{ "best_metric": null, "best_model_checkpoint": null, "epoch": 1.0, "eval_steps": 50, "global_step": 478, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.02092050209205021, "grad_norm": 9.137621785616465, "learning_rate": 1.0416666666666667e-07, "logits/chosen": -2.6498842239379883, "logits/rejected": -2.6401047706604004, "logps/chosen": -294.3343505859375, "logps/rejected": -289.1372985839844, "loss": 0.6931, "rewards/accuracies": 0.41874998807907104, "rewards/chosen": -0.0002774398017209023, "rewards/margins": 0.00026476362836547196, "rewards/rejected": -0.0005422033718787134, "step": 10 }, { "epoch": 0.04184100418410042, "grad_norm": 8.382944921921224, "learning_rate": 2.0833333333333333e-07, "logits/chosen": -2.6785335540771484, "logits/rejected": -2.6339826583862305, "logps/chosen": -294.88311767578125, "logps/rejected": -276.8221130371094, "loss": 0.6926, "rewards/accuracies": 0.5562499761581421, "rewards/chosen": 0.000260740373050794, "rewards/margins": 0.001288960687816143, "rewards/rejected": -0.0010282204020768404, "step": 20 }, { "epoch": 0.06276150627615062, "grad_norm": 8.348013142376631, "learning_rate": 3.1249999999999997e-07, "logits/chosen": -2.642605781555176, "logits/rejected": -2.6058638095855713, "logps/chosen": -282.0052185058594, "logps/rejected": -247.48208618164062, "loss": 0.6882, "rewards/accuracies": 0.6937500238418579, "rewards/chosen": 0.006200692616403103, "rewards/margins": 0.009530355222523212, "rewards/rejected": -0.0033296621404588223, "step": 30 }, { "epoch": 0.08368200836820083, "grad_norm": 8.873941277767704, "learning_rate": 4.1666666666666667e-07, "logits/chosen": -2.6004221439361572, "logits/rejected": -2.5648019313812256, "logps/chosen": -263.68304443359375, "logps/rejected": -269.95159912109375, "loss": 0.6776, "rewards/accuracies": 0.65625, "rewards/chosen": 0.0284257885068655, "rewards/margins": 0.026932990178465843, "rewards/rejected": 0.001492797746323049, "step": 40 }, { "epoch": 0.10460251046025104, "grad_norm": 9.18858486699177, "learning_rate": 4.999733114418725e-07, "logits/chosen": -2.5885472297668457, "logits/rejected": -2.5707509517669678, "logps/chosen": -269.8135070800781, "logps/rejected": -265.47064208984375, "loss": 0.6557, "rewards/accuracies": 0.706250011920929, "rewards/chosen": 0.018491629511117935, "rewards/margins": 0.0841543972492218, "rewards/rejected": -0.06566276401281357, "step": 50 }, { "epoch": 0.10460251046025104, "eval_logits/chosen": -2.649272918701172, "eval_logits/rejected": -2.611584424972534, "eval_logps/chosen": -263.7303466796875, "eval_logps/rejected": -275.50445556640625, "eval_loss": 0.6479206681251526, "eval_rewards/accuracies": 0.70703125, "eval_rewards/chosen": -0.011004311963915825, "eval_rewards/margins": 0.11741933971643448, "eval_rewards/rejected": -0.12842364609241486, "eval_runtime": 102.963, "eval_samples_per_second": 19.424, "eval_steps_per_second": 0.311, "step": 50 }, { "epoch": 0.12552301255230125, "grad_norm": 12.415021202754499, "learning_rate": 4.990398100856366e-07, "logits/chosen": -2.608935832977295, "logits/rejected": -2.5622076988220215, "logps/chosen": -316.8729553222656, "logps/rejected": -300.7294616699219, "loss": 0.6376, "rewards/accuracies": 