{ "best_metric": null, "best_model_checkpoint": null, "epoch": 0.999253545658124, "eval_steps": 100, "global_step": 753, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0013270299411130464, "grad_norm": 1.7421875, "learning_rate": 6.578947368421053e-08, "logits/chosen": -3.0980052947998047, "logits/rejected": -3.127007007598877, "logps/chosen": -425.274169921875, "logps/rejected": -373.2780456542969, "loss": 0.6931, "rewards/accuracies": 0.0, "rewards/chosen": 0.0, "rewards/margins": 0.0, "rewards/rejected": 0.0, "step": 1 }, { "epoch": 0.013270299411130464, "grad_norm": 1.765625, "learning_rate": 6.578947368421053e-07, "logits/chosen": -3.072850227355957, "logits/rejected": -3.0915706157684326, "logps/chosen": -421.3843994140625, "logps/rejected": -350.215087890625, "loss": 0.6919, "rewards/accuracies": 0.6041666865348816, "rewards/chosen": 0.0024290236178785563, "rewards/margins": 0.0028885826468467712, "rewards/rejected": -0.00045955937821418047, "step": 10 }, { "epoch": 0.026540598822260928, "grad_norm": 1.6875, "learning_rate": 1.3157894736842106e-06, "logits/chosen": -3.0872464179992676, "logits/rejected": -3.0978829860687256, "logps/chosen": -352.447021484375, "logps/rejected": -364.5257873535156, "loss": 0.6816, "rewards/accuracies": 0.956250011920929, "rewards/chosen": 0.015185330994427204, "rewards/margins": 0.021979082375764847, "rewards/rejected": -0.006793751381337643, "step": 20 }, { "epoch": 0.039810898233391394, "grad_norm": 1.53125, "learning_rate": 1.973684210526316e-06, "logits/chosen": -3.055328130722046, "logits/rejected": -3.0872576236724854, "logps/chosen": -374.4645080566406, "logps/rejected": -364.3621520996094, "loss": 0.6547, "rewards/accuracies": 0.9937499761581421, "rewards/chosen": 0.05001773685216904, "rewards/margins": 0.07953666895627975, "rewards/rejected": -0.029518935829401016, "step": 30 }, { "epoch": 0.053081197644521856, "grad_norm": 1.6640625, "learning_rate": 2.631578947368421e-06, "logits/chosen": -3.054077625274658, "logits/rejected": -3.090534210205078, "logps/chosen": -394.09552001953125, "logps/rejected": -394.32818603515625, "loss": 0.611, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": 0.0969473272562027, "rewards/margins": 0.17953996360301971, "rewards/rejected": -0.08259265124797821, "step": 40 }, { "epoch": 0.06635149705565231, "grad_norm": 1.796875, "learning_rate": 3.289473684210527e-06, "logits/chosen": -3.0039563179016113, "logits/rejected": -3.0520451068878174, "logps/chosen": -376.13897705078125, "logps/rejected": -405.67303466796875, "loss": 0.5412, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": 0.12179754674434662, "rewards/margins": 0.3453168272972107, "rewards/rejected": -0.22351928055286407, "step": 50 }, { "epoch": 0.07962179646678279, "grad_norm": 1.7890625, "learning_rate": 3.947368421052632e-06, "logits/chosen": -2.974642515182495, "logits/rejected": -3.0201048851013184, "logps/chosen": -418.08135986328125, "logps/rejected": -455.23846435546875, "loss": 0.4179, "rewards/accuracies": 0.987500011920929, "rewards/chosen": 0.048082564026117325, "rewards/margins": 0.6676187515258789, "rewards/rejected": -0.6195362210273743, "step": 60 }, { "epoch": 0.09289209587791325, "grad_norm": 1.34375, "learning_rate": 4.605263157894737e-06, "logits/chosen": -2.86934757232666, "logits/rejected": -2.9434354305267334, "logps/chosen": -400.07568359375, "logps/rejected": -482.0399475097656, "loss": 0.3225, "rewards/accuracies": 0.981249988079071, "rewards/chosen": -0.1442013680934906, "rewards/margins": 1.0631463527679443, "rewards/rejected": -1.2073477506637573, "step": 70 }, { "epoch": 0.10616239528904371, "grad_norm": 1.359375, "learning_rate": 4.9995693346469565e-06, "logits/chosen": -2.6817970275878906, "logits/rejected": -2.778237819671631, "logps/chosen": -460.60394287109375, "logps/rejected": -621.35498046875, "loss": 0.2331, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -0.620966374874115, "rewards/margins": 1.6919386386871338, "rewards/rejected": -2.3129050731658936, "step": 80 }, { "epoch": 0.11943269470017417, "grad_norm": 1.8671875, "learning_rate": 4.994726053293703e-06, "logits/chosen": -2.5432372093200684, "logits/rejected": -2.636983633041382, "logps/chosen": -504.8564453125, "logps/rejected": -734.1746826171875, "loss": 0.1809, "rewards/accuracies": 0.981249988079071, "rewards/chosen": -1.099843978881836, "rewards/margins": 2.6429855823516846, "rewards/rejected": -3.7428295612335205, "step": 90 }, { "epoch": 0.13270299411130462, "grad_norm": 1.140625, "learning_rate": 