{ "best_global_step": null, "best_metric": null, "best_model_checkpoint": null, "epoch": 0.9989528795811519, "eval_steps": 200, "global_step": 477, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0020942408376963353, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.02793481945991516, "fcm_dpo/q_t": 0.500069797039032, "grad_norm": 28.588966369628906, "learning_rate": 0.0, "logits/chosen": -0.5898098945617676, "logits/rejected": -0.604260265827179, "logps/chosen": -275.28570556640625, "logps/ref_chosen": -275.2312927246094, "logps/ref_rejected": -222.9380340576172, "logps/rejected": -222.96453857421875, "loss": 5.5463, "margin_dpo/margin_mean": -0.02793477475643158, "margin_dpo/margin_std": 0.5724214911460876, "step": 1 }, { "epoch": 0.004188481675392671, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.014312177896499634, "fcm_dpo/q_t": 0.4999642074108124, "grad_norm": 27.87852668762207, "learning_rate": 1.0416666666666666e-08, "logits/chosen": -0.6574729681015015, "logits/rejected": -0.6464410424232483, "logps/chosen": -264.7165222167969, "logps/ref_chosen": -264.7611083984375, "logps/ref_rejected": -242.5597686767578, "logps/rejected": -242.52951049804688, "loss": 5.5446, "margin_dpo/margin_mean": 0.014312252402305603, "margin_dpo/margin_std": 0.6423971652984619, "step": 2 }, { "epoch": 0.0062827225130890054, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.016503050923347473, "fcm_dpo/q_t": 0.5000412464141846, "grad_norm": 25.815200805664062, "learning_rate": 2.083333333333333e-08, "logits/chosen": -0.6840308308601379, "logits/rejected": -0.7351540327072144, "logps/chosen": -274.10198974609375, "logps/ref_chosen": -274.1018981933594, "logps/ref_rejected": -286.5882568359375, "logps/rejected": -286.5718688964844, "loss": 5.5459, "margin_dpo/margin_mean": -0.016506418585777283, "margin_dpo/margin_std": 0.6721947193145752, "step": 3 }, { "epoch": 0.008376963350785341, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.011121675372123718, "fcm_dpo/q_t": 0.5000278353691101, "grad_norm": 32.197547912597656, "learning_rate": 3.125e-08, "logits/chosen": -0.6205219626426697, "logits/rejected": -0.6148541569709778, "logps/chosen": -329.8299255371094, "logps/ref_chosen": -329.8382568359375, "logps/ref_rejected": -303.2850646972656, "logps/rejected": -303.26556396484375, "loss": 5.5457, "margin_dpo/margin_mean": -0.011121734976768494, "margin_dpo/margin_std": 0.7242615222930908, "step": 4 }, { "epoch": 0.010471204188481676, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.055549949407577515, "fcm_dpo/q_t": 0.5001388788223267, "grad_norm": 29.520116806030273, "learning_rate": 4.166666666666666e-08, "logits/chosen": -0.5685195922851562, "logits/rejected": -0.5843265652656555, "logps/chosen": -301.7572021484375, "logps/ref_chosen": -301.7389221191406, "logps/ref_rejected": -274.7654724121094, "logps/rejected": -274.7281799316406, "loss": 5.5475, "margin_dpo/margin_mean": -0.05554959177970886, "margin_dpo/margin_std": 0.8365975022315979, "step": 5 }, { "epoch": 0.012565445026178011, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.024648115038871765, "fcm_dpo/q_t": 0.49993836879730225, "grad_norm": 28.228336334228516, "learning_rate": 5.208333333333333e-08, "logits/chosen": -0.6921972036361694, "logits/rejected": -0.6560333967208862, "logps/chosen": -285.6768493652344, "logps/ref_chosen": -285.6946716308594, "logps/ref_rejected": -245.8200225830078, "logps/rejected": -245.8268585205078, "loss": 5.5442, "margin_dpo/margin_mean": 0.024649202823638916, "margin_dpo/margin_std": 0.704728901386261, "step": 6 }, { "epoch": 0.014659685863874346, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.06559588015079498, "fcm_dpo/q_t": 0.5001639127731323, "grad_norm": 28.551637649536133, "learning_rate": 6.25e-08, "logits/chosen": -0.5759603381156921, "logits/rejected": -0.6086133718490601, "logps/chosen": -264.6998596191406, "logps/ref_chosen": -264.65545654296875, "logps/ref_rejected": -253.10305786132812, "logps/rejected": -253.0819091796875, "loss": 5.5478, "margin_dpo/margin_mean": -0.06559541821479797, "margin_dpo/margin_std": 0.6130937933921814, "step": 7 }, { "epoch": 0.016753926701570682, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.028539016842842102, "fcm_dpo/q_t": 0.49992865324020386, "grad_norm": 30.765186309814453, "learning_rate": 7.291666666666667e-08, "logits/chosen": -0.6858725547790527, "logits/rejected": -0.6940003037452698, "logps/chosen": -354.1387023925781, "logps/ref_chosen": -354.1887512207031, "logps/ref_rejected": -282.9112243652344, "logps/rejected": -282.8896484375, "loss": 5.5441, "margin_dpo/margin_mean": 0.02854001522064209, "margin_dpo/margin_std": 0.676353394985199, "step": 8 }, { "epoch": 0.018848167539267015, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.09649568796157837, "fcm_dpo/q_t": 0.500241219997406, "grad_norm": 27.911169052124023, "learning_rate": 8.333333333333333e-08, "logits/chosen": -0.6349136829376221, "logits/rejected": -0.6529811024665833, "logps/chosen": -285.5809326171875, "logps/ref_chosen": -285.5502014160156, "logps/ref_rejected": -267.99664306640625, "logps/rejected": -267.9308776855469, "loss": 5.5491, "margin_dpo/margin_mean": -0.09649543464183807, "margin_dpo/margin_std": 0.650297224521637, "step": 9 }, { "epoch": 0.020942408376963352, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.019019827246665955, "fcm_dpo/q_t": 0.4999524652957916, "grad_norm": 26.725051879882812, "learning_rate": 9.375e-08, "logits/chosen": -0.6896336674690247, "logits/rejected": -0.6844220757484436, "logps/chosen": -251.85069274902344, "logps/ref_chosen": -251.91238403320312, "logps/ref_rejected": -226.45260620117188, "logps/rejected": -226.40988159179688, "loss": 5.5445, "margin_dpo/margin_mean": 0.019019678235054016, "margin_dpo/margin_std": 0.6863389015197754, "step": 10 }, { "epoch": 0.023036649214659685, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.03224755823612213, "fcm_dpo/q_t": 0.5000805854797363, "grad_norm": 28.949674606323242, "learning_rate": 1.0416666666666667e-07, "logits/chosen": -0.5962199568748474, "logits/rejected": -0.6516202092170715, "logps/chosen": -301.0753479003906, "logps/ref_chosen": -301.08343505859375, "logps/ref_rejected": -259.546630859375, "logps/rejected": -259.5062561035156, "loss": 5.5465, "margin_dpo/margin_mean": -0.032247647643089294, "margin_dpo/margin_std": 0.6402597427368164, "step": 11 }, { "epoch": 0.025130890052356022, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.1078348308801651, "fcm_dpo/q_t": 0.4997304081916809, "grad_norm": 30.103239059448242, "learning_rate": 1.1458333333333332e-07, "logits/chosen": -0.577997624874115, "logits/rejected": -0.5377943515777588, "logps/chosen": -287.52789306640625, "logps/ref_chosen": -287.548095703125, "logps/ref_rejected": -277.37945556640625, "logps/rejected": -277.46710205078125, "loss": 5.5409, "margin_dpo/margin_mean": 0.10783536732196808, "margin_dpo/margin_std": 0.7473784685134888, "step": 12 }, { "epoch": 0.027225130890052355, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.10099650919437408, "fcm_dpo/q_t": 0.49974748492240906, "grad_norm": 27.242509841918945, "learning_rate": 1.25e-07, "logits/chosen": -0.6565054059028625, "logits/rejected": -0.6640400290489197, "logps/chosen": -270.64508056640625, "logps/ref_chosen": -270.6664123535156, "logps/ref_rejected": -274.6546936035156, "logps/rejected": -274.7343444824219, "loss": 5.5412, "margin_dpo/margin_mean": 0.1009967178106308, "margin_dpo/margin_std": 0.676527738571167, "step": 13 }, { "epoch": 0.02931937172774869, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.02157723903656006, "fcm_dpo/q_t": 0.5000539422035217, "grad_norm": 28.256738662719727, "learning_rate": 1.3541666666666666e-07, "logits/chosen": -0.62161785364151, "logits/rejected": -0.6502420902252197, "logps/chosen": -281.6171569824219, "logps/ref_chosen": -281.59320068359375, "logps/ref_rejected": -263.52215576171875, "logps/rejected": -263.5245056152344, "loss": 5.5461, "margin_dpo/margin_mean": -0.021577835083007812, "margin_dpo/margin_std": 0.698745846748352, "step": 14 }, { "epoch": 0.031413612565445025, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.05705377459526062, "fcm_dpo/q_t": 0.4998573660850525, "grad_norm": 30.230676651000977, "learning_rate": 1.4583333333333335e-07, "logits/chosen": -0.6337839365005493, "logits/rejected": -0.6448744535446167, "logps/chosen": -298.41180419921875, "logps/ref_chosen": -298.45343017578125, "logps/ref_rejected": -227.17832946777344, "logps/rejected": -227.19375610351562, "loss": 5.5429, "margin_dpo/margin_mean": 0.05705252289772034, "margin_dpo/margin_std": 0.6579793691635132, "step": 15 }, { "epoch": 0.033507853403141365, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.08925226330757141, "fcm_dpo/q_t": 0.4997768998146057, "grad_norm": 30.15053939819336, "learning_rate": 1.5624999999999999e-07, "logits/chosen": -0.6029733419418335, "logits/rejected": -0.6001195907592773, "logps/chosen": -293.8938293457031, "logps/ref_chosen": -293.96661376953125, "logps/ref_rejected": -250.78443908691406, "logps/rejected": -250.80093383789062, "loss": 5.5417, "margin_dpo/margin_mean": 0.0892522782087326, "margin_dpo/margin_std": 0.725162148475647, "step": 16 }, { "epoch": 0.0356020942408377, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.04146842658519745, "fcm_dpo/q_t": 0.4998963177204132, "grad_norm": 27.696533203125, "learning_rate": 1.6666666666666665e-07, "logits/chosen": -0.5717862248420715, "logits/rejected": -0.5980989336967468, "logps/chosen": -262.3448791503906, "logps/ref_chosen": -262.39398193359375, "logps/ref_rejected": -248.500244140625, "logps/rejected": -248.49258422851562, "loss": 5.5436, "margin_dpo/margin_mean": 0.041468992829322815, "margin_dpo/margin_std": 0.6475922465324402, "step": 17 }, { "epoch": 0.03769633507853403, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.0832410454750061, "fcm_dpo/q_t": 0.4997918903827667, "grad_norm": 29.68451690673828, "learning_rate": 1.7708333333333334e-07, "logits/chosen": -0.6132781505584717, "logits/rejected": -0.6193888187408447, "logps/chosen": -293.6752014160156, "logps/ref_chosen": -293.709228515625, "logps/ref_rejected": -274.5875244140625, "logps/rejected": -274.63677978515625, "loss": 5.5419, "margin_dpo/margin_mean": 0.0832410603761673, "margin_dpo/margin_std": 0.6698707938194275, "step": 18 }, { "epoch": 0.039790575916230364, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.008256226778030396, "fcm_dpo/q_t": 0.4999793469905853, "grad_norm": 28.128124237060547, "learning_rate": 1.875e-07, "logits/chosen": -0.6208982467651367, "logits/rejected": -0.6168572902679443, "logps/chosen": -280.22943115234375, "logps/ref_chosen": -280.26568603515625, "logps/ref_rejected": -259.9742736816406, "logps/rejected": -259.9462890625, "loss": 5.5449, "margin_dpo/margin_mean": 0.008254945278167725, "margin_dpo/margin_std": 0.6279901266098022, "step": 19 }, { "epoch": 0.041884816753926704, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.07622025907039642, "fcm_dpo/q_t": 0.4998094439506531, "grad_norm": 29.674253463745117, "learning_rate": 1.9791666666666664e-07, "logits/chosen": -0.6202477812767029, "logits/rejected": -0.6521307826042175, "logps/chosen": -303.79779052734375, "logps/ref_chosen": -303.8954162597656, "logps/ref_rejected": -260.214599609375, "logps/rejected": -260.1932067871094, "loss": 5.5422, "margin_dpo/margin_mean": 0.07622019946575165, "margin_dpo/margin_std": 0.7053053379058838, "step": 20 }, { "epoch": 0.04397905759162304, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.10494795441627502, "fcm_dpo/q_t": 0.4997376799583435, "grad_norm": 35.037410736083984, "learning_rate": 2.0833333333333333e-07, "logits/chosen": -0.6494017243385315, "logits/rejected": -0.6782788038253784, "logps/chosen": -301.4276123046875, "logps/ref_chosen": -301.5334777832031, "logps/ref_rejected": -280.28900146484375, "logps/rejected": -280.2880554199219, "loss": 5.541, "margin_dpo/margin_mean": 0.10494862496852875, "margin_dpo/margin_std": 0.7810779213905334, "step": 21 }, { "epoch": 0.04607329842931937, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.06129841506481171, "fcm_dpo/q_t": 0.49984675645828247, "grad_norm": 25.279335021972656, "learning_rate": 2.1875e-07, "logits/chosen": -0.6577863693237305, "logits/rejected": -0.6595051288604736, "logps/chosen": -259.9060974121094, "logps/ref_chosen": -259.9951477050781, "logps/ref_rejected": -243.0721435546875, "logps/rejected": -243.04441833496094, "loss": 5.5428, "margin_dpo/margin_mean": 0.061297982931137085, "margin_dpo/margin_std": 0.7572940587997437, "step": 22 }, { "epoch": 0.048167539267015703, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.15443375706672668, "fcm_dpo/q_t": 0.4996139407157898, "grad_norm": 28.133821487426758, "learning_rate": 