{ "best_metric": null, "best_model_checkpoint": null, "epoch": 1.995276334435522, "eval_steps": 50, "global_step": 352, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "dpo_loss": 0.6931471824645996, "epoch": 0.005668398677373642, "grad_norm": 134.13658937705554, "learning_rate": 1.3888888888888888e-07, "logits": -1.3147305250167847, "logps": -88.0877456665039, "loss": 0.6931, "objective": 0.6931471824645996, "ranking_idealized": 0.5208333134651184, "ranking_idealized_expo": 0.5208333134651184, "ranking_simple": 0.5208333134651184, "regularize": 0.6931471824645996, "step": 1 }, { "dpo_loss": 0.6998100280761719, "epoch": 0.02834199338686821, "grad_norm": 131.14061254964616, "learning_rate": 6.944444444444446e-07, "logits": -1.3682328462600708, "logps": -84.04930877685547, "loss": 0.6964, "objective": 0.6998100280761719, "ranking_idealized": 0.5572916865348816, "ranking_idealized_expo": 0.546875, "ranking_simple": 0.546875, "regularize": 0.6998100280761719, "step": 5 }, { "dpo_loss": 0.6729087829589844, "epoch": 0.05668398677373642, "grad_norm": 189.53532499532858, "learning_rate": 1.3888888888888892e-06, "logits": -1.4502278566360474, "logps": -82.76358795166016, "loss": 0.705, "objective": 0.6729087829589844, "ranking_idealized": 0.5166666507720947, "ranking_idealized_expo": 0.5083333253860474, "ranking_simple": 0.512499988079071, "regularize": 0.6729087829589844, "step": 10 }, { "dpo_loss": 0.7919988632202148, "epoch": 0.08502598016060463, "grad_norm": 126.17436690763179, "learning_rate": 2.0833333333333334e-06, "logits": -1.4298707246780396, "logps": -82.53976440429688, "loss": 0.7233, "objective": 0.7919988632202148, "ranking_idealized": 0.512499988079071, "ranking_idealized_expo": 0.5041666626930237, "ranking_simple": 0.5083333253860474, "regularize": 0.7919988632202148, "step": 15 }, { "dpo_loss": 0.8708963990211487, "epoch": 0.11336797354747284, "grad_norm": 140.5214110989739, "learning_rate": 2.7777777777777783e-06, "logits": -1.4114222526550293, "logps": -84.63013458251953, "loss": 0.7455, "objective": 0.8708963990211487, "ranking_idealized": 0.5416666865348816, "ranking_idealized_expo": 0.5375000238418579, "ranking_simple": 0.5416666865348816, "regularize": 0.8708963990211487, "step": 20 }, { "dpo_loss": 0.8143303990364075, "epoch": 0.14170996693434104, "grad_norm": 147.98734815991946, "learning_rate": 3.4722222222222224e-06, "logits": -1.409764051437378, "logps": -83.78079986572266, "loss": 0.8653, "objective": 0.8143303990364075, "ranking_idealized": 0.5, "ranking_idealized_expo": 0.4958333373069763, "ranking_simple": 0.5, "regularize": 0.8143303990364075, "step": 25 }, { "dpo_loss": 1.0153591632843018, "epoch": 0.17005196032120926, "grad_norm": 119.43129724391213, "learning_rate": 4.166666666666667e-06, "logits": -1.3823612928390503, "logps": -84.5549087524414, "loss": 0.9219, "objective": 1.0153591632843018, "ranking_idealized": 0.5541666746139526, "ranking_idealized_expo": 0.5416666865348816, "ranking_simple": 0.550000011920929, "regularize": 1.0153591632843018, "step": 30 }, { "dpo_loss": 0.9491478204727173, "epoch": 0.19839395370807747, "grad_norm": 113.54025404025931, "learning_rate": 4.861111111111111e-06, "logits": -1.4168187379837036, "logps": -80.30852508544922, "loss": 0.9044, "objective": 0.9491478204727173, "ranking_idealized": 0.5083333253860474, "ranking_idealized_expo": 0.5041666626930237, "ranking_simple": 0.5541666746139526, "regularize": 0.9491478204727173, "step": 35 }, { "dpo_loss": 0.7457751035690308, "epoch": 0.22673594709494568, "grad_norm": 124.00613435721294, "learning_rate": 4.998023493068255e-06, "logits": -1.3939456939697266, "logps": -80.9621810913086, "loss": 0.9343, "objective": 0.7457751035690308, "ranking_idealized": 0.5249999761581421, "ranking_idealized_expo": 0.5208333134651184, "ranking_simple": 0.5833333134651184, "regularize": 0.7457751035690308, "step": 40 }, { "dpo_loss": 0.8915522694587708, "epoch": 0.25507794048181387, "grad_norm": 97.2880714247293, "learning_rate": 4.989999289644993e-06, "logits": -1.4818440675735474, "logps": -82.34064483642578, "loss": 0.9239, "objective": 0.8915522694587708, "ranking_idealized": 0.5458333492279053, "ranking_idealized_expo": 0.5375000238418579, "ranking_simple": 