4.984511621268103e-06, "logits/chosen": -2.420698642730713, "logits/rejected": -2.541872501373291, "logps/chosen": -544.0470581054688, "logps/rejected": -866.2991943359375, "loss": 0.1029, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -1.5227975845336914, "rewards/margins": 3.385746479034424, "rewards/rejected": -4.908544063568115, "step": 100 }, { "epoch": 0.13270299411130462, "eval_logits/chosen": -2.3704655170440674, "eval_logits/rejected": -2.4775025844573975, "eval_logps/chosen": -579.9114990234375, "eval_logps/rejected": -892.3373413085938, "eval_loss": 0.10992002487182617, "eval_rewards/accuracies": 0.9679104685783386, "eval_rewards/chosen": -1.8067275285720825, "eval_rewards/margins": 3.5615758895874023, "eval_rewards/rejected": -5.368303298950195, "eval_runtime": 828.869, "eval_samples_per_second": 6.465, "eval_steps_per_second": 1.617, "step": 100 }, { "epoch": 0.1459732935224351, "grad_norm": 1.2421875, "learning_rate": 4.968948030264743e-06, "logits/chosen": -2.349208354949951, "logits/rejected": -2.439763069152832, "logps/chosen": -572.8814697265625, "logps/rejected": -938.94091796875, "loss": 0.085, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -1.6822025775909424, "rewards/margins": 4.082047939300537, "rewards/rejected": -5.764250755310059, "step": 110 }, { "epoch": 0.15924359293356558, "grad_norm": 2.046875, "learning_rate": 4.948068788729238e-06, "logits/chosen": -2.156416177749634, "logits/rejected": -2.215620756149292, "logps/chosen": -626.0386962890625, "logps/rejected": -1036.8035888671875, "loss": 0.0874, "rewards/accuracies": 0.9624999761581421, "rewards/chosen": -2.1521973609924316, "rewards/margins": 4.6609344482421875, "rewards/rejected": -6.813131809234619, "step": 120 }, { "epoch": 0.17251389234469602, "grad_norm": 1.8515625, "learning_rate": 4.921918849714475e-06, "logits/chosen": -2.1903834342956543, "logits/rejected": -2.2045347690582275, "logps/chosen": -673.7794189453125, "logps/rejected": -1226.3961181640625, "loss": 0.0559, "rewards/accuracies": 0.9937499761581421, "rewards/chosen": -2.5693655014038086, "rewards/margins": 5.772281646728516, "rewards/rejected": -8.341647148132324, "step": 130 }, { "epoch": 0.1857841917558265, "grad_norm": 1.0078125, "learning_rate": 4.890554514096592e-06, "logits/chosen": -2.1850454807281494, "logits/rejected": -2.19503116607666, "logps/chosen": -688.4675903320312, "logps/rejected": -1202.0238037109375, "loss": 0.0699, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -2.7659270763397217, "rewards/margins": 5.731719970703125, "rewards/rejected": -8.497647285461426, "step": 140 }, { "epoch": 0.19905449116695695, "grad_norm": 4.96875, "learning_rate": 4.854043309359063e-06, "logits/chosen": -2.045557737350464, "logits/rejected": -2.022369146347046, "logps/chosen": -797.4768676757812, "logps/rejected": -1449.0230712890625, "loss": 0.0569, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -3.9922683238983154, "rewards/margins": 6.881680488586426, "rewards/rejected": -10.87394905090332, "step": 150 }, { "epoch": 0.21232479057808742, "grad_norm": 2.5, "learning_rate": 4.8124638442058856e-06, "logits/chosen": -2.0948424339294434, "logits/rejected": -2.136207103729248, "logps/chosen": -731.327392578125, "logps/rejected": -1373.2447509765625, "loss": 0.05, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -3.1236228942871094, "rewards/margins": 6.655406951904297, "rewards/rejected": -9.77902889251709, "step": 160 }, { "epoch": 0.22559508998921787, "grad_norm": 1.265625, "learning_rate": 4.765905639316861e-06, "logits/chosen": -1.9219977855682373, "logits/rejected": -1.9046119451522827, "logps/chosen": -720.5792236328125, "logps/rejected": -1352.9737548828125, "loss": 0.0472, "rewards/accuracies": 0.9937499761581421, "rewards/chosen": -3.072010040283203, "rewards/margins": 6.776447296142578, "rewards/rejected": -9.848457336425781, "step": 170 }, { "epoch": 0.23886538940034835, "grad_norm": 2.234375, "learning_rate": 4.7144689346093814e-06, "logits/chosen": -1.9168952703475952, "logits/rejected": -1.8547157049179077, "logps/chosen": -825.1336059570312, "logps/rejected": -1616.734375, "loss": 0.0578, "rewards/accuracies": 0.9624999761581421, "rewards/chosen": -4.402952671051025, "rewards/margins": 8.200037956237793, "rewards/rejected": -12.602991104125977, "step": 180 }, { "epoch": 0.2521356888114788, "grad_norm": 2.296875, "learning_rate": 4.65826447342166e-06, "logits/chosen": -1.9839226007461548, "logits/rejected": -1.9680023193359375, "logps/chosen": -803.8746948242188, "logps/rejected": -1493.1898193359375, "loss": 0.0497, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -3.822723865509033, "rewards/margins": 