chosen": -2.533379077911377,
|
|
"logits/rejected": -1.5070923566818237,
|
|
"logps/chosen": -98.47068786621094,
|
|
"logps/ref_chosen": -97.89152526855469,
|
|
"logps/ref_rejected": -100.19171142578125,
|
|
"logps/rejected": -103.68728637695312,
|
|
"loss": 1.1831,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.04744531959295273,
|
|
"rewards/margins": 0.23507529497146606,
|
|
"rewards/rejected": -0.282520592212677,
|
|
"step": 72
|
|
},
|
|
{
|
|
"epoch": 0.10719530102790015,
|
|
"grad_norm": 24.58971405029297,
|
|
"kl/avg_steps": 0.40625,
|
|
"kl/beta": 0.08103866130113602,
|
|
"kl/n_epsilon_steps": 0.296875,
|
|
"kl/p_epsilon_steps": 0.703125,
|
|
"learning_rate": 4.999703557245192e-07,
|
|
"logits/chosen": -2.8086342811584473,
|
|
"logits/rejected": -1.789698600769043,
|
|
"logps/chosen": -96.5920181274414,
|
|
"logps/ref_chosen": -95.7690200805664,
|
|
"logps/ref_rejected": -95.93243408203125,
|
|
"logps/rejected": -99.84471130371094,
|
|
"loss": 1.1906,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -0.06719671189785004,
|
|
"rewards/margins": 0.24726057052612305,
|
|
"rewards/rejected": -0.3144572973251343,
|
|
"step": 73
|
|
},
|
|
{
|
|
"epoch": 0.10866372980910426,
|
|
"grad_norm": 27.501953125,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.080710768699646,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 4.999472998758977e-07,
|
|
"logits/chosen": -2.912767171859741,
|
|
"logits/rejected": -1.9761888980865479,
|
|
"logps/chosen": -79.79828643798828,
|
|
"logps/ref_chosen": -78.80839538574219,
|
|
"logps/ref_rejected": -101.64676666259766,
|
|
"logps/rejected": -106.22779083251953,
|
|
"loss": 1.1518,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -0.08012489974498749,
|
|
"rewards/margins": 0.2865511476993561,
|
|
"rewards/rejected": -0.3666760325431824,
|
|
"step": 74
|
|
},
|
|
{
|
|
"epoch": 0.11013215859030837,
|
|
"grad_norm": 29.259719848632812,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.08030922710895538,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 4.999176576834721e-07,
|
|
"logits/chosen": -2.8685462474823,
|
|
"logits/rejected": -1.5099756717681885,
|
|
"logps/chosen": -79.43289947509766,
|
|
"logps/ref_chosen": -78.28185272216797,
|
|
"logps/ref_rejected": -115.40311431884766,
|
|
"logps/rejected": -121.06864929199219,
|
|
"loss": 1.1034,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.09254170209169388,
|
|
"rewards/margins": 0.35883650183677673,
|
|
"rewards/rejected": -0.45137819647789,
|
|
"step": 75
|
|
},
|
|
{
|
|
"epoch": 0.11160058737151249,
|
|
"grad_norm": 19.229440689086914,
|
|
"kl/avg_steps": 0.375,
|
|
"kl/beta": 0.079884834587574,
|
|
"kl/n_epsilon_steps": 0.3125,
|
|
"kl/p_epsilon_steps": 0.6875,
|
|
"learning_rate": 4.998814299283415e-07,
|
|
"logits/chosen": -3.326892375946045,
|
|
"logits/rejected": -2.0114529132843018,
|
|
"logps/chosen": -89.3749008178711,
|
|
"logps/ref_chosen": -87.87714385986328,
|
|
"logps/ref_rejected": -85.71968078613281,
|
|
"logps/rejected": -90.23574829101562,
|
|
"loss": 1.1945,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.12006325274705887,
|
|
"rewards/margins": 0.23822948336601257,
|
|
"rewards/rejected": -0.35829272866249084,
|
|
"step": 76
|
|
},
|
|
{
|
|
"epoch": 0.1130690161527166,
|
|
"grad_norm": 22.950105667114258,
|
|
"kl/avg_steps": 0.4375,
|
|
"kl/beta": 0.07958638668060303,
|
|
"kl/n_epsilon_steps": 0.28125,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 4.998386175651409e-07,
|
|
"logits/chosen": -2.7979376316070557,
|
|
"logits/rejected": -1.765808343887329,
|
|
"logps/chosen": -101.35743713378906,
|
|
"logps/ref_chosen": -99.70034790039062,
|
|
"logps/ref_rejected": -98.20576477050781,
|
|
"logps/rejected": -103.30995178222656,
|
|
"loss": 1.1752,
|
|
"rewards/accuracies": 0.71875,
|
|
"rewards/chosen": -0.1325981616973877,
|
|
"rewards/margins": 0.2707356810569763,
|
|
"rewards/rejected": -0.4033338725566864,
|
|
"step": 77
|
|
},
|
|
{
|
|
"epoch": 0.1145374449339207,
|
|
"grad_norm": 22.01249885559082,
|
|
"kl/avg_steps": 0.484375,
|
|
"kl/beta": 0.07923971116542816,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 4.997892217220159e-07,
|
|
"logits/chosen": -2.7685134410858154,
|
|
"logits/rejected": -1.9960670471191406,
|
|
"logps/chosen": -91.00945281982422,
|
|
"logps/ref_chosen": -90.29670715332031,
|
|
"logps/ref_rejected": -91.13772583007812,
|
|
"logps/rejected": -95.54225158691406,
|
|
"loss": 1.1464,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.05697726085782051,
|
|
"rewards/margins": 0.2894305884838104,
|
|
"rewards/rejected": -0.34640786051750183,
|
|
"step": 78
|
|
},
|
|
{
|
|
"epoch": 0.11600587371512482,
|
|
"grad_norm": 24.0369815826416,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.07885774970054626,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 4.997332437005931e-07,
|
|
"logits/chosen": -3.548121452331543,
|
|
"logits/rejected": -2.20076847076416,
|
|
"logps/chosen": -87.27758026123047,
|
|
"logps/ref_chosen": -86.37832641601562,
|
|
"logps/ref_rejected": -94.10777282714844,
|
|
"logps/rejected": -99.85430908203125,
|
|
"loss": 1.0944,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.0716433972120285,
|
|
"rewards/margins": 0.37777939438819885,
|
|
"rewards/rejected": -0.44942277669906616,
|
|
"step": 79
|
|
},
|
|
{
|
|
"epoch": 0.11747430249632893,
|
|
"grad_norm": 22.031478881835938,
|
|
"kl/avg_steps": 0.40625,
|
|
"kl/beta": 0.07839228957891464,
|
|
"kl/n_epsilon_steps": 0.296875,
|
|
"kl/p_epsilon_steps": 0.703125,
|
|
"learning_rate": 4.996706849759452e-07,
|
|
"logits/chosen": -3.074854850769043,
|
|
"logits/rejected": -2.0579161643981934,
|
|
"logps/chosen": -95.61634826660156,
|
|
"logps/ref_chosen": -93.97032165527344,
|
|
"logps/ref_rejected": -92.57441711425781,
|
|
"logps/rejected": -98.0230941772461,
|
|
"loss": 1.1543,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.12929072976112366,
|
|
"rewards/margins": 0.29479026794433594,
|
|
"rewards/rejected": -0.424081027507782,
|
|
"step": 80
|
|
},
|
|
{
|
|
"epoch": 0.11894273127753303,
|
|
"grad_norm": 24.160234451293945,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.078075110912323,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 4.996015471965529e-07,
|
|
"logits/chosen": -3.137648105621338,
|
|
"logits/rejected": -1.6095255613327026,
|
|
"logps/chosen": -100.93592834472656,
|
|
"logps/ref_chosen": -99.83012390136719,
|
|
"logps/ref_rejected": -133.67245483398438,
|
|
"logps/rejected": -140.25442504882812,
|
|
"loss": 1.0766,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -0.0865008607506752,
|
|
"rewards/margins": 0.42291849851608276,
|
|
"rewards/rejected": -0.509419322013855,
|
|
"step": 81
|
|
},
|
|
{
|
|
"epoch": 0.12041116005873716,
|
|
"grad_norm": 21.8071346282959,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.07766252756118774,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 4.995258321842611e-07,
|
|
"logits/chosen": -3.389235019683838,
|
|
"logits/rejected": -2.2665112018585205,
|
|
"logps/chosen": -85.08958435058594,
|
|
"logps/ref_chosen": -83.04598236083984,
|
|
"logps/ref_rejected": -94.52595520019531,
|
|
"logps/rejected": -100.36309814453125,
|
|
"loss": 1.1567,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.15859441459178925,
|
|
"rewards/margins": 0.29103851318359375,
|
|
"rewards/rejected": -0.4496329426765442,
|
|
"step": 82
|
|
},
|
|
{
|
|
"epoch": 0.12187958883994127,
|
|
"grad_norm": 21.79849624633789,
|
|
"kl/avg_steps": 0.28125,
|
|
"kl/beta": 0.07725212723016739,
|
|
"kl/n_epsilon_steps": 0.359375,
|
|
"kl/p_epsilon_steps": 0.640625,
|
|
"learning_rate": 4.994435419342304e-07,
|
|
"logits/chosen": -3.9037327766418457,
|
|
"logits/rejected": -2.2527427673339844,
|
|
"logps/chosen": -94.64817810058594,
|
|
"logps/ref_chosen": -92.17621612548828,
|
|
"logps/ref_rejected": -107.74464416503906,
|
|
"logps/rejected": -114.48917388916016,
|
|
"loss": 1.1507,
|
|
"rewards/accuracies": 0.703125,
|
|
"rewards/chosen": -0.19215711951255798,
|
|
"rewards/margins": 0.3255422115325928,
|
|
"rewards/rejected": -0.5176993012428284,
|
|
"step": 83
|
|
},
|
|
{
|
|
"epoch": 0.12334801762114538,
|
|
"grad_norm": 22.36842918395996,
|
|
"kl/avg_steps": 0.3125,
|
|
"kl/beta": 0.07703546434640884,
|
|
"kl/n_epsilon_steps": 0.34375,
|
|
"kl/p_epsilon_steps": 0.65625,
|
|
"learning_rate": 4.993546786148857e-07,
|
|
"logits/chosen": -3.4791769981384277,
|
|
"logits/rejected": -2.393409490585327,
|
|
"logps/chosen": -104.68441772460938,
|
|
"logps/ref_chosen": -101.5264892578125,
|
|
"logps/ref_rejected": -92.42608642578125,
|
|
"logps/rejected": -98.19878387451172,
|
|
"loss": 1.2405,
|
|
"rewards/accuracies": 0.6875,
|
|
"rewards/chosen": -0.24369969964027405,
|
|
"rewards/margins": 0.19829508662223816,
|
|
"rewards/rejected": -0.4419947564601898,
|
|
"step": 84
|
|
},
|
|
{
|
|
"epoch": 0.12481644640234948,
|
|
"grad_norm": 20.850772857666016,
|
|
"kl/avg_steps": 0.25,
|
|
"kl/beta": 0.07679548114538193,
|
|
"kl/n_epsilon_steps": 0.375,
|
|
"kl/p_epsilon_steps": 0.625,
|
|
"learning_rate": 4.992592445678582e-07,
|
|
"logits/chosen": -3.586297035217285,
|
|
"logits/rejected": -2.9060792922973633,
|
|
"logps/chosen": -98.8546371459961,
|
|
"logps/ref_chosen": -96.12738037109375,
|
|
"logps/ref_rejected": -85.05519104003906,
|
|
"logps/rejected": -91.33653259277344,
|
|
"loss": 1.1934,
|
|
"rewards/accuracies": 0.734375,
|
|
"rewards/chosen": -0.21086883544921875,
|
|
"rewards/margins": 0.2689524292945862,
|
|
"rewards/rejected": -0.4798212945461273,
|
|
"step": 85
|
|
},
|
|
{
|
|
"epoch": 0.1262848751835536,
|
|
"grad_norm": 18.46939468383789,
|
|
"kl/avg_steps": 0.40625,
|
|
"kl/beta": 0.0766039714217186,
|
|
"kl/n_epsilon_steps": 0.296875,
|
|
"kl/p_epsilon_steps": 0.703125,
|
|
"learning_rate": 4.991572423079235e-07,
|
|
"logits/chosen": -3.4410600662231445,
|
|
"logits/rejected": -2.2581920623779297,
|
|
"logps/chosen": -83.9306640625,
|
|
"logps/ref_chosen": -81.70426940917969,
|
|
"logps/ref_rejected": -93.6554946899414,
|
|
"logps/rejected": -100.36227416992188,
|
|
"loss": 1.1471,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.17061173915863037,
|
|
"rewards/margins": 0.3391328752040863,
|
|
"rewards/rejected": -0.5097446441650391,
|
|
"step": 86
|
|
},
|
|
{
|
|
"epoch": 0.1277533039647577,
|
|
"grad_norm": 21.282371520996094,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.07629402726888657,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 4.990486745229364e-07,
|
|
"logits/chosen": -4.2626729011535645,
|
|
"logits/rejected": -2.455996513366699,
|
|
"logps/chosen": -95.24873352050781,
|
|
"logps/ref_chosen": -92.68596649169922,
|
|
"logps/ref_rejected": -102.91818237304688,
|
|
"logps/rejected": -110.37586212158203,
|
|
"loss": 1.137,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.19558626413345337,
|
|
"rewards/margins": 0.36860454082489014,
|
|
"rewards/rejected": -0.5641908049583435,
|
|
"step": 87
|
|
},
|
|
{
|
|
"epoch": 0.12922173274596183,
|
|
"grad_norm": 19.734506607055664,
|
|
"kl/avg_steps": 0.34375,
|
|
"kl/beta": 0.07589085400104523,
|
|
"kl/n_epsilon_steps": 0.328125,
|
|
"kl/p_epsilon_steps": 0.671875,
|
|
"learning_rate": 4.989335440737586e-07,
|
|
"logits/chosen": -3.0765509605407715,
|
|
"logits/rejected": -2.2642664909362793,
|
|
"logps/chosen": -103.91253662109375,
|
|
"logps/ref_chosen": -100.76298522949219,
|
|
"logps/ref_rejected": -113.15037536621094,
|
|
"logps/rejected": -120.1460189819336,
|
|
"loss": 1.1828,
|
|
"rewards/accuracies": 0.734375,
|
|
"rewards/chosen": -0.239786297082901,
|
|
"rewards/margins": 0.28768616914749146,
|
|
"rewards/rejected": -0.5274724960327148,
|
|
"step": 88
|
|
},
|
|
{
|
|
"epoch": 0.13069016152716592,
|
|
"grad_norm": 20.043725967407227,
|
|
"kl/avg_steps": 0.34375,
|
|
"kl/beta": 0.07563087344169617,
|
|
"kl/n_epsilon_steps": 0.328125,
|
|
"kl/p_epsilon_steps": 0.671875,
|
|
"learning_rate": 4.988118539941847e-07,
|
|
"logits/chosen": -3.5043540000915527,
|
|
"logits/rejected": -2.6482343673706055,
|
|
"logps/chosen": -92.20272827148438,
|
|
"logps/ref_chosen": -89.69108581542969,
|
|
"logps/ref_rejected": -88.56832885742188,
|
|
"logps/rejected": -95.34589385986328,
|
|
"loss": 1.1584,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -0.19024960696697235,
|
|
"rewards/margins": 0.3187325596809387,
|
|
"rewards/rejected": -0.5089821815490723,
|
|
"step": 89
|
|
},
|
|
{
|
|
"epoch": 0.13215859030837004,
|
|
"grad_norm": 22.877426147460938,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.07537178695201874,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 4.986836074908615e-07,
|
|
"logits/chosen": -4.08919620513916,
|
|
"logits/rejected": -2.59071683883667,
|
|
"logps/chosen": -83.66722106933594,
|
|
"logps/ref_chosen": -81.38255310058594,
|
|
"logps/ref_rejected": -117.77714538574219,
|
|
"logps/rejected": -125.76116180419922,
|
|
"loss": 1.0847,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.1728699803352356,
|
|
"rewards/margins": 0.4245910942554474,
|
|
"rewards/rejected": -0.5974611043930054,
|
|
"step": 90
|
|
},
|
|
{
|
|
"epoch": 0.13362701908957417,
|
|
"grad_norm": 21.175289154052734,
|
|
"kl/avg_steps": 0.21875,
|
|
"kl/beta": 0.07497348636388779,
|
|
"kl/n_epsilon_steps": 0.390625,
|
|
"kl/p_epsilon_steps": 0.609375,
|
|
"learning_rate": 4.985488079432037e-07,
|
|
"logits/chosen": -3.6395578384399414,
|
|
"logits/rejected": -2.8058741092681885,
|
|
"logps/chosen": -99.98194885253906,
|
|
"logps/ref_chosen": -97.22188568115234,
|
|
"logps/ref_rejected": -92.97674560546875,
|
|
"logps/rejected": -100.44671630859375,
|
|
"loss": 1.1414,
|
|
"rewards/accuracies": 0.703125,
|
|
"rewards/chosen": -0.20845842361450195,
|
|
"rewards/margins": 0.3486502170562744,
|
|
"rewards/rejected": -0.5571086406707764,
|
|
"step": 91
|
|
},
|
|
{
|
|
"epoch": 0.13509544787077826,
|
|
"grad_norm": 19.441743850708008,
|
|
"kl/avg_steps": 0.28125,
|
|
"kl/beta": 0.07480984181165695,
|
|
"kl/n_epsilon_steps": 0.359375,
|
|
"kl/p_epsilon_steps": 0.640625,
|
|
"learning_rate": 4.984074589033043e-07,
|
|
"logits/chosen": -4.114226341247559,
|
|
"logits/rejected": -3.050344944000244,
|
|
"logps/chosen": -87.49899291992188,
|
|
"logps/ref_chosen": -84.5302734375,
|
|
"logps/ref_rejected": -84.5013198852539,
|
|
"logps/rejected": -91.48877716064453,
|
|
"loss": 1.1718,
|
|
"rewards/accuracies": 0.703125,
|
|
"rewards/chosen": -0.22265848517417908,
|
|
"rewards/margins": 0.29672500491142273,
|
|
"rewards/rejected": -0.5193834900856018,
|
|
"step": 92
|
|
},
|
|
{
|
|
"epoch": 0.13656387665198239,
|
|
"grad_norm": 20.867734909057617,
|
|
"kl/avg_steps": 0.28125,
|
|
"kl/beta": 0.07460002601146698,
|
|
"kl/n_epsilon_steps": 0.359375,
|
|
"kl/p_epsilon_steps": 0.640625,
|
|
"learning_rate": 4.982595640958425e-07,
|
|
"logits/chosen": -3.9874632358551025,
|
|
"logits/rejected": -2.371685266494751,
|
|
"logps/chosen": -93.55166625976562,
|
|
"logps/ref_chosen": -90.25043487548828,
|
|
"logps/ref_rejected": -84.09422302246094,
|
|
"logps/rejected": -91.32198333740234,
|
|
"loss": 1.192,
|
|
"rewards/accuracies": 0.71875,
|
|
"rewards/chosen": -0.24706237018108368,
|
|
"rewards/margins": 0.2884610891342163,
|
|
"rewards/rejected": -0.535523533821106,
|
|
"step": 93
|
|
},
|
|
{
|
|
"epoch": 0.13803230543318648,
|
|
"grad_norm": 22.021520614624023,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.07439080625772476,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 4.98105127417984e-07,
|
|
"logits/chosen": -4.104582786560059,
|
|
"logits/rejected": -2.78466796875,
|
|
"logps/chosen": -95.84749603271484,
|
|
"logps/ref_chosen": -92.4542236328125,
|
|
"logps/ref_rejected": -105.24728393554688,
|
|
"logps/rejected": -114.69227600097656,
|
|
"loss": 1.0808,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.25266966223716736,
|
|
"rewards/margins": 0.44442811608314514,
|
|
"rewards/rejected": -0.6970977783203125,
|
|
"step": 94
|
|
},
|
|
{
|
|
"epoch": 0.1395007342143906,
|
|
"grad_norm": 18.556907653808594,
|
|
"kl/avg_steps": 0.34375,
|
|
"kl/beta": 0.07399769127368927,
|
|
"kl/n_epsilon_steps": 0.328125,
|
|
"kl/p_epsilon_steps": 0.671875,
|
|
"learning_rate": 4.979441529392784e-07,
|
|
"logits/chosen": -4.07137393951416,
|
|
"logits/rejected": -2.9404728412628174,
|
|
"logps/chosen": -81.59260559082031,
|
|
"logps/ref_chosen": -78.87370300292969,
|
|
"logps/ref_rejected": -83.59121704101562,
|
|
"logps/rejected": -91.0462875366211,
|
|
"loss": 1.1379,
|
|
"rewards/accuracies": 0.703125,
|
|
"rewards/chosen": -0.201766699552536,
|
|
"rewards/margins": 0.346103310585022,
|
|
"rewards/rejected": -0.5478700399398804,
|
|
"step": 95
|
|
},
|
|
{
|
|
"epoch": 0.14096916299559473,
|
|
"grad_norm": 19.24999237060547,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.07374419271945953,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 4.977766449015534e-07,
|
|
"logits/chosen": -3.7676031589508057,
|
|
"logits/rejected": -2.4514598846435547,
|
|
"logps/chosen": -109.04351806640625,
|
|
"logps/ref_chosen": -106.5921630859375,
|
|
"logps/ref_rejected": -101.76802062988281,
|
|
"logps/rejected": -110.20415496826172,
|
|
"loss": 1.0757,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.18087545037269592,
|
|
"rewards/margins": 0.43662530183792114,
|
|
"rewards/rejected": -0.6175007224082947,
|
|
"step": 96
|
|
},
|
|
{
|
|
"epoch": 0.14243759177679882,
|
|
"grad_norm": 21.76212501525879,
|
|
"kl/avg_steps": 0.40625,
|
|
"kl/beta": 0.07340013235807419,
|
|
"kl/n_epsilon_steps": 0.296875,
|
|
"kl/p_epsilon_steps": 0.703125,
|
|
"learning_rate": 4.976026077188012e-07,
|
|
"logits/chosen": -4.675760746002197,
|
|
"logits/rejected": -3.565962791442871,
|
|
"logps/chosen": -92.28569030761719,
|
|
"logps/ref_chosen": -88.67988586425781,
|
|
"logps/ref_rejected": -84.81229400634766,
|
|
"logps/rejected": -93.66389465332031,
|
|
"loss": 1.1006,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.26550614833831787,
|
|
"rewards/margins": 0.380950003862381,
|
|
"rewards/rejected": -0.6464561223983765,
|
|
"step": 97
|
|
},
|
|
{
|
|
"epoch": 0.14390602055800295,
|
|
"grad_norm": 19.99883460998535,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.0731031522154808,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 4.974220459770639e-07,
|
|
"logits/chosen": -3.876922607421875,
|
|
"logits/rejected": -2.7262768745422363,
|
|
"logps/chosen": -95.04853057861328,
|
|
"logps/ref_chosen": -92.24249267578125,
|
|
"logps/ref_rejected": -101.51948547363281,
|
|
"logps/rejected": -110.45492553710938,
|
|
"loss": 1.0488,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -0.2048511505126953,
|
|
"rewards/margins": 0.4435645043849945,
|
|
"rewards/rejected": -0.6484156250953674,
|
|
"step": 98
|
|
},
|
|
{
|
|
"epoch": 0.14537444933920704,
|
|
"grad_norm": 20.251489639282227,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.0727168396115303,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 4.972349644343108e-07,
|
|
"logits/chosen": -4.384429454803467,
|
|
"logits/rejected": -3.071768045425415,
|
|
"logps/chosen": -74.27732849121094,
|
|
"logps/ref_chosen": -72.18464660644531,
|
|
"logps/ref_rejected": -91.88131713867188,
|
|
"logps/rejected": -100.81095123291016,
|
|
"loss": 1.0479,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.15243776142597198,
|
|
"rewards/margins": 0.49188998341560364,
|
|
"rewards/rejected": -0.6443277597427368,
|
|
"step": 99
|
|
},
|
|
{
|
|
"epoch": 0.14684287812041116,
|
|
"grad_norm": 17.54705047607422,
|
|
"kl/avg_steps": 0.40625,
|
|
"kl/beta": 0.07233257591724396,
|
|
"kl/n_epsilon_steps": 0.296875,
|
|
"kl/p_epsilon_steps": 0.703125,
|
|
"learning_rate": 4.970413680203148e-07,
|
|
"logits/chosen": -4.1084113121032715,
|
|
"logits/rejected": -2.772047758102417,
|
|
"logps/chosen": -92.06529235839844,
|
|
"logps/ref_chosen": -89.51382446289062,
|
|
"logps/ref_rejected": -81.21713256835938,
|
|
"logps/rejected": -88.53147888183594,
|
|
"loss": 1.1171,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.18433162569999695,
|
|
"rewards/margins": 0.3410602807998657,
|
|
"rewards/rejected": -0.5253919363021851,
|
|
"step": 100
|
|
},
|
|
{
|
|
"epoch": 0.14684287812041116,
|
|
"eval_kl/n_epsilon_steps": 0.3857020437717438,
|
|
"eval_kl/p_epsilon_steps": 0.6134417653083801,
|
|
"eval_logits/chosen": -3.108599901199341,
|
|
"eval_logits/rejected": -2.4245166778564453,
|
|
"eval_logps/chosen": -104.82538604736328,
|
|
"eval_logps/ref_chosen": -100.49356842041016,
|
|
"eval_logps/ref_rejected": -94.06775665283203,
|
|
"eval_logps/rejected": -101.79056549072266,
|
|
"eval_loss": 0.6201064586639404,
|
|
"eval_rewards/accuracies": 0.6840753555297852,
|
|
"eval_rewards/chosen": -0.31309446692466736,
|
|
"eval_rewards/margins": 0.24013973772525787,
|
|
"eval_rewards/rejected": -0.5532342195510864,
|
|
"eval_runtime": 48.5803,
|
|
"eval_samples_per_second": 48.147,
|
|
"eval_steps_per_second": 1.523,
|
|
"step": 100
|
|
},
|
|
{
|
|
"epoch": 0.14831130690161526,
|
|
"grad_norm": 21.208911895751953,
|
|
"kl/avg_steps": 0.34375,
|
|
"kl/beta": 0.07203991711139679,
|
|
"kl/n_epsilon_steps": 0.328125,
|
|
"kl/p_epsilon_steps": 0.671875,
|
|
"learning_rate": 4.968412618365215e-07,
|
|
"logits/chosen": -3.8228065967559814,
|
|
"logits/rejected": -3.2613918781280518,
|
|
"logps/chosen": -98.40846252441406,
|
|
"logps/ref_chosen": -94.4031982421875,
|
|
"logps/ref_rejected": -89.91001892089844,
|
|
"logps/rejected": -98.31871032714844,
|
|
"loss": 1.1745,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -0.2888886034488678,
|
|
"rewards/margins": 0.3126002252101898,
|
|
"rewards/rejected": -0.6014888286590576,
|
|
"step": 101
|
|
},
|
|
{
|
|
"epoch": 0.14977973568281938,
|
|
"grad_norm": 26.08466148376465,
|
|
"kl/avg_steps": 0.390625,
|
|
"kl/beta": 0.07179312407970428,
|
|
"kl/n_epsilon_steps": 0.296875,
|
|
"kl/p_epsilon_steps": 0.6875,
|
|
"learning_rate": 4.966346511559149e-07,
|
|
"logits/chosen": -4.024471282958984,
|
|
"logits/rejected": -3.1433403491973877,
|
|
"logps/chosen": -96.39299774169922,
|
|
"logps/ref_chosen": -91.60942840576172,
|
|
"logps/ref_rejected": -75.76891326904297,
|
|
"logps/rejected": -84.2006607055664,
|
|
"loss": 1.2292,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -0.3448712229728699,
|
|
"rewards/margins": 0.2573922276496887,
|
|
"rewards/rejected": -0.6022634506225586,
|
|
"step": 102
|
|
},
|
|
{
|
|
"epoch": 0.1512481644640235,
|
|
"grad_norm": 20.780960083007812,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.07151377201080322,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 4.964215414228785e-07,
|
|
"logits/chosen": -4.443181037902832,
|
|
"logits/rejected": -3.242924928665161,
|
|
"logps/chosen": -101.16336059570312,
|
|
"logps/ref_chosen": -97.40238952636719,
|
|
"logps/ref_rejected": -98.98168182373047,
|
|
"logps/rejected": -109.21807861328125,
|
|
"loss": 1.0673,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -0.26924505829811096,
|
|
"rewards/margins": 0.4573027193546295,
|
|
"rewards/rejected": -0.7265477776527405,
|
|
"step": 103
|
|
},
|
|
{
|
|
"epoch": 0.1527165932452276,
|
|
"grad_norm": 17.77613639831543,
|
|
"kl/avg_steps": 0.40625,
|
|
"kl/beta": 0.07113586366176605,
|
|
"kl/n_epsilon_steps": 0.296875,
|
|
"kl/p_epsilon_steps": 0.703125,
|
|
"learning_rate": 4.96201938253052e-07,
|
|
"logits/chosen": -4.106165885925293,
|
|
"logits/rejected": -3.3227481842041016,
|
|
"logps/chosen": -97.89046478271484,
|
|
"logps/ref_chosen": -94.85636138916016,
|
|
"logps/ref_rejected": -97.02668762207031,
|
|
"logps/rejected": -106.888427734375,
|
|
"loss": 1.0609,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.2158360481262207,
|
|
"rewards/margins": 0.4804786741733551,
|
|
"rewards/rejected": -0.6963146924972534,
|
|
"step": 104
|
|
},
|
|
{
|
|
"epoch": 0.15418502202643172,
|
|
"grad_norm": 20.64962387084961,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.07084804773330688,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 4.959758474331832e-07,
|
|
"logits/chosen": -4.99745512008667,
|
|
"logits/rejected": -3.3221869468688965,
|
|
"logps/chosen": -100.64320373535156,
|
|
"logps/ref_chosen": -96.75790405273438,
|
|
"logps/ref_rejected": -101.5774917602539,
|
|
"logps/rejected": -112.40327453613281,
|
|
"loss": 1.0544,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.2757706940174103,
|
|
"rewards/margins": 0.4858649969100952,
|
|
"rewards/rejected": -0.7616356611251831,
|
|
"step": 105
|
|
},
|
|
{
|
|
"epoch": 0.15565345080763582,
|
|
"grad_norm": 19.51259422302246,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.07047365605831146,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 4.957432749209755e-07,
|
|
"logits/chosen": -4.652019500732422,
|
|
"logits/rejected": -3.1131434440612793,
|
|
"logps/chosen": -96.30393981933594,
|
|
"logps/ref_chosen": -92.43791198730469,
|
|
"logps/ref_rejected": -91.00969696044922,
|
|
"logps/rejected": -101.36027526855469,
|
|
"loss": 1.0813,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -0.27257949113845825,
|
|
"rewards/margins": 0.4514220058917999,
|
|
"rewards/rejected": -0.7240015268325806,
|
|
"step": 106
|
|
},
|
|
{
|
|
"epoch": 0.15712187958883994,
|
|
"grad_norm": 19.533924102783203,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.07014484703540802,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 4.955042268449307e-07,
|
|
"logits/chosen": -3.9555153846740723,
|
|
"logits/rejected": -2.785179615020752,
|
|
"logps/chosen": -108.09788513183594,
|
|
"logps/ref_chosen": -105.12500762939453,
|
|
"logps/ref_rejected": -101.18250274658203,
|
|
"logps/rejected": -111.86402130126953,
|
|
"loss": 1.0554,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.20865142345428467,
|
|
"rewards/margins": 0.5342296361923218,
|
|
"rewards/rejected": -0.7428811192512512,
|
|
"step": 107
|
|
},
|
|
{
|
|
"epoch": 0.15859030837004406,
|
|
"grad_norm": 19.818805694580078,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.0697958692908287,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 4.952587095041881e-07,
|
|
"logits/chosen": -5.145913124084473,
|
|
"logits/rejected": -3.347729206085205,
|
|
"logps/chosen": -93.46691131591797,
|
|
"logps/ref_chosen": -89.20936584472656,
|
|
"logps/ref_rejected": -101.83784484863281,
|
|
"logps/rejected": -113.4327163696289,
|
|
"loss": 1.0533,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -0.29724550247192383,
|
|
"rewards/margins": 0.5062732100486755,
|
|
"rewards/rejected": -0.8035186529159546,
|
|
"step": 108
|
|
},
|
|
{
|
|
"epoch": 0.16005873715124816,
|
|
"grad_norm": 18.633304595947266,
|
|
"kl/avg_steps": 0.375,
|
|
"kl/beta": 0.06947022676467896,
|
|
"kl/n_epsilon_steps": 0.3125,
|
|
"kl/p_epsilon_steps": 0.6875,
|
|
"learning_rate": 4.95006729368358e-07,
|
|
"logits/chosen": -4.46284294128418,
|
|
"logits/rejected": -3.8937301635742188,
|
|
"logps/chosen": -103.74415588378906,
|
|
"logps/ref_chosen": -100.18708801269531,
|
|
"logps/ref_rejected": -103.53739929199219,
|
|
"logps/rejected": -116.03982543945312,
|
|
"loss": 0.9893,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -0.24741317331790924,
|
|
"rewards/margins": 0.6145913004875183,
|
|
"rewards/rejected": -0.8620044589042664,
|
|
"step": 109
|
|
},
|
|
{
|
|
"epoch": 0.16152716593245228,
|
|
"grad_norm": 22.954145431518555,
|
|
"kl/avg_steps": 0.3125,
|
|
"kl/beta": 0.06921068578958511,
|
|
"kl/n_epsilon_steps": 0.34375,
|
|
"kl/p_epsilon_steps": 0.65625,
|
|
"learning_rate": 4.947482930773511e-07,
|
|
"logits/chosen": -4.040766716003418,
|
|
"logits/rejected": -3.4033002853393555,
|
|
"logps/chosen": -111.32714080810547,
|
|
"logps/ref_chosen": -106.89572143554688,
|
|
"logps/ref_rejected": -85.81120300292969,
|
|
"logps/rejected": -95.40037536621094,
|
|
"loss": 1.1493,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -0.30703774094581604,
|
|
"rewards/margins": 0.3521472215652466,
|
|
"rewards/rejected": -0.6591849327087402,
|
|
"step": 110
|
|
},
|
|
{
|
|
"epoch": 0.16299559471365638,
|
|
"grad_norm": 17.981557846069336,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.06899508088827133,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.944834074412042e-07,
|
|
"logits/chosen": -4.594253063201904,
|
|
"logits/rejected": -3.7376368045806885,
|
|
"logps/chosen": -97.20205688476562,
|
|
"logps/ref_chosen": -93.3293685913086,
|
|
"logps/ref_rejected": -104.177490234375,
|
|
"logps/rejected": -118.46589660644531,
|
|
"loss": 0.9406,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.2670831084251404,
|
|
"rewards/margins": 0.7102484703063965,
|
|
"rewards/rejected": -0.9773316383361816,
|
|
"step": 111
|
|
},
|
|
{
|
|
"epoch": 0.1644640234948605,
|
|
"grad_norm": 19.202064514160156,
|
|
"kl/avg_steps": 0.1875,
|
|
"kl/beta": 0.06860914826393127,
|
|
"kl/n_epsilon_steps": 0.40625,
|
|
"kl/p_epsilon_steps": 0.59375,
|
|
"learning_rate": 4.942120794399002e-07,
|
|
"logits/chosen": -5.043312072753906,
|
|
"logits/rejected": -4.008922576904297,
|
|
"logps/chosen": -87.14385986328125,
|
|
"logps/ref_chosen": -82.7470474243164,
|
|
"logps/ref_rejected": -71.55845642089844,
|
|
"logps/rejected": -81.11387634277344,
|
|
"loss": 1.1448,
|
|
"rewards/accuracies": 0.71875,
|
|
"rewards/chosen": -0.3032221496105194,
|
|
"rewards/margins": 0.3490729331970215,
|
|
"rewards/rejected": -0.6522951126098633,
|
|
"step": 112
|
|
},
|
|
{
|
|
"epoch": 0.16593245227606462,
|
|
"grad_norm": 22.44894790649414,
|
|
"kl/avg_steps": 0.40625,
|
|
"kl/beta": 0.06848075240850449,
|
|
"kl/n_epsilon_steps": 0.296875,
|
|
"kl/p_epsilon_steps": 0.703125,
|
|
"learning_rate": 4.939343162231841e-07,
|
|
"logits/chosen": -4.6388139724731445,
|
|
"logits/rejected": -3.630887031555176,
|
|
"logps/chosen": -103.29286193847656,
|
|
"logps/ref_chosen": -97.89949798583984,
|
|
"logps/ref_rejected": -84.41960144042969,
|
|
"logps/rejected": -95.4296875,
|
|
"loss": 1.1132,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.37001150846481323,
|
|
"rewards/margins": 0.38021430373191833,
|
|
"rewards/rejected": -0.7502257823944092,
|
|
"step": 113
|
|
},
|
|
{
|
|
"epoch": 0.16740088105726872,
|
|
"grad_norm": 22.647991180419922,
|
|
"kl/avg_steps": 0.34375,
|
|
"kl/beta": 0.06820367276668549,
|
|
"kl/n_epsilon_steps": 0.328125,
|
|
"kl/p_epsilon_steps": 0.671875,
|
|
"learning_rate": 4.936501251103751e-07,
|
|
"logits/chosen": -5.111181259155273,
|
|
"logits/rejected": -3.6170125007629395,
|
|
"logps/chosen": -101.57315063476562,
|
|
"logps/ref_chosen": -96.4951171875,
|
|
"logps/ref_rejected": -92.01536560058594,
|
|
"logps/rejected": -105.01116180419922,
|
|
"loss": 1.0716,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.34747976064682007,
|
|
"rewards/margins": 0.533342719078064,
|
|
"rewards/rejected": -0.880822479724884,
|
|
"step": 114
|
|
},
|
|
{
|
|
"epoch": 0.16886930983847284,
|
|
"grad_norm": 20.238996505737305,
|
|
"kl/avg_steps": 0.4375,
|
|
"kl/beta": 0.06797002255916595,
|
|
"kl/n_epsilon_steps": 0.28125,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 4.933595135901732e-07,
|
|
"logits/chosen": -4.847101211547852,
|
|
"logits/rejected": -3.5432467460632324,
|
|
"logps/chosen": -108.37738800048828,
|
|
"logps/ref_chosen": -103.18013000488281,
|
|
"logps/ref_rejected": -105.71595764160156,
|
|
"logps/rejected": -118.87684631347656,
|
|
"loss": 1.0422,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -0.3536038398742676,
|
|
"rewards/margins": 0.5348291993141174,
|
|
"rewards/rejected": -0.8884330987930298,
|
|
"step": 115
|
|
},
|
|
{
|
|
"epoch": 0.17033773861967694,
|
|
"grad_norm": 17.519254684448242,
|
|
"kl/avg_steps": 0.375,
|
|
"kl/beta": 0.0676739513874054,
|
|
"kl/n_epsilon_steps": 0.3125,
|
|
"kl/p_epsilon_steps": 0.6875,
|
|
"learning_rate": 4.930624893204624e-07,
|
|
"logits/chosen": -5.042938232421875,
|
|
"logits/rejected": -3.4571943283081055,
|
|
"logps/chosen": -85.59718322753906,
|
|
"logps/ref_chosen": -81.15837097167969,
|
|
"logps/ref_rejected": -86.30586242675781,
|
|
"logps/rejected": -97.48081970214844,
|
|
"loss": 1.064,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -0.300454318523407,
|
|
"rewards/margins": 0.45105841755867004,
|
|
"rewards/rejected": -0.7515127062797546,
|
|
"step": 116
|
|
},
|
|
{
|
|
"epoch": 0.17180616740088106,
|
|
"grad_norm": 29.290815353393555,
|
|
"kl/avg_steps": 0.3125,
|
|
"kl/beta": 0.06742112338542938,
|
|
"kl/n_epsilon_steps": 0.34375,
|
|
"kl/p_epsilon_steps": 0.65625,
|
|
"learning_rate": 4.927590601281083e-07,
|
|
"logits/chosen": -4.475282192230225,
|
|
"logits/rejected": -3.8478426933288574,
|
|
"logps/chosen": -102.51126098632812,
|
|
"logps/ref_chosen": -97.0809555053711,
|
|
"logps/ref_rejected": -75.1399154663086,
|
|
"logps/rejected": -85.75031280517578,
|
|
"loss": 1.1723,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -0.36769068241119385,
|
|
"rewards/margins": 0.34401029348373413,
|
|
"rewards/rejected": -0.7117009162902832,
|
|
"step": 117
|
|
},
|
|
{
|
|
"epoch": 0.17327459618208516,
|
|
"grad_norm": 16.853479385375977,
|
|
"kl/avg_steps": 0.484375,
|
|
"kl/beta": 0.0672110915184021,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 4.924492340087524e-07,
|
|
"logits/chosen": -4.661401748657227,
|
|
"logits/rejected": -3.8089747428894043,
|
|
"logps/chosen": -90.43417358398438,
|
|
"logps/ref_chosen": -86.5362319946289,
|
|
"logps/ref_rejected": -82.93563842773438,
|
|
"logps/rejected": -94.443603515625,
|
|
"loss": 1.0216,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.2619098126888275,
|
|
"rewards/margins": 0.506035327911377,
|
|
"rewards/rejected": -0.7679451107978821,
|
|
"step": 118
|
|
},
|
|
{
|
|
"epoch": 0.17474302496328928,
|
|
"grad_norm": 20.742483139038086,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.06688710302114487,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 4.92133019126601e-07,
|
|
"logits/chosen": -5.028365135192871,
|
|
"logits/rejected": -4.106209754943848,
|
|
"logps/chosen": -110.28329467773438,
|
|
"logps/ref_chosen": -104.60908508300781,
|
|
"logps/ref_rejected": -109.1578140258789,
|
|
"logps/rejected": -122.87940979003906,
|
|
"loss": 1.0755,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.38021495938301086,
|
|
"rewards/margins": 0.5309704542160034,
|
|
"rewards/rejected": -0.9111853837966919,
|
|
"step": 119
|
|
},
|
|
{
|
|
"epoch": 0.1762114537444934,
|
|
"grad_norm": 25.90955924987793,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.06655433028936386,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 4.918104238142103e-07,
|
|
"logits/chosen": -4.781673431396484,
|
|
"logits/rejected": -3.775351047515869,
|
|
"logps/chosen": -119.28671264648438,
|
|
"logps/ref_chosen": -114.68742370605469,
|
|
"logps/ref_rejected": -111.18586730957031,
|
|
"logps/rejected": -126.28087615966797,
|
|
"loss": 1.0054,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.30640918016433716,
|
|
"rewards/margins": 0.6896119117736816,
|
|
"rewards/rejected": -0.9960211515426636,
|
|
"step": 120
|
|
},
|
|
{
|
|
"epoch": 0.1776798825256975,
|
|
"grad_norm": 18.882291793823242,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.06616149842739105,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 4.91481456572267e-07,
|
|
"logits/chosen": -5.0012335777282715,
|
|
"logits/rejected": -3.5916976928710938,
|
|
"logps/chosen": -96.17632293701172,
|
|
"logps/ref_chosen": -92.11666107177734,
|
|
"logps/ref_rejected": -114.972900390625,
|
|
"logps/rejected": -128.61309814453125,
|
|
"loss": 0.9533,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.26827573776245117,
|
|
"rewards/margins": 0.6271148920059204,
|
|
"rewards/rejected": -0.8953906297683716,
|
|
"step": 121
|
|
},
|
|
{
|
|
"epoch": 0.17914831130690162,
|
|
"grad_norm": 18.61724281311035,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.0658118724822998,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 4.911461260693638e-07,
|
|
"logits/chosen": -4.714388847351074,
|
|
"logits/rejected": -3.9871459007263184,
|
|
"logps/chosen": -89.19412994384766,
|
|
"logps/ref_chosen": -84.87757110595703,
|
|
"logps/ref_rejected": -111.0829849243164,
|
|
"logps/rejected": -126.41008758544922,
|
|
"loss": 0.9203,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.2842973470687866,
|
|
"rewards/margins": 0.716621994972229,
|
|
"rewards/rejected": -1.0009193420410156,
|
|
"step": 122
|
|
},
|
|
{
|
|
"epoch": 0.18061674008810572,
|
|
"grad_norm": 18.020599365234375,
|
|
"kl/avg_steps": 0.3125,
|
|
"kl/beta": 0.06548444926738739,
|
|
"kl/n_epsilon_steps": 0.34375,
|
|
"kl/p_epsilon_steps": 0.65625,
|
|
"learning_rate": 4.908044411417711e-07,
|
|
"logits/chosen": -4.689582824707031,
|
|
"logits/rejected": -3.9507789611816406,
|
|
"logps/chosen": -103.35702514648438,
|
|
"logps/ref_chosen": -98.47874450683594,
|
|
"logps/ref_rejected": -93.85417175292969,
|
|
"logps/rejected": -106.42095947265625,
|
|
"loss": 1.0567,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -0.3202221989631653,
|
|
"rewards/margins": 0.49699968099594116,
|
|
"rewards/rejected": -0.8172218799591064,
|
|
"step": 123
|
|
},
|
|
{
|
|
"epoch": 0.18208516886930984,
|
|
"grad_norm": 18.883220672607422,
|
|
"kl/avg_steps": 0.375,
|
|
"kl/beta": 0.06528045237064362,
|
|
"kl/n_epsilon_steps": 0.3125,
|
|
"kl/p_epsilon_steps": 0.6875,
|
|
"learning_rate": 4.904564107932048e-07,
|
|
"logits/chosen": -4.445045471191406,
|
|
"logits/rejected": -3.482326030731201,
|
|
"logps/chosen": -107.31734466552734,
|
|
"logps/ref_chosen": -102.76290130615234,
|
|
"logps/ref_rejected": -123.005615234375,
|
|
"logps/rejected": -137.15272521972656,
|
|
"loss": 1.0345,
|
|
"rewards/accuracies": 0.703125,
|
|
"rewards/chosen": -0.29794853925704956,
|
|
"rewards/margins": 0.6185939908027649,
|
|
"rewards/rejected": -0.9165425300598145,
|
|
"step": 124
|
|
},
|
|
{
|
|
"epoch": 0.18355359765051396,
|
|
"grad_norm": 21.531150817871094,
|
|
"kl/avg_steps": 0.28125,
|
|
"kl/beta": 0.06503656506538391,
|
|
"kl/n_epsilon_steps": 0.359375,
|
|
"kl/p_epsilon_steps": 0.640625,
|
|
"learning_rate": 4.90102044194588e-07,
|
|
"logits/chosen": -4.837682247161865,
|
|
"logits/rejected": -4.35373592376709,
|
|
"logps/chosen": -96.40164947509766,
|
|
"logps/ref_chosen": -90.42627716064453,
|
|
"logps/ref_rejected": -89.24901580810547,
|
|
"logps/rejected": -102.02460479736328,
|
|
"loss": 1.1176,
|
|
"rewards/accuracies": 0.65625,
|
|
"rewards/chosen": -0.3898157477378845,
|
|
"rewards/margins": 0.43594902753829956,
|
|
"rewards/rejected": -0.8257647752761841,
|
|
"step": 125
|
|
},
|
|
{
|
|
"epoch": 0.18502202643171806,
|
|
"grad_norm": 18.599184036254883,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.06485415995121002,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 4.897413506838102e-07,
|
|
"logits/chosen": -4.741695404052734,
|
|
"logits/rejected": -3.933988094329834,
|
|
"logps/chosen": -99.16655731201172,
|
|
"logps/ref_chosen": -94.42288208007812,
|
|
"logps/ref_rejected": -103.88156127929688,
|
|
"logps/rejected": -118.57122039794922,
|
|
"loss": 0.94,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -0.30715566873550415,
|
|
"rewards/margins": 0.6382361650466919,
|
|
"rewards/rejected": -0.9453917741775513,
|
|
"step": 126
|
|
},
|
|
{
|
|
"epoch": 0.18649045521292218,
|
|
"grad_norm": 20.271377563476562,
|
|
"kl/avg_steps": 0.28125,
|
|
"kl/beta": 0.06447136402130127,
|
|
"kl/n_epsilon_steps": 0.359375,
|
|
"kl/p_epsilon_steps": 0.640625,
|
|
"learning_rate": 4.89374339765481e-07,
|
|
"logits/chosen": -5.095927715301514,
|
|
"logits/rejected": -3.9779820442199707,
|
|
"logps/chosen": -89.53435516357422,
|
|
"logps/ref_chosen": -84.75105285644531,
|
|
"logps/ref_rejected": -83.54523468017578,
|
|
"logps/rejected": -95.63729858398438,
|
|
"loss": 1.0846,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -0.3097778856754303,
|
|
"rewards/margins": 0.4655241370201111,
|
|
"rewards/rejected": -0.7753020524978638,
|
|
"step": 127
|
|
},
|
|
{
|
|
"epoch": 0.18795888399412627,
|
|
"grad_norm": 22.813669204711914,
|
|
"kl/avg_steps": 0.328125,
|
|
"kl/beta": 0.06429054588079453,
|
|
"kl/n_epsilon_steps": 0.328125,
|
|
"kl/p_epsilon_steps": 0.65625,
|
|
"learning_rate": 4.890010211106795e-07,
|
|
"logits/chosen": -4.714390754699707,
|
|
"logits/rejected": -3.6033260822296143,
|
|
"logps/chosen": -95.26617431640625,
|
|
"logps/ref_chosen": -90.61964416503906,
|
|
"logps/ref_rejected": -83.10347747802734,
|
|
"logps/rejected": -95.3511734008789,
|
|
"loss": 1.1062,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -0.29941076040267944,
|
|
"rewards/margins": 0.482503741979599,
|
|
"rewards/rejected": -0.7819145321846008,
|
|
"step": 128
|
|
},
|
|
{
|
|
"epoch": 0.1894273127753304,
|
|
"grad_norm": 18.353992462158203,
|
|
"kl/avg_steps": 0.25,
|
|
"kl/beta": 0.06408028304576874,
|
|
"kl/n_epsilon_steps": 0.375,
|
|
"kl/p_epsilon_steps": 0.625,
|
|
"learning_rate": 4.88621404556699e-07,
|
|
"logits/chosen": -4.422687530517578,
|
|
"logits/rejected": -3.368746757507324,
|
|
"logps/chosen": -104.02716064453125,
|
|
"logps/ref_chosen": -98.91870880126953,
|
|
"logps/ref_rejected": -103.28419494628906,
|
|
"logps/rejected": -116.28022003173828,
|
|
"loss": 1.0793,
|
|
"rewards/accuracies": 0.6875,
|
|
"rewards/chosen": -0.32816505432128906,
|
|
"rewards/margins": 0.4989137649536133,
|
|
"rewards/rejected": -0.8270788192749023,
|
|
"step": 129
|
|
},
|
|
{
|
|
"epoch": 0.19089574155653452,
|
|
"grad_norm": 16.212928771972656,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.06392048299312592,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.882355001067891e-07,
|
|
"logits/chosen": -5.384788513183594,
|
|
"logits/rejected": -4.525052070617676,
|
|
"logps/chosen": -77.16912841796875,
|
|
"logps/ref_chosen": -73.18368530273438,
|
|
"logps/ref_rejected": -88.02154541015625,
|
|
"logps/rejected": -102.19236755371094,
|
|
"loss": 0.9753,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.2548287510871887,
|
|
"rewards/margins": 0.6435878276824951,
|
|
"rewards/rejected": -0.8984166383743286,
|
|
"step": 130
|
|
},
|
|
{
|
|
"epoch": 0.19236417033773862,
|
|
"grad_norm": 18.157773971557617,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.06356293708086014,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.878433179298909e-07,
|
|
"logits/chosen": -5.078728675842285,
|
|
"logits/rejected": -3.907397747039795,
|
|
"logps/chosen": -77.2987289428711,
|
|
"logps/ref_chosen": -73.9564437866211,
|
|
"logps/ref_rejected": -94.45582580566406,
|
|
"logps/rejected": -107.02981567382812,
|
|
"loss": 1.0071,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -0.21306927502155304,
|
|
"rewards/margins": 0.5797268748283386,
|
|
"rewards/rejected": -0.7927961349487305,
|
|
"step": 131
|
|
},
|
|
{
|
|
"epoch": 0.19383259911894274,
|
|
"grad_norm": 21.45513153076172,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.06320739537477493,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 4.874448683603694e-07,
|
|
"logits/chosen": -5.31199836730957,
|
|
"logits/rejected": -4.291113376617432,
|
|
"logps/chosen": -97.30502319335938,
|
|
"logps/ref_chosen": -92.43948364257812,
|
|
"logps/ref_rejected": -94.13251495361328,
|
|
"logps/rejected": -108.19319152832031,
|
|
"loss": 1.0126,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.3079445958137512,
|
|
"rewards/margins": 0.5743677020072937,
|
|
"rewards/rejected": -0.8823122978210449,
|
|
"step": 132
|
|
},
|
|
{
|
|
"epoch": 0.19530102790014683,
|
|
"grad_norm": 17.602924346923828,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.06283431500196457,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 4.870401618977415e-07,
|
|
"logits/chosen": -5.221474647521973,
|
|
"logits/rejected": -4.042792320251465,
|
|
"logps/chosen": -97.0290298461914,
|
|
"logps/ref_chosen": -93.11666107177734,
|
|
"logps/ref_rejected": -102.09491729736328,
|
|
"logps/rejected": -115.60842895507812,
|
|
"loss": 0.9767,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.24647247791290283,
|
|
"rewards/margins": 0.5967481732368469,
|
|
"rewards/rejected": -0.843220591545105,
|
|
"step": 133
|
|
},
|
|
{
|
|
"epoch": 0.19676945668135096,
|
|
"grad_norm": 16.935972213745117,
|
|
"kl/avg_steps": 0.4375,
|
|
"kl/beta": 0.06252171099185944,
|
|
"kl/n_epsilon_steps": 0.28125,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 4.866292092063986e-07,
|
|
"logits/chosen": -5.074636936187744,
|
|
"logits/rejected": -4.259199142456055,
|
|
"logps/chosen": -87.7016830444336,
|
|
"logps/ref_chosen": -83.55012512207031,
|
|
"logps/ref_rejected": -93.8815689086914,
|
|
"logps/rejected": -108.22308349609375,
|
|
"loss": 0.9528,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.26006680727005005,
|
|
"rewards/margins": 0.6304125785827637,
|
|
"rewards/rejected": -0.8904794454574585,
|
|
"step": 134
|
|
},
|
|
{
|
|
"epoch": 0.19823788546255505,
|
|
"grad_norm": 21.895069122314453,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.06224936991930008,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.862120211153265e-07,
|
|
"logits/chosen": -4.999515533447266,
|
|
"logits/rejected": -3.6168813705444336,
|
|
"logps/chosen": -81.46526336669922,
|
|
"logps/ref_chosen": -77.58262634277344,
|
|
"logps/ref_rejected": -121.05648803710938,
|
|
"logps/rejected": -134.99432373046875,
|
|
"loss": 0.9797,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.24222984910011292,
|
|
"rewards/margins": 0.6191209554672241,
|
|
"rewards/rejected": -0.8613507747650146,
|
|
"step": 135
|
|
},
|
|
{
|
|
"epoch": 0.19970631424375918,
|
|
"grad_norm": 17.361907958984375,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.06190117448568344,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 4.857886086178193e-07,
|
|
"logits/chosen": -5.6198625564575195,
|
|
"logits/rejected": -4.077550888061523,
|
|
"logps/chosen": -98.3825454711914,
|
|
"logps/ref_chosen": -94.1727294921875,
|
|
"logps/ref_rejected": -103.26428985595703,
|
|
"logps/rejected": -117.48695373535156,
|
|
"loss": 0.9543,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.2606146037578583,
|
|
"rewards/margins": 0.613581120967865,
|
|
"rewards/rejected": -0.8741957545280457,
|
|
"step": 136
|
|
},
|
|
{
|
|
"epoch": 0.2011747430249633,
|
|
"grad_norm": 18.447507858276367,
|
|
"kl/avg_steps": 0.8125,
|
|
"kl/beta": 0.06161236763000488,
|
|
"kl/n_epsilon_steps": 0.09375,
|
|
"kl/p_epsilon_steps": 0.90625,
|
|
"learning_rate": 4.853589828711902e-07,
|
|
"logits/chosen": -5.679508686065674,
|
|
"logits/rejected": -4.036385536193848,
|
|
"logps/chosen": -81.5108642578125,
|
|
"logps/ref_chosen": -78.11874389648438,
|
|
"logps/ref_rejected": -117.92237091064453,
|
|
"logps/rejected": -136.72076416015625,
|
|
"loss": 0.8065,
|
|
"rewards/accuracies": 0.921875,
|
|
"rewards/chosen": -0.20868739485740662,
|
|
"rewards/margins": 0.9392319321632385,
|
|
"rewards/rejected": -1.1479192972183228,
|
|
"step": 137
|
|
},
|
|
{
|
|
"epoch": 0.2026431718061674,
|
|
"grad_norm": 16.60499382019043,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.061115801334381104,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.849231551964771e-07,
|
|
"logits/chosen": -4.7099151611328125,
|
|
"logits/rejected": -4.268791198730469,
|
|
"logps/chosen": -98.61029815673828,
|
|
"logps/ref_chosen": -92.82994079589844,
|
|
"logps/ref_rejected": -98.64839935302734,
|
|
"logps/rejected": -113.98179626464844,
|
|
"loss": 0.992,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.35308200120925903,
|
|
"rewards/margins": 0.5774024724960327,
|
|
"rewards/rejected": -0.9304844737052917,
|
|
"step": 138
|
|
},
|
|
{
|
|
"epoch": 0.20411160058737152,
|
|
"grad_norm": 16.9788875579834,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.06077394634485245,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 4.844811370781446e-07,
|
|
"logits/chosen": -5.217746734619141,
|
|
"logits/rejected": -4.438928127288818,
|
|
"logps/chosen": -82.84307861328125,
|
|
"logps/ref_chosen": -79.00301361083984,
|
|
"logps/ref_rejected": -85.82733154296875,
|
|
"logps/rejected": -100.77876281738281,
|
|
"loss": 0.9595,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -0.23357324302196503,
|
|
"rewards/margins": 0.6675229072570801,
|
|
"rewards/rejected": -0.9010961055755615,
|
|
"step": 139
|
|
},
|
|
{
|
|
"epoch": 0.2055800293685756,
|
|
"grad_norm": 15.984687805175781,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.06037771701812744,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 4.840329401637809e-07,
|
|
"logits/chosen": -5.573225975036621,
|
|
"logits/rejected": -4.482275485992432,
|
|
"logps/chosen": -92.46583557128906,
|
|
"logps/ref_chosen": -88.35244750976562,
|
|
"logps/ref_rejected": -90.57477569580078,
|
|
"logps/rejected": -104.81709289550781,
|
|
"loss": 0.9655,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.24829351902008057,
|
|
"rewards/margins": 0.6051296591758728,
|
|
"rewards/rejected": -0.8534231781959534,
|
|
"step": 140
|
|
},
|
|
{
|
|
"epoch": 0.20704845814977973,
|
|
"grad_norm": 22.27216911315918,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.06007733196020126,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 4.83578576263792e-07,
|
|
"logits/chosen": -5.062009811401367,
|
|
"logits/rejected": -4.248960018157959,
|
|
"logps/chosen": -112.61961364746094,
|
|
"logps/ref_chosen": -107.22016906738281,
|
|
"logps/ref_rejected": -104.71223449707031,
|
|
"logps/rejected": -121.64073181152344,
|
|
"loss": 0.9503,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -0.3238122761249542,
|
|
"rewards/margins": 0.6848909854888916,
|
|
"rewards/rejected": -1.0087032318115234,
|
|
"step": 141
|
|
},
|
|
{
|
|
"epoch": 0.20851688693098386,
|
|
"grad_norm": 16.56783676147461,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.05968564376235008,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.83118057351089e-07,
|
|
"logits/chosen": -5.5315375328063965,
|
|
"logits/rejected": -4.753322601318359,
|
|
"logps/chosen": -94.61272430419922,
|
|
"logps/ref_chosen": -88.72894287109375,
|
|
"logps/ref_rejected": -99.73038482666016,
|
|
"logps/rejected": -116.76524353027344,
|
|
"loss": 0.9498,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.34996697306632996,
|
|
"rewards/margins": 0.6583119630813599,
|
|
"rewards/rejected": -1.0082789659500122,
|
|
"step": 142
|
|
},
|
|
{
|
|
"epoch": 0.20998531571218795,
|
|
"grad_norm": 16.937776565551758,
|
|
"kl/avg_steps": 0.375,
|
|
"kl/beta": 0.05935179069638252,
|
|
"kl/n_epsilon_steps": 0.3125,
|
|
"kl/p_epsilon_steps": 0.6875,
|
|
"learning_rate": 4.826513955607734e-07,
|
|
"logits/chosen": -5.914580345153809,
|
|
"logits/rejected": -4.424531936645508,
|
|
"logps/chosen": -89.47615051269531,
|
|
"logps/ref_chosen": -85.50457763671875,
|
|
"logps/ref_rejected": -85.8056640625,
|
|
"logps/rejected": -99.1048583984375,
|
|
"loss": 1.0174,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.2356262505054474,
|
|
"rewards/margins": 0.547943115234375,
|
|
"rewards/rejected": -0.7835693359375,
|
|
"step": 143
|
|
},
|
|
{
|
|
"epoch": 0.21145374449339208,
|
|
"grad_norm": 22.844528198242188,
|
|
"kl/avg_steps": 0.4375,
|
|
"kl/beta": 0.05913005396723747,
|
|
"kl/n_epsilon_steps": 0.28125,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 4.821786031898176e-07,
|
|
"logits/chosen": -5.4832234382629395,
|
|
"logits/rejected": -4.616265296936035,
|
|
"logps/chosen": -109.42718505859375,
|
|
"logps/ref_chosen": -104.68708801269531,
|
|
"logps/ref_rejected": -87.71908569335938,
|
|
"logps/rejected": -101.95695495605469,
|
|
"loss": 1.0494,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.2811855673789978,
|
|
"rewards/margins": 0.5548486113548279,
|
|
"rewards/rejected": -0.8360341787338257,
|
|
"step": 144
|
|
},
|
|
{
|
|
"epoch": 0.21292217327459617,
|
|
"grad_norm": 17.371044158935547,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.058872487396001816,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 4.816996926967401e-07,
|
|
"logits/chosen": -5.808310031890869,
|
|
"logits/rejected": -4.368202209472656,
|
|
"logps/chosen": -92.62870788574219,
|
|
"logps/ref_chosen": -88.51988983154297,
|
|
"logps/ref_rejected": -83.76513671875,
|
|
"logps/rejected": -99.01193237304688,
|
|
"loss": 0.948,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.24210438132286072,
|
|
"rewards/margins": 0.6486882567405701,
|
|
"rewards/rejected": -0.8907926678657532,
|
|
"step": 145
|
|
},
|
|
{
|
|
"epoch": 0.2143906020558003,
|
|
"grad_norm": 17.89594841003418,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.05857958644628525,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 4.812146767012779e-07,
|
|
"logits/chosen": -5.770585060119629,
|
|
"logits/rejected": -4.961735725402832,
|
|
"logps/chosen": -107.27189636230469,
|
|
"logps/ref_chosen": -102.9058609008789,
|
|
"logps/ref_rejected": -88.9493637084961,
|
|
"logps/rejected": -105.66189575195312,
|
|
"loss": 0.9083,
|
|
"rewards/accuracies": 0.890625,
|
|
"rewards/chosen": -0.2555050253868103,
|
|
"rewards/margins": 0.715686559677124,
|
|
"rewards/rejected": -0.9711916446685791,
|
|
"step": 146
|
|
},
|
|
{
|
|
"epoch": 0.21585903083700442,
|
|
"grad_norm": 16.04936408996582,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.058233823627233505,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.807235679840536e-07,
|
|
"logits/chosen": -6.048250198364258,
|
|
"logits/rejected": -5.216745853424072,
|
|