0.5791666507720947, "regularize": 0.8915522694587708, "step": 45 }, { "dpo_loss": 1.0888134241104126, "epoch": 0.2834199338686821, "grad_norm": 91.38901931507316, "learning_rate": 4.975823666181256e-06, "logits": -1.4309653043746948, "logps": -81.77391815185547, "loss": 1.0653, "objective": 1.0888134241104126, "ranking_idealized": 0.4333333373069763, "ranking_idealized_expo": 0.42500001192092896, "ranking_simple": 0.49166667461395264, "regularize": 1.0888134241104126, "step": 50 }, { "epoch": 0.2834199338686821, "eval_dpo_loss": 1.2502731084823608, "eval_logits": -1.4768881797790527, "eval_logps": -88.49180603027344, "eval_loss": 1.1910849809646606, "eval_objective": 1.2502731084823608, "eval_ranking_idealized": 0.5144628286361694, "eval_ranking_idealized_expo": 0.5092975497245789, "eval_ranking_simple": 0.5185950398445129, "eval_regularize": 1.2502731084823608, "eval_runtime": 260.3702, "eval_samples_per_second": 22.238, "eval_steps_per_second": 0.929, "step": 50 }, { "dpo_loss": 1.2924085855484009, "epoch": 0.3117619272555503, "grad_norm": 109.38466563530532, "learning_rate": 4.955531642853404e-06, "logits": -1.5091335773468018, "logps": -80.97488403320312, "loss": 1.2167, "objective": 1.2924085855484009, "ranking_idealized": 0.5458333492279053, "ranking_idealized_expo": 0.5416666865348816, "ranking_simple": 0.5666666626930237, "regularize": 1.2924085855484009, "step": 55 }, { "dpo_loss": 0.9506889581680298, "epoch": 0.3401039206424185, "grad_norm": 107.55617292574053, "learning_rate": 4.929173350101025e-06, "logits": -1.4120985269546509, "logps": -76.76266479492188, "loss": 1.0814, "objective": 0.9506889581680298, "ranking_idealized": 0.5208333134651184, "ranking_idealized_expo": 0.512499988079071, "ranking_simple": 0.5791666507720947, "regularize": 0.9506889581680298, "step": 60 }, { "dpo_loss": 1.035185694694519, "epoch": 0.3684459140292867, "grad_norm": 91.54638630267517, "learning_rate": 4.896813904782162e-06, "logits": -1.2802261114120483, "logps": -76.30724334716797, "loss": 1.1451, "objective": 1.035185694694519, "ranking_idealized": 0.5083333253860474, "ranking_idealized_expo": 0.5083333253860474, "ranking_simple": 0.5458333492279053, "regularize": 1.035185694694519, "step": 65 }, { "dpo_loss": 0.8953732252120972, "epoch": 0.39678790741615494, "grad_norm": 96.83949989428851, "learning_rate": 4.858533249305337e-06, "logits": -1.3047806024551392, "logps": -76.541259765625, "loss": 1.1303, "objective": 0.8953732252120972, "ranking_idealized": 0.5166666507720947, "ranking_idealized_expo": 0.5166666507720947, "ranking_simple": 0.5541666746139526, "regularize": 0.8953732252120972, "step": 70 }, { "dpo_loss": 1.0582144260406494, "epoch": 0.42512990080302315, "grad_norm": 92.62658273289433, "learning_rate": 4.814425954135786e-06, "logits": -1.303709864616394, "logps": -77.87187194824219, "loss": 1.1604, "objective": 1.0582144260406494, "ranking_idealized": 0.5458333492279053, "ranking_idealized_expo": 0.5458333492279053, "ranking_simple": 0.5916666388511658, "regularize": 1.0582144260406494, "step": 75 }, { "dpo_loss": 1.3123459815979004, "epoch": 0.45347189418989137, "grad_norm": 97.93404166357784, "learning_rate": 4.764600984163809e-06, "logits": -1.3597302436828613, "logps": -78.56843566894531, "loss": 1.2168, "objective": 1.3123459815979004, "ranking_idealized": 0.5166666507720947, "ranking_idealized_expo": 0.5166666507720947, "ranking_simple": 0.5791666507720947, "regularize": 1.3123459815979004, "step": 80 }, { "dpo_loss": 0.8359044790267944, "epoch": 0.4818138875767596, "grad_norm": 77.07769499983887, "learning_rate": 4.709181429512391e-06, "logits": -1.3809471130371094, "logps": -76.21597290039062, "loss": 1.0305, "objective": 0.8359044790267944, "ranking_idealized": 0.5249999761581421, "ranking_idealized_expo": 0.5249999761581421, "ranking_simple": 0.574999988079071, "regularize": 0.8359044790267944, "step": 85 }, { "dpo_loss": 1.3955554962158203, "epoch": 0.5101558809636277, "grad_norm": 81.93127920303778, "learning_rate": 4.648304201449153e-06, "logits": -1.3072587251663208, "logps": -76.73876953125, "loss": 1.0962, "objective": 1.3955554962158203, "ranking_idealized": 0.4833333194255829, "ranking_idealized_expo": 