0.65625, "rewards/chosen": -0.07160066068172455, "rewards/margins": 0.12479270994663239, "rewards/rejected": -0.19639338552951813, "step": 60 }, { "epoch": 0.14644351464435146, "grad_norm": 16.95023916970366, "learning_rate": 4.967775735898179e-07, "logits/chosen": -2.5255544185638428, "logits/rejected": -2.4770090579986572, "logps/chosen": -310.1564636230469, "logps/rejected": -286.7000427246094, "loss": 0.6174, "rewards/accuracies": 0.6625000238418579, "rewards/chosen": -0.1686697006225586, "rewards/margins": 0.18098047375679016, "rewards/rejected": -0.34965020418167114, "step": 70 }, { "epoch": 0.16736401673640167, "grad_norm": 14.651689743135485, "learning_rate": 4.931986719649298e-07, "logits/chosen": -2.3215534687042236, "logits/rejected": -2.257579803466797, "logps/chosen": -285.29833984375, "logps/rejected": -311.03729248046875, "loss": 0.6044, "rewards/accuracies": 0.675000011920929, "rewards/chosen": -0.07682778686285019, "rewards/margins": 0.2939576804637909, "rewards/rejected": -0.3707854449748993, "step": 80 }, { "epoch": 0.18828451882845187, "grad_norm": 19.539798910302995, "learning_rate": 4.883222001996351e-07, "logits/chosen": -1.3052003383636475, "logits/rejected": -1.1468216180801392, "logps/chosen": -294.7815246582031, "logps/rejected": -303.13427734375, "loss": 0.5866, "rewards/accuracies": 0.737500011920929, "rewards/chosen": -0.328985333442688, "rewards/margins": 0.45616284012794495, "rewards/rejected": -0.7851482629776001, "step": 90 }, { "epoch": 0.20920502092050208, "grad_norm": 25.641340302712948, "learning_rate": 4.821741763807186e-07, "logits/chosen": -0.7993453145027161, "logits/rejected": -0.40382108092308044, "logps/chosen": -333.15521240234375, "logps/rejected": -387.1316833496094, "loss": 0.5639, "rewards/accuracies": 0.7437499761581421, "rewards/chosen": -0.5871611833572388, "rewards/margins": 0.6322518587112427, "rewards/rejected": -1.2194130420684814, "step": 100 }, { "epoch": 0.20920502092050208, "eval_logits/chosen": -0.9011980295181274, "eval_logits/rejected": -0.5617944598197937, "eval_logps/chosen": -323.5882873535156, "eval_logps/rejected": -383.63751220703125, "eval_loss": 0.5651677846908569, "eval_rewards/accuracies": 0.73828125, "eval_rewards/chosen": -0.6095837354660034, "eval_rewards/margins": 0.6001702547073364, "eval_rewards/rejected": -1.2097539901733398, "eval_runtime": 101.4152, "eval_samples_per_second": 19.721, "eval_steps_per_second": 0.316, "step": 100 }, { "epoch": 0.2301255230125523, "grad_norm": 21.689227912073658, "learning_rate": 4.747874028753375e-07, "logits/chosen": -0.8007869720458984, "logits/rejected": -0.4384717047214508, "logps/chosen": -323.02410888671875, "logps/rejected": -343.52508544921875, "loss": 0.5585, "rewards/accuracies": 0.65625, "rewards/chosen": -0.5073341727256775, "rewards/margins": 0.517856240272522, "rewards/rejected": -1.0251905918121338, "step": 110 }, { "epoch": 0.2510460251046025, "grad_norm": 40.51442433382909, "learning_rate": 4.662012913161997e-07, "logits/chosen": 0.0625094398856163, "logits/rejected": 0.6715419292449951, "logps/chosen": -367.49664306640625, "logps/rejected": -383.600830078125, "loss": 0.5396, "rewards/accuracies": 0.768750011920929, "rewards/chosen": -0.6050666570663452, "rewards/margins": 