7.312263488769531, "rewards/rejected": -11.134986877441406, "step": 190 }, { "epoch": 0.26540598822260925, "grad_norm": 8.8125, "learning_rate": 4.597413264082086e-06, "logits/chosen": -1.9621204137802124, "logits/rejected": -1.9449526071548462, "logps/chosen": -757.6912841796875, "logps/rejected": -1455.258544921875, "loss": 0.042, "rewards/accuracies": 0.9750000238418579, "rewards/chosen": -3.6233069896698, "rewards/margins": 7.341341495513916, "rewards/rejected": -10.964648246765137, "step": 200 }, { "epoch": 0.26540598822260925, "eval_logits/chosen": -1.9894516468048096, "eval_logits/rejected": -1.9743856191635132, "eval_logps/chosen": -750.5289306640625, "eval_logps/rejected": -1423.288330078125, "eval_loss": 0.04295578598976135, "eval_rewards/accuracies": 0.9828358292579651, "eval_rewards/chosen": -3.5129029750823975, "eval_rewards/margins": 7.164910793304443, "eval_rewards/rejected": -10.677813529968262, "eval_runtime": 830.8607, "eval_samples_per_second": 6.45, "eval_steps_per_second": 1.613, "step": 200 }, { "epoch": 0.27867628763373975, "grad_norm": 9.6875, "learning_rate": 4.5320463193780265e-06, "logits/chosen": -1.9735777378082275, "logits/rejected": -1.9742670059204102, "logps/chosen": -789.7581176757812, "logps/rejected": -1441.5927734375, "loss": 0.0416, "rewards/accuracies": 0.9937499761581421, "rewards/chosen": -3.7721946239471436, "rewards/margins": 6.881577968597412, "rewards/rejected": -10.653772354125977, "step": 210 }, { "epoch": 0.2919465870448702, "grad_norm": 2.71875, "learning_rate": 4.462304374485005e-06, "logits/chosen": -1.910851240158081, "logits/rejected": -1.8877220153808594, "logps/chosen": -795.4382934570312, "logps/rejected": -1552.471923828125, "loss": 0.0435, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -4.075817584991455, "rewards/margins": 7.9385833740234375, "rewards/rejected": -12.014400482177734, "step": 220 }, { "epoch": 0.30521688645600065, "grad_norm": 0.9296875, "learning_rate": 4.388337583963563e-06, "logits/chosen": -1.8420207500457764, "logits/rejected": -1.7979652881622314, "logps/chosen": -831.4200439453125, "logps/rejected": -1719.750244140625, "loss": 0.0253, "rewards/accuracies": 1.0, "rewards/chosen": -4.229403018951416, "rewards/margins": 9.30135440826416, "rewards/rejected": -13.530756950378418, "step": 230 }, { "epoch": 0.31848718586713115, "grad_norm": 1.15625, "learning_rate": 4.310305198476161e-06, "logits/chosen": -1.813542366027832, "logits/rejected": -1.7659708261489868, "logps/chosen": -770.3541259765625, "logps/rejected": -1698.886474609375, "loss": 0.0336, "rewards/accuracies": 0.9937499761581421, "rewards/chosen": -3.893308162689209, "rewards/margins": 9.639238357543945, "rewards/rejected": -13.532546997070312, "step": 240 }, { "epoch": 0.3317574852782616, "grad_norm": 2.03125, "learning_rate": 4.228375221920147e-06, "logits/chosen": -1.8265107870101929, "logits/rejected": -1.765747308731079, "logps/chosen": -721.0618896484375, "logps/rejected": -1561.09033203125, "loss": 0.0269, "rewards/accuracies": 0.9937499761581421, "rewards/chosen": -3.446256160736084, "rewards/margins": 8.8670015335083, "rewards/rejected": -12.313258171081543, "step": 250 }, { "epoch": 0.34502778468939205, "grad_norm": 3.875, "learning_rate": 4.142724049715005e-06, "logits/chosen": -1.7385963201522827, "logits/rejected": -1.5829213857650757, "logps/chosen": -922.6043701171875, "logps/rejected": -2136.38525390625, "loss": 0.0346, "rewards/accuracies": 1.0, "rewards/chosen": -4.981525897979736, "rewards/margins": 11.962621688842773, "rewards/rejected": -16.94414710998535, "step": 260 }, { "epoch": 0.3582980841005225, "grad_norm": 5.5, "learning_rate": 4.053536089022624e-06, "logits/chosen": -1.841082215309143, "logits/rejected": -1.7555363178253174, "logps/chosen": -756.6123657226562, "logps/rejected": -1619.0562744140625, "loss": 0.0434, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -3.6720402240753174, "rewards/margins": 8.794347763061523, "rewards/rejected": -12.466386795043945, "step": 270 }, { "epoch": 0.371568383511653, "grad_norm": 3.1875, "learning_rate": 3.961003361718272e-06, "logits/chosen": -1.8635780811309814, "logits/rejected": -1.7543712854385376, "logps/chosen": -745.083251953125, "logps/rejected": -1585.915283203125, "loss": 0.0333, "rewards/accuracies": 0.9937499761581421, "rewards/chosen": -3.244401216506958, "rewards/margins": 8.643343925476074, "rewards/rejected": -11.887744903564453, "step": 280 }, { "epoch": 0.38483868292278345, "grad_norm": 16.5, "learning_rate": 