0.4791666567325592, "ranking_simple": 0.5083333253860474, "regularize": 1.3955554962158203, "step": 90 }, { "dpo_loss": 0.788605272769928, "epoch": 0.538497874350496, "grad_norm": 82.37111366044779, "learning_rate": 4.582119694153834e-06, "logits": -1.3767579793930054, "logps": -75.73047637939453, "loss": 0.9772, "objective": 0.788605272769928, "ranking_idealized": 0.5458333492279053, "ranking_idealized_expo": 0.5416666865348816, "ranking_simple": 0.625, "regularize": 0.788605272769928, "step": 95 }, { "dpo_loss": 0.9317588210105896, "epoch": 0.5668398677373642, "grad_norm": 91.36835540199432, "learning_rate": 4.510791413176912e-06, "logits": -1.4641989469528198, "logps": -78.29794311523438, "loss": 1.0433, "objective": 0.9317588210105896, "ranking_idealized": 0.5458333492279053, "ranking_idealized_expo": 0.5416666865348816, "ranking_simple": 0.612500011920929, "regularize": 0.9317588210105896, "step": 100 }, { "epoch": 0.5668398677373642, "eval_dpo_loss": 1.6223913431167603, "eval_logits": -1.4624882936477661, "eval_logps": -82.03755187988281, "eval_loss": 1.6200817823410034, "eval_objective": 1.6223913431167603, "eval_ranking_idealized": 0.5144628286361694, "eval_ranking_idealized_expo": 0.5092975497245789, "eval_ranking_simple": 0.5227272510528564, "eval_regularize": 1.6223913431167603, "eval_runtime": 258.5945, "eval_samples_per_second": 22.39, "eval_steps_per_second": 0.936, "step": 100 }, { "dpo_loss": 0.8804260492324829, "epoch": 0.5951818611242324, "grad_norm": 82.87732854480973, "learning_rate": 4.4344955715072344e-06, "logits": -1.3303104639053345, "logps": -77.949951171875, "loss": 1.0654, "objective": 0.8804260492324829, "ranking_idealized": 0.5458333492279053, "ranking_idealized_expo": 0.5333333611488342, "ranking_simple": 0.625, "regularize": 0.8804260492324829, "step": 105 }, { "dpo_loss": 1.1445485353469849, "epoch": 0.6235238545111006, "grad_norm": 95.04576924962743, "learning_rate": 4.353420654246546e-06, "logits": -1.3625129461288452, "logps": -76.01655578613281, "loss": 1.0183, "objective": 1.1445485353469849, "ranking_idealized": 0.4625000059604645, "ranking_idealized_expo": 0.44999998807907104, "ranking_simple": 0.5375000238418579, "regularize": 1.1445485353469849, "step": 110 }, { "dpo_loss": 0.9711818099021912, "epoch": 0.6518658478979689, "grad_norm": 75.33077814515555, "learning_rate": 4.267766952966369e-06, "logits": -1.3338966369628906, "logps": -75.58766174316406, "loss": 0.9873, "objective": 0.9711818099021912, "ranking_idealized": 0.4833333194255829, "ranking_idealized_expo": 0.4833333194255829, "ranking_simple": 0.550000011920929, "regularize": 0.9711818099021912, "step": 115 }, { "dpo_loss": 1.1687123775482178, "epoch": 0.680207841284837, "grad_norm": 79.26564985536876, "learning_rate": 4.177746070897593e-06, "logits": -1.2709105014801025, "logps": -74.96260833740234, "loss": 0.9884, "objective": 1.1687123775482178, "ranking_idealized": 0.5375000238418579, "ranking_idealized_expo": 0.5333333611488342, "ranking_simple": 0.6041666865348816, "regularize": 1.1687123775482178, "step": 120 }, { "dpo_loss": 1.1565426588058472, "epoch": 0.7085498346717053, "grad_norm": 74.43601140108132, "learning_rate": 4.083580400175153e-06, "logits": -1.2869479656219482, "logps": -73.08058166503906, "loss": 0.978, "objective": 1.1565426588058472, "ranking_idealized": 0.5583333373069763, "ranking_idealized_expo": 0.550000011920929, "ranking_simple": 0.6083333492279053, "regularize": 1.1565426588058472, "step": 125 }, { "dpo_loss": 0.999773383140564, "epoch": 0.7368918280585735, "grad_norm": 84.99903815462908, "learning_rate": 3.985502572429277e-06, "logits": -1.3930357694625854, "logps": -71.53678131103516, "loss": 0.9699, "objective": 0.999773383140564, "ranking_idealized": 0.4625000059604645, "ranking_idealized_expo": 0.44999998807907104, "ranking_simple": 0.5416666865348816, "regularize": 0.999773383140564, "step": 130 }, { "dpo_loss": 0.8479234576225281, "epoch": 0.7652338214454416, "grad_norm": 79.12645461805546, "learning_rate": 3.8837548840805395e-06, "logits": -1.3405685424804688, "logps": -74.6470947265625, "loss": 0.906, "objective": 0.8479234576225281, "ranking_idealized": 0.5375000238418579, "ranking_idealized_expo": 