0.6428467631340027, "rewards/rejected": -1.2479135990142822, "step": 120 }, { "epoch": 0.2719665271966527, "grad_norm": 31.04938830573247, "learning_rate": 4.5646165232345103e-07, "logits/chosen": 0.35010597109794617, "logits/rejected": 1.0627130270004272, "logps/chosen": -369.3751220703125, "logps/rejected": -391.6081237792969, "loss": 0.5421, "rewards/accuracies": 0.737500011920929, "rewards/chosen": -0.6456856727600098, "rewards/margins": 0.6666911840438843, "rewards/rejected": -1.312376856803894, "step": 130 }, { "epoch": 0.2928870292887029, "grad_norm": 23.124000981376877, "learning_rate": 4.456204510851956e-07, "logits/chosen": 1.3312932252883911, "logits/rejected": 1.6921491622924805, "logps/chosen": -328.2577209472656, "logps/rejected": -374.11846923828125, "loss": 0.5318, "rewards/accuracies": 0.7124999761581421, "rewards/chosen": -0.7673874497413635, "rewards/margins": 0.5369923710823059, "rewards/rejected": -1.3043798208236694, "step": 140 }, { "epoch": 0.3138075313807531, "grad_norm": 28.62578465081787, "learning_rate": 4.337355301007335e-07, "logits/chosen": 0.7725589871406555, "logits/rejected": 1.4066698551177979, "logps/chosen": -374.64288330078125, "logps/rejected": -426.3128356933594, "loss": 0.5323, "rewards/accuracies": 0.737500011920929, "rewards/chosen": -0.7278563380241394, "rewards/margins": 0.7114464044570923, "rewards/rejected": -1.439302682876587, "step": 150 }, { "epoch": 0.3138075313807531, "eval_logits/chosen": -0.09001250565052032, "eval_logits/rejected": 0.4792196750640869, "eval_logps/chosen": -317.6140441894531, "eval_logps/rejected": -391.6695556640625, "eval_loss": 0.5405200123786926, "eval_rewards/accuracies": 0.76171875, "eval_rewards/chosen": -0.5498415231704712, "eval_rewards/margins": 0.7402328252792358, "eval_rewards/rejected": -1.290074348449707, "eval_runtime": 102.6926, "eval_samples_per_second": 19.476, "eval_steps_per_second": 0.312, "step": 150 }, { "epoch": 0.33472803347280333, "grad_norm": 26.58643912067179, "learning_rate": 4.2087030056579986e-07, "logits/chosen": -0.5741851925849915, "logits/rejected": 0.09496372938156128, "logps/chosen": -313.5487976074219, "logps/rejected": -365.291015625, "loss": 0.5399, "rewards/accuracies": 0.75, "rewards/chosen": -0.5437325835227966, "rewards/margins": 0.6792270541191101, "rewards/rejected": -1.2229596376419067, "step": 160 }, { "epoch": 0.35564853556485354, "grad_norm": 23.464961776068243, "learning_rate": 4.070934040463998e-07, "logits/chosen": -0.5381526947021484, "logits/rejected": 0.14733538031578064, "logps/chosen": -354.4422912597656, "logps/rejected": -430.18682861328125, "loss": 0.5326, "rewards/accuracies": 0.75, "rewards/chosen": -0.7926353216171265, "rewards/margins": 0.8019927740097046, "rewards/rejected": -1.594628095626831, "step": 170 }, { "epoch": 0.37656903765690375, "grad_norm": 22.467354382291056, "learning_rate": 3.9247834624635404e-07, "logits/chosen": -0.20919005572795868, "logits/rejected": 0.49986928701400757, "logps/chosen": -371.064697265625, "logps/rejected": -408.84149169921875, "loss": 0.5358, "rewards/accuracies": 0.706250011920929, "rewards/chosen": -0.9278782606124878, "rewards/margins": 0.5680198669433594, "rewards/rejected": -1.4958980083465576, "step": 180 }, { "epoch": 