2.2916666666666663e-07, "logits/chosen": -0.6106438636779785, "logits/rejected": -0.6437903046607971, "logps/chosen": -282.0675048828125, "logps/ref_chosen": -282.1807556152344, "logps/ref_rejected": -265.0758056640625, "logps/rejected": -265.11700439453125, "loss": 5.5391, "margin_dpo/margin_mean": 0.1544336974620819, "margin_dpo/margin_std": 0.8530293107032776, "step": 23 }, { "epoch": 0.050261780104712044, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.24844954907894135, "fcm_dpo/q_t": 0.49937885999679565, "grad_norm": 29.64801597595215, "learning_rate": 2.3958333333333335e-07, "logits/chosen": -0.6590055823326111, "logits/rejected": -0.5712109208106995, "logps/chosen": -300.9676513671875, "logps/ref_chosen": -301.17962646484375, "logps/ref_rejected": -302.12786865234375, "logps/rejected": -302.1643371582031, "loss": 5.5353, "margin_dpo/margin_mean": 0.2484499216079712, "margin_dpo/margin_std": 0.7506996393203735, "step": 24 }, { "epoch": 0.05235602094240838, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.15574315190315247, "fcm_dpo/q_t": 0.49961066246032715, "grad_norm": 26.359050750732422, "learning_rate": 2.5e-07, "logits/chosen": -0.5938785076141357, "logits/rejected": -0.6017901301383972, "logps/chosen": -246.56497192382812, "logps/ref_chosen": -246.74649047851562, "logps/ref_rejected": -235.55638122558594, "logps/rejected": -235.53057861328125, "loss": 5.539, "margin_dpo/margin_mean": 0.15574237704277039, "margin_dpo/margin_std": 0.7371436953544617, "step": 25 }, { "epoch": 0.05445026178010471, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.15086300671100616, "fcm_dpo/q_t": 0.4996228516101837, "grad_norm": 28.607776641845703, "learning_rate": 2.604166666666667e-07, "logits/chosen": -0.6509619951248169, "logits/rejected": -0.6650726795196533, "logps/chosen": -281.9699401855469, "logps/ref_chosen": -282.1955871582031, "logps/ref_rejected": -235.3135528564453, "logps/rejected": -235.23878479003906, "loss": 5.5392, "margin_dpo/margin_mean": 0.1508627086877823, "margin_dpo/margin_std": 0.83431077003479, "step": 26 }, { "epoch": 0.05654450261780105, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.20935215055942535, "fcm_dpo/q_t": 0.4994766116142273, "grad_norm": 27.72915267944336, "learning_rate": 2.708333333333333e-07, "logits/chosen": -0.6564357280731201, "logits/rejected": -0.6764754056930542, "logps/chosen": -323.56011962890625, "logps/ref_chosen": -323.8563537597656, "logps/ref_rejected": -245.968017578125, "logps/rejected": -245.8811492919922, "loss": 5.5369, "margin_dpo/margin_mean": 0.20935235917568207, "margin_dpo/margin_std": 0.7932536005973816, "step": 27 }, { "epoch": 0.05863874345549738, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.18029074370861053, "fcm_dpo/q_t": 0.4995492696762085, "grad_norm": 26.313339233398438, "learning_rate": 2.8125e-07, "logits/chosen": -0.6219736337661743, "logits/rejected": -0.6311923861503601, "logps/chosen": -247.9824981689453, "logps/ref_chosen": -248.24673461914062, "logps/ref_rejected": -240.0382080078125, "logps/rejected": -239.95425415039062, "loss": 5.5381, "margin_dpo/margin_mean": 0.18029029667377472, "margin_dpo/margin_std": 0.8702433705329895, "step": 28 }, { "epoch": 0.060732984293193716, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.23714801669120789, "fcm_dpo/q_t": 0.4994071424007416, "grad_norm": 29.599367141723633, "learning_rate": 2.916666666666667e-07, "logits/chosen": -0.5967488884925842, "logits/rejected": -0.6196838617324829, "logps/chosen": -317.91094970703125, "logps/ref_chosen": -318.2564392089844, "logps/ref_rejected": -286.75848388671875, "logps/rejected": -286.650146484375, "loss": 5.5358, "margin_dpo/margin_mean": 0.23714756965637207, "margin_dpo/margin_std": 0.8616628646850586, "step": 29 }, { "epoch": 0.06282722513089005, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.2596416473388672, "fcm_dpo/q_t": 0.4993509352207184, "grad_norm": 28.952259063720703, "learning_rate": 3.020833333333333e-07, "logits/chosen": -0.5958232283592224, "logits/rejected": -0.6129447221755981, "logps/chosen": -252.67791748046875, "logps/ref_chosen": -253.0491485595703, "logps/ref_rejected": -261.30029296875, "logps/rejected": -261.188720703125, "loss": 5.5349, "margin_dpo/margin_mean": 0.2596414089202881, "margin_dpo/margin_std": 0.9200783967971802, "step": 30 }, { "epoch": 0.06492146596858639, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.31609046459198, "fcm_dpo/q_t": 0.49920982122421265, "grad_norm": 25.55852699279785, "learning_rate": 3.1249999999999997e-07, "logits/chosen": -0.6523265838623047, "logits/rejected": -0.6868293881416321, "logps/chosen": -247.73052978515625, "logps/ref_chosen": -248.15301513671875, "logps/ref_rejected": -203.17703247070312, "logps/rejected": -203.07061767578125, "loss": 5.5327, "margin_dpo/margin_mean": 0.316089928150177, "margin_dpo/margin_std": 1.0166584253311157, "step": 31 }, { "epoch": 0.06701570680628273, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.20415076613426208, "fcm_dpo/q_t": 0.49948960542678833, "grad_norm": 29.79863166809082, "learning_rate": 3.2291666666666666e-07, "logits/chosen": -0.6170097589492798, "logits/rejected": -0.6221901774406433, "logps/chosen": -305.0362548828125, "logps/ref_chosen": -305.5399475097656, "logps/ref_rejected": -267.6527099609375, "logps/rejected": -267.3531799316406, "loss": 5.5371, "margin_dpo/margin_mean": 0.20415127277374268, "margin_dpo/margin_std": 0.90806645154953, "step": 32 }, { "epoch": 0.06910994764397906, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.4111286997795105, "fcm_dpo/q_t": 0.49897223711013794, "grad_norm": 28.34008026123047, "learning_rate": 3.333333333333333e-07, "logits/chosen": -0.6372838020324707, "logits/rejected": -0.6494584083557129, "logps/chosen": -285.63262939453125, "logps/ref_chosen": -286.2335205078125, "logps/ref_rejected": -255.38748168945312, "logps/rejected": -255.19772338867188, "loss": 5.5289, "margin_dpo/margin_mean": 0.41112855076789856, "margin_dpo/margin_std": 1.0520519018173218, "step": 33 }, { "epoch": 0.0712041884816754, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.7322409152984619, "fcm_dpo/q_t": 0.49816954135894775, "grad_norm": 31.091548919677734, "learning_rate": 3.4375e-07, "logits/chosen": -0.6315101385116577, "logits/rejected": -0.6417458653450012, "logps/chosen": -340.8562927246094, "logps/ref_chosen": -341.5920104980469, "logps/ref_rejected": -278.8866882324219, "logps/rejected": -278.8831787109375, "loss": 5.5161, "margin_dpo/margin_mean": 0.7322409749031067, "margin_dpo/margin_std": 1.1406543254852295, "step": 34 }, { "epoch": 0.07329842931937172, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.43320757150650024, "fcm_dpo/q_t": 0.49891701340675354, "grad_norm": 26.510805130004883, "learning_rate": 3.541666666666667e-07, "logits/chosen": -0.6248103380203247, "logits/rejected": -0.6445828080177307, "logps/chosen": -264.3880310058594, "logps/ref_chosen": -265.0795593261719, "logps/ref_rejected": -264.4876708984375, "logps/rejected": -264.2293701171875, "loss": 5.528, "margin_dpo/margin_mean": 0.4332066774368286, "margin_dpo/margin_std": 1.2199397087097168, "step": 35 }, { "epoch": 0.07539267015706806, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.6611357927322388, "fcm_dpo/q_t": 0.49834728240966797, "grad_norm": 31.89276885986328, "learning_rate": 3.645833333333333e-07, "logits/chosen": -0.6019506454467773, "logits/rejected": -0.6194975972175598, "logps/chosen": -296.47509765625, "logps/ref_chosen": -297.3261413574219, "logps/ref_rejected": -282.09515380859375, "logps/rejected": -281.9052734375, "loss": 5.519, "margin_dpo/margin_mean": 0.6611351370811462, "margin_dpo/margin_std": 1.3979815244674683, "step": 36 }, { "epoch": 0.0774869109947644, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.5799505710601807, "fcm_dpo/q_t": 0.4985502362251282, "grad_norm": 30.829147338867188, "learning_rate": 3.75e-07, "logits/chosen": -0.5941240191459656, "logits/rejected": -0.6079503297805786, "logps/chosen": -313.24591064453125, "logps/ref_chosen": -314.0340270996094, "logps/ref_rejected": -299.3437805175781, "logps/rejected": -299.1355895996094, "loss": 5.5223, "margin_dpo/margin_mean": 0.579948902130127, "margin_dpo/margin_std": 1.535064458847046, "step": 37 }, { "epoch": 0.07958115183246073, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.6605011820793152, "fcm_dpo/q_t": 0.49834883213043213, "grad_norm": 28.47408676147461, "learning_rate": 3.8541666666666665e-07, "logits/chosen": -0.6379660964012146, "logits/rejected": -0.6501216292381287, "logps/chosen": -281.49658203125, "logps/ref_chosen": -282.54119873046875, "logps/ref_rejected": -269.7773132324219, "logps/rejected": -269.3931884765625, "loss": 5.5191, "margin_dpo/margin_mean": 0.6605010032653809, "margin_dpo/margin_std": 1.4664862155914307, "step": 38 }, { "epoch": 0.08167539267015707, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.1984353065490723, "fcm_dpo/q_t": 0.49700427055358887, "grad_norm": 29.65317726135254, "learning_rate": 3.958333333333333e-07, "logits/chosen": -0.6123558878898621, "logits/rejected": -0.6260115504264832, "logps/chosen": -275.47564697265625, "logps/ref_chosen": -276.7729187011719, "logps/ref_rejected": -249.95889282226562, "logps/rejected": -249.8600311279297, "loss": 5.4977, "margin_dpo/margin_mean": 1.198434829711914, "margin_dpo/margin_std": 1.7556183338165283, "step": 39 }, { "epoch": 0.08376963350785341, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.7856162786483765, "fcm_dpo/q_t": 0.49803632497787476, "grad_norm": 27.415355682373047, "learning_rate": 4.0625e-07, "logits/chosen": -0.623888373374939, "logits/rejected": -0.6597940921783447, "logps/chosen": -283.21929931640625, "logps/ref_chosen": -284.30706787109375, "logps/ref_rejected": -244.4459991455078, "logps/rejected": -244.14381408691406, "loss": 5.5142, "margin_dpo/margin_mean": 0.7856165170669556, "margin_dpo/margin_std": 1.7853032350540161, "step": 40 }, { "epoch": 0.08586387434554973, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.7766204476356506, "fcm_dpo/q_t": 0.49805861711502075, "grad_norm": 30.24774742126465, "learning_rate": 4.1666666666666667e-07, "logits/chosen": -0.6275569200515747, "logits/rejected": -0.6529408097267151, "logps/chosen": -292.7530822753906, "logps/ref_chosen": -293.8151550292969, "logps/ref_rejected": -252.16815185546875, "logps/rejected": -251.88272094726562, "loss": 5.5145, "margin_dpo/margin_mean": 0.7766217589378357, "margin_dpo/margin_std": 1.7000435590744019, "step": 41 }, { "epoch": 0.08795811518324607, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.9282649755477905, "fcm_dpo/q_t": 0.49767956137657166, "grad_norm": 27.3597412109375, "learning_rate": 4.270833333333333e-07, "logits/chosen": -0.6338014602661133, "logits/rejected": -0.6516577005386353, "logps/chosen": -251.51942443847656, "logps/ref_chosen": -252.76023864746094, "logps/ref_rejected": -261.0414733886719, "logps/rejected": -260.7288818359375, "loss": 5.5087, "margin_dpo/margin_mean": 0.928265392780304, "margin_dpo/margin_std": 2.2327194213867188, "step": 42 }, { "epoch": 0.09005235602094241, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.1566113233566284, "fcm_dpo/q_t": 0.4971088171005249, "grad_norm": 29.90976905822754, "learning_rate": 4.375e-07, "logits/chosen": -0.5936682820320129, "logits/rejected": -0.609065055847168, "logps/chosen": -315.6018981933594, "logps/ref_chosen": -316.8347473144531, "logps/ref_rejected": -273.7649230957031, "logps/rejected": -273.68865966796875, "loss": 5.4996, "margin_dpo/margin_mean": 1.1566113233566284, "margin_dpo/margin_std": 2.265800952911377, "step": 43 }, { "epoch": 0.09214659685863874, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.5960203409194946, "fcm_dpo/q_t": 0.49601075053215027, "grad_norm": 30.911487579345703, "learning_rate": 4.479166666666667e-07, "logits/chosen": -0.6010478734970093, "logits/rejected": -0.6001090407371521, "logps/chosen": -285.37725830078125, "logps/ref_chosen": -286.8757019042969, "logps/ref_rejected": -282.4681396484375, "logps/rejected": -282.56573486328125, "loss": 5.4826, "margin_dpo/margin_mean": 1.5960203409194946, "margin_dpo/margin_std": 3.0894060134887695, "step": 44 }, { "epoch": 0.09424083769633508, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.3747289180755615, "fcm_dpo/q_t": 0.49656426906585693, "grad_norm": 28.888906478881836, "learning_rate": 4.5833333333333327e-07, "logits/chosen": -0.6945655345916748, "logits/rejected": -0.7191615104675293, "logps/chosen": -322.55853271484375, "logps/ref_chosen": -324.2633972167969, "logps/ref_rejected": -293.09466552734375, "logps/rejected": -292.7645263671875, "loss": 5.4912, "margin_dpo/margin_mean": 1.3747289180755615, "margin_dpo/margin_std": 2.859666109085083, "step": 45 }, { "epoch": 0.09633507853403141, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.5242335796356201, "fcm_dpo/q_t": 