"logps/chosen": -93.50482177734375,
|
|
"logps/ref_chosen": -89.26219177246094,
|
|
"logps/ref_rejected": -76.50011444091797,
|
|
"logps/rejected": -93.5650405883789,
|
|
"loss": 0.8803,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.2465764731168747,
|
|
"rewards/margins": 0.7393079996109009,
|
|
"rewards/rejected": -0.9858845472335815,
|
|
"step": 147
|
|
},
|
|
{
|
|
"epoch": 0.2173274596182085,
|
|
"grad_norm": 17.10126304626465,
|
|
"kl/avg_steps": 0.375,
|
|
"kl/beta": 0.05790809169411659,
|
|
"kl/n_epsilon_steps": 0.3125,
|
|
"kl/p_epsilon_steps": 0.6875,
|
|
"learning_rate": 4.802263794862384e-07,
|
|
"logits/chosen": -5.5003767013549805,
|
|
"logits/rejected": -3.983274459838867,
|
|
"logps/chosen": -94.35877990722656,
|
|
"logps/ref_chosen": -91.76287841796875,
|
|
"logps/ref_rejected": -110.38552856445312,
|
|
"logps/rejected": -124.3576889038086,
|
|
"loss": 0.962,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.15071691572666168,
|
|
"rewards/margins": 0.6519370079040527,
|
|
"rewards/rejected": -0.8026539087295532,
|
|
"step": 148
|
|
},
|
|
{
|
|
"epoch": 0.21879588839941264,
|
|
"grad_norm": 17.85907745361328,
|
|
"kl/avg_steps": 0.4375,
|
|
"kl/beta": 0.057691749185323715,
|
|
"kl/n_epsilon_steps": 0.28125,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 4.797231243092118e-07,
|
|
"logits/chosen": -5.346034049987793,
|
|
"logits/rejected": -4.409138202667236,
|
|
"logps/chosen": -98.8444595336914,
|
|
"logps/ref_chosen": -93.4840087890625,
|
|
"logps/ref_rejected": -105.64369201660156,
|
|
"logps/rejected": -121.99969482421875,
|
|
"loss": 0.9992,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.3102826178073883,
|
|
"rewards/margins": 0.6273058652877808,
|
|
"rewards/rejected": -0.9375884532928467,
|
|
"step": 149
|
|
},
|
|
{
|
|
"epoch": 0.22026431718061673,
|
|
"grad_norm": 15.73752498626709,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.05744044482707977,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.792138157142157e-07,
|
|
"logits/chosen": -5.399765968322754,
|
|
"logits/rejected": -4.2668867111206055,
|
|
"logps/chosen": -79.96965026855469,
|
|
"logps/ref_chosen": -75.7786865234375,
|
|
"logps/ref_rejected": -90.02676391601562,
|
|
"logps/rejected": -105.93204498291016,
|
|
"loss": 0.9614,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -0.24165549874305725,
|
|
"rewards/margins": 0.6654115915298462,
|
|
"rewards/rejected": -0.907067060470581,
|
|
"step": 150
|
|
},
|
|
{
|
|
"epoch": 0.22173274596182085,
|
|
"grad_norm": 19.85856056213379,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.05711914971470833,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 4.786984671220053e-07,
|
|
"logits/chosen": -5.871485710144043,
|
|
"logits/rejected": -4.4503984451293945,
|
|
"logps/chosen": -108.49866485595703,
|
|
"logps/ref_chosen": -102.73422241210938,
|
|
"logps/ref_rejected": -106.03781127929688,
|
|
"logps/rejected": -123.03112030029297,
|
|
"loss": 0.9827,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.3297105133533478,
|
|
"rewards/margins": 0.633903980255127,
|
|
"rewards/rejected": -0.9636145234107971,
|
|
"step": 151
|
|
},
|
|
{
|
|
"epoch": 0.22320117474302498,
|
|
"grad_norm": 16.81842803955078,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.056834977120161057,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 4.78177092112495e-07,
|
|
"logits/chosen": -4.908111572265625,
|
|
"logits/rejected": -3.836758613586426,
|
|
"logps/chosen": -100.57263946533203,
|
|
"logps/ref_chosen": -97.59074401855469,
|
|
"logps/ref_rejected": -110.49913024902344,
|
|
"logps/rejected": -126.90365600585938,
|
|
"loss": 0.8941,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.1701817512512207,
|
|
"rewards/margins": 0.7548890113830566,
|
|
"rewards/rejected": -0.9250707626342773,
|
|
"step": 152
|
|
},
|
|
{
|
|
"epoch": 0.22466960352422907,
|
|
"grad_norm": 17.356094360351562,
|
|
"kl/avg_steps": 0.4375,
|
|
"kl/beta": 0.05649951100349426,
|
|
"kl/n_epsilon_steps": 0.28125,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 4.776497044244016e-07,
|
|
"logits/chosen": -5.573543548583984,
|
|
"logits/rejected": -4.830375671386719,
|
|
"logps/chosen": -92.43611907958984,
|
|
"logps/ref_chosen": -87.02560424804688,
|
|
"logps/ref_rejected": -91.06853485107422,
|
|
"logps/rejected": -106.13377380371094,
|
|
"loss": 1.0443,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -0.3063035011291504,
|
|
"rewards/margins": 0.5388628244400024,
|
|
"rewards/rejected": -0.8451663255691528,
|
|
"step": 153
|
|
},
|
|
{
|
|
"epoch": 0.2261380323054332,
|
|
"grad_norm": 15.954586029052734,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.056253399699926376,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 4.771163179548808e-07,
|
|
"logits/chosen": -5.686001300811768,
|
|
"logits/rejected": -4.457275390625,
|
|
"logps/chosen": -94.85427856445312,
|
|
"logps/ref_chosen": -91.10566711425781,
|
|
"logps/ref_rejected": -111.37238311767578,
|
|
"logps/rejected": -130.63352966308594,
|
|
"loss": 0.8199,
|
|
"rewards/accuracies": 0.90625,
|
|
"rewards/chosen": -0.21082431077957153,
|
|
"rewards/margins": 0.8645293116569519,
|
|
"rewards/rejected": -1.0753536224365234,
|
|
"step": 154
|
|
},
|
|
{
|
|
"epoch": 0.2276064610866373,
|
|
"grad_norm": 23.60336685180664,
|
|
"kl/avg_steps": 0.4375,
|
|
"kl/beta": 0.05599094182252884,
|
|
"kl/n_epsilon_steps": 0.28125,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 4.7657694675916247e-07,
|
|
"logits/chosen": -5.547731399536133,
|
|
"logits/rejected": -5.115335464477539,
|
|
"logps/chosen": -100.05500793457031,
|
|
"logps/ref_chosen": -94.56097412109375,
|
|
"logps/ref_rejected": -84.77359008789062,
|
|
"logps/rejected": -102.24840545654297,
|
|
"loss": 1.0092,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.3092682957649231,
|
|
"rewards/margins": 0.6626745462417603,
|
|
"rewards/rejected": -0.9719428420066833,
|
|
"step": 155
|
|
},
|
|
{
|
|
"epoch": 0.2290748898678414,
|
|
"grad_norm": 19.952051162719727,
|
|
"kl/avg_steps": 0.359375,
|
|
"kl/beta": 0.055747050791978836,
|
|
"kl/n_epsilon_steps": 0.3125,
|
|
"kl/p_epsilon_steps": 0.671875,
|
|
"learning_rate": 4.7603160505017893e-07,
|
|
"logits/chosen": -5.7792768478393555,
|
|
"logits/rejected": -4.890432357788086,
|
|
"logps/chosen": -101.51638793945312,
|
|
"logps/ref_chosen": -94.65538787841797,
|
|
"logps/ref_rejected": -84.41934204101562,
|
|
"logps/rejected": -101.98014831542969,
|
|
"loss": 1.0161,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -0.38445067405700684,
|
|
"rewards/margins": 0.5892693400382996,
|
|
"rewards/rejected": -0.9737200736999512,
|
|
"step": 156
|
|
},
|
|
{
|
|
"epoch": 0.2305433186490455,
|
|
"grad_norm": 17.198551177978516,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.055547427386045456,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 4.7548030719819154e-07,
|
|
"logits/chosen": -6.013505458831787,
|
|
"logits/rejected": -4.734508037567139,
|
|
"logps/chosen": -103.44956970214844,
|
|
"logps/ref_chosen": -97.01239013671875,
|
|
"logps/ref_rejected": -111.70133972167969,
|
|
"logps/rejected": -132.464111328125,
|
|
"loss": 0.8564,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -0.35640445351600647,
|
|
"rewards/margins": 0.7874443531036377,
|
|
"rewards/rejected": -1.1438487768173218,
|
|
"step": 157
|
|
},
|
|
{
|
|
"epoch": 0.23201174743024963,
|
|
"grad_norm": 16.71400260925293,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.05516814440488815,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 4.7492306773041136e-07,
|
|
"logits/chosen": -5.91513204574585,
|
|
"logits/rejected": -4.732603073120117,
|
|
"logps/chosen": -100.4566650390625,
|
|
"logps/ref_chosen": -94.71295928955078,
|
|
"logps/ref_rejected": -120.93733978271484,
|
|
"logps/rejected": -141.5926971435547,
|
|
"loss": 0.8512,
|
|
"rewards/accuracies": 0.90625,
|
|
"rewards/chosen": -0.316554993391037,
|
|
"rewards/margins": 0.814699113368988,
|
|
"rewards/rejected": -1.1312541961669922,
|
|
"step": 158
|
|
},
|
|
{
|
|
"epoch": 0.23348017621145375,
|
|
"grad_norm": 15.837252616882324,
|
|
"kl/avg_steps": 0.484375,
|
|
"kl/beta": 0.0548425167798996,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 4.743599013306165e-07,
|
|
"logits/chosen": -6.067984104156494,
|
|
"logits/rejected": -5.724274158477783,
|
|
"logps/chosen": -113.19328308105469,
|
|
"logps/ref_chosen": -108.2182388305664,
|
|
"logps/ref_rejected": -94.83901977539062,
|
|
"logps/rejected": -113.71253204345703,
|
|
"loss": 0.9148,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.2735292613506317,
|
|
"rewards/margins": 0.7541267275810242,
|
|
"rewards/rejected": -1.0276559591293335,
|
|
"step": 159
|
|
},
|
|
{
|
|
"epoch": 0.23494860499265785,
|
|
"grad_norm": 15.891210556030273,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.054578155279159546,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 4.737908228387656e-07,
|
|
"logits/chosen": -6.017996788024902,
|
|
"logits/rejected": -4.622193336486816,
|
|
"logps/chosen": -102.32695770263672,
|
|
"logps/ref_chosen": -98.16340637207031,
|
|
"logps/ref_rejected": -102.2809829711914,
|
|
"logps/rejected": -119.59290313720703,
|
|
"loss": 0.9092,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.22743016481399536,
|
|
"rewards/margins": 0.7104052305221558,
|
|
"rewards/rejected": -0.9378353953361511,
|
|
"step": 160
|
|
},
|
|
{
|
|
"epoch": 0.23641703377386197,
|
|
"grad_norm": 17.31453514099121,
|
|
"kl/avg_steps": 0.4375,
|
|
"kl/beta": 0.054289739578962326,
|
|
"kl/n_epsilon_steps": 0.28125,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 4.7321584725060594e-07,
|
|
"logits/chosen": -6.470645904541016,
|
|
"logits/rejected": -5.285548210144043,
|
|
"logps/chosen": -92.63033294677734,
|
|
"logps/ref_chosen": -86.59187316894531,
|
|
"logps/ref_rejected": -89.01643371582031,
|
|
"logps/rejected": -107.30801391601562,
|
|
"loss": 0.9714,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -0.32883331179618835,
|
|
"rewards/margins": 0.6572731733322144,
|
|
"rewards/rejected": -0.9861065149307251,
|
|
"step": 161
|
|
},
|
|
{
|
|
"epoch": 0.23788546255506607,
|
|
"grad_norm": 15.398876190185547,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.054053258150815964,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.7263498971727905e-07,
|
|
"logits/chosen": -5.194693088531494,
|
|
"logits/rejected": -4.573456764221191,
|
|
"logps/chosen": -102.532470703125,
|
|
"logps/ref_chosen": -97.40288543701172,
|
|
"logps/ref_rejected": -95.86923217773438,
|
|
"logps/rejected": -114.22633361816406,
|
|
"loss": 0.9167,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -0.278072714805603,
|
|
"rewards/margins": 0.7081333994865417,
|
|
"rewards/rejected": -0.9862060546875,
|
|
"step": 162
|
|
},
|
|
{
|
|
"epoch": 0.2393538913362702,
|
|
"grad_norm": 13.848814964294434,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.053750909864902496,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.720482655449212e-07,
|
|
"logits/chosen": -6.277469635009766,
|
|
"logits/rejected": -5.564417839050293,
|
|
"logps/chosen": -87.13270568847656,
|
|
"logps/ref_chosen": -82.1753921508789,
|
|
"logps/ref_rejected": -82.86473846435547,
|
|
"logps/rejected": -103.97894287109375,
|
|
"loss": 0.8482,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -0.26623615622520447,
|
|
"rewards/margins": 0.8599917888641357,
|
|
"rewards/rejected": -1.1262279748916626,
|
|
"step": 163
|
|
},
|
|
{
|
|
"epoch": 0.24082232011747431,
|
|
"grad_norm": 17.024089813232422,
|
|
"kl/avg_steps": 0.3125,
|
|
"kl/beta": 0.05345025286078453,
|
|
"kl/n_epsilon_steps": 0.34375,
|
|
"kl/p_epsilon_steps": 0.65625,
|
|
"learning_rate": 4.714556901942599e-07,
|
|
"logits/chosen": -5.856521129608154,
|
|
"logits/rejected": -4.963005542755127,
|
|
"logps/chosen": -98.60881805419922,
|
|
"logps/ref_chosen": -90.4742660522461,
|
|
"logps/ref_rejected": -87.02809143066406,
|
|
"logps/rejected": -106.09263610839844,
|
|
"loss": 1.0038,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.4347628355026245,
|
|
"rewards/margins": 0.5785181522369385,
|
|
"rewards/rejected": -1.013280987739563,
|
|
"step": 164
|
|
},
|
|
{
|
|
"epoch": 0.2422907488986784,
|
|
"grad_norm": 14.025198936462402,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.05328373983502388,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 4.708572792802069e-07,
|
|
"logits/chosen": -5.403741359710693,
|
|
"logits/rejected": -5.085051536560059,
|
|
"logps/chosen": -92.47659301757812,
|
|
"logps/ref_chosen": -87.313232421875,
|
|
"logps/ref_rejected": -81.902587890625,
|
|
"logps/rejected": -102.87800598144531,
|
|
"loss": 0.857,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -0.27433332800865173,
|
|
"rewards/margins": 0.8343935012817383,
|
|
"rewards/rejected": -1.1087267398834229,
|
|
"step": 165
|
|
},
|
|
{
|
|
"epoch": 0.24375917767988253,
|
|
"grad_norm": 15.63122844696045,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.052952785044908524,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 4.702530485714461e-07,
|
|
"logits/chosen": -6.5487165451049805,
|
|
"logits/rejected": -5.4730329513549805,
|
|
"logps/chosen": -90.94509887695312,
|
|
"logps/ref_chosen": -81.5860595703125,
|
|
"logps/ref_rejected": -103.16517639160156,
|
|
"logps/rejected": -127.57855224609375,
|
|
"loss": 0.9066,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.49418699741363525,
|
|
"rewards/margins": 0.7889156341552734,
|
|
"rewards/rejected": -1.2831027507781982,
|
|
"step": 166
|
|
},
|
|
{
|
|
"epoch": 0.24522760646108663,
|
|
"grad_norm": 14.752199172973633,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.05264023318886757,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 4.6964301399001877e-07,
|
|
"logits/chosen": -6.093404769897461,
|
|
"logits/rejected": -5.437514305114746,
|
|
"logps/chosen": -91.57254028320312,
|
|
"logps/ref_chosen": -83.56167602539062,
|
|
"logps/ref_rejected": -99.05305480957031,
|
|
"logps/rejected": -123.61196899414062,
|
|
"loss": 0.8129,
|
|
"rewards/accuracies": 0.890625,
|
|
"rewards/chosen": -0.4190889000892639,
|
|
"rewards/margins": 0.8631209135055542,
|
|
"rewards/rejected": -1.282209873199463,
|
|
"step": 167
|
|
},
|
|
{
|
|
"epoch": 0.24669603524229075,
|
|
"grad_norm": 16.830984115600586,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.05232952535152435,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.690271916109034e-07,
|
|
"logits/chosen": -6.080756664276123,
|
|
"logits/rejected": -5.517510414123535,
|
|
"logps/chosen": -87.75483703613281,
|
|
"logps/ref_chosen": -79.93124389648438,
|
|
"logps/ref_rejected": -80.10216522216797,
|
|
"logps/rejected": -102.02044677734375,
|
|
"loss": 0.9032,
|
|
"rewards/accuracies": 0.890625,
|
|
"rewards/chosen": -0.4084423780441284,
|
|
"rewards/margins": 0.7297256588935852,
|
|
"rewards/rejected": -1.1381680965423584,
|
|
"step": 168
|
|
},
|
|
{
|
|
"epoch": 0.24816446402349487,
|
|
"grad_norm": 15.615654945373535,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.052036818116903305,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 4.6840559766159235e-07,
|
|
"logits/chosen": -5.397192001342773,
|
|
"logits/rejected": -5.175437927246094,
|
|
"logps/chosen": -89.30410766601562,
|
|
"logps/ref_chosen": -82.90370178222656,
|
|
"logps/ref_rejected": -90.12191009521484,
|
|
"logps/rejected": -112.40554809570312,
|
|
"loss": 0.8642,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.33170753717422485,
|
|
"rewards/margins": 0.8178310394287109,
|
|
"rewards/rejected": -1.149538516998291,
|
|
"step": 169
|
|
},
|
|
{
|
|
"epoch": 0.24963289280469897,
|
|
"grad_norm": 16.472681045532227,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.05171360820531845,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.6777824852166437e-07,
|
|
"logits/chosen": -5.991590976715088,
|
|
"logits/rejected": -5.608396053314209,
|
|
"logps/chosen": -83.11984252929688,
|
|
"logps/ref_chosen": -74.98197174072266,
|
|
"logps/ref_rejected": -82.16560363769531,
|
|
"logps/rejected": -104.49899291992188,
|
|
"loss": 0.8871,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.4205918610095978,
|
|
"rewards/margins": 0.7263666987419128,
|
|
"rewards/rejected": -1.146958589553833,
|
|
"step": 170
|
|
},
|
|
{
|
|
"epoch": 0.2511013215859031,
|
|
"grad_norm": 19.757944107055664,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.05142434686422348,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 4.6714516072235273e-07,
|
|
"logits/chosen": -5.451948165893555,
|
|
"logits/rejected": -4.568760871887207,
|
|
"logps/chosen": -128.7462158203125,
|
|
"logps/ref_chosen": -118.97853088378906,
|
|
"logps/ref_rejected": -115.7693099975586,
|
|
"logps/rejected": -141.45156860351562,
|
|
"loss": 0.9253,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.5009539127349854,
|
|
"rewards/margins": 0.8095153570175171,
|
|
"rewards/rejected": -1.3104692697525024,
|
|
"step": 171
|
|
},
|
|
{
|
|
"epoch": 0.2525697503671072,
|
|
"grad_norm": 14.202917098999023,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.05118441954255104,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 4.6650635094610966e-07,
|
|
"logits/chosen": -6.403740882873535,
|
|
"logits/rejected": -5.487905025482178,
|
|
"logps/chosen": -111.73493957519531,
|
|
"logps/ref_chosen": -101.15379333496094,
|
|
"logps/ref_rejected": -91.00567626953125,
|
|
"logps/rejected": -116.28710174560547,
|
|
"loss": 0.8464,
|
|
"rewards/accuracies": 0.890625,
|
|
"rewards/chosen": -0.5407711267471313,
|
|
"rewards/margins": 0.7465265989303589,
|
|
"rewards/rejected": -1.2872977256774902,
|
|
"step": 172
|
|
},
|
|
{
|
|
"epoch": 0.2540381791483113,
|
|
"grad_norm": 14.85136604309082,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.05094561353325844,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 4.6586183602616687e-07,
|
|
"logits/chosen": -5.471677780151367,
|
|
"logits/rejected": -5.009877681732178,
|
|
"logps/chosen": -109.14393615722656,
|
|
"logps/ref_chosen": -99.05668640136719,
|
|
"logps/ref_rejected": -83.8507308959961,
|
|
"logps/rejected": -106.2504653930664,
|
|
"loss": 0.9611,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.5136967897415161,
|
|
"rewards/margins": 0.620842456817627,
|
|
"rewards/rejected": -1.1345391273498535,
|
|
"step": 173
|
|
},
|
|
{
|
|
"epoch": 0.2555066079295154,
|
|
"grad_norm": 16.939958572387695,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.050707921385765076,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 4.652116329460919e-07,
|
|
"logits/chosen": -6.058362007141113,
|
|
"logits/rejected": -5.3862104415893555,
|
|
"logps/chosen": -90.09979248046875,
|
|
"logps/ref_chosen": -80.40890502929688,
|
|
"logps/ref_rejected": -108.20171356201172,
|
|
"logps/rejected": -134.08053588867188,
|
|
"loss": 0.8609,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -0.49034202098846436,
|
|
"rewards/margins": 0.8122279644012451,
|
|
"rewards/rejected": -1.3025699853897095,
|
|
"step": 174
|
|
},
|
|
{
|
|
"epoch": 0.25697503671071953,
|
|
"grad_norm": 16.927989959716797,
|
|
"kl/avg_steps": 0.78125,
|
|
"kl/beta": 0.050392962992191315,
|
|
"kl/n_epsilon_steps": 0.109375,
|
|
"kl/p_epsilon_steps": 0.890625,
|
|
"learning_rate": 4.645557588393406e-07,
|
|
"logits/chosen": -6.3116278648376465,
|
|
"logits/rejected": -5.949856758117676,
|
|
"logps/chosen": -95.58826446533203,
|
|
"logps/ref_chosen": -82.71649169921875,
|
|
"logps/ref_rejected": -96.66712951660156,
|
|
"logps/rejected": -125.78546142578125,
|
|
"loss": 0.8423,
|
|
"rewards/accuracies": 0.890625,
|
|
"rewards/chosen": -0.6450769305229187,
|
|
"rewards/margins": 0.8101252317428589,
|
|
"rewards/rejected": -1.4552021026611328,
|
|
"step": 175
|
|
},
|
|
{
|
|
"epoch": 0.25844346549192365,
|
|
"grad_norm": 15.526463508605957,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.050002321600914,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 4.638942309888058e-07,
|
|
"logits/chosen": -6.382523059844971,
|
|
"logits/rejected": -5.265742778778076,
|
|
"logps/chosen": -90.09957885742188,
|
|
"logps/ref_chosen": -79.62930297851562,
|
|
"logps/ref_rejected": -102.6885986328125,
|
|
"logps/rejected": -129.64431762695312,
|
|
"loss": 0.9033,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -0.5223788619041443,
|
|
"rewards/margins": 0.8148936629295349,
|
|
"rewards/rejected": -1.3372724056243896,
|
|
"step": 176
|
|
},
|
|
{
|
|
"epoch": 0.2599118942731278,
|
|
"grad_norm": 14.051129341125488,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.0497380867600441,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 4.6322706682636137e-07,
|
|
"logits/chosen": -5.991172790527344,
|
|
"logits/rejected": -5.463325500488281,
|
|
"logps/chosen": -103.93224334716797,
|
|
"logps/ref_chosen": -92.56109619140625,
|
|
"logps/ref_rejected": -103.33838653564453,
|
|
"logps/rejected": -133.99789428710938,
|
|
"loss": 0.807,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -0.5626887083053589,
|
|
"rewards/margins": 0.9501281380653381,
|
|
"rewards/rejected": -1.5128169059753418,
|
|
"step": 177
|
|
},
|
|
{
|
|
"epoch": 0.26138032305433184,
|
|
"grad_norm": 16.428043365478516,
|
|
"kl/avg_steps": 0.75,
|
|
"kl/beta": 0.049413807690143585,
|
|
"kl/n_epsilon_steps": 0.125,
|
|
"kl/p_epsilon_steps": 0.875,
|
|
"learning_rate": 4.6255428393240354e-07,
|
|
"logits/chosen": -6.9524335861206055,
|
|
"logits/rejected": -6.058289527893066,
|
|
"logps/chosen": -94.86000061035156,
|
|
"logps/ref_chosen": -84.26579284667969,
|
|
"logps/ref_rejected": -110.79730987548828,
|
|
"logps/rejected": -143.80020141601562,
|
|
"loss": 0.7237,
|
|
"rewards/accuracies": 0.9375,
|
|
"rewards/chosen": -0.5214236974716187,
|
|
"rewards/margins": 1.096163272857666,
|
|
"rewards/rejected": -1.6175870895385742,
|
|
"step": 178
|
|
},
|
|
{
|
|
"epoch": 0.26284875183553597,
|
|
"grad_norm": 18.81154441833496,
|
|
"kl/avg_steps": 0.4375,
|
|
"kl/beta": 0.04904596507549286,
|
|
"kl/n_epsilon_steps": 0.28125,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 4.6187590003538724e-07,
|
|
"logits/chosen": -6.702276229858398,
|
|
"logits/rejected": -5.42424201965332,
|
|
"logps/chosen": -106.93254089355469,
|
|
"logps/ref_chosen": -93.4479751586914,
|
|
"logps/ref_rejected": -94.25540924072266,
|
|
"logps/rejected": -122.918212890625,
|
|
"loss": 0.9762,
|
|
"rewards/accuracies": 0.734375,
|
|
"rewards/chosen": -0.6620521545410156,
|
|
"rewards/margins": 0.735353946685791,
|
|
"rewards/rejected": -1.3974061012268066,
|
|
"step": 179
|
|
},
|
|
{
|
|
"epoch": 0.2643171806167401,
|
|
"grad_norm": 19.117578506469727,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.04883232340216637,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 4.611919330113591e-07,
|
|
"logits/chosen": -5.92899751663208,
|
|
"logits/rejected": -5.633153915405273,
|
|
"logps/chosen": -100.06759643554688,
|
|
"logps/ref_chosen": -88.83554077148438,
|
|
"logps/ref_rejected": -103.90696716308594,
|
|
"logps/rejected": -133.78744506835938,
|
|
"loss": 0.8081,
|
|
"rewards/accuracies": 0.890625,
|
|
"rewards/chosen": -0.5459901094436646,
|
|
"rewards/margins": 0.901505172252655,
|
|
"rewards/rejected": -1.4474952220916748,
|
|
"step": 180
|
|
},
|
|
{
|
|
"epoch": 0.2657856093979442,
|
|
"grad_norm": 16.675270080566406,
|
|
"kl/avg_steps": 0.453125,
|
|
"kl/beta": 0.04849889129400253,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 4.605024008834863e-07,
|
|
"logits/chosen": -6.083570957183838,
|
|
"logits/rejected": -5.588318824768066,
|
|
"logps/chosen": -91.0196762084961,
|
|
"logps/ref_chosen": -79.23396301269531,
|
|
"logps/ref_rejected": -69.17442321777344,
|
|
"logps/rejected": -95.4429931640625,
|
|
"loss": 0.9401,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.5707411766052246,
|
|
"rewards/margins": 0.6956329345703125,
|
|
"rewards/rejected": -1.266374111175537,
|
|
"step": 181
|
|
},
|
|
{
|
|
"epoch": 0.26725403817914833,
|
|
"grad_norm": 15.468544006347656,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.04828012362122536,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.598073218215817e-07,
|
|
"logits/chosen": -6.867135047912598,
|
|
"logits/rejected": -6.089150428771973,
|
|
"logps/chosen": -90.71196746826172,
|
|
"logps/ref_chosen": -76.08218383789062,
|
|
"logps/ref_rejected": -94.9686050415039,
|
|
"logps/rejected": -128.43321228027344,
|
|
"loss": 0.8622,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -0.7034212350845337,
|
|
"rewards/margins": 0.8995457887649536,
|
|
"rewards/rejected": -1.6029670238494873,
|
|
"step": 182
|
|
},
|
|
{
|
|
"epoch": 0.2687224669603524,
|
|
"grad_norm": 16.670425415039062,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.04801006615161896,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 4.5910671414162484e-07,
|
|
"logits/chosen": -7.112674713134766,
|
|
"logits/rejected": -6.326689720153809,
|
|
"logps/chosen": -108.50236511230469,
|
|
"logps/ref_chosen": -93.90790557861328,
|
|
"logps/ref_rejected": -83.56061553955078,
|
|
"logps/rejected": -117.24481201171875,
|
|
"loss": 0.8151,
|
|
"rewards/accuracies": 0.890625,
|
|
"rewards/chosen": -0.6983296871185303,
|
|
"rewards/margins": 0.9077068567276001,
|
|
"rewards/rejected": -1.6060365438461304,
|
|
"step": 183
|
|
},
|
|
{
|
|
"epoch": 0.2701908957415565,
|
|
"grad_norm": 17.39129066467285,
|
|
"kl/avg_steps": 0.375,
|
|
"kl/beta": 0.04772669076919556,
|
|
"kl/n_epsilon_steps": 0.3125,
|
|
"kl/p_epsilon_steps": 0.6875,
|
|
"learning_rate": 4.5840059630527985e-07,
|
|
"logits/chosen": -5.867307662963867,
|
|
"logits/rejected": -5.439145565032959,
|
|
"logps/chosen": -100.19184875488281,
|
|
"logps/ref_chosen": -86.86790466308594,
|
|
"logps/ref_rejected": -84.28456115722656,
|
|
"logps/rejected": -110.23704528808594,
|
|
"loss": 1.0505,
|
|
"rewards/accuracies": 0.734375,
|
|
"rewards/chosen": -0.6346135139465332,
|
|
"rewards/margins": 0.5953242778778076,
|
|
"rewards/rejected": -1.2299377918243408,
|
|
"step": 184
|
|
},
|
|
{
|
|
"epoch": 0.27165932452276065,
|
|
"grad_norm": 19.491111755371094,
|
|
"kl/avg_steps": 0.4375,
|
|
"kl/beta": 0.047548383474349976,
|
|
"kl/n_epsilon_steps": 0.28125,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 4.5768898691940836e-07,
|
|
"logits/chosen": -6.329668045043945,
|
|
"logits/rejected": -5.343301773071289,
|
|
"logps/chosen": -108.25900268554688,
|
|
"logps/ref_chosen": -94.02142333984375,
|
|
"logps/ref_rejected": -81.45668029785156,
|
|
"logps/rejected": -111.11482238769531,
|
|
"loss": 0.9684,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.6773566007614136,
|
|
"rewards/margins": 0.7247765064239502,
|
|
"rewards/rejected": -1.4021331071853638,
|
|
"step": 185
|
|
},
|
|
{
|
|
"epoch": 0.27312775330396477,
|
|
"grad_norm": 18.766021728515625,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.04734126478433609,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 4.5697190473557947e-07,
|
|
"logits/chosen": -6.936692237854004,
|
|
"logits/rejected": -6.2103986740112305,
|
|
"logps/chosen": -131.09014892578125,
|
|
"logps/ref_chosen": -114.19168090820312,
|
|
"logps/ref_rejected": -93.33912658691406,
|
|
"logps/rejected": -127.2985610961914,
|
|
"loss": 0.8852,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.7987364530563354,
|
|
"rewards/margins": 0.7991558313369751,
|
|
"rewards/rejected": -1.5978922843933105,
|
|
"step": 186
|
|
},
|
|
{
|
|
"epoch": 0.2745961820851689,
|
|
"grad_norm": 16.06877899169922,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.047120388597249985,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 4.5624936864957555e-07,
|
|
"logits/chosen": -6.3317413330078125,
|
|
"logits/rejected": -5.483089447021484,
|
|
"logps/chosen": -91.86463928222656,
|
|
"logps/ref_chosen": -76.86955261230469,
|
|
"logps/ref_rejected": -88.16494750976562,
|
|
"logps/rejected": -118.53182983398438,
|
|
"loss": 0.968,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.705847978591919,
|
|
"rewards/margins": 0.7159624099731445,
|
|
"rewards/rejected": -1.4218103885650635,
|
|
"step": 187
|
|
},
|
|
{
|
|
"epoch": 0.27606461086637296,
|
|
"grad_norm": 15.785407066345215,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.046871382743120193,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 4.5552139770089454e-07,
|
|
"logits/chosen": -6.618991851806641,
|
|
"logits/rejected": -5.915453910827637,
|
|
"logps/chosen": -90.27525329589844,
|
|
"logps/ref_chosen": -79.19479370117188,
|
|
"logps/ref_rejected": -96.82890319824219,
|
|
"logps/rejected": -125.84371948242188,
|
|
"loss": 0.8615,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -0.5177306532859802,
|
|
"rewards/margins": 0.8314113616943359,
|
|
"rewards/rejected": -1.349142074584961,
|
|
"step": 188
|
|
},
|
|
{
|
|
"epoch": 0.2775330396475771,
|
|
"grad_norm": 14.651037216186523,
|
|
"kl/avg_steps": 0.34375,
|
|
"kl/beta": 0.04658025875687599,
|
|
"kl/n_epsilon_steps": 0.328125,
|
|
"kl/p_epsilon_steps": 0.671875,
|
|
"learning_rate": 4.5478801107224794e-07,
|
|
"logits/chosen": -6.843825817108154,
|
|
"logits/rejected": -6.071965217590332,
|
|
"logps/chosen": -98.51889038085938,
|
|
"logps/ref_chosen": -83.8927993774414,
|
|
"logps/ref_rejected": -79.46589660644531,
|
|
"logps/rejected": -112.61798095703125,
|
|
"loss": 0.8845,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.6822503805160522,
|
|
"rewards/margins": 0.8548742532730103,
|
|
"rewards/rejected": -1.5371246337890625,
|
|
"step": 189
|
|
},
|
|
{
|
|
"epoch": 0.2790014684287812,
|
|
"grad_norm": 15.543427467346191,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.046420685946941376,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 4.5404922808905543e-07,
|
|
"logits/chosen": -6.244542121887207,
|
|
"logits/rejected": -5.1248064041137695,
|
|
"logps/chosen": -114.46714782714844,
|
|
"logps/ref_chosen": -103.44146728515625,
|
|
"logps/ref_rejected": -109.21503448486328,
|
|
"logps/rejected": -141.81443786621094,
|
|
"loss": 0.7663,
|
|
"rewards/accuracies": 0.90625,
|
|
"rewards/chosen": -0.5097974538803101,
|
|
"rewards/margins": 0.991597056388855,
|
|
"rewards/rejected": -1.501394510269165,
|
|
"step": 190
|
|
},
|
|
{
|
|
"epoch": 0.28046989720998533,
|
|
"grad_norm": 16.704288482666016,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.04611803591251373,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 4.5330506821893565e-07,
|
|
"logits/chosen": -5.9573469161987305,
|
|
"logits/rejected": -5.739363193511963,
|
|
"logps/chosen": -110.06816864013672,
|
|
"logps/ref_chosen": -99.78551483154297,
|
|
"logps/ref_rejected": -116.00028228759766,
|
|
"logps/rejected": -149.94467163085938,
|
|
"loss": 0.7528,
|
|
"rewards/accuracies": 0.890625,
|
|
"rewards/chosen": -0.47380760312080383,
|
|
"rewards/margins": 1.0805836915969849,
|
|
"rewards/rejected": -1.5543913841247559,
|
|
"step": 191
|
|
},
|
|
{
|
|
"epoch": 0.28193832599118945,
|
|
"grad_norm": 19.49271583557129,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.04584582895040512,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 4.5255555107119336e-07,
|
|
"logits/chosen": -6.447579860687256,
|
|
"logits/rejected": -5.825846195220947,
|
|
"logps/chosen": -119.15298461914062,
|
|
"logps/ref_chosen": -100.86032104492188,
|
|
"logps/ref_rejected": -109.44500732421875,
|
|
"logps/rejected": -144.4326934814453,
|
|
"loss": 0.9386,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.8361877202987671,
|
|
"rewards/margins": 0.757125973701477,
|
|
"rewards/rejected": -1.5933136940002441,
|
|
"step": 192
|
|
},
|
|
{
|
|
"epoch": 0.2834067547723935,
|
|
"grad_norm": 19.705093383789062,
|
|
"kl/avg_steps": 0.25,
|
|
"kl/beta": 0.04560355842113495,
|
|
"kl/n_epsilon_steps": 0.375,
|
|
"kl/p_epsilon_steps": 0.625,
|
|
"learning_rate": 4.5180069639630236e-07,
|
|
"logits/chosen": -6.4181036949157715,
|
|
"logits/rejected": -5.755351543426514,
|
|
"logps/chosen": -107.38533020019531,
|
|
"logps/ref_chosen": -92.93831634521484,
|
|
"logps/ref_rejected": -83.10917663574219,
|
|
"logps/rejected": -111.87471771240234,
|
|
"loss": 1.05,
|
|
"rewards/accuracies": 0.71875,
|
|
"rewards/chosen": -0.6607075929641724,
|
|
"rewards/margins": 0.6442556381225586,
|
|
"rewards/rejected": -1.304963231086731,
|
|
"step": 193
|
|
},
|
|
{
|
|
"epoch": 0.28487518355359764,
|
|
"grad_norm": 14.911462783813477,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.0454898327589035,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.510405240853854e-07,
|
|
"logits/chosen": -6.82811164855957,
|
|
"logits/rejected": -6.5324788093566895,
|
|
"logps/chosen": -79.15022277832031,
|
|
"logps/ref_chosen": -63.79157257080078,
|
|
"logps/ref_rejected": -68.72665405273438,
|
|
"logps/rejected": -100.59823608398438,
|
|
"loss": 0.9448,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.6952458620071411,
|
|
"rewards/margins": 0.7425950765609741,
|
|
"rewards/rejected": -1.4378409385681152,
|
|
"step": 194
|
|
},
|
|
{
|
|
"epoch": 0.28634361233480177,
|
|
"grad_norm": 15.496999740600586,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.04523538425564766,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 4.5027505416968985e-07,
|
|
"logits/chosen": -6.802285194396973,
|
|
"logits/rejected": -5.815672874450684,
|
|
"logps/chosen": -98.27217102050781,
|
|
"logps/ref_chosen": -83.88545989990234,
|
|
"logps/ref_rejected": -104.00125122070312,
|
|
"logps/rejected": -139.27413940429688,
|
|
"loss": 0.847,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -0.6492324471473694,
|
|
"rewards/margins": 0.9344033002853394,
|
|
"rewards/rejected": -1.583635687828064,
|
|
"step": 195
|
|
},
|
|
{
|
|
"epoch": 0.2878120411160059,
|
|
"grad_norm": 13.717185974121094,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.04495441913604736,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.495043068200599e-07,
|
|
"logits/chosen": -6.458456993103027,
|
|
"logits/rejected": -5.85124397277832,
|
|
"logps/chosen": -105.92664337158203,
|
|
"logps/ref_chosen": -89.4013671875,
|
|
"logps/ref_rejected": -83.34735107421875,
|
|
"logps/rejected": -120.28113555908203,
|
|
"loss": 0.8099,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.7399119138717651,
|
|
"rewards/margins": 0.908238410949707,
|
|
"rewards/rejected": -1.6481503248214722,
|
|
"step": 196
|
|
},
|
|
{
|
|
"epoch": 0.28928046989721,
|
|
"grad_norm": 16.101377487182617,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.04470296576619148,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.4872830234640493e-07,
|
|
"logits/chosen": -6.566864013671875,
|
|
"logits/rejected": -5.954580783843994,
|
|
"logps/chosen": -103.89794921875,
|
|
"logps/ref_chosen": -86.8406982421875,
|
|
"logps/ref_rejected": -82.79289245605469,
|
|
"logps/rejected": -112.60846710205078,
|
|
"loss": 1.0239,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.7603064179420471,
|
|
"rewards/margins": 0.5637940168380737,
|
|
"rewards/rejected": -1.3241004943847656,
|
|
"step": 197
|
|
},
|
|
{
|
|
"epoch": 0.2907488986784141,
|
|
"grad_norm": 18.52680206298828,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.04445291683077812,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 4.479470611971645e-07,
|
|
"logits/chosen": -6.684327125549316,
|
|
"logits/rejected": -5.88149881362915,
|
|
"logps/chosen": -104.97913360595703,
|
|
"logps/ref_chosen": -85.04824829101562,
|
|
"logps/ref_rejected": -104.10365295410156,
|
|
"logps/rejected": -143.3640594482422,
|
|
"loss": 0.8573,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.8809682130813599,
|
|
"rewards/margins": 0.8499820232391357,
|
|
"rewards/rejected": -1.730950117111206,
|
|
"step": 198
|
|
},
|
|
{
|
|
"epoch": 0.2922173274596182,
|
|
"grad_norm": 15.971412658691406,
|
|
"kl/avg_steps": 0.4375,
|
|
"kl/beta": 0.044135693460702896,
|
|
"kl/n_epsilon_steps": 0.28125,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 4.471606039587695e-07,
|
|
"logits/chosen": -6.563417434692383,
|
|
"logits/rejected": -6.149653434753418,
|
|
"logps/chosen": -101.11582946777344,
|
|
"logps/ref_chosen": -81.78669738769531,
|
|
"logps/ref_rejected": -90.790771484375,
|
|
"logps/rejected": -126.85987091064453,
|
|
"loss": 0.9335,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.8515444993972778,
|
|
"rewards/margins": 0.7307862043380737,
|
|
"rewards/rejected": -1.5823307037353516,
|
|
"step": 199
|
|
},
|
|
{
|
|
"epoch": 0.2936857562408223,
|
|
"grad_norm": 14.888577461242676,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.043943438678979874,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 4.4636895135509966e-07,
|
|
"logits/chosen": -6.618537425994873,
|
|
"logits/rejected": -5.955951690673828,
|
|
"logps/chosen": -102.80973052978516,
|
|
"logps/ref_chosen": -84.26809692382812,
|
|
"logps/ref_rejected": -86.96786499023438,
|
|
"logps/rejected": -125.29328918457031,
|
|
"loss": 0.9014,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.8119258284568787,
|
|
"rewards/margins": 0.8592541813850403,
|
|
"rewards/rejected": -1.6711798906326294,
|
|
"step": 200
|
|
},
|
|
{
|
|
"epoch": 0.2936857562408223,
|
|
"eval_kl/n_epsilon_steps": 0.3069349229335785,
|
|
"eval_kl/p_epsilon_steps": 0.690496563911438,
|
|
"eval_logits/chosen": -5.910821914672852,
|
|
"eval_logits/rejected": -5.565565586090088,
|
|
"eval_logps/chosen": -122.01994323730469,
|
|
"eval_logps/ref_chosen": -100.49356842041016,
|
|
"eval_logps/ref_rejected": -94.06775665283203,
|
|
"eval_logps/rejected": -130.100341796875,
|
|
"eval_loss": 0.5126060247421265,
|
|
"eval_rewards/accuracies": 0.7619863152503967,
|
|
"eval_rewards/chosen": -0.9392141103744507,
|
|
"eval_rewards/margins": 0.6259253621101379,
|
|
"eval_rewards/rejected": -1.5651392936706543,
|
|
"eval_runtime": 48.5322,
|
|
"eval_samples_per_second": 48.195,
|
|
"eval_steps_per_second": 1.525,
|
|
"step": 200
|
|
},
|
|
{
|
|
"epoch": 0.29515418502202645,
|
|
"grad_norm": 16.643579483032227,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.04368406534194946,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 4.455721242469372e-07,
|
|
"logits/chosen": -6.567927360534668,
|
|
"logits/rejected": -5.76133394241333,
|
|
"logps/chosen": -127.57327270507812,
|
|
"logps/ref_chosen": -107.77249145507812,
|
|
"logps/ref_rejected": -119.79248046875,
|
|
"logps/rejected": -158.6796875,
|
|
"loss": 0.9275,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.8622275590896606,
|
|
"rewards/margins": 0.8243715763092041,
|
|
"rewards/rejected": -1.6865991353988647,
|
|
"step": 201
|
|
},
|
|
{
|
|
"epoch": 0.2966226138032305,
|
|
"grad_norm": 15.5465669631958,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.043466731905937195,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 4.4477014363141755e-07,
|
|
"logits/chosen": -6.6011576652526855,
|
|
"logits/rejected": -5.9799299240112305,
|
|
"logps/chosen": -95.47408294677734,
|
|
"logps/ref_chosen": -75.97245025634766,
|
|
"logps/ref_rejected": -94.4599838256836,
|
|
"logps/rejected": -130.75558471679688,
|
|
"loss": 0.947,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -0.8463935256004333,
|
|
"rewards/margins": 0.7217241525650024,
|
|
"rewards/rejected": -1.568117618560791,
|
|
"step": 202
|
|
},
|
|
{
|
|
"epoch": 0.29809104258443464,
|
|
"grad_norm": 14.089229583740234,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.04325047880411148,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.439630306414758e-07,
|
|
"logits/chosen": -6.752559661865234,
|
|
"logits/rejected": -5.879279136657715,
|
|
"logps/chosen": -111.82502746582031,
|
|
"logps/ref_chosen": -94.96715545654297,
|
|
"logps/ref_rejected": -92.8876724243164,
|
|
"logps/rejected": -129.54229736328125,
|
|
"loss": 0.8472,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -0.7273061275482178,
|
|
"rewards/margins": 0.8475444316864014,
|
|
"rewards/rejected": -1.5748505592346191,
|
|
"step": 203
|
|
},
|
|
{
|
|
"epoch": 0.29955947136563876,
|
|
"grad_norm": 17.907960891723633,
|
|
"kl/avg_steps": 0.40625,
|
|
"kl/beta": 0.043008554726839066,
|
|
"kl/n_epsilon_steps": 0.296875,
|
|
"kl/p_epsilon_steps": 0.703125,
|
|
"learning_rate": 4.431508065452897e-07,
|
|
"logits/chosen": -6.687747955322266,
|
|
"logits/rejected": -5.982107162475586,
|
|
"logps/chosen": -137.6444091796875,
|
|
"logps/ref_chosen": -116.35719299316406,
|
|
"logps/ref_rejected": -93.39759063720703,
|
|
"logps/rejected": -131.52774047851562,
|
|
"loss": 0.9609,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -0.9145333170890808,
|
|
"rewards/margins": 0.7152736186981201,
|
|
"rewards/rejected": -1.6298069953918457,
|
|
"step": 204
|
|
},
|
|
{
|
|
"epoch": 0.3010279001468429,
|
|
"grad_norm": 12.563446998596191,
|
|
"kl/avg_steps": 0.75,
|
|
"kl/beta": 0.04283454269170761,
|
|
"kl/n_epsilon_steps": 0.125,
|
|
"kl/p_epsilon_steps": 0.875,
|
|
"learning_rate": 4.4233349274571974e-07,
|
|
"logits/chosen": -6.747779846191406,
|
|
"logits/rejected": -6.664907932281494,
|
|
"logps/chosen": -108.34004211425781,
|
|
"logps/ref_chosen": -88.85934448242188,
|
|
"logps/ref_rejected": -91.8544921875,
|
|
"logps/rejected": -133.15499877929688,
|
|
"loss": 0.7808,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -0.8293547630310059,
|
|
"rewards/margins": 0.9246822595596313,
|
|
"rewards/rejected": -1.7540370225906372,
|
|
"step": 205
|
|
},
|
|
{
|
|
"epoch": 0.302496328928047,
|
|
"grad_norm": 17.306621551513672,
|
|
"kl/avg_steps": 0.78125,
|
|
"kl/beta": 0.042515672743320465,
|
|
"kl/n_epsilon_steps": 0.109375,
|
|
"kl/p_epsilon_steps": 0.890625,
|
|
"learning_rate": 4.415111107797445e-07,
|
|
"logits/chosen": -6.731910705566406,
|
|
"logits/rejected": -5.665759086608887,
|
|
"logps/chosen": -96.40692138671875,
|
|
"logps/ref_chosen": -76.54634857177734,
|
|
"logps/ref_rejected": -102.95314025878906,
|
|
"logps/rejected": -142.6748809814453,
|
|
"loss": 0.8253,
|
|
"rewards/accuracies": 0.890625,
|
|
"rewards/chosen": -0.8388286828994751,
|
|
"rewards/margins": 0.8355581164360046,
|
|
"rewards/rejected": -1.6743868589401245,
|
|
"step": 206
|
|
},
|
|
{
|
|
"epoch": 0.3039647577092511,
|
|
"grad_norm": 19.9998722076416,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.04218609631061554,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 4.4068368231789365e-07,
|
|
"logits/chosen": -7.154547214508057,
|
|
"logits/rejected": -6.694057464599609,
|
|
"logps/chosen": -107.92840576171875,
|
|
"logps/ref_chosen": -86.23164367675781,
|
|
"logps/ref_rejected": -90.65512084960938,
|
|
"logps/rejected": -133.81866455078125,
|
|
"loss": 0.9093,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.9145298600196838,
|
|
"rewards/margins": 0.8951901197433472,
|
|
"rewards/rejected": -1.8097199201583862,
|
|
"step": 207
|
|
},
|
|
{
|
|
"epoch": 0.3054331864904552,
|
|
"grad_norm": 15.674067497253418,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.041976213455200195,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 4.398512291636768e-07,
|
|
"logits/chosen": -7.054866313934326,
|
|
"logits/rejected": -6.2353129386901855,
|
|
"logps/chosen": -115.6429672241211,
|
|
"logps/ref_chosen": -94.1595458984375,
|
|
"logps/ref_rejected": -100.96233367919922,
|
|
"logps/rejected": -145.07806396484375,
|
|
"loss": 0.7998,
|
|
"rewards/accuracies": 0.921875,
|
|
"rewards/chosen": -0.8975973725318909,
|
|
"rewards/margins": 0.9404553771018982,
|
|
"rewards/rejected": -1.838052749633789,
|
|
"step": 208
|
|
},
|
|
{
|
|
"epoch": 0.3069016152716593,
|
|
"grad_norm": 15.056414604187012,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.04170254245400429,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 4.3901377325300857e-07,
|
|
"logits/chosen": -7.251114845275879,
|
|
"logits/rejected": -6.478391647338867,
|
|
"logps/chosen": -102.76752471923828,
|
|
"logps/ref_chosen": -84.17056274414062,
|
|
"logps/ref_rejected": -87.61955261230469,
|
|
"logps/rejected": -125.9391860961914,
|
|
"loss": 0.8985,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.7734057903289795,
|
|
"rewards/margins": 0.8138402104377747,
|
|
"rewards/rejected": -1.5872459411621094,
|
|
"step": 209
|
|
},
|
|
{
|
|
"epoch": 0.30837004405286345,
|
|
"grad_norm": 13.82249641418457,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.04149506613612175,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 4.381713366536311e-07,
|
|
"logits/chosen": -7.367747783660889,
|
|
"logits/rejected": -6.7037177085876465,
|
|
"logps/chosen": -101.83394622802734,
|
|
"logps/ref_chosen": -81.17117309570312,
|
|
"logps/ref_rejected": -84.17478942871094,
|
|
"logps/rejected": -123.829833984375,
|
|
"loss": 0.8996,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.8534192442893982,
|
|
"rewards/margins": 0.7803501486778259,
|
|
"rewards/rejected": -1.6337693929672241,
|
|
"step": 210
|
|
},
|
|
{
|
|
"epoch": 0.30983847283406757,
|
|
"grad_norm": 16.27776527404785,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.04123733192682266,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 4.373239415645323e-07,
|
|
"logits/chosen": -6.783376216888428,
|
|
"logits/rejected": -6.2379889488220215,
|
|
"logps/chosen": -131.35140991210938,
|
|
"logps/ref_chosen": -108.71271514892578,
|
|
"logps/ref_rejected": -93.55564880371094,
|
|
"logps/rejected": -135.39334106445312,
|
|
"loss": 0.9141,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.9316354990005493,
|
|
"rewards/margins": 0.7831017971038818,
|
|
"rewards/rejected": -1.7147372961044312,
|
|
"step": 211
|
|
},
|
|
{
|
|
"epoch": 0.31130690161527164,
|
|
"grad_norm": 13.126346588134766,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.0410449355840683,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 4.3647161031536086e-07,
|
|
"logits/chosen": -6.987194061279297,
|
|
"logits/rejected": -6.6716461181640625,
|
|
"logps/chosen": -113.453125,
|
|
"logps/ref_chosen": -98.36194610595703,
|
|
"logps/ref_rejected": -109.88999938964844,
|
|
"logps/rejected": -151.75265502929688,
|
|
"loss": 0.7363,
|
|
"rewards/accuracies": 0.90625,
|
|
"rewards/chosen": -0.6175893545150757,
|
|
"rewards/margins": 1.0888441801071167,
|
|
"rewards/rejected": -1.706433653831482,
|
|
"step": 212
|
|
},
|
|
{
|
|
"epoch": 0.31277533039647576,
|
|
"grad_norm": 19.004343032836914,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.04082803428173065,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.3561436536583774e-07,
|
|
"logits/chosen": -6.994483947753906,
|
|
"logits/rejected": -6.408557891845703,
|
|
"logps/chosen": -126.75518798828125,
|
|
"logps/ref_chosen": -108.05531311035156,
|
|
"logps/ref_rejected": -100.14414978027344,
|
|
"logps/rejected": -140.40818786621094,
|
|
"loss": 0.8605,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.7606232166290283,
|
|
"rewards/margins": 0.8708195686340332,
|
|
"rewards/rejected": -1.6314427852630615,
|
|
"step": 213
|
|
},
|
|
{
|
|
"epoch": 0.3142437591776799,
|
|
"grad_norm": 15.15538501739502,
|
|
"kl/avg_steps": 0.4375,
|
|
"kl/beta": 0.04059966281056404,
|
|
"kl/n_epsilon_steps": 0.28125,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 4.3475222930516473e-07,
|
|
"logits/chosen": -6.964527130126953,
|
|
"logits/rejected": -6.679478645324707,
|
|
"logps/chosen": -98.05276489257812,
|
|
"logps/ref_chosen": -77.80473327636719,
|
|
"logps/ref_rejected": -84.27578735351562,
|
|
"logps/rejected": -123.21199798583984,
|
|
"loss": 0.9841,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -0.8206988573074341,
|
|
"rewards/margins": 0.749920129776001,
|
|
"rewards/rejected": -1.570618987083435,
|
|
"step": 214
|
|
},
|
|
{
|
|
"epoch": 0.315712187958884,
|
|
"grad_norm": 13.474177360534668,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.04042281210422516,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 4.3388522485142885e-07,
|
|
"logits/chosen": -7.422246932983398,
|
|
"logits/rejected": -6.752878189086914,
|
|
"logps/chosen": -101.85633087158203,
|
|
"logps/ref_chosen": -85.1138916015625,
|
|
"logps/ref_rejected": -96.86151885986328,
|
|
"logps/rejected": -138.35052490234375,
|
|
"loss": 0.7734,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -0.6723982095718384,
|
|
"rewards/margins": 0.9908286333084106,
|
|
"rewards/rejected": -1.663226842880249,
|
|
"step": 215
|
|
},
|
|
{
|
|
"epoch": 0.31718061674008813,
|
|
"grad_norm": 13.187141418457031,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.04014680162072182,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 4.330133748510036e-07,
|
|
"logits/chosen": -6.856760501861572,
|
|
"logits/rejected": -6.676458835601807,
|
|
"logps/chosen": -99.6475830078125,
|
|
"logps/ref_chosen": -80.5923080444336,
|
|
"logps/ref_rejected": -81.41983795166016,
|
|
"logps/rejected": -121.41978454589844,
|
|
"loss": 0.8502,
|
|
"rewards/accuracies": 0.890625,
|
|
"rewards/chosen": -0.7613588571548462,
|
|
"rewards/margins": 0.832861065864563,
|
|
"rewards/rejected": -1.5942199230194092,
|
|
"step": 216
|
|
},
|
|
{
|
|
"epoch": 0.3186490455212922,
|
|
"grad_norm": 17.514558792114258,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.03993465006351471,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 4.3213670227794757e-07,
|
|
"logits/chosen": -7.025016784667969,
|
|
"logits/rejected": -6.148534774780273,
|
|
"logps/chosen": -115.89637756347656,
|
|
"logps/ref_chosen": -93.47257995605469,
|
|
"logps/ref_rejected": -103.488525390625,
|
|
"logps/rejected": -144.48422241210938,
|
|
"loss": 0.9483,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.8944262266159058,
|
|
"rewards/margins": 0.7329256534576416,
|
|
"rewards/rejected": -1.6273518800735474,
|
|
"step": 217
|
|
},
|
|
{
|
|
"epoch": 0.3201174743024963,
|
|
"grad_norm": 15.729193687438965,
|
|
"kl/avg_steps": 0.453125,
|
|
"kl/beta": 0.03974832966923714,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 4.3125523023339815e-07,
|
|
"logits/chosen": -7.055582046508789,
|
|
"logits/rejected": -6.02012300491333,
|
|
"logps/chosen": -110.00785827636719,
|
|
"logps/ref_chosen": -89.05883026123047,
|
|
"logps/ref_rejected": -94.30680847167969,
|
|
"logps/rejected": -136.88497924804688,
|
|
"loss": 0.8984,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.8316381573677063,
|
|
"rewards/margins": 0.8504889607429504,
|
|
"rewards/rejected": -1.6821269989013672,
|
|
"step": 218
|
|
},
|
|
{
|
|
"epoch": 0.32158590308370044,
|
|
"grad_norm": 19.58609390258789,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.039569031447172165,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 4.303689819449636e-07,
|
|
"logits/chosen": -7.102532386779785,
|
|
"logits/rejected": -6.606064796447754,
|
|
"logps/chosen": -120.11994171142578,
|
|
"logps/ref_chosen": -101.00733947753906,
|
|
"logps/ref_rejected": -92.46794128417969,
|
|
"logps/rejected": -131.8204803466797,
|
|
"loss": 0.9475,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.7558479309082031,
|
|
"rewards/margins": 0.7920831441879272,
|
|
"rewards/rejected": -1.5479310750961304,
|
|
"step": 219
|
|
},
|
|
{
|
|
"epoch": 0.32305433186490456,
|
|
"grad_norm": 16.06678581237793,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.03938441723585129,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 4.2947798076611047e-07,
|
|
"logits/chosen": -6.876911163330078,
|
|
"logits/rejected": -6.016563415527344,
|
|
"logps/chosen": -115.22142028808594,
|
|
"logps/ref_chosen": -95.53721618652344,
|
|
"logps/ref_rejected": -94.30703735351562,
|
|
"logps/rejected": -132.008544921875,
|
|
"loss": 0.9298,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -0.7732564210891724,
|
|
"rewards/margins": 0.702507734298706,
|
|
"rewards/rejected": -1.4757641553878784,
|
|
"step": 220
|
|
},
|
|
{
|
|
"epoch": 0.3245227606461087,
|
|
"grad_norm": 14.65071964263916,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.039200663566589355,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 4.285822501755485e-07,
|
|
"logits/chosen": -6.791396141052246,
|
|
"logits/rejected": -5.949629783630371,
|
|
"logps/chosen": -101.76954650878906,
|
|
"logps/ref_chosen": -82.84486389160156,
|
|
"logps/ref_rejected": -110.81179809570312,
|
|
"logps/rejected": -156.64260864257812,
|
|
"loss": 0.7544,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -0.7377252578735352,
|
|
"rewards/margins": 1.0444551706314087,
|
|
"rewards/rejected": -1.7821805477142334,
|
|
"step": 221
|
|
},
|
|
{
|
|
"epoch": 0.32599118942731276,
|
|
"grad_norm": 20.610990524291992,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.03893300145864487,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 4.276818137766118e-07,
|
|
"logits/chosen": -7.152358055114746,
|
|
"logits/rejected": -6.381607532501221,
|
|
"logps/chosen": -118.88861083984375,
|
|
"logps/ref_chosen": -95.14198303222656,
|
|
"logps/ref_rejected": -106.80441284179688,
|
|
"logps/rejected": -152.92138671875,
|
|
"loss": 0.9338,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.9227837920188904,
|
|
"rewards/margins": 0.8608850240707397,
|
|
"rewards/rejected": -1.7836687564849854,
|
|
"step": 222
|
|
},
|
|
{
|
|
"epoch": 0.3274596182085169,
|
|
"grad_norm": 15.403647422790527,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.03872726112604141,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 4.2677669529663686e-07,
|
|
"logits/chosen": -7.370273590087891,
|
|
"logits/rejected": -6.7063889503479,