3.8653250909670815e-06, "logits/chosen": -1.6911243200302124, "logits/rejected": -1.550756573677063, "logps/chosen": -840.0848388671875, "logps/rejected": -1973.161376953125, "loss": 0.0493, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -4.447511196136475, "rewards/margins": 11.780040740966797, "rewards/rejected": -16.227550506591797, "step": 290 }, { "epoch": 0.3981089823339139, "grad_norm": 1.5859375, "learning_rate": 3.7667072722961363e-06, "logits/chosen": -1.8192800283432007, "logits/rejected": -1.7625439167022705, "logps/chosen": -772.19189453125, "logps/rejected": -1698.5767822265625, "loss": 0.0278, "rewards/accuracies": 0.9937499761581421, "rewards/chosen": -3.4313864707946777, "rewards/margins": 9.797091484069824, "rewards/rejected": -13.228475570678711, "step": 300 }, { "epoch": 0.3981089823339139, "eval_logits/chosen": -1.8191460371017456, "eval_logits/rejected": -1.7454465627670288, "eval_logps/chosen": -772.5892944335938, "eval_logps/rejected": -1707.0360107421875, "eval_loss": 0.03441401198506355, "eval_rewards/accuracies": 0.9828358292579651, "eval_rewards/chosen": -3.733506202697754, "eval_rewards/margins": 9.781785011291504, "eval_rewards/rejected": -13.51529312133789, "eval_runtime": 829.9568, "eval_samples_per_second": 6.457, "eval_steps_per_second": 1.615, "step": 300 }, { "epoch": 0.41137928174504435, "grad_norm": 6.5625, "learning_rate": 3.665362230085646e-06, "logits/chosen": -1.755239725112915, "logits/rejected": -1.6191673278808594, "logps/chosen": -831.6585693359375, "logps/rejected": -1947.4847412109375, "loss": 0.0316, "rewards/accuracies": 0.9937499761581421, "rewards/chosen": -4.389147758483887, "rewards/margins": 11.429269790649414, "rewards/rejected": -15.8184175491333, "step": 310 }, { "epoch": 0.42464958115617485, "grad_norm": 2.9375, "learning_rate": 3.5615081604340905e-06, "logits/chosen": -1.812110185623169, "logits/rejected": -1.6317228078842163, "logps/chosen": -835.9284057617188, "logps/rejected": -2101.62646484375, "loss": 0.04, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -4.403895378112793, "rewards/margins": 12.521451950073242, "rewards/rejected": -16.92534828186035, "step": 320 }, { "epoch": 0.4379198805673053, "grad_norm": 1.3125, "learning_rate": 3.4553686613815436e-06, "logits/chosen": -1.9260807037353516, "logits/rejected": -1.830836534500122, "logps/chosen": -688.6129760742188, "logps/rejected": -1593.781982421875, "loss": 0.0215, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -3.1708381175994873, "rewards/margins": 8.699010848999023, "rewards/rejected": -11.869850158691406, "step": 330 }, { "epoch": 0.45119017997843575, "grad_norm": 1.8515625, "learning_rate": 3.3471722515025986e-06, "logits/chosen": -1.8425785303115845, "logits/rejected": -1.7366511821746826, "logps/chosen": -812.2515258789062, "logps/rejected": -1760.351806640625, "loss": 0.0311, "rewards/accuracies": 0.981249988079071, "rewards/chosen": -4.064314842224121, "rewards/margins": 10.192426681518555, "rewards/rejected": -14.256741523742676, "step": 340 }, { "epoch": 0.46446047938956625, "grad_norm": 1.25, "learning_rate": 3.2371518779053744e-06, "logits/chosen": -1.9980113506317139, "logits/rejected": -1.9302339553833008, "logps/chosen": -760.2137451171875, "logps/rejected": -1788.198486328125, "loss": 0.0314, "rewards/accuracies": 1.0, "rewards/chosen": -3.4233691692352295, "rewards/margins": 10.155590057373047, "rewards/rejected": -13.578959465026855, "step": 350 }, { "epoch": 0.4777307788006967, "grad_norm": 2.125, "learning_rate": 3.1255444146958845e-06, "logits/chosen": -1.8775854110717773, "logits/rejected": -1.7533352375030518, "logps/chosen": -726.8655395507812, "logps/rejected": -1770.5924072265625, "loss": 0.0329, "rewards/accuracies": 0.981249988079071, "rewards/chosen": -3.306746244430542, "rewards/margins": 10.716817855834961, "rewards/rejected": -14.023564338684082, "step": 360 }, { "epoch": 0.49100107821182715, "grad_norm": 2.578125, "learning_rate": 3.0125901529875612e-06, "logits/chosen": -1.854692816734314, "logits/rejected": -1.7414394617080688, "logps/chosen": -820.2403564453125, "logps/rejected": -1938.7720947265625, "loss": 0.0401, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -3.7310116291046143, "rewards/margins": 12.33505630493164, "rewards/rejected": -16.06606674194336, "step": 370 }, { "epoch": 0.5042713776229576, "grad_norm": 1.4765625, "learning_rate": 2.898532283553963e-06, "logits/chosen": -1.8446468114852905, "logits/rejected": -1.6768661737442017, "logps/chosen": -750.1494750976562, "logps/rejected": -1983.754638671875, "loss": 0.0151, "rewards/accuracies": 0.9937499761581421, "rewards/chosen": -3.6851119995117188, "rewards/margins": 