0.49619045853614807, "grad_norm": 30.363393783569336, "learning_rate": 4.6874999999999996e-07, "logits/chosen": -0.6261583566665649, "logits/rejected": -0.6395462155342102, "logps/chosen": -296.660888671875, "logps/ref_chosen": -298.3357238769531, "logps/ref_rejected": -267.66204833984375, "logps/rejected": -267.51141357421875, "loss": 5.4852, "margin_dpo/margin_mean": 1.5242342948913574, "margin_dpo/margin_std": 2.66381573677063, "step": 46 }, { "epoch": 0.09842931937172775, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.9342592358589172, "fcm_dpo/q_t": 0.49766474962234497, "grad_norm": 26.27431869506836, "learning_rate": 4.791666666666667e-07, "logits/chosen": -0.6025406718254089, "logits/rejected": -0.6231379508972168, "logps/chosen": -261.08758544921875, "logps/ref_chosen": -262.5669250488281, "logps/ref_rejected": -258.70989990234375, "logps/rejected": -258.1648254394531, "loss": 5.5091, "margin_dpo/margin_mean": 0.9342588186264038, "margin_dpo/margin_std": 3.3158531188964844, "step": 47 }, { "epoch": 0.10052356020942409, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.5198674201965332, "fcm_dpo/q_t": 0.49620240926742554, "grad_norm": 27.544031143188477, "learning_rate": 4.895833333333333e-07, "logits/chosen": -0.6120340824127197, "logits/rejected": -0.6370552182197571, "logps/chosen": -267.6554870605469, "logps/ref_chosen": -269.4932556152344, "logps/ref_rejected": -241.888916015625, "logps/rejected": -241.5709991455078, "loss": 5.4857, "margin_dpo/margin_mean": 1.5198673009872437, "margin_dpo/margin_std": 2.993640422821045, "step": 48 }, { "epoch": 0.10261780104712041, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.6754217147827148, "fcm_dpo/q_t": 0.49581262469291687, "grad_norm": 27.63544273376465, "learning_rate": 5e-07, "logits/chosen": -0.6648722887039185, "logits/rejected": -0.649385929107666, "logps/chosen": -255.69033813476562, "logps/ref_chosen": -257.8844909667969, "logps/ref_rejected": -256.8912048339844, "logps/rejected": -256.3724670410156, "loss": 5.4797, "margin_dpo/margin_mean": 1.6754218339920044, "margin_dpo/margin_std": 3.419741630554199, "step": 49 }, { "epoch": 0.10471204188481675, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.1309083700180054, "fcm_dpo/q_t": 0.4971740245819092, "grad_norm": 28.015737533569336, "learning_rate": 4.999932966293553e-07, "logits/chosen": -0.6291318535804749, "logits/rejected": -0.6510509252548218, "logps/chosen": -299.6326599121094, "logps/ref_chosen": -301.62884521484375, "logps/ref_rejected": -298.2716064453125, "logps/rejected": -297.40631103515625, "loss": 5.5017, "margin_dpo/margin_mean": 1.1309083700180054, "margin_dpo/margin_std": 3.866288423538208, "step": 50 }, { "epoch": 0.1068062827225131, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.9230157136917114, "fcm_dpo/q_t": 0.49519532918930054, "grad_norm": 29.120933532714844, "learning_rate": 4.999731868769026e-07, "logits/chosen": -0.6170083284378052, "logits/rejected": -0.6109319925308228, "logps/chosen": -267.2721862792969, "logps/ref_chosen": -269.37237548828125, "logps/ref_rejected": -297.0167541503906, "logps/rejected": -296.8395690917969, "loss": 5.4711, "margin_dpo/margin_mean": 1.9230164289474487, "margin_dpo/margin_std": 4.57507848739624, "step": 51 }, { "epoch": 0.10890052356020942, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 3.091094493865967, "fcm_dpo/q_t": 0.4922794699668884, "grad_norm": 30.30776023864746, "learning_rate": 4.99939671821067e-07, "logits/chosen": -0.6616694331169128, "logits/rejected": -0.6685453653335571, "logps/chosen": -304.083251953125, "logps/ref_chosen": -306.9028015136719, "logps/ref_rejected": -281.24737548828125, "logps/rejected": -281.5189208984375, "loss": 5.4251, "margin_dpo/margin_mean": 3.0910940170288086, "margin_dpo/margin_std": 4.87042760848999, "step": 52 }, { "epoch": 0.11099476439790576, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 2.1641690731048584, "fcm_dpo/q_t": 0.4945937991142273, "grad_norm": 29.97233772277832, "learning_rate": 4.998927532591591e-07, "logits/chosen": -0.6528322100639343, "logits/rejected": -0.6935089230537415, "logps/chosen": -283.1597595214844, "logps/ref_chosen": -285.9759521484375, "logps/ref_rejected": -273.9073486328125, "logps/rejected": -273.2553405761719, "loss": 5.4621, "margin_dpo/margin_mean": 2.164168119430542, "margin_dpo/margin_std": 5.062405586242676, "step": 53 }, { "epoch": 0.1130890052356021, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.8245185613632202, "fcm_dpo/q_t": 0.4954439401626587, "grad_norm": 26.48580551147461, "learning_rate": 4.998324337072792e-07, "logits/chosen": -0.698631763458252, "logits/rejected": -0.7079989314079285, "logps/chosen": -303.81121826171875, "logps/ref_chosen": -306.504638671875, "logps/ref_rejected": -272.67431640625, "logps/rejected": -271.8053894042969, "loss": 5.4757, "margin_dpo/margin_mean": 1.824518084526062, "margin_dpo/margin_std": 5.36137580871582, "step": 54 }, { "epoch": 0.11518324607329843, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 2.2451283931732178, "fcm_dpo/q_t": 0.494392067193985, "grad_norm": 24.85249137878418, "learning_rate": 4.997587164001815e-07, "logits/chosen": -0.6399841904640198, "logits/rejected": -0.6441481113433838, "logps/chosen": -220.5376434326172, "logps/ref_chosen": -222.33013916015625, "logps/ref_rejected": -206.59571838378906, "logps/rejected": -207.04835510253906, "loss": 5.4583, "margin_dpo/margin_mean": 2.245128870010376, "margin_dpo/margin_std": 4.873081207275391, "step": 55 }, { "epoch": 0.11727748691099477, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 3.144686698913574, "fcm_dpo/q_t": 0.4921509027481079, "grad_norm": 27.462751388549805, "learning_rate": 4.996716052911017e-07, "logits/chosen": -0.6191872358322144, "logits/rejected": -0.6348315477371216, "logps/chosen": -247.60455322265625, "logps/ref_chosen": -250.47816467285156, "logps/ref_rejected": -228.25848388671875, "logps/rejected": -228.52955627441406, "loss": 5.4246, "margin_dpo/margin_mean": 3.144686698913574, "margin_dpo/margin_std": 5.893349647521973, "step": 56 }, { "epoch": 0.1193717277486911, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 4.213151454925537, "fcm_dpo/q_t": 0.48947957158088684, "grad_norm": 30.760799407958984, "learning_rate": 4.99571105051544e-07, "logits/chosen": -0.7002225518226624, "logits/rejected": -0.6711569428443909, "logps/chosen": -311.23614501953125, "logps/ref_chosen": -315.1195373535156, "logps/ref_rejected": -272.755615234375, "logps/rejected": -273.08538818359375, "loss": 5.3819, "margin_dpo/margin_mean": 4.213151454925537, "margin_dpo/margin_std": 5.628840446472168, "step": 57 }, { "epoch": 0.12146596858638743, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 2.343231439590454, "fcm_dpo/q_t": 0.4941456913948059, "grad_norm": 27.43522834777832, "learning_rate": 4.994572210710314e-07, "logits/chosen": -0.6193488836288452, "logits/rejected": -0.6438428163528442, "logps/chosen": -262.74346923828125, "logps/ref_chosen": -265.1816711425781, "logps/ref_rejected": -268.2203369140625, "logps/rejected": -268.1253662109375, "loss": 5.4556, "margin_dpo/margin_mean": 2.343231201171875, "margin_dpo/margin_std": 5.691303253173828, "step": 58 }, { "epoch": 0.12356020942408377, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.5638056993484497, "fcm_dpo/q_t": 0.4960971474647522, "grad_norm": 29.727506637573242, "learning_rate": 4.993299594568162e-07, "logits/chosen": -0.6027001142501831, "logits/rejected": -0.595492959022522, "logps/chosen": -284.2311096191406, "logps/ref_chosen": -286.35394287109375, "logps/ref_rejected": -260.6757507324219, "logps/rejected": -260.11669921875, "loss": 5.4886, "margin_dpo/margin_mean": 1.5638054609298706, "margin_dpo/margin_std": 6.958956241607666, "step": 59 }, { "epoch": 0.1256544502617801, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 3.373918056488037, "fcm_dpo/q_t": 0.4915754795074463, "grad_norm": 27.950477600097656, "learning_rate": 4.991893270335525e-07, "logits/chosen": -0.6873760223388672, "logits/rejected": -0.7125962376594543, "logps/chosen": -255.85675048828125, "logps/ref_chosen": -258.74859619140625, "logps/ref_rejected": -255.04893493652344, "logps/rejected": -255.53097534179688, "loss": 5.4167, "margin_dpo/margin_mean": 3.3739190101623535, "margin_dpo/margin_std": 7.061320781707764, "step": 60 }, { "epoch": 0.12774869109947645, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 3.1508805751800537, "fcm_dpo/q_t": 0.4921364188194275, "grad_norm": 29.970989227294922, "learning_rate": 4.990353313429303e-07, "logits/chosen": -0.6524519920349121, "logits/rejected": -0.6696836948394775, "logps/chosen": -275.4600524902344, "logps/ref_chosen": -278.4678955078125, "logps/ref_rejected": -252.02720642089844, "logps/rejected": -252.170166015625, "loss": 5.4269, "margin_dpo/margin_mean": 3.1508798599243164, "margin_dpo/margin_std": 7.672506809234619, "step": 61 }, { "epoch": 0.12984293193717278, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 3.335423469543457, "fcm_dpo/q_t": 0.4916812479496002, "grad_norm": 26.69674301147461, "learning_rate": 4.988679806432711e-07, "logits/chosen": -0.6011494994163513, "logits/rejected": -0.6419218182563782, "logps/chosen": -269.00103759765625, "logps/ref_chosen": -272.92431640625, "logps/ref_rejected": -260.7935485839844, "logps/rejected": -260.2056884765625, "loss": 5.4187, "margin_dpo/margin_mean": 3.3354249000549316, "margin_dpo/margin_std": 7.46706485748291, "step": 62 }, { "epoch": 0.1319371727748691, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 3.5092599391937256, "fcm_dpo/q_t": 0.4912412762641907, "grad_norm": 28.325027465820312, "learning_rate": 4.986872839090852e-07, "logits/chosen": -0.6640771627426147, "logits/rejected": -0.6706352233886719, "logps/chosen": -273.694580078125, "logps/ref_chosen": -277.0889892578125, "logps/ref_rejected": -273.3413391113281, "logps/rejected": -273.4562072753906, "loss": 5.4118, "margin_dpo/margin_mean": 3.509258985519409, "margin_dpo/margin_std": 7.294081687927246, "step": 63 }, { "epoch": 0.13403141361256546, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 4.153190612792969, "fcm_dpo/q_t": 0.489638090133667, "grad_norm": 28.40380859375, "learning_rate": 4.9849325083059e-07, "logits/chosen": -0.6381584405899048, "logits/rejected": -0.6338601112365723, "logps/chosen": -279.7315368652344, "logps/ref_chosen": -283.8244934082031, "logps/ref_rejected": -263.29351806640625, "logps/rejected": -263.3537902832031, "loss": 5.3882, "margin_dpo/margin_mean": 4.1531901359558105, "margin_dpo/margin_std": 8.279900550842285, "step": 64 }, { "epoch": 0.13612565445026178, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 3.30912184715271, "fcm_dpo/q_t": 0.4917401969432831, "grad_norm": 27.938804626464844, "learning_rate": 4.982858918131906e-07, "logits/chosen": -0.7063099145889282, "logits/rejected": -0.674233078956604, "logps/chosen": -261.4156188964844, "logps/ref_chosen": -264.8699645996094, "logps/ref_rejected": -268.5076904296875, "logps/rejected": -268.3625183105469, "loss": 5.4195, "margin_dpo/margin_mean": 3.3091213703155518, "margin_dpo/margin_std": 7.275178909301758, "step": 65 }, { "epoch": 0.1382198952879581, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 4.092304229736328, "fcm_dpo/q_t": 0.48980259895324707, "grad_norm": 28.009225845336914, "learning_rate": 4.980652179769217e-07, "logits/chosen": -0.6750394105911255, "logits/rejected": -0.6948624849319458, "logps/chosen": -269.80352783203125, "logps/ref_chosen": -272.9283142089844, "logps/ref_rejected": -280.94696044921875, "logps/rejected": -281.9145202636719, "loss": 5.3937, "margin_dpo/margin_mean": 4.09230375289917, "margin_dpo/margin_std": 10.11854362487793, "step": 66 }, { "epoch": 0.14031413612565444, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 3.718635320663452, "fcm_dpo/q_t": 0.4907230734825134, "grad_norm": 25.479549407958984, "learning_rate": 4.978312411558517e-07, "logits/chosen": -0.6987938284873962, "logits/rejected": -0.7324379086494446, "logps/chosen": -262.1153564453125, "logps/ref_chosen": -266.18695068359375, "logps/ref_rejected": -250.17405700683594, "logps/rejected": -249.82113647460938, "loss": 5.4061, "margin_dpo/margin_mean": 3.718634843826294, "margin_dpo/margin_std": 8.839105606079102, "step": 67 }, { "epoch": 0.1424083769633508, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 4.682402610778809, "fcm_dpo/q_t": 0.48835432529449463, "grad_norm": 27.78085708618164, "learning_rate": 4.975839738974473e-07, "logits/chosen": -0.6808131337165833, "logits/rejected": -0.6937141418457031, "logps/chosen": -294.8902587890625, "logps/ref_chosen": -297.9385986328125, "logps/ref_rejected": -261.5141296386719, "logps/rejected": -263.148193359375, "loss": 5.3723, "margin_dpo/margin_mean": 4.682403564453125, "margin_dpo/margin_std": 10.64492130279541, "step": 