0.5291666388511658, "ranking_simple": 0.5625, "regularize": 0.8479234576225281, "step": 135 }, { "dpo_loss": 0.7356327176094055, "epoch": 0.7935758148323099, "grad_norm": 73.93634362309558, "learning_rate": 3.7785886977585562e-06, "logits": -1.2152502536773682, "logps": -74.95491027832031, "loss": 0.8578, "objective": 0.7356327176094055, "ranking_idealized": 0.5166666507720947, "ranking_idealized_expo": 0.5041666626930237, "ranking_simple": 0.5708333253860474, "regularize": 0.7356327176094055, "step": 140 }, { "dpo_loss": 0.6884726285934448, "epoch": 0.821917808219178, "grad_norm": 77.59609177667399, "learning_rate": 3.6702638213230344e-06, "logits": -1.2230737209320068, "logps": -75.2629165649414, "loss": 0.7952, "objective": 0.6884726285934448, "ranking_idealized": 0.5, "ranking_idealized_expo": 0.4833333194255829, "ranking_simple": 0.5708333253860474, "regularize": 0.6884726285934448, "step": 145 }, { "dpo_loss": 0.8187524676322937, "epoch": 0.8502598016060463, "grad_norm": 65.33005864175458, "learning_rate": 3.5590478660213214e-06, "logits": -1.2330371141433716, "logps": -74.38772583007812, "loss": 0.8546, "objective": 0.8187524676322937, "ranking_idealized": 0.550000011920929, "ranking_idealized_expo": 0.5375000238418579, "ranking_simple": 0.6166666746139526, "regularize": 0.8187524676322937, "step": 150 }, { "epoch": 0.8502598016060463, "eval_dpo_loss": 1.70005202293396, "eval_logits": -1.2569595575332642, "eval_logps": -82.16104125976562, "eval_loss": 1.6978113651275635, "eval_objective": 1.70005202293396, "eval_ranking_idealized": 0.5144628286361694, "eval_ranking_idealized_expo": 0.5092975497245789, "eval_ranking_simple": 0.5258264541625977, "eval_regularize": 1.70005202293396, "eval_runtime": 259.5265, "eval_samples_per_second": 22.31, "eval_steps_per_second": 0.932, "step": 150 }, { "dpo_loss": 0.837838351726532, "epoch": 0.8786017949929145, "grad_norm": 69.73971801640664, "learning_rate": 3.4452155853680457e-06, "logits": -1.1960196495056152, "logps": -75.7232894897461, "loss": 0.7339, "objective": 0.837838351726532, "ranking_idealized": 0.5291666388511658, "ranking_idealized_expo": 0.512499988079071, "ranking_simple": 0.6291666626930237, "regularize": 0.837838351726532, "step": 155 }, { "dpo_loss": 0.7973706722259521, "epoch": 0.9069437883797827, "grad_norm": 67.79186806814941, "learning_rate": 3.32904819638017e-06, "logits": -1.068854808807373, "logps": -78.05864715576172, "loss": 0.8754, "objective": 0.7973706722259521, "ranking_idealized": 0.4958333373069763, "ranking_idealized_expo": 0.4958333373069763, "ranking_simple": 0.5958333611488342, "regularize": 0.7973706722259521, "step": 160 }, { "dpo_loss": 0.8812463283538818, "epoch": 0.9352857817666509, "grad_norm": 66.78755174418093, "learning_rate": 3.2108326848442507e-06, "logits": -1.1002658605575562, "logps": -78.5975112915039, "loss": 0.7891, "objective": 0.8812463283538818, "ranking_idealized": 0.5333333611488342, "ranking_idealized_expo": 0.5208333134651184, "ranking_simple": 0.5916666388511658, "regularize": 0.8812463283538818, "step": 165 }, { "dpo_loss": 1.0159343481063843, "epoch": 0.9636277751535192, "grad_norm": 64.59296669922323, "learning_rate": 3.090861096332263e-06, "logits": -1.1763144731521606, "logps": -78.24954223632812, "loss": 0.7872, "objective": 1.0159343481063843, "ranking_idealized": 0.5208333134651184, "ranking_idealized_expo": 0.5083333253860474, "ranking_simple": 0.6000000238418579, "regularize": 1.0159343481063843, "step": 170 }, { "dpo_loss": 0.835219144821167, "epoch": 0.9919697685403873, "grad_norm": 63.856080085057854, "learning_rate": 2.9694298147174566e-06, "logits": -1.1786764860153198, "logps": -77.57752990722656, "loss": 0.7333, "objective": 0.835219144821167, "ranking_idealized": 0.5791666507720947, "ranking_idealized_expo": 0.574999988079071, "ranking_simple": 0.675000011920929, "regularize": 0.835219144821167, "step": 175 }, { "dpo_loss": 0.4763469994068146, "epoch": 1.0203117619272555, "grad_norm": 46.80871403438143, "learning_rate": 2.8468388299726714e-06, "logits": -1.2518482208251953, "logps": -77.97151947021484, "loss": 0.4442, "objective": 0.4763469994068146, "ranking_idealized": 0.5416666865348816, "ranking_idealized_expo": 