|
|
"logps/chosen": -108.16079711914062,
|
|
"logps/ref_chosen": -85.57511138916016,
|
|
"logps/ref_rejected": -86.45238494873047,
|
|
"logps/rejected": -126.88802337646484,
|
|
"loss": 0.98,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.8735491037368774,
|
|
"rewards/margins": 0.6830679178237915,
|
|
"rewards/rejected": -1.556617021560669,
|
|
"step": 223
|
|
},
|
|
{
|
|
"epoch": 0.328928046989721,
|
|
"grad_norm": 15.034847259521484,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.03853458911180496,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 4.2586691858633747e-07,
|
|
"logits/chosen": -7.058864593505859,
|
|
"logits/rejected": -6.6526618003845215,
|
|
"logps/chosen": -104.06519317626953,
|
|
"logps/ref_chosen": -82.72380065917969,
|
|
"logps/ref_rejected": -82.59538269042969,
|
|
"logps/rejected": -126.1408462524414,
|
|
"loss": 0.867,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.820472002029419,
|
|
"rewards/margins": 0.8463258147239685,
|
|
"rewards/rejected": -1.6667978763580322,
|
|
"step": 224
|
|
},
|
|
{
|
|
"epoch": 0.3303964757709251,
|
|
"grad_norm": 15.859506607055664,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.03834287449717522,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 4.249525076191759e-07,
|
|
"logits/chosen": -7.314513206481934,
|
|
"logits/rejected": -6.8365983963012695,
|
|
"logps/chosen": -119.15274810791016,
|
|
"logps/ref_chosen": -95.67768096923828,
|
|
"logps/ref_rejected": -105.09687805175781,
|
|
"logps/rejected": -153.17977905273438,
|
|
"loss": 0.8512,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.8980911374092102,
|
|
"rewards/margins": 0.9331176280975342,
|
|
"rewards/rejected": -1.83120858669281,
|
|
"step": 225
|
|
},
|
|
{
|
|
"epoch": 0.33186490455212925,
|
|
"grad_norm": 13.422561645507812,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.03814025595784187,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 4.2403348649073167e-07,
|
|
"logits/chosen": -7.025399684906006,
|
|
"logits/rejected": -6.684216499328613,
|
|
"logps/chosen": -112.96669006347656,
|
|
"logps/ref_chosen": -93.46092987060547,
|
|
"logps/ref_rejected": -86.7017593383789,
|
|
"logps/rejected": -128.74534606933594,
|
|
"loss": 0.8457,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -0.7408524751663208,
|
|
"rewards/margins": 0.8516495227813721,
|
|
"rewards/rejected": -1.5925018787384033,
|
|
"step": 226
|
|
},
|
|
{
|
|
"epoch": 0.3333333333333333,
|
|
"grad_norm": 15.427087783813477,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.03790335729718208,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.2310987941806615e-07,
|
|
"logits/chosen": -7.346820831298828,
|
|
"logits/rejected": -6.7165141105651855,
|
|
"logps/chosen": -119.78524017333984,
|
|
"logps/ref_chosen": -92.81427001953125,
|
|
"logps/ref_rejected": -104.73692321777344,
|
|
"logps/rejected": -154.31568908691406,
|
|
"loss": 0.906,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -1.0186035633087158,
|
|
"rewards/margins": 0.8465676307678223,
|
|
"rewards/rejected": -1.8651710748672485,
|
|
"step": 227
|
|
},
|
|
{
|
|
"epoch": 0.33480176211453744,
|
|
"grad_norm": 15.720854759216309,
|
|
"kl/avg_steps": 0.40625,
|
|
"kl/beta": 0.03769134357571602,
|
|
"kl/n_epsilon_steps": 0.296875,
|
|
"kl/p_epsilon_steps": 0.703125,
|
|
"learning_rate": 4.2218171073908463e-07,
|
|
"logits/chosen": -7.082067489624023,
|
|
"logits/rejected": -6.396081924438477,
|
|
"logps/chosen": -118.57168579101562,
|
|
"logps/ref_chosen": -94.03712463378906,
|
|
"logps/ref_rejected": -96.02151489257812,
|
|
"logps/rejected": -138.6416015625,
|
|
"loss": 0.962,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.9233343601226807,
|
|
"rewards/margins": 0.6742476224899292,
|
|
"rewards/rejected": -1.5975819826126099,
|
|
"step": 228
|
|
},
|
|
{
|
|
"epoch": 0.33627019089574156,
|
|
"grad_norm": 13.65562915802002,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.03753884509205818,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.212490049118951e-07,
|
|
"logits/chosen": -7.236542224884033,
|
|
"logits/rejected": -6.673727989196777,
|
|
"logps/chosen": -116.85358428955078,
|
|
"logps/ref_chosen": -95.57766723632812,
|
|
"logps/ref_rejected": -89.17379760742188,
|
|
"logps/rejected": -139.57473754882812,
|
|
"loss": 0.7638,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -0.7962503433227539,
|
|
"rewards/margins": 1.0827782154083252,
|
|
"rewards/rejected": -1.879028558731079,
|
|
"step": 229
|
|
},
|
|
{
|
|
"epoch": 0.3377386196769457,
|
|
"grad_norm": 15.202421188354492,
|
|
"kl/avg_steps": 0.78125,
|
|
"kl/beta": 0.037328869104385376,
|
|
"kl/n_epsilon_steps": 0.109375,
|
|
"kl/p_epsilon_steps": 0.890625,
|
|
"learning_rate": 4.203117865141635e-07,
|
|
"logits/chosen": -7.399197578430176,
|
|
"logits/rejected": -6.820221900939941,
|
|
"logps/chosen": -86.61280822753906,
|
|
"logps/ref_chosen": -63.713626861572266,
|
|
"logps/ref_rejected": -91.9087142944336,
|
|
"logps/rejected": -143.0418243408203,
|
|
"loss": 0.7439,
|
|
"rewards/accuracies": 0.90625,
|
|
"rewards/chosen": -0.8491979241371155,
|
|
"rewards/margins": 1.0426983833312988,
|
|
"rewards/rejected": -1.8918962478637695,
|
|
"step": 230
|
|
},
|
|
{
|
|
"epoch": 0.3392070484581498,
|
|
"grad_norm": 13.081562995910645,
|
|
"kl/avg_steps": 0.671875,
|
|
"kl/beta": 0.03703949600458145,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 4.1937008024246625e-07,
|
|
"logits/chosen": -6.668990135192871,
|
|
"logits/rejected": -6.165530681610107,
|
|
"logps/chosen": -121.447998046875,
|
|
"logps/ref_chosen": -95.45567321777344,
|
|
"logps/ref_rejected": -80.95568084716797,
|
|
"logps/rejected": -130.36236572265625,
|
|
"loss": 0.8274,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -0.957029402256012,
|
|
"rewards/margins": 0.8586543798446655,
|
|
"rewards/rejected": -1.8156838417053223,
|
|
"step": 231
|
|
},
|
|
{
|
|
"epoch": 0.3406754772393539,
|
|
"grad_norm": 15.00186538696289,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.03679230064153671,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 4.1842391091163933e-07,
|
|
"logits/chosen": -6.989046096801758,
|
|
"logits/rejected": -6.184296607971191,
|
|
"logps/chosen": -122.46676635742188,
|
|
"logps/ref_chosen": -96.89726257324219,
|
|
"logps/ref_rejected": -89.76461791992188,
|
|
"logps/rejected": -136.52182006835938,
|
|
"loss": 0.9199,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.9383091926574707,
|
|
"rewards/margins": 0.7705283164978027,
|
|
"rewards/rejected": -1.7088375091552734,
|
|
"step": 232
|
|
},
|
|
{
|
|
"epoch": 0.342143906020558,
|
|
"grad_norm": 14.02349853515625,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.03660925105214119,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 4.174733034541245e-07,
|
|
"logits/chosen": -7.088686943054199,
|
|
"logits/rejected": -6.224997520446777,
|
|
"logps/chosen": -116.73382568359375,
|
|
"logps/ref_chosen": -89.05032348632812,
|
|
"logps/ref_rejected": -112.75917053222656,
|
|
"logps/rejected": -169.069580078125,
|
|
"loss": 0.7772,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.0092403888702393,
|
|
"rewards/margins": 1.036684513092041,
|
|
"rewards/rejected": -2.045924663543701,
|
|
"step": 233
|
|
},
|
|
{
|
|
"epoch": 0.3436123348017621,
|
|
"grad_norm": 13.894185066223145,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.03638186678290367,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 4.165182829193126e-07,
|
|
"logits/chosen": -7.294958114624023,
|
|
"logits/rejected": -6.297677040100098,
|
|
"logps/chosen": -100.55630493164062,
|
|
"logps/ref_chosen": -74.318359375,
|
|
"logps/ref_rejected": -106.38758850097656,
|
|
"logps/rejected": -162.41140747070312,
|
|
"loss": 0.7907,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -0.9510073661804199,
|
|
"rewards/margins": 1.071781873703003,
|
|
"rewards/rejected": -2.022789239883423,
|
|
"step": 234
|
|
},
|
|
{
|
|
"epoch": 0.34508076358296624,
|
|
"grad_norm": 13.863997459411621,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.03615589067339897,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 4.1555887447288255e-07,
|
|
"logits/chosen": -6.75037956237793,
|
|
"logits/rejected": -6.180291175842285,
|
|
"logps/chosen": -129.61538696289062,
|
|
"logps/ref_chosen": -98.217041015625,
|
|
"logps/ref_rejected": -97.24677276611328,
|
|
"logps/rejected": -151.29678344726562,
|
|
"loss": 0.8641,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -1.1314321756362915,
|
|
"rewards/margins": 0.810757577419281,
|
|
"rewards/rejected": -1.9421896934509277,
|
|
"step": 235
|
|
},
|
|
{
|
|
"epoch": 0.3465491923641703,
|
|
"grad_norm": 14.032482147216797,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.03596482798457146,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 4.1459510339613946e-07,
|
|
"logits/chosen": -6.762874126434326,
|
|
"logits/rejected": -6.526078224182129,
|
|
"logps/chosen": -105.3929672241211,
|
|
"logps/ref_chosen": -78.83773040771484,
|
|
"logps/ref_rejected": -109.06343078613281,
|
|
"logps/rejected": -163.83856201171875,
|
|
"loss": 0.8264,
|
|
"rewards/accuracies": 0.921875,
|
|
"rewards/chosen": -0.9487680196762085,
|
|
"rewards/margins": 1.0044116973876953,
|
|
"rewards/rejected": -1.9531795978546143,
|
|
"step": 236
|
|
},
|
|
{
|
|
"epoch": 0.34801762114537443,
|
|
"grad_norm": 15.7308931350708,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.03570817783474922,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 4.136269950853473e-07,
|
|
"logits/chosen": -7.501216888427734,
|
|
"logits/rejected": -6.719086647033691,
|
|
"logps/chosen": -115.78558349609375,
|
|
"logps/ref_chosen": -85.21128845214844,
|
|
"logps/ref_rejected": -99.90999603271484,
|
|
"logps/rejected": -153.35638427734375,
|
|
"loss": 0.9932,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -1.0902290344238281,
|
|
"rewards/margins": 0.8065685033798218,
|
|
"rewards/rejected": -1.89679753780365,
|
|
"step": 237
|
|
},
|
|
{
|
|
"epoch": 0.34948604992657856,
|
|
"grad_norm": 13.518217086791992,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.03554157540202141,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 4.126545750510605e-07,
|
|
"logits/chosen": -6.873219013214111,
|
|
"logits/rejected": -5.906381607055664,
|
|
"logps/chosen": -112.39120483398438,
|
|
"logps/ref_chosen": -78.73123168945312,
|
|
"logps/ref_rejected": -95.41840362548828,
|
|
"logps/rejected": -153.98170471191406,
|
|
"loss": 0.8772,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -1.190969705581665,
|
|
"rewards/margins": 0.8749915361404419,
|
|
"rewards/rejected": -2.0659611225128174,
|
|
"step": 238
|
|
},
|
|
{
|
|
"epoch": 0.3509544787077827,
|
|
"grad_norm": 14.817635536193848,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.035320818424224854,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 4.116778689174514e-07,
|
|
"logits/chosen": -7.087869644165039,
|
|
"logits/rejected": -6.191803932189941,
|
|
"logps/chosen": -124.30506134033203,
|
|
"logps/ref_chosen": -92.60093688964844,
|
|
"logps/ref_rejected": -100.51769256591797,
|
|
"logps/rejected": -155.46278381347656,
|
|
"loss": 0.9094,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -1.117043375968933,
|
|
"rewards/margins": 0.8112146854400635,
|
|
"rewards/rejected": -1.9282580614089966,
|
|
"step": 239
|
|
},
|
|
{
|
|
"epoch": 0.3524229074889868,
|
|
"grad_norm": 17.138147354125977,
|
|
"kl/avg_steps": 0.375,
|
|
"kl/beta": 0.03514509275555611,
|
|
"kl/n_epsilon_steps": 0.3125,
|
|
"kl/p_epsilon_steps": 0.6875,
|
|
"learning_rate": 4.106969024216348e-07,
|
|
"logits/chosen": -7.254019260406494,
|
|
"logits/rejected": -7.064877986907959,
|
|
"logps/chosen": -118.43206787109375,
|
|
"logps/ref_chosen": -86.15977478027344,
|
|
"logps/ref_rejected": -80.45567321777344,
|
|
"logps/rejected": -133.68617248535156,
|
|
"loss": 0.9385,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -1.1331913471221924,
|
|
"rewards/margins": 0.7280229926109314,
|
|
"rewards/rejected": -1.8612143993377686,
|
|
"step": 240
|
|
},
|
|
{
|
|
"epoch": 0.35389133627019087,
|
|
"grad_norm": 13.10658073425293,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.03501379117369652,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 4.097117014129903e-07,
|
|
"logits/chosen": -7.0530242919921875,
|
|
"logits/rejected": -6.54688835144043,
|
|
"logps/chosen": -128.86328125,
|
|
"logps/ref_chosen": -101.04594421386719,
|
|
"logps/ref_rejected": -94.04934692382812,
|
|
"logps/rejected": -158.0999298095703,
|
|
"loss": 0.6851,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -0.9694840312004089,
|
|
"rewards/margins": 1.2560572624206543,
|
|
"rewards/rejected": -2.225541353225708,
|
|
"step": 241
|
|
},
|
|
{
|
|
"epoch": 0.355359765051395,
|
|
"grad_norm": 14.712688446044922,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.034774716943502426,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 4.087222918524807e-07,
|
|
"logits/chosen": -7.671010494232178,
|
|
"logits/rejected": -7.0201215744018555,
|
|
"logps/chosen": -129.89794921875,
|
|
"logps/ref_chosen": -95.67266082763672,
|
|
"logps/ref_rejected": -90.65454864501953,
|
|
"logps/rejected": -150.88523864746094,
|
|
"loss": 0.8351,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.184476613998413,
|
|
"rewards/margins": 0.8943547606468201,
|
|
"rewards/rejected": -2.078831434249878,
|
|
"step": 242
|
|
},
|
|
{
|
|
"epoch": 0.3568281938325991,
|
|
"grad_norm": 14.254800796508789,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.03453727439045906,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.07728699811968e-07,
|
|
"logits/chosen": -7.436939239501953,
|
|
"logits/rejected": -7.066817283630371,
|
|
"logps/chosen": -130.82611083984375,
|
|
"logps/ref_chosen": -98.03140258789062,
|
|
"logps/ref_rejected": -83.18806457519531,
|
|
"logps/rejected": -141.62124633789062,
|
|
"loss": 0.8947,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -1.1293984651565552,
|
|
"rewards/margins": 0.8747692704200745,
|
|
"rewards/rejected": -2.0041677951812744,
|
|
"step": 243
|
|
},
|
|
{
|
|
"epoch": 0.35829662261380324,
|
|
"grad_norm": 14.085189819335938,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.03434408828616142,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 4.067309514735267e-07,
|
|
"logits/chosen": -7.297283172607422,
|
|
"logits/rejected": -6.48216438293457,
|
|
"logps/chosen": -123.65585327148438,
|
|
"logps/ref_chosen": -88.89391326904297,
|
|
"logps/ref_rejected": -102.57278442382812,
|
|
"logps/rejected": -163.68592834472656,
|
|
"loss": 0.8616,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.187152624130249,
|
|
"rewards/margins": 0.8951038122177124,
|
|
"rewards/rejected": -2.082256317138672,
|
|
"step": 244
|
|
},
|
|
{
|
|
"epoch": 0.35976505139500736,
|
|
"grad_norm": 13.635170936584473,
|
|
"kl/avg_steps": 0.78125,
|
|
"kl/beta": 0.03412017226219177,
|
|
"kl/n_epsilon_steps": 0.109375,
|
|
"kl/p_epsilon_steps": 0.890625,
|
|
"learning_rate": 4.057290731287531e-07,
|
|
"logits/chosen": -7.1143999099731445,
|
|
"logits/rejected": -6.517737865447998,
|
|
"logps/chosen": -133.73660278320312,
|
|
"logps/ref_chosen": -104.19400024414062,
|
|
"logps/ref_rejected": -92.65645599365234,
|
|
"logps/rejected": -151.42526245117188,
|
|
"loss": 0.777,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -1.0023081302642822,
|
|
"rewards/margins": 0.9868428707122803,
|
|
"rewards/rejected": -1.9891510009765625,
|
|
"step": 245
|
|
},
|
|
{
|
|
"epoch": 0.36123348017621143,
|
|
"grad_norm": 15.270983695983887,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.03385567665100098,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 4.047230911780736e-07,
|
|
"logits/chosen": -6.8056464195251465,
|
|
"logits/rejected": -6.480011940002441,
|
|
"logps/chosen": -141.06942749023438,
|
|
"logps/ref_chosen": -103.21904754638672,
|
|
"logps/ref_rejected": -90.9922103881836,
|
|
"logps/rejected": -151.3291015625,
|
|
"loss": 0.9225,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -1.277728796005249,
|
|
"rewards/margins": 0.7526397705078125,
|
|
"rewards/rejected": -2.0303685665130615,
|
|
"step": 246
|
|
},
|
|
{
|
|
"epoch": 0.36270190895741555,
|
|
"grad_norm": 12.317371368408203,
|
|
"kl/avg_steps": 0.78125,
|
|
"kl/beta": 0.03367676958441734,
|
|
"kl/n_epsilon_steps": 0.109375,
|
|
"kl/p_epsilon_steps": 0.890625,
|
|
"learning_rate": 4.0371303213004814e-07,
|
|
"logits/chosen": -7.767994403839111,
|
|
"logits/rejected": -7.044826507568359,
|
|
"logps/chosen": -121.5998306274414,
|
|
"logps/ref_chosen": -86.99436950683594,
|
|
"logps/ref_rejected": -111.33802795410156,
|
|
"logps/rejected": -179.18167114257812,
|
|
"loss": 0.7459,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -1.1581556797027588,
|
|
"rewards/margins": 1.1077954769134521,
|
|
"rewards/rejected": -2.265951156616211,
|
|
"step": 247
|
|
},
|
|
{
|
|
"epoch": 0.3641703377386197,
|
|
"grad_norm": 15.562211990356445,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.03341570869088173,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.0269892260067197e-07,
|
|
"logits/chosen": -7.3634748458862305,
|
|
"logits/rejected": -6.64450740814209,
|
|
"logps/chosen": -113.81005859375,
|
|
"logps/ref_chosen": -74.7855224609375,
|
|
"logps/ref_rejected": -98.27689361572266,
|
|
"logps/rejected": -160.82879638671875,
|
|
"loss": 0.9066,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -1.2984592914581299,
|
|
"rewards/margins": 0.7774021625518799,
|
|
"rewards/rejected": -2.0758614540100098,
|
|
"step": 248
|
|
},
|
|
{
|
|
"epoch": 0.3656387665198238,
|
|
"grad_norm": 17.970985412597656,
|
|
"kl/avg_steps": 0.4375,
|
|
"kl/beta": 0.0332287959754467,
|
|
"kl/n_epsilon_steps": 0.28125,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 4.0168078931267426e-07,
|
|
"logits/chosen": -7.139558792114258,
|
|
"logits/rejected": -6.51041316986084,
|
|
"logps/chosen": -137.87208557128906,
|
|
"logps/ref_chosen": -95.70379638671875,
|
|
"logps/ref_rejected": -87.14646911621094,
|
|
"logps/rejected": -151.0142059326172,
|
|
"loss": 0.9861,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -1.3996614217758179,
|
|
"rewards/margins": 0.7120518684387207,
|
|
"rewards/rejected": -2.111713409423828,
|
|
"step": 249
|
|
},
|
|
{
|
|
"epoch": 0.3671071953010279,
|
|
"grad_norm": 18.755630493164062,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.03308405354619026,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.006586590948141e-07,
|
|
"logits/chosen": -7.073131561279297,
|
|
"logits/rejected": -6.734502792358398,
|
|
"logps/chosen": -153.10398864746094,
|
|
"logps/ref_chosen": -114.05220794677734,
|
|
"logps/ref_rejected": -81.08768463134766,
|
|
"logps/rejected": -142.95217895507812,
|
|
"loss": 0.9481,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -1.287900686264038,
|
|
"rewards/margins": 0.7461932897567749,
|
|
"rewards/rejected": -2.0340938568115234,
|
|
"step": 250
|
|
},
|
|
{
|
|
"epoch": 0.368575624082232,
|
|
"grad_norm": 15.883668899536133,
|
|
"kl/avg_steps": 0.28125,
|
|
"kl/beta": 0.03289899602532387,
|
|
"kl/n_epsilon_steps": 0.359375,
|
|
"kl/p_epsilon_steps": 0.640625,
|
|
"learning_rate": 3.9963255888117325e-07,
|
|
"logits/chosen": -7.458626747131348,
|
|
"logits/rejected": -6.964792251586914,
|
|
"logps/chosen": -135.24278259277344,
|
|
"logps/ref_chosen": -97.71128845214844,
|
|
"logps/ref_rejected": -83.52742004394531,
|
|
"logps/rejected": -142.99111938476562,
|
|
"loss": 0.9872,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -1.2341444492340088,
|
|
"rewards/margins": 0.7130569219589233,
|
|
"rewards/rejected": -1.9472013711929321,
|
|
"step": 251
|
|
},
|
|
{
|
|
"epoch": 0.3700440528634361,
|
|
"grad_norm": 16.21406364440918,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.03280672803521156,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 3.9860251571044666e-07,
|
|
"logits/chosen": -7.022328853607178,
|
|
"logits/rejected": -6.755680084228516,
|
|
"logps/chosen": -145.88766479492188,
|
|
"logps/ref_chosen": -108.9861068725586,
|
|
"logps/ref_rejected": -91.56424713134766,
|
|
"logps/rejected": -151.3180389404297,
|
|
"loss": 0.93,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -1.2076178789138794,
|
|
"rewards/margins": 0.7408928871154785,
|
|
"rewards/rejected": -1.948510766029358,
|
|
"step": 252
|
|
},
|
|
{
|
|
"epoch": 0.37151248164464024,
|
|
"grad_norm": 15.019343376159668,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.03264351189136505,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 3.9756855672522986e-07,
|
|
"logits/chosen": -6.749828815460205,
|
|
"logits/rejected": -6.487143516540527,
|
|
"logps/chosen": -140.8556365966797,
|
|
"logps/ref_chosen": -100.21630859375,
|
|
"logps/ref_rejected": -105.67670440673828,
|
|
"logps/rejected": -168.1231689453125,
|
|
"loss": 0.9468,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -1.3202366828918457,
|
|
"rewards/margins": 0.7031639218330383,
|
|
"rewards/rejected": -2.0234005451202393,
|
|
"step": 253
|
|
},
|
|
{
|
|
"epoch": 0.37298091042584436,
|
|
"grad_norm": 14.38194751739502,
|
|
"kl/avg_steps": 0.359375,
|
|
"kl/beta": 0.032440755516290665,
|
|
"kl/n_epsilon_steps": 0.3125,
|
|
"kl/p_epsilon_steps": 0.671875,
|
|
"learning_rate": 3.965307091713037e-07,
|
|
"logits/chosen": -7.187813758850098,
|
|
"logits/rejected": -6.099554061889648,
|
|
"logps/chosen": -138.41323852539062,
|
|
"logps/ref_chosen": -98.73518371582031,
|
|
"logps/ref_rejected": -93.73825073242188,
|
|
"logps/rejected": -154.52410888671875,
|
|
"loss": 0.971,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -1.2845338582992554,
|
|
"rewards/margins": 0.676956057548523,
|
|
"rewards/rejected": -1.9614899158477783,
|
|
"step": 254
|
|
},
|
|
{
|
|
"epoch": 0.3744493392070485,
|
|
"grad_norm": 14.393223762512207,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.03232458978891373,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 3.954890003969163e-07,
|
|
"logits/chosen": -7.347094535827637,
|
|
"logits/rejected": -6.871660232543945,
|
|
"logps/chosen": -131.79598999023438,
|
|
"logps/ref_chosen": -90.382568359375,
|
|
"logps/ref_rejected": -97.07625579833984,
|
|
"logps/rejected": -166.75144958496094,
|
|
"loss": 0.8306,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.3301056623458862,
|
|
"rewards/margins": 0.9035694599151611,
|
|
"rewards/rejected": -2.233675003051758,
|
|
"step": 255
|
|
},
|
|
{
|
|
"epoch": 0.37591776798825255,
|
|
"grad_norm": 14.8760986328125,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.03209391236305237,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 3.944434578520628e-07,
|
|
"logits/chosen": -7.483323097229004,
|
|
"logits/rejected": -6.505797386169434,
|
|
"logps/chosen": -126.28778076171875,
|
|
"logps/ref_chosen": -88.7528076171875,
|
|
"logps/ref_rejected": -98.49382781982422,
|
|
"logps/rejected": -164.5001220703125,
|
|
"loss": 0.8817,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -1.1998153924942017,
|
|
"rewards/margins": 0.9028624296188354,
|
|
"rewards/rejected": -2.102677822113037,
|
|
"step": 256
|
|
},
|
|
{
|
|
"epoch": 0.37738619676945667,
|
|
"grad_norm": 15.880992889404297,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.03188467025756836,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 3.933941090877615e-07,
|
|
"logits/chosen": -7.26761531829834,
|
|
"logits/rejected": -6.992372512817383,
|
|
"logps/chosen": -122.74734497070312,
|
|
"logps/ref_chosen": -82.80352783203125,
|
|
"logps/ref_rejected": -85.8677978515625,
|
|
"logps/rejected": -153.03878784179688,
|
|
"loss": 0.8809,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.2686723470687866,
|
|
"rewards/margins": 0.8583362698554993,
|
|
"rewards/rejected": -2.1270086765289307,
|
|
"step": 257
|
|
},
|
|
{
|
|
"epoch": 0.3788546255506608,
|
|
"grad_norm": 14.805852890014648,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.031706321984529495,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 3.923409817553284e-07,
|
|
"logits/chosen": -7.284233093261719,
|
|
"logits/rejected": -6.5939788818359375,
|
|
"logps/chosen": -129.7963409423828,
|
|
"logps/ref_chosen": -90.187744140625,
|
|
"logps/ref_rejected": -103.47068786621094,
|
|
"logps/rejected": -169.39410400390625,
|
|
"loss": 0.8631,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -1.2512186765670776,
|
|
"rewards/margins": 0.8250081539154053,
|
|
"rewards/rejected": -2.0762269496917725,
|
|
"step": 258
|
|
},
|
|
{
|
|
"epoch": 0.3803230543318649,
|
|
"grad_norm": 15.601424217224121,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.03152897208929062,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 3.9128410360564793e-07,
|
|
"logits/chosen": -6.8948163986206055,
|
|
"logits/rejected": -6.679216384887695,
|
|
"logps/chosen": -133.95770263671875,
|
|
"logps/ref_chosen": -90.77254486083984,
|
|
"logps/ref_rejected": -94.58816528320312,
|
|
"logps/rejected": -159.41954040527344,
|
|
"loss": 0.9968,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -1.3569505214691162,
|
|
"rewards/margins": 0.6738921403884888,
|
|
"rewards/rejected": -2.0308425426483154,
|
|
"step": 259
|
|
},
|
|
{
|
|
"epoch": 0.38179148311306904,
|
|
"grad_norm": 16.810916900634766,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.03135261312127113,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 3.9022350248844246e-07,
|
|
"logits/chosen": -7.622888088226318,
|
|
"logits/rejected": -6.941500186920166,
|
|
"logps/chosen": -118.55705261230469,
|
|
"logps/ref_chosen": -75.59269714355469,
|
|
"logps/ref_rejected": -100.84554290771484,
|
|
"logps/rejected": -164.71981811523438,
|
|
"loss": 1.0562,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -1.3429489135742188,
|
|
"rewards/margins": 0.6456452012062073,
|
|
"rewards/rejected": -1.9885942935943604,
|
|
"step": 260
|
|
},
|
|
{
|
|
"epoch": 0.3832599118942731,
|
|
"grad_norm": 13.099295616149902,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.031196629628539085,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 3.891592063515376e-07,
|
|
"logits/chosen": -7.312009811401367,
|
|
"logits/rejected": -7.067169666290283,
|
|
"logps/chosen": -129.3551025390625,
|
|
"logps/ref_chosen": -94.25491333007812,
|
|
"logps/ref_rejected": -93.65699768066406,
|
|
"logps/rejected": -158.41415405273438,
|
|
"loss": 0.8049,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.0897305011749268,
|
|
"rewards/margins": 0.9158782958984375,
|
|
"rewards/rejected": -2.0056087970733643,
|
|
"step": 261
|
|
},
|
|
{
|
|
"epoch": 0.38472834067547723,
|
|
"grad_norm": 13.834671020507812,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.03099323809146881,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 3.880912432401264e-07,
|
|
"logits/chosen": -7.694709777832031,
|
|
"logits/rejected": -6.795375823974609,
|
|
"logps/chosen": -123.33172607421875,
|
|
"logps/ref_chosen": -85.26730346679688,
|
|
"logps/ref_rejected": -90.82609558105469,
|
|
"logps/rejected": -156.57522583007812,
|
|
"loss": 0.8587,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -1.1732381582260132,
|
|
"rewards/margins": 0.8486050367355347,
|
|
"rewards/rejected": -2.021843194961548,
|
|
"step": 262
|
|
},
|
|
{
|
|
"epoch": 0.38619676945668135,
|
|
"grad_norm": 13.023946762084961,
|
|
"kl/avg_steps": 0.75,
|
|
"kl/beta": 0.030781613662838936,
|
|
"kl/n_epsilon_steps": 0.125,
|
|
"kl/p_epsilon_steps": 0.875,
|
|
"learning_rate": 3.870196412960302e-07,
|
|
"logits/chosen": -6.901250839233398,
|
|
"logits/rejected": -6.722340106964111,
|
|
"logps/chosen": -127.14653015136719,
|
|
"logps/ref_chosen": -95.75790405273438,
|
|
"logps/ref_rejected": -101.83377075195312,
|
|
"logps/rejected": -169.84298706054688,
|
|
"loss": 0.7435,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -0.9609699845314026,
|
|
"rewards/margins": 1.114938497543335,
|
|
"rewards/rejected": -2.075908660888672,
|
|
"step": 263
|
|
},
|
|
{
|
|
"epoch": 0.3876651982378855,
|
|
"grad_norm": 13.047411918640137,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.030552471056580544,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 3.8594442875695665e-07,
|
|
"logits/chosen": -7.175429344177246,
|
|
"logits/rejected": -6.558530330657959,
|
|
"logps/chosen": -124.76637268066406,
|
|
"logps/ref_chosen": -90.6226577758789,
|
|
"logps/ref_rejected": -100.32554626464844,
|
|
"logps/rejected": -164.76083374023438,
|
|
"loss": 0.8463,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.0397236347198486,
|
|
"rewards/margins": 0.9148914217948914,
|
|
"rewards/rejected": -1.9546151161193848,
|
|
"step": 264
|
|
},
|
|
{
|
|
"epoch": 0.3891336270190896,
|
|
"grad_norm": 12.617591857910156,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.030381573364138603,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 3.848656339557562e-07,
|
|
"logits/chosen": -6.967649936676025,
|
|
"logits/rejected": -6.795099258422852,
|
|
"logps/chosen": -128.9173583984375,
|
|
"logps/ref_chosen": -92.37232971191406,
|
|
"logps/ref_rejected": -94.62757110595703,
|
|
"logps/rejected": -160.58474731445312,
|
|
"loss": 0.8792,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -1.1058483123779297,
|
|
"rewards/margins": 0.8837213516235352,
|
|
"rewards/rejected": -1.9895697832107544,
|
|
"step": 265
|
|
},
|
|
{
|
|
"epoch": 0.39060205580029367,
|
|
"grad_norm": 15.880788803100586,
|
|
"kl/avg_steps": 0.34375,
|
|
"kl/beta": 0.030211633071303368,
|
|
"kl/n_epsilon_steps": 0.328125,
|
|
"kl/p_epsilon_steps": 0.671875,
|
|
"learning_rate": 3.8378328531967507e-07,
|
|
"logits/chosen": -7.422516345977783,
|
|
"logits/rejected": -7.111500263214111,
|
|
"logps/chosen": -143.80422973632812,
|
|
"logps/ref_chosen": -102.20002746582031,
|
|
"logps/ref_rejected": -74.36642456054688,
|
|
"logps/rejected": -141.06625366210938,
|
|
"loss": 0.9138,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -1.2558131217956543,
|
|
"rewards/margins": 0.7501237392425537,
|
|
"rewards/rejected": -2.005937099456787,
|
|
"step": 266
|
|
},
|
|
{
|
|
"epoch": 0.3920704845814978,
|
|
"grad_norm": 13.670655250549316,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.0301081370562315,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 3.8269741136960646e-07,
|
|
"logits/chosen": -7.119365692138672,
|
|
"logits/rejected": -6.709759712219238,
|
|
"logps/chosen": -139.10177612304688,
|
|
"logps/ref_chosen": -104.28599548339844,
|
|
"logps/ref_rejected": -95.98719024658203,
|
|
"logps/rejected": -161.63311767578125,
|
|
"loss": 0.7978,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -1.04282808303833,
|
|
"rewards/margins": 0.9189317226409912,
|
|
"rewards/rejected": -1.9617595672607422,
|
|
"step": 267
|
|
},
|
|
{
|
|
"epoch": 0.3935389133627019,
|
|
"grad_norm": 13.244879722595215,
|
|
"kl/avg_steps": 0.75,
|
|
"kl/beta": 0.029911840334534645,
|
|
"kl/n_epsilon_steps": 0.125,
|
|
"kl/p_epsilon_steps": 0.875,
|
|
"learning_rate": 3.8160804071933894e-07,
|
|
"logits/chosen": -7.436177730560303,
|
|
"logits/rejected": -6.697609901428223,
|
|
"logps/chosen": -121.36919403076172,
|
|
"logps/ref_chosen": -86.69622039794922,
|
|
"logps/ref_rejected": -109.19183349609375,
|
|
"logps/rejected": -176.08946228027344,
|
|
"loss": 0.7894,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -1.030745029449463,
|
|
"rewards/margins": 0.9536941051483154,
|
|
"rewards/rejected": -1.9844391345977783,
|
|
"step": 268
|
|
},
|
|
{
|
|
"epoch": 0.39500734214390604,
|
|
"grad_norm": 13.288008689880371,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.02968917228281498,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 3.8051520207480204e-07,
|
|
"logits/chosen": -7.45705509185791,
|
|
"logits/rejected": -6.695377826690674,
|
|
"logps/chosen": -140.38177490234375,
|
|
"logps/ref_chosen": -104.97181701660156,
|
|
"logps/ref_rejected": -112.4764633178711,
|
|
"logps/rejected": -181.06021118164062,
|
|
"loss": 0.8031,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.0468546152114868,
|
|
"rewards/margins": 0.9742782115936279,
|
|
"rewards/rejected": -2.0211329460144043,
|
|
"step": 269
|
|
},
|
|
{
|
|
"epoch": 0.3964757709251101,
|
|
"grad_norm": 13.801987648010254,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.029523104429244995,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 3.794189242333106e-07,
|
|
"logits/chosen": -7.553506851196289,
|
|
"logits/rejected": -6.664368152618408,
|
|
"logps/chosen": -139.99703979492188,
|
|
"logps/ref_chosen": -101.07383728027344,
|
|
"logps/ref_rejected": -117.75289916992188,
|
|
"logps/rejected": -186.19866943359375,
|
|
"loss": 0.9174,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -1.1459414958953857,
|
|
"rewards/margins": 0.8607551455497742,
|
|
"rewards/rejected": -2.0066967010498047,
|
|
"step": 270
|
|
},
|
|
{
|
|
"epoch": 0.39794419970631423,
|
|
"grad_norm": 12.916444778442383,
|
|
"kl/avg_steps": 0.75,
|
|
"kl/beta": 0.029357966035604477,
|
|
"kl/n_epsilon_steps": 0.125,
|
|
"kl/p_epsilon_steps": 0.875,
|
|
"learning_rate": 3.7831923608280514e-07,
|
|
"logits/chosen": -6.938437461853027,
|
|
"logits/rejected": -6.5058183670043945,
|
|
"logps/chosen": -129.179443359375,
|
|
"logps/ref_chosen": -96.72459411621094,
|
|
"logps/ref_rejected": -98.5244140625,
|
|
"logps/rejected": -166.06207275390625,
|
|
"loss": 0.8123,
|
|
"rewards/accuracies": 0.890625,
|
|
"rewards/chosen": -0.9474867582321167,
|
|
"rewards/margins": 1.0177432298660278,
|
|
"rewards/rejected": -1.9652299880981445,
|
|
"step": 271
|
|
},
|
|
{
|
|
"epoch": 0.39941262848751835,
|
|
"grad_norm": 14.018041610717773,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.02913942001760006,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 3.772161666010912e-07,
|
|
"logits/chosen": -7.514509677886963,
|
|
"logits/rejected": -7.041294574737549,
|
|
"logps/chosen": -108.3265380859375,
|
|
"logps/ref_chosen": -80.97721862792969,
|
|
"logps/ref_rejected": -108.55535888671875,
|
|
"logps/rejected": -169.4478759765625,
|
|
"loss": 0.8282,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -0.7937983870506287,
|
|
"rewards/margins": 0.9667574167251587,
|
|
"rewards/rejected": -1.7605557441711426,
|
|
"step": 272
|
|
},
|
|
{
|
|
"epoch": 0.4008810572687225,
|
|
"grad_norm": 13.580171585083008,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.028958430513739586,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 3.761097448550755e-07,
|
|
"logits/chosen": -7.261774063110352,
|
|
"logits/rejected": -6.644618988037109,
|
|
"logps/chosen": -128.26251220703125,
|
|
"logps/ref_chosen": -92.22460174560547,
|
|
"logps/ref_rejected": -97.3630599975586,
|
|
"logps/rejected": -168.29635620117188,
|
|
"loss": 0.8615,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.0409189462661743,
|
|
"rewards/margins": 0.9985566139221191,
|
|
"rewards/rejected": -2.039475440979004,
|
|
"step": 273
|
|
},
|
|
{
|
|
"epoch": 0.4023494860499266,
|
|
"grad_norm": 15.386717796325684,
|
|
"kl/avg_steps": 0.75,
|
|
"kl/beta": 0.028787503018975258,
|
|
"kl/n_epsilon_steps": 0.125,
|
|
"kl/p_epsilon_steps": 0.875,
|
|
"learning_rate": 3.75e-07,
|
|
"logits/chosen": -7.7138991355896,
|
|
"logits/rejected": -6.974665641784668,
|
|
"logps/chosen": -119.94572448730469,
|
|
"logps/ref_chosen": -82.2608871459961,
|
|
"logps/ref_rejected": -83.87699127197266,
|
|
"logps/rejected": -149.46990966796875,
|
|
"loss": 0.8819,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -1.0782254934310913,
|
|
"rewards/margins": 0.7943964004516602,
|
|
"rewards/rejected": -1.8726218938827515,
|
|
"step": 274
|
|
},
|
|
{
|
|
"epoch": 0.40381791483113066,
|
|
"grad_norm": 15.320517539978027,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.028573205694556236,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 3.738869612786737e-07,
|
|
"logits/chosen": -7.628604888916016,
|
|
"logits/rejected": -6.829320907592773,
|
|
"logps/chosen": -111.30049896240234,
|
|
"logps/ref_chosen": -79.68695831298828,
|
|
"logps/ref_rejected": -98.7509765625,
|
|
"logps/rejected": -160.444580078125,
|
|
"loss": 0.9047,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.8999394178390503,
|
|
"rewards/margins": 0.8488346338272095,
|
|
"rewards/rejected": -1.7487740516662598,
|
|
"step": 275
|
|
},
|
|
{
|
|
"epoch": 0.4052863436123348,
|
|
"grad_norm": 15.482325553894043,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.028369300067424774,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 3.7277065802070204e-07,
|
|
"logits/chosen": -7.228045463562012,
|
|
"logits/rejected": -6.846656799316406,
|
|
"logps/chosen": -126.2287826538086,
|
|
"logps/ref_chosen": -86.53970336914062,
|
|
"logps/ref_rejected": -77.85394287109375,
|
|
"logps/rejected": -146.60427856445312,
|
|
"loss": 0.9138,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.1226071119308472,
|
|
"rewards/margins": 0.8151148557662964,
|
|
"rewards/rejected": -1.9377222061157227,
|
|
"step": 276
|
|
},
|
|
{
|
|
"epoch": 0.4067547723935389,
|
|
"grad_norm": 15.835052490234375,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.02821938507258892,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 3.71651119641714e-07,
|
|
"logits/chosen": -7.30094051361084,
|
|
"logits/rejected": -6.272882461547852,
|
|
"logps/chosen": -123.2606430053711,
|
|
"logps/ref_chosen": -84.24411010742188,
|
|
"logps/ref_rejected": -98.83421325683594,
|
|
"logps/rejected": -164.99749755859375,
|
|
"loss": 0.898,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -1.0951545238494873,
|
|
"rewards/margins": 0.7574270963668823,
|
|
"rewards/rejected": -1.8525817394256592,
|
|
"step": 277
|
|
},
|
|
{
|
|
"epoch": 0.40822320117474303,
|
|
"grad_norm": 13.882027626037598,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.028035402297973633,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 3.705283756425872e-07,
|
|
"logits/chosen": -7.213944435119629,
|
|
"logits/rejected": -6.768559455871582,
|
|
"logps/chosen": -120.13731384277344,
|
|
"logps/ref_chosen": -82.431884765625,
|
|
"logps/ref_rejected": -97.85691833496094,
|
|
"logps/rejected": -164.93141174316406,
|
|
"loss": 0.9088,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -1.0521965026855469,
|
|
"rewards/margins": 0.8135128021240234,
|
|
"rewards/rejected": -1.8657093048095703,
|
|
"step": 278
|
|
},
|
|
{
|
|
"epoch": 0.40969162995594716,
|
|
"grad_norm": 13.638445854187012,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.02785261906683445,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 3.6940245560867e-07,
|
|
"logits/chosen": -7.466989517211914,
|
|
"logits/rejected": -7.357805252075195,
|
|
"logps/chosen": -125.54033660888672,
|
|
"logps/ref_chosen": -85.16799926757812,
|
|
"logps/ref_rejected": -94.12664794921875,
|
|
"logps/rejected": -165.55284118652344,
|
|
"loss": 0.8719,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -1.119484305381775,
|
|
"rewards/margins": 0.8550732135772705,
|
|
"rewards/rejected": -1.9745573997497559,
|
|
"step": 279
|
|
},
|
|
{
|
|
"epoch": 0.4111600587371512,
|
|
"grad_norm": 13.675751686096191,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.027688222005963326,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 3.6827338920900253e-07,
|
|
"logits/chosen": -7.190787315368652,
|
|
"logits/rejected": -6.597271919250488,
|
|
"logps/chosen": -123.91621398925781,
|
|
"logps/ref_chosen": -85.85641479492188,
|
|
"logps/ref_rejected": -104.11859130859375,
|
|
"logps/rejected": -177.7848663330078,
|
|
"loss": 0.7947,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -1.0486705303192139,
|
|
"rewards/margins": 0.9753453731536865,
|
|
"rewards/rejected": -2.0240159034729004,
|
|
"step": 280
|
|
},
|
|
{
|
|
"epoch": 0.41262848751835535,
|
|
"grad_norm": 11.921332359313965,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.02751624397933483,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 3.6714120619553435e-07,
|
|
"logits/chosen": -7.7915472984313965,
|
|
"logits/rejected": -6.586763381958008,
|
|
"logps/chosen": -125.37873840332031,
|
|
"logps/ref_chosen": -86.55081939697266,
|
|
"logps/ref_rejected": -88.62866973876953,
|
|
"logps/rejected": -164.1964874267578,
|
|
"loss": 0.8278,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.0626280307769775,
|
|
"rewards/margins": 0.9996333718299866,
|
|
"rewards/rejected": -2.0622613430023193,
|
|
"step": 281
|
|
},
|
|
{
|
|
"epoch": 0.41409691629955947,
|
|
"grad_norm": 11.661370277404785,
|
|
"kl/avg_steps": 0.8125,
|
|
"kl/beta": 0.027328362688422203,
|
|
"kl/n_epsilon_steps": 0.09375,
|
|
"kl/p_epsilon_steps": 0.90625,
|
|
"learning_rate": 3.660059364023408e-07,
|
|
"logits/chosen": -7.430984973907471,
|
|
"logits/rejected": -6.962902545928955,
|
|
"logps/chosen": -143.371826171875,
|
|
"logps/ref_chosen": -105.10511016845703,
|
|
"logps/ref_rejected": -102.85336303710938,
|
|
"logps/rejected": -183.70306396484375,
|
|
"loss": 0.6931,
|
|
"rewards/accuracies": 0.953125,
|
|
"rewards/chosen": -1.037402868270874,
|
|
"rewards/margins": 1.152178168296814,
|
|
"rewards/rejected": -2.1895809173583984,
|
|
"step": 282
|
|
},
|
|
{
|
|
"epoch": 0.4155653450807636,
|
|
"grad_norm": 17.020105361938477,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.02710810862481594,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 3.6486760974483685e-07,
|
|
"logits/chosen": -7.325111389160156,
|
|
"logits/rejected": -6.981871604919434,
|
|
"logps/chosen": -139.34780883789062,
|
|
"logps/ref_chosen": -95.05259704589844,
|
|
"logps/ref_rejected": -103.54454803466797,
|
|
"logps/rejected": -177.7033233642578,
|
|
"loss": 0.9734,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -1.1950119733810425,
|
|
"rewards/margins": 0.7993054389953613,
|
|
"rewards/rejected": -1.9943175315856934,
|
|
"step": 283
|
|
},
|
|
{
|
|
"epoch": 0.4170337738619677,
|
|
"grad_norm": 11.532405853271484,
|
|
"kl/avg_steps": 0.78125,
|
|
"kl/beta": 0.02693137153983116,
|
|
"kl/n_epsilon_steps": 0.109375,
|
|
"kl/p_epsilon_steps": 0.890625,
|
|
"learning_rate": 3.6372625621898863e-07,
|
|
"logits/chosen": -7.086446762084961,
|
|
"logits/rejected": -6.687787055969238,
|
|
"logps/chosen": -124.68060302734375,
|
|
"logps/ref_chosen": -87.6664810180664,
|
|
"logps/ref_rejected": -98.75103759765625,
|
|
"logps/rejected": -177.91339111328125,
|
|
"loss": 0.6944,
|
|
"rewards/accuracies": 0.921875,
|
|
"rewards/chosen": -0.9900509119033813,
|
|
"rewards/margins": 1.1234304904937744,
|
|
"rewards/rejected": -2.1134815216064453,
|
|
"step": 284
|
|
},
|
|
{
|
|
"epoch": 0.4185022026431718,
|
|
"grad_norm": 18.86406898498535,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.026722600683569908,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 3.625819059005228e-07,
|
|
"logits/chosen": -7.379509449005127,
|
|
"logits/rejected": -7.0911407470703125,
|
|
"logps/chosen": -137.629150390625,
|
|
"logps/ref_chosen": -94.43303680419922,
|
|
"logps/ref_rejected": -104.07194519042969,
|
|
"logps/rejected": -176.97714233398438,
|
|
"loss": 0.8825,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.1481690406799316,
|
|
"rewards/margins": 0.7845076322555542,
|
|
"rewards/rejected": -1.9326767921447754,
|
|
"step": 285
|
|
},
|
|
{
|
|
"epoch": 0.4199706314243759,
|
|
"grad_norm": 14.356172561645508,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.026540137827396393,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 3.614345889441346e-07,
|
|
"logits/chosen": -7.472691535949707,
|
|
"logits/rejected": -7.013616561889648,
|
|
"logps/chosen": -149.8736572265625,
|
|
"logps/ref_chosen": -103.72039794921875,
|
|
"logps/ref_rejected": -96.25775909423828,
|
|
"logps/rejected": -179.59671020507812,
|
|
"loss": 0.8014,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.2212910652160645,
|
|
"rewards/margins": 0.9765151143074036,
|
|
"rewards/rejected": -2.1978063583374023,
|
|
"step": 286
|
|
},
|
|
{
|
|
"epoch": 0.42143906020558003,
|
|
"grad_norm": 15.217528343200684,
|
|
"kl/avg_steps": 0.34375,
|
|
"kl/beta": 0.026391685009002686,
|
|
"kl/n_epsilon_steps": 0.328125,
|
|
"kl/p_epsilon_steps": 0.671875,
|
|
"learning_rate": 3.6028433558269275e-07,
|
|
"logits/chosen": -7.295309066772461,
|
|
"logits/rejected": -6.699875831604004,
|
|
"logps/chosen": -147.86685180664062,
|
|
"logps/ref_chosen": -93.88988494873047,
|
|
"logps/ref_rejected": -83.33365631103516,
|
|
"logps/rejected": -162.0262451171875,
|
|
"loss": 1.0332,
|
|
"rewards/accuracies": 0.71875,
|
|
"rewards/chosen": -1.4231257438659668,
|
|
"rewards/margins": 0.6437419652938843,
|
|
"rewards/rejected": -2.0668678283691406,
|
|
"step": 287
|
|
},
|
|
{
|
|
"epoch": 0.42290748898678415,
|
|
"grad_norm": 14.122811317443848,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.026301274076104164,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 3.5913117612644327e-07,
|
|
"logits/chosen": -7.697492599487305,
|
|
"logits/rejected": -7.256779670715332,
|
|
"logps/chosen": -139.82644653320312,
|
|
"logps/ref_chosen": -88.15602111816406,
|
|
"logps/ref_rejected": -93.28195190429688,
|
|
"logps/rejected": -183.26959228515625,
|
|
"loss": 0.816,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.3554046154022217,
|
|
"rewards/margins": 0.9966922998428345,
|
|
"rewards/rejected": -2.3520970344543457,
|
|
"step": 288
|
|
},
|
|
{
|
|
"epoch": 0.4243759177679883,
|
|
"grad_norm": 13.091045379638672,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.02616228722035885,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 3.5797514096221024e-07,
|
|
"logits/chosen": -7.527945518493652,
|
|
"logits/rejected": -7.149833679199219,
|
|
"logps/chosen": -126.64739227294922,
|
|
"logps/ref_chosen": -75.39292907714844,
|
|
"logps/ref_rejected": -93.15428161621094,
|
|
"logps/rejected": -185.3885498046875,
|
|
"loss": 0.8297,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -1.3352906703948975,
|
|
"rewards/margins": 1.0608309507369995,
|
|
"rewards/rejected": -2.3961217403411865,
|
|
"step": 289
|
|
},
|
|
{
|
|
"epoch": 0.42584434654919234,
|
|
"grad_norm": 14.055940628051758,
|
|
"kl/avg_steps": 0.8125,
|
|
"kl/beta": 0.026007864624261856,
|
|
"kl/n_epsilon_steps": 0.09375,
|
|
"kl/p_epsilon_steps": 0.90625,
|
|
"learning_rate": 3.568162605525952e-07,
|
|
"logits/chosen": -7.388311386108398,
|
|
"logits/rejected": -7.0829925537109375,
|
|
"logps/chosen": -144.79820251464844,
|
|
"logps/ref_chosen": -88.0419692993164,
|
|
"logps/ref_rejected": -123.21215057373047,
|
|
"logps/rejected": -221.4034423828125,
|
|
"loss": 0.8033,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -1.4665530920028687,
|
|
"rewards/margins": 1.0655090808868408,
|
|
"rewards/rejected": -2.53206205368042,
|
|
"step": 290
|
|
},
|
|
{
|
|
"epoch": 0.42731277533039647,
|
|
"grad_norm": 14.552366256713867,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.0257982537150383,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 3.5565456543517485e-07,
|
|
"logits/chosen": -7.654292106628418,
|
|
"logits/rejected": -6.923820972442627,
|
|
"logps/chosen": -148.91311645507812,
|
|
"logps/ref_chosen": -94.09524536132812,
|
|
"logps/ref_rejected": -96.05006408691406,
|
|
"logps/rejected": -183.22817993164062,
|
|
"loss": 0.9127,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -1.408342719078064,
|
|
"rewards/margins": 0.8246374130249023,
|
|
"rewards/rejected": -2.232980251312256,
|
|
"step": 291
|
|
},
|
|
{
|
|
"epoch": 0.4287812041116006,
|
|
"grad_norm": 12.810431480407715,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.025638015940785408,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 3.5449008622169583e-07,
|
|
"logits/chosen": -7.4952850341796875,
|
|
"logits/rejected": -7.096066951751709,
|
|
"logps/chosen": -142.94708251953125,
|
|
"logps/ref_chosen": -88.25041198730469,
|
|
"logps/ref_rejected": -96.41764068603516,
|
|
"logps/rejected": -189.70559692382812,
|
|
"loss": 0.7522,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -1.3947566747665405,
|
|
"rewards/margins": 0.9789954423904419,
|
|
"rewards/rejected": -2.3737521171569824,
|
|
"step": 292
|
|
},
|
|
{
|
|
"epoch": 0.4302496328928047,
|
|
"grad_norm": 13.038102149963379,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.025462958961725235,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 3.5332285359726846e-07,
|
|
"logits/chosen": -7.2822585105896,
|
|
"logits/rejected": -6.593290328979492,
|
|
"logps/chosen": -146.07388305664062,
|
|
"logps/ref_chosen": -87.37654876708984,
|
|
"logps/ref_rejected": -85.75579833984375,
|
|
"logps/rejected": -172.91517639160156,
|
|
"loss": 0.9942,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -1.4899965524673462,
|
|
"rewards/margins": 0.7153399586677551,
|
|
"rewards/rejected": -2.205336570739746,
|
|
"step": 293
|
|
},
|
|
{
|
|
"epoch": 0.43171806167400884,
|
|
"grad_norm": 14.321362495422363,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.02532840147614479,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 3.5215289831955786e-07,
|
|
"logits/chosen": -6.959033966064453,
|
|
"logits/rejected": -6.0898590087890625,
|
|
"logps/chosen": -134.58004760742188,
|
|
"logps/ref_chosen": -73.5079574584961,
|
|
"logps/ref_rejected": -88.08877563476562,
|
|
"logps/rejected": -174.16912841796875,
|
|
"loss": 1.0529,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -1.5407367944717407,
|
|
"rewards/margins": 0.6243571043014526,
|
|
"rewards/rejected": -2.1650938987731934,
|
|
"step": 294
|
|
},
|
|
{
|
|
"epoch": 0.4331864904552129,
|
|
"grad_norm": 13.053278923034668,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.025194555521011353,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 3.509802512179737e-07,
|
|
"logits/chosen": -7.271920680999756,
|
|
"logits/rejected": -7.100852012634277,
|
|
"logps/chosen": -136.8726043701172,
|
|
"logps/ref_chosen": -77.76548767089844,
|
|
"logps/ref_rejected": -94.24726867675781,
|
|
"logps/rejected": -189.65672302246094,
|
|
"loss": 0.8606,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.4831101894378662,
|
|
"rewards/margins": 0.9041118025779724,
|
|
"rewards/rejected": -2.3872218132019043,
|
|
"step": 295
|
|
},
|
|
{
|
|
"epoch": 0.434654919236417,
|
|
"grad_norm": 13.810120582580566,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.02504584565758705,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 3.498049431928577e-07,
|
|
"logits/chosen": -7.517086505889893,
|
|
"logits/rejected": -7.0501275062561035,
|
|
"logps/chosen": -161.45822143554688,
|
|
"logps/ref_chosen": -97.85641479492188,
|
|
"logps/ref_rejected": -100.81631469726562,
|
|
"logps/rejected": -203.63685607910156,
|
|
"loss": 0.8194,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.586834192276001,
|
|
"rewards/margins": 0.9714199304580688,
|
|
"rewards/rejected": -2.5582542419433594,
|
|
"step": 296
|
|
},
|
|
{
|
|
"epoch": 0.43612334801762115,
|
|
"grad_norm": 14.663503646850586,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.024905750527977943,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 3.486270052146694e-07,
|
|
"logits/chosen": -7.523492813110352,
|
|
"logits/rejected": -6.788856506347656,
|
|
"logps/chosen": -154.54638671875,
|
|
"logps/ref_chosen": -88.56583404541016,
|
|
"logps/ref_rejected": -101.55656433105469,
|
|
"logps/rejected": -200.7189178466797,
|
|
"loss": 0.8867,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.6361867189407349,
|
|
"rewards/margins": 0.8170878887176514,
|
|
"rewards/rejected": -2.453274726867676,
|
|
"step": 297
|
|
},
|
|
{
|
|
"epoch": 0.43759177679882527,
|
|
"grad_norm": 16.382911682128906,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.02476643957197666,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 3.474464683231698e-07,
|
|
"logits/chosen": -7.475191593170166,
|
|
"logits/rejected": -6.408700942993164,
|
|
"logps/chosen": -154.0912322998047,
|
|
"logps/ref_chosen": -94.88043975830078,
|
|
"logps/ref_rejected": -122.31101989746094,
|
|
"logps/rejected": -221.15391540527344,
|
|
"loss": 0.8289,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.460301160812378,
|
|
"rewards/margins": 0.9708787202835083,
|
|
"rewards/rejected": -2.431180000305176,
|
|
"step": 298
|
|
},
|
|
{
|
|
"epoch": 0.4390602055800294,
|
|
"grad_norm": 12.90149974822998,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.024620257318019867,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 3.462633636266041e-07,
|
|
"logits/chosen": -7.417859077453613,
|
|
"logits/rejected": -6.741024971008301,
|
|
"logps/chosen": -142.339111328125,
|
|
"logps/ref_chosen": -80.40835571289062,
|
|
"logps/ref_rejected": -89.53716278076172,
|
|
"logps/rejected": -193.05116271972656,
|
|
"loss": 0.785,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -1.5161972045898438,
|
|
"rewards/margins": 1.0127980709075928,
|
|
"rewards/rejected": -2.5289955139160156,
|
|
"step": 299
|
|
},
|
|
{
|
|
"epoch": 0.44052863436123346,
|
|
"grad_norm": 14.46535587310791,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.024444561451673508,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 3.4507772230088147e-07,
|
|
"logits/chosen": -7.5312700271606445,
|
|
"logits/rejected": -7.016364097595215,
|
|
"logps/chosen": -153.97088623046875,
|
|
"logps/ref_chosen": -88.15890502929688,
|
|
"logps/ref_rejected": -100.93919372558594,
|