68 }, { "epoch": 0.14450261780104712, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 6.0140380859375, "fcm_dpo/q_t": 0.48501238226890564, "grad_norm": 28.59217643737793, "learning_rate": 4.97323429461901e-07, "logits/chosen": -0.6912984251976013, "logits/rejected": -0.7229157090187073, "logps/chosen": -261.6896667480469, "logps/ref_chosen": -265.6175231933594, "logps/ref_rejected": -236.8287353515625, "logps/rejected": -238.91488647460938, "loss": 5.3182, "margin_dpo/margin_mean": 6.014038562774658, "margin_dpo/margin_std": 9.668503761291504, "step": 69 }, { "epoch": 0.14659685863874344, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 5.854832649230957, "fcm_dpo/q_t": 0.48542001843452454, "grad_norm": 28.611467361450195, "learning_rate": 4.970496218214204e-07, "logits/chosen": -0.6638342142105103, "logits/rejected": -0.6955633759498596, "logps/chosen": -291.9162902832031, "logps/ref_chosen": -296.2259216308594, "logps/ref_rejected": -254.68496704101562, "logps/rejected": -256.2301330566406, "loss": 5.3275, "margin_dpo/margin_mean": 5.854830741882324, "margin_dpo/margin_std": 11.032973289489746, "step": 70 }, { "epoch": 0.1486910994764398, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 5.318414211273193, "fcm_dpo/q_t": 0.48674529790878296, "grad_norm": 28.393184661865234, "learning_rate": 4.967625656594781e-07, "logits/chosen": -0.6533457040786743, "logits/rejected": -0.6424985527992249, "logps/chosen": -283.6885681152344, "logps/ref_chosen": -288.92724609375, "logps/ref_rejected": -278.6405334472656, "logps/rejected": -278.72027587890625, "loss": 5.3521, "margin_dpo/margin_mean": 5.318414688110352, "margin_dpo/margin_std": 12.410322189331055, "step": 71 }, { "epoch": 0.15078534031413612, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 5.140360355377197, "fcm_dpo/q_t": 0.48718106746673584, "grad_norm": 28.1318359375, "learning_rate": 4.964622763700252e-07, "logits/chosen": -0.6930279731750488, "logits/rejected": -0.7042161226272583, "logps/chosen": -233.69351196289062, "logps/ref_chosen": -237.0452880859375, "logps/ref_rejected": -252.7946319580078, "logps/rejected": -254.58322143554688, "loss": 5.3546, "margin_dpo/margin_mean": 5.140359878540039, "margin_dpo/margin_std": 10.367585182189941, "step": 72 }, { "epoch": 0.15287958115183245, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 4.793428897857666, "fcm_dpo/q_t": 0.48804718255996704, "grad_norm": 27.781818389892578, "learning_rate": 4.961487700566646e-07, "logits/chosen": -0.6512953042984009, "logits/rejected": -0.6683382987976074, "logps/chosen": -268.6753845214844, "logps/ref_chosen": -273.0531005859375, "logps/ref_rejected": -246.8330841064453, "logps/rejected": -247.24880981445312, "loss": 5.372, "margin_dpo/margin_mean": 4.793429374694824, "margin_dpo/margin_std": 12.23983383178711, "step": 73 }, { "epoch": 0.1549738219895288, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 4.438821792602539, "fcm_dpo/q_t": 0.4889482259750366, "grad_norm": 30.2846736907959, "learning_rate": 4.958220635317885e-07, "logits/chosen": -0.7208205461502075, "logits/rejected": -0.6990806460380554, "logps/chosen": -338.8956298828125, "logps/ref_chosen": -342.2818908691406, "logps/ref_rejected": -330.0293884277344, "logps/rejected": -331.0819091796875, "loss": 5.3847, "margin_dpo/margin_mean": 4.438822269439697, "margin_dpo/margin_std": 11.789268493652344, "step": 74 }, { "epoch": 0.15706806282722513, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 6.654679298400879, "fcm_dpo/q_t": 0.4834195673465729, "grad_norm": 29.386520385742188, "learning_rate": 4.954821743156767e-07, "logits/chosen": -0.6316950917243958, "logits/rejected": -0.632973849773407, "logps/chosen": -262.26519775390625, "logps/ref_chosen": -266.8641662597656, "logps/ref_rejected": -276.8699951171875, "logps/rejected": -278.92572021484375, "loss": 5.2967, "margin_dpo/margin_mean": 6.654679775238037, "margin_dpo/margin_std": 10.98399829864502, "step": 75 }, { "epoch": 0.15916230366492146, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 7.279214859008789, "fcm_dpo/q_t": 0.48191386461257935, "grad_norm": 29.20782470703125, "learning_rate": 4.951291206355559e-07, "logits/chosen": -0.7203505039215088, "logits/rejected": -0.7281088829040527, "logps/chosen": -277.01470947265625, "logps/ref_chosen": -281.174560546875, "logps/ref_rejected": -263.6067199707031, "logps/rejected": -266.72607421875, "loss": 5.2772, "margin_dpo/margin_mean": 7.279214859008789, "margin_dpo/margin_std": 12.870685577392578, "step": 76 }, { "epoch": 0.1612565445026178, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 5.79923152923584, "fcm_dpo/q_t": 0.4856007397174835, "grad_norm": 32.94340133666992, "learning_rate": 4.947629214246236e-07, "logits/chosen": -0.5545709729194641, "logits/rejected": -0.5626035928726196, "logps/chosen": -302.3695068359375, "logps/ref_chosen": -306.09527587890625, "logps/ref_rejected": -253.49569702148438, "logps/rejected": -255.5691375732422, "loss": 5.3375, "margin_dpo/margin_mean": 5.79923152923584, "margin_dpo/margin_std": 13.905439376831055, "step": 77 }, { "epoch": 0.16335078534031414, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 8.666434288024902, "fcm_dpo/q_t": 0.478497177362442, "grad_norm": 29.634822845458984, "learning_rate": 4.943835963210323e-07, "logits/chosen": -0.677687406539917, "logits/rejected": -0.6719750165939331, "logps/chosen": -253.00672912597656, "logps/ref_chosen": -256.90234375, "logps/ref_rejected": -211.57154846191406, "logps/rejected": -216.3424072265625, "loss": 5.2278, "margin_dpo/margin_mean": 8.666435241699219, "margin_dpo/margin_std": 14.475006103515625, "step": 78 }, { "epoch": 0.16544502617801046, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 8.294734954833984, "fcm_dpo/q_t": 0.47941556572914124, "grad_norm": 29.799663543701172, "learning_rate": 4.939911656668361e-07, "logits/chosen": -0.6619881987571716, "logits/rejected": -0.6830765604972839, "logps/chosen": -263.20843505859375, "logps/ref_chosen": -266.2735595703125, "logps/ref_rejected": -251.57257080078125, "logps/rejected": -256.80218505859375, "loss": 5.2444, "margin_dpo/margin_mean": 8.294734001159668, "margin_dpo/margin_std": 15.015451431274414, "step": 79 }, { "epoch": 0.16753926701570682, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 6.695713043212891, "fcm_dpo/q_t": 0.48338207602500916, "grad_norm": 28.94121742248535, "learning_rate": 4.935856505068998e-07, "logits/chosen": -0.666534423828125, "logits/rejected": -0.6961286067962646, "logps/chosen": -285.9317932128906, "logps/ref_chosen": -287.8509826660156, "logps/ref_rejected": -256.0766296386719, "logps/rejected": -260.8531494140625, "loss": 5.3016, "margin_dpo/margin_mean": 6.695713043212891, "margin_dpo/margin_std": 13.224631309509277, "step": 80 }, { "epoch": 0.16963350785340314, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 7.167357444763184, "fcm_dpo/q_t": 0.482227087020874, "grad_norm": 28.103904724121094, "learning_rate": 4.93167072587771e-07, "logits/chosen": -0.6446385383605957, "logits/rejected": -0.6438944935798645, "logps/chosen": -266.1433410644531, "logps/ref_chosen": -268.5232238769531, "logps/ref_rejected": -237.81137084960938, "logps/rejected": -242.59884643554688, "loss": 5.2934, "margin_dpo/margin_mean": 7.167357921600342, "margin_dpo/margin_std": 16.743844985961914, "step": 81 }, { "epoch": 0.17172774869109947, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 7.904582977294922, "fcm_dpo/q_t": 0.480448454618454, "grad_norm": 27.851560592651367, "learning_rate": 4.92735454356513e-07, "logits/chosen": -0.7309600710868835, "logits/rejected": -0.7388311624526978, "logps/chosen": -276.9690246582031, "logps/ref_chosen": -279.36395263671875, "logps/ref_rejected": -236.51365661621094, "logps/rejected": -242.02334594726562, "loss": 5.2636, "margin_dpo/margin_mean": 7.904582977294922, "margin_dpo/margin_std": 15.73293685913086, "step": 82 }, { "epoch": 0.17382198952879582, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 8.677565574645996, "fcm_dpo/q_t": 0.4784577488899231, "grad_norm": 30.75320816040039, "learning_rate": 4.922908189595017e-07, "logits/chosen": -0.6892061233520508, "logits/rejected": -0.6728148460388184, "logps/chosen": -273.9486083984375, "logps/ref_chosen": -274.21923828125, "logps/ref_rejected": -276.2212219238281, "logps/rejected": -284.628173828125, "loss": 5.2424, "margin_dpo/margin_mean": 8.67756462097168, "margin_dpo/margin_std": 18.219993591308594, "step": 83 }, { "epoch": 0.17591623036649215, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 7.32714319229126, "fcm_dpo/q_t": 0.48182541131973267, "grad_norm": 29.90275764465332, "learning_rate": 4.918331902411841e-07, "logits/chosen": -0.7316781282424927, "logits/rejected": -0.7458916902542114, "logps/chosen": -293.8858337402344, "logps/ref_chosen": -294.3975524902344, "logps/ref_rejected": -279.81884765625, "logps/rejected": -286.63421630859375, "loss": 5.2887, "margin_dpo/margin_mean": 7.327144145965576, "margin_dpo/margin_std": 16.910655975341797, "step": 84 }, { "epoch": 0.17801047120418848, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 5.808302879333496, "fcm_dpo/q_t": 0.48558884859085083, "grad_norm": 29.378389358520508, "learning_rate": 4.913625927427995e-07, "logits/chosen": -0.6668398380279541, "logits/rejected": -0.6751588582992554, "logps/chosen": -245.18560791015625, "logps/ref_chosen": -243.66220092773438, "logps/ref_rejected": -263.9421691894531, "logps/rejected": -271.2738952636719, "loss": 5.3432, "margin_dpo/margin_mean": 5.808302402496338, "margin_dpo/margin_std": 15.501203536987305, "step": 85 }, { "epoch": 0.18010471204188483, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 8.855512619018555, "fcm_dpo/q_t": 0.47801530361175537, "grad_norm": 34.95083236694336, "learning_rate": 4.908790517010636e-07, "logits/chosen": -0.6866583228111267, "logits/rejected": -0.6797590255737305, "logps/chosen": -308.28826904296875, "logps/ref_chosen": -309.4306945800781, "logps/ref_rejected": -290.91278076171875, "logps/rejected": -298.6258544921875, "loss": 5.2308, "margin_dpo/margin_mean": 8.855510711669922, "margin_dpo/margin_std": 17.44927406311035, "step": 86 }, { "epoch": 0.18219895287958116, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 10.284191131591797, "fcm_dpo/q_t": 0.4746079742908478, "grad_norm": 29.6323184967041, "learning_rate": 4.903825930468148e-07, "logits/chosen": -0.7473527193069458, "logits/rejected": -0.7407711744308472, "logps/chosen": -278.1808166503906, "logps/ref_chosen": -278.0277099609375, "logps/ref_rejected": -245.70123291015625, "logps/rejected": -256.13848876953125, "loss": 5.1829, "margin_dpo/margin_mean": 10.284191131591797, "margin_dpo/margin_std": 18.875999450683594, "step": 87 }, { "epoch": 0.18429319371727748, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 8.930685997009277, "fcm_dpo/q_t": 0.47792893648147583, "grad_norm": 28.799327850341797, "learning_rate": 4.898732434036243e-07, "logits/chosen": -0.7647247314453125, "logits/rejected": -0.7806797027587891, "logps/chosen": -268.60662841796875, "logps/ref_chosen": -266.5148010253906, "logps/ref_rejected": -265.90081787109375, "logps/rejected": -276.9233093261719, "loss": 5.2343, "margin_dpo/margin_mean": 8.930685997009277, "margin_dpo/margin_std": 19.163759231567383, "step": 88 }, { "epoch": 0.18638743455497384, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 9.23874282836914, "fcm_dpo/q_t": 0.4771117866039276, "grad_norm": 30.643220901489258, "learning_rate": 4.893510300863676e-07, "logits/chosen": -0.7400345802307129, "logits/rejected": -0.7301192283630371, "logps/chosen": -265.6995544433594, "logps/ref_chosen": -265.6893005371094, "logps/ref_rejected": -251.49314880371094, "logps/rejected": -260.7421569824219, "loss": 5.2214, "margin_dpo/margin_mean": 9.238741874694824, "margin_dpo/margin_std": 18.232431411743164, "step": 89 }, { "epoch": 0.18848167539267016, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 8.30599594116211, "fcm_dpo/q_t": 0.4794096052646637, "grad_norm": 29.981016159057617, "learning_rate": 4.8881598109976e-07, "logits/chosen": -0.7483118176460266, "logits/rejected": -0.7575773000717163, "logps/chosen": -308.5583801269531, "logps/ref_chosen": -307.4250183105469, "logps/ref_rejected": -265.7172546386719, "logps/rejected": -275.1565856933594, "loss": 5.2543, "margin_dpo/margin_mean": 8.305996894836426, "margin_dpo/margin_std": 17.779386520385742, "step": 90 }, { "epoch": 0.1905759162303665, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 9.575455665588379, "fcm_dpo/q_t": 0.47638604044914246, "grad_norm": 33.001312255859375, "learning_rate": 4.882681251368548e-07, "logits/chosen": -0.6704283356666565, "logits/rejected": -0.6869698166847229, "logps/chosen": -237.86941528320312, "logps/ref_chosen": -235.74098205566406, "logps/ref_rejected": -226.6428985595703, "logps/rejected": -238.34674072265625, "loss": 