0.39748953974895396, "grad_norm": 29.254730102983988, "learning_rate": 3.7710310482256523e-07, "logits/chosen": -0.3338702619075775, "logits/rejected": 0.38479360938072205, "logps/chosen": -319.56805419921875, "logps/rejected": -378.23590087890625, "loss": 0.5412, "rewards/accuracies": 0.7562500238418579, "rewards/chosen": -0.709225058555603, "rewards/margins": 0.857193648815155, "rewards/rejected": -1.5664187669754028, "step": 190 }, { "epoch": 0.41841004184100417, "grad_norm": 21.764790734639657, "learning_rate": 3.610497133404795e-07, "logits/chosen": -0.4434884190559387, "logits/rejected": 0.06593348830938339, "logps/chosen": -343.6837463378906, "logps/rejected": -397.1659240722656, "loss": 0.536, "rewards/accuracies": 0.7124999761581421, "rewards/chosen": -0.6647871136665344, "rewards/margins": 0.6501897573471069, "rewards/rejected": -1.314976692199707, "step": 200 }, { "epoch": 0.41841004184100417, "eval_logits/chosen": -0.019472889602184296, "eval_logits/rejected": 0.65250563621521, "eval_logps/chosen": -326.44696044921875, "eval_logps/rejected": -400.973388671875, "eval_loss": 0.5353578329086304, "eval_rewards/accuracies": 0.765625, "eval_rewards/chosen": -0.6381703019142151, "eval_rewards/margins": 0.7449426054954529, "eval_rewards/rejected": -1.3831130266189575, "eval_runtime": 102.2662, "eval_samples_per_second": 19.557, "eval_steps_per_second": 0.313, "step": 200 }, { "epoch": 0.4393305439330544, "grad_norm": 27.745510528591602, "learning_rate": 3.4440382358952115e-07, "logits/chosen": 1.1720086336135864, "logits/rejected": 1.8830058574676514, "logps/chosen": -403.1050109863281, "logps/rejected": -435.21942138671875, "loss": 0.5366, "rewards/accuracies": 0.7124999761581421, "rewards/chosen": -0.988387405872345, "rewards/margins": 0.5626603960990906, "rewards/rejected": -1.5510478019714355, "step": 210 }, { "epoch": 0.4602510460251046, "grad_norm": 22.068359980003752, "learning_rate": 3.272542485937368e-07, "logits/chosen": 1.7870858907699585, "logits/rejected": 2.4022154808044434, "logps/chosen": -351.79974365234375, "logps/rejected": -414.64129638671875, "loss": 0.5367, "rewards/accuracies": 0.7250000238418579, "rewards/chosen": -0.945723831653595, "rewards/margins": 0.6799818277359009, "rewards/rejected": -1.6257057189941406, "step": 220 }, { "epoch": 0.4811715481171548, "grad_norm": 24.112210218904135, "learning_rate": 3.096924887558854e-07, "logits/chosen": 1.5326052904129028, "logits/rejected": 1.9995750188827515, "logps/chosen": -340.270263671875, "logps/rejected": -413.39385986328125, "loss": 0.494, "rewards/accuracies": 0.768750011920929, "rewards/chosen": -0.8984026908874512, "rewards/margins": 0.7524574995040894, "rewards/rejected": -1.650860071182251, "step": 230 }, { "epoch": 0.502092050209205, "grad_norm": 27.75263813297474, "learning_rate": 2.9181224366319943e-07, "logits/chosen": 0.5852566361427307, "logits/rejected": 1.34577214717865, "logps/chosen": -346.68267822265625, "logps/rejected": -395.8317565917969, "loss": 0.5068, "rewards/accuracies": 0.731249988079071, "rewards/chosen": -0.6660369634628296, "rewards/margins": 0.7894813418388367, "rewards/rejected": -1.4555184841156006, "step": 240 }, { "epoch": 0.5230125523012552, "grad_norm": 34.032720640859, "learning_rate": 