|
"logps/rejected": -202.05059814453125,
|
|
"loss": 0.912,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.6017869710922241,
|
|
"rewards/margins": 0.8526525497436523,
|
|
"rewards/rejected": -2.454439640045166,
|
|
"step": 300
|
|
},
|
|
{
|
|
"epoch": 0.44052863436123346,
|
|
"eval_kl/n_epsilon_steps": 0.24700342118740082,
|
|
"eval_kl/p_epsilon_steps": 0.7508561611175537,
|
|
"eval_logits/chosen": -7.007012367248535,
|
|
"eval_logits/rejected": -6.721932888031006,
|
|
"eval_logps/chosen": -167.77142333984375,
|
|
"eval_logps/ref_chosen": -100.49356842041016,
|
|
"eval_logps/ref_rejected": -94.06775665283203,
|
|
"eval_logps/rejected": -191.32177734375,
|
|
"eval_loss": 0.4874841272830963,
|
|
"eval_rewards/accuracies": 0.7786815166473389,
|
|
"eval_rewards/chosen": -1.6289423704147339,
|
|
"eval_rewards/margins": 0.7190282940864563,
|
|
"eval_rewards/rejected": -2.347970724105835,
|
|
"eval_runtime": 48.5566,
|
|
"eval_samples_per_second": 48.171,
|
|
"eval_steps_per_second": 1.524,
|
|
"step": 300
|
|
},
|
|
{
|
|
"epoch": 0.4419970631424376,
|
|
"grad_norm": 13.402172088623047,
|
|
"kl/avg_steps": 0.609375,
|
|
"kl/beta": 0.024292731657624245,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 3.4388957558875316e-07,
|
|
"logits/chosen": -7.663355827331543,
|
|
"logits/rejected": -6.919793128967285,
|
|
"logps/chosen": -146.97164916992188,
|
|
"logps/ref_chosen": -85.50491333007812,
|
|
"logps/ref_rejected": -98.9168472290039,
|
|
"logps/rejected": -203.77188110351562,
|
|
"loss": 0.7859,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.4873508214950562,
|
|
"rewards/margins": 1.0426502227783203,
|
|
"rewards/rejected": -2.530000925064087,
|
|
"step": 301
|
|
},
|
|
{
|
|
"epoch": 0.4434654919236417,
|
|
"grad_norm": 15.27524185180664,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.024145595729351044,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 3.426989547989902e-07,
|
|
"logits/chosen": -7.2973504066467285,
|
|
"logits/rejected": -6.5177412033081055,
|
|
"logps/chosen": -142.61880493164062,
|
|
"logps/ref_chosen": -79.45040893554688,
|
|
"logps/ref_rejected": -103.05909729003906,
|
|
"logps/rejected": -201.1916046142578,
|
|
"loss": 0.9129,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.5182812213897705,
|
|
"rewards/margins": 0.8340890407562256,
|
|
"rewards/rejected": -2.352370262145996,
|
|
"step": 302
|
|
},
|
|
{
|
|
"epoch": 0.44493392070484583,
|
|
"grad_norm": 12.639718055725098,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.023995622992515564,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 3.4150589130555773e-07,
|
|
"logits/chosen": -7.532470703125,
|
|
"logits/rejected": -7.15334939956665,
|
|
"logps/chosen": -152.356689453125,
|
|
"logps/ref_chosen": -95.59829711914062,
|
|
"logps/ref_rejected": -92.0645751953125,
|
|
"logps/rejected": -189.04776000976562,
|
|
"loss": 0.8158,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -1.356937289237976,
|
|
"rewards/margins": 0.9546318650245667,
|
|
"rewards/rejected": -2.3115692138671875,
|
|
"step": 303
|
|
},
|
|
{
|
|
"epoch": 0.44640234948604995,
|
|
"grad_norm": 13.857175827026367,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.02386140264570713,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 3.403104165467883e-07,
|
|
"logits/chosen": -7.706064224243164,
|
|
"logits/rejected": -7.303454399108887,
|
|
"logps/chosen": -156.93124389648438,
|
|
"logps/ref_chosen": -99.4531478881836,
|
|
"logps/ref_rejected": -98.08226013183594,
|
|
"logps/rejected": -197.8302001953125,
|
|
"loss": 0.7929,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.365452766418457,
|
|
"rewards/margins": 0.9975064396858215,
|
|
"rewards/rejected": -2.362959146499634,
|
|
"step": 304
|
|
},
|
|
{
|
|
"epoch": 0.447870778267254,
|
|
"grad_norm": 13.020018577575684,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.02372056059539318,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 3.391125620245535e-07,
|
|
"logits/chosen": -7.775443077087402,
|
|
"logits/rejected": -6.882397174835205,
|
|
"logps/chosen": -162.6031494140625,
|
|
"logps/ref_chosen": -103.26339721679688,
|
|
"logps/ref_rejected": -97.82897186279297,
|
|
"logps/rejected": -199.2515869140625,
|
|
"loss": 0.8001,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.4017119407653809,
|
|
"rewards/margins": 0.9873659610748291,
|
|
"rewards/rejected": -2.38907790184021,
|
|
"step": 305
|
|
},
|
|
{
|
|
"epoch": 0.44933920704845814,
|
|
"grad_norm": 12.147577285766602,
|
|
"kl/avg_steps": 0.84375,
|
|
"kl/beta": 0.023580551147460938,
|
|
"kl/n_epsilon_steps": 0.078125,
|
|
"kl/p_epsilon_steps": 0.921875,
|
|
"learning_rate": 3.3791235930343417e-07,
|
|
"logits/chosen": -7.674917697906494,
|
|
"logits/rejected": -7.210516452789307,
|
|
"logps/chosen": -152.4990234375,
|
|
"logps/ref_chosen": -100.3038330078125,
|
|
"logps/ref_rejected": -92.23086547851562,
|
|
"logps/rejected": -195.33135986328125,
|
|
"loss": 0.6474,
|
|
"rewards/accuracies": 0.9375,
|
|
"rewards/chosen": -1.2215442657470703,
|
|
"rewards/margins": 1.1882264614105225,
|
|
"rewards/rejected": -2.4097707271575928,
|
|
"step": 306
|
|
},
|
|
{
|
|
"epoch": 0.45080763582966227,
|
|
"grad_norm": 14.133451461791992,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.023383256047964096,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 3.367098400098881e-07,
|
|
"logits/chosen": -7.329036235809326,
|
|
"logits/rejected": -7.163274765014648,
|
|
"logps/chosen": -155.960205078125,
|
|
"logps/ref_chosen": -96.64976501464844,
|
|
"logps/ref_rejected": -93.34860229492188,
|
|
"logps/rejected": -189.1453857421875,
|
|
"loss": 0.9218,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -1.3819317817687988,
|
|
"rewards/margins": 0.8423360586166382,
|
|
"rewards/rejected": -2.2242679595947266,
|
|
"step": 307
|
|
},
|
|
{
|
|
"epoch": 0.4522760646108664,
|
|
"grad_norm": 15.209975242614746,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.023259688168764114,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 3.355050358314172e-07,
|
|
"logits/chosen": -7.033696174621582,
|
|
"logits/rejected": -7.267401218414307,
|
|
"logps/chosen": -146.19422912597656,
|
|
"logps/ref_chosen": -85.72467041015625,
|
|
"logps/ref_rejected": -86.42318725585938,
|
|
"logps/rejected": -183.3817596435547,
|
|
"loss": 0.9237,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -1.399101734161377,
|
|
"rewards/margins": 0.8382473587989807,
|
|
"rewards/rejected": -2.237349033355713,
|
|
"step": 308
|
|
},
|
|
{
|
|
"epoch": 0.45374449339207046,
|
|
"grad_norm": 14.02563762664795,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.023115217685699463,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 3.3429797851573183e-07,
|
|
"logits/chosen": -7.613210201263428,
|
|
"logits/rejected": -7.04583215713501,
|
|
"logps/chosen": -148.60951232910156,
|
|
"logps/ref_chosen": -83.15145874023438,
|
|
"logps/ref_rejected": -86.58602142333984,
|
|
"logps/rejected": -190.06564331054688,
|
|
"loss": 0.8829,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.5072109699249268,
|
|
"rewards/margins": 0.8685282468795776,
|
|
"rewards/rejected": -2.375739097595215,
|
|
"step": 309
|
|
},
|
|
{
|
|
"epoch": 0.4552129221732746,
|
|
"grad_norm": 13.904105186462402,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.022978780791163445,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 3.3308869986991487e-07,
|
|
"logits/chosen": -7.28669548034668,
|
|
"logits/rejected": -7.22123908996582,
|
|
"logps/chosen": -148.03106689453125,
|
|
"logps/ref_chosen": -89.73799133300781,
|
|
"logps/ref_rejected": -87.07083129882812,
|
|
"logps/rejected": -181.75921630859375,
|
|
"loss": 0.9143,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -1.3350245952606201,
|
|
"rewards/margins": 0.826156735420227,
|
|
"rewards/rejected": -2.1611814498901367,
|
|
"step": 310
|
|
},
|
|
{
|
|
"epoch": 0.4566813509544787,
|
|
"grad_norm": 13.777924537658691,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.02285735122859478,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 3.3187723175958346e-07,
|
|
"logits/chosen": -7.765219688415527,
|
|
"logits/rejected": -6.731596946716309,
|
|
"logps/chosen": -159.7617645263672,
|
|
"logps/ref_chosen": -88.52644348144531,
|
|
"logps/ref_rejected": -82.26608276367188,
|
|
"logps/rejected": -199.7523651123047,
|
|
"loss": 0.7578,
|
|
"rewards/accuracies": 0.890625,
|
|
"rewards/chosen": -1.6187620162963867,
|
|
"rewards/margins": 1.046243667602539,
|
|
"rewards/rejected": -2.665005683898926,
|
|
"step": 311
|
|
},
|
|
{
|
|
"epoch": 0.4581497797356828,
|
|
"grad_norm": 13.674013137817383,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.022708328440785408,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 3.306636061080487e-07,
|
|
"logits/chosen": -7.522071838378906,
|
|
"logits/rejected": -6.754973411560059,
|
|
"logps/chosen": -146.90805053710938,
|
|
"logps/ref_chosen": -80.40069580078125,
|
|
"logps/ref_rejected": -82.4471435546875,
|
|
"logps/rejected": -192.24771118164062,
|
|
"loss": 0.8414,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -1.5027642250061035,
|
|
"rewards/margins": 0.9715795516967773,
|
|
"rewards/rejected": -2.474343776702881,
|
|
"step": 312
|
|
},
|
|
{
|
|
"epoch": 0.45961820851688695,
|
|
"grad_norm": 12.186444282531738,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.022553274407982826,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 3.2944785489547537e-07,
|
|
"logits/chosen": -7.390935897827148,
|
|
"logits/rejected": -6.827203750610352,
|
|
"logps/chosen": -146.6564483642578,
|
|
"logps/ref_chosen": -84.14190673828125,
|
|
"logps/ref_rejected": -93.42684936523438,
|
|
"logps/rejected": -194.61727905273438,
|
|
"loss": 0.9267,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -1.4026517868041992,
|
|
"rewards/margins": 0.8612587451934814,
|
|
"rewards/rejected": -2.2639105319976807,
|
|
"step": 313
|
|
},
|
|
{
|
|
"epoch": 0.461086637298091,
|
|
"grad_norm": 13.97519588470459,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.02242712303996086,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 3.2823001015803857e-07,
|
|
"logits/chosen": -7.827272415161133,
|
|
"logits/rejected": -7.375405311584473,
|
|
"logps/chosen": -151.27520751953125,
|
|
"logps/ref_chosen": -88.80972290039062,
|
|
"logps/ref_rejected": -104.15892028808594,
|
|
"logps/rejected": -207.26962280273438,
|
|
"loss": 0.872,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -1.3944957256317139,
|
|
"rewards/margins": 0.9006831049919128,
|
|
"rewards/rejected": -2.2951788902282715,
|
|
"step": 314
|
|
},
|
|
{
|
|
"epoch": 0.46255506607929514,
|
|
"grad_norm": 14.756415367126465,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.022287823259830475,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 3.270101039870797e-07,
|
|
"logits/chosen": -7.049970626831055,
|
|
"logits/rejected": -6.526283264160156,
|
|
"logps/chosen": -143.4265594482422,
|
|
"logps/ref_chosen": -80.793701171875,
|
|
"logps/ref_rejected": -92.13711547851562,
|
|
"logps/rejected": -195.54171752929688,
|
|
"loss": 0.9216,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -1.391645073890686,
|
|
"rewards/margins": 0.8969128131866455,
|
|
"rewards/rejected": -2.288558006286621,
|
|
"step": 315
|
|
},
|
|
{
|
|
"epoch": 0.46402349486049926,
|
|
"grad_norm": 14.08152961730957,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.022163156419992447,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 3.2578816852826086e-07,
|
|
"logits/chosen": -7.215094089508057,
|
|
"logits/rejected": -7.132869720458984,
|
|
"logps/chosen": -150.78274536132812,
|
|
"logps/ref_chosen": -84.18425750732422,
|
|
"logps/ref_rejected": -105.6908187866211,
|
|
"logps/rejected": -212.91693115234375,
|
|
"loss": 0.8837,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.467881679534912,
|
|
"rewards/margins": 0.8902303576469421,
|
|
"rewards/rejected": -2.358112096786499,
|
|
"step": 316
|
|
},
|
|
{
|
|
"epoch": 0.4654919236417034,
|
|
"grad_norm": 14.905621528625488,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.02203233912587166,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 3.2456423598071783e-07,
|
|
"logits/chosen": -7.734880447387695,
|
|
"logits/rejected": -7.02924919128418,
|
|
"logps/chosen": -152.81094360351562,
|
|
"logps/ref_chosen": -87.87348937988281,
|
|
"logps/ref_rejected": -108.26519775390625,
|
|
"logps/rejected": -228.77639770507812,
|
|
"loss": 0.7381,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -1.4228007793426514,
|
|
"rewards/margins": 1.2112901210784912,
|
|
"rewards/rejected": -2.6340909004211426,
|
|
"step": 317
|
|
},
|
|
{
|
|
"epoch": 0.4669603524229075,
|
|
"grad_norm": 11.828646659851074,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.021888693794608116,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 3.233383385962115e-07,
|
|
"logits/chosen": -7.722089767456055,
|
|
"logits/rejected": -7.081110000610352,
|
|
"logps/chosen": -161.36456298828125,
|
|
"logps/ref_chosen": -98.20553588867188,
|
|
"logps/ref_rejected": -88.13629150390625,
|
|
"logps/rejected": -199.14776611328125,
|
|
"loss": 0.7892,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.3737818002700806,
|
|
"rewards/margins": 1.0358906984329224,
|
|
"rewards/rejected": -2.409672498703003,
|
|
"step": 318
|
|
},
|
|
{
|
|
"epoch": 0.4684287812041116,
|
|
"grad_norm": 13.124372482299805,
|
|
"kl/avg_steps": 0.78125,
|
|
"kl/beta": 0.021732492372393608,
|
|
"kl/n_epsilon_steps": 0.109375,
|
|
"kl/p_epsilon_steps": 0.890625,
|
|
"learning_rate": 3.2211050867827805e-07,
|
|
"logits/chosen": -7.879185199737549,
|
|
"logits/rejected": -7.01790714263916,
|
|
"logps/chosen": -137.73611450195312,
|
|
"logps/ref_chosen": -82.50337219238281,
|
|
"logps/ref_rejected": -118.65068054199219,
|
|
"logps/rejected": -224.25790405273438,
|
|
"loss": 0.7808,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -1.1925673484802246,
|
|
"rewards/margins": 1.082844853401184,
|
|
"rewards/rejected": -2.2754123210906982,
|
|
"step": 319
|
|
},
|
|
{
|
|
"epoch": 0.4698972099853157,
|
|
"grad_norm": 12.478285789489746,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.021564023569226265,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 3.208807785813777e-07,
|
|
"logits/chosen": -7.674609184265137,
|
|
"logits/rejected": -7.378990173339844,
|
|
"logps/chosen": -144.46971130371094,
|
|
"logps/ref_chosen": -87.1372299194336,
|
|
"logps/ref_rejected": -103.22412109375,
|
|
"logps/rejected": -203.762451171875,
|
|
"loss": 0.8435,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.2286326885223389,
|
|
"rewards/margins": 0.9210659861564636,
|
|
"rewards/rejected": -2.149698495864868,
|
|
"step": 320
|
|
},
|
|
{
|
|
"epoch": 0.4713656387665198,
|
|
"grad_norm": 13.07551097869873,
|
|
"kl/avg_steps": 0.78125,
|
|
"kl/beta": 0.021416783332824707,
|
|
"kl/n_epsilon_steps": 0.109375,
|
|
"kl/p_epsilon_steps": 0.890625,
|
|
"learning_rate": 3.1964918071004217e-07,
|
|
"logits/chosen": -7.491329193115234,
|
|
"logits/rejected": -7.139033317565918,
|
|
"logps/chosen": -157.41146850585938,
|
|
"logps/ref_chosen": -93.10466766357422,
|
|
"logps/ref_rejected": -97.4681396484375,
|
|
"logps/rejected": -210.389892578125,
|
|
"loss": 0.8056,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.3688414096832275,
|
|
"rewards/margins": 1.0292859077453613,
|
|
"rewards/rejected": -2.398127555847168,
|
|
"step": 321
|
|
},
|
|
{
|
|
"epoch": 0.47283406754772395,
|
|
"grad_norm": 15.268094062805176,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.021250760182738304,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 3.184157475180207e-07,
|
|
"logits/chosen": -7.795929908752441,
|
|
"logits/rejected": -7.272746562957764,
|
|
"logps/chosen": -152.91873168945312,
|
|
"logps/ref_chosen": -92.19291687011719,
|
|
"logps/ref_rejected": -100.89321899414062,
|
|
"logps/rejected": -216.01556396484375,
|
|
"loss": 0.7797,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.2838683128356934,
|
|
"rewards/margins": 1.1429917812347412,
|
|
"rewards/rejected": -2.4268603324890137,
|
|
"step": 322
|
|
},
|
|
{
|
|
"epoch": 0.47430249632892807,
|
|
"grad_norm": 12.563841819763184,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.021112211048603058,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 3.171805115074251e-07,
|
|
"logits/chosen": -7.563092231750488,
|
|
"logits/rejected": -7.294485092163086,
|
|
"logps/chosen": -148.90322875976562,
|
|
"logps/ref_chosen": -80.27210235595703,
|
|
"logps/ref_rejected": -80.30335235595703,
|
|
"logps/rejected": -197.24813842773438,
|
|
"loss": 0.7955,
|
|
"rewards/accuracies": 0.890625,
|
|
"rewards/chosen": -1.440096139907837,
|
|
"rewards/margins": 1.0090997219085693,
|
|
"rewards/rejected": -2.449195623397827,
|
|
"step": 323
|
|
},
|
|
{
|
|
"epoch": 0.47577092511013214,
|
|
"grad_norm": 13.184640884399414,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.020961550995707512,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 3.1594350522787295e-07,
|
|
"logits/chosen": -7.81708288192749,
|
|
"logits/rejected": -6.640934944152832,
|
|
"logps/chosen": -165.7332763671875,
|
|
"logps/ref_chosen": -95.72935485839844,
|
|
"logps/ref_rejected": -94.16924285888672,
|
|
"logps/rejected": -216.46131896972656,
|
|
"loss": 0.732,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -1.4592244625091553,
|
|
"rewards/margins": 1.0842043161392212,
|
|
"rewards/rejected": -2.543428897857666,
|
|
"step": 324
|
|
},
|
|
{
|
|
"epoch": 0.47723935389133626,
|
|
"grad_norm": 11.3136625289917,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.020818423479795456,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 3.147047612756302e-07,
|
|
"logits/chosen": -7.779555320739746,
|
|
"logits/rejected": -7.204484939575195,
|
|
"logps/chosen": -164.93191528320312,
|
|
"logps/ref_chosen": -98.70687103271484,
|
|
"logps/ref_rejected": -76.40809631347656,
|
|
"logps/rejected": -192.03103637695312,
|
|
"loss": 0.7957,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.37131929397583,
|
|
"rewards/margins": 1.0167847871780396,
|
|
"rewards/rejected": -2.38810396194458,
|
|
"step": 325
|
|
},
|
|
{
|
|
"epoch": 0.4787077826725404,
|
|
"grad_norm": 13.992250442504883,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.020682694390416145,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 3.134643122927519e-07,
|
|
"logits/chosen": -7.7820024490356445,
|
|
"logits/rejected": -7.323631286621094,
|
|
"logps/chosen": -180.78939819335938,
|
|
"logps/ref_chosen": -104.66130065917969,
|
|
"logps/ref_rejected": -86.82537841796875,
|
|
"logps/rejected": -208.27098083496094,
|
|
"loss": 0.8558,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.567757487297058,
|
|
"rewards/margins": 0.9265294671058655,
|
|
"rewards/rejected": -2.4942867755889893,
|
|
"step": 326
|
|
},
|
|
{
|
|
"epoch": 0.4801762114537445,
|
|
"grad_norm": 13.753683090209961,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.020554229617118835,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 3.1222219096622264e-07,
|
|
"logits/chosen": -7.905759811401367,
|
|
"logits/rejected": -7.24397611618042,
|
|
"logps/chosen": -180.28341674804688,
|
|
"logps/ref_chosen": -107.44024658203125,
|
|
"logps/ref_rejected": -117.05323028564453,
|
|
"logps/rejected": -237.6690216064453,
|
|
"loss": 0.8317,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.491244912147522,
|
|
"rewards/margins": 0.9712578058242798,
|
|
"rewards/rejected": -2.4625027179718018,
|
|
"step": 327
|
|
},
|
|
{
|
|
"epoch": 0.48164464023494863,
|
|
"grad_norm": 13.748778343200684,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.020439259707927704,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 3.1097843002709427e-07,
|
|
"logits/chosen": -7.3603315353393555,
|
|
"logits/rejected": -6.873122215270996,
|
|
"logps/chosen": -162.29519653320312,
|
|
"logps/ref_chosen": -86.6106948852539,
|
|
"logps/ref_rejected": -97.40188598632812,
|
|
"logps/rejected": -229.42196655273438,
|
|
"loss": 0.7407,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -1.5384535789489746,
|
|
"rewards/margins": 1.1390830278396606,
|
|
"rewards/rejected": -2.6775364875793457,
|
|
"step": 328
|
|
},
|
|
{
|
|
"epoch": 0.4831130690161527,
|
|
"grad_norm": 14.800250053405762,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.020299699157476425,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 3.0973306224962437e-07,
|
|
"logits/chosen": -7.575889587402344,
|
|
"logits/rejected": -7.213668346405029,
|
|
"logps/chosen": -178.1090850830078,
|
|
"logps/ref_chosen": -100.61308288574219,
|
|
"logps/ref_rejected": -104.95370483398438,
|
|
"logps/rejected": -230.13726806640625,
|
|
"loss": 0.8164,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.564449667930603,
|
|
"rewards/margins": 0.9571825861930847,
|
|
"rewards/rejected": -2.521632194519043,
|
|
"step": 329
|
|
},
|
|
{
|
|
"epoch": 0.4845814977973568,
|
|
"grad_norm": 12.028066635131836,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.020161090418696404,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 3.084861204504122e-07,
|
|
"logits/chosen": -7.624053955078125,
|
|
"logits/rejected": -7.438431739807129,
|
|
"logps/chosen": -147.8223876953125,
|
|
"logps/ref_chosen": -76.95843505859375,
|
|
"logps/ref_rejected": -92.81494140625,
|
|
"logps/rejected": -214.4041290283203,
|
|
"loss": 0.7539,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -1.4205009937286377,
|
|
"rewards/margins": 1.0114681720733643,
|
|
"rewards/rejected": -2.431969165802002,
|
|
"step": 330
|
|
},
|
|
{
|
|
"epoch": 0.48604992657856094,
|
|
"grad_norm": 16.355430603027344,
|
|
"kl/avg_steps": 0.75,
|
|
"kl/beta": 0.02002342976629734,
|
|
"kl/n_epsilon_steps": 0.125,
|
|
"kl/p_epsilon_steps": 0.875,
|
|
"learning_rate": 3.072376374875335e-07,
|
|
"logits/chosen": -7.7047834396362305,
|
|
"logits/rejected": -7.323886394500732,
|
|
"logps/chosen": -158.1719512939453,
|
|
"logps/ref_chosen": -83.0594253540039,
|
|
"logps/ref_rejected": -94.13113403320312,
|
|
"logps/rejected": -222.07827758789062,
|
|
"loss": 0.7566,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -1.4940705299377441,
|
|
"rewards/margins": 1.0462760925292969,
|
|
"rewards/rejected": -2.540346622467041,
|
|
"step": 331
|
|
},
|
|
{
|
|
"epoch": 0.48751835535976507,
|
|
"grad_norm": 13.534795761108398,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.019874371588230133,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 3.059876462596758e-07,
|
|
"logits/chosen": -7.760369300842285,
|
|
"logits/rejected": -7.029866695404053,
|
|
"logps/chosen": -159.60678100585938,
|
|
"logps/ref_chosen": -81.93089294433594,
|
|
"logps/ref_rejected": -82.84590148925781,
|
|
"logps/rejected": -202.23422241210938,
|
|
"loss": 0.9013,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.5361754894256592,
|
|
"rewards/margins": 0.8190126419067383,
|
|
"rewards/rejected": -2.3551878929138184,
|
|
"step": 332
|
|
},
|
|
{
|
|
"epoch": 0.4889867841409692,
|
|
"grad_norm": 13.899153709411621,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.01975092850625515,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 3.0473617970527015e-07,
|
|
"logits/chosen": -7.564597129821777,
|
|
"logits/rejected": -7.067813396453857,
|
|
"logps/chosen": -177.69625854492188,
|
|
"logps/ref_chosen": -88.53257751464844,
|
|
"logps/ref_rejected": -102.46601867675781,
|
|
"logps/rejected": -244.46804809570312,
|
|
"loss": 0.8778,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -1.7536228895187378,
|
|
"rewards/margins": 1.0312126874923706,
|
|
"rewards/rejected": -2.7848353385925293,
|
|
"step": 333
|
|
},
|
|
{
|
|
"epoch": 0.49045521292217326,
|
|
"grad_norm": 16.076087951660156,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.019640451297163963,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 3.034832708016243e-07,
|
|
"logits/chosen": -7.2526021003723145,
|
|
"logits/rejected": -7.094786643981934,
|
|
"logps/chosen": -180.3138427734375,
|
|
"logps/ref_chosen": -97.75778198242188,
|
|
"logps/ref_rejected": -100.00584411621094,
|
|
"logps/rejected": -230.16831970214844,
|
|
"loss": 0.9017,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.6136884689331055,
|
|
"rewards/margins": 0.9234945774078369,
|
|
"rewards/rejected": -2.5371832847595215,
|
|
"step": 334
|
|
},
|
|
{
|
|
"epoch": 0.4919236417033774,
|
|
"grad_norm": 15.462376594543457,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.019506344571709633,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 3.022289525640531e-07,
|
|
"logits/chosen": -7.538198471069336,
|
|
"logits/rejected": -6.875842571258545,
|
|
"logps/chosen": -179.6153106689453,
|
|
"logps/ref_chosen": -96.37603759765625,
|
|
"logps/ref_rejected": -94.92401123046875,
|
|
"logps/rejected": -222.1100616455078,
|
|
"loss": 0.9078,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -1.618840217590332,
|
|
"rewards/margins": 0.8465075492858887,
|
|
"rewards/rejected": -2.4653477668762207,
|
|
"step": 335
|
|
},
|
|
{
|
|
"epoch": 0.4933920704845815,
|
|
"grad_norm": 14.706995010375977,
|
|
"kl/avg_steps": 0.75,
|
|
"kl/beta": 0.01940326578915119,
|
|
"kl/n_epsilon_steps": 0.125,
|
|
"kl/p_epsilon_steps": 0.875,
|
|
"learning_rate": 3.009732580450086e-07,
|
|
"logits/chosen": -7.663290500640869,
|
|
"logits/rejected": -7.155807971954346,
|
|
"logps/chosen": -170.34237670898438,
|
|
"logps/ref_chosen": -88.52952575683594,
|
|
"logps/ref_rejected": -109.55577850341797,
|
|
"logps/rejected": -245.36219787597656,
|
|
"loss": 0.7966,
|
|
"rewards/accuracies": 0.890625,
|
|
"rewards/chosen": -1.5775220394134521,
|
|
"rewards/margins": 1.0359567403793335,
|
|
"rewards/rejected": -2.6134791374206543,
|
|
"step": 336
|
|
},
|
|
{
|
|
"epoch": 0.4948604992657856,
|
|
"grad_norm": 15.293198585510254,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.019258825108408928,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 2.9971622033320914e-07,
|
|
"logits/chosen": -7.815160274505615,
|
|
"logits/rejected": -7.3971452713012695,
|
|
"logps/chosen": -179.1387939453125,
|
|
"logps/ref_chosen": -97.31238555908203,
|
|
"logps/ref_rejected": -110.13134765625,
|
|
"logps/rejected": -241.39308166503906,
|
|
"loss": 0.8903,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.5696876049041748,
|
|
"rewards/margins": 0.9406921863555908,
|
|
"rewards/rejected": -2.5103797912597656,
|
|
"step": 337
|
|
},
|
|
{
|
|
"epoch": 0.49632892804698975,
|
|
"grad_norm": 15.433863639831543,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.019145149737596512,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 2.984578725527675e-07,
|
|
"logits/chosen": -7.460289478302002,
|
|
"logits/rejected": -7.160378456115723,
|
|
"logps/chosen": -167.595947265625,
|
|
"logps/ref_chosen": -83.63011932373047,
|
|
"logps/ref_rejected": -93.69235229492188,
|
|
"logps/rejected": -226.01394653320312,
|
|
"loss": 0.8439,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.5996140241622925,
|
|
"rewards/margins": 0.9147864580154419,
|
|
"rewards/rejected": -2.5144004821777344,
|
|
"step": 338
|
|
},
|
|
{
|
|
"epoch": 0.4977973568281938,
|
|
"grad_norm": 14.1244535446167,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.01902623660862446,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 2.9719824786231796e-07,
|
|
"logits/chosen": -7.783293724060059,
|
|
"logits/rejected": -7.273715496063232,
|
|
"logps/chosen": -184.80462646484375,
|
|
"logps/ref_chosen": -99.85182189941406,
|
|
"logps/ref_rejected": -109.0743637084961,
|
|
"logps/rejected": -246.49473571777344,
|
|
"loss": 0.8648,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -1.609981656074524,
|
|
"rewards/margins": 0.985941469669342,
|
|
"rewards/rejected": -2.5959229469299316,
|
|
"step": 339
|
|
},
|
|
{
|
|
"epoch": 0.49926578560939794,
|
|
"grad_norm": 17.34671974182129,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.018913934007287025,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 2.959373794541426e-07,
|
|
"logits/chosen": -7.614503383636475,
|
|
"logits/rejected": -7.28403377532959,
|
|
"logps/chosen": -177.3800811767578,
|
|
"logps/ref_chosen": -84.97396087646484,
|
|
"logps/ref_rejected": -78.80033874511719,
|
|
"logps/rejected": -218.79473876953125,
|
|
"loss": 0.8763,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.739735722541809,
|
|
"rewards/margins": 0.8897112607955933,
|
|
"rewards/rejected": -2.6294469833374023,
|
|
"step": 340
|
|
},
|
|
{
|
|
"epoch": 0.5007342143906021,
|
|
"grad_norm": 16.43499755859375,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.01879645697772503,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 2.946753005532965e-07,
|
|
"logits/chosen": -7.702646732330322,
|
|
"logits/rejected": -6.986645698547363,
|
|
"logps/chosen": -184.0854949951172,
|
|
"logps/ref_chosen": -85.26399230957031,
|
|
"logps/ref_rejected": -106.08595275878906,
|
|
"logps/rejected": -250.41880798339844,
|
|
"loss": 0.9302,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -1.850089430809021,
|
|
"rewards/margins": 0.8445804119110107,
|
|
"rewards/rejected": -2.694669723510742,
|
|
"step": 341
|
|
},
|
|
{
|
|
"epoch": 0.5022026431718062,
|
|
"grad_norm": 14.287416458129883,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.01869712769985199,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 2.934120444167326e-07,
|
|
"logits/chosen": -7.695643424987793,
|
|
"logits/rejected": -6.729315280914307,
|
|
"logps/chosen": -178.91830444335938,
|
|
"logps/ref_chosen": -101.03860473632812,
|
|
"logps/ref_rejected": -98.89851379394531,
|
|
"logps/rejected": -240.55776977539062,
|
|
"loss": 0.7283,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -1.4495775699615479,
|
|
"rewards/margins": 1.179455280303955,
|
|
"rewards/rejected": -2.629033088684082,
|
|
"step": 342
|
|
},
|
|
{
|
|
"epoch": 0.5036710719530103,
|
|
"grad_norm": 15.677153587341309,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.018580997362732887,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 2.9214764433242476e-07,
|
|
"logits/chosen": -7.71726131439209,
|
|
"logits/rejected": -7.144956588745117,
|
|
"logps/chosen": -166.2335205078125,
|
|
"logps/ref_chosen": -83.74736785888672,
|
|
"logps/ref_rejected": -116.57908630371094,
|
|
"logps/rejected": -253.29689025878906,
|
|
"loss": 0.8041,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -1.5235764980316162,
|
|
"rewards/margins": 0.996279776096344,
|
|
"rewards/rejected": -2.5198564529418945,
|
|
"step": 343
|
|
},
|
|
{
|
|
"epoch": 0.5051395007342144,
|
|
"grad_norm": 14.323365211486816,
|
|
"kl/avg_steps": 0.640625,
|
|
"kl/beta": 0.018448399379849434,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 2.9088213361849126e-07,
|
|
"logits/chosen": -7.541003227233887,
|
|
"logits/rejected": -7.094578266143799,
|
|
"logps/chosen": -167.88778686523438,
|
|
"logps/ref_chosen": -78.25396728515625,
|
|
"logps/ref_rejected": -95.4287109375,
|
|
"logps/rejected": -237.73025512695312,
|
|
"loss": 0.8489,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.6444833278656006,
|
|
"rewards/margins": 0.9605084657669067,
|
|
"rewards/rejected": -2.604991912841797,
|
|
"step": 344
|
|
},
|
|
{
|
|
"epoch": 0.5066079295154186,
|
|
"grad_norm": 14.38456916809082,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.01833096705377102,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 2.896155456223163e-07,
|
|
"logits/chosen": -7.9548492431640625,
|
|
"logits/rejected": -7.6490678787231445,
|
|
"logps/chosen": -184.28465270996094,
|
|
"logps/ref_chosen": -98.0794906616211,
|
|
"logps/ref_rejected": -104.03477478027344,
|
|
"logps/rejected": -248.8415985107422,
|
|
"loss": 0.7793,
|
|
"rewards/accuracies": 0.890625,
|
|
"rewards/chosen": -1.5716257095336914,
|
|
"rewards/margins": 1.0628182888031006,
|
|
"rewards/rejected": -2.634444236755371,
|
|
"step": 345
|
|
},
|
|
{
|
|
"epoch": 0.5080763582966226,
|
|
"grad_norm": 13.954184532165527,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.01820015348494053,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 2.883479137196714e-07,
|
|
"logits/chosen": -7.322576522827148,
|
|
"logits/rejected": -6.97154426574707,
|
|
"logps/chosen": -179.28921508789062,
|
|
"logps/ref_chosen": -80.68348693847656,
|
|
"logps/ref_rejected": -85.89260864257812,
|
|
"logps/rejected": -232.02542114257812,
|
|
"loss": 0.9413,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -1.788151502609253,
|
|
"rewards/margins": 0.8540264368057251,
|
|
"rewards/rejected": -2.6421780586242676,
|
|
"step": 346
|
|
},
|
|
{
|
|
"epoch": 0.5095447870778267,
|
|
"grad_norm": 14.986981391906738,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.01810397580265999,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 2.8707927131383614e-07,
|
|
"logits/chosen": -7.665702819824219,
|
|
"logits/rejected": -7.370832920074463,
|
|
"logps/chosen": -187.2706298828125,
|
|
"logps/ref_chosen": -93.67938232421875,
|
|
"logps/ref_rejected": -97.54251861572266,
|
|
"logps/rejected": -237.28759765625,
|
|
"loss": 0.9517,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -1.686805009841919,
|
|
"rewards/margins": 0.8248369693756104,
|
|
"rewards/rejected": -2.5116419792175293,
|
|
"step": 347
|
|
},
|
|
{
|
|
"epoch": 0.5110132158590308,
|
|
"grad_norm": 13.239946365356445,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.01800270937383175,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 2.858096518347179e-07,
|
|
"logits/chosen": -7.554695129394531,
|
|
"logits/rejected": -7.127823829650879,
|
|
"logps/chosen": -173.94375610351562,
|
|
"logps/ref_chosen": -90.25224304199219,
|
|
"logps/ref_rejected": -95.77027893066406,
|
|
"logps/rejected": -228.76617431640625,
|
|
"loss": 0.888,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.4998319149017334,
|
|
"rewards/margins": 0.8769406676292419,
|
|
"rewards/rejected": -2.376772403717041,
|
|
"step": 348
|
|
},
|
|
{
|
|
"epoch": 0.5124816446402349,
|
|
"grad_norm": 13.889335632324219,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.0178797859698534,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 2.845390887379706e-07,
|
|
"logits/chosen": -7.729681968688965,
|
|
"logits/rejected": -7.328579902648926,
|
|
"logps/chosen": -180.9906463623047,
|
|
"logps/ref_chosen": -92.05809020996094,
|
|
"logps/ref_rejected": -103.96527099609375,
|
|
"logps/rejected": -241.1156463623047,
|
|
"loss": 0.906,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -1.5828135013580322,
|
|
"rewards/margins": 0.8517956137657166,
|
|
"rewards/rejected": -2.4346094131469727,
|
|
"step": 349
|
|
},
|
|
{
|
|
"epoch": 0.5139500734214391,
|
|
"grad_norm": 15.606201171875,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.017774252220988274,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 2.8326761550411346e-07,
|
|
"logits/chosen": -7.431205749511719,
|
|
"logits/rejected": -7.113109588623047,
|
|
"logps/chosen": -183.2895050048828,
|
|
"logps/ref_chosen": -87.05419921875,
|
|
"logps/ref_rejected": -95.72193145751953,
|
|
"logps/rejected": -232.35379028320312,
|
|
"loss": 0.9483,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -1.7016650438308716,
|
|
"rewards/margins": 0.7095808982849121,
|
|
"rewards/rejected": -2.411245822906494,
|
|
"step": 350
|
|
},
|
|
{
|
|
"epoch": 0.5154185022026432,
|
|
"grad_norm": 13.189620018005371,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.01766934059560299,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 2.819952656376487e-07,
|
|
"logits/chosen": -8.007095336914062,
|
|
"logits/rejected": -7.268853187561035,
|
|
"logps/chosen": -181.0399169921875,
|
|
"logps/ref_chosen": -89.16903686523438,
|
|
"logps/ref_rejected": -108.29801940917969,
|
|
"logps/rejected": -259.3538818359375,
|
|
"loss": 0.8124,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.6149550676345825,
|
|
"rewards/margins": 1.0333154201507568,
|
|
"rewards/rejected": -2.64827036857605,
|
|
"step": 351
|
|
},
|
|
{
|
|
"epoch": 0.5168869309838473,
|
|
"grad_norm": 14.201725959777832,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.017554141581058502,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 2.8072207266617854e-07,
|
|
"logits/chosen": -7.791421890258789,
|
|
"logits/rejected": -7.136171340942383,
|
|
"logps/chosen": -183.04286193847656,
|
|
"logps/ref_chosen": -96.81147766113281,
|
|
"logps/ref_rejected": -84.40482330322266,
|
|
"logps/rejected": -225.66738891601562,
|
|
"loss": 0.8255,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.509531021118164,
|
|
"rewards/margins": 0.955288290977478,
|
|
"rewards/rejected": -2.4648194313049316,
|
|
"step": 352
|
|
},
|
|
{
|
|
"epoch": 0.5183553597650514,
|
|
"grad_norm": 15.365931510925293,
|
|
"kl/avg_steps": 0.40625,
|
|
"kl/beta": 0.017455952242016792,
|
|
"kl/n_epsilon_steps": 0.296875,
|
|
"kl/p_epsilon_steps": 0.703125,
|
|
"learning_rate": 2.794480701395219e-07,
|
|
"logits/chosen": -7.713462829589844,
|
|
"logits/rejected": -7.477531433105469,
|
|
"logps/chosen": -183.22384643554688,
|
|
"logps/ref_chosen": -85.36248016357422,
|
|
"logps/ref_rejected": -87.36732482910156,
|
|
"logps/rejected": -219.9366455078125,
|
|
"loss": 1.0508,
|
|
"rewards/accuracies": 0.734375,
|
|
"rewards/chosen": -1.7041988372802734,
|
|
"rewards/margins": 0.5970733761787415,
|
|
"rewards/rejected": -2.301272392272949,
|
|
"step": 353
|
|
},
|
|
{
|
|
"epoch": 0.5198237885462555,
|
|
"grad_norm": 12.326685905456543,
|
|
"kl/avg_steps": 0.8125,
|
|
"kl/beta": 0.017385324463248253,
|
|
"kl/n_epsilon_steps": 0.09375,
|
|
"kl/p_epsilon_steps": 0.90625,
|
|
"learning_rate": 2.781732916288303e-07,
|
|
"logits/chosen": -7.71798038482666,
|
|
"logits/rejected": -7.087685585021973,
|
|
"logps/chosen": -183.48825073242188,
|
|
"logps/ref_chosen": -95.8011703491211,
|
|
"logps/ref_rejected": -94.86614227294922,
|
|
"logps/rejected": -236.8675537109375,
|
|
"loss": 0.7853,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -1.5137109756469727,
|
|
"rewards/margins": 0.9342291951179504,
|
|
"rewards/rejected": -2.4479401111602783,
|
|
"step": 354
|
|
},
|
|
{
|
|
"epoch": 0.5212922173274597,
|
|
"grad_norm": 12.55382251739502,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.017245206981897354,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 2.7689777072570284e-07,
|
|
"logits/chosen": -7.723712921142578,
|
|
"logits/rejected": -7.181628704071045,
|
|
"logps/chosen": -172.11546325683594,
|
|
"logps/ref_chosen": -86.185546875,
|
|
"logps/ref_rejected": -88.4615707397461,
|
|
"logps/rejected": -227.43336486816406,
|
|
"loss": 0.8456,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.473813772201538,
|
|
"rewards/margins": 0.904168426990509,
|
|
"rewards/rejected": -2.3779821395874023,
|
|
"step": 355
|
|
},
|
|
{
|
|
"epoch": 0.5227606461086637,
|
|
"grad_norm": 15.903926849365234,
|
|
"kl/avg_steps": 0.40625,
|
|
"kl/beta": 0.01712745428085327,
|
|
"kl/n_epsilon_steps": 0.296875,
|
|
"kl/p_epsilon_steps": 0.703125,
|
|
"learning_rate": 2.7562154104130176e-07,
|
|
"logits/chosen": -7.718483924865723,
|
|
"logits/rejected": -7.080849647521973,
|
|
"logps/chosen": -191.98709106445312,
|
|
"logps/ref_chosen": -90.818115234375,
|
|
"logps/ref_rejected": -82.1273193359375,
|
|
"logps/rejected": -222.52052307128906,
|
|
"loss": 1.0336,
|
|
"rewards/accuracies": 0.71875,
|
|
"rewards/chosen": -1.7288204431533813,
|
|
"rewards/margins": 0.6621589660644531,
|
|
"rewards/rejected": -2.390979290008545,
|
|
"step": 356
|
|
},
|
|
{
|
|
"epoch": 0.5242290748898678,
|
|
"grad_norm": 12.080792427062988,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.017058156430721283,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 2.7434463620546594e-07,
|
|
"logits/chosen": -7.890803337097168,
|
|
"logits/rejected": -7.257773399353027,
|
|
"logps/chosen": -171.85879516601562,
|
|
"logps/ref_chosen": -87.50475311279297,
|
|
"logps/ref_rejected": -92.24937438964844,
|
|
"logps/rejected": -239.12313842773438,
|
|
"loss": 0.7488,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -1.4316484928131104,
|
|
"rewards/margins": 1.055083990097046,
|
|
"rewards/rejected": -2.4867324829101562,
|
|
"step": 357
|
|
},
|
|
{
|
|
"epoch": 0.5256975036710719,
|
|
"grad_norm": 13.29261589050293,
|
|
"kl/avg_steps": 0.546875,
|
|
"kl/beta": 0.016946941614151,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 2.730670898658255e-07,
|
|
"logits/chosen": -7.43376350402832,
|
|
"logits/rejected": -6.966874122619629,
|
|
"logps/chosen": -190.7314453125,
|
|
"logps/ref_chosen": -95.05546569824219,
|
|
"logps/ref_rejected": -96.47611999511719,
|
|
"logps/rejected": -237.2618408203125,
|
|
"loss": 0.9431,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -1.6134233474731445,
|
|
"rewards/margins": 0.7554316520690918,
|
|
"rewards/rejected": -2.3688549995422363,
|
|
"step": 358
|
|
},
|
|
{
|
|
"epoch": 0.527165932452276,
|
|
"grad_norm": 15.180391311645508,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.016854766756296158,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 2.717889356869146e-07,
|
|
"logits/chosen": -7.312591552734375,
|
|
"logits/rejected": -6.956157684326172,
|
|
"logps/chosen": -192.33929443359375,
|
|
"logps/ref_chosen": -87.37332916259766,
|
|
"logps/ref_rejected": -88.77188110351562,
|
|
"logps/rejected": -236.61764526367188,
|
|
"loss": 1.0136,
|
|
"rewards/accuracies": 0.71875,
|
|
"rewards/chosen": -1.762831687927246,
|
|
"rewards/margins": 0.7133128046989441,
|
|
"rewards/rejected": -2.476144313812256,
|
|
"step": 359
|
|
},
|
|
{
|
|
"epoch": 0.5286343612334802,
|
|
"grad_norm": 14.518013954162598,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.01677091233432293,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 2.7051020734928443e-07,
|
|
"logits/chosen": -7.531591415405273,
|
|
"logits/rejected": -6.866611480712891,
|
|
"logps/chosen": -175.69329833984375,
|
|
"logps/ref_chosen": -75.22344970703125,
|
|
"logps/ref_rejected": -76.093994140625,
|
|
"logps/rejected": -229.7479248046875,
|
|
"loss": 0.8437,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -1.6761372089385986,
|
|
"rewards/margins": 0.8820022344589233,
|
|
"rewards/rejected": -2.5581398010253906,
|
|
"step": 360
|
|
},
|
|
{
|
|
"epoch": 0.5301027900146843,
|
|
"grad_norm": 14.980257987976074,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.016661571338772774,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 2.6923093854861593e-07,
|
|
"logits/chosen": -7.7249908447265625,
|
|
"logits/rejected": -7.067028999328613,
|
|
"logps/chosen": -183.44650268554688,
|
|
"logps/ref_chosen": -81.95333099365234,
|
|
"logps/ref_rejected": -96.67919921875,
|
|
"logps/rejected": -245.79881286621094,
|
|
"loss": 0.9577,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -1.684104561805725,
|
|
"rewards/margins": 0.7838806509971619,
|
|
"rewards/rejected": -2.467985153198242,
|
|
"step": 361
|
|
},
|
|
{
|
|
"epoch": 0.5315712187958884,
|
|
"grad_norm": 12.824585914611816,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.016568373888731003,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 2.679511629948319e-07,
|
|
"logits/chosen": -7.730887413024902,
|
|
"logits/rejected": -7.379080772399902,
|
|
"logps/chosen": -192.10179138183594,
|
|
"logps/ref_chosen": -89.33971405029297,
|
|
"logps/ref_rejected": -110.5019760131836,
|
|
"logps/rejected": -265.67584228515625,
|
|
"loss": 0.8735,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -1.6923720836639404,
|
|
"rewards/margins": 0.8581586480140686,
|
|
"rewards/rejected": -2.5505309104919434,
|
|
"step": 362
|
|
},
|
|
{
|
|
"epoch": 0.5330396475770925,
|
|
"grad_norm": 12.172687530517578,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.01645524427294731,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 2.6667091441120816e-07,
|
|
"logits/chosen": -7.627842903137207,
|
|
"logits/rejected": -7.226386070251465,
|
|
"logps/chosen": -177.34544372558594,
|
|
"logps/ref_chosen": -80.7750015258789,
|
|
"logps/ref_rejected": -79.43267822265625,
|
|
"logps/rejected": -243.47921752929688,
|
|
"loss": 0.7656,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -1.5798821449279785,
|
|
"rewards/margins": 1.0982314348220825,
|
|
"rewards/rejected": -2.6781134605407715,
|
|
"step": 363
|
|
},
|
|
{
|
|
"epoch": 0.5345080763582967,
|
|
"grad_norm": 14.342798233032227,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.01634288765490055,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 2.6539022653348575e-07,
|
|
"logits/chosen": -7.7837934494018555,
|
|
"logits/rejected": -6.654484748840332,
|
|
"logps/chosen": -174.86444091796875,
|
|
"logps/ref_chosen": -76.52960205078125,
|
|
"logps/ref_rejected": -97.56173706054688,
|
|
"logps/rejected": -247.8596649169922,
|
|
"loss": 0.8462,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -1.5991710424423218,
|
|
"rewards/margins": 0.8397619724273682,
|
|
"rewards/rejected": -2.4389328956604004,
|
|
"step": 364
|
|
},
|
|
{
|
|
"epoch": 0.5359765051395007,
|
|
"grad_norm": 13.283844947814941,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.01624137908220291,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 2.641091331089811e-07,
|
|
"logits/chosen": -7.504325866699219,
|
|
"logits/rejected": -7.238739967346191,
|
|
"logps/chosen": -178.31759643554688,
|
|
"logps/ref_chosen": -80.54489135742188,
|
|
"logps/ref_rejected": -99.32316589355469,
|
|
"logps/rejected": -254.62168884277344,
|
|
"loss": 0.8603,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.5808205604553223,
|
|
"rewards/margins": 0.9229766130447388,
|
|
"rewards/rejected": -2.5037970542907715,
|
|
"step": 365
|
|
},
|
|
{
|
|
"epoch": 0.5374449339207048,
|
|
"grad_norm": 12.054483413696289,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.01613548956811428,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 2.6282766789569736e-07,
|
|
"logits/chosen": -7.318140983581543,
|
|
"logits/rejected": -6.966005325317383,
|
|
"logps/chosen": -168.7039337158203,
|
|
"logps/ref_chosen": -72.23806762695312,
|
|
"logps/ref_rejected": -91.67182922363281,
|
|
"logps/rejected": -242.06005859375,
|
|
"loss": 0.86,
|
|
"rewards/accuracies": 0.921875,
|
|
"rewards/chosen": -1.5466276407241821,
|
|
"rewards/margins": 0.8613663911819458,
|
|
"rewards/rejected": -2.407994031906128,
|
|
"step": 366
|
|
},
|
|
{
|
|
"epoch": 0.5389133627019089,
|
|
"grad_norm": 12.747227668762207,
|
|
"kl/avg_steps": 0.828125,
|
|
"kl/beta": 0.016020342707633972,
|
|
"kl/n_epsilon_steps": 0.078125,
|
|
"kl/p_epsilon_steps": 0.90625,
|
|
"learning_rate": 2.615458646614349e-07,
|
|
"logits/chosen": -7.577718734741211,
|
|
"logits/rejected": -7.4753642082214355,
|
|
"logps/chosen": -179.722412109375,
|
|
"logps/ref_chosen": -85.42201232910156,
|
|
"logps/ref_rejected": -83.51779174804688,
|
|
"logps/rejected": -234.3914031982422,
|
|
"loss": 0.8254,
|
|
"rewards/accuracies": 0.90625,
|
|
"rewards/chosen": -1.4997714757919312,
|
|
"rewards/margins": 0.896426796913147,
|
|
"rewards/rejected": -2.396198272705078,
|
|
"step": 367
|
|
},
|
|
{
|
|
"epoch": 0.540381791483113,
|
|
"grad_norm": 14.037300109863281,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.015888763591647148,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 2.6026375718290083e-07,
|
|
"logits/chosen": -7.514585494995117,
|
|
"logits/rejected": -7.210457801818848,
|
|
"logps/chosen": -165.94821166992188,
|
|
"logps/ref_chosen": -73.84112548828125,
|
|
"logps/ref_rejected": -104.69528198242188,
|
|
"logps/rejected": -256.14404296875,
|
|
"loss": 0.8403,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.4564285278320312,
|
|
"rewards/margins": 0.9321895241737366,
|
|
"rewards/rejected": -2.388618230819702,
|
|
"step": 368
|
|
},
|
|
{
|
|
"epoch": 0.5418502202643172,
|
|
"grad_norm": 13.097122192382812,
|
|
"kl/avg_steps": 0.75,
|
|
"kl/beta": 0.015790076926350594,
|
|
"kl/n_epsilon_steps": 0.125,
|
|
"kl/p_epsilon_steps": 0.875,
|
|
"learning_rate": 2.589813792448196e-07,
|
|
"logits/chosen": -7.481395721435547,
|
|
"logits/rejected": -6.605921745300293,
|
|
"logps/chosen": -196.92771911621094,
|
|
"logps/ref_chosen": -98.46196746826172,
|
|
"logps/ref_rejected": -99.15672302246094,
|
|
"logps/rejected": -256.7950439453125,
|
|
"loss": 0.7803,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -1.545214056968689,
|
|
"rewards/margins": 0.9250272512435913,
|
|
"rewards/rejected": -2.4702415466308594,
|
|
"step": 369
|
|
},
|
|
{
|
|
"epoch": 0.5433186490455213,
|
|
"grad_norm": 13.142091751098633,
|
|
"kl/avg_steps": 0.40625,
|
|
"kl/beta": 0.015672532841563225,
|
|
"kl/n_epsilon_steps": 0.296875,
|
|
"kl/p_epsilon_steps": 0.703125,
|
|
"learning_rate": 2.5769876463904263e-07,
|
|
"logits/chosen": -7.735294342041016,
|
|
"logits/rejected": -7.384430885314941,
|
|
"logps/chosen": -204.67718505859375,
|
|
"logps/ref_chosen": -96.77574157714844,
|
|
"logps/ref_rejected": -102.74411010742188,
|
|
"logps/rejected": -253.85340881347656,
|
|
"loss": 1.002,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -1.687301754951477,
|
|
"rewards/margins": 0.6685256361961365,
|
|
"rewards/rejected": -2.3558273315429688,
|
|
"step": 370
|
|
},
|
|
{
|
|
"epoch": 0.5447870778267254,
|
|
"grad_norm": 14.304271697998047,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.015609120018780231,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 2.5641594716365744e-07,
|
|
"logits/chosen": -7.651013374328613,
|
|
"logits/rejected": -7.241464614868164,
|
|
"logps/chosen": -206.90415954589844,
|
|
"logps/ref_chosen": -102.99020385742188,
|
|
"logps/ref_rejected": -105.35063171386719,
|
|
"logps/rejected": -252.56182861328125,
|
|
"loss": 1.0035,
|
|
"rewards/accuracies": 0.734375,
|
|
"rewards/chosen": -1.6160762310028076,
|
|
"rewards/margins": 0.6670252084732056,
|
|
"rewards/rejected": -2.2831013202667236,
|
|
"step": 371
|
|
},
|
|
{
|
|
"epoch": 0.5462555066079295,
|
|
"grad_norm": 11.53231430053711,
|
|
"kl/avg_steps": 0.8125,
|
|
"kl/beta": 0.0155362943187356,
|
|
"kl/n_epsilon_steps": 0.09375,
|
|
"kl/p_epsilon_steps": 0.90625,
|
|
"learning_rate": 2.551329606220976e-07,
|
|
"logits/chosen": -7.6374592781066895,
|
|
"logits/rejected": -7.359759330749512,
|
|
"logps/chosen": -181.00729370117188,
|
|
"logps/ref_chosen": -91.02623748779297,
|
|
"logps/ref_rejected": -85.6102294921875,
|
|
"logps/rejected": -244.81317138671875,
|
|
"loss": 0.7449,
|
|
"rewards/accuracies": 0.890625,
|
|
"rewards/chosen": -1.3880116939544678,
|
|
"rewards/margins": 1.0640125274658203,
|
|
"rewards/rejected": -2.452024221420288,
|
|
"step": 372
|
|
},
|
|
{
|
|
"epoch": 0.5477239353891337,
|
|
"grad_norm": 14.019618034362793,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.015411078929901123,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 2.538498388222517e-07,
|
|
"logits/chosen": -7.477275848388672,
|
|
"logits/rejected": -7.463842868804932,
|
|
"logps/chosen": -201.16326904296875,
|
|
"logps/ref_chosen": -100.07307434082031,
|
|
"logps/ref_rejected": -91.11634826660156,
|
|
"logps/rejected": -247.42393493652344,
|
|
"loss": 0.8783,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.5517187118530273,
|
|
"rewards/margins": 0.8413230180740356,
|
|
"rewards/rejected": -2.3930416107177734,
|
|
"step": 373
|
|
},
|
|
{
|
|
"epoch": 0.5491923641703378,
|
|
"grad_norm": 11.548641204833984,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.015315357595682144,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 2.525666155755725e-07,
|
|
"logits/chosen": -7.55961799621582,
|
|
"logits/rejected": -7.308844089508057,
|
|
"logps/chosen": -188.9845733642578,
|
|
"logps/ref_chosen": -98.01234436035156,
|
|
"logps/ref_rejected": -100.75224304199219,
|
|
"logps/rejected": -249.3828582763672,
|
|
"loss": 0.8132,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.3845758438110352,
|
|
"rewards/margins": 0.8738114833831787,
|
|
"rewards/rejected": -2.258387327194214,
|
|
"step": 374
|
|
},
|
|
{
|
|
"epoch": 0.5506607929515418,
|
|
"grad_norm": 12.984651565551758,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.015210784040391445,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 2.512833246961859e-07,
|
|
"logits/chosen": -7.9140520095825195,
|
|
"logits/rejected": -7.584393501281738,
|
|
"logps/chosen": -180.4878387451172,
|
|
"logps/ref_chosen": -84.39627838134766,
|
|
"logps/ref_rejected": -95.7188491821289,
|
|
"logps/rejected": -254.66061401367188,
|
|
"loss": 0.8142,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.4533175230026245,
|
|
"rewards/margins": 0.9453805685043335,
|
|
"rewards/rejected": -2.398698091506958,
|
|
"step": 375
|
|
},
|
|
{
|
|
"epoch": 0.5521292217327459,
|
|
"grad_norm": 12.451373100280762,
|
|
"kl/avg_steps": 0.75,
|
|
"kl/beta": 0.015102236531674862,
|
|
"kl/n_epsilon_steps": 0.125,
|
|
"kl/p_epsilon_steps": 0.875,
|
|
"learning_rate": 2.5e-07,
|
|
"logits/chosen": -7.593746662139893,
|
|
"logits/rejected": -7.532279014587402,
|
|
"logps/chosen": -187.0894775390625,
|
|
"logps/ref_chosen": -95.73770141601562,
|
|
"logps/ref_rejected": -111.97331237792969,
|
|
"logps/rejected": -271.1904296875,
|
|
"loss": 0.7789,
|
|
"rewards/accuracies": 0.890625,
|
|
"rewards/chosen": -1.3709272146224976,
|
|
"rewards/margins": 1.0137239694595337,
|
|