5.2137, "margin_dpo/margin_mean": 9.575454711914062, "margin_dpo/margin_std": 19.24980926513672, "step": 91 }, { "epoch": 0.19267015706806281, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 10.090764999389648, "fcm_dpo/q_t": 0.475067675113678, "grad_norm": 33.0310173034668, "learning_rate": 4.877074915775048e-07, "logits/chosen": -0.7317771911621094, "logits/rejected": -0.7149327993392944, "logps/chosen": -286.60858154296875, "logps/ref_chosen": -283.4475402832031, "logps/ref_rejected": -273.134033203125, "logps/rejected": -286.38580322265625, "loss": 5.1997, "margin_dpo/margin_mean": 10.090764999389648, "margin_dpo/margin_std": 21.30156707763672, "step": 92 }, { "epoch": 0.19476439790575917, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 9.687959671020508, "fcm_dpo/q_t": 0.4760984778404236, "grad_norm": 28.729263305664062, "learning_rate": 4.871341104867864e-07, "logits/chosen": -0.7301860451698303, "logits/rejected": -0.7536556124687195, "logps/chosen": -235.7203826904297, "logps/ref_chosen": -233.33714294433594, "logps/ref_rejected": -230.54273986816406, "logps/rejected": -242.61395263671875, "loss": 5.2079, "margin_dpo/margin_mean": 9.687959671020508, "margin_dpo/margin_std": 19.470970153808594, "step": 93 }, { "epoch": 0.1968586387434555, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 9.32542610168457, "fcm_dpo/q_t": 0.47693106532096863, "grad_norm": 31.511959075927734, "learning_rate": 4.865480126133871e-07, "logits/chosen": -0.6908999681472778, "logits/rejected": -0.7126898169517517, "logps/chosen": -297.0324401855469, "logps/ref_chosen": -294.6528015136719, "logps/ref_rejected": -283.657958984375, "logps/rejected": -295.3630065917969, "loss": 5.2322, "margin_dpo/margin_mean": 9.325427055358887, "margin_dpo/margin_std": 21.890119552612305, "step": 94 }, { "epoch": 0.19895287958115182, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 10.206320762634277, "fcm_dpo/q_t": 0.4748663306236267, "grad_norm": 33.60295867919922, "learning_rate": 4.859492293879573e-07, "logits/chosen": -0.7269071936607361, "logits/rejected": -0.7481443881988525, "logps/chosen": -314.99395751953125, "logps/ref_chosen": -311.6697082519531, "logps/ref_rejected": -262.7471923828125, "logps/rejected": -276.27777099609375, "loss": 5.2042, "margin_dpo/margin_mean": 10.206320762634277, "margin_dpo/margin_std": 22.42072296142578, "step": 95 }, { "epoch": 0.20104712041884817, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 10.304360389709473, "fcm_dpo/q_t": 0.4746476709842682, "grad_norm": 34.0047721862793, "learning_rate": 4.853377929214243e-07, "logits/chosen": -0.6979155540466309, "logits/rejected": -0.7102517485618591, "logps/chosen": -287.2977294921875, "logps/ref_chosen": -282.55596923828125, "logps/ref_rejected": -242.71588134765625, "logps/rejected": -257.7619934082031, "loss": 5.2021, "margin_dpo/margin_mean": 10.304360389709473, "margin_dpo/margin_std": 23.36362648010254, "step": 96 }, { "epoch": 0.2031413612565445, "fcm_dpo/beta": 0.009999998845160007, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 12.058700561523438, "fcm_dpo/q_t": 0.47029611468315125, "grad_norm": 32.09419250488281, "learning_rate": 4.847137360032699e-07, "logits/chosen": -0.7502722144126892, "logits/rejected": -0.7376055717468262, "logps/chosen": -307.92425537109375, "logps/ref_chosen": -303.57781982421875, "logps/ref_rejected": -264.22491455078125, "logps/rejected": -280.6300964355469, "loss": 5.1301, "margin_dpo/margin_mean": 12.05870246887207, "margin_dpo/margin_std": 22.335546493530273, "step": 97 }, { "epoch": 0.20523560209424083, "fcm_dpo/beta": 0.010166899301111698, "fcm_dpo/delta": 0.03736456483602524, "fcm_dpo/margin": 12.270679473876953, "fcm_dpo/q_t": 0.4697565734386444, "grad_norm": 35.15353012084961, "learning_rate": 4.84077092099773e-07, "logits/chosen": -0.7689064145088196, "logits/rejected": -0.7808451056480408, "logps/chosen": -291.73529052734375, "logps/ref_chosen": -286.8303527832031, "logps/ref_rejected": -278.08331298828125, "logps/rejected": -295.2589416503906, "loss": 5.1298, "margin_dpo/margin_mean": 12.270679473876953, "margin_dpo/margin_std": 22.752973556518555, "step": 98 }, { "epoch": 0.20732984293193718, "fcm_dpo/beta": 0.010414928197860718, "fcm_dpo/delta": 0.021457988768815994, "fcm_dpo/margin": 12.633167266845703, "fcm_dpo/q_t": 0.46785154938697815, "grad_norm": 33.9055290222168, "learning_rate": 4.834278953522137e-07, "logits/chosen": -0.7366681098937988, "logits/rejected": -0.7495771646499634, "logps/chosen": -285.1845703125, "logps/ref_chosen": -279.92120361328125, "logps/ref_rejected": -250.3365478515625, "logps/rejected": -268.2330627441406, "loss": 5.121, "margin_dpo/margin_mean": 12.633166313171387, "margin_dpo/margin_std": 27.084667205810547, "step": 99 }, { "epoch": 0.2094240837696335, "fcm_dpo/beta": 0.010724691674113274, "fcm_dpo/delta": 0.03860073536634445, "fcm_dpo/margin": 12.622820854187012, "fcm_dpo/q_t": 0.4671909511089325, "grad_norm": 37.84049987792969, "learning_rate": 4.827661805750437e-07, "logits/chosen": -0.785658597946167, "logits/rejected": -0.7990057468414307, "logps/chosen": -304.5786437988281, "logps/ref_chosen": -296.8276672363281, "logps/ref_rejected": -275.56146240234375, "logps/rejected": -295.9351806640625, "loss": 5.0963, "margin_dpo/margin_mean": 12.622820854187012, "margin_dpo/margin_std": 24.152620315551758, "step": 100 }, { "epoch": 0.21151832460732983, "fcm_dpo/beta": 0.010941779240965843, "fcm_dpo/delta": 0.03345339000225067, "fcm_dpo/margin": 15.069441795349121, "fcm_dpo/q_t": 0.4600520133972168, "grad_norm": 35.19692611694336, "learning_rate": 4.820919832540181e-07, "logits/chosen": -0.7683415412902832, "logits/rejected": -0.7820852398872375, "logps/chosen": -257.931396484375, "logps/ref_chosen": -252.74203491210938, "logps/ref_rejected": -276.4185485839844, "logps/rejected": -296.6773986816406, "loss": 5.0018, "margin_dpo/margin_mean": 15.069441795349121, "margin_dpo/margin_std": 26.246978759765625, "step": 101 }, { "epoch": 0.2136125654450262, "fcm_dpo/beta": 0.011548048816621304, "fcm_dpo/delta": 0.08071689307689667, "fcm_dpo/margin": 14.844532012939453, "fcm_dpo/q_t": 0.45827770233154297, "grad_norm": 37.058868408203125, "learning_rate": 4.814053395442932e-07, "logits/chosen": -0.7501267194747925, "logits/rejected": -0.7462649941444397, "logps/chosen": -224.66371154785156, "logps/ref_chosen": -219.5537109375, "logps/ref_rejected": -231.90853881835938, "logps/rejected": -251.86306762695312, "loss": 4.9782, "margin_dpo/margin_mean": 14.844533920288086, "margin_dpo/margin_std": 25.003633499145508, "step": 102 }, { "epoch": 0.2157068062827225, "fcm_dpo/beta": 0.01229560561478138, "fcm_dpo/delta": 0.06470398604869843, "fcm_dpo/margin": 13.61020278930664, "fcm_dpo/q_t": 0.4601737856864929, "grad_norm": 39.8343391418457, "learning_rate": 4.807062862684873e-07, "logits/chosen": -0.7709028124809265, "logits/rejected": -0.7678197026252747, "logps/chosen": -264.3942565917969, "logps/ref_chosen": -259.6750793457031, "logps/ref_rejected": -278.7400817871094, "logps/rejected": -297.0694885253906, "loss": 5.0107, "margin_dpo/margin_mean": 13.61020278930664, "margin_dpo/margin_std": 25.62775993347168, "step": 103 }, { "epoch": 0.21780104712041884, "fcm_dpo/beta": 0.01250280998647213, "fcm_dpo/delta": 0.023114658892154694, "fcm_dpo/margin": 10.093114852905273, "fcm_dpo/q_t": 0.4694807529449463, "grad_norm": 41.128822326660156, "learning_rate": 4.799948609147061e-07, "logits/chosen": -0.765766978263855, "logits/rejected": -0.771574079990387, "logps/chosen": -276.9814147949219, "logps/ref_chosen": -267.9741516113281, "logps/ref_rejected": -230.5306396484375, "logps/rejected": -249.6309814453125, "loss": 5.1671, "margin_dpo/margin_mean": 10.09311580657959, "margin_dpo/margin_std": 26.334136962890625, "step": 104 }, { "epoch": 0.2198952879581152, "fcm_dpo/beta": 0.012972178868949413, "fcm_dpo/delta": 0.06148073077201843, "fcm_dpo/margin": 20.653295516967773, "fcm_dpo/q_t": 0.4356263279914856, "grad_norm": 43.76247024536133, "learning_rate": 4.792711016345321e-07, "logits/chosen": -0.7641775608062744, "logits/rejected": -0.7758720517158508, "logps/chosen": -327.5227966308594, "logps/ref_chosen": -322.25482177734375, "logps/ref_rejected": -279.02978515625, "logps/rejected": -304.9510498046875, "loss": 4.6816, "margin_dpo/margin_mean": 20.653295516967773, "margin_dpo/margin_std": 27.019716262817383, "step": 105 }, { "epoch": 0.22198952879581152, "fcm_dpo/beta": 0.01369519904255867, "fcm_dpo/delta": 0.049693141132593155, "fcm_dpo/margin": 12.526891708374023, "fcm_dpo/q_t": 0.4589945077896118, "grad_norm": 50.769107818603516, "learning_rate": 4.785350472409791e-07, "logits/chosen": -0.7395763397216797, "logits/rejected": -0.7750528454780579, "logps/chosen": -308.6379089355469, "logps/ref_chosen": -296.15777587890625, "logps/ref_rejected": -266.2691650390625, "logps/rejected": -291.2762145996094, "loss": 5.0575, "margin_dpo/margin_mean": 12.52688980102539, "margin_dpo/margin_std": 29.427845001220703, "step": 106 }, { "epoch": 0.22408376963350785, "fcm_dpo/beta": 0.014684994705021381, "fcm_dpo/delta": 0.11428901553153992, "fcm_dpo/margin": 19.656145095825195, "fcm_dpo/q_t": 0.431806743144989, "grad_norm": 52.60951232910156, "learning_rate": 4.777867372064105e-07, "logits/chosen": -0.780152440071106, "logits/rejected": -0.7732595205307007, "logps/chosen": -311.265869140625, "logps/ref_chosen": -306.996337890625, "logps/ref_rejected": -296.79412841796875, "logps/rejected": -320.7198181152344, "loss": 4.6426, "margin_dpo/margin_mean": 19.656145095825195, "margin_dpo/margin_std": 27.61237335205078, "step": 107 }, { "epoch": 0.2261780104712042, "fcm_dpo/beta": 0.015535693615674973, "fcm_dpo/delta": 0.054226309061050415, "fcm_dpo/margin": 18.248409271240234, "fcm_dpo/q_t": 0.4343521296977997, "grad_norm": 162.06214904785156, "learning_rate": 4.770262116604223e-07, "logits/chosen": -0.7690107822418213, "logits/rejected": -0.781175434589386, "logps/chosen": -300.4916687011719, "logps/ref_chosen": -295.1526794433594, "logps/ref_rejected": -235.974853515625, "logps/rejected": -259.562255859375, "loss": 4.7324, "margin_dpo/margin_mean": 18.248409271240234, "margin_dpo/margin_std": 30.292400360107422, "step": 108 }, { "epoch": 0.22827225130890053, "fcm_dpo/beta": 0.016129182651638985, "fcm_dpo/delta": 0.03493024408817291, "fcm_dpo/margin": 20.186677932739258, "fcm_dpo/q_t": 0.4253409504890442, "grad_norm": 57.66378402709961, "learning_rate": 4.7625351138769166e-07, "logits/chosen": -0.7831578850746155, "logits/rejected": -0.7787821888923645, "logps/chosen": -333.96337890625, "logps/ref_chosen": -325.9248046875, "logps/ref_rejected": -279.15423583984375, "logps/rejected": -307.3795166015625, "loss": 4.6025, "margin_dpo/margin_mean": 20.186674118041992, "margin_dpo/margin_std": 31.418649673461914, "step": 109 }, { "epoch": 0.23036649214659685, "fcm_dpo/beta": 0.016638344153761864, "fcm_dpo/delta": 0.03626322001218796, "fcm_dpo/margin": 18.989017486572266, "fcm_dpo/q_t": 0.42738083004951477, "grad_norm": 53.80938720703125, "learning_rate": 4.75468677825789e-07, "logits/chosen": -0.8024147748947144, "logits/rejected": -0.7910180687904358, "logps/chosen": -282.9509582519531, "logps/ref_chosen": -274.439208984375, "logps/ref_rejected": -260.0552062988281, "logps/rejected": -287.5560302734375, "loss": 4.6842, "margin_dpo/margin_mean": 18.989017486572266, "margin_dpo/margin_std": 33.461700439453125, "step": 110 }, { "epoch": 0.2324607329842932, "fcm_dpo/beta": 0.016925374045968056, "fcm_dpo/delta": 0.004534606821835041, "fcm_dpo/margin": 20.09602928161621, "fcm_dpo/q_t": 0.4227144718170166, "grad_norm": 59.05382537841797, "learning_rate": 4.7467175306295647e-07, "logits/chosen": -0.8369516134262085, "logits/rejected": -0.8165923953056335, "logps/chosen": -338.4478454589844, "logps/ref_chosen": -329.2361755371094, "logps/ref_rejected": -287.82830810546875, "logps/rejected": -317.135986328125, "loss": 4.6353, "margin_dpo/margin_mean": 20.09602928161621, "margin_dpo/margin_std": 33.968441009521484, "step": 111 }, { "epoch": 0.23455497382198953, "fcm_dpo/beta": 0.016810838133096695, "fcm_dpo/delta": 0.002743429969996214, "fcm_dpo/margin": 12.950404167175293, "fcm_dpo/q_t": 0.45069649815559387, "grad_norm": 67.04423522949219, "learning_rate": 4.7386277983585053e-07, "logits/chosen": -0.7323642373085022, "logits/rejected": -0.762016773223877, "logps/chosen": -271.5980529785156, "logps/ref_chosen": -257.0593566894531, "logps/ref_rejected": -272.9595031738281, "logps/rejected": -300.4485778808594, "loss": 5.1077, "margin_dpo/margin_mean": 12.950403213500977, "margin_dpo/margin_std": 35.27339172363281, "step": 112 }, { "epoch": 0.23664921465968586, "fcm_dpo/beta": 0.016340255737304688, "fcm_dpo/delta": -0.055206988006830215, "fcm_dpo/margin": 23.974803924560547, "fcm_dpo/q_t": 