2.7370891215954565e-07, "logits/chosen": 1.4959585666656494, "logits/rejected": 2.515637159347534, "logps/chosen": -362.95111083984375, "logps/rejected": -426.1146545410156, "loss": 0.5163, "rewards/accuracies": 0.71875, "rewards/chosen": -1.0264902114868164, "rewards/margins": 0.8269758224487305, "rewards/rejected": -1.8534657955169678, "step": 250 }, { "epoch": 0.5230125523012552, "eval_logits/chosen": 1.7619643211364746, "eval_logits/rejected": 2.488286018371582, "eval_logps/chosen": -373.8662414550781, "eval_logps/rejected": -458.7007751464844, "eval_loss": 0.5184882879257202, "eval_rewards/accuracies": 0.73828125, "eval_rewards/chosen": -1.112363576889038, "eval_rewards/margins": 0.8480234742164612, "eval_rewards/rejected": -1.9603869915008545, "eval_runtime": 103.2706, "eval_samples_per_second": 19.367, "eval_steps_per_second": 0.31, "step": 250 }, { "epoch": 0.5439330543933054, "grad_norm": 24.412359940634452, "learning_rate": 2.55479083351317e-07, "logits/chosen": 1.4174010753631592, "logits/rejected": 2.5714869499206543, "logps/chosen": -396.62353515625, "logps/rejected": -448.77813720703125, "loss": 0.5172, "rewards/accuracies": 0.6875, "rewards/chosen": -1.168769121170044, "rewards/margins": 0.7762032747268677, "rewards/rejected": -1.9449723958969116, "step": 260 }, { "epoch": 0.5648535564853556, "grad_norm": 27.094533941945745, "learning_rate": 2.3722002126275822e-07, "logits/chosen": 0.924826979637146, "logits/rejected": 2.0330166816711426, "logps/chosen": -410.7693786621094, "logps/rejected": -429.46466064453125, "loss": 0.5165, "rewards/accuracies": 0.731249988079071, "rewards/chosen": -0.9795888662338257, "rewards/margins": 0.6443194150924683, "rewards/rejected": -1.623908281326294, "step": 270 }, { "epoch": 0.5857740585774058, "grad_norm": 25.27079656796927, "learning_rate": 2.19029145890313e-07, "logits/chosen": 1.045302391052246, "logits/rejected": 1.8793681859970093, "logps/chosen": -354.340576171875, "logps/rejected": -408.4857177734375, "loss": 0.5031, "rewards/accuracies": 0.7562500238418579, "rewards/chosen": -0.960013210773468, "rewards/margins": 0.848355770111084, "rewards/rejected": -1.8083690404891968, "step": 280 }, { "epoch": 0.606694560669456, "grad_norm": 33.50688260105116, "learning_rate": 2.0100351342479216e-07, "logits/chosen": 0.9456893801689148, "logits/rejected": 2.003875255584717, "logps/chosen": -404.9028015136719, "logps/rejected": -436.6299743652344, "loss": 0.496, "rewards/accuracies": 0.800000011920929, "rewards/chosen": -0.9352100491523743, "rewards/margins": 0.8176320791244507, "rewards/rejected": -1.7528421878814697, "step": 290 }, { "epoch": 0.6276150627615062, "grad_norm": 31.055623020499947, "learning_rate": 1.8323929841460178e-07, "logits/chosen": 1.6724220514297485, "logits/rejected": 2.5003108978271484, "logps/chosen": -367.1915283203125, "logps/rejected": -454.2134704589844, "loss": 0.5018, "rewards/accuracies": 0.6875, "rewards/chosen": -1.1671972274780273, "rewards/margins": 0.8203157186508179, "rewards/rejected": -1.9875129461288452, "step": 300 }, { "epoch": 0.6276150627615062, "eval_logits/chosen": 1.3198010921478271, "eval_logits/rejected": 2.090496778488159, "eval_logps/chosen": -355.89239501953125, "eval_logps/rejected": -443.9043884277344, "eval_loss": 