12.596293449401855, "rewards/rejected": -16.281402587890625, "step": 380 }, { "epoch": 0.517541677034088, "grad_norm": 3.78125, "learning_rate": 2.783616373238507e-06, "logits/chosen": -1.808468222618103, "logits/rejected": -1.6193923950195312, "logps/chosen": -817.4727783203125, "logps/rejected": -1965.7611083984375, "loss": 0.0232, "rewards/accuracies": 1.0, "rewards/chosen": -4.224337577819824, "rewards/margins": 11.751169204711914, "rewards/rejected": -15.975506782531738, "step": 390 }, { "epoch": 0.5308119764452185, "grad_norm": 1.3359375, "learning_rate": 2.6680898362485126e-06, "logits/chosen": -1.9052226543426514, "logits/rejected": -1.790858507156372, "logps/chosen": -760.7432861328125, "logps/rejected": -1788.7734375, "loss": 0.0223, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -3.564972400665283, "rewards/margins": 10.65664291381836, "rewards/rejected": -14.2216157913208, "step": 400 }, { "epoch": 0.5308119764452185, "eval_logits/chosen": -1.918375015258789, "eval_logits/rejected": -1.8019860982894897, "eval_logps/chosen": -764.7831420898438, "eval_logps/rejected": -1732.62890625, "eval_loss": 0.030813412740826607, "eval_rewards/accuracies": 0.9858208894729614, "eval_rewards/chosen": -3.655445098876953, "eval_rewards/margins": 10.11577320098877, "eval_rewards/rejected": -13.771217346191406, "eval_runtime": 830.1888, "eval_samples_per_second": 6.455, "eval_steps_per_second": 1.614, "step": 400 }, { "epoch": 0.544082275856349, "grad_norm": 0.76953125, "learning_rate": 2.55220140147187e-06, "logits/chosen": -1.919586420059204, "logits/rejected": -1.729554533958435, "logps/chosen": -761.3343505859375, "logps/rejected": -1788.455322265625, "loss": 0.0345, "rewards/accuracies": 0.9937499761581421, "rewards/chosen": -3.7783074378967285, "rewards/margins": 10.44767951965332, "rewards/rejected": -14.225985527038574, "step": 410 }, { "epoch": 0.5573525752674795, "grad_norm": 0.7890625, "learning_rate": 2.4362005769631985e-06, "logits/chosen": -1.8321539163589478, "logits/rejected": -1.664629340171814, "logps/chosen": -864.9075317382812, "logps/rejected": -2111.5986328125, "loss": 0.0269, "rewards/accuracies": 0.9937499761581421, "rewards/chosen": -4.227338790893555, "rewards/margins": 13.148414611816406, "rewards/rejected": -17.375751495361328, "step": 420 }, { "epoch": 0.57062287467861, "grad_norm": 0.6796875, "learning_rate": 2.320337112752459e-06, "logits/chosen": -1.767311692237854, "logits/rejected": -1.6259944438934326, "logps/chosen": -846.6857299804688, "logps/rejected": -1985.771240234375, "loss": 0.0251, "rewards/accuracies": 1.0, "rewards/chosen": -4.594606399536133, "rewards/margins": 11.99472427368164, "rewards/rejected": -16.58932876586914, "step": 430 }, { "epoch": 0.5838931740897404, "grad_norm": 2.21875, "learning_rate": 2.2048604631325896e-06, "logits/chosen": -1.9071296453475952, "logits/rejected": -1.7615699768066406, "logps/chosen": -740.1785888671875, "logps/rejected": -1816.4349365234375, "loss": 0.0218, "rewards/accuracies": 1.0, "rewards/chosen": -3.5860729217529297, "rewards/margins": 11.016485214233398, "rewards/rejected": -14.602559089660645, "step": 440 }, { "epoch": 0.5971634735008708, "grad_norm": 1.3359375, "learning_rate": 2.0900192495838617e-06, "logits/chosen": -1.8986167907714844, "logits/rejected": -1.749355673789978, "logps/chosen": -749.77294921875, "logps/rejected": -1782.9183349609375, "loss": 0.0303, "rewards/accuracies": 0.9937499761581421, "rewards/chosen": -3.5921592712402344, "rewards/margins": 10.537958145141602, "rewards/rejected": -14.130119323730469, "step": 450 }, { "epoch": 0.6104337729120013, "grad_norm": 5.6875, "learning_rate": 1.976060725491293e-06, "logits/chosen": -1.8696537017822266, "logits/rejected": -1.6489883661270142, "logps/chosen": -817.318359375, "logps/rejected": -2079.50732421875, "loss": 0.0169, "rewards/accuracies": 1.0, "rewards/chosen": -4.171055316925049, "rewards/margins": 12.784004211425781, "rewards/rejected": -16.955059051513672, "step": 460 }, { "epoch": 0.6237040723231317, "grad_norm": 0.466796875, "learning_rate": 1.8632302438075618e-06, "logits/chosen": -1.7880550622940063, "logits/rejected": -1.5018467903137207, "logps/chosen": -904.4384765625, "logps/rejected": -2307.56103515625, "loss": 0.0164, "rewards/accuracies": 0.9937499761581421, "rewards/chosen": -4.95499324798584, "rewards/margins": 14.273486137390137, "rewards/rejected": -19.22848129272461, "step": 470 }, { "epoch": 0.6369743717342623, "grad_norm": 0.8046875, "learning_rate": 