0.41226544976234436, "grad_norm": 59.851139068603516, "learning_rate": 4.7304180152725024e-07, "logits/chosen": -0.7996099591255188, "logits/rejected": -0.8028220534324646, "logps/chosen": -297.7246398925781, "logps/ref_chosen": -286.0416564941406, "logps/ref_rejected": -270.374267578125, "logps/rejected": -306.0320739746094, "loss": 4.5109, "margin_dpo/margin_mean": 23.974803924560547, "margin_dpo/margin_std": 38.330360412597656, "step": 113 }, { "epoch": 0.2387434554973822, "fcm_dpo/beta": 0.01674072816967964, "fcm_dpo/delta": 0.04583574831485748, "fcm_dpo/margin": 13.43380355834961, "fcm_dpo/q_t": 0.44763025641441345, "grad_norm": 63.377777099609375, "learning_rate": 4.7220886216373085e-07, "logits/chosen": -0.844326913356781, "logits/rejected": -0.8458607196807861, "logps/chosen": -275.0863037109375, "logps/ref_chosen": -260.0084533691406, "logps/ref_rejected": -246.67190551757812, "logps/rejected": -275.18353271484375, "loss": 5.0337, "margin_dpo/margin_mean": 13.433801651000977, "margin_dpo/margin_std": 34.03593444824219, "step": 114 }, { "epoch": 0.24083769633507854, "fcm_dpo/beta": 0.01667495258152485, "fcm_dpo/delta": 0.022258806973695755, "fcm_dpo/margin": 13.424970626831055, "fcm_dpo/q_t": 0.45159363746643066, "grad_norm": 71.67733001708984, "learning_rate": 4.7136400641330245e-07, "logits/chosen": -0.8421843647956848, "logits/rejected": -0.8031139969825745, "logps/chosen": -315.6531677246094, "logps/ref_chosen": -299.4229736328125, "logps/ref_rejected": -272.1186828613281, "logps/rejected": -301.77386474609375, "loss": 5.0871, "margin_dpo/margin_mean": 13.424969673156738, "margin_dpo/margin_std": 37.24310302734375, "step": 115 }, { "epoch": 0.24293193717277486, "fcm_dpo/beta": 0.01778252422809601, "fcm_dpo/delta": 0.06407217681407928, "fcm_dpo/margin": 17.605409622192383, "fcm_dpo/q_t": 0.4284539520740509, "grad_norm": 64.84068298339844, "learning_rate": 4.70507279583015e-07, "logits/chosen": -0.8544034957885742, "logits/rejected": -0.8170161843299866, "logps/chosen": -291.0081481933594, "logps/ref_chosen": -279.263916015625, "logps/ref_rejected": -253.6192169189453, "logps/rejected": -282.9688720703125, "loss": 4.744, "margin_dpo/margin_mean": 17.60540771484375, "margin_dpo/margin_std": 33.50556182861328, "step": 116 }, { "epoch": 0.2450261780104712, "fcm_dpo/beta": 0.017902228981256485, "fcm_dpo/delta": -0.05058136209845543, "fcm_dpo/margin": 19.721925735473633, "fcm_dpo/q_t": 0.41956445574760437, "grad_norm": 74.53150177001953, "learning_rate": 4.6963872761652834e-07, "logits/chosen": -0.8143470287322998, "logits/rejected": -0.8171401619911194, "logps/chosen": -273.25616455078125, "logps/ref_chosen": -259.2248840332031, "logps/ref_rejected": -229.3042755126953, "logps/rejected": -263.0574951171875, "loss": 4.6214, "margin_dpo/margin_mean": 19.721923828125, "margin_dpo/margin_std": 30.71808624267578, "step": 117 }, { "epoch": 0.24712041884816754, "fcm_dpo/beta": 0.01726110279560089, "fcm_dpo/delta": -0.024961750954389572, "fcm_dpo/margin": 22.500560760498047, "fcm_dpo/q_t": 0.4144339859485626, "grad_norm": 71.345458984375, "learning_rate": 4.687583970916486e-07, "logits/chosen": -0.7954049110412598, "logits/rejected": -0.7841992378234863, "logps/chosen": -285.67327880859375, "logps/ref_chosen": -267.0707092285156, "logps/ref_rejected": -272.7322082519531, "logps/rejected": -313.8353271484375, "loss": 4.6027, "margin_dpo/margin_mean": 22.50056266784668, "margin_dpo/margin_std": 39.91600036621094, "step": 118 }, { "epoch": 0.24921465968586387, "fcm_dpo/beta": 0.017213156446814537, "fcm_dpo/delta": 0.016311757266521454, "fcm_dpo/margin": 16.51951026916504, "fcm_dpo/q_t": 0.43743908405303955, "grad_norm": 75.18436431884766, "learning_rate": 4.6786633521783005e-07, "logits/chosen": -0.8822775483131409, "logits/rejected": -0.8836984634399414, "logps/chosen": -347.19122314453125, "logps/ref_chosen": -324.6766357421875, "logps/ref_rejected": -306.0322265625, "logps/rejected": -345.0663146972656, "loss": 4.9579, "margin_dpo/margin_mean": 16.519508361816406, "margin_dpo/margin_std": 38.816429138183594, "step": 119 }, { "epoch": 0.2513089005235602, "fcm_dpo/beta": 0.017053598538041115, "fcm_dpo/delta": -0.027553297579288483, "fcm_dpo/margin": 17.9389591217041, "fcm_dpo/q_t": 0.43243440985679626, "grad_norm": 66.44527435302734, "learning_rate": 4.669625898336438e-07, "logits/chosen": -0.8248355388641357, "logits/rejected": -0.839232325553894, "logps/chosen": -334.0506286621094, "logps/ref_chosen": -315.2617492675781, "logps/ref_rejected": -265.32501220703125, "logps/rejected": -302.0528564453125, "loss": 4.869, "margin_dpo/margin_mean": 17.9389591217041, "margin_dpo/margin_std": 38.128028869628906, "step": 120 }, { "epoch": 0.2534031413612565, "fcm_dpo/beta": 0.01718924008309841, "fcm_dpo/delta": 0.06284664571285248, "fcm_dpo/margin": 14.12536334991455, "fcm_dpo/q_t": 0.4474913775920868, "grad_norm": 73.3908462524414, "learning_rate": 4.6604720940421207e-07, "logits/chosen": -0.8356261253356934, "logits/rejected": -0.8507847785949707, "logps/chosen": -246.18655395507812, "logps/ref_chosen": -222.99609375, "logps/ref_rejected": -226.92860412597656, "logps/rejected": -264.244384765625, "loss": 5.0392, "margin_dpo/margin_mean": 14.125362396240234, "margin_dpo/margin_std": 37.465171813964844, "step": 121 }, { "epoch": 0.2554973821989529, "fcm_dpo/beta": 0.017536453902721405, "fcm_dpo/delta": 0.005643364042043686, "fcm_dpo/margin": 18.017147064208984, "fcm_dpo/q_t": 0.4305107891559601, "grad_norm": 70.16292572021484, "learning_rate": 4.651202430186092e-07, "logits/chosen": -0.89496248960495, "logits/rejected": -0.8574784994125366, "logps/chosen": -299.5643310546875, "logps/ref_chosen": -276.02630615234375, "logps/ref_rejected": -277.97418212890625, "logps/rejected": -319.52935791015625, "loss": 4.8959, "margin_dpo/margin_mean": 18.017147064208984, "margin_dpo/margin_std": 41.55096435546875, "step": 122 }, { "epoch": 0.25759162303664923, "fcm_dpo/beta": 0.01756231114268303, "fcm_dpo/delta": -0.03471625596284866, "fcm_dpo/margin": 23.458398818969727, "fcm_dpo/q_t": 0.40726321935653687, "grad_norm": 73.8258056640625, "learning_rate": 4.6418174038722924e-07, "logits/chosen": -0.8168793320655823, "logits/rejected": -0.8113296031951904, "logps/chosen": -346.0722351074219, "logps/ref_chosen": -328.1546325683594, "logps/ref_rejected": -280.6911315917969, "logps/rejected": -322.06719970703125, "loss": 4.5056, "margin_dpo/margin_mean": 23.458398818969727, "margin_dpo/margin_std": 37.95882034301758, "step": 123 }, { "epoch": 0.25968586387434556, "fcm_dpo/beta": 0.017159288749098778, "fcm_dpo/delta": 0.005674887448549271, "fcm_dpo/margin": 19.5401668548584, "fcm_dpo/q_t": 0.42332780361175537, "grad_norm": 76.41766357421875, "learning_rate": 4.6323175183912023e-07, "logits/chosen": -0.8458907604217529, "logits/rejected": -0.8133480548858643, "logps/chosen": -295.9541015625, "logps/ref_chosen": -275.6961975097656, "logps/ref_rejected": -225.361572265625, "logps/rejected": -265.1595764160156, "loss": 4.7102, "margin_dpo/margin_mean": 19.540164947509766, "margin_dpo/margin_std": 35.17292785644531, "step": 124 }, { "epoch": 0.2617801047120419, "fcm_dpo/beta": 0.017188817262649536, "fcm_dpo/delta": -0.03201991319656372, "fcm_dpo/margin": 18.89151382446289, "fcm_dpo/q_t": 0.43093550205230713, "grad_norm": 80.25518035888672, "learning_rate": 4.6227032831928483e-07, "logits/chosen": -0.8065889477729797, "logits/rejected": -0.7634490728378296, "logps/chosen": -299.75115966796875, "logps/ref_chosen": -278.06976318359375, "logps/ref_rejected": -265.63873291015625, "logps/rejected": -306.21160888671875, "loss": 4.9278, "margin_dpo/margin_mean": 18.89151382446289, "margin_dpo/margin_std": 44.717071533203125, "step": 125 }, { "epoch": 0.2638743455497382, "fcm_dpo/beta": 0.01680522970855236, "fcm_dpo/delta": 0.005955208092927933, "fcm_dpo/margin": 20.775150299072266, "fcm_dpo/q_t": 0.4213416278362274, "grad_norm": 78.756103515625, "learning_rate": 4.612975213859487e-07, "logits/chosen": -0.8176358938217163, "logits/rejected": -0.8314664959907532, "logps/chosen": -339.98236083984375, "logps/ref_chosen": -321.3960876464844, "logps/ref_rejected": -285.37664794921875, "logps/rejected": -324.7380676269531, "loss": 4.6984, "margin_dpo/margin_mean": 20.775150299072266, "margin_dpo/margin_std": 39.32340621948242, "step": 126 }, { "epoch": 0.26596858638743454, "fcm_dpo/beta": 0.01681065373122692, "fcm_dpo/delta": -0.07578819990158081, "fcm_dpo/margin": 24.14862823486328, "fcm_dpo/q_t": 0.4106398820877075, "grad_norm": 79.24022674560547, "learning_rate": 4.603133832077953e-07, "logits/chosen": -0.8952986598014832, "logits/rejected": -0.8467426300048828, "logps/chosen": -322.97332763671875, "logps/ref_chosen": -306.55877685546875, "logps/ref_rejected": -274.8651428222656, "logps/rejected": -315.4283447265625, "loss": 4.5653, "margin_dpo/margin_mean": 24.14862632751465, "margin_dpo/margin_std": 39.488861083984375, "step": 127 }, { "epoch": 0.2680628272251309, "fcm_dpo/beta": 0.015498126856982708, "fcm_dpo/delta": -0.057604167610406876, "fcm_dpo/margin": 27.36457061767578, "fcm_dpo/q_t": 0.40523380041122437, "grad_norm": 59.929752349853516, "learning_rate": 4.5931796656116837e-07, "logits/chosen": -0.7822437286376953, "logits/rejected": -0.783670961856842, "logps/chosen": -276.9512634277344, "logps/ref_chosen": -265.3973693847656, "logps/ref_rejected": -250.9737548828125, "logps/rejected": -289.8922119140625, "loss": 4.4608, "margin_dpo/margin_mean": 27.364572525024414, "margin_dpo/margin_std": 43.219398498535156, "step": 128 }, { "epoch": 0.27015706806282724, "fcm_dpo/beta": 0.015127632766962051, "fcm_dpo/delta": -0.0237518809735775, "fcm_dpo/margin": 24.08989906311035, "fcm_dpo/q_t": 0.4185941815376282, "grad_norm": 64.50642395019531, "learning_rate": 4.5831132482724193e-07, "logits/chosen": -0.8073078989982605, "logits/rejected": -0.809032678604126, "logps/chosen": -315.49456787109375, "logps/ref_chosen": -303.158447265625, "logps/ref_rejected": -275.9891052246094, "logps/rejected": -312.4151306152344, "loss": 4.622, "margin_dpo/margin_mean": 24.089900970458984, "margin_dpo/margin_std": 42.023170471191406, "step": 129 }, { "epoch": 0.27225130890052357, "fcm_dpo/beta": 0.015055293217301369, "fcm_dpo/delta": 0.02385116182267666, "fcm_dpo/margin": 23.168506622314453, "fcm_dpo/q_t": 0.42154157161712646, "grad_norm": 69.50651550292969, "learning_rate": 4.5729351198915705e-07, "logits/chosen": -0.7816063761711121, "logits/rejected": -0.8228086829185486, "logps/chosen": -300.2805480957031, "logps/ref_chosen": -286.4073486328125, "logps/ref_rejected": -294.38665771484375, "logps/rejected": -331.42840576171875, "loss": 4.6517, "margin_dpo/margin_mean": 23.168506622314453, "margin_dpo/margin_std": 41.75603103637695, "step": 130 }, { "epoch": 0.2743455497382199, "fcm_dpo/beta": 0.01558979507535696, "fcm_dpo/delta": 0.03358662500977516, "fcm_dpo/margin": 19.491979598999023, "fcm_dpo/q_t": 0.4319177567958832, "grad_norm": 82.41316986083984, "learning_rate": 4.5626458262912735e-07, "logits/chosen": -0.855750560760498, "logits/rejected": -0.8051818013191223, "logps/chosen": -327.0265808105469, "logps/ref_chosen": -311.5650634765625, "logps/ref_rejected": -291.62432861328125, "logps/rejected": -326.577880859375, "loss": 4.863, "margin_dpo/margin_mean": 19.491979598999023, "margin_dpo/margin_std": 42.24031066894531, "step": 131 }, { "epoch": 0.2764397905759162, "fcm_dpo/beta": 0.015864521265029907, "fcm_dpo/delta": -0.031978119164705276, "fcm_dpo/margin": 27.09807014465332, "fcm_dpo/q_t": 0.4063335061073303, "grad_norm": 95.45539093017578, "learning_rate": 4.5522459192551166e-07, "logits/chosen": -0.8062165379524231, "logits/rejected": -0.7902382612228394, "logps/chosen": -281.4801025390625, "logps/ref_chosen": -270.0818176269531, "logps/ref_rejected": -284.3084411621094, "logps/rejected": -322.8048095703125, "loss": 4.5304, "margin_dpo/margin_mean": 27.098068237304688, "margin_dpo/margin_std": 45.48096466064453, "step": 132 }, { "epoch": 0.27853403141361255, "fcm_dpo/beta": 0.015226936899125576, "fcm_dpo/delta": -0.009090131148695946, "fcm_dpo/margin": 24.486173629760742, "fcm_dpo/q_t": 0.41608279943466187, "grad_norm": 59.91682434082031, "learning_rate": 4.541735956498554e-07, "logits/chosen": -0.8516502976417542, "logits/rejected": -0.8550176024436951, "logps/chosen": -298.2793884277344, "logps/ref_chosen": -285.6213684082031, "logps/ref_rejected": -251.19386291503906, "logps/rejected": -288.3380432128906, "loss": 4.607, "margin_dpo/margin_mean": 24.486173629760742, "margin_dpo/margin_std": 42.33135223388672, "step": 133 }, { "epoch": 0.2806282722513089, "fcm_dpo/beta": 0.015293042175471783, "fcm_dpo/delta": 0.0003851773217320442, "fcm_dpo/margin": 19.430198669433594, "fcm_dpo/q_t": 0.43152743577957153, "grad_norm": 61.967952728271484, "learning_rate": 