0.5108399987220764, "eval_rewards/accuracies": 0.7578125, "eval_rewards/chosen": -0.9326243996620178, "eval_rewards/margins": 0.879798412322998, "eval_rewards/rejected": -1.8124228715896606, "eval_runtime": 103.5061, "eval_samples_per_second": 19.323, "eval_steps_per_second": 0.309, "step": 300 }, { "epoch": 0.6485355648535565, "grad_norm": 26.914602226083172, "learning_rate": 1.6583128063291573e-07, "logits/chosen": 1.2696157693862915, "logits/rejected": 2.2371857166290283, "logps/chosen": -391.40081787109375, "logps/rejected": -433.94384765625, "loss": 0.5165, "rewards/accuracies": 0.762499988079071, "rewards/chosen": -0.8740479350090027, "rewards/margins": 0.8631827235221863, "rewards/rejected": -1.737230658531189, "step": 310 }, { "epoch": 0.6694560669456067, "grad_norm": 23.39436395049965, "learning_rate": 1.488723393865766e-07, "logits/chosen": 0.8202553987503052, "logits/rejected": 1.8485848903656006, "logps/chosen": -368.58477783203125, "logps/rejected": -417.188720703125, "loss": 0.5097, "rewards/accuracies": 0.7749999761581421, "rewards/chosen": -0.8171142339706421, "rewards/margins": 0.8511916995048523, "rewards/rejected": -1.6683059930801392, "step": 320 }, { "epoch": 0.6903765690376569, "grad_norm": 24.464796036697045, "learning_rate": 1.3245295796480788e-07, "logits/chosen": 0.7323199510574341, "logits/rejected": 1.601609230041504, "logps/chosen": -404.5626525878906, "logps/rejected": -442.4979553222656, "loss": 0.5021, "rewards/accuracies": 0.737500011920929, "rewards/chosen": -1.0547947883605957, "rewards/margins": 0.6202768087387085, "rewards/rejected": -1.6750714778900146, "step": 330 }, { "epoch": 0.7112970711297071, "grad_norm": 28.602197351940898, "learning_rate": 1.1666074087171627e-07, "logits/chosen": 1.1269073486328125, "logits/rejected": 2.291718006134033, "logps/chosen": -386.19696044921875, "logps/rejected": -446.26080322265625, "loss": 0.51, "rewards/accuracies": 0.7562500238418579, "rewards/chosen": -1.0688059329986572, "rewards/margins": 0.8116031885147095, "rewards/rejected": -1.8804088830947876, "step": 340 }, { "epoch": 0.7322175732217573, "grad_norm": 27.032175742818502, "learning_rate": 1.0157994641835734e-07, "logits/chosen": 1.3312422037124634, "logits/rejected": 2.5307185649871826, "logps/chosen": -407.82183837890625, "logps/rejected": -435.91339111328125, "loss": 0.4999, "rewards/accuracies": 0.768750011920929, "rewards/chosen": -1.1246039867401123, "rewards/margins": 0.9464033246040344, "rewards/rejected": -2.071007490158081, "step": 350 }, { "epoch": 0.7322175732217573, "eval_logits/chosen": 1.2352690696716309, "eval_logits/rejected": 2.0403130054473877, "eval_logps/chosen": -366.191650390625, "eval_logps/rejected": -457.5764465332031, "eval_loss": 0.5094150900840759, "eval_rewards/accuracies": 0.74609375, "eval_rewards/chosen": -1.035617470741272, "eval_rewards/margins": 0.9135259389877319, "eval_rewards/rejected": -1.949143409729004, "eval_runtime": 101.7074, "eval_samples_per_second": 19.664, "eval_steps_per_second": 0.315, "step": 350 }, { "epoch": 0.7531380753138075, "grad_norm": 27.108638768821233, "learning_rate": 8.729103716819111e-08, "logits/chosen": 1.3057953119277954, "logits/rejected": 