0.5416666865348816, "ranking_simple": 0.6499999761581421, "regularize": 0.4763469994068146, "step": 180 }, { "dpo_loss": 0.4130065143108368, "epoch": 1.0486537553141237, "grad_norm": 50.334391778688, "learning_rate": 2.723390997059943e-06, "logits": -1.1434574127197266, "logps": -77.29036712646484, "loss": 0.3885, "objective": 0.4130065143108368, "ranking_idealized": 0.5041666626930237, "ranking_idealized_expo": 0.4833333194255829, "ranking_simple": 0.6041666865348816, "regularize": 0.4130065143108368, "step": 185 }, { "dpo_loss": 0.41978922486305237, "epoch": 1.076995748700992, "grad_norm": 50.91052129602807, "learning_rate": 2.599391287742315e-06, "logits": -1.1852554082870483, "logps": -77.81266784667969, "loss": 0.3795, "objective": 0.41978922486305237, "ranking_idealized": 0.5583333373069763, "ranking_idealized_expo": 0.550000011920929, "ranking_simple": 0.6499999761581421, "regularize": 0.41978922486305237, "step": 190 }, { "dpo_loss": 0.3017837405204773, "epoch": 1.10533774208786, "grad_norm": 47.94860845225541, "learning_rate": 2.4751460371661763e-06, "logits": -1.139602541923523, "logps": -76.76433563232422, "loss": 0.3311, "objective": 0.3017837405204773, "ranking_idealized": 0.5708333253860474, "ranking_idealized_expo": 0.5666666626930237, "ranking_simple": 0.675000011920929, "regularize": 0.3017837405204773, "step": 195 }, { "dpo_loss": 0.2089507132768631, "epoch": 1.1336797354747283, "grad_norm": 49.07634030170392, "learning_rate": 2.3509621870754505e-06, "logits": -1.179439902305603, "logps": -76.19169616699219, "loss": 0.3915, "objective": 0.2089507132768631, "ranking_idealized": 0.5083333253860474, "ranking_idealized_expo": 0.5083333253860474, "ranking_simple": 0.6291666626930237, "regularize": 0.2089507132768631, "step": 200 }, { "epoch": 1.1336797354747283, "eval_dpo_loss": 1.7483144998550415, "eval_logits": -1.200191617012024, "eval_logps": -81.72872924804688, "eval_loss": 1.722193717956543, "eval_objective": 1.7483144998550415, "eval_ranking_idealized": 0.5144628286361694, "eval_ranking_idealized_expo": 0.5092975497245789, "eval_ranking_simple": 0.5154958963394165, "eval_regularize": 1.7483144998550415, "eval_runtime": 258.9538, "eval_samples_per_second": 22.359, "eval_steps_per_second": 0.935, "step": 200 }, { "dpo_loss": 0.4261940121650696, "epoch": 1.1620217288615966, "grad_norm": 36.61392014542111, "learning_rate": 2.2271465275271985e-06, "logits": -1.1415462493896484, "logps": -76.67549896240234, "loss": 0.3966, "objective": 0.4261940121650696, "ranking_idealized": 0.4958333373069763, "ranking_idealized_expo": 0.48750001192092896, "ranking_simple": 0.6041666865348816, "regularize": 0.4261940121650696, "step": 205 }, { "dpo_loss": 0.26957443356513977, "epoch": 1.1903637222484649, "grad_norm": 44.00744245316499, "learning_rate": 2.1040049389819628e-06, "logits": -1.0984385013580322, "logps": -76.03750610351562, "loss": 0.3734, "objective": 0.26957443356513977, "ranking_idealized": 0.5083333253860474, "ranking_idealized_expo": 0.4958333373069763, "ranking_simple": 0.6166666746139526, "regularize": 0.26957443356513977, "step": 210 }, { "dpo_loss": 0.4371963143348694, "epoch": 1.2187057156353331, "grad_norm": 45.87158978742655, "learning_rate": 1.9818416366412277e-06, "logits": -1.0474963188171387, "logps": -75.47949981689453, "loss": 0.354, "objective": 0.4371963143348694, "ranking_idealized": 0.48750001192092896, "ranking_idealized_expo": 0.4833333194255829, "ranking_simple": 0.625, "regularize": 0.4371963143348694, "step": 215 }, { "dpo_loss": 0.37918028235435486, "epoch": 1.2470477090222012, "grad_norm": 51.193635889386584, "learning_rate": 1.8609584188988135e-06, "logits": -0.9162617325782776, "logps": -75.1710433959961, "loss": 0.4083, "objective": 0.37918028235435486, "ranking_idealized": 0.6041666865348816, "ranking_idealized_expo": 0.6000000238418579, "ranking_simple": 0.6875, "regularize": 0.37918028235435486, "step": 220 }, { "dpo_loss": 0.3700805604457855, "epoch": 1.2753897024090695, "grad_norm": 44.89792607105144, "learning_rate": 1.7416539217628792e-06, "logits": -0.9836326241493225, "logps": -76.86432647705078, "loss": 0.4221, "objective": 0.3700805604457855, "ranking_idealized": 