4.5311165016389914e-07, "logits/chosen": -0.8527645468711853, "logits/rejected": -0.8533920049667358, "logps/chosen": -341.70172119140625, "logps/ref_chosen": -318.92083740234375, "logps/ref_rejected": -293.1894836425781, "logps/rejected": -335.4005432128906, "loss": 4.8138, "margin_dpo/margin_mean": 19.430200576782227, "margin_dpo/margin_std": 38.5192985534668, "step": 134 }, { "epoch": 0.28272251308900526, "fcm_dpo/beta": 0.015089647844433784, "fcm_dpo/delta": -0.019326163455843925, "fcm_dpo/margin": 22.623905181884766, "fcm_dpo/q_t": 0.4212813675403595, "grad_norm": 82.32707214355469, "learning_rate": 4.520388124165564e-07, "logits/chosen": -0.7468621134757996, "logits/rejected": -0.7879329323768616, "logps/chosen": -311.9961853027344, "logps/ref_chosen": -292.8217468261719, "logps/ref_rejected": -269.2896728515625, "logps/rejected": -311.0880432128906, "loss": 4.6979, "margin_dpo/margin_mean": 22.623905181884766, "margin_dpo/margin_std": 40.231651306152344, "step": 135 }, { "epoch": 0.2848167539267016, "fcm_dpo/beta": 0.014683660119771957, "fcm_dpo/delta": -0.03437824174761772, "fcm_dpo/margin": 25.665916442871094, "fcm_dpo/q_t": 0.4163571298122406, "grad_norm": 78.79556274414062, "learning_rate": 4.5095513994085974e-07, "logits/chosen": -0.808585524559021, "logits/rejected": -0.8038730621337891, "logps/chosen": -292.9419860839844, "logps/ref_chosen": -272.8525390625, "logps/ref_rejected": -252.68202209472656, "logps/rejected": -298.4374084472656, "loss": 4.6275, "margin_dpo/margin_mean": 25.665916442871094, "margin_dpo/margin_std": 45.39299392700195, "step": 136 }, { "epoch": 0.2869109947643979, "fcm_dpo/beta": 0.014740757644176483, "fcm_dpo/delta": 0.006549544632434845, "fcm_dpo/margin": 22.0756893157959, "fcm_dpo/q_t": 0.4280848503112793, "grad_norm": 67.095458984375, "learning_rate": 4.498606908508753e-07, "logits/chosen": -0.8490076661109924, "logits/rejected": -0.8347117304801941, "logps/chosen": -323.59130859375, "logps/ref_chosen": -300.7522277832031, "logps/ref_rejected": -286.1935119628906, "logps/rejected": -331.1082458496094, "loss": 4.7574, "margin_dpo/margin_mean": 22.07569122314453, "margin_dpo/margin_std": 43.516868591308594, "step": 137 }, { "epoch": 0.28900523560209423, "fcm_dpo/beta": 0.01454901322722435, "fcm_dpo/delta": -0.0002657044678926468, "fcm_dpo/margin": 24.885833740234375, "fcm_dpo/q_t": 0.42203420400619507, "grad_norm": 63.38542556762695, "learning_rate": 4.487555238385862e-07, "logits/chosen": -0.7799475789070129, "logits/rejected": -0.7601336240768433, "logps/chosen": -309.79620361328125, "logps/ref_chosen": -288.9369812011719, "logps/ref_rejected": -263.7076416015625, "logps/rejected": -309.45269775390625, "loss": 4.7118, "margin_dpo/margin_mean": 24.885831832885742, "margin_dpo/margin_std": 49.14541244506836, "step": 138 }, { "epoch": 0.29109947643979056, "fcm_dpo/beta": 0.014541015960276127, "fcm_dpo/delta": 0.017182698473334312, "fcm_dpo/margin": 16.95909309387207, "fcm_dpo/q_t": 0.4451315104961395, "grad_norm": 65.62071228027344, "learning_rate": 4.476396981707453e-07, "logits/chosen": -0.8039760589599609, "logits/rejected": -0.8343940377235413, "logps/chosen": -290.50616455078125, "logps/ref_chosen": -270.0443115234375, "logps/ref_rejected": -267.3226013183594, "logps/rejected": -304.7435302734375, "loss": 5.0197, "margin_dpo/margin_mean": 16.95909309387207, "margin_dpo/margin_std": 42.556758880615234, "step": 139 }, { "epoch": 0.2931937172774869, "fcm_dpo/beta": 0.014604137279093266, "fcm_dpo/delta": -0.03599924594163895, "fcm_dpo/margin": 26.53323745727539, "fcm_dpo/q_t": 0.4121074080467224, "grad_norm": 72.81695556640625, "learning_rate": 4.4651327368569684e-07, "logits/chosen": -0.8876050710678101, "logits/rejected": -0.856640636920929, "logps/chosen": -302.76434326171875, "logps/ref_chosen": -282.9555969238281, "logps/ref_rejected": -251.17181396484375, "logps/rejected": -297.5137939453125, "loss": 4.495, "margin_dpo/margin_mean": 26.53323745727539, "margin_dpo/margin_std": 40.214508056640625, "step": 140 }, { "epoch": 0.29528795811518327, "fcm_dpo/beta": 0.01404612883925438, "fcm_dpo/delta": -0.04051290452480316, "fcm_dpo/margin": 29.221651077270508, "fcm_dpo/q_t": 0.4060809910297394, "grad_norm": 59.52610778808594, "learning_rate": 4.453763107901675e-07, "logits/chosen": -0.8155990242958069, "logits/rejected": -0.8117971420288086, "logps/chosen": -314.1397399902344, "logps/ref_chosen": -296.3001708984375, "logps/ref_rejected": -279.8486633300781, "logps/rejected": -326.9098815917969, "loss": 4.4761, "margin_dpo/margin_mean": 29.221649169921875, "margin_dpo/margin_std": 45.766151428222656, "step": 141 }, { "epoch": 0.2973821989528796, "fcm_dpo/beta": 0.01387177873402834, "fcm_dpo/delta": -0.0016752388328313828, "fcm_dpo/margin": 23.28960418701172, "fcm_dpo/q_t": 0.42782607674598694, "grad_norm": 62.42996597290039, "learning_rate": 4.4422887045602674e-07, "logits/chosen": -0.8308397531509399, "logits/rejected": -0.8334007263183594, "logps/chosen": -321.20721435546875, "logps/ref_chosen": -300.56585693359375, "logps/ref_rejected": -231.43316650390625, "logps/rejected": -275.3641357421875, "loss": 4.8108, "margin_dpo/margin_mean": 23.28960418701172, "margin_dpo/margin_std": 48.74913024902344, "step": 142 }, { "epoch": 0.2994764397905759, "fcm_dpo/beta": 0.01358854304999113, "fcm_dpo/delta": -0.025667672976851463, "fcm_dpo/margin": 28.89480209350586, "fcm_dpo/q_t": 0.4115155041217804, "grad_norm": 58.460384368896484, "learning_rate": 4.4307101421701755e-07, "logits/chosen": -0.8297260403633118, "logits/rejected": -0.8115658760070801, "logps/chosen": -318.246337890625, "logps/ref_chosen": -296.73236083984375, "logps/ref_rejected": -266.45257568359375, "logps/rejected": -316.8614196777344, "loss": 4.4811, "margin_dpo/margin_mean": 28.894800186157227, "margin_dpo/margin_std": 43.21503448486328, "step": 143 }, { "epoch": 0.30157068062827225, "fcm_dpo/beta": 0.01343129575252533, "fcm_dpo/delta": -0.006690308451652527, "fcm_dpo/margin": 23.162403106689453, "fcm_dpo/q_t": 0.43112730979919434, "grad_norm": 63.20121383666992, "learning_rate": 4.419028041654559e-07, "logits/chosen": -0.878182590007782, "logits/rejected": -0.864521324634552, "logps/chosen": -323.18084716796875, "logps/ref_chosen": -298.843994140625, "logps/ref_rejected": -266.120849609375, "logps/rejected": -313.6200866699219, "loss": 4.8203, "margin_dpo/margin_mean": 23.162403106689453, "margin_dpo/margin_std": 48.521446228027344, "step": 144 }, { "epoch": 0.3036649214659686, "fcm_dpo/beta": 0.012913926504552364, "fcm_dpo/delta": -0.08783036470413208, "fcm_dpo/margin": 31.798524856567383, "fcm_dpo/q_t": 0.4062502682209015, "grad_norm": 59.772727966308594, "learning_rate": 4.4072430294890166e-07, "logits/chosen": -0.8632069826126099, "logits/rejected": -0.8655521273612976, "logps/chosen": -296.8944091796875, "logps/ref_chosen": -275.7528381347656, "logps/ref_rejected": -214.74807739257812, "logps/rejected": -267.68817138671875, "loss": 4.4332, "margin_dpo/margin_mean": 31.798524856567383, "margin_dpo/margin_std": 45.571964263916016, "step": 145 }, { "epoch": 0.3057591623036649, "fcm_dpo/beta": 0.01257642637938261, "fcm_dpo/delta": -0.011518063023686409, "fcm_dpo/margin": 30.71509552001953, "fcm_dpo/q_t": 0.412579208612442, "grad_norm": 73.49806213378906, "learning_rate": 4.395355737667985e-07, "logits/chosen": -0.8509379625320435, "logits/rejected": -0.8461902737617493, "logps/chosen": -308.1560363769531, "logps/ref_chosen": -277.09820556640625, "logps/ref_rejected": -265.41046142578125, "logps/rejected": -327.1833190917969, "loss": 4.493, "margin_dpo/margin_mean": 30.71509552001953, "margin_dpo/margin_std": 45.785736083984375, "step": 146 }, { "epoch": 0.3078534031413613, "fcm_dpo/beta": 0.01252201572060585, "fcm_dpo/delta": -0.038640253245830536, "fcm_dpo/margin": 23.62162971496582, "fcm_dpo/q_t": 0.4322171211242676, "grad_norm": 72.45833587646484, "learning_rate": 4.3833668036708483e-07, "logits/chosen": -0.8362576365470886, "logits/rejected": -0.8324323892593384, "logps/chosen": -325.21533203125, "logps/ref_chosen": -291.4185791015625, "logps/ref_rejected": -253.43051147460938, "logps/rejected": -310.84893798828125, "loss": 4.8467, "margin_dpo/margin_mean": 23.621631622314453, "margin_dpo/margin_std": 47.34939193725586, "step": 147 }, { "epoch": 0.3099476439790576, "fcm_dpo/beta": 0.012431098148226738, "fcm_dpo/delta": 0.056200552731752396, "fcm_dpo/margin": 23.877735137939453, "fcm_dpo/q_t": 0.43532854318618774, "grad_norm": 64.35401916503906, "learning_rate": 4.3712768704277524e-07, "logits/chosen": -0.8974530100822449, "logits/rejected": -0.8999665379524231, "logps/chosen": -275.08160400390625, "logps/ref_chosen": -236.74850463867188, "logps/ref_rejected": -231.4674072265625, "logps/rejected": -293.67816162109375, "loss": 4.865, "margin_dpo/margin_mean": 23.877735137939453, "margin_dpo/margin_std": 53.183143615722656, "step": 148 }, { "epoch": 0.31204188481675393, "fcm_dpo/beta": 0.012634760700166225, "fcm_dpo/delta": -0.0065022800117731094, "fcm_dpo/margin": 32.123836517333984, "fcm_dpo/q_t": 0.4069136083126068, "grad_norm": 76.46810150146484, "learning_rate": 4.3590865862851263e-07, "logits/chosen": -0.8669228553771973, "logits/rejected": -0.8517608046531677, "logps/chosen": -360.485595703125, "logps/ref_chosen": -319.9284973144531, "logps/ref_rejected": -308.20233154296875, "logps/rejected": -380.88330078125, "loss": 4.3739, "margin_dpo/margin_mean": 32.123836517333984, "margin_dpo/margin_std": 43.10367202758789, "step": 149 }, { "epoch": 0.31413612565445026, "fcm_dpo/beta": 0.012791416607797146, "fcm_dpo/delta": 0.025544868782162666, "fcm_dpo/margin": 24.68185806274414, "fcm_dpo/q_t": 0.4278425872325897, "grad_norm": 66.41766357421875, "learning_rate": 4.346796604970912e-07, "logits/chosen": -0.8340862989425659, "logits/rejected": -0.8192029595375061, "logps/chosen": -324.4963073730469, "logps/ref_chosen": -276.3182373046875, "logps/ref_rejected": -273.02215576171875, "logps/rejected": -345.8821105957031, "loss": 4.734, "margin_dpo/margin_mean": 24.68185806274414, "margin_dpo/margin_std": 46.04266357421875, "step": 150 }, { "epoch": 0.3162303664921466, "fcm_dpo/beta": 0.012728921137750149, "fcm_dpo/delta": -0.04246210679411888, "fcm_dpo/margin": 34.62612533569336, "fcm_dpo/q_t": 0.40107759833335876, "grad_norm": 93.15609741210938, "learning_rate": 4.3344075855595097e-07, "logits/chosen": -0.8574318289756775, "logits/rejected": -0.8527010679244995, "logps/chosen": -344.9148254394531, "logps/ref_chosen": -297.31280517578125, "logps/ref_rejected": -266.1003723144531, "logps/rejected": -348.3284912109375, "loss": 4.3664, "margin_dpo/margin_mean": 34.62612533569336, "margin_dpo/margin_std": 49.68989562988281, "step": 151 }, { "epoch": 0.3183246073298429, "fcm_dpo/beta": 0.012182039208710194, "fcm_dpo/delta": -0.0288827046751976, "fcm_dpo/margin": 30.36262321472168, "fcm_dpo/q_t": 0.4171737730503082, "grad_norm": 97.927978515625, "learning_rate": 4.3219201924364323e-07, "logits/chosen": -0.8693718314170837, "logits/rejected": -0.8675329685211182, "logps/chosen": -310.8759765625, "logps/ref_chosen": -270.2470397949219, "logps/ref_rejected": -269.7749328613281, "logps/rejected": -340.7665100097656, "loss": 4.5782, "margin_dpo/margin_mean": 30.36262321472168, "margin_dpo/margin_std": 49.9327278137207, "step": 152 }, { "epoch": 0.3204188481675393, "fcm_dpo/beta": 0.01159929670393467, "fcm_dpo/delta": -0.1059812605381012, "fcm_dpo/margin": 43.1450080871582, "fcm_dpo/q_t": 0.3860260844230652, "grad_norm": 74.16168212890625, "learning_rate": 4.309335095262675e-07, "logits/chosen": -0.8602730631828308, "logits/rejected": -0.8485139608383179, "logps/chosen": -318.30548095703125, "logps/ref_chosen": -273.779052734375, "logps/ref_rejected": -280.9530944824219, "logps/rejected": -368.62457275390625, "loss": 4.1014, "margin_dpo/margin_mean": 43.1450080871582, "margin_dpo/margin_std": 48.90004348754883, "step": 153 }, { "epoch": 0.3225130890052356, "fcm_dpo/beta": 0.010896118357777596, "fcm_dpo/delta": -0.05441652983427048, "fcm_dpo/margin": 35.19595718383789, "fcm_dpo/q_t": 0.41286537051200867, "grad_norm": 69.13275909423828, "learning_rate": 4.2966529689388064e-07, "logits/chosen": -0.8889592289924622, "logits/rejected": -0.8696526288986206, "logps/chosen": -331.47686767578125, "logps/ref_chosen": -289.9031982421875, "logps/ref_rejected": -261.5166320800781, "logps/rejected": -338.2862854003906, "loss": 4.5747, "margin_dpo/margin_mean": 35.19595718383789, "margin_dpo/margin_std": 58.20466232299805, "step": 154 }, { "epoch": 0.32460732984293195, "fcm_dpo/beta": 0.010797424241900444, "fcm_dpo/delta": 0.02106129191815853, "fcm_dpo/margin": 30.947917938232422, "fcm_dpo/q_t": 0.42601996660232544, "grad_norm": 76.80484008789062, "learning_rate": 