2.3061721324920654, "logps/chosen": -390.41973876953125, "logps/rejected": -483.9664001464844, "loss": 0.4746, "rewards/accuracies": 0.75, "rewards/chosen": -1.1125240325927734, "rewards/margins": 0.8978655934333801, "rewards/rejected": -2.0103893280029297, "step": 360 }, { "epoch": 0.7740585774058577, "grad_norm": 23.348973436681856, "learning_rate": 7.387025063449081e-08, "logits/chosen": 1.0233042240142822, "logits/rejected": 1.8939619064331055, "logps/chosen": -426.7222595214844, "logps/rejected": -498.1934509277344, "loss": 0.5237, "rewards/accuracies": 0.793749988079071, "rewards/chosen": -0.9773656725883484, "rewards/margins": 1.0110007524490356, "rewards/rejected": -1.9883663654327393, "step": 370 }, { "epoch": 0.7949790794979079, "grad_norm": 23.77690961130278, "learning_rate": 6.138919252022435e-08, "logits/chosen": 0.7093733549118042, "logits/rejected": 1.795745849609375, "logps/chosen": -405.4716796875, "logps/rejected": -439.33441162109375, "loss": 0.5086, "rewards/accuracies": 0.7250000238418579, "rewards/chosen": -0.994636058807373, "rewards/margins": 0.8113524317741394, "rewards/rejected": -1.8059885501861572, "step": 380 }, { "epoch": 0.8158995815899581, "grad_norm": 24.44927335880225, "learning_rate": 4.991445467064689e-08, "logits/chosen": 1.5607655048370361, "logits/rejected": 1.7856521606445312, "logps/chosen": -352.96942138671875, "logps/rejected": -442.06964111328125, "loss": 0.5034, "rewards/accuracies": 0.7250000238418579, "rewards/chosen": -1.0203418731689453, "rewards/margins": 0.8515297770500183, "rewards/rejected": -1.8718715906143188, "step": 390 }, { "epoch": 0.8368200836820083, "grad_norm": 22.777855221169585, "learning_rate": 3.9507259776993954e-08, "logits/chosen": 0.8675743341445923, "logits/rejected": 2.180140972137451, "logps/chosen": -412.081787109375, "logps/rejected": -439.1665954589844, "loss": 0.4966, "rewards/accuracies": 0.7562500238418579, "rewards/chosen": -0.9012048840522766, "rewards/margins": 0.8984438180923462, "rewards/rejected": -1.799648642539978, "step": 400 }, { "epoch": 0.8368200836820083, "eval_logits/chosen": 1.2642048597335815, "eval_logits/rejected": 2.022581100463867, "eval_logps/chosen": -361.9197998046875, "eval_logps/rejected": -454.93212890625, "eval_loss": 0.5065781474113464, "eval_rewards/accuracies": 0.7578125, "eval_rewards/chosen": -0.9928989410400391, "eval_rewards/margins": 0.9298015236854553, "eval_rewards/rejected": -1.92270028591156, "eval_runtime": 103.4115, "eval_samples_per_second": 19.34, "eval_steps_per_second": 0.309, "step": 400 }, { "epoch": 0.8577405857740585, "grad_norm": 41.05467932326275, "learning_rate": 3.022313472693447e-08, "logits/chosen": 1.1478136777877808, "logits/rejected": 2.6675543785095215, "logps/chosen": -409.5799865722656, "logps/rejected": -468.4205627441406, "loss": 0.4807, "rewards/accuracies": 0.768750011920929, "rewards/chosen": -0.9972062110900879, "rewards/margins": 1.0250569581985474, "rewards/rejected": -2.0222630500793457, "step": 410 }, { "epoch": 0.8786610878661087, "grad_norm": 29.37147689145298, "learning_rate": 2.2111614344599684e-08, "logits/chosen": 1.628156304359436, "logits/rejected": 2.2244324684143066, "logps/chosen": -386.01971435546875, "logps/rejected": -460.0570373535156, "loss": 