1.7517707288075617e-06, "logits/chosen": -1.7752134799957275, "logits/rejected": -1.5753581523895264, "logps/chosen": -829.3978271484375, "logps/rejected": -2190.319091796875, "loss": 0.0281, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -4.3248748779296875, "rewards/margins": 13.842320442199707, "rewards/rejected": -18.16719627380371, "step": 480 }, { "epoch": 0.6502446711453927, "grad_norm": 5.03125, "learning_rate": 1.6419221530719062e-06, "logits/chosen": -1.8694307804107666, "logits/rejected": -1.6351697444915771, "logps/chosen": -807.4894409179688, "logps/rejected": -2210.259765625, "loss": 0.0224, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -4.219082832336426, "rewards/margins": 13.85973834991455, "rewards/rejected": -18.078821182250977, "step": 490 }, { "epoch": 0.6635149705565232, "grad_norm": 5.25, "learning_rate": 1.5339210208254345e-06, "logits/chosen": -1.796460509300232, "logits/rejected": -1.5865790843963623, "logps/chosen": -825.5130615234375, "logps/rejected": -2155.32177734375, "loss": 0.0378, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -4.26907205581665, "rewards/margins": 13.452142715454102, "rewards/rejected": -17.721214294433594, "step": 500 }, { "epoch": 0.6635149705565232, "eval_logits/chosen": -1.8649603128433228, "eval_logits/rejected": -1.6923589706420898, "eval_logps/chosen": -799.422119140625, "eval_logps/rejected": -1988.354248046875, "eval_loss": 0.029703186824917793, "eval_rewards/accuracies": 0.9850746393203735, "eval_rewards/chosen": -4.001834392547607, "eval_rewards/margins": 12.326638221740723, "eval_rewards/rejected": -16.328474044799805, "eval_runtime": 830.0926, "eval_samples_per_second": 6.456, "eval_steps_per_second": 1.614, "step": 500 }, { "epoch": 0.6767852699676536, "grad_norm": 2.53125, "learning_rate": 1.4279998587430944e-06, "logits/chosen": -1.9209420680999756, "logits/rejected": -1.7078396081924438, "logps/chosen": -800.6217041015625, "logps/rejected": -2088.8896484375, "loss": 0.0344, "rewards/accuracies": 1.0, "rewards/chosen": -3.650468349456787, "rewards/margins": 13.276763916015625, "rewards/rejected": -16.927234649658203, "step": 510 }, { "epoch": 0.6900555693787841, "grad_norm": 0.44921875, "learning_rate": 1.3243867153195033e-06, "logits/chosen": -1.9058411121368408, "logits/rejected": -1.7081434726715088, "logps/chosen": -756.5093994140625, "logps/rejected": -1919.1995849609375, "loss": 0.0323, "rewards/accuracies": 1.0, "rewards/chosen": -3.5202860832214355, "rewards/margins": 11.93966007232666, "rewards/rejected": -15.459945678710938, "step": 520 }, { "epoch": 0.7033258687899145, "grad_norm": 10.4375, "learning_rate": 1.2233046698800343e-06, "logits/chosen": -1.9104827642440796, "logits/rejected": -1.6314716339111328, "logps/chosen": -794.9130859375, "logps/rejected": -2090.015380859375, "loss": 0.0287, "rewards/accuracies": 0.9937499761581421, "rewards/chosen": -3.8667774200439453, "rewards/margins": 13.207371711730957, "rewards/rejected": -17.074146270751953, "step": 530 }, { "epoch": 0.716596168201045, "grad_norm": 1.3125, "learning_rate": 1.124971352290545e-06, "logits/chosen": -1.8662798404693604, "logits/rejected": -1.6738961935043335, "logps/chosen": -800.027587890625, "logps/rejected": -2160.337158203125, "loss": 0.0261, "rewards/accuracies": 0.981249988079071, "rewards/chosen": -4.005196571350098, "rewards/margins": 13.762578964233398, "rewards/rejected": -17.76777458190918, "step": 540 }, { "epoch": 0.7298664676121756, "grad_norm": 1.3046875, "learning_rate": 1.0295984743997911e-06, "logits/chosen": -1.8652830123901367, "logits/rejected": -1.6665403842926025, "logps/chosen": -794.8709106445312, "logps/rejected": -1995.255859375, "loss": 0.03, "rewards/accuracies": 0.9937499761581421, "rewards/chosen": -3.9191575050354004, "rewards/margins": 12.423823356628418, "rewards/rejected": -16.34298324584961, "step": 550 }, { "epoch": 0.743136767023306, "grad_norm": 0.470703125, "learning_rate": 9.37391374223355e-07, "logits/chosen": -1.8799976110458374, "logits/rejected": -1.6955276727676392, "logps/chosen": -788.1275634765625, "logps/rejected": -1998.5726318359375, "loss": 0.0196, "rewards/accuracies": 1.0, "rewards/chosen": -3.821469783782959, "rewards/margins": 12.33929443359375, "rewards/rejected": -16.160762786865234, "step": 560 }, { "epoch": 0.7564070664344364, "grad_norm": 1.6171875, "learning_rate": 8.48548573850449e-07, "logits/chosen": -1.8499343395233154, "logits/rejected": -1.6717488765716553, "logps/chosen": -778.8970947265625, "logps/rejected": -1945.9202880859375, "loss": 0.0362, "rewards/accuracies": 0.9937499761581421, "rewards/chosen": -3.6539416313171387, "rewards/margins": 