"rewards/rejected": -2.3846511840820312,
|
|
"step": 376
|
|
},
|
|
{
|
|
"epoch": 0.55359765051395,
|
|
"grad_norm": 14.200995445251465,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.014989812858402729,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 2.487166753038141e-07,
|
|
"logits/chosen": -7.635054588317871,
|
|
"logits/rejected": -7.0941009521484375,
|
|
"logps/chosen": -178.06936645507812,
|
|
"logps/ref_chosen": -78.92694854736328,
|
|
"logps/ref_rejected": -105.89892578125,
|
|
"logps/rejected": -275.2508544921875,
|
|
"loss": 0.7926,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.4772100448608398,
|
|
"rewards/margins": 1.0408310890197754,
|
|
"rewards/rejected": -2.5180411338806152,
|
|
"step": 377
|
|
},
|
|
{
|
|
"epoch": 0.5550660792951542,
|
|
"grad_norm": 13.540968894958496,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.014887461438775063,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 2.4743338442442754e-07,
|
|
"logits/chosen": -7.783210754394531,
|
|
"logits/rejected": -7.034327507019043,
|
|
"logps/chosen": -172.66668701171875,
|
|
"logps/ref_chosen": -75.37240600585938,
|
|
"logps/ref_rejected": -93.70409393310547,
|
|
"logps/rejected": -250.10272216796875,
|
|
"loss": 0.8581,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.4412063360214233,
|
|
"rewards/margins": 0.869320273399353,
|
|
"rewards/rejected": -2.3105268478393555,
|
|
"step": 378
|
|
},
|
|
{
|
|
"epoch": 0.5565345080763583,
|
|
"grad_norm": 12.696690559387207,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.014785809442400932,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 2.461501611777483e-07,
|
|
"logits/chosen": -7.8199920654296875,
|
|
"logits/rejected": -6.8525214195251465,
|
|
"logps/chosen": -195.7200927734375,
|
|
"logps/ref_chosen": -88.85568237304688,
|
|
"logps/ref_rejected": -121.38114929199219,
|
|
"logps/rejected": -284.5279541015625,
|
|
"loss": 0.9442,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -1.5753943920135498,
|
|
"rewards/margins": 0.8215144276618958,
|
|
"rewards/rejected": -2.396908760070801,
|
|
"step": 379
|
|
},
|
|
{
|
|
"epoch": 0.5580029368575624,
|
|
"grad_norm": 14.155634880065918,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.014707674272358418,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 2.4486703937790243e-07,
|
|
"logits/chosen": -7.855627059936523,
|
|
"logits/rejected": -7.288210868835449,
|
|
"logps/chosen": -197.644775390625,
|
|
"logps/ref_chosen": -79.83321380615234,
|
|
"logps/ref_rejected": -110.74475860595703,
|
|
"logps/rejected": -291.2830810546875,
|
|
"loss": 0.8692,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.7239048480987549,
|
|
"rewards/margins": 0.9118264317512512,
|
|
"rewards/rejected": -2.6357314586639404,
|
|
"step": 380
|
|
},
|
|
{
|
|
"epoch": 0.5594713656387665,
|
|
"grad_norm": 14.164735794067383,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.014611784368753433,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 2.435840528363426e-07,
|
|
"logits/chosen": -7.805541038513184,
|
|
"logits/rejected": -6.962313652038574,
|
|
"logps/chosen": -190.10198974609375,
|
|
"logps/ref_chosen": -86.56683349609375,
|
|
"logps/ref_rejected": -84.416748046875,
|
|
"logps/rejected": -243.78826904296875,
|
|
"loss": 0.8887,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.5055629014968872,
|
|
"rewards/margins": 0.8068733811378479,
|
|
"rewards/rejected": -2.312436580657959,
|
|
"step": 381
|
|
},
|
|
{
|
|
"epoch": 0.5609397944199707,
|
|
"grad_norm": 12.716059684753418,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.01452102791517973,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 2.4230123536095745e-07,
|
|
"logits/chosen": -7.854497909545898,
|
|
"logits/rejected": -7.399716377258301,
|
|
"logps/chosen": -207.37265014648438,
|
|
"logps/ref_chosen": -104.46180725097656,
|
|
"logps/ref_rejected": -114.60223388671875,
|
|
"logps/rejected": -278.6050109863281,
|
|
"loss": 0.8502,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.4875770807266235,
|
|
"rewards/margins": 0.8772394061088562,
|
|
"rewards/rejected": -2.364816427230835,
|
|
"step": 382
|
|
},
|
|
{
|
|
"epoch": 0.5624082232011748,
|
|
"grad_norm": 13.573335647583008,
|
|
"kl/avg_steps": 0.4375,
|
|
"kl/beta": 0.014435318298637867,
|
|
"kl/n_epsilon_steps": 0.28125,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 2.4101862075518037e-07,
|
|
"logits/chosen": -7.561490058898926,
|
|
"logits/rejected": -7.388404369354248,
|
|
"logps/chosen": -192.78128051757812,
|
|
"logps/ref_chosen": -82.11607360839844,
|
|
"logps/ref_rejected": -98.77716064453125,
|
|
"logps/rejected": -262.2416076660156,
|
|
"loss": 0.927,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -1.593533992767334,
|
|
"rewards/margins": 0.7530485987663269,
|
|
"rewards/rejected": -2.3465826511383057,
|
|
"step": 383
|
|
},
|
|
{
|
|
"epoch": 0.5638766519823789,
|
|
"grad_norm": 14.46293830871582,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.014372439123690128,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 2.397362428170992e-07,
|
|
"logits/chosen": -7.834491729736328,
|
|
"logits/rejected": -7.562644004821777,
|
|
"logps/chosen": -197.14675903320312,
|
|
"logps/ref_chosen": -80.58525085449219,
|
|
"logps/ref_rejected": -91.44789123535156,
|
|
"logps/rejected": -262.77398681640625,
|
|
"loss": 0.9637,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.6679697036743164,
|
|
"rewards/margins": 0.7766201496124268,
|
|
"rewards/rejected": -2.444589853286743,
|
|
"step": 384
|
|
},
|
|
{
|
|
"epoch": 0.5653450807635829,
|
|
"grad_norm": 13.538987159729004,
|
|
"kl/avg_steps": 0.546875,
|
|
"kl/beta": 0.01428316906094551,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 2.3845413533856514e-07,
|
|
"logits/chosen": -7.669012069702148,
|
|
"logits/rejected": -7.240923881530762,
|
|
"logps/chosen": -192.1517333984375,
|
|
"logps/ref_chosen": -92.76016235351562,
|
|
"logps/ref_rejected": -84.45997619628906,
|
|
"logps/rejected": -247.5975341796875,
|
|
"loss": 0.8706,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -1.4141170978546143,
|
|
"rewards/margins": 0.8999189138412476,
|
|
"rewards/rejected": -2.3140358924865723,
|
|
"step": 385
|
|
},
|
|
{
|
|
"epoch": 0.566813509544787,
|
|
"grad_norm": 11.718225479125977,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.01420548278838396,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 2.3717233210430254e-07,
|
|
"logits/chosen": -7.931042671203613,
|
|
"logits/rejected": -7.277120590209961,
|
|
"logps/chosen": -188.76788330078125,
|
|
"logps/ref_chosen": -85.35896301269531,
|
|
"logps/ref_rejected": -96.99065399169922,
|
|
"logps/rejected": -262.10107421875,
|
|
"loss": 0.8706,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.461796522140503,
|
|
"rewards/margins": 0.8657737970352173,
|
|
"rewards/rejected": -2.3275704383850098,
|
|
"step": 386
|
|
},
|
|
{
|
|
"epoch": 0.5682819383259912,
|
|
"grad_norm": 12.556093215942383,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.014112867414951324,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 2.3589086689101889e-07,
|
|
"logits/chosen": -7.733986854553223,
|
|
"logits/rejected": -6.640727519989014,
|
|
"logps/chosen": -217.92800903320312,
|
|
"logps/ref_chosen": -102.68814086914062,
|
|
"logps/ref_rejected": -99.42858123779297,
|
|
"logps/rejected": -274.15191650390625,
|
|
"loss": 0.8604,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.6170012950897217,
|
|
"rewards/margins": 0.8299874067306519,
|
|
"rewards/rejected": -2.446988582611084,
|
|
"step": 387
|
|
},
|
|
{
|
|
"epoch": 0.5697503671071953,
|
|
"grad_norm": 12.432695388793945,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.014012155123054981,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 2.3460977346651428e-07,
|
|
"logits/chosen": -7.944313049316406,
|
|
"logits/rejected": -7.454949855804443,
|
|
"logps/chosen": -197.45010375976562,
|
|
"logps/ref_chosen": -83.58523559570312,
|
|
"logps/ref_rejected": -110.10037231445312,
|
|
"logps/rejected": -292.4563293457031,
|
|
"loss": 0.7893,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -1.5865106582641602,
|
|
"rewards/margins": 0.9495990872383118,
|
|
"rewards/rejected": -2.5361099243164062,
|
|
"step": 388
|
|
},
|
|
{
|
|
"epoch": 0.5712187958883994,
|
|
"grad_norm": 13.698887825012207,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.013916479423642159,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 2.3332908558879177e-07,
|
|
"logits/chosen": -7.64278507232666,
|
|
"logits/rejected": -7.23593807220459,
|
|
"logps/chosen": -209.8504180908203,
|
|
"logps/ref_chosen": -89.822265625,
|
|
"logps/ref_rejected": -84.85377502441406,
|
|
"logps/rejected": -264.9825134277344,
|
|
"loss": 0.9298,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -1.663960337638855,
|
|
"rewards/margins": 0.8260251879692078,
|
|
"rewards/rejected": -2.489985704421997,
|
|
"step": 389
|
|
},
|
|
{
|
|
"epoch": 0.5726872246696035,
|
|
"grad_norm": 12.986654281616211,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.013842938467860222,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 2.320488370051681e-07,
|
|
"logits/chosen": -7.480320453643799,
|
|
"logits/rejected": -7.105579853057861,
|
|
"logps/chosen": -198.42169189453125,
|
|
"logps/ref_chosen": -77.53244018554688,
|
|
"logps/ref_rejected": -92.58392333984375,
|
|
"logps/rejected": -268.6641540527344,
|
|
"loss": 0.971,
|
|
"rewards/accuracies": 0.734375,
|
|
"rewards/chosen": -1.6698600053787231,
|
|
"rewards/margins": 0.7543532252311707,
|
|
"rewards/rejected": -2.424213171005249,
|
|
"step": 390
|
|
},
|
|
{
|
|
"epoch": 0.5741556534508077,
|
|
"grad_norm": 13.51944351196289,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.013778353109955788,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 2.3076906145138405e-07,
|
|
"logits/chosen": -7.788546085357666,
|
|
"logits/rejected": -6.940990447998047,
|
|
"logps/chosen": -198.9241943359375,
|
|
"logps/ref_chosen": -88.2329330444336,
|
|
"logps/ref_rejected": -93.00756072998047,
|
|
"logps/rejected": -264.8871765136719,
|
|
"loss": 0.9043,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.5192664861679077,
|
|
"rewards/margins": 0.8330180048942566,
|
|
"rewards/rejected": -2.3522844314575195,
|
|
"step": 391
|
|
},
|
|
{
|
|
"epoch": 0.5756240822320118,
|
|
"grad_norm": 13.640396118164062,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.013697026297450066,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 2.294897926507156e-07,
|
|
"logits/chosen": -7.636374473571777,
|
|
"logits/rejected": -7.049764156341553,
|
|
"logps/chosen": -198.85385131835938,
|
|
"logps/ref_chosen": -87.74441528320312,
|
|
"logps/ref_rejected": -108.57071685791016,
|
|
"logps/rejected": -287.69525146484375,
|
|
"loss": 0.8565,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.5143227577209473,
|
|
"rewards/margins": 0.9206820726394653,
|
|
"rewards/rejected": -2.435004711151123,
|
|
"step": 392
|
|
},
|
|
{
|
|
"epoch": 0.5770925110132159,
|
|
"grad_norm": 11.787603378295898,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.013607725501060486,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 2.2821106431308543e-07,
|
|
"logits/chosen": -7.288888931274414,
|
|
"logits/rejected": -6.870842933654785,
|
|
"logps/chosen": -176.33932495117188,
|
|
"logps/ref_chosen": -76.56381225585938,
|
|
"logps/ref_rejected": -79.72618865966797,
|
|
"logps/rejected": -242.21641540527344,
|
|
"loss": 0.8682,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -1.3515516519546509,
|
|
"rewards/margins": 0.8435693979263306,
|
|
"rewards/rejected": -2.1951210498809814,
|
|
"step": 393
|
|
},
|
|
{
|
|
"epoch": 0.57856093979442,
|
|
"grad_norm": 12.555571556091309,
|
|
"kl/avg_steps": 0.703125,
|
|
"kl/beta": 0.013523206114768982,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 2.2693291013417452e-07,
|
|
"logits/chosen": -7.870112895965576,
|
|
"logits/rejected": -7.220252513885498,
|
|
"logps/chosen": -188.681640625,
|
|
"logps/ref_chosen": -84.69866180419922,
|
|
"logps/ref_rejected": -97.457763671875,
|
|
"logps/rejected": -268.5348815917969,
|
|
"loss": 0.8111,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.3974303007125854,
|
|
"rewards/margins": 0.8974814414978027,
|
|
"rewards/rejected": -2.2949118614196777,
|
|
"step": 394
|
|
},
|
|
{
|
|
"epoch": 0.580029368575624,
|
|
"grad_norm": 14.517809867858887,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.013428784906864166,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 2.2565536379453404e-07,
|
|
"logits/chosen": -7.730129718780518,
|
|
"logits/rejected": -7.53689432144165,
|
|
"logps/chosen": -195.88299560546875,
|
|
"logps/ref_chosen": -88.88763427734375,
|
|
"logps/ref_rejected": -91.0455322265625,
|
|
"logps/rejected": -254.3079376220703,
|
|
"loss": 1.0664,
|
|
"rewards/accuracies": 0.71875,
|
|
"rewards/chosen": -1.4323430061340332,
|
|
"rewards/margins": 0.7445914149284363,
|
|
"rewards/rejected": -2.1769344806671143,
|
|
"step": 395
|
|
},
|
|
{
|
|
"epoch": 0.5814977973568282,
|
|
"grad_norm": 16.962797164916992,
|
|
"kl/avg_steps": 0.40625,
|
|
"kl/beta": 0.01335367001593113,
|
|
"kl/n_epsilon_steps": 0.296875,
|
|
"kl/p_epsilon_steps": 0.703125,
|
|
"learning_rate": 2.2437845895869825e-07,
|
|
"logits/chosen": -7.620447158813477,
|
|
"logits/rejected": -7.031801223754883,
|
|
"logps/chosen": -218.27194213867188,
|
|
"logps/ref_chosen": -102.47132873535156,
|
|
"logps/ref_rejected": -93.23478698730469,
|
|
"logps/rejected": -254.623046875,
|
|
"loss": 1.0318,
|
|
"rewards/accuracies": 0.734375,
|
|
"rewards/chosen": -1.542496919631958,
|
|
"rewards/margins": 0.6003559827804565,
|
|
"rewards/rejected": -2.142852783203125,
|
|
"step": 396
|
|
},
|
|
{
|
|
"epoch": 0.5829662261380323,
|
|
"grad_norm": 13.833749771118164,
|
|
"kl/avg_steps": 0.578125,
|
|
"kl/beta": 0.013299640268087387,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 2.2310222927429716e-07,
|
|
"logits/chosen": -8.157541275024414,
|
|
"logits/rejected": -7.536031246185303,
|
|
"logps/chosen": -204.63792419433594,
|
|
"logps/ref_chosen": -101.04441833496094,
|
|
"logps/ref_rejected": -108.8511734008789,
|
|
"logps/rejected": -296.58837890625,
|
|
"loss": 0.794,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.3734837770462036,
|
|
"rewards/margins": 1.1063416004180908,
|
|
"rewards/rejected": -2.479825496673584,
|
|
"step": 397
|
|
},
|
|
{
|
|
"epoch": 0.5844346549192364,
|
|
"grad_norm": 13.908939361572266,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.01322319358587265,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 2.2182670837116972e-07,
|
|
"logits/chosen": -7.908001899719238,
|
|
"logits/rejected": -7.516959190368652,
|
|
"logps/chosen": -204.48220825195312,
|
|
"logps/ref_chosen": -95.04922485351562,
|
|
"logps/ref_rejected": -114.00563049316406,
|
|
"logps/rejected": -286.8526611328125,
|
|
"loss": 0.9283,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -1.4388189315795898,
|
|
"rewards/margins": 0.82853102684021,
|
|
"rewards/rejected": -2.2673499584198,
|
|
"step": 398
|
|
},
|
|
{
|
|
"epoch": 0.5859030837004405,
|
|
"grad_norm": 12.327967643737793,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.013136982917785645,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 2.2055192986047804e-07,
|
|
"logits/chosen": -7.728894233703613,
|
|
"logits/rejected": -7.285490989685059,
|
|
"logps/chosen": -196.3619384765625,
|
|
"logps/ref_chosen": -96.63853454589844,
|
|
"logps/ref_rejected": -83.53645324707031,
|
|
"logps/rejected": -262.048828125,
|
|
"loss": 0.7418,
|
|
"rewards/accuracies": 0.921875,
|
|
"rewards/chosen": -1.3026535511016846,
|
|
"rewards/margins": 1.0243358612060547,
|
|
"rewards/rejected": -2.3269894123077393,
|
|
"step": 399
|
|
},
|
|
{
|
|
"epoch": 0.5873715124816447,
|
|
"grad_norm": 14.875067710876465,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.013043234124779701,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 2.192779273338215e-07,
|
|
"logits/chosen": -8.027582168579102,
|
|
"logits/rejected": -7.236846923828125,
|
|
"logps/chosen": -204.7100067138672,
|
|
"logps/ref_chosen": -105.30314636230469,
|
|
"logps/ref_rejected": -108.6555404663086,
|
|
"logps/rejected": -291.34063720703125,
|
|
"loss": 0.7868,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -1.2911700010299683,
|
|
"rewards/margins": 1.073652744293213,
|
|
"rewards/rejected": -2.3648228645324707,
|
|
"step": 400
|
|
},
|
|
{
|
|
"epoch": 0.5873715124816447,
|
|
"eval_kl/n_epsilon_steps": 0.23416095972061157,
|
|
"eval_kl/p_epsilon_steps": 0.7645547986030579,
|
|
"eval_logits/chosen": -7.477985382080078,
|
|
"eval_logits/rejected": -7.118831157684326,
|
|
"eval_logps/chosen": -222.89894104003906,
|
|
"eval_logps/ref_chosen": -100.49356842041016,
|
|
"eval_logps/ref_rejected": -94.06775665283203,
|
|
"eval_logps/rejected": -269.13983154296875,
|
|
"eval_loss": 0.49379390478134155,
|
|
"eval_rewards/accuracies": 0.7791095972061157,
|
|
"eval_rewards/chosen": -1.5799260139465332,
|
|
"eval_rewards/margins": 0.673462986946106,
|
|
"eval_rewards/rejected": -2.2533888816833496,
|
|
"eval_runtime": 48.5073,
|
|
"eval_samples_per_second": 48.22,
|
|
"eval_steps_per_second": 1.526,
|
|
"step": 400
|
|
},
|
|
{
|
|
"epoch": 0.5888399412628488,
|
|
"grad_norm": 15.558944702148438,
|
|
"kl/avg_steps": 0.375,
|
|
"kl/beta": 0.012954174540936947,
|
|
"kl/n_epsilon_steps": 0.3125,
|
|
"kl/p_epsilon_steps": 0.6875,
|
|
"learning_rate": 2.1800473436235136e-07,
|
|
"logits/chosen": -7.620119094848633,
|
|
"logits/rejected": -7.406888961791992,
|
|
"logps/chosen": -197.28976440429688,
|
|
"logps/ref_chosen": -84.28916931152344,
|
|
"logps/ref_rejected": -90.943115234375,
|
|
"logps/rejected": -255.7905731201172,
|
|
"loss": 1.0429,
|
|
"rewards/accuracies": 0.703125,
|
|
"rewards/chosen": -1.4610803127288818,
|
|
"rewards/margins": 0.6627606749534607,
|
|
"rewards/rejected": -2.123840808868408,
|
|
"step": 401
|
|
},
|
|
{
|
|
"epoch": 0.5903083700440529,
|
|
"grad_norm": 12.638237953186035,
|
|
"kl/avg_steps": 0.78125,
|
|
"kl/beta": 0.012905777432024479,
|
|
"kl/n_epsilon_steps": 0.109375,
|
|
"kl/p_epsilon_steps": 0.890625,
|
|
"learning_rate": 2.1673238449588665e-07,
|
|
"logits/chosen": -7.716352939605713,
|
|
"logits/rejected": -7.249260902404785,
|
|
"logps/chosen": -185.138916015625,
|
|
"logps/ref_chosen": -83.59312438964844,
|
|
"logps/ref_rejected": -87.81027221679688,
|
|
"logps/rejected": -284.30706787109375,
|
|
"loss": 0.6486,
|
|
"rewards/accuracies": 0.90625,
|
|
"rewards/chosen": -1.3018509149551392,
|
|
"rewards/margins": 1.2127395868301392,
|
|
"rewards/rejected": -2.5145905017852783,
|
|
"step": 402
|
|
},
|
|
{
|
|
"epoch": 0.591776798825257,
|
|
"grad_norm": 11.707761764526367,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.012805732898414135,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 2.154609112620295e-07,
|
|
"logits/chosen": -7.694974899291992,
|
|
"logits/rejected": -6.938326835632324,
|
|
"logps/chosen": -179.35401916503906,
|
|
"logps/ref_chosen": -73.75308227539062,
|
|
"logps/ref_rejected": -83.92012786865234,
|
|
"logps/rejected": -260.35247802734375,
|
|
"loss": 0.8488,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.3442716598510742,
|
|
"rewards/margins": 0.8966612219810486,
|
|
"rewards/rejected": -2.2409329414367676,
|
|
"step": 403
|
|
},
|
|
{
|
|
"epoch": 0.593245227606461,
|
|
"grad_norm": 17.042625427246094,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.01271434873342514,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 2.1419034816528218e-07,
|
|
"logits/chosen": -7.690559387207031,
|
|
"logits/rejected": -7.393105983734131,
|
|
"logps/chosen": -203.4947509765625,
|
|
"logps/ref_chosen": -79.67617797851562,
|
|
"logps/ref_rejected": -84.280517578125,
|
|
"logps/rejected": -271.59332275390625,
|
|
"loss": 0.9664,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.567918300628662,
|
|
"rewards/margins": 0.7968522906303406,
|
|
"rewards/rejected": -2.3647704124450684,
|
|
"step": 404
|
|
},
|
|
{
|
|
"epoch": 0.5947136563876652,
|
|
"grad_norm": 11.477922439575195,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.012639302760362625,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 2.129207286861638e-07,
|
|
"logits/chosen": -7.6790571212768555,
|
|
"logits/rejected": -7.126956462860107,
|
|
"logps/chosen": -224.55743408203125,
|
|
"logps/ref_chosen": -96.46195220947266,
|
|
"logps/ref_rejected": -92.87071228027344,
|
|
"logps/rejected": -283.11083984375,
|
|
"loss": 0.8803,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.6104973554611206,
|
|
"rewards/margins": 0.7770987749099731,
|
|
"rewards/rejected": -2.3875961303710938,
|
|
"step": 405
|
|
},
|
|
{
|
|
"epoch": 0.5961820851688693,
|
|
"grad_norm": 16.791955947875977,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.012556898407638073,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 2.1165208628032861e-07,
|
|
"logits/chosen": -7.518543720245361,
|
|
"logits/rejected": -7.182583808898926,
|
|
"logps/chosen": -199.13026428222656,
|
|
"logps/ref_chosen": -78.13396453857422,
|
|
"logps/ref_rejected": -98.28359985351562,
|
|
"logps/rejected": -276.6565246582031,
|
|
"loss": 0.9891,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -1.5131518840789795,
|
|
"rewards/margins": 0.7109516859054565,
|
|
"rewards/rejected": -2.2241034507751465,
|
|
"step": 406
|
|
},
|
|
{
|
|
"epoch": 0.5976505139500734,
|
|
"grad_norm": 14.285913467407227,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.012482781894505024,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 2.1038445437768375e-07,
|
|
"logits/chosen": -7.507756233215332,
|
|
"logits/rejected": -6.921146392822266,
|
|
"logps/chosen": -212.05181884765625,
|
|
"logps/ref_chosen": -84.01283264160156,
|
|
"logps/ref_rejected": -82.78103637695312,
|
|
"logps/rejected": -264.0316162109375,
|
|
"loss": 0.9939,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -1.5919959545135498,
|
|
"rewards/margins": 0.6562482118606567,
|
|
"rewards/rejected": -2.248244285583496,
|
|
"step": 407
|
|
},
|
|
{
|
|
"epoch": 0.5991189427312775,
|
|
"grad_norm": 13.731730461120605,
|
|
"kl/avg_steps": 0.421875,
|
|
"kl/beta": 0.012412958778440952,
|
|
"kl/n_epsilon_steps": 0.28125,
|
|
"kl/p_epsilon_steps": 0.703125,
|
|
"learning_rate": 2.0911786638150872e-07,
|
|
"logits/chosen": -7.530098915100098,
|
|
"logits/rejected": -7.0652008056640625,
|
|
"logps/chosen": -224.00033569335938,
|
|
"logps/ref_chosen": -104.46175384521484,
|
|
"logps/ref_rejected": -96.37218475341797,
|
|
"logps/rejected": -270.20428466796875,
|
|
"loss": 1.011,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -1.480905294418335,
|
|
"rewards/margins": 0.6653140783309937,
|
|
"rewards/rejected": -2.146219253540039,
|
|
"step": 408
|
|
},
|
|
{
|
|
"epoch": 0.6005873715124816,
|
|
"grad_norm": 12.978145599365234,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.012360811233520508,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 2.0785235566757517e-07,
|
|
"logits/chosen": -7.490418434143066,
|
|
"logits/rejected": -7.061441898345947,
|
|
"logps/chosen": -229.72506713867188,
|
|
"logps/ref_chosen": -97.66830444335938,
|
|
"logps/ref_rejected": -90.04584503173828,
|
|
"logps/rejected": -275.54852294921875,
|
|
"loss": 1.0303,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.6262993812561035,
|
|
"rewards/margins": 0.6515868306159973,
|
|
"rewards/rejected": -2.277886152267456,
|
|
"step": 409
|
|
},
|
|
{
|
|
"epoch": 0.6020558002936858,
|
|
"grad_norm": 12.724774360656738,
|
|
"kl/avg_steps": 0.578125,
|
|
"kl/beta": 0.01229167077690363,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 2.065879555832674e-07,
|
|
"logits/chosen": -7.461033821105957,
|
|
"logits/rejected": -7.218528747558594,
|
|
"logps/chosen": -205.697265625,
|
|
"logps/ref_chosen": -76.46923828125,
|
|
"logps/ref_rejected": -88.64064025878906,
|
|
"logps/rejected": -275.3063049316406,
|
|
"loss": 0.9507,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -1.5810291767120361,
|
|
"rewards/margins": 0.697364091873169,
|
|
"rewards/rejected": -2.278393268585205,
|
|
"step": 410
|
|
},
|
|
{
|
|
"epoch": 0.6035242290748899,
|
|
"grad_norm": 11.978557586669922,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.012221017852425575,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 2.0532469944670343e-07,
|
|
"logits/chosen": -7.684117317199707,
|
|
"logits/rejected": -7.11949348449707,
|
|
"logps/chosen": -221.50474548339844,
|
|
"logps/ref_chosen": -87.16630554199219,
|
|
"logps/ref_rejected": -87.09603118896484,
|
|
"logps/rejected": -281.2148742675781,
|
|
"loss": 0.9201,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -1.6353148221969604,
|
|
"rewards/margins": 0.7224369049072266,
|
|
"rewards/rejected": -2.3577518463134766,
|
|
"step": 411
|
|
},
|
|
{
|
|
"epoch": 0.604992657856094,
|
|
"grad_norm": 13.545778274536133,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.012156437151134014,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 2.0406262054585738e-07,
|
|
"logits/chosen": -7.71500301361084,
|
|
"logits/rejected": -6.960906982421875,
|
|
"logps/chosen": -205.3089599609375,
|
|
"logps/ref_chosen": -78.94734191894531,
|
|
"logps/ref_rejected": -106.10554504394531,
|
|
"logps/rejected": -289.611572265625,
|
|
"loss": 1.0053,
|
|
"rewards/accuracies": 0.71875,
|
|
"rewards/chosen": -1.5305845737457275,
|
|
"rewards/margins": 0.6853532791137695,
|
|
"rewards/rejected": -2.215937614440918,
|
|
"step": 412
|
|
},
|
|
{
|
|
"epoch": 0.6064610866372981,
|
|
"grad_norm": 12.726061820983887,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.012099719606339931,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 2.0280175213768205e-07,
|
|
"logits/chosen": -7.195650100708008,
|
|
"logits/rejected": -7.105502128601074,
|
|
"logps/chosen": -232.97862243652344,
|
|
"logps/ref_chosen": -95.69471740722656,
|
|
"logps/ref_rejected": -107.96085357666016,
|
|
"logps/rejected": -308.6137390136719,
|
|
"loss": 0.9401,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.650810956954956,
|
|
"rewards/margins": 0.7573432922363281,
|
|
"rewards/rejected": -2.4081544876098633,
|
|
"step": 413
|
|
},
|
|
{
|
|
"epoch": 0.6079295154185022,
|
|
"grad_norm": 16.81880760192871,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.012013372965157032,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 2.0154212744723247e-07,
|
|
"logits/chosen": -7.222269058227539,
|
|
"logits/rejected": -6.936489105224609,
|
|
"logps/chosen": -214.6531524658203,
|
|
"logps/ref_chosen": -88.27667236328125,
|
|
"logps/ref_rejected": -92.87004089355469,
|
|
"logps/rejected": -278.8741455078125,
|
|
"loss": 0.9978,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -1.5135799646377563,
|
|
"rewards/margins": 0.7068269848823547,
|
|
"rewards/rejected": -2.220407009124756,
|
|
"step": 414
|
|
},
|
|
{
|
|
"epoch": 0.6093979441997063,
|
|
"grad_norm": 13.74764633178711,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.011949889361858368,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 2.002837796667909e-07,
|
|
"logits/chosen": -7.991650104522705,
|
|
"logits/rejected": -7.442322731018066,
|
|
"logps/chosen": -242.48207092285156,
|
|
"logps/ref_chosen": -108.91590118408203,
|
|
"logps/ref_rejected": -107.47135925292969,
|
|
"logps/rejected": -295.0639953613281,
|
|
"loss": 1.0108,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -1.5901520252227783,
|
|
"rewards/margins": 0.6369077563285828,
|
|
"rewards/rejected": -2.227059841156006,
|
|
"step": 415
|
|
},
|
|
{
|
|
"epoch": 0.6108663729809104,
|
|
"grad_norm": 11.695809364318848,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.01188304740935564,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 1.990267419549914e-07,
|
|
"logits/chosen": -7.476058006286621,
|
|
"logits/rejected": -7.21858024597168,
|
|
"logps/chosen": -224.93905639648438,
|
|
"logps/ref_chosen": -93.39888000488281,
|
|
"logps/ref_rejected": -97.6729736328125,
|
|
"logps/rejected": -302.91412353515625,
|
|
"loss": 0.875,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -1.556352138519287,
|
|
"rewards/margins": 0.8654446601867676,
|
|
"rewards/rejected": -2.4217967987060547,
|
|
"step": 416
|
|
},
|
|
{
|
|
"epoch": 0.6123348017621145,
|
|
"grad_norm": 12.070055961608887,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.011816578917205334,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 1.9777104743594686e-07,
|
|
"logits/chosen": -7.844758033752441,
|
|
"logits/rejected": -6.70277214050293,
|
|
"logps/chosen": -208.16903686523438,
|
|
"logps/ref_chosen": -83.53533172607422,
|
|
"logps/ref_rejected": -74.44184112548828,
|
|
"logps/rejected": -266.1756591796875,
|
|
"loss": 0.8896,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.4651763439178467,
|
|
"rewards/margins": 0.7836226224899292,
|
|
"rewards/rejected": -2.2487990856170654,
|
|
"step": 417
|
|
},
|
|
{
|
|
"epoch": 0.6138032305433186,
|
|
"grad_norm": 13.669163703918457,
|
|
"kl/avg_steps": 0.84375,
|
|
"kl/beta": 0.011735894717276096,
|
|
"kl/n_epsilon_steps": 0.078125,
|
|
"kl/p_epsilon_steps": 0.921875,
|
|
"learning_rate": 1.965167291983757e-07,
|
|
"logits/chosen": -7.78868293762207,
|
|
"logits/rejected": -7.488645553588867,
|
|
"logps/chosen": -225.50465393066406,
|
|
"logps/ref_chosen": -108.22152709960938,
|
|
"logps/ref_rejected": -111.8646469116211,
|
|
"logps/rejected": -311.86407470703125,
|
|
"loss": 0.7851,
|
|
"rewards/accuracies": 0.921875,
|
|
"rewards/chosen": -1.3652478456497192,
|
|
"rewards/margins": 0.9605357646942139,
|
|
"rewards/rejected": -2.3257837295532227,
|
|
"step": 418
|
|
},
|
|
{
|
|
"epoch": 0.6152716593245228,
|
|
"grad_norm": 11.460182189941406,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.011637701652944088,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 1.9526382029472988e-07,
|
|
"logits/chosen": -7.678118705749512,
|
|
"logits/rejected": -7.43233585357666,
|
|
"logps/chosen": -223.0901641845703,
|
|
"logps/ref_chosen": -97.18328094482422,
|
|
"logps/ref_rejected": -98.18531799316406,
|
|
"logps/rejected": -296.13623046875,
|
|
"loss": 0.8759,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.4567546844482422,
|
|
"rewards/margins": 0.8287121057510376,
|
|
"rewards/rejected": -2.2854669094085693,
|
|
"step": 419
|
|
},
|
|
{
|
|
"epoch": 0.6167400881057269,
|
|
"grad_norm": 14.639280319213867,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.011554652824997902,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 1.9401235374032425e-07,
|
|
"logits/chosen": -7.911190509796143,
|
|
"logits/rejected": -6.9983086585998535,
|
|
"logps/chosen": -243.3070831298828,
|
|
"logps/ref_chosen": -114.30847930908203,
|
|
"logps/ref_rejected": -75.68356323242188,
|
|
"logps/rejected": -270.84930419921875,
|
|
"loss": 0.9212,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -1.4852039813995361,
|
|
"rewards/margins": 0.7554365396499634,
|
|
"rewards/rejected": -2.24064040184021,
|
|
"step": 420
|
|
},
|
|
{
|
|
"epoch": 0.618208516886931,
|
|
"grad_norm": 11.352572441101074,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.011490020900964737,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 1.9276236251246653e-07,
|
|
"logits/chosen": -7.578913688659668,
|
|
"logits/rejected": -6.917729377746582,
|
|
"logps/chosen": -215.33929443359375,
|
|
"logps/ref_chosen": -85.87985229492188,
|
|
"logps/ref_rejected": -96.33648681640625,
|
|
"logps/rejected": -295.12469482421875,
|
|
"loss": 0.8958,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -1.4818620681762695,
|
|
"rewards/margins": 0.7875877618789673,
|
|
"rewards/rejected": -2.2694497108459473,
|
|
"step": 421
|
|
},
|
|
{
|
|
"epoch": 0.6196769456681351,
|
|
"grad_norm": 11.291938781738281,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.011425751261413097,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 1.9151387954958792e-07,
|
|
"logits/chosen": -7.927703857421875,
|
|
"logits/rejected": -7.495624542236328,
|
|
"logps/chosen": -224.50631713867188,
|
|
"logps/ref_chosen": -100.48060607910156,
|
|
"logps/ref_rejected": -94.40821838378906,
|
|
"logps/rejected": -286.848388671875,
|
|
"loss": 0.8898,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.409227728843689,
|
|
"rewards/margins": 0.7730013132095337,
|
|
"rewards/rejected": -2.1822290420532227,
|
|
"step": 422
|
|
},
|
|
{
|
|
"epoch": 0.6211453744493393,
|
|
"grad_norm": 12.197092056274414,
|
|
"kl/avg_steps": 0.4375,
|
|
"kl/beta": 0.011347736231982708,
|
|
"kl/n_epsilon_steps": 0.28125,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 1.902669377503756e-07,
|
|
"logits/chosen": -7.66268253326416,
|
|
"logits/rejected": -7.711108207702637,
|
|
"logps/chosen": -198.95823669433594,
|
|
"logps/ref_chosen": -78.44993591308594,
|
|
"logps/ref_rejected": -92.04652404785156,
|
|
"logps/rejected": -276.6893310546875,
|
|
"loss": 0.9664,
|
|
"rewards/accuracies": 0.71875,
|
|
"rewards/chosen": -1.364768624305725,
|
|
"rewards/margins": 0.718655526638031,
|
|
"rewards/rejected": -2.0834240913391113,
|
|
"step": 423
|
|
},
|
|
{
|
|
"epoch": 0.6226138032305433,
|
|
"grad_norm": 11.235798835754395,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.011298305355012417,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 1.890215699729057e-07,
|
|
"logits/chosen": -7.801157474517822,
|
|
"logits/rejected": -6.635551452636719,
|
|
"logps/chosen": -220.01161193847656,
|
|
"logps/ref_chosen": -87.6423568725586,
|
|
"logps/ref_rejected": -72.36566162109375,
|
|
"logps/rejected": -272.5384521484375,
|
|
"loss": 0.9078,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -1.4883897304534912,
|
|
"rewards/margins": 0.7573517560958862,
|
|
"rewards/rejected": -2.245741367340088,
|
|
"step": 424
|
|
},
|
|
{
|
|
"epoch": 0.6240822320117474,
|
|
"grad_norm": 13.440584182739258,
|
|
"kl/avg_steps": 0.1875,
|
|
"kl/beta": 0.011231618002057076,
|
|
"kl/n_epsilon_steps": 0.40625,
|
|
"kl/p_epsilon_steps": 0.59375,
|
|
"learning_rate": 1.8777780903377732e-07,
|
|
"logits/chosen": -7.429314613342285,
|
|
"logits/rejected": -7.256626605987549,
|
|
"logps/chosen": -215.4922332763672,
|
|
"logps/ref_chosen": -78.51979064941406,
|
|
"logps/ref_rejected": -102.74864196777344,
|
|
"logps/rejected": -285.57012939453125,
|
|
"loss": 1.1393,
|
|
"rewards/accuracies": 0.625,
|
|
"rewards/chosen": -1.5395094156265259,
|
|
"rewards/margins": 0.5070829391479492,
|
|
"rewards/rejected": -2.0465922355651855,
|
|
"step": 425
|
|
},
|
|
{
|
|
"epoch": 0.6255506607929515,
|
|
"grad_norm": 13.104548454284668,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.011210598051548004,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 1.8653568770724803e-07,
|
|
"logits/chosen": -7.60335636138916,
|
|
"logits/rejected": -7.2415266036987305,
|
|
"logps/chosen": -227.1387939453125,
|
|
"logps/ref_chosen": -108.50582885742188,
|
|
"logps/ref_rejected": -88.300048828125,
|
|
"logps/rejected": -274.4301452636719,
|
|
"loss": 0.8943,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.3247995376586914,
|
|
"rewards/margins": 0.7480140924453735,
|
|
"rewards/rejected": -2.0728135108947754,
|
|
"step": 426
|
|
},
|
|
{
|
|
"epoch": 0.6270190895741556,
|
|
"grad_norm": 10.724087715148926,
|
|
"kl/avg_steps": 0.75,
|
|
"kl/beta": 0.01114789117127657,
|
|
"kl/n_epsilon_steps": 0.125,
|
|
"kl/p_epsilon_steps": 0.875,
|
|
"learning_rate": 1.8529523872436977e-07,
|
|
"logits/chosen": -7.953921794891357,
|
|
"logits/rejected": -7.36940860748291,
|
|
"logps/chosen": -200.22837829589844,
|
|
"logps/ref_chosen": -99.12046813964844,
|
|
"logps/ref_rejected": -85.724609375,
|
|
"logps/rejected": -250.37628173828125,
|
|
"loss": 0.9039,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -1.119832158088684,
|
|
"rewards/margins": 0.7005788087844849,
|
|
"rewards/rejected": -1.820410966873169,
|
|
"step": 427
|
|
},
|
|
{
|
|
"epoch": 0.6284875183553598,
|
|
"grad_norm": 12.104276657104492,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.011064904741942883,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 1.8405649477212697e-07,
|
|
"logits/chosen": -7.528921604156494,
|
|
"logits/rejected": -7.260353088378906,
|
|
"logps/chosen": -243.97320556640625,
|
|
"logps/ref_chosen": -105.96925354003906,
|
|
"logps/ref_rejected": -109.1021728515625,
|
|
"logps/rejected": -297.7687683105469,
|
|
"loss": 1.0721,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -1.520464301109314,
|
|
"rewards/margins": 0.5525679588317871,
|
|
"rewards/rejected": -2.0730321407318115,
|
|
"step": 428
|
|
},
|
|
{
|
|
"epoch": 0.6299559471365639,
|
|
"grad_norm": 13.0979642868042,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.011003012768924236,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 1.828194884925749e-07,
|
|
"logits/chosen": -7.794681072235107,
|
|
"logits/rejected": -7.093453407287598,
|
|
"logps/chosen": -237.86270141601562,
|
|
"logps/ref_chosen": -113.54486846923828,
|
|
"logps/ref_rejected": -98.24201965332031,
|
|
"logps/rejected": -292.80511474609375,
|
|
"loss": 0.9249,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -1.363885760307312,
|
|
"rewards/margins": 0.7638142108917236,
|
|
"rewards/rejected": -2.127699851989746,
|
|
"step": 429
|
|
},
|
|
{
|
|
"epoch": 0.631424375917768,
|
|
"grad_norm": 12.785799980163574,
|
|
"kl/avg_steps": 0.4375,
|
|
"kl/beta": 0.010944867506623268,
|
|
"kl/n_epsilon_steps": 0.28125,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 1.8158425248197928e-07,
|
|
"logits/chosen": -7.379518032073975,
|
|
"logits/rejected": -7.428651809692383,
|
|
"logps/chosen": -216.0718994140625,
|
|
"logps/ref_chosen": -91.31936645507812,
|
|
"logps/ref_rejected": -110.1096420288086,
|
|
"logps/rejected": -288.0528564453125,
|
|
"loss": 1.0248,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -1.361114501953125,
|
|
"rewards/margins": 0.5744451284408569,
|
|
"rewards/rejected": -1.9355595111846924,
|
|
"step": 430
|
|
},
|
|
{
|
|
"epoch": 0.6328928046989721,
|
|
"grad_norm": 11.655879974365234,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.010897193104028702,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 1.8035081928995788e-07,
|
|
"logits/chosen": -7.7428998947143555,
|
|
"logits/rejected": -7.222956657409668,
|
|
"logps/chosen": -215.51998901367188,
|
|
"logps/ref_chosen": -93.18122100830078,
|
|
"logps/ref_rejected": -98.13226318359375,
|
|
"logps/rejected": -283.9945983886719,
|
|
"loss": 0.9558,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -1.3294377326965332,
|
|
"rewards/margins": 0.6835219860076904,
|
|
"rewards/rejected": -2.0129597187042236,
|
|
"step": 431
|
|
},
|
|
{
|
|
"epoch": 0.6343612334801763,
|
|
"grad_norm": 10.997950553894043,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.010842978022992611,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 1.791192214186223e-07,
|
|
"logits/chosen": -7.891312122344971,
|
|
"logits/rejected": -7.409452438354492,
|
|
"logps/chosen": -215.7032928466797,
|
|
"logps/ref_chosen": -104.43478393554688,
|
|
"logps/ref_rejected": -105.08955383300781,
|
|
"logps/rejected": -292.071533203125,
|
|
"loss": 0.8684,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.2002313137054443,
|
|
"rewards/margins": 0.8118045330047607,
|
|
"rewards/rejected": -2.012035846710205,
|
|
"step": 432
|
|
},
|
|
{
|
|
"epoch": 0.6358296622613803,
|
|
"grad_norm": 11.87765884399414,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.010775630362331867,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 1.7788949132172193e-07,
|
|
"logits/chosen": -7.532526969909668,
|
|
"logits/rejected": -7.0036725997924805,
|
|
"logps/chosen": -220.77711486816406,
|
|
"logps/ref_chosen": -89.84322357177734,
|
|
"logps/ref_rejected": -101.73345947265625,
|
|
"logps/rejected": -292.6389465332031,
|
|
"loss": 1.001,
|
|
"rewards/accuracies": 0.71875,
|
|
"rewards/chosen": -1.4064010381698608,
|
|
"rewards/margins": 0.6378804445266724,
|
|
"rewards/rejected": -2.044281482696533,
|
|
"step": 433
|
|
},
|
|
{
|
|
"epoch": 0.6372980910425844,
|
|
"grad_norm": 11.362030982971191,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.0107253547757864,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 1.7666166140378853e-07,
|
|
"logits/chosen": -7.701730728149414,
|
|
"logits/rejected": -7.350008010864258,
|
|
"logps/chosen": -219.51129150390625,
|
|
"logps/ref_chosen": -97.6925277709961,
|
|
"logps/ref_rejected": -84.09130096435547,
|
|
"logps/rejected": -268.685302734375,
|
|
"loss": 0.984,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -1.3009238243103027,
|
|
"rewards/margins": 0.6648129820823669,
|
|
"rewards/rejected": -1.9657368659973145,
|
|
"step": 434
|
|
},
|
|
{
|
|
"epoch": 0.6387665198237885,
|
|
"grad_norm": 11.840472221374512,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.010671994648873806,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 1.7543576401928218e-07,
|
|
"logits/chosen": -7.7169880867004395,
|
|
"logits/rejected": -7.244067192077637,
|
|
"logps/chosen": -203.26406860351562,
|
|
"logps/ref_chosen": -86.17192077636719,
|
|
"logps/ref_rejected": -93.751708984375,
|
|
"logps/rejected": -278.71820068359375,
|
|
"loss": 0.9314,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.2436943054199219,
|
|
"rewards/margins": 0.7158711552619934,
|
|
"rewards/rejected": -1.9595654010772705,
|
|
"step": 435
|
|
},
|
|
{
|
|
"epoch": 0.6402349486049926,
|
|
"grad_norm": 11.141986846923828,
|
|
"kl/avg_steps": 0.78125,
|
|
"kl/beta": 0.010602416470646858,
|
|
"kl/n_epsilon_steps": 0.109375,
|
|
"kl/p_epsilon_steps": 0.890625,
|
|
"learning_rate": 1.742118314717391e-07,
|
|
"logits/chosen": -8.073395729064941,
|
|
"logits/rejected": -6.879586219787598,
|
|
"logps/chosen": -200.92079162597656,
|
|
"logps/ref_chosen": -105.78710174560547,
|
|
"logps/ref_rejected": -88.62471008300781,
|
|
"logps/rejected": -263.154296875,
|
|
"loss": 0.8215,
|
|
"rewards/accuracies": 0.90625,
|
|
"rewards/chosen": -1.0024278163909912,
|
|
"rewards/margins": 0.8327617645263672,
|
|
"rewards/rejected": -1.8351895809173584,
|
|
"step": 436
|
|
},
|
|
{
|
|
"epoch": 0.6417033773861968,
|
|
"grad_norm": 12.68950366973877,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.010520227253437042,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 1.7298989601292036e-07,
|
|
"logits/chosen": -7.771686553955078,
|
|
"logits/rejected": -7.039131164550781,
|
|
"logps/chosen": -210.84364318847656,
|
|
"logps/ref_chosen": -96.06204223632812,
|
|
"logps/ref_rejected": -89.01220703125,
|
|
"logps/rejected": -258.4656677246094,
|
|
"loss": 0.9969,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -1.2023859024047852,
|
|
"rewards/margins": 0.5679023265838623,
|
|
"rewards/rejected": -1.770288109779358,
|
|
"step": 437
|
|
},
|
|
{
|
|
"epoch": 0.6431718061674009,
|
|
"grad_norm": 11.616890907287598,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.01046463381499052,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 1.7176998984196144e-07,
|
|
"logits/chosen": -7.707596302032471,
|
|
"logits/rejected": -7.402861595153809,
|
|
"logps/chosen": -212.05523681640625,
|
|
"logps/ref_chosen": -101.85537719726562,
|
|
"logps/ref_rejected": -89.4476547241211,
|
|
"logps/rejected": -262.99102783203125,
|
|
"loss": 0.949,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -1.149721622467041,
|
|
"rewards/margins": 0.6547319889068604,
|
|
"rewards/rejected": -1.8044536113739014,
|
|
"step": 438
|
|
},
|
|
{
|
|
"epoch": 0.644640234948605,
|
|
"grad_norm": 12.508461952209473,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.010409334674477577,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 1.7055214510452458e-07,
|
|
"logits/chosen": -7.787407875061035,
|
|
"logits/rejected": -7.194408416748047,
|
|
"logps/chosen": -205.33200073242188,
|
|
"logps/ref_chosen": -81.75563049316406,
|
|
"logps/ref_rejected": -90.58635711669922,
|
|
"logps/rejected": -273.9890441894531,
|
|
"loss": 0.9918,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -1.2794837951660156,
|
|
"rewards/margins": 0.6150201559066772,
|
|
"rewards/rejected": -1.8945040702819824,
|
|
"step": 439
|
|
},
|
|
{
|
|
"epoch": 0.6461086637298091,
|
|
"grad_norm": 10.161545753479004,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.010344680398702621,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 1.6933639389195134e-07,
|
|
"logits/chosen": -7.840778827667236,
|
|
"logits/rejected": -7.435771465301514,
|
|
"logps/chosen": -216.48208618164062,
|
|
"logps/ref_chosen": -105.64108276367188,
|
|
"logps/ref_rejected": -103.40100860595703,
|
|
"logps/rejected": -284.5960388183594,
|
|
"loss": 0.923,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -1.1412960290908813,
|
|
"rewards/margins": 0.7194013595581055,
|
|
"rewards/rejected": -1.8606973886489868,
|
|
"step": 440
|
|
},
|
|
{
|
|
"epoch": 0.6475770925110133,
|
|
"grad_norm": 10.330644607543945,
|
|
"kl/avg_steps": 0.84375,
|
|
"kl/beta": 0.010280427522957325,
|
|
"kl/n_epsilon_steps": 0.078125,
|
|
"kl/p_epsilon_steps": 0.921875,
|
|
"learning_rate": 1.681227682404166e-07,
|
|
"logits/chosen": -7.877140522003174,
|
|
"logits/rejected": -6.723335266113281,
|
|
"logps/chosen": -221.08547973632812,
|
|
"logps/ref_chosen": -91.529541015625,
|
|
"logps/ref_rejected": -103.619384765625,
|
|
"logps/rejected": -306.0940856933594,
|
|
"loss": 0.8764,
|
|
"rewards/accuracies": 0.890625,
|
|
"rewards/chosen": -1.3216805458068848,
|
|
"rewards/margins": 0.7420125007629395,
|
|
"rewards/rejected": -2.063693046569824,
|
|
"step": 441
|
|
},
|
|
{
|
|
"epoch": 0.6490455212922174,
|
|
"grad_norm": 11.554139137268066,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.010194412432610989,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 1.669113001300851e-07,
|
|
"logits/chosen": -7.492166996002197,
|
|
"logits/rejected": -6.9776811599731445,
|
|
"logps/chosen": -216.06195068359375,
|
|
"logps/ref_chosen": -84.77755737304688,
|
|
"logps/ref_rejected": -83.82415008544922,
|
|
"logps/rejected": -295.6746826171875,
|
|
"loss": 0.8427,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -1.3306344747543335,
|
|
"rewards/margins": 0.8125271797180176,
|
|
"rewards/rejected": -2.1431615352630615,
|
|
"step": 442
|
|
},
|
|
{
|
|
"epoch": 0.6505139500734214,
|
|
"grad_norm": 11.208428382873535,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.010124804452061653,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 1.6570202148426815e-07,
|
|
"logits/chosen": -7.71527099609375,
|
|
"logits/rejected": -7.359766960144043,
|
|
"logps/chosen": -227.578125,
|
|
"logps/ref_chosen": -102.64927673339844,
|
|
"logps/ref_rejected": -93.03807067871094,
|
|
"logps/rejected": -277.5597229003906,
|
|
"loss": 0.9816,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.2582169771194458,
|
|
"rewards/margins": 0.5959169864654541,
|
|
"rewards/rejected": -1.8541338443756104,
|
|
"step": 443
|
|
},
|
|
{
|
|
"epoch": 0.6519823788546255,
|
|
"grad_norm": 12.04550552368164,
|
|
"kl/avg_steps": 0.4375,
|
|
"kl/beta": 0.010061916895210743,
|
|
"kl/n_epsilon_steps": 0.28125,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 1.6449496416858282e-07,
|
|
"logits/chosen": -7.471601963043213,
|
|
"logits/rejected": -7.507413387298584,
|
|
"logps/chosen": -221.87020874023438,
|
|
"logps/ref_chosen": -87.91971588134766,
|
|
"logps/ref_rejected": -103.32345581054688,
|
|
"logps/rejected": -295.9997253417969,
|
|
"loss": 1.0177,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -1.34379243850708,
|
|
"rewards/margins": 0.5834265947341919,
|
|
"rewards/rejected": -1.9272189140319824,
|
|
"step": 444
|
|
},
|
|
{
|
|
"epoch": 0.6534508076358296,
|
|
"grad_norm": 10.32401180267334,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.010018087923526764,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 1.6329015999011182e-07,
|
|
"logits/chosen": -7.7422380447387695,
|
|
"logits/rejected": -7.081811904907227,
|
|
"logps/chosen": -216.97213745117188,
|
|
"logps/ref_chosen": -101.40087127685547,
|
|
"logps/ref_rejected": -99.03790283203125,
|
|
"logps/rejected": -282.439208984375,
|
|
"loss": 0.925,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.1529053449630737,
|
|
"rewards/margins": 0.6718940138816833,
|
|
"rewards/rejected": -1.8247992992401123,
|
|
"step": 445
|
|
},
|
|
{
|
|
"epoch": 0.6549192364170338,
|
|
"grad_norm": 12.130196571350098,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.009958956390619278,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 1.6208764069656578e-07,
|
|
"logits/chosen": -7.809151649475098,
|
|
"logits/rejected": -7.551025390625,
|
|
"logps/chosen": -199.480224609375,
|
|
"logps/ref_chosen": -87.42234802246094,
|
|
"logps/ref_rejected": -106.70075988769531,
|
|
"logps/rejected": -281.7625732421875,
|
|
"loss": 0.9615,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -1.1098341941833496,
|
|
"rewards/margins": 0.6203456521034241,
|
|
"rewards/rejected": -1.7301799058914185,
|
|
"step": 446
|
|
},
|
|
{
|
|
"epoch": 0.6563876651982379,
|
|
"grad_norm": 11.679533958435059,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.009903251193463802,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 1.608874379754465e-07,
|
|
"logits/chosen": -7.548259735107422,
|
|
"logits/rejected": -7.2763800621032715,
|
|
"logps/chosen": -197.861083984375,
|
|
"logps/ref_chosen": -83.6152572631836,
|
|
"logps/ref_rejected": -104.91239929199219,
|
|
"logps/rejected": -294.75396728515625,
|
|
"loss": 0.9079,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.1265549659729004,
|
|
"rewards/margins": 0.7398759126663208,
|
|
"rewards/rejected": -1.8664308786392212,
|
|
"step": 447
|
|
},
|
|
{
|
|
"epoch": 0.657856093979442,
|
|
"grad_norm": 11.890493392944336,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.00983868446201086,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 1.5968958345321177e-07,
|
|
"logits/chosen": -7.886592864990234,
|
|
"logits/rejected": -7.5247802734375,
|
|
"logps/chosen": -214.25633239746094,
|
|
"logps/ref_chosen": -92.5757827758789,
|
|
"logps/ref_rejected": -107.68977355957031,
|
|
"logps/rejected": -305.142578125,
|
|
"loss": 0.9455,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -1.1929302215576172,
|
|
"rewards/margins": 0.7362387180328369,
|
|
"rewards/rejected": -1.929168939590454,
|
|
"step": 448
|
|
},
|
|
{
|
|
"epoch": 0.6593245227606461,
|
|
"grad_norm": 10.460795402526855,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.009783651679754257,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 1.584941086944423e-07,
|
|
"logits/chosen": -7.6187238693237305,
|
|
"logits/rejected": -7.113336086273193,
|
|
"logps/chosen": -223.71267700195312,
|
|
"logps/ref_chosen": -102.39893341064453,
|
|
"logps/ref_rejected": -95.14886474609375,
|
|
"logps/rejected": -293.65252685546875,
|
|
"loss": 0.9078,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -1.1828500032424927,
|
|
"rewards/margins": 0.7462369799613953,
|
|
"rewards/rejected": -1.9290869235992432,
|
|
"step": 449
|
|
},
|
|
{
|
|
"epoch": 0.6607929515418502,
|
|
"grad_norm": 11.509246826171875,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.009731950238347054,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 1.573010452010098e-07,
|
|
"logits/chosen": -8.000127792358398,
|
|
"logits/rejected": -7.75508451461792,
|
|
"logps/chosen": -191.9837646484375,
|
|
"logps/ref_chosen": -86.99285888671875,
|
|
"logps/ref_rejected": -108.53203582763672,
|
|
"logps/rejected": -290.04888916015625,
|
|
"loss": 0.9059,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -1.0168712139129639,
|
|
"rewards/margins": 0.7357749938964844,
|
|
"rewards/rejected": -1.7526460886001587,
|
|
"step": 450
|
|
},
|
|
{
|
|
"epoch": 0.6622613803230544,
|
|
"grad_norm": 16.418062210083008,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.009662501513957977,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 1.5611042441124687e-07,
|
|
"logits/chosen": -7.632309913635254,
|
|
"logits/rejected": -7.299648761749268,
|
|
"logps/chosen": -206.88548278808594,
|
|
"logps/ref_chosen": -86.81128692626953,
|
|
"logps/ref_rejected": -79.8555908203125,
|
|
"logps/rejected": -262.9779357910156,
|
|
"loss": 1.0328,
|
|
"rewards/accuracies": 0.734375,
|
|
"rewards/chosen": -1.157651424407959,
|
|
"rewards/margins": 0.6012389659881592,
|
|
"rewards/rejected": -1.7588902711868286,
|
|
"step": 451
|
|
},
|
|
{
|
|
"epoch": 0.6637298091042585,
|
|
"grad_norm": 10.160775184631348,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.00961741991341114,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 1.549222776991186e-07,
|
|
"logits/chosen": -7.714634895324707,
|
|
"logits/rejected": -7.272608757019043,
|
|
"logps/chosen": -185.05380249023438,
|
|
"logps/ref_chosen": -79.379638671875,
|
|
"logps/ref_rejected": -103.71539306640625,
|
|
"logps/rejected": -277.22369384765625,
|
|
"loss": 0.9212,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.0104700326919556,
|
|
"rewards/margins": 0.6453733444213867,
|
|
"rewards/rejected": -1.6558433771133423,
|
|
"step": 452
|
|
},
|
|
{
|
|
"epoch": 0.6651982378854625,
|
|
"grad_norm": 10.804041862487793,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.009551751427352428,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 1.5373663637339584e-07,