0.5291666388511658, "ranking_idealized_expo": 0.5291666388511658, "ranking_simple": 0.6833333373069763, "regularize": 0.3700805604457855, "step": 225 }, { "dpo_loss": 0.4244202673435211, "epoch": 1.3037316957959377, "grad_norm": 42.464451303213224, "learning_rate": 1.6242228810904393e-06, "logits": -0.9664452075958252, "logps": -76.29109954833984, "loss": 0.3812, "objective": 0.4244202673435211, "ranking_idealized": 0.512499988079071, "ranking_idealized_expo": 0.5083333253860474, "ranking_simple": 0.612500011920929, "regularize": 0.4244202673435211, "step": 230 }, { "dpo_loss": 0.28174513578414917, "epoch": 1.3320736891828058, "grad_norm": 47.58560313825458, "learning_rate": 1.508955404457015e-06, "logits": -1.0931307077407837, "logps": -73.45219421386719, "loss": 0.3941, "objective": 0.28174513578414917, "ranking_idealized": 0.4958333373069763, "ranking_idealized_expo": 0.4833333194255829, "ranking_simple": 0.6499999761581421, "regularize": 0.28174513578414917, "step": 235 }, { "dpo_loss": 0.36350587010383606, "epoch": 1.360415682569674, "grad_norm": 51.30864554663432, "learning_rate": 1.3961362544602215e-06, "logits": -1.066832423210144, "logps": -75.47599792480469, "loss": 0.3706, "objective": 0.36350587010383606, "ranking_idealized": 0.48750001192092896, "ranking_idealized_expo": 0.4791666567325592, "ranking_simple": 0.6458333134651184, "regularize": 0.36350587010383606, "step": 240 }, { "dpo_loss": 0.3202449679374695, "epoch": 1.3887576759565423, "grad_norm": 40.382330440406214, "learning_rate": 1.2860441452278578e-06, "logits": -1.0695809125900269, "logps": -75.72514343261719, "loss": 0.3752, "objective": 0.3202449679374695, "ranking_idealized": 0.5249999761581421, "ranking_idealized_expo": 0.5166666507720947, "ranking_simple": 0.6499999761581421, "regularize": 0.3202449679374695, "step": 245 }, { "dpo_loss": 0.2569473087787628, "epoch": 1.4170996693434104, "grad_norm": 45.39780331035215, "learning_rate": 1.1789510538684524e-06, "logits": -1.0426526069641113, "logps": -75.82499694824219, "loss": 0.361, "objective": 0.2569473087787628, "ranking_idealized": 0.4791666567325592, "ranking_idealized_expo": 0.47083333134651184, "ranking_simple": 0.6541666388511658, "regularize": 0.2569473087787628, "step": 250 }, { "epoch": 1.4170996693434104, "eval_dpo_loss": 1.6762409210205078, "eval_logits": -1.155253529548645, "eval_logps": -81.39386749267578, "eval_loss": 1.647857427597046, "eval_objective": 1.6762409210205078, "eval_ranking_idealized": 0.5144628286361694, "eval_ranking_idealized_expo": 0.5092975497245789, "eval_ranking_simple": 0.5268595218658447, "eval_regularize": 1.6762409210205078, "eval_runtime": 259.2372, "eval_samples_per_second": 22.335, "eval_steps_per_second": 0.934, "step": 250 }, { "dpo_loss": 0.2646530568599701, "epoch": 1.4454416627302786, "grad_norm": 43.51583035745083, "learning_rate": 1.0751215485652644e-06, "logits": -1.0492957830429077, "logps": -76.46492004394531, "loss": 0.3288, "objective": 0.2646530568599701, "ranking_idealized": 0.5666666626930237, "ranking_idealized_expo": 0.5458333492279053, "ranking_simple": 0.6666666865348816, "regularize": 0.2646530568599701, "step": 255 }, { "dpo_loss": 0.26657938957214355, "epoch": 1.473783656117147, "grad_norm": 40.706877805351375, "learning_rate": 9.74812134973689e-07, "logits": -1.0810118913650513, "logps": -77.01248931884766, "loss": 0.2812, "objective": 0.26657938957214355, "ranking_idealized": 0.49166667461395264, "ranking_idealized_expo": 0.48750001192092896, "ranking_simple": 0.6416666507720947, "regularize": 0.26657938957214355, "step": 260 }, { "dpo_loss": 0.2822701334953308, "epoch": 1.5021256495040152, "grad_norm": 54.897750443597026, "learning_rate": 8.78270622536716e-07, "logits": -1.0692379474639893, "logps": -75.91280364990234, "loss": 0.3101, "objective": 0.2822701334953308, "ranking_idealized": 0.5083333253860474, "ranking_idealized_expo": 0.5041666626930237, "ranking_simple": 0.637499988079071, "regularize": 0.2822701334953308, "step": 265 }, { "dpo_loss": 0.3819098472595215, "epoch": 1.5304676428908834, "grad_norm": 54.244126776548505, "learning_rate": 7.857355122839674e-07, "logits": -1.0702418088912964, "logps": -77.07707977294922, "loss": 0.3015, "objective": 