12.185694694519043, "rewards/rejected": -15.839635848999023, "step": 570 }, { "epoch": 0.7696773658455669, "grad_norm": 2.46875, "learning_rate": 7.632613520254159e-07, "logits/chosen": -1.9424989223480225, "logits/rejected": -1.8162052631378174, "logps/chosen": -791.0227661132812, "logps/rejected": -1805.901123046875, "loss": 0.035, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -3.653287410736084, "rewards/margins": 11.017667770385742, "rewards/rejected": -14.670953750610352, "step": 580 }, { "epoch": 0.7829476652566973, "grad_norm": 0.94921875, "learning_rate": 6.817133323241757e-07, "logits/chosen": -1.9590984582901, "logits/rejected": -1.754612922668457, "logps/chosen": -768.0120239257812, "logps/rejected": -2016.432373046875, "loss": 0.0217, "rewards/accuracies": 0.9937499761581421, "rewards/chosen": -3.4281082153320312, "rewards/margins": 12.803730964660645, "rewards/rejected": -16.23183822631836, "step": 590 }, { "epoch": 0.7962179646678278, "grad_norm": 2.203125, "learning_rate": 6.040800878122655e-07, "logits/chosen": -1.9362680912017822, "logits/rejected": -1.8174184560775757, "logps/chosen": -807.0682373046875, "logps/rejected": -1890.33984375, "loss": 0.0352, "rewards/accuracies": 1.0, "rewards/chosen": -3.6727454662323, "rewards/margins": 11.275753021240234, "rewards/rejected": -14.948498725891113, "step": 600 }, { "epoch": 0.7962179646678278, "eval_logits/chosen": -1.8977892398834229, "eval_logits/rejected": -1.7437201738357544, "eval_logps/chosen": -780.2752075195312, "eval_logps/rejected": -1919.8118896484375, "eval_loss": 0.02784702554345131, "eval_rewards/accuracies": 0.983582079410553, "eval_rewards/chosen": -3.810366153717041, "eval_rewards/margins": 11.832680702209473, "eval_rewards/rejected": -15.643047332763672, "eval_runtime": 829.9624, "eval_samples_per_second": 6.457, "eval_steps_per_second": 1.615, "step": 600 }, { "epoch": 0.8094882640789582, "grad_norm": 1.171875, "learning_rate": 5.305287630356363e-07, "logits/chosen": -1.8633983135223389, "logits/rejected": -1.6604722738265991, "logps/chosen": -751.8145141601562, "logps/rejected": -1894.828857421875, "loss": 0.0299, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -3.7778866291046143, "rewards/margins": 11.620625495910645, "rewards/rejected": -15.39851188659668, "step": 610 }, { "epoch": 0.8227585634900887, "grad_norm": 0.37109375, "learning_rate": 4.612177141580876e-07, "logits/chosen": -1.8507238626480103, "logits/rejected": -1.6502704620361328, "logps/chosen": -739.475341796875, "logps/rejected": -1800.6304931640625, "loss": 0.028, "rewards/accuracies": 0.9937499761581421, "rewards/chosen": -3.580106735229492, "rewards/margins": 10.93178939819336, "rewards/rejected": -14.511896133422852, "step": 620 }, { "epoch": 0.8360288629012192, "grad_norm": 2.375, "learning_rate": 3.962961680200927e-07, "logits/chosen": -1.9084770679473877, "logits/rejected": -1.7507511377334595, "logps/chosen": -785.728515625, "logps/rejected": -1896.435791015625, "loss": 0.0222, "rewards/accuracies": 0.9937499761581421, "rewards/chosen": -3.7701239585876465, "rewards/margins": 11.418277740478516, "rewards/rejected": -15.18840217590332, "step": 630 }, { "epoch": 0.8492991623123497, "grad_norm": 1.96875, "learning_rate": 3.3590390085308457e-07, "logits/chosen": -1.9369093179702759, "logits/rejected": -1.782634973526001, "logps/chosen": -792.2256469726562, "logps/rejected": -1962.2239990234375, "loss": 0.0246, "rewards/accuracies": 0.981249988079071, "rewards/chosen": -3.8103549480438232, "rewards/margins": 11.93663501739502, "rewards/rejected": -15.746989250183105, "step": 640 }, { "epoch": 0.8625694617234801, "grad_norm": 0.373046875, "learning_rate": 2.801709373409248e-07, "logits/chosen": -1.8602193593978882, "logits/rejected": -1.7032759189605713, "logps/chosen": -804.9444580078125, "logps/rejected": -2019.756103515625, "loss": 0.0177, "rewards/accuracies": 0.9937499761581421, "rewards/chosen": -3.675445556640625, "rewards/margins": 12.645570755004883, "rewards/rejected": -16.321016311645508, "step": 650 }, { "epoch": 0.8758397611346106, "grad_norm": 1.90625, "learning_rate": 2.2921727067647032e-07, "logits/chosen": -1.9662708044052124, "logits/rejected": -1.7979133129119873, "logps/chosen": -745.5474243164062, "logps/rejected": -1877.61328125, "loss": 0.0226, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -3.5280399322509766, "rewards/margins": 11.784358024597168, "rewards/rejected": -15.312397956848145, "step": 660 }, { "epoch": 0.889110060545741, "grad_norm": 15.375, "learning_rate": 