|
|
"logits/chosen": -7.6690826416015625,
|
|
"logits/rejected": -7.041682243347168,
|
|
"logps/chosen": -213.16781616210938,
|
|
"logps/ref_chosen": -87.6951904296875,
|
|
"logps/ref_rejected": -90.0582275390625,
|
|
"logps/rejected": -297.2564697265625,
|
|
"loss": 0.8581,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -1.191428303718567,
|
|
"rewards/margins": 0.7726534605026245,
|
|
"rewards/rejected": -1.964081883430481,
|
|
"step": 453
|
|
},
|
|
{
|
|
"epoch": 0.6666666666666666,
|
|
"grad_norm": 11.586421966552734,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.009486531838774681,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 1.5255353167683017e-07,
|
|
"logits/chosen": -7.6572160720825195,
|
|
"logits/rejected": -7.268322944641113,
|
|
"logps/chosen": -224.0977325439453,
|
|
"logps/ref_chosen": -89.56623840332031,
|
|
"logps/ref_rejected": -92.92105102539062,
|
|
"logps/rejected": -293.582275390625,
|
|
"loss": 1.0254,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -1.273888349533081,
|
|
"rewards/margins": 0.6185543537139893,
|
|
"rewards/rejected": -1.8924428224563599,
|
|
"step": 454
|
|
},
|
|
{
|
|
"epoch": 0.6681350954478708,
|
|
"grad_norm": 10.893000602722168,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.00944227073341608,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 1.5137299478533064e-07,
|
|
"logits/chosen": -7.797979354858398,
|
|
"logits/rejected": -7.566769599914551,
|
|
"logps/chosen": -199.83328247070312,
|
|
"logps/ref_chosen": -77.6299819946289,
|
|
"logps/ref_rejected": -118.97795104980469,
|
|
"logps/rejected": -308.0276184082031,
|
|
"loss": 0.9778,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -1.15059232711792,
|
|
"rewards/margins": 0.6233835220336914,
|
|
"rewards/rejected": -1.7739757299423218,
|
|
"step": 455
|
|
},
|
|
{
|
|
"epoch": 0.6696035242290749,
|
|
"grad_norm": 12.390460014343262,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.009395293891429901,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 1.5019505680714232e-07,
|
|
"logits/chosen": -7.607115745544434,
|
|
"logits/rejected": -7.611271858215332,
|
|
"logps/chosen": -215.07440185546875,
|
|
"logps/ref_chosen": -89.61686706542969,
|
|
"logps/ref_rejected": -109.5597152709961,
|
|
"logps/rejected": -298.5582580566406,
|
|
"loss": 0.9811,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -1.1745624542236328,
|
|
"rewards/margins": 0.5899935960769653,
|
|
"rewards/rejected": -1.7645561695098877,
|
|
"step": 456
|
|
},
|
|
{
|
|
"epoch": 0.671071953010279,
|
|
"grad_norm": 11.532391548156738,
|
|
"kl/avg_steps": 0.8125,
|
|
"kl/beta": 0.009342741221189499,
|
|
"kl/n_epsilon_steps": 0.09375,
|
|
"kl/p_epsilon_steps": 0.90625,
|
|
"learning_rate": 1.4901974878202627e-07,
|
|
"logits/chosen": -7.829406261444092,
|
|
"logits/rejected": -7.343839168548584,
|
|
"logps/chosen": -193.81298828125,
|
|
"logps/ref_chosen": -87.32168579101562,
|
|
"logps/ref_rejected": -90.76660919189453,
|
|
"logps/rejected": -280.0379638671875,
|
|
"loss": 0.842,
|
|
"rewards/accuracies": 0.921875,
|
|
"rewards/chosen": -0.9881317615509033,
|
|
"rewards/margins": 0.765188992023468,
|
|
"rewards/rejected": -1.7533208131790161,
|
|
"step": 457
|
|
},
|
|
{
|
|
"epoch": 0.6725403817914831,
|
|
"grad_norm": 11.981277465820312,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.009267443791031837,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 1.4784710168044212e-07,
|
|
"logits/chosen": -7.395938873291016,
|
|
"logits/rejected": -7.573344707489014,
|
|
"logps/chosen": -213.52423095703125,
|
|
"logps/ref_chosen": -93.52044677734375,
|
|
"logps/ref_rejected": -103.36898803710938,
|
|
"logps/rejected": -286.9295654296875,
|
|
"loss": 1.024,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -1.1065789461135864,
|
|
"rewards/margins": 0.581390380859375,
|
|
"rewards/rejected": -1.687969446182251,
|
|
"step": 458
|
|
},
|
|
{
|
|
"epoch": 0.6740088105726872,
|
|
"grad_norm": 9.090399742126465,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.009212742559611797,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 1.466771464027316e-07,
|
|
"logits/chosen": -7.5142059326171875,
|
|
"logits/rejected": -6.964491844177246,
|
|
"logps/chosen": -190.7766571044922,
|
|
"logps/ref_chosen": -75.68820190429688,
|
|
"logps/ref_rejected": -92.17048645019531,
|
|
"logps/rejected": -284.36572265625,
|
|
"loss": 0.9252,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -1.05517578125,
|
|
"rewards/margins": 0.7023875713348389,
|
|
"rewards/rejected": -1.7575633525848389,
|
|
"step": 459
|
|
},
|
|
{
|
|
"epoch": 0.6754772393538914,
|
|
"grad_norm": 12.827125549316406,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.009152678772807121,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 1.4550991377830423e-07,
|
|
"logits/chosen": -7.7618255615234375,
|
|
"logits/rejected": -7.418609619140625,
|
|
"logps/chosen": -201.06686401367188,
|
|
"logps/ref_chosen": -81.11788940429688,
|
|
"logps/ref_rejected": -110.31238555908203,
|
|
"logps/rejected": -300.1006164550781,
|
|
"loss": 0.9702,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -1.0923173427581787,
|
|
"rewards/margins": 0.6314802765846252,
|
|
"rewards/rejected": -1.7237976789474487,
|
|
"step": 460
|
|
},
|
|
{
|
|
"epoch": 0.6769456681350955,
|
|
"grad_norm": 11.000265121459961,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.009093005210161209,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 1.4434543456482518e-07,
|
|
"logits/chosen": -7.253688335418701,
|
|
"logits/rejected": -6.825348854064941,
|
|
"logps/chosen": -227.41180419921875,
|
|
"logps/ref_chosen": -81.58352661132812,
|
|
"logps/ref_rejected": -93.87710571289062,
|
|
"logps/rejected": -295.16571044921875,
|
|
"loss": 1.0504,
|
|
"rewards/accuracies": 0.734375,
|
|
"rewards/chosen": -1.3213638067245483,
|
|
"rewards/margins": 0.4980071187019348,
|
|
"rewards/rejected": -1.819370985031128,
|
|
"step": 461
|
|
},
|
|
{
|
|
"epoch": 0.6784140969162996,
|
|
"grad_norm": 10.546554565429688,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.00905058067291975,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 1.4318373944740484e-07,
|
|
"logits/chosen": -7.895207405090332,
|
|
"logits/rejected": -7.166084289550781,
|
|
"logps/chosen": -205.47802734375,
|
|
"logps/ref_chosen": -94.19855499267578,
|
|
"logps/ref_rejected": -85.63162994384766,
|
|
"logps/rejected": -272.15692138671875,
|
|
"loss": 0.9154,
|
|
"rewards/accuracies": 0.90625,
|
|
"rewards/chosen": -1.00137460231781,
|
|
"rewards/margins": 0.6736320853233337,
|
|
"rewards/rejected": -1.675006628036499,
|
|
"step": 462
|
|
},
|
|
{
|
|
"epoch": 0.6798825256975036,
|
|
"grad_norm": 12.333318710327148,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.008985994383692741,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 1.4202485903778976e-07,
|
|
"logits/chosen": -7.807669639587402,
|
|
"logits/rejected": -7.118095397949219,
|
|
"logps/chosen": -193.8429718017578,
|
|
"logps/ref_chosen": -85.92474365234375,
|
|
"logps/ref_rejected": -96.90184020996094,
|
|
"logps/rejected": -292.2165222167969,
|
|
"loss": 0.8797,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.9655375480651855,
|
|
"rewards/margins": 0.7759765386581421,
|
|
"rewards/rejected": -1.7415142059326172,
|
|
"step": 463
|
|
},
|
|
{
|
|
"epoch": 0.6813509544787077,
|
|
"grad_norm": 14.331416130065918,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.008927407674491405,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 1.4086882387355658e-07,
|
|
"logits/chosen": -7.688302993774414,
|
|
"logits/rejected": -7.788723945617676,
|
|
"logps/chosen": -212.67556762695312,
|
|
"logps/ref_chosen": -79.68920135498047,
|
|
"logps/ref_rejected": -107.29232025146484,
|
|
"logps/rejected": -310.6097717285156,
|
|
"loss": 0.9686,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.1799356937408447,
|
|
"rewards/margins": 0.6203708648681641,
|
|
"rewards/rejected": -1.8003066778182983,
|
|
"step": 464
|
|
},
|
|
{
|
|
"epoch": 0.6828193832599119,
|
|
"grad_norm": 11.430578231811523,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.008863699622452259,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 1.3971566441730714e-07,
|
|
"logits/chosen": -7.3866472244262695,
|
|
"logits/rejected": -7.172554016113281,
|
|
"logps/chosen": -213.5987548828125,
|
|
"logps/ref_chosen": -91.8602294921875,
|
|
"logps/ref_rejected": -118.71000671386719,
|
|
"logps/rejected": -325.18511962890625,
|
|
"loss": 0.8775,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -1.073095440864563,
|
|
"rewards/margins": 0.7431429624557495,
|
|
"rewards/rejected": -1.8162384033203125,
|
|
"step": 465
|
|
},
|
|
{
|
|
"epoch": 0.684287812041116,
|
|
"grad_norm": 8.964788436889648,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.0088004469871521,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 1.3856541105586545e-07,
|
|
"logits/chosen": -7.547239780426025,
|
|
"logits/rejected": -7.197183609008789,
|
|
"logps/chosen": -195.26803588867188,
|
|
"logps/ref_chosen": -84.70140075683594,
|
|
"logps/ref_rejected": -96.05084228515625,
|
|
"logps/rejected": -294.0559387207031,
|
|
"loss": 0.8759,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -0.9667361974716187,
|
|
"rewards/margins": 0.7610065937042236,
|
|
"rewards/rejected": -1.7277427911758423,
|
|
"step": 466
|
|
},
|
|
{
|
|
"epoch": 0.6857562408223201,
|
|
"grad_norm": 11.361374855041504,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.008737645111978054,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 1.3741809409947729e-07,
|
|
"logits/chosen": -8.130990028381348,
|
|
"logits/rejected": -7.5615644454956055,
|
|
"logps/chosen": -244.66354370117188,
|
|
"logps/ref_chosen": -109.29832458496094,
|
|
"logps/ref_rejected": -108.8436508178711,
|
|
"logps/rejected": -320.74517822265625,
|
|
"loss": 0.9698,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.1772236824035645,
|
|
"rewards/margins": 0.6603525876998901,
|
|
"rewards/rejected": -1.8375762701034546,
|
|
"step": 467
|
|
},
|
|
{
|
|
"epoch": 0.6872246696035242,
|
|
"grad_norm": 11.03893756866455,
|
|
"kl/avg_steps": 0.4375,
|
|
"kl/beta": 0.008688771165907383,
|
|
"kl/n_epsilon_steps": 0.28125,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 1.362737437810114e-07,
|
|
"logits/chosen": -7.599516868591309,
|
|
"logits/rejected": -7.804562091827393,
|
|
"logps/chosen": -210.38668823242188,
|
|
"logps/ref_chosen": -98.32164764404297,
|
|
"logps/ref_rejected": -106.68048095703125,
|
|
"logps/rejected": -282.5191345214844,
|
|
"loss": 1.0436,
|
|
"rewards/accuracies": 0.6875,
|
|
"rewards/chosen": -0.9716976881027222,
|
|
"rewards/margins": 0.5471584796905518,
|
|
"rewards/rejected": -1.518856167793274,
|
|
"step": 468
|
|
},
|
|
{
|
|
"epoch": 0.6886930983847284,
|
|
"grad_norm": 12.710182189941406,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.008650923147797585,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 1.351323902551631e-07,
|
|
"logits/chosen": -7.760175704956055,
|
|
"logits/rejected": -7.189078330993652,
|
|
"logps/chosen": -229.18951416015625,
|
|
"logps/ref_chosen": -96.76420593261719,
|
|
"logps/ref_rejected": -109.59500885009766,
|
|
"logps/rejected": -324.5375061035156,
|
|
"loss": 0.9145,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.1398154497146606,
|
|
"rewards/margins": 0.7056005597114563,
|
|
"rewards/rejected": -1.8454160690307617,
|
|
"step": 469
|
|
},
|
|
{
|
|
"epoch": 0.6901615271659325,
|
|
"grad_norm": 11.018048286437988,
|
|
"kl/avg_steps": 0.78125,
|
|
"kl/beta": 0.008591854013502598,
|
|
"kl/n_epsilon_steps": 0.109375,
|
|
"kl/p_epsilon_steps": 0.890625,
|
|
"learning_rate": 1.339940635976592e-07,
|
|
"logits/chosen": -7.906304359436035,
|
|
"logits/rejected": -7.4226555824279785,
|
|
"logps/chosen": -208.847900390625,
|
|
"logps/ref_chosen": -83.49665832519531,
|
|
"logps/ref_rejected": -88.48578643798828,
|
|
"logps/rejected": -289.9870300292969,
|
|
"loss": 0.9357,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -1.0692801475524902,
|
|
"rewards/margins": 0.6468397378921509,
|
|
"rewards/rejected": -1.7161200046539307,
|
|
"step": 470
|
|
},
|
|
{
|
|
"epoch": 0.6916299559471366,
|
|
"grad_norm": 10.446932792663574,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.008525250479578972,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 1.3285879380446563e-07,
|
|
"logits/chosen": -7.087899208068848,
|
|
"logits/rejected": -6.836267948150635,
|
|
"logps/chosen": -233.13998413085938,
|
|
"logps/ref_chosen": -88.47430419921875,
|
|
"logps/ref_rejected": -90.48171997070312,
|
|
"logps/rejected": -307.9644470214844,
|
|
"loss": 0.9497,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -1.2270324230194092,
|
|
"rewards/margins": 0.6135708093643188,
|
|
"rewards/rejected": -1.8406031131744385,
|
|
"step": 471
|
|
},
|
|
{
|
|
"epoch": 0.6930983847283406,
|
|
"grad_norm": 9.13286304473877,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.008467039093375206,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 1.317266107909975e-07,
|
|
"logits/chosen": -7.951723098754883,
|
|
"logits/rejected": -7.220186233520508,
|
|
"logps/chosen": -254.59384155273438,
|
|
"logps/ref_chosen": -125.23369598388672,
|
|
"logps/ref_rejected": -121.05349731445312,
|
|
"logps/rejected": -340.85113525390625,
|
|
"loss": 0.8455,
|
|
"rewards/accuracies": 0.890625,
|
|
"rewards/chosen": -1.0882880687713623,
|
|
"rewards/margins": 0.757997989654541,
|
|
"rewards/rejected": -1.8462860584259033,
|
|
"step": 472
|
|
},
|
|
{
|
|
"epoch": 0.6945668135095447,
|
|
"grad_norm": 11.645308494567871,
|
|
"kl/avg_steps": 0.40625,
|
|
"kl/beta": 0.008406616747379303,
|
|
"kl/n_epsilon_steps": 0.296875,
|
|
"kl/p_epsilon_steps": 0.703125,
|
|
"learning_rate": 1.3059754439133002e-07,
|
|
"logits/chosen": -7.586986064910889,
|
|
"logits/rejected": -7.300940990447998,
|
|
"logps/chosen": -243.3128662109375,
|
|
"logps/ref_chosen": -95.61137390136719,
|
|
"logps/ref_rejected": -88.15115356445312,
|
|
"logps/rejected": -292.8554992675781,
|
|
"loss": 1.0855,
|
|
"rewards/accuracies": 0.71875,
|
|
"rewards/chosen": -1.2393798828125,
|
|
"rewards/margins": 0.4723225235939026,
|
|
"rewards/rejected": -1.711702585220337,
|
|
"step": 473
|
|
},
|
|
{
|
|
"epoch": 0.6960352422907489,
|
|
"grad_norm": 12.829643249511719,
|
|
"kl/avg_steps": 0.4375,
|
|
"kl/beta": 0.008372602984309196,
|
|
"kl/n_epsilon_steps": 0.28125,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 1.2947162435741277e-07,
|
|
"logits/chosen": -7.622153282165527,
|
|
"logits/rejected": -7.345100402832031,
|
|
"logps/chosen": -225.4574737548828,
|
|
"logps/ref_chosen": -81.47975158691406,
|
|
"logps/ref_rejected": -96.46562957763672,
|
|
"logps/rejected": -293.2532958984375,
|
|
"loss": 1.1116,
|
|
"rewards/accuracies": 0.734375,
|
|
"rewards/chosen": -1.2031913995742798,
|
|
"rewards/margins": 0.43547606468200684,
|
|
"rewards/rejected": -1.6386675834655762,
|
|
"step": 474
|
|
},
|
|
{
|
|
"epoch": 0.697503671071953,
|
|
"grad_norm": 9.961170196533203,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.008336132392287254,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 1.2834888035828596e-07,
|
|
"logits/chosen": -7.716796875,
|
|
"logits/rejected": -7.385658264160156,
|
|
"logps/chosen": -182.11282348632812,
|
|
"logps/ref_chosen": -74.19598388671875,
|
|
"logps/ref_rejected": -94.69242095947266,
|
|
"logps/rejected": -285.2503662109375,
|
|
"loss": 0.9232,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.8945390582084656,
|
|
"rewards/margins": 0.6813703775405884,
|
|
"rewards/rejected": -1.5759094953536987,
|
|
"step": 475
|
|
},
|
|
{
|
|
"epoch": 0.6989720998531571,
|
|
"grad_norm": 12.75778579711914,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.008284355513751507,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 1.2722934197929802e-07,
|
|
"logits/chosen": -7.788543224334717,
|
|
"logits/rejected": -7.032910346984863,
|
|
"logps/chosen": -201.364013671875,
|
|
"logps/ref_chosen": -71.97109985351562,
|
|
"logps/ref_rejected": -80.26224517822266,
|
|
"logps/rejected": -281.6759033203125,
|
|
"loss": 0.9888,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -1.067098617553711,
|
|
"rewards/margins": 0.5897184610366821,
|
|
"rewards/rejected": -1.656817078590393,
|
|
"step": 476
|
|
},
|
|
{
|
|
"epoch": 0.7004405286343612,
|
|
"grad_norm": 10.86681079864502,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.008235457353293896,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 1.2611303872132631e-07,
|
|
"logits/chosen": -7.685525894165039,
|
|
"logits/rejected": -7.131807327270508,
|
|
"logps/chosen": -231.5677032470703,
|
|
"logps/ref_chosen": -105.00555419921875,
|
|
"logps/ref_rejected": -81.87843322753906,
|
|
"logps/rejected": -285.14166259765625,
|
|
"loss": 0.9673,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -1.0373156070709229,
|
|
"rewards/margins": 0.6240187883377075,
|
|
"rewards/rejected": -1.6613343954086304,
|
|
"step": 477
|
|
},
|
|
{
|
|
"epoch": 0.7019089574155654,
|
|
"grad_norm": 9.899713516235352,
|
|
"kl/avg_steps": 0.75,
|
|
"kl/beta": 0.008179225027561188,
|
|
"kl/n_epsilon_steps": 0.125,
|
|
"kl/p_epsilon_steps": 0.875,
|
|
"learning_rate": 1.2500000000000005e-07,
|
|
"logits/chosen": -7.471836090087891,
|
|
"logits/rejected": -7.25750732421875,
|
|
"logps/chosen": -198.8499755859375,
|
|
"logps/ref_chosen": -76.7882080078125,
|
|
"logps/ref_rejected": -90.43994140625,
|
|
"logps/rejected": -291.9015808105469,
|
|
"loss": 0.9417,
|
|
"rewards/accuracies": 0.90625,
|
|
"rewards/chosen": -0.9919052124023438,
|
|
"rewards/margins": 0.6421196460723877,
|
|
"rewards/rejected": -1.6340248584747314,
|
|
"step": 478
|
|
},
|
|
{
|
|
"epoch": 0.7033773861967695,
|
|
"grad_norm": 12.096019744873047,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.008118337951600552,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 1.2389025514492456e-07,
|
|
"logits/chosen": -7.843846321105957,
|
|
"logits/rejected": -7.429350852966309,
|
|
"logps/chosen": -211.1195526123047,
|
|
"logps/ref_chosen": -81.3623046875,
|
|
"logps/ref_rejected": -101.09114074707031,
|
|
"logps/rejected": -299.9999694824219,
|
|
"loss": 0.9958,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -1.0483417510986328,
|
|
"rewards/margins": 0.5544079542160034,
|
|
"rewards/rejected": -1.6027498245239258,
|
|
"step": 479
|
|
},
|
|
{
|
|
"epoch": 0.7048458149779736,
|
|
"grad_norm": 12.593810081481934,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.008067913353443146,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 1.227838333989088e-07,
|
|
"logits/chosen": -8.081748962402344,
|
|
"logits/rejected": -7.16196346282959,
|
|
"logps/chosen": -240.56443786621094,
|
|
"logps/ref_chosen": -96.7739028930664,
|
|
"logps/ref_rejected": -86.40473937988281,
|
|
"logps/rejected": -296.2643127441406,
|
|
"loss": 1.0276,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -1.1555697917938232,
|
|
"rewards/margins": 0.5263204574584961,
|
|
"rewards/rejected": -1.6818903684616089,
|
|
"step": 480
|
|
},
|
|
{
|
|
"epoch": 0.7063142437591777,
|
|
"grad_norm": 11.46150016784668,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.008027774281799793,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 1.2168076391719489e-07,
|
|
"logits/chosen": -7.622033596038818,
|
|
"logits/rejected": -6.947890281677246,
|
|
"logps/chosen": -233.6854705810547,
|
|
"logps/ref_chosen": -91.670166015625,
|
|
"logps/ref_rejected": -98.69490051269531,
|
|
"logps/rejected": -314.57574462890625,
|
|
"loss": 0.99,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -1.1354894638061523,
|
|
"rewards/margins": 0.5857840180397034,
|
|
"rewards/rejected": -1.721273422241211,
|
|
"step": 481
|
|
},
|
|
{
|
|
"epoch": 0.7077826725403817,
|
|
"grad_norm": 11.263471603393555,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.007980390451848507,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 1.2058107576668938e-07,
|
|
"logits/chosen": -7.860917091369629,
|
|
"logits/rejected": -7.225180625915527,
|
|
"logps/chosen": -235.95123291015625,
|
|
"logps/ref_chosen": -98.52011108398438,
|
|
"logps/ref_rejected": -94.8294448852539,
|
|
"logps/rejected": -304.33599853515625,
|
|
"loss": 0.9871,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -1.0913074016571045,
|
|
"rewards/margins": 0.5683927536010742,
|
|
"rewards/rejected": -1.6596999168395996,
|
|
"step": 482
|
|
},
|
|
{
|
|
"epoch": 0.7092511013215859,
|
|
"grad_norm": 9.014310836791992,
|
|
"kl/avg_steps": 0.75,
|
|
"kl/beta": 0.00793328694999218,
|
|
"kl/n_epsilon_steps": 0.125,
|
|
"kl/p_epsilon_steps": 0.875,
|
|
"learning_rate": 1.194847979251979e-07,
|
|
"logits/chosen": -8.008612632751465,
|
|
"logits/rejected": -7.4816975593566895,
|
|
"logps/chosen": -248.36256408691406,
|
|
"logps/ref_chosen": -107.11860656738281,
|
|
"logps/ref_rejected": -101.11499786376953,
|
|
"logps/rejected": -319.85467529296875,
|
|
"loss": 0.9526,
|
|
"rewards/accuracies": 0.90625,
|
|
"rewards/chosen": -1.1138007640838623,
|
|
"rewards/margins": 0.6078551411628723,
|
|
"rewards/rejected": -1.7216558456420898,
|
|
"step": 483
|
|
},
|
|
{
|
|
"epoch": 0.71071953010279,
|
|
"grad_norm": 8.998906135559082,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.007874229922890663,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 1.1839195928066101e-07,
|
|
"logits/chosen": -7.730357646942139,
|
|
"logits/rejected": -6.921565055847168,
|
|
"logps/chosen": -220.87344360351562,
|
|
"logps/ref_chosen": -86.97991943359375,
|
|
"logps/ref_rejected": -90.72367095947266,
|
|
"logps/rejected": -307.90826416015625,
|
|
"loss": 0.9462,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -1.049012303352356,
|
|
"rewards/margins": 0.6481724381446838,
|
|
"rewards/rejected": -1.6971848011016846,
|
|
"step": 484
|
|
},
|
|
{
|
|
"epoch": 0.7121879588839941,
|
|
"grad_norm": 9.495361328125,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.007820464670658112,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 1.1730258863039347e-07,
|
|
"logits/chosen": -7.978307247161865,
|
|
"logits/rejected": -7.098984241485596,
|
|
"logps/chosen": -215.27700805664062,
|
|
"logps/ref_chosen": -94.05874633789062,
|
|
"logps/ref_rejected": -108.56297302246094,
|
|
"logps/rejected": -324.5331726074219,
|
|
"loss": 0.8722,
|
|
"rewards/accuracies": 0.90625,
|
|
"rewards/chosen": -0.9427545070648193,
|
|
"rewards/margins": 0.733292281627655,
|
|
"rewards/rejected": -1.6760468482971191,
|
|
"step": 485
|
|
},
|
|
{
|
|
"epoch": 0.7136563876651982,
|
|
"grad_norm": 11.05057144165039,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.007767065893858671,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 1.1621671468032493e-07,
|
|
"logits/chosen": -7.851730823516846,
|
|
"logits/rejected": -7.302485466003418,
|
|
"logps/chosen": -233.9942626953125,
|
|
"logps/ref_chosen": -93.74588012695312,
|
|
"logps/ref_rejected": -98.07064819335938,
|
|
"logps/rejected": -322.165771484375,
|
|
"loss": 0.9519,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -1.0858381986618042,
|
|
"rewards/margins": 0.643623411655426,
|
|
"rewards/rejected": -1.7294615507125854,
|
|
"step": 486
|
|
},
|
|
{
|
|
"epoch": 0.7151248164464024,
|
|
"grad_norm": 10.726838111877441,
|
|
"kl/avg_steps": 0.78125,
|
|
"kl/beta": 0.0077260215766727924,
|
|
"kl/n_epsilon_steps": 0.109375,
|
|
"kl/p_epsilon_steps": 0.890625,
|
|
"learning_rate": 1.1513436604424378e-07,
|
|
"logits/chosen": -7.636082649230957,
|
|
"logits/rejected": -7.332340240478516,
|
|
"logps/chosen": -224.526123046875,
|
|
"logps/ref_chosen": -88.0335693359375,
|
|
"logps/ref_rejected": -98.47209930419922,
|
|
"logps/rejected": -311.92681884765625,
|
|
"loss": 0.9663,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -1.047347068786621,
|
|
"rewards/margins": 0.5880845785140991,
|
|
"rewards/rejected": -1.6354316473007202,
|
|
"step": 487
|
|
},
|
|
{
|
|
"epoch": 0.7165932452276065,
|
|
"grad_norm": 11.892740249633789,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.007666129618883133,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 1.1405557124304335e-07,
|
|
"logits/chosen": -7.434564590454102,
|
|
"logits/rejected": -6.9452362060546875,
|
|
"logps/chosen": -213.1976776123047,
|
|
"logps/ref_chosen": -84.78964233398438,
|
|
"logps/ref_rejected": -90.2734603881836,
|
|
"logps/rejected": -299.0603942871094,
|
|
"loss": 0.935,
|
|
"rewards/accuracies": 0.890625,
|
|
"rewards/chosen": -0.9786466360092163,
|
|
"rewards/margins": 0.6095430850982666,
|
|
"rewards/rejected": -1.5881898403167725,
|
|
"step": 488
|
|
},
|
|
{
|
|
"epoch": 0.7180616740088106,
|
|
"grad_norm": 9.741364479064941,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.007616148795932531,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 1.1298035870396985e-07,
|
|
"logits/chosen": -7.642593860626221,
|
|
"logits/rejected": -7.163801670074463,
|
|
"logps/chosen": -229.92735290527344,
|
|
"logps/ref_chosen": -90.46929931640625,
|
|
"logps/ref_rejected": -86.39761352539062,
|
|
"logps/rejected": -301.63934326171875,
|
|
"loss": 0.982,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -1.0582023859024048,
|
|
"rewards/margins": 0.5706866979598999,
|
|
"rewards/rejected": -1.6288890838623047,
|
|
"step": 489
|
|
},
|
|
{
|
|
"epoch": 0.7195301027900147,
|
|
"grad_norm": 10.637678146362305,
|
|
"kl/avg_steps": 0.453125,
|
|
"kl/beta": 0.0075759016908705235,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 1.1190875675987355e-07,
|
|
"logits/chosen": -7.2537336349487305,
|
|
"logits/rejected": -7.196831226348877,
|
|
"logps/chosen": -230.93482971191406,
|
|
"logps/ref_chosen": -85.32012939453125,
|
|
"logps/ref_rejected": -115.99385070800781,
|
|
"logps/rejected": -327.2318115234375,
|
|
"loss": 1.094,
|
|
"rewards/accuracies": 0.734375,
|
|
"rewards/chosen": -1.1003369092941284,
|
|
"rewards/margins": 0.48968762159347534,
|
|
"rewards/rejected": -1.590024471282959,
|
|
"step": 490
|
|
},
|
|
{
|
|
"epoch": 0.7209985315712188,
|
|
"grad_norm": 9.847221374511719,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.00754172820597887,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 1.1084079364846241e-07,
|
|
"logits/chosen": -7.813044548034668,
|
|
"logits/rejected": -7.093764305114746,
|
|
"logps/chosen": -229.64122009277344,
|
|
"logps/ref_chosen": -86.14351654052734,
|
|
"logps/ref_rejected": -80.67945861816406,
|
|
"logps/rejected": -291.98773193359375,
|
|
"loss": 1.0378,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -1.0794801712036133,
|
|
"rewards/margins": 0.5048484802246094,
|
|
"rewards/rejected": -1.5843286514282227,
|
|
"step": 491
|
|
},
|
|
{
|
|
"epoch": 0.7224669603524229,
|
|
"grad_norm": 13.724920272827148,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.007504207547754049,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 1.097764975115576e-07,
|
|
"logits/chosen": -7.806938171386719,
|
|
"logits/rejected": -7.273170471191406,
|
|
"logps/chosen": -213.5839080810547,
|
|
"logps/ref_chosen": -81.10757446289062,
|
|
"logps/ref_rejected": -80.75199890136719,
|
|
"logps/rejected": -268.00067138671875,
|
|
"loss": 1.151,
|
|
"rewards/accuracies": 0.71875,
|
|
"rewards/chosen": -0.9921280145645142,
|
|
"rewards/margins": 0.40435731410980225,
|
|
"rewards/rejected": -1.3964853286743164,
|
|
"step": 492
|
|
},
|
|
{
|
|
"epoch": 0.723935389133627,
|
|
"grad_norm": 11.94544506072998,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.007469195406883955,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 1.0871589639435203e-07,
|
|
"logits/chosen": -7.833661079406738,
|
|
"logits/rejected": -6.987953186035156,
|
|
"logps/chosen": -245.13128662109375,
|
|
"logps/ref_chosen": -112.20733642578125,
|
|
"logps/ref_rejected": -93.60719299316406,
|
|
"logps/rejected": -293.9502868652344,
|
|
"loss": 1.0192,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.9897406697273254,
|
|
"rewards/margins": 0.49745315313339233,
|
|
"rewards/rejected": -1.4871938228607178,
|
|
"step": 493
|
|
},
|
|
{
|
|
"epoch": 0.7254038179148311,
|
|
"grad_norm": 10.316105842590332,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.007427416276186705,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 1.0765901824467166e-07,
|
|
"logits/chosen": -7.544450759887695,
|
|
"logits/rejected": -7.313390731811523,
|
|
"logps/chosen": -200.50823974609375,
|
|
"logps/ref_chosen": -73.11489868164062,
|
|
"logps/ref_rejected": -92.16300201416016,
|
|
"logps/rejected": -312.24517822265625,
|
|
"loss": 0.9054,
|
|
"rewards/accuracies": 0.890625,
|
|
"rewards/chosen": -0.940959632396698,
|
|
"rewards/margins": 0.6807957291603088,
|
|
"rewards/rejected": -1.6217553615570068,
|
|
"step": 494
|
|
},
|
|
{
|
|
"epoch": 0.7268722466960352,
|
|
"grad_norm": 10.488041877746582,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.0073744128458201885,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 1.0660589091223854e-07,
|
|
"logits/chosen": -7.922597408294678,
|
|
"logits/rejected": -7.498479843139648,
|
|
"logps/chosen": -228.8547821044922,
|
|
"logps/ref_chosen": -99.52032470703125,
|
|
"logps/ref_rejected": -97.93089294433594,
|
|
"logps/rejected": -302.8748779296875,
|
|
"loss": 1.0093,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.950130820274353,
|
|
"rewards/margins": 0.5504207611083984,
|
|
"rewards/rejected": -1.5005515813827515,
|
|
"step": 495
|
|
},
|
|
{
|
|
"epoch": 0.7283406754772394,
|
|
"grad_norm": 10.726445198059082,
|
|
"kl/avg_steps": 0.4375,
|
|
"kl/beta": 0.007328609004616737,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.703125,
|
|
"learning_rate": 1.0555654214793722e-07,
|
|
"logits/chosen": -7.784862995147705,
|
|
"logits/rejected": -7.454352378845215,
|
|
"logps/chosen": -250.06234741210938,
|
|
"logps/ref_chosen": -107.85675048828125,
|
|
"logps/ref_rejected": -92.77056121826172,
|
|
"logps/rejected": -301.3846740722656,
|
|
"loss": 1.0384,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -1.039567232131958,
|
|
"rewards/margins": 0.48080918192863464,
|
|
"rewards/rejected": -1.520376443862915,
|
|
"step": 496
|
|
},
|
|
{
|
|
"epoch": 0.7298091042584435,
|
|
"grad_norm": 10.815900802612305,
|
|
"kl/avg_steps": 0.3125,
|
|
"kl/beta": 0.007296686060726643,
|
|
"kl/n_epsilon_steps": 0.34375,
|
|
"kl/p_epsilon_steps": 0.65625,
|
|
"learning_rate": 1.0451099960308374e-07,
|
|
"logits/chosen": -7.584544658660889,
|
|
"logits/rejected": -7.411238670349121,
|
|
"logps/chosen": -249.58370971679688,
|
|
"logps/ref_chosen": -92.08322143554688,
|
|
"logps/ref_rejected": -81.79503631591797,
|
|
"logps/rejected": -286.0672607421875,
|
|
"loss": 1.1655,
|
|
"rewards/accuracies": 0.703125,
|
|
"rewards/chosen": -1.1472631692886353,
|
|
"rewards/margins": 0.33593475818634033,
|
|
"rewards/rejected": -1.483197808265686,
|
|
"step": 497
|
|
},
|
|
{
|
|
"epoch": 0.7312775330396476,
|
|
"grad_norm": 10.606035232543945,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.007273954804986715,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 1.0346929082869641e-07,
|
|
"logits/chosen": -7.672981262207031,
|
|
"logits/rejected": -7.160597801208496,
|
|
"logps/chosen": -226.19503784179688,
|
|
"logps/ref_chosen": -98.19436645507812,
|
|
"logps/ref_rejected": -90.68746185302734,
|
|
"logps/rejected": -293.2069396972656,
|
|
"loss": 0.9923,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -0.9259578585624695,
|
|
"rewards/margins": 0.5358661413192749,
|
|
"rewards/rejected": -1.4618239402770996,
|
|
"step": 498
|
|
},
|
|
{
|
|
"epoch": 0.7327459618208517,
|
|
"grad_norm": 13.12460708618164,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.007224287837743759,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 1.0243144327477013e-07,
|
|
"logits/chosen": -7.342049598693848,
|
|
"logits/rejected": -7.139578342437744,
|
|
"logps/chosen": -219.22235107421875,
|
|
"logps/ref_chosen": -81.0399169921875,
|
|
"logps/ref_rejected": -106.92170715332031,
|
|
"logps/rejected": -319.46099853515625,
|
|
"loss": 1.0261,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.9939165711402893,
|
|
"rewards/margins": 0.5303443074226379,
|
|
"rewards/rejected": -1.5242608785629272,
|
|
"step": 499
|
|
},
|
|
{
|
|
"epoch": 0.7342143906020558,
|
|
"grad_norm": 11.147185325622559,
|
|
"kl/avg_steps": 0.4375,
|
|
"kl/beta": 0.007179416250437498,
|
|
"kl/n_epsilon_steps": 0.28125,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 1.0139748428955333e-07,
|
|
"logits/chosen": -7.811801433563232,
|
|
"logits/rejected": -7.598600387573242,
|
|
"logps/chosen": -237.38864135742188,
|
|
"logps/ref_chosen": -89.248046875,
|
|
"logps/ref_rejected": -100.41021728515625,
|
|
"logps/rejected": -317.2994079589844,
|
|
"loss": 1.0563,
|
|
"rewards/accuracies": 0.734375,
|
|
"rewards/chosen": -1.0610756874084473,
|
|
"rewards/margins": 0.48731058835983276,
|
|
"rewards/rejected": -1.5483863353729248,
|
|
"step": 500
|
|
},
|
|
{
|
|
"epoch": 0.7342143906020558,
|
|
"eval_kl/n_epsilon_steps": 0.25042808055877686,
|
|
"eval_kl/p_epsilon_steps": 0.7491438388824463,
|
|
"eval_logits/chosen": -7.587223529815674,
|
|
"eval_logits/rejected": -7.184981822967529,
|
|
"eval_logps/chosen": -242.73046875,
|
|
"eval_logps/ref_chosen": -100.49356842041016,
|
|
"eval_logps/ref_rejected": -94.06775665283203,
|
|
"eval_logps/rejected": -299.77655029296875,
|
|
"eval_loss": 0.5370330810546875,
|
|
"eval_rewards/accuracies": 0.7666952013969421,
|
|
"eval_rewards/chosen": -1.0134860277175903,
|
|
"eval_rewards/margins": 0.447692334651947,
|
|
"eval_rewards/rejected": -1.4611784219741821,
|
|
"eval_runtime": 48.5908,
|
|
"eval_samples_per_second": 48.137,
|
|
"eval_steps_per_second": 1.523,
|
|
"step": 500
|
|
},
|
|
{
|
|
"epoch": 0.73568281938326,
|
|
"grad_norm": 11.189581871032715,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.007148143369704485,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 1.0036744111882672e-07,
|
|
"logits/chosen": -7.5872368812561035,
|
|
"logits/rejected": -7.235992431640625,
|
|
"logps/chosen": -220.14817810058594,
|
|
"logps/ref_chosen": -90.596923828125,
|
|
"logps/ref_rejected": -85.2972640991211,
|
|
"logps/rejected": -288.4246520996094,
|
|
"loss": 1.0588,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -0.9249030947685242,
|
|
"rewards/margins": 0.5188544392585754,
|
|
"rewards/rejected": -1.4437575340270996,
|
|
"step": 501
|
|
},
|
|
{
|
|
"epoch": 0.737151248164464,
|
|
"grad_norm": 10.501534461975098,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.007114792708307505,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 9.934134090518592e-08,
|
|
"logits/chosen": -8.039237976074219,
|
|
"logits/rejected": -6.9175286293029785,
|
|
"logps/chosen": -235.90652465820312,
|
|
"logps/ref_chosen": -103.47073364257812,
|
|
"logps/ref_rejected": -89.04415893554688,
|
|
"logps/rejected": -287.358642578125,
|
|
"loss": 1.0789,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -0.9396613836288452,
|
|
"rewards/margins": 0.46196457743644714,
|
|
"rewards/rejected": -1.4016259908676147,
|
|
"step": 502
|
|
},
|
|
{
|
|
"epoch": 0.7386196769456681,
|
|
"grad_norm": 9.694924354553223,
|
|
"kl/avg_steps": 0.4375,
|
|
"kl/beta": 0.007079395931214094,
|
|
"kl/n_epsilon_steps": 0.28125,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 9.831921068732571e-08,
|
|
"logits/chosen": -7.841567039489746,
|
|
"logits/rejected": -7.385000228881836,
|
|
"logps/chosen": -222.0538787841797,
|
|
"logps/ref_chosen": -93.72999572753906,
|
|
"logps/ref_rejected": -87.60896301269531,
|
|
"logps/rejected": -291.72247314453125,
|
|
"loss": 1.0182,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -0.9058637619018555,
|
|
"rewards/margins": 0.5300005674362183,
|
|
"rewards/rejected": -1.4358644485473633,
|
|
"step": 503
|
|
},
|
|
{
|
|
"epoch": 0.7400881057268722,
|
|
"grad_norm": 12.8939208984375,
|
|
"kl/avg_steps": 0.75,
|
|
"kl/beta": 0.007048558443784714,
|
|
"kl/n_epsilon_steps": 0.125,
|
|
"kl/p_epsilon_steps": 0.875,
|
|
"learning_rate": 9.730107739932805e-08,
|
|
"logits/chosen": -8.119867324829102,
|
|
"logits/rejected": -7.42840576171875,
|
|
"logps/chosen": -221.08261108398438,
|
|
"logps/ref_chosen": -92.2918701171875,
|
|
"logps/ref_rejected": -109.03765106201172,
|
|
"logps/rejected": -328.2010498046875,
|
|
"loss": 0.9324,
|
|
"rewards/accuracies": 0.890625,
|
|
"rewards/chosen": -0.9024004340171814,
|
|
"rewards/margins": 0.629888117313385,
|
|
"rewards/rejected": -1.5322885513305664,
|
|
"step": 504
|
|
},
|
|
{
|
|
"epoch": 0.7415565345080763,
|
|
"grad_norm": 9.646526336669922,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.006996087729930878,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 9.628696786995188e-08,
|
|
"logits/chosen": -7.906380653381348,
|
|
"logits/rejected": -7.302044868469238,
|
|
"logps/chosen": -250.1861572265625,
|
|
"logps/ref_chosen": -106.27693176269531,
|
|
"logps/ref_rejected": -94.20611572265625,
|
|
"logps/rejected": -310.4810791015625,
|
|
"loss": 1.0176,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -1.0034873485565186,
|
|
"rewards/margins": 0.5002511143684387,
|
|
"rewards/rejected": -1.5037386417388916,
|
|
"step": 505
|
|
},
|
|
{
|
|
"epoch": 0.7430249632892805,
|
|
"grad_norm": 9.696473121643066,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.006959117483347654,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 9.527690882192635e-08,
|
|
"logits/chosen": -7.563765525817871,
|
|
"logits/rejected": -7.289196014404297,
|
|
"logps/chosen": -221.6510009765625,
|
|
"logps/ref_chosen": -81.44086456298828,
|
|
"logps/ref_rejected": -85.66439819335938,
|
|
"logps/rejected": -301.66949462890625,
|
|
"loss": 0.9974,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -0.9702832698822021,
|
|
"rewards/margins": 0.5212363600730896,
|
|
"rewards/rejected": -1.4915196895599365,
|
|
"step": 506
|
|
},
|
|
{
|
|
"epoch": 0.7444933920704846,
|
|
"grad_norm": 8.90442943572998,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.006913745775818825,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 9.427092687124691e-08,
|
|
"logits/chosen": -7.426417827606201,
|
|
"logits/rejected": -7.099169731140137,
|
|
"logps/chosen": -245.88427734375,
|
|
"logps/ref_chosen": -99.42694854736328,
|
|
"logps/ref_rejected": -102.25296020507812,
|
|
"logps/rejected": -328.4168701171875,
|
|
"loss": 1.0131,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -1.0063550472259521,
|
|
"rewards/margins": 0.5441376566886902,
|
|
"rewards/rejected": -1.550492763519287,
|
|
"step": 507
|
|
},
|
|
{
|
|
"epoch": 0.7459618208516887,
|
|
"grad_norm": 10.534308433532715,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.006864408031105995,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 9.326904852647344e-08,
|
|
"logits/chosen": -7.581225395202637,
|
|
"logits/rejected": -7.184242248535156,
|
|
"logps/chosen": -241.60638427734375,
|
|
"logps/ref_chosen": -97.9156494140625,
|
|
"logps/ref_rejected": -103.79782104492188,
|
|
"logps/rejected": -314.3401794433594,
|
|
"loss": 1.0875,
|
|
"rewards/accuracies": 0.734375,
|
|
"rewards/chosen": -0.9846047759056091,
|
|
"rewards/margins": 0.4526156783103943,
|
|
"rewards/rejected": -1.4372204542160034,
|
|
"step": 508
|
|
},
|
|
{
|
|
"epoch": 0.7474302496328928,
|
|
"grad_norm": 12.450897216796875,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.006832381244748831,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 9.227130018803195e-08,
|
|
"logits/chosen": -7.861898422241211,
|
|
"logits/rejected": -7.100181579589844,
|
|
"logps/chosen": -237.10333251953125,
|
|
"logps/ref_chosen": -100.72984313964844,
|
|
"logps/ref_rejected": -86.943359375,
|
|
"logps/rejected": -287.923095703125,
|
|
"loss": 1.0779,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.9274278879165649,
|
|
"rewards/margins": 0.43540412187576294,
|
|
"rewards/rejected": -1.3628320693969727,
|
|
"step": 509
|
|
},
|
|
{
|
|
"epoch": 0.748898678414097,
|
|
"grad_norm": 10.103314399719238,
|
|
"kl/avg_steps": 0.578125,
|
|
"kl/beta": 0.006789944134652615,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 9.127770814751932e-08,
|
|
"logits/chosen": -7.755180358886719,
|
|
"logits/rejected": -7.310698986053467,
|
|
"logps/chosen": -248.5462646484375,
|
|
"logps/ref_chosen": -89.16938781738281,
|
|
"logps/ref_rejected": -107.91940307617188,
|
|
"logps/rejected": -342.783447265625,
|
|
"loss": 1.0245,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -1.0782501697540283,
|
|
"rewards/margins": 0.5061511993408203,
|
|
"rewards/rejected": -1.5844013690948486,
|
|
"step": 510
|
|
},
|
|
{
|
|
"epoch": 0.750367107195301,
|
|
"grad_norm": 9.313636779785156,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.006750915199518204,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 9.028829858700973e-08,
|
|
"logits/chosen": -7.98992919921875,
|
|
"logits/rejected": -7.07249641418457,
|
|
"logps/chosen": -233.65809631347656,
|
|
"logps/ref_chosen": -89.74979400634766,
|
|
"logps/ref_rejected": -97.73127746582031,
|
|
"logps/rejected": -322.42425537109375,
|
|
"loss": 0.998,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.9661585092544556,
|
|
"rewards/margins": 0.539162278175354,
|
|
"rewards/rejected": -1.5053207874298096,
|
|
"step": 511
|
|
},
|
|
{
|
|
"epoch": 0.7518355359765051,
|
|
"grad_norm": 11.483407974243164,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.006704819854348898,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 8.930309757836516e-08,
|
|
"logits/chosen": -7.597411155700684,
|
|
"logits/rejected": -7.131270408630371,
|
|
"logps/chosen": -221.8423614501953,
|
|
"logps/ref_chosen": -85.73515319824219,
|
|
"logps/ref_rejected": -87.6248779296875,
|
|
"logps/rejected": -300.55499267578125,
|
|
"loss": 1.032,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.9088408350944519,
|
|
"rewards/margins": 0.5085346698760986,
|
|
"rewards/rejected": -1.4173755645751953,
|
|
"step": 512
|
|
},
|
|
{
|
|
"epoch": 0.7533039647577092,
|
|
"grad_norm": 10.936783790588379,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.006663174834102392,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 8.832213108254863e-08,
|
|
"logits/chosen": -7.599587917327881,
|
|
"logits/rejected": -7.238034725189209,
|
|
"logps/chosen": -240.4691162109375,
|
|
"logps/ref_chosen": -96.19051361083984,
|
|
"logps/ref_rejected": -82.91773986816406,
|
|
"logps/rejected": -295.57098388671875,
|
|
"loss": 1.0548,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.956098198890686,
|
|
"rewards/margins": 0.44991129636764526,
|
|
"rewards/rejected": -1.4060094356536865,
|
|
"step": 513
|
|
},
|
|
{
|
|
"epoch": 0.7547723935389133,
|
|
"grad_norm": 10.800374031066895,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.006617678329348564,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 8.734542494893954e-08,
|
|
"logits/chosen": -7.524253845214844,
|
|
"logits/rejected": -7.1176605224609375,
|
|
"logps/chosen": -235.91629028320312,
|
|
"logps/ref_chosen": -97.82363891601562,
|
|
"logps/ref_rejected": -85.36982727050781,
|
|
"logps/rejected": -294.8122253417969,
|
|
"loss": 1.0464,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -0.9116629958152771,
|
|
"rewards/margins": 0.46637022495269775,
|
|
"rewards/rejected": -1.37803316116333,
|
|
"step": 514
|
|
},
|
|
{
|
|
"epoch": 0.7562408223201175,
|
|
"grad_norm": 13.166977882385254,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.0065868026576936245,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 8.637300491465272e-08,
|
|
"logits/chosen": -7.724981307983398,
|
|
"logits/rejected": -7.439934730529785,
|
|
"logps/chosen": -219.55661010742188,
|
|
"logps/ref_chosen": -81.93355560302734,
|
|
"logps/ref_rejected": -93.5765609741211,
|
|
"logps/rejected": -294.6715087890625,
|
|
"loss": 1.091,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.9028570652008057,
|
|
"rewards/margins": 0.41230061650276184,
|
|
"rewards/rejected": -1.3151576519012451,
|
|
"step": 515
|
|
},
|
|
{
|
|
"epoch": 0.7577092511013216,
|
|
"grad_norm": 9.387166023254395,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.006547924131155014,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 8.540489660386064e-08,
|
|
"logits/chosen": -7.62106990814209,
|
|
"logits/rejected": -7.103853225708008,
|
|
"logps/chosen": -236.6641845703125,
|
|
"logps/ref_chosen": -97.09166717529297,
|
|
"logps/ref_rejected": -118.55667114257812,
|
|
"logps/rejected": -326.1341552734375,
|
|
"loss": 1.1026,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -0.9115662574768066,
|
|
"rewards/margins": 0.43881756067276,
|
|
"rewards/rejected": -1.3503837585449219,
|
|
"step": 516
|
|
},
|
|
{
|
|
"epoch": 0.7591776798825257,
|
|
"grad_norm": 9.433574676513672,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.006515347398817539,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 8.444112552711752e-08,
|
|
"logits/chosen": -7.664484024047852,
|
|
"logits/rejected": -7.389780521392822,
|
|
"logps/chosen": -241.97000122070312,
|
|
"logps/ref_chosen": -93.18656921386719,
|
|
"logps/ref_rejected": -95.5927963256836,
|
|
"logps/rejected": -327.4776611328125,
|
|
"loss": 1.0038,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.9653780460357666,
|
|
"rewards/margins": 0.534911036491394,
|
|
"rewards/rejected": -1.500288963317871,
|
|
"step": 517
|
|
},
|
|
{
|
|
"epoch": 0.7606461086637298,
|
|
"grad_norm": 11.14731502532959,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.006478903815150261,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 8.348171708068747e-08,
|
|
"logits/chosen": -7.792545318603516,
|
|
"logits/rejected": -7.39441442489624,
|
|
"logps/chosen": -241.240478515625,
|
|
"logps/ref_chosen": -87.0213623046875,
|
|
"logps/ref_rejected": -91.24455261230469,
|
|
"logps/rejected": -307.2711486816406,
|
|
"loss": 1.1116,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.9949862360954285,
|
|
"rewards/margins": 0.394919216632843,
|
|
"rewards/rejected": -1.3899054527282715,
|
|
"step": 518
|
|
},
|
|
{
|
|
"epoch": 0.762114537444934,
|
|
"grad_norm": 9.395498275756836,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.006442663725465536,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 8.25266965458755e-08,
|
|
"logits/chosen": -7.722479343414307,
|
|
"logits/rejected": -7.366518974304199,
|
|
"logps/chosen": -244.36495971679688,
|
|
"logps/ref_chosen": -100.68755340576172,
|
|
"logps/ref_rejected": -109.03042602539062,
|
|
"logps/rejected": -321.7382507324219,
|
|
"loss": 1.0488,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.9199578762054443,
|
|
"rewards/margins": 0.43974611163139343,
|
|
"rewards/rejected": -1.3597040176391602,
|
|
"step": 519
|
|
},
|
|
{
|
|
"epoch": 0.7635829662261381,
|
|
"grad_norm": 9.159811973571777,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.006400659214705229,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 8.15760890883607e-08,
|
|
"logits/chosen": -7.769089221954346,
|
|
"logits/rejected": -7.011660575866699,
|
|
"logps/chosen": -229.7172393798828,
|
|
"logps/ref_chosen": -98.77762603759766,
|
|
"logps/ref_rejected": -106.9457015991211,
|
|
"logps/rejected": -318.0968322753906,
|
|
"loss": 0.9971,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -0.834189772605896,
|
|
"rewards/margins": 0.507774829864502,
|
|
"rewards/rejected": -1.341964602470398,
|
|
"step": 520
|
|
},
|
|
{
|
|
"epoch": 0.7650513950073421,
|
|
"grad_norm": 9.642518043518066,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.006360903847962618,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 8.062991975753378e-08,
|
|
"logits/chosen": -7.651534080505371,
|
|
"logits/rejected": -7.242176055908203,
|
|
"logps/chosen": -224.09951782226562,
|
|
"logps/ref_chosen": -86.35108947753906,
|
|
"logps/ref_rejected": -90.39317321777344,
|
|
"logps/rejected": -306.9713134765625,
|
|
"loss": 1.0088,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.8716791868209839,
|
|
"rewards/margins": 0.4956836700439453,
|
|
"rewards/rejected": -1.3673629760742188,
|
|
"step": 521
|
|
},
|
|
{
|
|
"epoch": 0.7665198237885462,
|
|
"grad_norm": 9.185672760009766,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.006319432519376278,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 7.968821348583643e-08,
|
|
"logits/chosen": -7.3879499435424805,
|
|
"logits/rejected": -6.970926284790039,
|
|
"logps/chosen": -215.34011840820312,
|
|
"logps/ref_chosen": -76.008056640625,
|
|
"logps/ref_rejected": -74.44125366210938,
|
|
"logps/rejected": -289.88787841796875,
|
|
"loss": 1.0429,
|
|
"rewards/accuracies": 0.734375,
|
|
"rewards/chosen": -0.8779767751693726,
|
|
"rewards/margins": 0.47517913579940796,
|
|
"rewards/rejected": -1.3531558513641357,
|
|
"step": 522
|
|
},
|
|
{
|
|
"epoch": 0.7679882525697503,
|
|
"grad_norm": 11.464506149291992,
|
|
"kl/avg_steps": 0.4375,
|
|
"kl/beta": 0.006289948243647814,
|
|
"kl/n_epsilon_steps": 0.28125,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 7.875099508810484e-08,
|
|
"logits/chosen": -7.647948741912842,
|
|
"logits/rejected": -6.6883955001831055,
|
|
"logps/chosen": -255.38108825683594,
|
|
"logps/ref_chosen": -96.49627685546875,
|
|
"logps/ref_rejected": -91.0152816772461,
|
|
"logps/rejected": -302.5433349609375,
|
|
"loss": 1.1708,
|
|
"rewards/accuracies": 0.734375,
|
|
"rewards/chosen": -0.9972683787345886,
|
|
"rewards/margins": 0.3256107568740845,
|
|
"rewards/rejected": -1.3228791952133179,
|
|
"step": 523
|
|
},
|
|
{
|
|
"epoch": 0.7694566813509545,
|
|
"grad_norm": 10.063695907592773,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.006262549664825201,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 7.781828926091535e-08,
|
|
"logits/chosen": -8.02735710144043,
|
|
"logits/rejected": -7.3874831199646,
|
|
"logps/chosen": -248.78433227539062,
|
|
"logps/ref_chosen": -103.40550231933594,
|
|
"logps/ref_rejected": -87.91015625,
|
|
"logps/rejected": -306.3849792480469,
|
|
"loss": 1.0526,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.908073902130127,
|
|
"rewards/margins": 0.45238441228866577,
|
|
"rewards/rejected": -1.3604583740234375,
|
|
"step": 524
|
|
},
|
|
{
|
|
"epoch": 0.7709251101321586,
|
|
"grad_norm": 8.88698673248291,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.006233330816030502,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 7.689012058193384e-08,
|
|
"logits/chosen": -7.746663570404053,
|
|
"logits/rejected": -7.510030746459961,
|
|
"logps/chosen": -218.09078979492188,
|
|
"logps/ref_chosen": -84.20648193359375,
|
|
"logps/ref_rejected": -106.26158905029297,
|
|
"logps/rejected": -323.40283203125,
|
|
"loss": 1.0194,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -0.8309329748153687,
|
|
"rewards/margins": 0.513077974319458,
|
|
"rewards/rejected": -1.344010829925537,
|
|
"step": 525
|
|
},
|
|
{
|
|
"epoch": 0.7723935389133627,
|
|
"grad_norm": 8.236015319824219,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.006198464427143335,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 7.596651350926836e-08,
|
|
"logits/chosen": -7.848876476287842,
|
|
"logits/rejected": -7.135056495666504,
|
|
"logps/chosen": -246.24072265625,
|
|
"logps/ref_chosen": -102.85411071777344,
|
|
"logps/ref_rejected": -93.65145874023438,
|
|
"logps/rejected": -315.64605712890625,
|
|
"loss": 1.0393,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -0.8863099813461304,
|
|
"rewards/margins": 0.4812777638435364,
|
|
"rewards/rejected": -1.3675878047943115,
|
|
"step": 526
|
|
},
|
|
{
|
|
"epoch": 0.7738619676945668,
|
|
"grad_norm": 8.652771949768066,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.006167626474052668,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 7.504749238082414e-08,
|
|
"logits/chosen": -8.085288047790527,
|
|
"logits/rejected": -7.213015556335449,
|
|
"logps/chosen": -236.21170043945312,
|
|
"logps/ref_chosen": -96.76543426513672,
|
|
"logps/ref_rejected": -85.39518737792969,
|
|
"logps/rejected": -304.16583251953125,
|
|
"loss": 1.0093,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -0.8550451993942261,
|
|
"rewards/margins": 0.4838915765285492,
|
|
"rewards/rejected": -1.3389368057250977,
|
|
"step": 527
|
|
},
|
|
{
|
|
"epoch": 0.775330396475771,
|
|
"grad_norm": 8.880172729492188,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.006125513464212418,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 7.413308141366254e-08,
|
|
"logits/chosen": -8.002275466918945,
|
|
"logits/rejected": -7.597675800323486,
|
|
"logps/chosen": -225.98036193847656,
|
|
"logps/ref_chosen": -94.1649169921875,
|
|
"logps/ref_rejected": -100.372314453125,
|
|
"logps/rejected": -312.89617919921875,
|
|
"loss": 1.0225,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.8029056191444397,
|
|
"rewards/margins": 0.48855090141296387,
|
|
"rewards/rejected": -1.2914564609527588,
|
|
"step": 528
|
|
},
|
|
{
|
|
"epoch": 0.7767988252569751,
|
|
"grad_norm": 10.054601669311523,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.006083688233047724,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 7.322330470336313e-08,
|
|
"logits/chosen": -7.928413391113281,
|
|