0.3819098472595215, "ranking_idealized": 0.512499988079071, "ranking_idealized_expo": 0.5041666626930237, "ranking_simple": 0.6208333373069763, "regularize": 0.3819098472595215, "step": 270 }, { "dpo_loss": 0.3564348518848419, "epoch": 1.5588096362777515, "grad_norm": 47.03320315146794, "learning_rate": 6.974354076267081e-07, "logits": -1.0758289098739624, "logps": -77.3719482421875, "loss": 0.3224, "objective": 0.3564348518848419, "ranking_idealized": 0.5416666865348816, "ranking_idealized_expo": 0.5249999761581421, "ranking_simple": 0.6499999761581421, "regularize": 0.3564348518848419, "step": 275 }, { "dpo_loss": 0.3068177402019501, "epoch": 1.5871516296646198, "grad_norm": 44.12484186942547, "learning_rate": 6.135884496044245e-07, "logits": -1.0040802955627441, "logps": -75.53694915771484, "loss": 0.3271, "objective": 0.3068177402019501, "ranking_idealized": 0.5083333253860474, "ranking_idealized_expo": 0.5041666626930237, "ranking_simple": 0.637499988079071, "regularize": 0.3068177402019501, "step": 280 }, { "dpo_loss": 0.1998676210641861, "epoch": 1.615493623051488, "grad_norm": 43.046848040974425, "learning_rate": 5.344017779781835e-07, "logits": -1.1104124784469604, "logps": -77.22103118896484, "loss": 0.276, "objective": 0.1998676210641861, "ranking_idealized": 0.5249999761581421, "ranking_idealized_expo": 0.5166666507720947, "ranking_simple": 0.6499999761581421, "regularize": 0.1998676210641861, "step": 285 }, { "dpo_loss": 0.32884570956230164, "epoch": 1.643835616438356, "grad_norm": 41.71440224560557, "learning_rate": 4.6007101950209827e-07, "logits": -1.1282609701156616, "logps": -75.14998626708984, "loss": 0.3052, "objective": 0.32884570956230164, "ranking_idealized": 0.5666666626930237, "ranking_idealized_expo": 0.5625, "ranking_simple": 0.6541666388511658, "regularize": 0.32884570956230164, "step": 290 }, { "dpo_loss": 0.2726498246192932, "epoch": 1.6721776098252243, "grad_norm": 48.144178875127, "learning_rate": 3.907798046371139e-07, "logits": -1.040784239768982, "logps": -74.10509490966797, "loss": 0.3043, "objective": 0.2726498246192932, "ranking_idealized": 0.49166667461395264, "ranking_idealized_expo": 0.4791666567325592, "ranking_simple": 0.6333333253860474, "regularize": 0.2726498246192932, "step": 295 }, { "dpo_loss": 0.1931297481060028, "epoch": 1.7005196032120926, "grad_norm": 33.90759140524704, "learning_rate": 3.266993139010438e-07, "logits": -1.089760661125183, "logps": -75.90840148925781, "loss": 0.2219, "objective": 0.1931297481060028, "ranking_idealized": 0.5833333134651184, "ranking_idealized_expo": 0.574999988079071, "ranking_simple": 0.6833333373069763, "regularize": 0.1931297481060028, "step": 300 }, { "epoch": 1.7005196032120926, "eval_dpo_loss": 1.6152015924453735, "eval_logits": -1.1283313035964966, "eval_logps": -81.46894073486328, "eval_loss": 1.592815637588501, "eval_objective": 1.6152015924453735, "eval_ranking_idealized": 0.5144628286361694, "eval_ranking_idealized_expo": 0.5092975497245789, "eval_ranking_simple": 0.5247933864593506, "eval_regularize": 1.6152015924453735, "eval_runtime": 258.7786, "eval_samples_per_second": 22.374, "eval_steps_per_second": 0.935, "step": 300 }, { "dpo_loss": 0.3021364212036133, "epoch": 1.7288615965989607, "grad_norm": 44.912490110408946, "learning_rate": 2.6798785497556184e-07, "logits": -1.057539939880371, "logps": -76.72319030761719, "loss": 0.2844, "objective": 0.3021364212036133, "ranking_idealized": 0.5166666507720947, "ranking_idealized_expo": 0.5041666626930237, "ranking_simple": 0.6291666626930237, "regularize": 0.3021364212036133, "step": 305 }, { "dpo_loss": 0.26111310720443726, "epoch": 1.7572035899858292, "grad_norm": 44.74561046923631, "learning_rate": 2.1479047161491351e-07, "logits": -1.105485439300537, "logps": -74.68561553955078, "loss": 0.3067, "objective": 0.26111310720443726, "ranking_idealized": 0.5333333611488342, "ranking_idealized_expo": 0.5208333134651184, "ranking_simple": 0.6499999761581421, "regularize": 0.26111310720443726, "step": 310 }, { "dpo_loss": 0.3290327191352844, "epoch": 1.7855455833726972, "grad_norm": 42.09710061818234, "learning_rate": 1.6723858532249783e-07, "logits": -1.0454905033111572, "logps": -75.4970932006836, "loss": 