1.8315260421596925e-07, "logits/chosen": -1.9315414428710938, "logits/rejected": -1.7226520776748657, "logps/chosen": -760.8677978515625, "logps/rejected": -1978.8160400390625, "loss": 0.0338, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -3.70927357673645, "rewards/margins": 12.64477252960205, "rewards/rejected": -16.35404396057129, "step": 670 }, { "epoch": 0.9023803599568715, "grad_norm": 0.412109375, "learning_rate": 1.4207611528749e-07, "logits/chosen": -1.9155975580215454, "logits/rejected": -1.6965796947479248, "logps/chosen": -769.4285278320312, "logps/rejected": -1877.641845703125, "loss": 0.0319, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -3.767559766769409, "rewards/margins": 11.334938049316406, "rewards/rejected": -15.102499008178711, "step": 680 }, { "epoch": 0.9156506593680019, "grad_norm": 0.56640625, "learning_rate": 1.060762416619196e-07, "logits/chosen": -1.946947693824768, "logits/rejected": -1.7013956308364868, "logps/chosen": -765.5987548828125, "logps/rejected": -2080.13134765625, "loss": 0.0215, "rewards/accuracies": 0.981249988079071, "rewards/chosen": -3.7093491554260254, "rewards/margins": 12.91864013671875, "rewards/rejected": -16.627988815307617, "step": 690 }, { "epoch": 0.9289209587791325, "grad_norm": 0.515625, "learning_rate": 7.523049114624647e-08, "logits/chosen": -1.9617208242416382, "logits/rejected": -1.8319323062896729, "logps/chosen": -802.1334838867188, "logps/rejected": -1892.474609375, "loss": 0.0238, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -3.8938992023468018, "rewards/margins": 11.566261291503906, "rewards/rejected": -15.460162162780762, "step": 700 }, { "epoch": 0.9289209587791325, "eval_logits/chosen": -1.8936866521835327, "eval_logits/rejected": -1.737060785293579, "eval_logps/chosen": -788.9779663085938, "eval_logps/rejected": -1951.9310302734375, "eval_loss": 0.027877720072865486, "eval_rewards/accuracies": 0.9828358292579651, "eval_rewards/chosen": -3.897392511367798, "eval_rewards/margins": 12.06684684753418, "eval_rewards/rejected": -15.964240074157715, "eval_runtime": 830.3344, "eval_samples_per_second": 6.454, "eval_steps_per_second": 1.614, "step": 700 }, { "epoch": 0.942191258190263, "grad_norm": 0.92578125, "learning_rate": 4.9605274709082774e-08, "logits/chosen": -1.883724570274353, "logits/rejected": -1.7111318111419678, "logps/chosen": -796.9892578125, "logps/rejected": -2012.552001953125, "loss": 0.0267, "rewards/accuracies": 0.987500011920929, "rewards/chosen": -3.85164213180542, "rewards/margins": 12.14155101776123, "rewards/rejected": -15.993194580078125, "step": 710 }, { "epoch": 0.9554615576013934, "grad_norm": 6.9375, "learning_rate": 2.9255763497703373e-08, "logits/chosen": -1.8911021947860718, "logits/rejected": -1.6832084655761719, "logps/chosen": -790.8480224609375, "logps/rejected": -2087.356689453125, "loss": 0.0314, "rewards/accuracies": 1.0, "rewards/chosen": -3.734605312347412, "rewards/margins": 13.477132797241211, "rewards/rejected": -17.21173858642578, "step": 720 }, { "epoch": 0.9687318570125238, "grad_norm": 3.46875, "learning_rate": 1.42257700544432e-08, "logits/chosen": -1.9454247951507568, "logits/rejected": -1.7627389430999756, "logps/chosen": -774.6414794921875, "logps/rejected": -1878.0687255859375, "loss": 0.0266, "rewards/accuracies": 0.9937499761581421, "rewards/chosen": -3.7591774463653564, "rewards/margins": 11.635783195495605, "rewards/rejected": -15.3949613571167, "step": 730 }, { "epoch": 0.9820021564236543, "grad_norm": 0.6640625, "learning_rate": 4.547653988198619e-09, "logits/chosen": -1.8585050106048584, "logits/rejected": -1.6734154224395752, "logps/chosen": -771.71044921875, "logps/rejected": -1891.1732177734375, "loss": 0.0211, "rewards/accuracies": 1.0, "rewards/chosen": -3.8070855140686035, "rewards/margins": 11.580972671508789, "rewards/rejected": -15.38805866241455, "step": 740 }, { "epoch": 0.9952724558347847, "grad_norm": 0.46875, "learning_rate": 2.422523041178959e-10, "logits/chosen": -1.9344615936279297, "logits/rejected": -1.733974814414978, "logps/chosen": -803.4793090820312, "logps/rejected": -2031.9625244140625, "loss": 0.0295, "rewards/accuracies": 0.981249988079071, "rewards/chosen": -3.7710418701171875, "rewards/margins": 12.417600631713867, "rewards/rejected": -16.188644409179688, "step": 750 }, { "epoch": 0.999253545658124, "step": 753, "total_flos": 0.0, "train_loss": 0.0882907345950366, "train_runtime": 20220.3954, "train_samples_per_second": 2.385, "train_steps_per_second": 0.037 } ], "logging_steps": 10, "max_steps": 753, "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": 1, "trial_name": null, "trial_params": null }