"logits/rejected": -7.295429229736328,
|
|
"logps/chosen": -234.24697875976562,
|
|
"logps/ref_chosen": -80.77679443359375,
|
|
"logps/ref_rejected": -97.191162109375,
|
|
"logps/rejected": -321.6305847167969,
|
|
"loss": 1.0863,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -0.9308747053146362,
|
|
"rewards/margins": 0.42603152990341187,
|
|
"rewards/rejected": -1.3569061756134033,
|
|
"step": 529
|
|
},
|
|
{
|
|
"epoch": 0.7782672540381792,
|
|
"grad_norm": 8.395234107971191,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.006053421180695295,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 7.231818622338822e-08,
|
|
"logits/chosen": -7.8860321044921875,
|
|
"logits/rejected": -7.2712297439575195,
|
|
"logps/chosen": -221.32212829589844,
|
|
"logps/ref_chosen": -79.7757339477539,
|
|
"logps/ref_rejected": -92.05390167236328,
|
|
"logps/rejected": -306.6104736328125,
|
|
"loss": 1.0424,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.852878212928772,
|
|
"rewards/margins": 0.43727239966392517,
|
|
"rewards/rejected": -1.2901506423950195,
|
|
"step": 530
|
|
},
|
|
{
|
|
"epoch": 0.7797356828193832,
|
|
"grad_norm": 8.446939468383789,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.006013954523950815,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 7.141774982445147e-08,
|
|
"logits/chosen": -7.656224250793457,
|
|
"logits/rejected": -6.923822402954102,
|
|
"logps/chosen": -217.52377319335938,
|
|
"logps/ref_chosen": -79.31919860839844,
|
|
"logps/ref_rejected": -78.10832214355469,
|
|
"logps/rejected": -304.80413818359375,
|
|
"loss": 0.9847,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -0.8268610239028931,
|
|
"rewards/margins": 0.5261213183403015,
|
|
"rewards/rejected": -1.3529822826385498,
|
|
"step": 531
|
|
},
|
|
{
|
|
"epoch": 0.7812041116005873,
|
|
"grad_norm": 9.200221061706543,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.005972891114652157,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 7.052201923388953e-08,
|
|
"logits/chosen": -7.909399032592773,
|
|
"logits/rejected": -7.158049583435059,
|
|
"logps/chosen": -246.2357177734375,
|
|
"logps/ref_chosen": -101.63691711425781,
|
|
"logps/ref_rejected": -93.34539031982422,
|
|
"logps/rejected": -336.9823913574219,
|
|
"loss": 0.9473,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.859551191329956,
|
|
"rewards/margins": 0.5850256681442261,
|
|
"rewards/rejected": -1.4445768594741821,
|
|
"step": 532
|
|
},
|
|
{
|
|
"epoch": 0.7826725403817915,
|
|
"grad_norm": 9.506068229675293,
|
|
"kl/avg_steps": 0.4375,
|
|
"kl/beta": 0.0059357923455536366,
|
|
"kl/n_epsilon_steps": 0.28125,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 6.963101805503646e-08,
|
|
"logits/chosen": -7.598127365112305,
|
|
"logits/rejected": -6.821909427642822,
|
|
"logps/chosen": -232.10377502441406,
|
|
"logps/ref_chosen": -90.74664306640625,
|
|
"logps/ref_rejected": -82.88848876953125,
|
|
"logps/rejected": -282.5700378417969,
|
|
"loss": 1.1378,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -0.8360397219657898,
|
|
"rewards/margins": 0.3414258658885956,
|
|
"rewards/rejected": -1.177465558052063,
|
|
"step": 533
|
|
},
|
|
{
|
|
"epoch": 0.7841409691629956,
|
|
"grad_norm": 8.0197114944458,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.005909936036914587,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 6.874476976660184e-08,
|
|
"logits/chosen": -7.659628391265869,
|
|
"logits/rejected": -7.227793216705322,
|
|
"logps/chosen": -214.4866943359375,
|
|
"logps/ref_chosen": -86.63084411621094,
|
|
"logps/ref_rejected": -84.87196350097656,
|
|
"logps/rejected": -296.1422424316406,
|
|
"loss": 1.0025,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -0.7507482767105103,
|
|
"rewards/margins": 0.4875665009021759,
|
|
"rewards/rejected": -1.2383147478103638,
|
|
"step": 534
|
|
},
|
|
{
|
|
"epoch": 0.7856093979441997,
|
|
"grad_norm": 7.752889633178711,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.005867761559784412,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 6.786329772205246e-08,
|
|
"logits/chosen": -7.62947940826416,
|
|
"logits/rejected": -7.0962138175964355,
|
|
"logps/chosen": -216.57179260253906,
|
|
"logps/ref_chosen": -90.34539031982422,
|
|
"logps/ref_rejected": -100.7779769897461,
|
|
"logps/rejected": -315.5731201171875,
|
|
"loss": 1.0009,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.7379162907600403,
|
|
"rewards/margins": 0.5136837363243103,
|
|
"rewards/rejected": -1.2516000270843506,
|
|
"step": 535
|
|
},
|
|
{
|
|
"epoch": 0.7870778267254038,
|
|
"grad_norm": 8.29379940032959,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.005833127535879612,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 6.698662514899638e-08,
|
|
"logits/chosen": -7.761280059814453,
|
|
"logits/rejected": -7.13836669921875,
|
|
"logps/chosen": -211.44839477539062,
|
|
"logps/ref_chosen": -78.95956420898438,
|
|
"logps/ref_rejected": -93.08779907226562,
|
|
"logps/rejected": -308.13739013671875,
|
|
"loss": 1.0251,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -0.7697625756263733,
|
|
"rewards/margins": 0.47593456506729126,
|
|
"rewards/rejected": -1.245697021484375,
|
|
"step": 536
|
|
},
|
|
{
|
|
"epoch": 0.788546255506608,
|
|
"grad_norm": 7.392307758331299,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.005796896759420633,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 6.611477514857114e-08,
|
|
"logits/chosen": -7.517266273498535,
|
|
"logits/rejected": -6.534526824951172,
|
|
"logps/chosen": -200.18478393554688,
|
|
"logps/ref_chosen": -86.70939636230469,
|
|
"logps/ref_rejected": -78.90767669677734,
|
|
"logps/rejected": -286.97747802734375,
|
|
"loss": 0.9757,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -0.654714047908783,
|
|
"rewards/margins": 0.5424842834472656,
|
|
"rewards/rejected": -1.1971983909606934,
|
|
"step": 537
|
|
},
|
|
{
|
|
"epoch": 0.7900146842878121,
|
|
"grad_norm": 8.975393295288086,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.005760891363024712,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 6.524777069483525e-08,
|
|
"logits/chosen": -7.621641159057617,
|
|
"logits/rejected": -7.222417831420898,
|
|
"logps/chosen": -250.26470947265625,
|
|
"logps/ref_chosen": -98.44214630126953,
|
|
"logps/ref_rejected": -91.09465026855469,
|
|
"logps/rejected": -319.07464599609375,
|
|
"loss": 1.0446,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -0.8716259002685547,
|
|
"rewards/margins": 0.43390387296676636,
|
|
"rewards/rejected": -1.3055297136306763,
|
|
"step": 538
|
|
},
|
|
{
|
|
"epoch": 0.7914831130690162,
|
|
"grad_norm": 7.8592424392700195,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.005730448290705681,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 6.438563463416221e-08,
|
|
"logits/chosen": -7.682827949523926,
|
|
"logits/rejected": -7.03498649597168,
|
|
"logps/chosen": -227.2099609375,
|
|
"logps/ref_chosen": -90.93212127685547,
|
|
"logps/ref_rejected": -97.44546508789062,
|
|
"logps/rejected": -312.05865478515625,
|
|
"loss": 1.0608,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.7778847217559814,
|
|
"rewards/margins": 0.44326433539390564,
|
|
"rewards/rejected": -1.2211490869522095,
|
|
"step": 539
|
|
},
|
|
{
|
|
"epoch": 0.7929515418502202,
|
|
"grad_norm": 9.220181465148926,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.005696624517440796,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 6.352838968463919e-08,
|
|
"logits/chosen": -8.04260540008545,
|
|
"logits/rejected": -7.764565467834473,
|
|
"logps/chosen": -247.24325561523438,
|
|
"logps/ref_chosen": -107.8698959350586,
|
|
"logps/ref_rejected": -121.39349365234375,
|
|
"logps/rejected": -353.7125549316406,
|
|
"loss": 1.0036,
|
|
"rewards/accuracies": 0.890625,
|
|
"rewards/chosen": -0.789432942867279,
|
|
"rewards/margins": 0.5233358144760132,
|
|
"rewards/rejected": -1.3127686977386475,
|
|
"step": 540
|
|
},
|
|
{
|
|
"epoch": 0.7944199706314243,
|
|
"grad_norm": 8.543745040893555,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.00565597228705883,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 6.267605843546767e-08,
|
|
"logits/chosen": -7.653639793395996,
|
|
"logits/rejected": -7.352320671081543,
|
|
"logps/chosen": -264.3243713378906,
|
|
"logps/ref_chosen": -106.56748962402344,
|
|
"logps/ref_rejected": -109.06633758544922,
|
|
"logps/rejected": -329.9075927734375,
|
|
"loss": 1.114,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -0.8885458111763,
|
|
"rewards/margins": 0.3523477613925934,
|
|
"rewards/rejected": -1.2408936023712158,
|
|
"step": 541
|
|
},
|
|
{
|
|
"epoch": 0.7958883994126285,
|
|
"grad_norm": 8.278136253356934,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.005626083817332983,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 6.182866334636888e-08,
|
|
"logits/chosen": -7.408226013183594,
|
|
"logits/rejected": -7.1460490226745605,
|
|
"logps/chosen": -218.92977905273438,
|
|
"logps/ref_chosen": -87.36929321289062,
|
|
"logps/ref_rejected": -102.82034301757812,
|
|
"logps/rejected": -326.39044189453125,
|
|
"loss": 0.9943,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.7365444302558899,
|
|
"rewards/margins": 0.5116760730743408,
|
|
"rewards/rejected": -1.248220443725586,
|
|
"step": 542
|
|
},
|
|
{
|
|
"epoch": 0.7973568281938326,
|
|
"grad_norm": 9.519865989685059,
|
|
"kl/avg_steps": 0.375,
|
|
"kl/beta": 0.005589403212070465,
|
|
"kl/n_epsilon_steps": 0.3125,
|
|
"kl/p_epsilon_steps": 0.6875,
|
|
"learning_rate": 6.098622674699147e-08,
|
|
"logits/chosen": -7.630051612854004,
|
|
"logits/rejected": -7.440648078918457,
|
|
"logps/chosen": -244.11569213867188,
|
|
"logps/ref_chosen": -87.152587890625,
|
|
"logps/ref_rejected": -111.53172302246094,
|
|
"logps/rejected": -317.44921875,
|
|
"loss": 1.2137,
|
|
"rewards/accuracies": 0.6875,
|
|
"rewards/chosen": -0.8755909204483032,
|
|
"rewards/margins": 0.2687823474407196,
|
|
"rewards/rejected": -1.1443732976913452,
|
|
"step": 543
|
|
},
|
|
{
|
|
"epoch": 0.7988252569750367,
|
|
"grad_norm": 10.596867561340332,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.005568521562963724,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 6.01487708363232e-08,
|
|
"logits/chosen": -7.832944869995117,
|
|
"logits/rejected": -7.403214931488037,
|
|
"logps/chosen": -244.70623779296875,
|
|
"logps/ref_chosen": -86.14691162109375,
|
|
"logps/ref_rejected": -106.3939208984375,
|
|
"logps/rejected": -325.4195861816406,
|
|
"loss": 1.1329,
|
|
"rewards/accuracies": 0.734375,
|
|
"rewards/chosen": -0.8795119524002075,
|
|
"rewards/margins": 0.33259445428848267,
|
|
"rewards/rejected": -1.212106466293335,
|
|
"step": 544
|
|
},
|
|
{
|
|
"epoch": 0.8002936857562408,
|
|
"grad_norm": 8.951570510864258,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.005542540457099676,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 5.9316317682106294e-08,
|
|
"logits/chosen": -7.671846389770508,
|
|
"logits/rejected": -7.305430889129639,
|
|
"logps/chosen": -242.76431274414062,
|
|
"logps/ref_chosen": -96.42424011230469,
|
|
"logps/ref_rejected": -100.73405456542969,
|
|
"logps/rejected": -321.11676025390625,
|
|
"loss": 1.0893,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.8078324794769287,
|
|
"rewards/margins": 0.4048755168914795,
|
|
"rewards/rejected": -1.2127079963684082,
|
|
"step": 545
|
|
},
|
|
{
|
|
"epoch": 0.801762114537445,
|
|
"grad_norm": 8.22799301147461,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.005509825889021158,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 5.848888922025552e-08,
|
|
"logits/chosen": -7.779958724975586,
|
|
"logits/rejected": -7.305843353271484,
|
|
"logps/chosen": -233.88308715820312,
|
|
"logps/ref_chosen": -86.02112579345703,
|
|
"logps/ref_rejected": -87.37263488769531,
|
|
"logps/rejected": -306.00103759765625,
|
|
"loss": 1.0982,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.8111412525177002,
|
|
"rewards/margins": 0.3848496675491333,
|
|
"rewards/rejected": -1.195991039276123,
|
|
"step": 546
|
|
},
|
|
{
|
|
"epoch": 0.8032305433186491,
|
|
"grad_norm": 8.054706573486328,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.005479006562381983,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 5.7666507254280265e-08,
|
|
"logits/chosen": -7.915996074676514,
|
|
"logits/rejected": -7.356800556182861,
|
|
"logps/chosen": -244.9899444580078,
|
|
"logps/ref_chosen": -100.65100860595703,
|
|
"logps/ref_rejected": -97.9117431640625,
|
|
"logps/rejected": -322.5784912109375,
|
|
"loss": 1.0372,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -0.7865848541259766,
|
|
"rewards/margins": 0.43519890308380127,
|
|
"rewards/rejected": -1.2217838764190674,
|
|
"step": 547
|
|
},
|
|
{
|
|
"epoch": 0.8046989720998532,
|
|
"grad_norm": 9.502042770385742,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.005443285219371319,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 5.684919345471029e-08,
|
|
"logits/chosen": -7.732963562011719,
|
|
"logits/rejected": -7.511667251586914,
|
|
"logps/chosen": -241.53533935546875,
|
|
"logps/ref_chosen": -96.03010559082031,
|
|
"logps/ref_rejected": -98.41322326660156,
|
|
"logps/rejected": -305.8096618652344,
|
|
"loss": 1.1549,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -0.7902284860610962,
|
|
"rewards/margins": 0.3315829634666443,
|
|
"rewards/rejected": -1.1218115091323853,
|
|
"step": 548
|
|
},
|
|
{
|
|
"epoch": 0.8061674008810573,
|
|
"grad_norm": 9.200112342834473,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.00541620422154665,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 5.603696935852426e-08,
|
|
"logits/chosen": -7.810609340667725,
|
|
"logits/rejected": -7.074420928955078,
|
|
"logps/chosen": -215.589111328125,
|
|
"logps/ref_chosen": -83.45096588134766,
|
|
"logps/ref_rejected": -81.38023376464844,
|
|
"logps/rejected": -288.0585632324219,
|
|
"loss": 1.0911,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.7132003307342529,
|
|
"rewards/margins": 0.39838820695877075,
|
|
"rewards/rejected": -1.111588716506958,
|
|
"step": 549
|
|
},
|
|
{
|
|
"epoch": 0.8076358296622613,
|
|
"grad_norm": 8.178189277648926,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.00538256298750639,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 5.5229856368582376e-08,
|
|
"logits/chosen": -7.702338218688965,
|
|
"logits/rejected": -7.392082214355469,
|
|
"logps/chosen": -239.71810913085938,
|
|
"logps/ref_chosen": -89.59434509277344,
|
|
"logps/ref_rejected": -101.31549835205078,
|
|
"logps/rejected": -320.63720703125,
|
|
"loss": 1.0957,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -0.8037484884262085,
|
|
"rewards/margins": 0.36810043454170227,
|
|
"rewards/rejected": -1.171849012374878,
|
|
"step": 550
|
|
},
|
|
{
|
|
"epoch": 0.8091042584434655,
|
|
"grad_norm": 8.914412498474121,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.005347470287233591,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 5.4427875753062734e-08,
|
|
"logits/chosen": -7.659028053283691,
|
|
"logits/rejected": -7.469613075256348,
|
|
"logps/chosen": -226.66494750976562,
|
|
"logps/ref_chosen": -95.08216094970703,
|
|
"logps/ref_rejected": -117.7192611694336,
|
|
"logps/rejected": -333.44720458984375,
|
|
"loss": 1.0428,
|
|
"rewards/accuracies": 0.890625,
|
|
"rewards/chosen": -0.6989625692367554,
|
|
"rewards/margins": 0.4451354146003723,
|
|
"rewards/rejected": -1.144097924232483,
|
|
"step": 551
|
|
},
|
|
{
|
|
"epoch": 0.8105726872246696,
|
|
"grad_norm": 7.542394638061523,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.005309309810400009,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 5.363104864490034e-08,
|
|
"logits/chosen": -8.226886749267578,
|
|
"logits/rejected": -7.813612937927246,
|
|
"logps/chosen": -229.49818420410156,
|
|
"logps/ref_chosen": -101.40918731689453,
|
|
"logps/ref_rejected": -110.83843994140625,
|
|
"logps/rejected": -349.8272705078125,
|
|
"loss": 0.9501,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.6762351989746094,
|
|
"rewards/margins": 0.582294762134552,
|
|
"rewards/rejected": -1.2585299015045166,
|
|
"step": 552
|
|
},
|
|
{
|
|
"epoch": 0.8120411160058737,
|
|
"grad_norm": 8.038527488708496,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.005273057147860527,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 5.2839396041230415e-08,
|
|
"logits/chosen": -7.498537540435791,
|
|
"logits/rejected": -6.907330513000488,
|
|
"logps/chosen": -237.8641815185547,
|
|
"logps/ref_chosen": -97.96388244628906,
|
|
"logps/ref_rejected": -103.57719421386719,
|
|
"logps/rejected": -313.3408203125,
|
|
"loss": 1.1097,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -0.7355344295501709,
|
|
"rewards/margins": 0.3635327219963074,
|
|
"rewards/rejected": -1.0990670919418335,
|
|
"step": 553
|
|
},
|
|
{
|
|
"epoch": 0.8135095447870778,
|
|
"grad_norm": 7.692688941955566,
|
|
"kl/avg_steps": 0.40625,
|
|
"kl/beta": 0.005246823187917471,
|
|
"kl/n_epsilon_steps": 0.296875,
|
|
"kl/p_epsilon_steps": 0.703125,
|
|
"learning_rate": 5.205293880283551e-08,
|
|
"logits/chosen": -7.834980487823486,
|
|
"logits/rejected": -7.495749473571777,
|
|
"logps/chosen": -251.03372192382812,
|
|
"logps/ref_chosen": -103.68680572509766,
|
|
"logps/ref_rejected": -92.99210357666016,
|
|
"logps/rejected": -316.3883056640625,
|
|
"loss": 1.1078,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -0.7710244655609131,
|
|
"rewards/margins": 0.39337316155433655,
|
|
"rewards/rejected": -1.1643975973129272,
|
|
"step": 554
|
|
},
|
|
{
|
|
"epoch": 0.8149779735682819,
|
|
"grad_norm": 8.527905464172363,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.005225594155490398,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 5.127169765359515e-08,
|
|
"logits/chosen": -7.357694149017334,
|
|
"logits/rejected": -7.210930824279785,
|
|
"logps/chosen": -241.05128479003906,
|
|
"logps/ref_chosen": -86.20362854003906,
|
|
"logps/ref_rejected": -114.54719543457031,
|
|
"logps/rejected": -343.023193359375,
|
|
"loss": 1.1068,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -0.8064323663711548,
|
|
"rewards/margins": 0.37958627939224243,
|
|
"rewards/rejected": -1.1860185861587524,
|
|
"step": 555
|
|
},
|
|
{
|
|
"epoch": 0.8164464023494861,
|
|
"grad_norm": 7.89186954498291,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.005197979975491762,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 5.049569317994012e-08,
|
|
"logits/chosen": -7.9103193283081055,
|
|
"logits/rejected": -6.936588287353516,
|
|
"logps/chosen": -226.0050811767578,
|
|
"logps/ref_chosen": -88.40348815917969,
|
|
"logps/ref_rejected": -106.63912200927734,
|
|
"logps/rejected": -323.8564147949219,
|
|
"loss": 1.0752,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -0.7129172086715698,
|
|
"rewards/margins": 0.40871816873550415,
|
|
"rewards/rejected": -1.1216354370117188,
|
|
"step": 556
|
|
},
|
|
{
|
|
"epoch": 0.8179148311306902,
|
|
"grad_norm": 6.813101291656494,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.005170511547476053,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.9724945830310144e-08,
|
|
"logits/chosen": -7.78956413269043,
|
|
"logits/rejected": -7.387689590454102,
|
|
"logps/chosen": -245.75643920898438,
|
|
"logps/ref_chosen": -98.83523559570312,
|
|
"logps/ref_rejected": -115.64888763427734,
|
|
"logps/rejected": -350.8326110839844,
|
|
"loss": 1.0491,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -0.7568396925926208,
|
|
"rewards/margins": 0.45094895362854004,
|
|
"rewards/rejected": -1.2077887058258057,
|
|
"step": 557
|
|
},
|
|
{
|
|
"epoch": 0.8193832599118943,
|
|
"grad_norm": 8.840348243713379,
|
|
"kl/avg_steps": 0.8125,
|
|
"kl/beta": 0.005141590256243944,
|
|
"kl/n_epsilon_steps": 0.09375,
|
|
"kl/p_epsilon_steps": 0.90625,
|
|
"learning_rate": 4.8959475914614554e-08,
|
|
"logits/chosen": -8.046621322631836,
|
|
"logits/rejected": -7.308237075805664,
|
|
"logps/chosen": -254.44580078125,
|
|
"logps/ref_chosen": -106.01183319091797,
|
|
"logps/ref_rejected": -108.39131164550781,
|
|
"logps/rejected": -347.26983642578125,
|
|
"loss": 1.0219,
|
|
"rewards/accuracies": 0.890625,
|
|
"rewards/chosen": -0.7578614950180054,
|
|
"rewards/margins": 0.45999467372894287,
|
|
"rewards/rejected": -1.2178561687469482,
|
|
"step": 558
|
|
},
|
|
{
|
|
"epoch": 0.8208516886930984,
|
|
"grad_norm": 9.36201000213623,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.005100151523947716,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 4.8199303603697614e-08,
|
|
"logits/chosen": -7.8494062423706055,
|
|
"logits/rejected": -7.495527267456055,
|
|
"logps/chosen": -257.96856689453125,
|
|
"logps/ref_chosen": -100.42474365234375,
|
|
"logps/ref_rejected": -100.41270446777344,
|
|
"logps/rejected": -334.3748474121094,
|
|
"loss": 1.0914,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.79989093542099,
|
|
"rewards/margins": 0.38467034697532654,
|
|
"rewards/rejected": -1.1845612525939941,
|
|
"step": 559
|
|
},
|
|
{
|
|
"epoch": 0.8223201174743024,
|
|
"grad_norm": 10.975285530090332,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.005070047918707132,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 4.7444448928806615e-08,
|
|
"logits/chosen": -7.894330978393555,
|
|
"logits/rejected": -7.052946090698242,
|
|
"logps/chosen": -242.33444213867188,
|
|
"logps/ref_chosen": -87.63551330566406,
|
|
"logps/ref_rejected": -87.37985229492188,
|
|
"logps/rejected": -313.7993469238281,
|
|
"loss": 1.1114,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.781200647354126,
|
|
"rewards/margins": 0.35893329977989197,
|
|
"rewards/rejected": -1.1401338577270508,
|
|
"step": 560
|
|
},
|
|
{
|
|
"epoch": 0.8237885462555066,
|
|
"grad_norm": 7.341772556304932,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.005043255630880594,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 4.669493178106432e-08,
|
|
"logits/chosen": -7.702178001403809,
|
|
"logits/rejected": -7.204655647277832,
|
|
"logps/chosen": -235.06967163085938,
|
|
"logps/ref_chosen": -74.88140106201172,
|
|
"logps/ref_rejected": -106.04796600341797,
|
|
"logps/rejected": -336.55767822265625,
|
|
"loss": 1.104,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.8038696050643921,
|
|
"rewards/margins": 0.35057592391967773,
|
|
"rewards/rejected": -1.1544455289840698,
|
|
"step": 561
|
|
},
|
|
{
|
|
"epoch": 0.8252569750367107,
|
|
"grad_norm": 8.39021110534668,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.005010375287383795,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 4.5950771910944596e-08,
|
|
"logits/chosen": -7.776498794555664,
|
|
"logits/rejected": -7.434757232666016,
|
|
"logps/chosen": -246.33297729492188,
|
|
"logps/ref_chosen": -100.66322326660156,
|
|
"logps/ref_rejected": -102.24087524414062,
|
|
"logps/rejected": -314.5208740234375,
|
|
"loss": 1.1451,
|
|
"rewards/accuracies": 0.71875,
|
|
"rewards/chosen": -0.7275943756103516,
|
|
"rewards/margins": 0.32898658514022827,
|
|
"rewards/rejected": -1.056580901145935,
|
|
"step": 562
|
|
},
|
|
{
|
|
"epoch": 0.8267254038179148,
|
|
"grad_norm": 8.17624568939209,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.004985447973012924,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 4.521198892775202e-08,
|
|
"logits/chosen": -7.574672698974609,
|
|
"logits/rejected": -7.067543983459473,
|
|
"logps/chosen": -251.52645874023438,
|
|
"logps/ref_chosen": -96.25921630859375,
|
|
"logps/ref_rejected": -101.20503997802734,
|
|
"logps/rejected": -318.0636901855469,
|
|
"loss": 1.151,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -0.7714153528213501,
|
|
"rewards/margins": 0.30289867520332336,
|
|
"rewards/rejected": -1.0743141174316406,
|
|
"step": 563
|
|
},
|
|
{
|
|
"epoch": 0.8281938325991189,
|
|
"grad_norm": 8.074674606323242,
|
|
"kl/avg_steps": 0.578125,
|
|
"kl/beta": 0.0049606445245444775,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.447860229910544e-08,
|
|
"logits/chosen": -7.528448104858398,
|
|
"logits/rejected": -7.25858736038208,
|
|
"logps/chosen": -263.2978210449219,
|
|
"logps/ref_chosen": -112.96040344238281,
|
|
"logps/ref_rejected": -99.52803802490234,
|
|
"logps/rejected": -316.2669982910156,
|
|
"loss": 1.1213,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -0.7416489124298096,
|
|
"rewards/margins": 0.32538917660713196,
|
|
"rewards/rejected": -1.0670380592346191,
|
|
"step": 564
|
|
},
|
|
{
|
|
"epoch": 0.8296622613803231,
|
|
"grad_norm": 7.512981414794922,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.004932130686938763,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 4.375063135042445e-08,
|
|
"logits/chosen": -7.909860610961914,
|
|
"logits/rejected": -7.201497554779053,
|
|
"logps/chosen": -239.09127807617188,
|
|
"logps/ref_chosen": -95.36558532714844,
|
|
"logps/ref_rejected": -92.21737670898438,
|
|
"logps/rejected": -322.09124755859375,
|
|
"loss": 1.0586,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.7062567472457886,
|
|
"rewards/margins": 0.41997939348220825,
|
|
"rewards/rejected": -1.1262362003326416,
|
|
"step": 565
|
|
},
|
|
{
|
|
"epoch": 0.8311306901615272,
|
|
"grad_norm": 6.797919750213623,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.004906067159026861,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 4.3028095264420525e-08,
|
|
"logits/chosen": -7.698864936828613,
|
|
"logits/rejected": -7.01510763168335,
|
|
"logps/chosen": -248.45718383789062,
|
|
"logps/ref_chosen": -98.97320556640625,
|
|
"logps/ref_rejected": -110.82090759277344,
|
|
"logps/rejected": -340.752197265625,
|
|
"loss": 1.0825,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.7296649217605591,
|
|
"rewards/margins": 0.3899438977241516,
|
|
"rewards/rejected": -1.1196088790893555,
|
|
"step": 566
|
|
},
|
|
{
|
|
"epoch": 0.8325991189427313,
|
|
"grad_norm": 8.844162940979004,
|
|
"kl/avg_steps": 0.578125,
|
|
"kl/beta": 0.00487559475004673,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.231101308059165e-08,
|
|
"logits/chosen": -7.589968681335449,
|
|
"logits/rejected": -6.949653625488281,
|
|
"logps/chosen": -247.43397521972656,
|
|
"logps/ref_chosen": -92.7035903930664,
|
|
"logps/ref_rejected": -91.22431945800781,
|
|
"logps/rejected": -315.1552734375,
|
|
"loss": 1.1228,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.7517495155334473,
|
|
"rewards/margins": 0.3330436944961548,
|
|
"rewards/rejected": -1.084793210029602,
|
|
"step": 567
|
|
},
|
|
{
|
|
"epoch": 0.8340675477239354,
|
|
"grad_norm": 7.321277618408203,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.004847569856792688,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 4.1599403694720145e-08,
|
|
"logits/chosen": -7.3952226638793945,
|
|
"logits/rejected": -7.033419609069824,
|
|
"logps/chosen": -219.06063842773438,
|
|
"logps/ref_chosen": -73.13226318359375,
|
|
"logps/ref_rejected": -94.95265197753906,
|
|
"logps/rejected": -326.4876708984375,
|
|
"loss": 1.0596,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -0.7039542198181152,
|
|
"rewards/margins": 0.4104452133178711,
|
|
"rewards/rejected": -1.1143994331359863,
|
|
"step": 568
|
|
},
|
|
{
|
|
"epoch": 0.8355359765051396,
|
|
"grad_norm": 10.243264198303223,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.004818957298994064,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 4.089328585837512e-08,
|
|
"logits/chosen": -7.595402240753174,
|
|
"logits/rejected": -7.113461494445801,
|
|
"logps/chosen": -243.99252319335938,
|
|
"logps/ref_chosen": -89.21708679199219,
|
|
"logps/ref_rejected": -86.41105651855469,
|
|
"logps/rejected": -312.2991943359375,
|
|
"loss": 1.1345,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.7422770261764526,
|
|
"rewards/margins": 0.3377072811126709,
|
|
"rewards/rejected": -1.0799843072891235,
|
|
"step": 569
|
|
},
|
|
{
|
|
"epoch": 0.8370044052863436,
|
|
"grad_norm": 7.365958213806152,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.004789025988429785,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 4.019267817841834e-08,
|
|
"logits/chosen": -7.986109733581543,
|
|
"logits/rejected": -7.3613433837890625,
|
|
"logps/chosen": -233.00393676757812,
|
|
"logps/ref_chosen": -94.76277160644531,
|
|
"logps/ref_rejected": -88.84693145751953,
|
|
"logps/rejected": -320.5428466796875,
|
|
"loss": 1.029,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.6582140922546387,
|
|
"rewards/margins": 0.4426998496055603,
|
|
"rewards/rejected": -1.1009138822555542,
|
|
"step": 570
|
|
},
|
|
{
|
|
"epoch": 0.8384728340675477,
|
|
"grad_norm": 9.40793514251709,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.004756326321512461,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 3.9497599116513705e-08,
|
|
"logits/chosen": -7.549665451049805,
|
|
"logits/rejected": -7.4702606201171875,
|
|
"logps/chosen": -221.9232177734375,
|
|
"logps/ref_chosen": -80.8212890625,
|
|
"logps/ref_rejected": -97.01528930664062,
|
|
"logps/rejected": -323.5943603515625,
|
|
"loss": 1.0783,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.6683664917945862,
|
|
"rewards/margins": 0.40152615308761597,
|
|
"rewards/rejected": -1.0698926448822021,
|
|
"step": 571
|
|
},
|
|
{
|
|
"epoch": 0.8399412628487518,
|
|
"grad_norm": 7.180628776550293,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.004728252068161964,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 3.880806698864086e-08,
|
|
"logits/chosen": -7.4798078536987305,
|
|
"logits/rejected": -7.4275007247924805,
|
|
"logps/chosen": -221.84637451171875,
|
|
"logps/ref_chosen": -77.37992858886719,
|
|
"logps/ref_rejected": -90.56597900390625,
|
|
"logps/rejected": -319.32598876953125,
|
|
"loss": 1.0794,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.6801784038543701,
|
|
"rewards/margins": 0.39393994212150574,
|
|
"rewards/rejected": -1.0741183757781982,
|
|
"step": 572
|
|
},
|
|
{
|
|
"epoch": 0.8414096916299559,
|
|
"grad_norm": 7.156675338745117,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.004701804369688034,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 3.812409996461275e-08,
|
|
"logits/chosen": -7.6273932456970215,
|
|
"logits/rejected": -7.473799228668213,
|
|
"logps/chosen": -224.49549865722656,
|
|
"logps/ref_chosen": -83.7060775756836,
|
|
"logps/ref_rejected": -90.69746398925781,
|
|
"logps/rejected": -310.0919494628906,
|
|
"loss": 1.1016,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.6590225100517273,
|
|
"rewards/margins": 0.3648913502693176,
|
|
"rewards/rejected": -1.023913860321045,
|
|
"step": 573
|
|
},
|
|
{
|
|
"epoch": 0.8428781204111601,
|
|
"grad_norm": 8.27857494354248,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.004671149887144566,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 3.74457160675965e-08,
|
|
"logits/chosen": -7.626450538635254,
|
|
"logits/rejected": -7.274733066558838,
|
|
"logps/chosen": -234.46607971191406,
|
|
"logps/ref_chosen": -85.12948608398438,
|
|
"logps/ref_rejected": -99.05403137207031,
|
|
"logps/rejected": -330.6099548339844,
|
|
"loss": 1.0924,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.6954778432846069,
|
|
"rewards/margins": 0.37927812337875366,
|
|
"rewards/rejected": -1.074756145477295,
|
|
"step": 574
|
|
},
|
|
{
|
|
"epoch": 0.8443465491923642,
|
|
"grad_norm": 9.878924369812012,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.004645021632313728,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 3.677293317363864e-08,
|
|
"logits/chosen": -7.789747714996338,
|
|
"logits/rejected": -7.52932071685791,
|
|
"logps/chosen": -258.09442138671875,
|
|
"logps/ref_chosen": -105.77084350585938,
|
|
"logps/ref_rejected": -100.94944763183594,
|
|
"logps/rejected": -335.0401306152344,
|
|
"loss": 1.0999,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.7044973969459534,
|
|
"rewards/margins": 0.37485966086387634,
|
|
"rewards/rejected": -1.0793570280075073,
|
|
"step": 575
|
|
},
|
|
{
|
|
"epoch": 0.8458149779735683,
|
|
"grad_norm": 7.550034046173096,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.004617604892700911,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 3.6105769011194224e-08,
|
|
"logits/chosen": -7.752475738525391,
|
|
"logits/rejected": -7.319489479064941,
|
|
"logps/chosen": -234.95172119140625,
|
|
"logps/ref_chosen": -81.20452880859375,
|
|
"logps/ref_rejected": -105.40940856933594,
|
|
"logps/rejected": -339.11627197265625,
|
|
"loss": 1.1064,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -0.7074535489082336,
|
|
"rewards/margins": 0.36441028118133545,
|
|
"rewards/rejected": -1.0718638896942139,
|
|
"step": 576
|
|
},
|
|
{
|
|
"epoch": 0.8472834067547724,
|
|
"grad_norm": 6.790652275085449,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.00459320330992341,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 3.5444241160659304e-08,
|
|
"logits/chosen": -7.78272819519043,
|
|
"logits/rejected": -7.093866348266602,
|
|
"logps/chosen": -238.08453369140625,
|
|
"logps/ref_chosen": -101.46419525146484,
|
|
"logps/ref_rejected": -89.32637023925781,
|
|
"logps/rejected": -322.09918212890625,
|
|
"loss": 1.0486,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.6245477199554443,
|
|
"rewards/margins": 0.43618321418762207,
|
|
"rewards/rejected": -1.0607309341430664,
|
|
"step": 577
|
|
},
|
|
{
|
|
"epoch": 0.8487518355359766,
|
|
"grad_norm": 10.109602928161621,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.004563257098197937,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 3.478836705390808e-08,
|
|
"logits/chosen": -7.468972682952881,
|
|
"logits/rejected": -7.147921562194824,
|
|
"logps/chosen": -239.65093994140625,
|
|
"logps/ref_chosen": -77.29241943359375,
|
|
"logps/ref_rejected": -88.24701690673828,
|
|
"logps/rejected": -317.5960693359375,
|
|
"loss": 1.1523,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -0.7381951808929443,
|
|
"rewards/margins": 0.3014621138572693,
|
|
"rewards/rejected": -1.0396572351455688,
|
|
"step": 578
|
|
},
|
|
{
|
|
"epoch": 0.8502202643171806,
|
|
"grad_norm": 7.376382350921631,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.004539142828434706,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 3.41381639738331e-08,
|
|
"logits/chosen": -7.745573997497559,
|
|
"logits/rejected": -6.99274206161499,
|
|
"logps/chosen": -232.49017333984375,
|
|
"logps/ref_chosen": -87.34305572509766,
|
|
"logps/ref_rejected": -102.12100982666016,
|
|
"logps/rejected": -319.19024658203125,
|
|
"loss": 1.13,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.6563674211502075,
|
|
"rewards/margins": 0.3224019408226013,
|
|
"rewards/rejected": -0.9787693023681641,
|
|
"step": 579
|
|
},
|
|
{
|
|
"epoch": 0.8516886930983847,
|
|
"grad_norm": 7.219406604766846,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.004515156149864197,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 3.349364905389032e-08,
|
|
"logits/chosen": -7.58662223815918,
|
|
"logits/rejected": -7.382083892822266,
|
|
"logps/chosen": -229.9556121826172,
|
|
"logps/ref_chosen": -88.92879486083984,
|
|
"logps/ref_rejected": -88.69661712646484,
|
|
"logps/rejected": -315.407958984375,
|
|
"loss": 1.0857,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.6354615092277527,
|
|
"rewards/margins": 0.3822898864746094,
|
|
"rewards/rejected": -1.0177514553070068,
|
|
"step": 580
|
|
},
|
|
{
|
|
"epoch": 0.8531571218795888,
|
|
"grad_norm": 8.329227447509766,
|
|
"kl/avg_steps": 0.4375,
|
|
"kl/beta": 0.0044926926493644714,
|
|
"kl/n_epsilon_steps": 0.28125,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 3.285483927764726e-08,
|
|
"logits/chosen": -7.649691581726074,
|
|
"logits/rejected": -7.227565765380859,
|
|
"logps/chosen": -242.04396057128906,
|
|
"logps/ref_chosen": -92.22323608398438,
|
|
"logps/ref_rejected": -99.38943481445312,
|
|
"logps/rejected": -324.2420654296875,
|
|
"loss": 1.1356,
|
|
"rewards/accuracies": 0.734375,
|
|
"rewards/chosen": -0.671623945236206,
|
|
"rewards/margins": 0.33214688301086426,
|
|
"rewards/rejected": -1.0037708282470703,
|
|
"step": 581
|
|
},
|
|
{
|
|
"epoch": 0.8546255506607929,
|
|
"grad_norm": 6.972781181335449,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.004473122768104076,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 3.222175147833556e-08,
|
|
"logits/chosen": -7.825651168823242,
|
|
"logits/rejected": -7.724842548370361,
|
|
"logps/chosen": -217.5670928955078,
|
|
"logps/ref_chosen": -87.33561706542969,
|
|
"logps/ref_rejected": -115.1544189453125,
|
|
"logps/rejected": -317.5841064453125,
|
|
"loss": 1.1379,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.5807515382766724,
|
|
"rewards/margins": 0.3188899755477905,
|
|
"rewards/rejected": -0.8996415138244629,
|
|
"step": 582
|
|
},
|
|
{
|
|
"epoch": 0.856093979441997,
|
|
"grad_norm": 7.694782733917236,
|
|
"kl/avg_steps": 0.4375,
|
|
"kl/beta": 0.004448102321475744,
|
|
"kl/n_epsilon_steps": 0.28125,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 3.159440233840763e-08,
|
|
"logits/chosen": -7.3343915939331055,
|
|
"logits/rejected": -7.080375671386719,
|
|
"logps/chosen": -240.53501892089844,
|
|
"logps/ref_chosen": -83.71707153320312,
|
|
"logps/ref_rejected": -94.58465576171875,
|
|
"logps/rejected": -328.16815185546875,
|
|
"loss": 1.12,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -0.695499837398529,
|
|
"rewards/margins": 0.337327241897583,
|
|
"rewards/rejected": -1.0328271389007568,
|
|
"step": 583
|
|
},
|
|
{
|
|
"epoch": 0.8575624082232012,
|
|
"grad_norm": 7.240791320800781,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.0044287266209721565,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 3.0972808389096635e-08,
|
|
"logits/chosen": -7.599449157714844,
|
|
"logits/rejected": -7.185504913330078,
|
|
"logps/chosen": -232.9828338623047,
|
|
"logps/ref_chosen": -94.25955963134766,
|
|
"logps/ref_rejected": -104.48269653320312,
|
|
"logps/rejected": -325.73150634765625,
|
|
"loss": 1.104,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.6124042272567749,
|
|
"rewards/margins": 0.3608018755912781,
|
|
"rewards/rejected": -0.9732060432434082,
|
|
"step": 584
|
|
},
|
|
{
|
|
"epoch": 0.8590308370044053,
|
|
"grad_norm": 6.849031925201416,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.004405323415994644,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 3.035698600998121e-08,
|
|
"logits/chosen": -7.737574577331543,
|
|
"logits/rejected": -7.5866241455078125,
|
|
"logps/chosen": -234.18655395507812,
|
|
"logps/ref_chosen": -85.63675689697266,
|
|
"logps/ref_rejected": -90.28547668457031,
|
|
"logps/rejected": -332.9254150390625,
|
|
"loss": 1.0747,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -0.6510626077651978,
|
|
"rewards/margins": 0.4093397259712219,
|
|
"rewards/rejected": -1.0604023933410645,
|
|
"step": 585
|
|
},
|
|
{
|
|
"epoch": 0.8604992657856094,
|
|
"grad_norm": 7.319284439086914,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.0043752435594797134,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 2.974695142855388e-08,
|
|
"logits/chosen": -7.4545087814331055,
|
|
"logits/rejected": -6.863378524780273,
|
|
"logps/chosen": -246.67648315429688,
|
|
"logps/ref_chosen": -83.68115234375,
|
|
"logps/ref_rejected": -98.35916137695312,
|
|
"logps/rejected": -325.6482238769531,
|
|
"loss": 1.1595,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.7111270427703857,
|
|
"rewards/margins": 0.2777676582336426,
|
|
"rewards/rejected": -0.9888947606086731,
|
|
"step": 586
|
|
},
|
|
{
|
|
"epoch": 0.8619676945668135,
|
|
"grad_norm": 6.929393291473389,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.004353476222604513,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 2.9142720719793122e-08,
|
|
"logits/chosen": -7.802062511444092,
|
|
"logits/rejected": -6.934123992919922,
|
|
"logps/chosen": -210.05108642578125,
|
|
"logps/ref_chosen": -77.5775146484375,
|
|
"logps/ref_rejected": -89.3203353881836,
|
|
"logps/rejected": -292.1748352050781,
|
|
"loss": 1.1418,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.5736163258552551,
|
|
"rewards/margins": 0.30252429842948914,
|
|
"rewards/rejected": -0.8761405944824219,
|
|
"step": 587
|
|
},
|
|
{
|
|
"epoch": 0.8634361233480177,
|
|
"grad_norm": 7.931462287902832,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.00432643573731184,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 2.8544309805740018e-08,
|
|
"logits/chosen": -7.959348678588867,
|
|
"logits/rejected": -7.665192604064941,
|
|
"logps/chosen": -227.4091796875,
|
|
"logps/ref_chosen": -83.80426025390625,
|
|
"logps/ref_rejected": -113.11943054199219,
|
|
"logps/rejected": -323.88861083984375,
|
|
"loss": 1.168,
|
|
"rewards/accuracies": 0.703125,
|
|
"rewards/chosen": -0.6190832257270813,
|
|
"rewards/margins": 0.28650587797164917,
|
|
"rewards/rejected": -0.9055891036987305,
|
|
"step": 588
|
|
},
|
|
{
|
|
"epoch": 0.8649045521292217,
|
|
"grad_norm": 6.9884934425354,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.004302235785871744,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 2.7951734455078786e-08,
|
|
"logits/chosen": -7.721344947814941,
|
|
"logits/rejected": -6.941378593444824,
|
|
"logps/chosen": -250.2454376220703,
|
|
"logps/ref_chosen": -96.46501922607422,
|
|
"logps/ref_rejected": -116.03719329833984,
|
|
"logps/rejected": -356.3966064453125,
|
|
"loss": 1.0798,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -0.65859454870224,
|
|
"rewards/margins": 0.3684034049510956,
|
|
"rewards/rejected": -1.0269978046417236,
|
|
"step": 589
|
|
},
|
|
{
|
|
"epoch": 0.8663729809104258,
|
|
"grad_norm": 8.866425514221191,
|
|
"kl/avg_steps": 0.75,
|
|
"kl/beta": 0.0042755138128995895,
|
|
"kl/n_epsilon_steps": 0.125,
|
|
"kl/p_epsilon_steps": 0.875,
|
|
"learning_rate": 2.736501028272095e-08,
|
|
"logits/chosen": -7.883823394775391,
|
|
"logits/rejected": -7.474784851074219,
|
|
"logps/chosen": -230.829345703125,
|
|
"logps/ref_chosen": -88.4278564453125,
|
|
"logps/ref_rejected": -110.19876861572266,
|
|
"logps/rejected": -336.062744140625,
|
|
"loss": 1.0942,
|
|
"rewards/accuracies": 0.890625,
|
|
"rewards/chosen": -0.6050465106964111,
|
|
"rewards/margins": 0.3528992533683777,
|
|
"rewards/rejected": -0.957945704460144,
|
|
"step": 590
|
|
},
|
|
{
|
|
"epoch": 0.8678414096916299,
|
|
"grad_norm": 5.826635360717773,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.004243686329573393,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 2.678415274939408e-08,
|
|
"logits/chosen": -7.728172302246094,
|
|
"logits/rejected": -6.94543981552124,
|
|
"logps/chosen": -224.34735107421875,
|
|
"logps/ref_chosen": -87.84968566894531,
|
|
"logps/ref_rejected": -89.57919311523438,
|
|
"logps/rejected": -317.07794189453125,
|
|
"loss": 1.0725,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.577212929725647,
|
|
"rewards/margins": 0.3819907307624817,
|
|
"rewards/rejected": -0.9592036604881287,
|
|
"step": 591
|
|
},
|
|
{
|
|
"epoch": 0.869309838472834,
|
|
"grad_norm": 7.418654441833496,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.004218637943267822,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 2.6209177161234442e-08,
|
|
"logits/chosen": -7.696051597595215,
|
|
"logits/rejected": -7.094331741333008,
|
|
"logps/chosen": -235.08880615234375,
|
|
"logps/ref_chosen": -74.97573852539062,
|
|
"logps/ref_rejected": -81.55262756347656,
|
|
"logps/rejected": -313.0411376953125,
|
|
"loss": 1.1533,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -0.6733924150466919,
|
|
"rewards/margins": 0.2968708872795105,
|
|
"rewards/rejected": -0.9702633619308472,
|
|
"step": 592
|
|
},
|
|
{
|
|
"epoch": 0.8707782672540382,
|
|
"grad_norm": 10.643902778625488,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.0041963448747992516,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 2.564009866938349e-08,
|
|
"logits/chosen": -7.23436164855957,
|
|
"logits/rejected": -6.888458251953125,
|
|
"logps/chosen": -221.35968017578125,
|
|
"logps/ref_chosen": -72.23902893066406,
|
|
"logps/ref_rejected": -70.09652709960938,
|
|
"logps/rejected": -280.5572509765625,
|
|
"loss": 1.1869,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -0.6236478090286255,
|
|
"rewards/margins": 0.2537292242050171,
|
|
"rewards/rejected": -0.8773769736289978,
|
|
"step": 593
|
|
},
|
|
{
|
|
"epoch": 0.8722466960352423,
|
|
"grad_norm": 6.839757442474365,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.004175467416644096,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 2.5076932269588708e-08,
|
|
"logits/chosen": -7.658824920654297,
|
|
"logits/rejected": -7.013692378997803,
|
|
"logps/chosen": -226.31228637695312,
|
|
"logps/ref_chosen": -90.43771362304688,
|
|
"logps/ref_rejected": -92.267578125,
|
|
"logps/rejected": -318.3495788574219,
|
|
"loss": 1.0774,
|
|
"rewards/accuracies": 0.890625,
|
|
"rewards/chosen": -0.5646122694015503,
|
|
"rewards/margins": 0.37262603640556335,
|
|
"rewards/rejected": -0.937238335609436,
|
|
"step": 594
|
|
},
|
|
{
|
|
"epoch": 0.8737151248164464,
|
|
"grad_norm": 7.265661239624023,
|
|
"kl/avg_steps": 0.40625,
|
|
"kl/beta": 0.004149532876908779,
|
|
"kl/n_epsilon_steps": 0.296875,
|
|
"kl/p_epsilon_steps": 0.703125,
|
|
"learning_rate": 2.451969280180849e-08,
|
|
"logits/chosen": -7.426346778869629,
|
|
"logits/rejected": -6.822681427001953,
|
|
"logps/chosen": -224.98826599121094,
|
|
"logps/ref_chosen": -81.51480865478516,
|
|
"logps/ref_rejected": -86.52047729492188,
|
|
"logps/rejected": -306.0340576171875,
|
|
"loss": 1.1373,
|
|
"rewards/accuracies": 0.71875,
|
|
"rewards/chosen": -0.5944833159446716,
|
|
"rewards/margins": 0.31153303384780884,
|
|
"rewards/rejected": -0.9060162901878357,
|
|
"step": 595
|
|
},
|
|
{
|
|
"epoch": 0.8751835535976505,
|
|
"grad_norm": 7.548264503479004,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.004132743459194899,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 2.396839494982103e-08,
|
|
"logits/chosen": -7.338626384735107,
|
|
"logits/rejected": -6.9880876541137695,
|
|
"logps/chosen": -234.91661071777344,
|
|
"logps/ref_chosen": -88.20805358886719,
|
|
"logps/ref_rejected": -86.09375,
|
|
"logps/rejected": -299.5952453613281,
|
|
"loss": 1.1748,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -0.6046238541603088,
|
|
"rewards/margins": 0.27212288975715637,
|
|
"rewards/rejected": -0.8767467737197876,
|
|
"step": 596
|
|
},
|
|
{
|
|
"epoch": 0.8766519823788547,
|
|
"grad_norm": 6.239217758178711,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.004110904410481453,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 2.3423053240837514e-08,
|
|
"logits/chosen": -7.8197503089904785,
|
|
"logits/rejected": -7.281065940856934,
|
|
"logps/chosen": -227.93858337402344,
|
|
"logps/ref_chosen": -83.58998107910156,
|
|
"logps/ref_rejected": -106.94439697265625,
|
|
"logps/rejected": -340.60699462890625,
|
|
"loss": 1.0937,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.5899473428726196,
|
|
"rewards/margins": 0.36288982629776,
|
|
"rewards/rejected": -0.9528372287750244,
|
|
"step": 597
|
|
},
|
|
{
|
|
"epoch": 0.8781204111600588,
|
|
"grad_norm": 6.588911533355713,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.004081568215042353,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 2.2883682045119062e-08,
|
|
"logits/chosen": -7.641655921936035,
|
|
"logits/rejected": -7.323144912719727,
|
|
"logps/chosen": -227.22210693359375,
|
|
"logps/ref_chosen": -90.18262481689453,
|
|
"logps/ref_rejected": -94.7985610961914,
|
|
"logps/rejected": -304.92047119140625,
|
|
"loss": 1.1493,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -0.557396411895752,
|
|
"rewards/margins": 0.2942042946815491,
|
|
"rewards/rejected": -0.8516007661819458,
|
|
"step": 598
|
|
},
|
|
{
|
|
"epoch": 0.8795888399412628,
|
|
"grad_norm": 6.382659435272217,
|
|
"kl/avg_steps": 0.375,
|
|
"kl/beta": 0.004058737773448229,
|
|
"kl/n_epsilon_steps": 0.3125,
|
|
"kl/p_epsilon_steps": 0.6875,
|
|
"learning_rate": 2.2350295575598367e-08,
|
|
"logits/chosen": -7.610280990600586,
|
|
"logits/rejected": -7.40827751159668,
|
|
"logps/chosen": -218.54464721679688,
|
|
"logps/ref_chosen": -77.84227752685547,
|
|
"logps/ref_rejected": -89.1976318359375,
|
|
"logps/rejected": -296.90069580078125,
|
|
"loss": 1.1718,
|
|
"rewards/accuracies": 0.703125,
|
|
"rewards/chosen": -0.570677638053894,
|
|
"rewards/margins": 0.2681824266910553,
|
|
"rewards/rejected": -0.838860034942627,
|
|
"step": 599
|
|
},
|
|
{
|
|
"epoch": 0.8810572687224669,
|
|
"grad_norm": 5.930601119995117,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.004043574444949627,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 2.1822907887504932e-08,
|
|
"logits/chosen": -7.736866474151611,
|
|
"logits/rejected": -7.516294956207275,
|
|
"logps/chosen": -242.80270385742188,
|
|
"logps/ref_chosen": -92.18781280517578,
|
|
"logps/ref_rejected": -92.95429992675781,
|
|
"logps/rejected": -322.2016906738281,
|
|
"loss": 1.1296,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.6067217588424683,
|
|
"rewards/margins": 0.31417691707611084,
|
|
"rewards/rejected": -0.9208986759185791,
|
|
"step": 600
|
|
},
|
|
{
|
|
"epoch": 0.8810572687224669,
|
|
"eval_kl/n_epsilon_steps": 0.252996563911438,
|
|
"eval_kl/p_epsilon_steps": 0.7465753555297852,
|
|
"eval_logits/chosen": -7.617927551269531,
|
|
"eval_logits/rejected": -7.201268672943115,
|
|
"eval_logps/chosen": -250.90887451171875,
|
|
"eval_logps/ref_chosen": -100.49356842041016,
|
|
"eval_logps/ref_rejected": -94.06775665283203,
|
|
"eval_logps/rejected": -312.31573486328125,
|
|
"eval_loss": 0.5845144987106323,
|
|
"eval_rewards/accuracies": 0.7568492889404297,
|
|
"eval_rewards/chosen": -0.6029341220855713,
|
|
"eval_rewards/margins": 0.2691001296043396,
|
|
"eval_rewards/rejected": -0.8720343112945557,
|
|
"eval_runtime": 48.5484,
|
|
"eval_samples_per_second": 48.179,
|
|
"eval_steps_per_second": 1.524,
|
|
"step": 600
|
|
},
|
|
{
|
|
"epoch": 0.882525697503671,
|
|
"grad_norm": 6.323831081390381,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.004020956344902515,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 2.1301532877994742e-08,
|
|
"logits/chosen": -7.606431007385254,
|
|
"logits/rejected": -7.300946235656738,
|
|
"logps/chosen": -244.5936279296875,
|
|
"logps/ref_chosen": -89.21614074707031,
|
|
"logps/ref_rejected": -100.17054748535156,
|
|
"logps/rejected": -336.119873046875,
|
|
"loss": 1.1196,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.6223502159118652,
|
|
"rewards/margins": 0.32026663422584534,
|
|
"rewards/rejected": -0.9426168203353882,
|
|
"step": 601
|
|
},
|
|
{
|
|
"epoch": 0.8839941262848752,
|
|
"grad_norm": 7.207840919494629,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.003997222986072302,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 2.0786184285784298e-08,
|
|
"logits/chosen": -7.25759744644165,
|
|
"logits/rejected": -7.028449058532715,
|
|
"logps/chosen": -202.654296875,
|
|
"logps/ref_chosen": -80.05760192871094,
|
|
"logps/ref_rejected": -93.197509765625,
|
|
"logps/rejected": -315.4969482421875,
|
|
"loss": 1.0711,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.48817095160484314,
|
|
"rewards/margins": 0.3939117193222046,
|
|
"rewards/rejected": -0.8820826411247253,
|
|
"step": 602
|
|
},
|
|
{
|
|
"epoch": 0.8854625550660793,
|
|
"grad_norm": 6.062160968780518,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.003971162252128124,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 2.0276875690788204e-08,
|
|
"logits/chosen": -7.96243953704834,
|
|
"logits/rejected": -7.082343101501465,
|
|
"logps/chosen": -235.39010620117188,
|
|
"logps/ref_chosen": -102.30957794189453,
|
|
"logps/ref_rejected": -108.06884765625,
|
|
"logps/rejected": -326.241455078125,
|
|
"loss": 1.1139,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -0.526463508605957,
|
|
"rewards/margins": 0.33385905623435974,
|
|
"rewards/rejected": -0.8603225946426392,
|
|
"step": 603
|
|
},
|
|
{
|
|
"epoch": 0.8869309838472834,
|
|
"grad_norm": 6.03358268737793,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.0039489492774009705,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 1.977362051376158e-08,
|
|
"logits/chosen": -7.427624702453613,
|
|
"logits/rejected": -7.069514274597168,
|
|
"logps/chosen": -216.16348266601562,
|
|
"logps/ref_chosen": -78.17408752441406,
|
|
"logps/ref_rejected": -99.4961166381836,
|
|
"logps/rejected": -323.61181640625,
|
|
"loss": 1.1058,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.5420645475387573,
|
|
"rewards/margins": 0.33607620000839233,
|
|
"rewards/rejected": -0.8781407475471497,
|
|
"step": 604
|
|
},
|
|
{
|
|
"epoch": 0.8883994126284875,
|
|
"grad_norm": 6.287881374359131,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.003921985626220703,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 1.9276432015946446e-08,
|
|
"logits/chosen": -7.781911373138428,
|
|
"logits/rejected": -7.073343276977539,
|
|
"logps/chosen": -242.73826599121094,
|
|
"logps/ref_chosen": -94.77333068847656,
|
|
"logps/ref_rejected": -107.30490112304688,
|
|
"logps/rejected": -327.88031005859375,
|
|
"loss": 1.1539,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -0.5785641074180603,
|
|
"rewards/margins": 0.2811967134475708,