0.3075, "objective": 0.3290327191352844, "ranking_idealized": 0.5208333134651184, "ranking_idealized_expo": 0.5083333253860474, "ranking_simple": 0.6166666746139526, "regularize": 0.3290327191352844, "step": 315 }, { "dpo_loss": 0.3243291974067688, "epoch": 1.8138875767595655, "grad_norm": 44.554500492238475, "learning_rate": 1.2544967068054332e-07, "logits": -1.0948283672332764, "logps": -77.41373443603516, "loss": 0.3045, "objective": 0.3243291974067688, "ranking_idealized": 0.512499988079071, "ranking_idealized_expo": 0.5041666626930237, "ranking_simple": 0.6333333253860474, "regularize": 0.3243291974067688, "step": 320 }, { "dpo_loss": 0.2077714055776596, "epoch": 1.8422295701464337, "grad_norm": 43.51438298831344, "learning_rate": 8.952696513496756e-08, "logits": -1.1423591375350952, "logps": -78.48583221435547, "loss": 0.2996, "objective": 0.2077714055776596, "ranking_idealized": 0.5333333611488342, "ranking_idealized_expo": 0.5166666507720947, "ranking_simple": 0.675000011920929, "regularize": 0.2077714055776596, "step": 325 }, { "dpo_loss": 0.2971789240837097, "epoch": 1.8705715635333018, "grad_norm": 39.32410077316406, "learning_rate": 5.955921395237318e-08, "logits": -0.9325958490371704, "logps": -75.0538330078125, "loss": 0.3109, "objective": 0.2971789240837097, "ranking_idealized": 0.4749999940395355, "ranking_idealized_expo": 0.47083333134651184, "ranking_simple": 0.6208333373069763, "regularize": 0.2971789240837097, "step": 330 }, { "dpo_loss": 0.2692752778530121, "epoch": 1.89891355692017, "grad_norm": 32.47537766337789, "learning_rate": 3.5620450979259034e-08, "logits": -1.07979416847229, "logps": -75.161865234375, "loss": 0.2887, "objective": 0.2692752778530121, "ranking_idealized": 0.5, "ranking_idealized_expo": 0.4958333373069763, "ranking_simple": 0.6416666507720947, "regularize": 0.2692752778530121, "step": 335 }, { "dpo_loss": 0.28865328431129456, "epoch": 1.9272555503070383, "grad_norm": 39.17207203979263, "learning_rate": 1.7769815745066476e-08, "logits": -1.054652452468872, "logps": -77.40970611572266, "loss": 0.3272, "objective": 0.28865328431129456, "ranking_idealized": 0.5416666865348816, "ranking_idealized_expo": 0.5416666865348816, "ranking_simple": 0.675000011920929, "regularize": 0.28865328431129456, "step": 340 }, { "dpo_loss": 0.2623630166053772, "epoch": 1.9555975436939064, "grad_norm": 40.59066285763909, "learning_rate": 6.051407360895822e-09, "logits": -1.0662161111831665, "logps": -76.14335632324219, "loss": 0.2707, "objective": 0.2623630166053772, "ranking_idealized": 0.5333333611488342, "ranking_idealized_expo": 0.5333333611488342, "ranking_simple": 0.6541666388511658, "regularize": 0.2623630166053772, "step": 345 }, { "dpo_loss": 0.27006199955940247, "epoch": 1.9839395370807746, "grad_norm": 46.652605490125566, "learning_rate": 4.941755748361088e-10, "logits": -1.0103791952133179, "logps": -76.85175323486328, "loss": 0.2782, "objective": 0.27006199955940247, "ranking_idealized": 0.49166667461395264, "ranking_idealized_expo": 0.48750001192092896, "ranking_simple": 0.6041666865348816, "regularize": 0.27006199955940247, "step": 350 }, { "epoch": 1.9839395370807746, "eval_dpo_loss": 1.605821132659912, "eval_logits": -1.134600043296814, "eval_logps": -81.46167755126953, "eval_loss": 1.5849751234054565, "eval_objective": 1.605821132659912, "eval_ranking_idealized": 0.5144628286361694, "eval_ranking_idealized_expo": 0.5092975497245789, "eval_ranking_simple": 0.5216942429542542, "eval_regularize": 1.605821132659912, "eval_runtime": 259.036, "eval_samples_per_second": 22.352, "eval_steps_per_second": 0.934, "step": 350 }, { "epoch": 1.995276334435522, "step": 352, "total_flos": 0.0, "train_loss": 0.635454860092564, "train_runtime": 14182.6927, "train_samples_per_second": 7.164, "train_steps_per_second": 0.025 } ], "logging_steps": 5, "max_steps": 352, "num_input_tokens_seen": 0, "num_train_epochs": 2, "save_steps": 500, "stateful_callbacks": { "TrainerControl": { "args": { "should_epoch_stop": false, "should_evaluate": false, "should_log": false, "should_save": false, "should_training_stop": false }, "attributes": {} } }, "total_flos": 0.0, "train_batch_size": 4, "trial_name": null, "trial_params": null }