0.5114, "rewards/accuracies": 0.737500011920929, "rewards/chosen": -1.121294617652893, "rewards/margins": 0.8227565884590149, "rewards/rejected": -1.9440511465072632, "step": 420 }, { "epoch": 0.899581589958159, "grad_norm": 36.49027990223903, "learning_rate": 1.521597710086439e-08, "logits/chosen": 1.1463916301727295, "logits/rejected": 2.26253342628479, "logps/chosen": -402.85162353515625, "logps/rejected": -490.9013671875, "loss": 0.4991, "rewards/accuracies": 0.8125, "rewards/chosen": -1.1392345428466797, "rewards/margins": 1.0718799829483032, "rewards/rejected": -2.2111144065856934, "step": 430 }, { "epoch": 0.9205020920502092, "grad_norm": 25.14386947872744, "learning_rate": 9.57301420397924e-09, "logits/chosen": 1.6257444620132446, "logits/rejected": 2.745250701904297, "logps/chosen": -375.26214599609375, "logps/rejected": -446.528076171875, "loss": 0.4996, "rewards/accuracies": 0.762499988079071, "rewards/chosen": -1.0888123512268066, "rewards/margins": 0.9304329752922058, "rewards/rejected": -2.019245147705078, "step": 440 }, { "epoch": 0.9414225941422594, "grad_norm": 25.676457242831784, "learning_rate": 5.212833302556258e-09, "logits/chosen": 1.4634435176849365, "logits/rejected": 1.8727718591690063, "logps/chosen": -356.43170166015625, "logps/rejected": -456.1715393066406, "loss": 0.5198, "rewards/accuracies": 0.7875000238418579, "rewards/chosen": -1.0714185237884521, "rewards/margins": 0.8603706359863281, "rewards/rejected": -1.9317893981933594, "step": 450 }, { "epoch": 0.9414225941422594, "eval_logits/chosen": 1.3119983673095703, "eval_logits/rejected": 2.1024117469787598, "eval_logps/chosen": -363.6942138671875, "eval_logps/rejected": -458.2783508300781, "eval_loss": 0.505741536617279, "eval_rewards/accuracies": 0.76171875, "eval_rewards/chosen": -1.010642647743225, "eval_rewards/margins": 0.9455199241638184, "eval_rewards/rejected": -1.9561628103256226, "eval_runtime": 101.6569, "eval_samples_per_second": 19.674, "eval_steps_per_second": 0.315, "step": 450 }, { "epoch": 0.9623430962343096, "grad_norm": 26.320704397497966, "learning_rate": 2.158697848236607e-09, "logits/chosen": 1.4335782527923584, "logits/rejected": 2.41444730758667, "logps/chosen": -418.5895080566406, "logps/rejected": -466.94525146484375, "loss": 0.4754, "rewards/accuracies": 0.768750011920929, "rewards/chosen": -1.0504125356674194, "rewards/margins": 0.8714386820793152, "rewards/rejected": -1.921851396560669, "step": 460 }, { "epoch": 0.9832635983263598, "grad_norm": 24.390801418724564, "learning_rate": 4.269029751107489e-10, "logits/chosen": 1.4916856288909912, "logits/rejected": 2.8449251651763916, "logps/chosen": -409.98480224609375, "logps/rejected": -462.52374267578125, "loss": 0.4863, "rewards/accuracies": 0.7562500238418579, "rewards/chosen": -1.0590018033981323, "rewards/margins": 0.990155816078186, "rewards/rejected": -2.0491576194763184, "step": 470 }, { "epoch": 1.0, "step": 478, "total_flos": 0.0, "train_loss": 0.5417051395113, "train_runtime": 12463.9447, "train_samples_per_second": 4.905, "train_steps_per_second": 0.038 } ], "logging_steps": 10, "max_steps": 478, "num_input_tokens_seen": 0, "num_train_epochs": 1, "save_steps": 100, "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 }