|
|
"rewards/rejected": -0.8597608208656311,
|
|
"step": 605
|
|
},
|
|
{
|
|
"epoch": 0.8898678414096917,
|
|
"grad_norm": 6.576826572418213,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.003903687233105302,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 1.8785323298722093e-08,
|
|
"logits/chosen": -7.757646083831787,
|
|
"logits/rejected": -6.858328342437744,
|
|
"logps/chosen": -232.45693969726562,
|
|
"logps/ref_chosen": -87.7533950805664,
|
|
"logps/ref_rejected": -104.2422103881836,
|
|
"logps/rejected": -326.3280944824219,
|
|
"loss": 1.1412,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.5615659952163696,
|
|
"rewards/margins": 0.298251748085022,
|
|
"rewards/rejected": -0.8598177433013916,
|
|
"step": 606
|
|
},
|
|
{
|
|
"epoch": 0.8913362701908958,
|
|
"grad_norm": 6.907278060913086,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.0038782362826168537,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 1.8300307303259904e-08,
|
|
"logits/chosen": -7.828334808349609,
|
|
"logits/rejected": -6.711248397827148,
|
|
"logps/chosen": -229.9647216796875,
|
|
"logps/ref_chosen": -88.32904815673828,
|
|
"logps/ref_rejected": -86.76811218261719,
|
|
"logps/rejected": -314.5696716308594,
|
|
"loss": 1.1314,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -0.5482698082923889,
|
|
"rewards/margins": 0.3295517861843109,
|
|
"rewards/rejected": -0.8778215646743774,
|
|
"step": 607
|
|
},
|
|
{
|
|
"epoch": 0.8928046989720999,
|
|
"grad_norm": 7.614426612854004,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.0038589416071772575,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 1.7821396810182437e-08,
|
|
"logits/chosen": -7.646144866943359,
|
|
"logits/rejected": -7.090588569641113,
|
|
"logps/chosen": -215.21920776367188,
|
|
"logps/ref_chosen": -85.76937103271484,
|
|
"logps/ref_rejected": -100.23281860351562,
|
|
"logps/rejected": -319.41949462890625,
|
|
"loss": 1.1084,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.4974094033241272,
|
|
"rewards/margins": 0.34195607900619507,
|
|
"rewards/rejected": -0.8393654823303223,
|
|
"step": 608
|
|
},
|
|
{
|
|
"epoch": 0.8942731277533039,
|
|
"grad_norm": 6.746026992797852,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.0038337823934853077,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 1.7348604439226617e-08,
|
|
"logits/chosen": -7.802424907684326,
|
|
"logits/rejected": -7.451755523681641,
|
|
"logps/chosen": -232.28611755371094,
|
|
"logps/ref_chosen": -92.96656799316406,
|
|
"logps/ref_rejected": -95.91818237304688,
|
|
"logps/rejected": -322.77618408203125,
|
|
"loss": 1.1163,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.531781792640686,
|
|
"rewards/margins": 0.3316161632537842,
|
|
"rewards/rejected": -0.8633979558944702,
|
|
"step": 609
|
|
},
|
|
{
|
|
"epoch": 0.895741556534508,
|
|
"grad_norm": 6.536015510559082,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.0038099701050668955,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 1.6881942648911074e-08,
|
|
"logits/chosen": -7.566076278686523,
|
|
"logits/rejected": -7.088051795959473,
|
|
"logps/chosen": -228.5191650390625,
|
|
"logps/ref_chosen": -94.70028686523438,
|
|
"logps/ref_rejected": -89.68739318847656,
|
|
"logps/rejected": -314.64813232421875,
|
|
"loss": 1.1011,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -0.5073626637458801,
|
|
"rewards/margins": 0.3432729244232178,
|
|
"rewards/rejected": -0.8506356477737427,
|
|
"step": 610
|
|
},
|
|
{
|
|
"epoch": 0.8972099853157122,
|
|
"grad_norm": 7.131064414978027,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.003785130102187395,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 1.6421423736208e-08,
|
|
"logits/chosen": -7.605221748352051,
|
|
"logits/rejected": -7.5947184562683105,
|
|
"logps/chosen": -224.7576446533203,
|
|
"logps/ref_chosen": -86.78334045410156,
|
|
"logps/ref_rejected": -89.84307861328125,
|
|
"logps/rejected": -316.98431396484375,
|
|
"loss": 1.1105,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.52030348777771,
|
|
"rewards/margins": 0.3337860703468323,
|
|
"rewards/rejected": -0.8540895581245422,
|
|
"step": 611
|
|
},
|
|
{
|
|
"epoch": 0.8986784140969163,
|
|
"grad_norm": 6.337128639221191,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.003763957880437374,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 1.5967059836219042e-08,
|
|
"logits/chosen": -7.775234699249268,
|
|
"logits/rejected": -6.849664688110352,
|
|
"logps/chosen": -251.38430786132812,
|
|
"logps/ref_chosen": -101.02015686035156,
|
|
"logps/ref_rejected": -93.78302764892578,
|
|
"logps/rejected": -335.0631103515625,
|
|
"loss": 1.1023,
|
|
"rewards/accuracies": 0.90625,
|
|
"rewards/chosen": -0.5628792643547058,
|
|
"rewards/margins": 0.33847683668136597,
|
|
"rewards/rejected": -0.9013561010360718,
|
|
"step": 612
|
|
},
|
|
{
|
|
"epoch": 0.9001468428781204,
|
|
"grad_norm": 6.324493885040283,
|
|
"kl/avg_steps": 0.90625,
|
|
"kl/beta": 0.003738257335498929,
|
|
"kl/n_epsilon_steps": 0.046875,
|
|
"kl/p_epsilon_steps": 0.953125,
|
|
"learning_rate": 1.551886292185553e-08,
|
|
"logits/chosen": -8.078614234924316,
|
|
"logits/rejected": -7.592404365539551,
|
|
"logps/chosen": -213.39401245117188,
|
|
"logps/ref_chosen": -88.9886245727539,
|
|
"logps/ref_rejected": -109.99551391601562,
|
|
"logps/rejected": -333.4338073730469,
|
|
"loss": 1.0762,
|
|
"rewards/accuracies": 0.921875,
|
|
"rewards/chosen": -0.46113917231559753,
|
|
"rewards/margins": 0.3663785457611084,
|
|
"rewards/rejected": -0.8275177478790283,
|
|
"step": 613
|
|
},
|
|
{
|
|
"epoch": 0.9016152716593245,
|
|
"grad_norm": 7.8051652908325195,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.0037046836223453283,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 1.507684480352292e-08,
|
|
"logits/chosen": -7.316326141357422,
|
|
"logits/rejected": -7.080374240875244,
|
|
"logps/chosen": -230.80239868164062,
|
|
"logps/ref_chosen": -80.20005798339844,
|
|
"logps/ref_rejected": -109.86239624023438,
|
|
"logps/rejected": -340.70416259765625,
|
|
"loss": 1.1461,
|
|
"rewards/accuracies": 0.734375,
|
|
"rewards/chosen": -0.5560883283615112,
|
|
"rewards/margins": 0.29344964027404785,
|
|
"rewards/rejected": -0.8495379686355591,
|
|
"step": 614
|
|
},
|
|
{
|
|
"epoch": 0.9030837004405287,
|
|
"grad_norm": 6.5305585861206055,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.0036839614622294903,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 1.4641017128809801e-08,
|
|
"logits/chosen": -7.641972541809082,
|
|
"logits/rejected": -6.816801071166992,
|
|
"logps/chosen": -243.8656463623047,
|
|
"logps/ref_chosen": -100.43526458740234,
|
|
"logps/ref_rejected": -101.1800537109375,
|
|
"logps/rejected": -315.3383483886719,
|
|
"loss": 1.1669,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.525856077671051,
|
|
"rewards/margins": 0.2574092149734497,
|
|
"rewards/rejected": -0.7832653522491455,
|
|
"step": 615
|
|
},
|
|
{
|
|
"epoch": 0.9045521292217328,
|
|
"grad_norm": 5.881459712982178,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.003662216942757368,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 1.4211391382180637e-08,
|
|
"logits/chosen": -7.507961273193359,
|
|
"logits/rejected": -7.136435508728027,
|
|
"logps/chosen": -236.98391723632812,
|
|
"logps/ref_chosen": -92.49292755126953,
|
|
"logps/ref_rejected": -82.06065368652344,
|
|
"logps/rejected": -318.6792297363281,
|
|
"loss": 1.1124,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.5267339944839478,
|
|
"rewards/margins": 0.3333369195461273,
|
|
"rewards/rejected": -0.8600709438323975,
|
|
"step": 616
|
|
},
|
|
{
|
|
"epoch": 0.9060205580029369,
|
|
"grad_norm": 6.169250011444092,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.0036394703201949596,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 1.378797888467345e-08,
|
|
"logits/chosen": -7.898214817047119,
|
|
"logits/rejected": -7.082752227783203,
|
|
"logps/chosen": -241.96456909179688,
|
|
"logps/ref_chosen": -91.09699249267578,
|
|
"logps/ref_rejected": -70.41004943847656,
|
|
"logps/rejected": -294.9502868652344,
|
|
"loss": 1.1664,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -0.5471093654632568,
|
|
"rewards/margins": 0.26464053988456726,
|
|
"rewards/rejected": -0.8117498755455017,
|
|
"step": 617
|
|
},
|
|
{
|
|
"epoch": 0.9074889867841409,
|
|
"grad_norm": 7.049670696258545,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.003617988433688879,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 1.3370790793601371e-08,
|
|
"logits/chosen": -7.613478660583496,
|
|
"logits/rejected": -7.167911052703857,
|
|
"logps/chosen": -252.33189392089844,
|
|
"logps/ref_chosen": -102.02059936523438,
|
|
"logps/ref_rejected": -99.80119323730469,
|
|
"logps/rejected": -331.26397705078125,
|
|
"loss": 1.1462,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.5415487289428711,
|
|
"rewards/margins": 0.2898721694946289,
|
|
"rewards/rejected": -0.8314208388328552,
|
|
"step": 618
|
|
},
|
|
{
|
|
"epoch": 0.908957415565345,
|
|
"grad_norm": 6.142784118652344,
|
|
"kl/avg_steps": 0.375,
|
|
"kl/beta": 0.003597751259803772,
|
|
"kl/n_epsilon_steps": 0.3125,
|
|
"kl/p_epsilon_steps": 0.6875,
|
|
"learning_rate": 1.2959838102258535e-08,
|
|
"logits/chosen": -7.600045204162598,
|
|
"logits/rejected": -7.366483688354492,
|
|
"logps/chosen": -234.5894775390625,
|
|
"logps/ref_chosen": -89.74136352539062,
|
|
"logps/ref_rejected": -99.90138244628906,
|
|
"logps/rejected": -319.714111328125,
|
|
"loss": 1.1708,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -0.5201643705368042,
|
|
"rewards/margins": 0.26622116565704346,
|
|
"rewards/rejected": -0.7863855361938477,
|
|
"step": 619
|
|
},
|
|
{
|
|
"epoch": 0.9104258443465492,
|
|
"grad_norm": 5.838797092437744,
|
|
"kl/avg_steps": 0.40625,
|
|
"kl/beta": 0.003584309946745634,
|
|
"kl/n_epsilon_steps": 0.296875,
|
|
"kl/p_epsilon_steps": 0.703125,
|
|
"learning_rate": 1.2555131639630567e-08,
|
|
"logits/chosen": -8.057722091674805,
|
|
"logits/rejected": -7.399764060974121,
|
|
"logps/chosen": -225.7647247314453,
|
|
"logps/ref_chosen": -85.12431335449219,
|
|
"logps/ref_rejected": -85.41253662109375,
|
|
"logps/rejected": -307.8377380371094,
|
|
"loss": 1.1526,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -0.5036029815673828,
|
|
"rewards/margins": 0.28931373357772827,
|
|
"rewards/rejected": -0.7929166555404663,
|
|
"step": 620
|
|
},
|
|
{
|
|
"epoch": 0.9118942731277533,
|
|
"grad_norm": 5.572709083557129,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.0035698076244443655,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 1.2156682070109086e-08,
|
|
"logits/chosen": -7.758251190185547,
|
|
"logits/rejected": -7.50982666015625,
|
|
"logps/chosen": -229.43209838867188,
|
|
"logps/ref_chosen": -89.24842071533203,
|
|
"logps/ref_rejected": -95.46463775634766,
|
|
"logps/rejected": -315.90655517578125,
|
|
"loss": 1.1604,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -0.49918726086616516,
|
|
"rewards/margins": 0.28264692425727844,
|
|
"rewards/rejected": -0.7818341851234436,
|
|
"step": 621
|
|
},
|
|
{
|
|
"epoch": 0.9133627019089574,
|
|
"grad_norm": 6.386410713195801,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.0035531523171812296,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 1.1764499893210878e-08,
|
|
"logits/chosen": -7.883831977844238,
|
|
"logits/rejected": -6.893963813781738,
|
|
"logps/chosen": -242.18194580078125,
|
|
"logps/ref_chosen": -99.79413604736328,
|
|
"logps/ref_rejected": -90.82821655273438,
|
|
"logps/rejected": -317.67083740234375,
|
|
"loss": 1.1355,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.5042383670806885,
|
|
"rewards/margins": 0.29660987854003906,
|
|
"rewards/rejected": -0.8008482456207275,
|
|
"step": 622
|
|
},
|
|
{
|
|
"epoch": 0.9148311306901615,
|
|
"grad_norm": 7.438137531280518,
|
|
"kl/avg_steps": 0.40625,
|
|
"kl/beta": 0.0035332776606082916,
|
|
"kl/n_epsilon_steps": 0.296875,
|
|
"kl/p_epsilon_steps": 0.703125,
|
|
"learning_rate": 1.1378595443300998e-08,
|
|
"logits/chosen": -7.771790504455566,
|
|
"logits/rejected": -7.093353748321533,
|
|
"logps/chosen": -236.72573852539062,
|
|
"logps/ref_chosen": -90.45555114746094,
|
|
"logps/ref_rejected": -91.32276916503906,
|
|
"logps/rejected": -297.7119445800781,
|
|
"loss": 1.2219,
|
|
"rewards/accuracies": 0.703125,
|
|
"rewards/chosen": -0.5164909958839417,
|
|
"rewards/margins": 0.20907816290855408,
|
|
"rewards/rejected": -0.7255691289901733,
|
|
"step": 623
|
|
},
|
|
{
|
|
"epoch": 0.9162995594713657,
|
|
"grad_norm": 6.896797180175781,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.0035189816262573004,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 1.0998978889320582e-08,
|
|
"logits/chosen": -8.146064758300781,
|
|
"logits/rejected": -7.289480209350586,
|
|
"logps/chosen": -258.7445373535156,
|
|
"logps/ref_chosen": -109.87522888183594,
|
|
"logps/ref_rejected": -104.77320861816406,
|
|
"logps/rejected": -344.84759521484375,
|
|
"loss": 1.1213,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.5218902826309204,
|
|
"rewards/margins": 0.3172043263912201,
|
|
"rewards/rejected": -0.8390946388244629,
|
|
"step": 624
|
|
},
|
|
{
|
|
"epoch": 0.9177679882525698,
|
|
"grad_norm": 5.683692455291748,
|
|
"kl/avg_steps": 0.5,
|
|
"kl/beta": 0.0034992981236428022,
|
|
"kl/n_epsilon_steps": 0.25,
|
|
"kl/p_epsilon_steps": 0.75,
|
|
"learning_rate": 1.0625660234518913e-08,
|
|
"logits/chosen": -7.770914554595947,
|
|
"logits/rejected": -7.289626121520996,
|
|
"logps/chosen": -243.88046264648438,
|
|
"logps/ref_chosen": -87.16815948486328,
|
|
"logps/ref_rejected": -91.86148071289062,
|
|
"logps/rejected": -318.0606994628906,
|
|
"loss": 1.1839,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -0.5464307069778442,
|
|
"rewards/margins": 0.24018320441246033,
|
|
"rewards/rejected": -0.786613941192627,
|
|
"step": 625
|
|
},
|
|
{
|
|
"epoch": 0.9192364170337739,
|
|
"grad_norm": 7.084156513214111,
|
|
"kl/avg_steps": 0.40625,
|
|
"kl/beta": 0.0034818886779248714,
|
|
"kl/n_epsilon_steps": 0.296875,
|
|
"kl/p_epsilon_steps": 0.703125,
|
|
"learning_rate": 1.0258649316189721e-08,
|
|
"logits/chosen": -7.743939399719238,
|
|
"logits/rejected": -7.482439994812012,
|
|
"logps/chosen": -254.01075744628906,
|
|
"logps/ref_chosen": -104.22421264648438,
|
|
"logps/ref_rejected": -104.1774673461914,
|
|
"logps/rejected": -321.1068115234375,
|
|
"loss": 1.2005,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -0.5199881792068481,
|
|
"rewards/margins": 0.23050320148468018,
|
|
"rewards/rejected": -0.7504914402961731,
|
|
"step": 626
|
|
},
|
|
{
|
|
"epoch": 0.920704845814978,
|
|
"grad_norm": 5.33841609954834,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.0034678007941693068,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 9.897955805412e-09,
|
|
"logits/chosen": -7.591939926147461,
|
|
"logits/rejected": -7.3537187576293945,
|
|
"logps/chosen": -208.89431762695312,
|
|
"logps/ref_chosen": -74.93461608886719,
|
|
"logps/ref_rejected": -112.57289123535156,
|
|
"logps/rejected": -330.51983642578125,
|
|
"loss": 1.1507,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.46203625202178955,
|
|
"rewards/margins": 0.2876952886581421,
|
|
"rewards/rejected": -0.7497316002845764,
|
|
"step": 627
|
|
},
|
|
{
|
|
"epoch": 0.922173274596182,
|
|
"grad_norm": 6.301426410675049,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.0034462616313248873,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 9.543589206795238e-09,
|
|
"logits/chosen": -7.962158679962158,
|
|
"logits/rejected": -7.421542644500732,
|
|
"logps/chosen": -231.39364624023438,
|
|
"logps/ref_chosen": -93.69107818603516,
|
|
"logps/ref_rejected": -107.34395599365234,
|
|
"logps/rejected": -335.9172668457031,
|
|
"loss": 1.1251,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -0.4717589020729065,
|
|
"rewards/margins": 0.30952274799346924,
|
|
"rewards/rejected": -0.7812816500663757,
|
|
"step": 628
|
|
},
|
|
{
|
|
"epoch": 0.9236417033773862,
|
|
"grad_norm": 5.8255295753479,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.0034227303694933653,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 9.19555885822887e-09,
|
|
"logits/chosen": -7.680126190185547,
|
|
"logits/rejected": -7.379844665527344,
|
|
"logps/chosen": -253.8419952392578,
|
|
"logps/ref_chosen": -103.23037719726562,
|
|
"logps/ref_rejected": -97.16841888427734,
|
|
"logps/rejected": -326.94293212890625,
|
|
"loss": 1.1605,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.5139710903167725,
|
|
"rewards/margins": 0.2675333023071289,
|
|
"rewards/rejected": -0.7815043926239014,
|
|
"step": 629
|
|
},
|
|
{
|
|
"epoch": 0.9251101321585903,
|
|
"grad_norm": 7.221403121948242,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.003404643153771758,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 8.85387393063622e-09,
|
|
"logits/chosen": -7.831258773803711,
|
|
"logits/rejected": -7.15199089050293,
|
|
"logps/chosen": -232.68309020996094,
|
|
"logps/ref_chosen": -93.89755249023438,
|
|
"logps/ref_rejected": -89.3743896484375,
|
|
"logps/rejected": -283.4253234863281,
|
|
"loss": 1.2339,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -0.4714937210083008,
|
|
"rewards/margins": 0.1851492077112198,
|
|
"rewards/rejected": -0.6566429138183594,
|
|
"step": 630
|
|
},
|
|
{
|
|
"epoch": 0.9265785609397944,
|
|
"grad_norm": 5.580920219421387,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.003388758283108473,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 8.518543427732949e-09,
|
|
"logits/chosen": -7.815197944641113,
|
|
"logits/rejected": -7.3377485275268555,
|
|
"logps/chosen": -220.7677764892578,
|
|
"logps/ref_chosen": -87.77082061767578,
|
|
"logps/ref_rejected": -88.68241882324219,
|
|
"logps/rejected": -300.57403564453125,
|
|
"loss": 1.1612,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -0.4492769241333008,
|
|
"rewards/margins": 0.2641494572162628,
|
|
"rewards/rejected": -0.7134263515472412,
|
|
"step": 631
|
|
},
|
|
{
|
|
"epoch": 0.9280469897209985,
|
|
"grad_norm": 6.467636585235596,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.0033708508126437664,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 8.189576185789637e-09,
|
|
"logits/chosen": -7.605412483215332,
|
|
"logits/rejected": -7.2681169509887695,
|
|
"logps/chosen": -221.4307861328125,
|
|
"logps/ref_chosen": -88.62652587890625,
|
|
"logps/ref_rejected": -91.45091247558594,
|
|
"logps/rejected": -310.26220703125,
|
|
"loss": 1.1516,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -0.44607216119766235,
|
|
"rewards/margins": 0.2862231731414795,
|
|
"rewards/rejected": -0.7322953343391418,
|
|
"step": 632
|
|
},
|
|
{
|
|
"epoch": 0.9295154185022027,
|
|
"grad_norm": 6.414995193481445,
|
|
"kl/avg_steps": 0.28125,
|
|
"kl/beta": 0.00335303763858974,
|
|
"kl/n_epsilon_steps": 0.359375,
|
|
"kl/p_epsilon_steps": 0.640625,
|
|
"learning_rate": 7.866980873399015e-09,
|
|
"logits/chosen": -7.696202278137207,
|
|
"logits/rejected": -7.264307498931885,
|
|
"logps/chosen": -247.08786010742188,
|
|
"logps/ref_chosen": -81.37442016601562,
|
|
"logps/ref_rejected": -98.62571716308594,
|
|
"logps/rejected": -316.35614013671875,
|
|
"loss": 1.2589,
|
|
"rewards/accuracies": 0.671875,
|
|
"rewards/chosen": -0.5550001859664917,
|
|
"rewards/margins": 0.17108407616615295,
|
|
"rewards/rejected": -0.7260842323303223,
|
|
"step": 633
|
|
},
|
|
{
|
|
"epoch": 0.9309838472834068,
|
|
"grad_norm": 6.248257637023926,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.0033436338417232037,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 7.550765991247654e-09,
|
|
"logits/chosen": -7.881155490875244,
|
|
"logits/rejected": -7.148515701293945,
|
|
"logps/chosen": -250.07144165039062,
|
|
"logps/ref_chosen": -96.12284851074219,
|
|
"logps/ref_rejected": -112.84780883789062,
|
|
"logps/rejected": -334.539306640625,
|
|
"loss": 1.1935,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -0.5128281712532043,
|
|
"rewards/margins": 0.22365406155586243,
|
|
"rewards/rejected": -0.7364822030067444,
|
|
"step": 634
|
|
},
|
|
{
|
|
"epoch": 0.9324522760646109,
|
|
"grad_norm": 6.869739055633545,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.003325964557006955,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 7.240939871891699e-09,
|
|
"logits/chosen": -7.332803249359131,
|
|
"logits/rejected": -6.858902931213379,
|
|
"logps/chosen": -258.06280517578125,
|
|
"logps/ref_chosen": -98.68411254882812,
|
|
"logps/ref_rejected": -89.8991928100586,
|
|
"logps/rejected": -301.9739990234375,
|
|
"loss": 1.2457,
|
|
"rewards/accuracies": 0.71875,
|
|
"rewards/chosen": -0.5283612012863159,
|
|
"rewards/margins": 0.17254400253295898,
|
|
"rewards/rejected": -0.7009052634239197,
|
|
"step": 635
|
|
},
|
|
{
|
|
"epoch": 0.933920704845815,
|
|
"grad_norm": 5.351519584655762,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.003310446860268712,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 6.937510679537628e-09,
|
|
"logits/chosen": -7.650508880615234,
|
|
"logits/rejected": -6.78857946395874,
|
|
"logps/chosen": -234.77163696289062,
|
|
"logps/ref_chosen": -90.41796112060547,
|
|
"logps/ref_rejected": -87.70687866210938,
|
|
"logps/rejected": -316.6680603027344,
|
|
"loss": 1.1499,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.4755138158798218,
|
|
"rewards/margins": 0.27691105008125305,
|
|
"rewards/rejected": -0.7524248361587524,
|
|
"step": 636
|
|
},
|
|
{
|
|
"epoch": 0.9353891336270191,
|
|
"grad_norm": 5.5197577476501465,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.0032898851204663515,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 6.640486409826785e-09,
|
|
"logits/chosen": -7.380791664123535,
|
|
"logits/rejected": -7.242560386657715,
|
|
"logps/chosen": -227.92446899414062,
|
|
"logps/ref_chosen": -82.44971466064453,
|
|
"logps/ref_rejected": -104.02860260009766,
|
|
"logps/rejected": -339.26043701171875,
|
|
"loss": 1.1395,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -0.4761905372142792,
|
|
"rewards/margins": 0.29185453057289124,
|
|
"rewards/rejected": -0.7680450677871704,
|
|
"step": 637
|
|
},
|
|
{
|
|
"epoch": 0.9368575624082232,
|
|
"grad_norm": 5.237280368804932,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.0032704665791243315,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 6.349874889624962e-09,
|
|
"logits/chosen": -7.648863792419434,
|
|
"logits/rejected": -6.924067497253418,
|
|
"logps/chosen": -226.06134033203125,
|
|
"logps/ref_chosen": -91.92498779296875,
|
|
"logps/ref_rejected": -86.28703308105469,
|
|
"logps/rejected": -310.4959716796875,
|
|
"loss": 1.1318,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.43678027391433716,
|
|
"rewards/margins": 0.2914542555809021,
|
|
"rewards/rejected": -0.7282345294952393,
|
|
"step": 638
|
|
},
|
|
{
|
|
"epoch": 0.9383259911894273,
|
|
"grad_norm": 5.765613079071045,
|
|
"kl/avg_steps": 0.34375,
|
|
"kl/beta": 0.003251162823289633,
|
|
"kl/n_epsilon_steps": 0.328125,
|
|
"kl/p_epsilon_steps": 0.671875,
|
|
"learning_rate": 6.065683776815933e-09,
|
|
"logits/chosen": -7.779665470123291,
|
|
"logits/rejected": -6.977321624755859,
|
|
"logps/chosen": -281.9426574707031,
|
|
"logps/ref_chosen": -104.52755737304688,
|
|
"logps/ref_rejected": -81.4803466796875,
|
|
"logps/rejected": -305.64599609375,
|
|
"loss": 1.2695,
|
|
"rewards/accuracies": 0.71875,
|
|
"rewards/chosen": -0.5761454105377197,
|
|
"rewards/margins": 0.1490968018770218,
|
|
"rewards/rejected": -0.7252421975135803,
|
|
"step": 639
|
|
},
|
|
{
|
|
"epoch": 0.9397944199706314,
|
|
"grad_norm": 5.791350364685059,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.00324002536945045,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 5.7879205600998296e-09,
|
|
"logits/chosen": -7.959763526916504,
|
|
"logits/rejected": -7.383049964904785,
|
|
"logps/chosen": -245.22933959960938,
|
|
"logps/ref_chosen": -97.88526916503906,
|
|
"logps/ref_rejected": -112.70501708984375,
|
|
"logps/rejected": -343.93792724609375,
|
|
"loss": 1.1573,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -0.47469913959503174,
|
|
"rewards/margins": 0.26852816343307495,
|
|
"rewards/rejected": -0.7432272434234619,
|
|
"step": 640
|
|
},
|
|
{
|
|
"epoch": 0.9412628487518355,
|
|
"grad_norm": 5.238857269287109,
|
|
"kl/avg_steps": 0.3125,
|
|
"kl/beta": 0.0032169038895517588,
|
|
"kl/n_epsilon_steps": 0.34375,
|
|
"kl/p_epsilon_steps": 0.65625,
|
|
"learning_rate": 5.516592558795746e-09,
|
|
"logits/chosen": -7.708554267883301,
|
|
"logits/rejected": -7.541186332702637,
|
|
"logps/chosen": -232.91995239257812,
|
|
"logps/ref_chosen": -96.4456787109375,
|
|
"logps/ref_rejected": -95.13568878173828,
|
|
"logps/rejected": -317.35809326171875,
|
|
"loss": 1.1614,
|
|
"rewards/accuracies": 0.71875,
|
|
"rewards/chosen": -0.4384271800518036,
|
|
"rewards/margins": 0.27239149808883667,
|
|
"rewards/rejected": -0.7108187079429626,
|
|
"step": 641
|
|
},
|
|
{
|
|
"epoch": 0.9427312775330396,
|
|
"grad_norm": 5.874085903167725,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.003206882392987609,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 5.251706922648868e-09,
|
|
"logits/chosen": -7.545126914978027,
|
|
"logits/rejected": -7.102571487426758,
|
|
"logps/chosen": -256.21673583984375,
|
|
"logps/ref_chosen": -100.75984954833984,
|
|
"logps/ref_rejected": -114.70763397216797,
|
|
"logps/rejected": -347.91558837890625,
|
|
"loss": 1.1825,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -0.497159868478775,
|
|
"rewards/margins": 0.24616217613220215,
|
|
"rewards/rejected": -0.7433220148086548,
|
|
"step": 642
|
|
},
|
|
{
|
|
"epoch": 0.9441997063142438,
|
|
"grad_norm": 6.670008659362793,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.003189935814589262,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.993270631642038e-09,
|
|
"logits/chosen": -7.875416278839111,
|
|
"logits/rejected": -7.243396759033203,
|
|
"logps/chosen": -229.69638061523438,
|
|
"logps/ref_chosen": -84.74365997314453,
|
|
"logps/ref_rejected": -94.31842041015625,
|
|
"logps/rejected": -315.35858154296875,
|
|
"loss": 1.1822,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.4609188437461853,
|
|
"rewards/margins": 0.2396201640367508,
|
|
"rewards/rejected": -0.7005389928817749,
|
|
"step": 643
|
|
},
|
|
{
|
|
"epoch": 0.9456681350954479,
|
|
"grad_norm": 6.133624076843262,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.0031720928382128477,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 4.741290495811873e-09,
|
|
"logits/chosen": -7.913561820983887,
|
|
"logits/rejected": -7.103449821472168,
|
|
"logps/chosen": -230.85415649414062,
|
|
"logps/ref_chosen": -85.32275390625,
|
|
"logps/ref_rejected": -94.60861206054688,
|
|
"logps/rejected": -304.661865234375,
|
|
"loss": 1.2111,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.45997491478919983,
|
|
"rewards/margins": 0.20199283957481384,
|
|
"rewards/rejected": -0.6619677543640137,
|
|
"step": 644
|
|
},
|
|
{
|
|
"epoch": 0.947136563876652,
|
|
"grad_norm": 5.844942092895508,
|
|
"kl/avg_steps": 0.1875,
|
|
"kl/beta": 0.0031543495133519173,
|
|
"kl/n_epsilon_steps": 0.40625,
|
|
"kl/p_epsilon_steps": 0.59375,
|
|
"learning_rate": 4.495773155069299e-09,
|
|
"logits/chosen": -7.571622371673584,
|
|
"logits/rejected": -7.209532737731934,
|
|
"logps/chosen": -241.77206420898438,
|
|
"logps/ref_chosen": -82.59024047851562,
|
|
"logps/ref_rejected": -103.00375366210938,
|
|
"logps/rejected": -306.9005432128906,
|
|
"loss": 1.2771,
|
|
"rewards/accuracies": 0.625,
|
|
"rewards/chosen": -0.5023009777069092,
|
|
"rewards/margins": 0.13825571537017822,
|
|
"rewards/rejected": -0.6405566930770874,
|
|
"step": 645
|
|
},
|
|
{
|
|
"epoch": 0.9486049926578561,
|
|
"grad_norm": 5.194578647613525,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.0031484460923820734,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 4.256725079024553e-09,
|
|
"logits/chosen": -7.935370922088623,
|
|
"logits/rejected": -7.05921745300293,
|
|
"logps/chosen": -229.54640197753906,
|
|
"logps/ref_chosen": -91.1920394897461,
|
|
"logps/ref_rejected": -83.77833557128906,
|
|
"logps/rejected": -303.084716796875,
|
|
"loss": 1.1663,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.43383896350860596,
|
|
"rewards/margins": 0.2520383596420288,
|
|
"rewards/rejected": -0.6858773231506348,
|
|
"step": 646
|
|
},
|
|
{
|
|
"epoch": 0.9500734214390602,
|
|
"grad_norm": 5.19912576675415,
|
|
"kl/avg_steps": 0.40625,
|
|
"kl/beta": 0.00312986271455884,
|
|
"kl/n_epsilon_steps": 0.296875,
|
|
"kl/p_epsilon_steps": 0.703125,
|
|
"learning_rate": 4.024152566816791e-09,
|
|
"logits/chosen": -7.209627151489258,
|
|
"logits/rejected": -7.173065185546875,
|
|
"logps/chosen": -242.48236083984375,
|
|
"logps/ref_chosen": -88.84446716308594,
|
|
"logps/ref_rejected": -99.49832916259766,
|
|
"logps/rejected": -322.8554382324219,
|
|
"loss": 1.2089,
|
|
"rewards/accuracies": 0.703125,
|
|
"rewards/chosen": -0.47964316606521606,
|
|
"rewards/margins": 0.21512295305728912,
|
|
"rewards/rejected": -0.694766104221344,
|
|
"step": 647
|
|
},
|
|
{
|
|
"epoch": 0.9515418502202643,
|
|
"grad_norm": 5.339207172393799,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.003117199055850506,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 3.798061746947995e-09,
|
|
"logits/chosen": -7.534468173980713,
|
|
"logits/rejected": -7.395105838775635,
|
|
"logps/chosen": -227.42025756835938,
|
|
"logps/ref_chosen": -87.84810638427734,
|
|
"logps/ref_rejected": -104.67005920410156,
|
|
"logps/rejected": -343.5663146972656,
|
|
"loss": 1.127,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -0.4329220652580261,
|
|
"rewards/margins": 0.30612409114837646,
|
|
"rewards/rejected": -0.7390461564064026,
|
|
"step": 648
|
|
},
|
|
{
|
|
"epoch": 0.9530102790014684,
|
|
"grad_norm": 5.596925258636475,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.0030959146097302437,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 3.5784585771215235e-09,
|
|
"logits/chosen": -7.8632001876831055,
|
|
"logits/rejected": -7.287814140319824,
|
|
"logps/chosen": -217.3938446044922,
|
|
"logps/ref_chosen": -89.6925048828125,
|
|
"logps/ref_rejected": -88.70658111572266,
|
|
"logps/rejected": -297.97711181640625,
|
|
"loss": 1.1676,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -0.39361506700515747,
|
|
"rewards/margins": 0.24976572394371033,
|
|
"rewards/rejected": -0.6433808207511902,
|
|
"step": 649
|
|
},
|
|
{
|
|
"epoch": 0.9544787077826725,
|
|
"grad_norm": 5.370362281799316,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.0030785975977778435,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 3.3653488440851253e-09,
|
|
"logits/chosen": -7.673471450805664,
|
|
"logits/rejected": -7.417351722717285,
|
|
"logps/chosen": -241.04432678222656,
|
|
"logps/ref_chosen": -89.93060302734375,
|
|
"logps/ref_rejected": -102.61282348632812,
|
|
"logps/rejected": -332.0096435546875,
|
|
"loss": 1.1869,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.463270366191864,
|
|
"rewards/margins": 0.23782768845558167,
|
|
"rewards/rejected": -0.7010980844497681,
|
|
"step": 650
|
|
},
|
|
{
|
|
"epoch": 0.9559471365638766,
|
|
"grad_norm": 5.453701972961426,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.0030594756826758385,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 3.158738163478475e-09,
|
|
"logits/chosen": -7.594956398010254,
|
|
"logits/rejected": -6.981936454772949,
|
|
"logps/chosen": -208.45730590820312,
|
|
"logps/ref_chosen": -79.18731689453125,
|
|
"logps/ref_rejected": -105.93333435058594,
|
|
"logps/rejected": -328.37841796875,
|
|
"loss": 1.1412,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.3932613134384155,
|
|
"rewards/margins": 0.2819024920463562,
|
|
"rewards/rejected": -0.675163745880127,
|
|
"step": 651
|
|
},
|
|
{
|
|
"epoch": 0.9574155653450808,
|
|
"grad_norm": 4.904377460479736,
|
|
"kl/avg_steps": 0.6875,
|
|
"kl/beta": 0.003039528848603368,
|
|
"kl/n_epsilon_steps": 0.15625,
|
|
"kl/p_epsilon_steps": 0.84375,
|
|
"learning_rate": 2.9586319796851555e-09,
|
|
"logits/chosen": -7.993418216705322,
|
|
"logits/rejected": -7.6948065757751465,
|
|
"logps/chosen": -234.99612426757812,
|
|
"logps/ref_chosen": -101.79022979736328,
|
|
"logps/ref_rejected": -116.3245849609375,
|
|
"logps/rejected": -330.2041015625,
|
|
"loss": 1.1713,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -0.40263664722442627,
|
|
"rewards/margins": 0.24258247017860413,
|
|
"rewards/rejected": -0.6452191472053528,
|
|
"step": 652
|
|
},
|
|
{
|
|
"epoch": 0.9588839941262849,
|
|
"grad_norm": 5.360039234161377,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.003018774790689349,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 2.7650355656892166e-09,
|
|
"logits/chosen": -7.770810127258301,
|
|
"logits/rejected": -7.401587963104248,
|
|
"logps/chosen": -243.21240234375,
|
|
"logps/ref_chosen": -93.35359191894531,
|
|
"logps/ref_rejected": -109.12324523925781,
|
|
"logps/rejected": -335.51324462890625,
|
|
"loss": 1.1885,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.45035725831985474,
|
|
"rewards/margins": 0.22805194556713104,
|
|
"rewards/rejected": -0.678409218788147,
|
|
"step": 653
|
|
},
|
|
{
|
|
"epoch": 0.960352422907489,
|
|
"grad_norm": 7.331425666809082,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.003000956494361162,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 2.577954022936174e-09,
|
|
"logits/chosen": -7.537060260772705,
|
|
"logits/rejected": -7.259842872619629,
|
|
"logps/chosen": -241.98892211914062,
|
|
"logps/ref_chosen": -89.11553955078125,
|
|
"logps/ref_rejected": -104.91995239257812,
|
|
"logps/rejected": -323.5856628417969,
|
|
"loss": 1.2262,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -0.45708924531936646,
|
|
"rewards/margins": 0.1945749968290329,
|
|
"rewards/rejected": -0.6516642570495605,
|
|
"step": 654
|
|
},
|
|
{
|
|
"epoch": 0.9618208516886931,
|
|
"grad_norm": 5.144998550415039,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.0029850981663912535,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 2.397392281198729e-09,
|
|
"logits/chosen": -7.6725006103515625,
|
|
"logits/rejected": -7.398637771606445,
|
|
"logps/chosen": -220.708984375,
|
|
"logps/ref_chosen": -81.03610229492188,
|
|
"logps/ref_rejected": -102.80233764648438,
|
|
"logps/rejected": -322.0316162109375,
|
|
"loss": 1.1856,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.4156780540943146,
|
|
"rewards/margins": 0.23453345894813538,
|
|
"rewards/rejected": -0.65021151304245,
|
|
"step": 655
|
|
},
|
|
{
|
|
"epoch": 0.9632892804698973,
|
|
"grad_norm": 5.95790958404541,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.00296932365745306,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 2.223355098446622e-09,
|
|
"logits/chosen": -7.680576324462891,
|
|
"logits/rejected": -7.717093467712402,
|
|
"logps/chosen": -223.98876953125,
|
|
"logps/ref_chosen": -85.32534790039062,
|
|
"logps/ref_rejected": -118.33866882324219,
|
|
"logps/rejected": -357.14013671875,
|
|
"loss": 1.1311,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.4093048572540283,
|
|
"rewards/margins": 0.294216513633728,
|
|
"rewards/rejected": -0.7035213708877563,
|
|
"step": 656
|
|
},
|
|
{
|
|
"epoch": 0.9647577092511013,
|
|
"grad_norm": 6.30833625793457,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.002948133973404765,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 2.055847060721566e-09,
|
|
"logits/chosen": -7.752594947814941,
|
|
"logits/rejected": -7.609277725219727,
|
|
"logps/chosen": -226.81549072265625,
|
|
"logps/ref_chosen": -80.19772338867188,
|
|
"logps/ref_rejected": -102.581298828125,
|
|
"logps/rejected": -330.3614501953125,
|
|
"loss": 1.1867,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -0.4302343726158142,
|
|
"rewards/margins": 0.23619841039180756,
|
|
"rewards/rejected": -0.666432797908783,
|
|
"step": 657
|
|
},
|
|
{
|
|
"epoch": 0.9662261380323054,
|
|
"grad_norm": 5.188457012176514,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.002930732909590006,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 1.8948725820160662e-09,
|
|
"logits/chosen": -7.566948890686035,
|
|
"logits/rejected": -7.049898147583008,
|
|
"logps/chosen": -248.58145141601562,
|
|
"logps/ref_chosen": -94.634521484375,
|
|
"logps/ref_rejected": -101.63162231445312,
|
|
"logps/rejected": -325.5663146972656,
|
|
"loss": 1.2098,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.44880080223083496,
|
|
"rewards/margins": 0.202609121799469,
|
|
"rewards/rejected": -0.6514099836349487,
|
|
"step": 658
|
|
},
|
|
{
|
|
"epoch": 0.9676945668135095,
|
|
"grad_norm": 4.891656875610352,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.0029116251971572638,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 1.7404359041573723e-09,
|
|
"logits/chosen": -8.355998039245605,
|
|
"logits/rejected": -7.4629597663879395,
|
|
"logps/chosen": -246.71299743652344,
|
|
"logps/ref_chosen": -112.55587005615234,
|
|
"logps/ref_rejected": -93.9216079711914,
|
|
"logps/rejected": -315.93914794921875,
|
|
"loss": 1.1774,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.3891860246658325,
|
|
"rewards/margins": 0.25232329964637756,
|
|
"rewards/rejected": -0.6415092945098877,
|
|
"step": 659
|
|
},
|
|
{
|
|
"epoch": 0.9691629955947136,
|
|
"grad_norm": 7.11164665222168,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.0028953389264643192,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 1.592541096695571e-09,
|
|
"logits/chosen": -7.641197681427002,
|
|
"logits/rejected": -7.153841972351074,
|
|
"logps/chosen": -234.69586181640625,
|
|
"logps/ref_chosen": -93.37742614746094,
|
|
"logps/ref_rejected": -81.39482116699219,
|
|
"logps/rejected": -311.2170715332031,
|
|
"loss": 1.1675,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.4076111614704132,
|
|
"rewards/margins": 0.25319740176200867,
|
|
"rewards/rejected": -0.6608085632324219,
|
|
"step": 660
|
|
},
|
|
{
|
|
"epoch": 0.9706314243759178,
|
|
"grad_norm": 4.977914333343506,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.002878249390050769,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 1.4511920567963908e-09,
|
|
"logits/chosen": -7.581234931945801,
|
|
"logits/rejected": -6.8354692459106445,
|
|
"logps/chosen": -233.6283416748047,
|
|
"logps/ref_chosen": -87.85516357421875,
|
|
"logps/ref_rejected": -92.40330505371094,
|
|
"logps/rejected": -332.63751220703125,
|
|
"loss": 1.1572,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -0.41780516505241394,
|
|
"rewards/margins": 0.26863789558410645,
|
|
"rewards/rejected": -0.6864430904388428,
|
|
"step": 661
|
|
},
|
|
{
|
|
"epoch": 0.9720998531571219,
|
|
"grad_norm": 4.641286373138428,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.002859483938664198,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 1.3163925091384532e-09,
|
|
"logits/chosen": -7.912067413330078,
|
|
"logits/rejected": -7.321425437927246,
|
|
"logps/chosen": -243.25115966796875,
|
|
"logps/ref_chosen": -102.77980041503906,
|
|
"logps/ref_rejected": -95.22531127929688,
|
|
"logps/rejected": -322.36285400390625,
|
|
"loss": 1.1704,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.4002068340778351,
|
|
"rewards/margins": 0.24505510926246643,
|
|
"rewards/rejected": -0.6452619433403015,
|
|
"step": 662
|
|
},
|
|
{
|
|
"epoch": 0.973568281938326,
|
|
"grad_norm": 5.062544822692871,
|
|
"kl/avg_steps": 0.46875,
|
|
"kl/beta": 0.002842606045305729,
|
|
"kl/n_epsilon_steps": 0.265625,
|
|
"kl/p_epsilon_steps": 0.734375,
|
|
"learning_rate": 1.1881460058152382e-09,
|
|
"logits/chosen": -7.826132774353027,
|
|
"logits/rejected": -7.58699369430542,
|
|
"logps/chosen": -235.74203491210938,
|
|
"logps/ref_chosen": -96.34658813476562,
|
|
"logps/ref_rejected": -120.52645111083984,
|
|
"logps/rejected": -340.11309814453125,
|
|
"loss": 1.1945,
|
|
"rewards/accuracies": 0.75,
|
|
"rewards/chosen": -0.39547061920166016,
|
|
"rewards/margins": 0.2249811291694641,
|
|
"rewards/rejected": -0.6204517483711243,
|
|
"step": 663
|
|
},
|
|
{
|
|
"epoch": 0.9750367107195301,
|
|
"grad_norm": 5.462507724761963,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.002829343546181917,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 1.066455926241383e-09,
|
|
"logits/chosen": -7.908196449279785,
|
|
"logits/rejected": -7.449959754943848,
|
|
"logps/chosen": -245.00120544433594,
|
|
"logps/ref_chosen": -91.84242248535156,
|
|
"logps/ref_rejected": -111.83668518066406,
|
|
"logps/rejected": -338.63543701171875,
|
|
"loss": 1.2061,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -0.43186432123184204,
|
|
"rewards/margins": 0.20575222373008728,
|
|
"rewards/rejected": -0.6376165151596069,
|
|
"step": 664
|
|
},
|
|
{
|
|
"epoch": 0.9765051395007343,
|
|
"grad_norm": 4.550708293914795,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.0028143920935690403,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 9.513254770636137e-10,
|
|
"logits/chosen": -7.470752716064453,
|
|
"logits/rejected": -7.1154465675354,
|
|
"logps/chosen": -217.97030639648438,
|
|
"logps/ref_chosen": -88.18618774414062,
|
|
"logps/ref_rejected": -91.9120101928711,
|
|
"logps/rejected": -303.5909423828125,
|
|
"loss": 1.1848,
|
|
"rewards/accuracies": 0.890625,
|
|
"rewards/chosen": -0.36321356892585754,
|
|
"rewards/margins": 0.22789615392684937,
|
|
"rewards/rejected": -0.5911097526550293,
|
|
"step": 665
|
|
},
|
|
{
|
|
"epoch": 0.9779735682819384,
|
|
"grad_norm": 6.450153350830078,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.002794307889416814,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 8.427576920763956e-10,
|
|
"logits/chosen": -7.702958583831787,
|
|
"logits/rejected": -7.515384674072266,
|
|
"logps/chosen": -249.63475036621094,
|
|
"logps/ref_chosen": -100.97460174560547,
|
|
"logps/ref_rejected": -101.24992370605469,
|
|
"logps/rejected": -327.71697998046875,
|
|
"loss": 1.2025,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -0.41387584805488586,
|
|
"rewards/margins": 0.2145766168832779,
|
|
"rewards/rejected": -0.628452479839325,
|
|
"step": 666
|
|
},
|
|
{
|
|
"epoch": 0.9794419970631424,
|
|
"grad_norm": 4.641995429992676,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.002779541537165642,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 7.407554321417764e-10,
|
|
"logits/chosen": -7.869577884674072,
|
|
"logits/rejected": -7.049417495727539,
|
|
"logps/chosen": -256.4490661621094,
|
|
"logps/ref_chosen": -97.5711669921875,
|
|
"logps/ref_rejected": -93.58476257324219,
|
|
"logps/rejected": -339.6527404785156,
|
|
"loss": 1.1737,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.4399784505367279,
|
|
"rewards/margins": 0.23969195783138275,
|
|
"rewards/rejected": -0.6796703934669495,
|
|
"step": 667
|
|
},
|
|
{
|
|
"epoch": 0.9809104258443465,
|
|
"grad_norm": 6.144704341888428,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.002763135591521859,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 6.453213851142225e-10,
|
|
"logits/chosen": -8.091211318969727,
|
|
"logits/rejected": -7.701225280761719,
|
|
"logps/chosen": -242.12380981445312,
|
|
"logps/ref_chosen": -102.5750503540039,
|
|
"logps/ref_rejected": -108.81768798828125,
|
|
"logps/rejected": -345.40899658203125,
|
|
"loss": 1.1573,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.38403064012527466,
|
|
"rewards/margins": 0.2652904689311981,
|
|
"rewards/rejected": -0.6493211388587952,
|
|
"step": 668
|
|
},
|
|
{
|
|
"epoch": 0.9823788546255506,
|
|
"grad_norm": 5.321188926696777,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.0027485338505357504,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 5.564580657695939e-10,
|
|
"logits/chosen": -7.704500198364258,
|
|
"logits/rejected": -6.9445672035217285,
|
|
"logps/chosen": -217.85406494140625,
|
|
"logps/ref_chosen": -89.49478149414062,
|
|
"logps/ref_rejected": -82.51950073242188,
|
|
"logps/rejected": -305.6316223144531,
|
|
"loss": 1.1609,
|
|
"rewards/accuracies": 0.8125,
|
|
"rewards/chosen": -0.3517535924911499,
|
|
"rewards/margins": 0.2574951946735382,
|
|
"rewards/rejected": -0.6092487573623657,
|
|
"step": 669
|
|
},
|
|
{
|
|
"epoch": 0.9838472834067548,
|
|
"grad_norm": 4.645162582397461,
|
|
"kl/avg_steps": 0.75,
|
|
"kl/beta": 0.0027323109097778797,
|
|
"kl/n_epsilon_steps": 0.125,
|
|
"kl/p_epsilon_steps": 0.875,
|
|
"learning_rate": 4.741678157389739e-10,
|
|
"logits/chosen": -7.779049873352051,
|
|
"logits/rejected": -7.034740447998047,
|
|
"logps/chosen": -223.52005004882812,
|
|
"logps/ref_chosen": -95.45459747314453,
|
|
"logps/ref_rejected": -101.53292846679688,
|
|
"logps/rejected": -332.3464050292969,
|
|
"loss": 1.1435,
|
|
"rewards/accuracies": 0.890625,
|
|
"rewards/chosen": -0.3478962182998657,
|
|
"rewards/margins": 0.2775384485721588,
|
|
"rewards/rejected": -0.6254346370697021,
|
|
"step": 670
|
|
},
|
|
{
|
|
"epoch": 0.9853157121879589,
|
|
"grad_norm": 4.930266380310059,
|
|
"kl/avg_steps": 0.5625,
|
|
"kl/beta": 0.0027119710575789213,
|
|
"kl/n_epsilon_steps": 0.21875,
|
|
"kl/p_epsilon_steps": 0.78125,
|
|
"learning_rate": 3.9845280344705245e-10,
|
|
"logits/chosen": -7.417923927307129,
|
|
"logits/rejected": -7.285045623779297,
|
|
"logps/chosen": -239.85293579101562,
|
|
"logps/ref_chosen": -82.12312316894531,
|
|
"logps/ref_rejected": -90.21969604492188,
|
|
"logps/rejected": -317.8214111328125,
|
|
"loss": 1.2244,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.42587053775787354,
|
|
"rewards/margins": 0.18708932399749756,
|
|
"rewards/rejected": -0.6129598617553711,
|
|
"step": 671
|
|
},
|
|
{
|
|
"epoch": 0.986784140969163,
|
|
"grad_norm": 5.901381969451904,
|
|
"kl/avg_steps": 0.4375,
|
|
"kl/beta": 0.0026968014426529408,
|
|
"kl/n_epsilon_steps": 0.28125,
|
|
"kl/p_epsilon_steps": 0.71875,
|
|
"learning_rate": 3.293150240547549e-10,
|
|
"logits/chosen": -7.819128513336182,
|
|
"logits/rejected": -7.268359184265137,
|
|
"logps/chosen": -247.38026428222656,
|
|
"logps/ref_chosen": -90.0619125366211,
|
|
"logps/ref_rejected": -100.45323181152344,
|
|
"logps/rejected": -320.5558166503906,
|
|
"loss": 1.2405,
|
|
"rewards/accuracies": 0.734375,
|
|
"rewards/chosen": -0.423230916261673,
|
|
"rewards/margins": 0.1669987291097641,
|
|
"rewards/rejected": -0.5902296304702759,
|
|
"step": 672
|
|
},
|
|
{
|
|
"epoch": 0.9882525697503671,
|
|
"grad_norm": 5.085114479064941,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.002685054438188672,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 2.6675629940689504e-10,
|
|
"logits/chosen": -7.903676986694336,
|
|
"logits/rejected": -7.309889793395996,
|
|
"logps/chosen": -218.5710906982422,
|
|
"logps/ref_chosen": -79.26315307617188,
|
|
"logps/ref_rejected": -91.34925079345703,
|
|
"logps/rejected": -302.2262268066406,
|
|
"loss": 1.2187,
|
|
"rewards/accuracies": 0.78125,
|
|
"rewards/chosen": -0.3727704882621765,
|
|
"rewards/margins": 0.18984678387641907,
|
|
"rewards/rejected": -0.562617301940918,
|
|
"step": 673
|
|
},
|
|
{
|
|
"epoch": 0.9897209985315712,
|
|
"grad_norm": 4.563065052032471,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.0026708655059337616,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 2.1077827798404725e-10,
|
|
"logits/chosen": -7.53197717666626,
|
|
"logits/rejected": -7.05259895324707,
|
|
"logps/chosen": -206.57798767089844,
|
|
"logps/ref_chosen": -75.45831298828125,
|
|
"logps/ref_rejected": -76.20362854003906,
|
|
"logps/rejected": -302.875,
|
|
"loss": 1.1642,
|
|
"rewards/accuracies": 0.890625,
|
|
"rewards/chosen": -0.34820300340652466,
|
|
"rewards/margins": 0.2524449825286865,
|
|
"rewards/rejected": -0.600648045539856,
|
|
"step": 674
|
|
},
|
|
{
|
|
"epoch": 0.9911894273127754,
|
|
"grad_norm": 5.335831165313721,
|
|
"kl/avg_steps": 0.53125,
|
|
"kl/beta": 0.0026518055237829685,
|
|
"kl/n_epsilon_steps": 0.234375,
|
|
"kl/p_epsilon_steps": 0.765625,
|
|
"learning_rate": 1.6138243485910863e-10,
|
|
"logits/chosen": -7.707300186157227,
|
|
"logits/rejected": -7.3889970779418945,
|
|
"logps/chosen": -227.656494140625,
|
|
"logps/ref_chosen": -79.90953063964844,
|
|
"logps/ref_rejected": -81.21824645996094,
|
|
"logps/rejected": -313.17236328125,
|
|
"loss": 1.1993,
|
|
"rewards/accuracies": 0.765625,
|
|
"rewards/chosen": -0.3907131552696228,
|
|
"rewards/margins": 0.2203899621963501,
|
|
"rewards/rejected": -0.6111031174659729,
|
|
"step": 675
|
|
},
|
|
{
|
|
"epoch": 0.9926578560939795,
|
|
"grad_norm": 4.61472749710083,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.0026377923786640167,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 1.1857007165852472e-10,
|
|
"logits/chosen": -8.043817520141602,
|
|
"logits/rejected": -7.283336639404297,
|
|
"logps/chosen": -254.79995727539062,
|
|
"logps/ref_chosen": -98.17111206054688,
|
|
"logps/ref_rejected": -95.024658203125,
|
|
"logps/rejected": -336.66851806640625,
|
|
"loss": 1.1933,
|
|
"rewards/accuracies": 0.796875,
|
|
"rewards/chosen": -0.41136687994003296,
|
|
"rewards/margins": 0.2214387059211731,
|
|
"rewards/rejected": -0.632805585861206,
|
|
"step": 676
|
|
},
|
|
{
|
|
"epoch": 0.9941262848751835,
|
|
"grad_norm": 4.470887660980225,
|
|
"kl/avg_steps": 0.59375,
|
|
"kl/beta": 0.002622222760692239,
|
|
"kl/n_epsilon_steps": 0.203125,
|
|
"kl/p_epsilon_steps": 0.796875,
|
|
"learning_rate": 8.23423165278725e-11,
|
|
"logits/chosen": -7.727552890777588,
|
|
"logits/rejected": -6.984161376953125,
|
|
"logps/chosen": -230.60610961914062,
|
|
"logps/ref_chosen": -91.37928009033203,
|
|
"logps/ref_rejected": -82.87776947021484,
|
|
"logps/rejected": -308.8211669921875,
|
|
"loss": 1.186,
|
|
"rewards/accuracies": 0.84375,
|
|
"rewards/chosen": -0.3634149134159088,
|
|
"rewards/margins": 0.22486907243728638,
|
|
"rewards/rejected": -0.5882839560508728,
|
|
"step": 677
|
|
},
|
|
{
|
|
"epoch": 0.9955947136563876,
|
|
"grad_norm": 4.6202616691589355,
|
|
"kl/avg_steps": 0.71875,
|
|
"kl/beta": 0.002606745343655348,
|
|
"kl/n_epsilon_steps": 0.140625,
|
|
"kl/p_epsilon_steps": 0.859375,
|
|
"learning_rate": 5.270012410216185e-11,
|
|
"logits/chosen": -7.629189491271973,
|
|
"logits/rejected": -6.984808921813965,
|
|
"logps/chosen": -208.89874267578125,
|
|
"logps/ref_chosen": -75.64586639404297,
|
|
"logps/ref_rejected": -86.96611022949219,
|
|
"logps/rejected": -323.3565979003906,
|
|
"loss": 1.1519,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -0.3453512489795685,
|
|
"rewards/margins": 0.265934556722641,
|
|
"rewards/rejected": -0.6112858057022095,
|
|
"step": 678
|
|
},
|
|
{
|
|
"epoch": 0.9970631424375918,
|
|
"grad_norm": 4.485384464263916,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.0025881431065499783,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 2.9644275480772416e-11,
|
|
"logits/chosen": -7.64573860168457,
|
|
"logits/rejected": -6.9826436042785645,
|
|
"logps/chosen": -232.62774658203125,
|
|
"logps/ref_chosen": -80.77344512939453,
|
|
"logps/ref_rejected": -82.87850189208984,
|
|
"logps/rejected": -317.65594482421875,
|
|
"loss": 1.1954,
|
|
"rewards/accuracies": 0.875,
|
|
"rewards/chosen": -0.39101141691207886,
|
|
"rewards/margins": 0.2123621702194214,
|
|
"rewards/rejected": -0.6033735871315002,
|
|
"step": 679
|
|
},
|
|
{
|
|
"epoch": 0.9985315712187959,
|
|
"grad_norm": 4.077805042266846,
|
|
"kl/avg_steps": 0.625,
|
|
"kl/beta": 0.0025712691713124514,
|
|
"kl/n_epsilon_steps": 0.1875,
|
|
"kl/p_epsilon_steps": 0.8125,
|
|
"learning_rate": 1.31753782067201e-11,
|
|
"logits/chosen": -7.682188987731934,
|
|
"logits/rejected": -7.54879093170166,
|
|
"logps/chosen": -261.8695373535156,
|
|
"logps/ref_chosen": -107.68292999267578,
|
|
"logps/ref_rejected": -116.09486389160156,
|
|
"logps/rejected": -359.0973815917969,
|
|
"loss": 1.1887,
|
|
"rewards/accuracies": 0.859375,
|
|
"rewards/chosen": -0.3943979740142822,
|
|
"rewards/margins": 0.22572645545005798,
|
|
"rewards/rejected": -0.6201244592666626,
|
|
"step": 680
|
|
},
|
|
{
|
|
"epoch": 1.0,
|
|
"grad_norm": 4.475795269012451,
|
|
"kl/avg_steps": 0.65625,
|
|
"kl/beta": 0.002555298386141658,
|
|
"kl/n_epsilon_steps": 0.171875,
|
|
"kl/p_epsilon_steps": 0.828125,
|
|
"learning_rate": 3.2938662507808745e-12,
|
|
"logits/chosen": -7.837874412536621,
|
|
"logits/rejected": -7.355884552001953,
|
|
"logps/chosen": -231.38394165039062,
|
|
"logps/ref_chosen": -93.01106262207031,
|
|
"logps/ref_rejected": -94.82217407226562,
|
|
"logps/rejected": -311.75775146484375,
|
|
"loss": 1.2117,
|
|
"rewards/accuracies": 0.828125,
|
|
"rewards/chosen": -0.3517065942287445,
|
|
"rewards/margins": 0.19827213883399963,
|
|
"rewards/rejected": -0.5499787330627441,
|
|
"step": 681
|
|
},
|
|
{
|
|
"epoch": 1.0,
|
|
"step": 681,
|
|
"total_flos": 0.0,
|
|
"train_loss": 1.0212064078200755,
|
|
"train_runtime": 3308.3388,
|
|
"train_samples_per_second": 13.178,
|
|
"train_steps_per_second": 0.206
|
|
}
|
|
],
|
|
"logging_steps": 1,
|
|
"max_steps": 681,
|
|
"num_input_tokens_seen": 0,
|
|
"num_train_epochs": 1,
|
|
"save_steps": 200,
|
|
"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": 8,
|
|
"trial_name": null,
|
|
"trial_params": null
|
|
}
|