4.2838744935687716e-07, "logits/chosen": -0.8353626132011414, "logits/rejected": -0.8337759375572205, "logps/chosen": -328.4182434082031, "logps/ref_chosen": -285.8612060546875, "logps/ref_rejected": -300.1272888183594, "logps/rejected": -373.63226318359375, "loss": 4.6686, "margin_dpo/margin_mean": 30.947919845581055, "margin_dpo/margin_std": 54.85850143432617, "step": 155 }, { "epoch": 0.3267015706806283, "fcm_dpo/beta": 0.010698702186346054, "fcm_dpo/delta": -0.09174332022666931, "fcm_dpo/margin": 43.385841369628906, "fcm_dpo/q_t": 0.3962228298187256, "grad_norm": 61.48815155029297, "learning_rate": 4.271000354423425e-07, "logits/chosen": -0.8628277778625488, "logits/rejected": -0.8613920211791992, "logps/chosen": -312.9440002441406, "logps/ref_chosen": -279.0354919433594, "logps/ref_rejected": -244.2198486328125, "logps/rejected": -321.5141906738281, "loss": 4.3529, "margin_dpo/margin_mean": 43.38584518432617, "margin_dpo/margin_std": 63.006675720214844, "step": 156 }, { "epoch": 0.3287958115183246, "fcm_dpo/beta": 0.010013958439230919, "fcm_dpo/delta": -0.013928013853728771, "fcm_dpo/margin": 33.83953094482422, "fcm_dpo/q_t": 0.422372430562973, "grad_norm": 63.8918571472168, "learning_rate": 4.258031241903777e-07, "logits/chosen": -0.9240726232528687, "logits/rejected": -0.9229752421379089, "logps/chosen": -305.9951171875, "logps/ref_chosen": -270.830322265625, "logps/ref_rejected": -259.08319091796875, "logps/rejected": -328.08758544921875, "loss": 4.618, "margin_dpo/margin_mean": 33.83953094482422, "margin_dpo/margin_std": 55.008338928222656, "step": 157 }, { "epoch": 0.3308900523560209, "fcm_dpo/beta": 0.010049000382423401, "fcm_dpo/delta": -0.018396489322185516, "fcm_dpo/margin": 36.628936767578125, "fcm_dpo/q_t": 0.4149770736694336, "grad_norm": 68.94039916992188, "learning_rate": 4.2449678515039743e-07, "logits/chosen": -0.8562602996826172, "logits/rejected": -0.8422697186470032, "logps/chosen": -323.416748046875, "logps/ref_chosen": -289.9663391113281, "logps/ref_rejected": -271.335693359375, "logps/rejected": -341.4150390625, "loss": 4.5504, "margin_dpo/margin_mean": 36.628936767578125, "margin_dpo/margin_std": 56.774017333984375, "step": 158 }, { "epoch": 0.33298429319371725, "fcm_dpo/beta": 0.01005611103028059, "fcm_dpo/delta": 0.04300171136856079, "fcm_dpo/margin": 23.479984283447266, "fcm_dpo/q_t": 0.4478485584259033, "grad_norm": 74.50262451171875, "learning_rate": 4.2318108837739986e-07, "logits/chosen": -0.9407361745834351, "logits/rejected": -0.8954955339431763, "logps/chosen": -359.2971496582031, "logps/ref_chosen": -321.37835693359375, "logps/ref_rejected": -250.45652770996094, "logps/rejected": -311.85528564453125, "loss": 5.0421, "margin_dpo/margin_mean": 23.4799861907959, "margin_dpo/margin_std": 59.843711853027344, "step": 159 }, { "epoch": 0.33507853403141363, "fcm_dpo/beta": 0.010049490258097649, "fcm_dpo/delta": -0.03839590400457382, "fcm_dpo/margin": 43.411376953125, "fcm_dpo/q_t": 0.39909377694129944, "grad_norm": 58.37568664550781, "learning_rate": 4.218561044282098e-07, "logits/chosen": -0.8778911828994751, "logits/rejected": -0.883208155632019, "logps/chosen": -304.81475830078125, "logps/ref_chosen": -276.28350830078125, "logps/ref_rejected": -262.7477722167969, "logps/rejected": -334.6904296875, "loss": 4.2303, "margin_dpo/margin_mean": 43.411376953125, "margin_dpo/margin_std": 48.92018508911133, "step": 160 }, { "epoch": 0.33717277486910996, "fcm_dpo/beta": 0.009891795925796032, "fcm_dpo/delta": -0.013076556846499443, "fcm_dpo/margin": 36.996002197265625, "fcm_dpo/q_t": 0.4181225299835205, "grad_norm": 63.23640441894531, "learning_rate": 4.2052190435769554e-07, "logits/chosen": -0.8851248621940613, "logits/rejected": -0.8717115521430969, "logps/chosen": -345.7448425292969, "logps/ref_chosen": -310.4927978515625, "logps/ref_rejected": -250.25347900390625, "logps/rejected": -322.50146484375, "loss": 4.5767, "margin_dpo/margin_mean": 36.996002197265625, "margin_dpo/margin_std": 60.53446960449219, "step": 161 }, { "epoch": 0.3392670157068063, "fcm_dpo/beta": 0.009680384770035744, "fcm_dpo/delta": 0.001748579554259777, "fcm_dpo/margin": 34.674312591552734, "fcm_dpo/q_t": 0.4228632152080536, "grad_norm": 58.60459518432617, "learning_rate": 4.1917855971495763e-07, "logits/chosen": -0.8581404089927673, "logits/rejected": -0.8511423468589783, "logps/chosen": -327.4073791503906, "logps/ref_chosen": -296.1105041503906, "logps/ref_rejected": -253.4247589111328, "logps/rejected": -319.3959655761719, "loss": 4.6121, "margin_dpo/margin_mean": 34.674312591552734, "margin_dpo/margin_std": 54.75847625732422, "step": 162 }, { "epoch": 0.3413612565445026, "fcm_dpo/beta": 0.010051360353827477, "fcm_dpo/delta": 0.02088327519595623, "fcm_dpo/margin": 37.680484771728516, "fcm_dpo/q_t": 0.41251957416534424, "grad_norm": 72.17606353759766, "learning_rate": 4.1782614253949255e-07, "logits/chosen": -0.8856139779090881, "logits/rejected": -0.8851478099822998, "logps/chosen": -326.783935546875, "logps/ref_chosen": -293.4999084472656, "logps/ref_rejected": -266.7116394042969, "logps/rejected": -337.67620849609375, "loss": 4.4711, "margin_dpo/margin_mean": 37.68048858642578, "margin_dpo/margin_std": 53.57781982421875, "step": 163 }, { "epoch": 0.34345549738219894, "fcm_dpo/beta": 0.010161195881664753, "fcm_dpo/delta": 0.015082177706062794, "fcm_dpo/margin": 34.78863525390625, "fcm_dpo/q_t": 0.42124950885772705, "grad_norm": 82.29058074951172, "learning_rate": 4.164647253573289e-07, "logits/chosen": -0.8554076552391052, "logits/rejected": -0.8738577365875244, "logps/chosen": -307.874755859375, "logps/ref_chosen": -267.04949951171875, "logps/ref_rejected": -215.9768829345703, "logps/rejected": -291.59075927734375, "loss": 4.6376, "margin_dpo/margin_mean": 34.788639068603516, "margin_dpo/margin_std": 61.51546859741211, "step": 164 }, { "epoch": 0.34554973821989526, "fcm_dpo/beta": 0.010130844078958035, "fcm_dpo/delta": 0.030348509550094604, "fcm_dpo/margin": 25.893747329711914, "fcm_dpo/q_t": 0.44016268849372864, "grad_norm": 80.56285858154297, "learning_rate": 4.1509438117713863e-07, "logits/chosen": -0.8746411204338074, "logits/rejected": -0.8493326902389526, "logps/chosen": -309.4532470703125, "logps/ref_chosen": -278.06146240234375, "logps/ref_rejected": -260.4288635253906, "logps/rejected": -317.71435546875, "loss": 4.862, "margin_dpo/margin_mean": 25.893747329711914, "margin_dpo/margin_std": 52.21025848388672, "step": 165 }, { "epoch": 0.34764397905759165, "fcm_dpo/beta": 0.010512001812458038, "fcm_dpo/delta": 0.04555278643965721, "fcm_dpo/margin": 31.762598037719727, "fcm_dpo/q_t": 0.42573946714401245, "grad_norm": 66.32711029052734, "learning_rate": 4.137151834863213e-07, "logits/chosen": -0.8395042419433594, "logits/rejected": -0.8069786429405212, "logps/chosen": -304.6323547363281, "logps/ref_chosen": -275.9490661621094, "logps/ref_rejected": -232.13473510742188, "logps/rejected": -292.58062744140625, "loss": 4.6734, "margin_dpo/margin_mean": 31.762596130371094, "margin_dpo/margin_std": 57.40185546875, "step": 166 }, { "epoch": 0.34973821989528797, "fcm_dpo/beta": 0.01087398361414671, "fcm_dpo/delta": 0.006117178127169609, "fcm_dpo/margin": 33.77849578857422, "fcm_dpo/q_t": 0.4176112115383148, "grad_norm": 79.66935729980469, "learning_rate": 4.123272062470633e-07, "logits/chosen": -0.8249646425247192, "logits/rejected": -0.8123272061347961, "logps/chosen": -314.0303955078125, "logps/ref_chosen": -280.5514221191406, "logps/ref_rejected": -255.2896728515625, "logps/rejected": -322.5471496582031, "loss": 4.6344, "margin_dpo/margin_mean": 33.77849578857422, "margin_dpo/margin_std": 60.363433837890625, "step": 167 }, { "epoch": 0.3518324607329843, "fcm_dpo/beta": 0.01047597173601389, "fcm_dpo/delta": -0.048771023750305176, "fcm_dpo/margin": 38.110443115234375, "fcm_dpo/q_t": 0.40823835134506226, "grad_norm": 148.15753173828125, "learning_rate": 4.1093052389237174e-07, "logits/chosen": -0.80986487865448, "logits/rejected": -0.7853665351867676, "logps/chosen": -350.20916748046875, "logps/ref_chosen": -315.7982177734375, "logps/ref_rejected": -291.48406982421875, "logps/rejected": -364.0054626464844, "loss": 4.5115, "margin_dpo/margin_mean": 38.110443115234375, "margin_dpo/margin_std": 59.79719161987305, "step": 168 }, { "epoch": 0.3539267015706806, "fcm_dpo/beta": 0.009861285798251629, "fcm_dpo/delta": -0.12296908348798752, "fcm_dpo/margin": 52.32954406738281, "fcm_dpo/q_t": 0.3826444447040558, "grad_norm": 73.20549774169922, "learning_rate": 4.0952521132208267e-07, "logits/chosen": -0.8107226490974426, "logits/rejected": -0.8243895173072815, "logps/chosen": -291.12841796875, "logps/ref_chosen": -261.06427001953125, "logps/ref_rejected": -235.40663146972656, "logps/rejected": -317.80035400390625, "loss": 4.0596, "margin_dpo/margin_mean": 52.32954406738281, "margin_dpo/margin_std": 56.9708251953125, "step": 169 }, { "epoch": 0.35602094240837695, "fcm_dpo/beta": 0.009434648789465427, "fcm_dpo/delta": 0.022008666768670082, "fcm_dpo/margin": 40.11225128173828, "fcm_dpo/q_t": 0.41667625308036804, "grad_norm": 67.7850570678711, "learning_rate": 4.081113438988443e-07, "logits/chosen": -0.7921425700187683, "logits/rejected": -0.7939149141311646, "logps/chosen": -343.00537109375, "logps/ref_chosen": -308.96722412109375, "logps/ref_rejected": -263.8466796875, "logps/rejected": -337.9970703125, "loss": 4.59, "margin_dpo/margin_mean": 40.11225509643555, "margin_dpo/margin_std": 70.82965850830078, "step": 170 }, { "epoch": 0.3581151832460733, "fcm_dpo/beta": 0.009155556559562683, "fcm_dpo/delta": -0.0953391045331955, "fcm_dpo/margin": 49.351959228515625, "fcm_dpo/q_t": 0.3953308165073395, "grad_norm": 91.96127319335938, "learning_rate": 4.0668899744407567e-07, "logits/chosen": -0.8171353936195374, "logits/rejected": -0.8306329250335693, "logps/chosen": -289.5184326171875, "logps/ref_chosen": -258.8890380859375, "logps/ref_rejected": -262.19140625, "logps/rejected": -342.1727294921875, "loss": 4.2024, "margin_dpo/margin_mean": 49.351959228515625, "margin_dpo/margin_std": 55.003509521484375, "step": 171 }, { "epoch": 0.36020942408376966, "fcm_dpo/beta": 0.009115180931985378, "fcm_dpo/delta": 0.04742582514882088, "fcm_dpo/margin": 25.906282424926758, "fcm_dpo/q_t": 0.44478750228881836, "grad_norm": 70.82637786865234, "learning_rate": 4.0525824823390043e-07, "logits/chosen": -0.8440933227539062, "logits/rejected": -0.8594982624053955, "logps/chosen": -378.7886047363281, "logps/ref_chosen": -339.0223388671875, "logps/ref_rejected": -295.78759765625, "logps/rejected": -361.46014404296875, "loss": 4.9617, "margin_dpo/margin_mean": 25.906280517578125, "margin_dpo/margin_std": 60.43056106567383, "step": 172 }, { "epoch": 0.362303664921466, "fcm_dpo/beta": 0.009056363254785538, "fcm_dpo/delta": -0.02586848847568035, "fcm_dpo/margin": 36.4200553894043, "fcm_dpo/q_t": 0.42598268389701843, "grad_norm": 67.26022338867188, "learning_rate": 4.0381917299505686e-07, "logits/chosen": -0.8703533411026001, "logits/rejected": -0.8670443296432495, "logps/chosen": -341.75244140625, "logps/ref_chosen": -300.1114501953125, "logps/ref_rejected": -273.78460693359375, "logps/rejected": -351.84564208984375, "loss": 4.6727, "margin_dpo/margin_mean": 36.4200553894043, "margin_dpo/margin_std": 62.181610107421875, "step": 173 }, { "epoch": 0.3643979057591623, "fcm_dpo/beta": 0.009070999920368195, "fcm_dpo/delta": 0.0008946415036916733, "fcm_dpo/margin": 43.95625686645508, "fcm_dpo/q_t": 0.4071962535381317, "grad_norm": 80.60259246826172, "learning_rate": 4.0237184890078243e-07, "logits/chosen": -0.8544137477874756, "logits/rejected": -0.8370189070701599, "logps/chosen": -381.8381042480469, "logps/ref_chosen": -335.0538635253906, "logps/ref_rejected": -257.4646911621094, "logps/rejected": -348.2052001953125, "loss": 4.3964, "margin_dpo/margin_mean": 43.95625686645508, "margin_dpo/margin_std": 59.383609771728516, "step": 174 }, { "epoch": 0.36649214659685864, "fcm_dpo/beta": 0.00893239676952362, "fcm_dpo/delta": -0.05215398222208023, "fcm_dpo/margin": 39.2378044128418, "fcm_dpo/q_t": 0.423565149307251, "grad_norm": 96.69216918945312, "learning_rate": 4.00916353566676e-07, "logits/chosen": -0.8526287078857422, "logits/rejected": -0.8526190519332886, "logps/chosen": -346.5810241699219, "logps/ref_chosen": -284.39556884765625, "logps/ref_rejected": -283.3876647949219, "logps/rejected": -384.81085205078125, "loss": 4.6754, "margin_dpo/margin_mean": 39.2378044128418, "margin_dpo/margin_std": 69.70426177978516, "step": 175 }, { "epoch": 0.36858638743455496, "fcm_dpo/beta": 0.00879521295428276, "fcm_dpo/delta": 0.0073929643258452415, "fcm_dpo/margin": 32.00230026245117, "fcm_dpo/q_t": 0.43624645471572876, "grad_norm": 77.65434265136719, "learning_rate": 3.994527650465352e-07, "logits/chosen": -0.8234915137290955, "logits/rejected": -0.8285030722618103, "logps/chosen": -316.092529296875, "logps/ref_chosen": -251.81280517578125, "logps/ref_rejected": -242.05328369140625, "logps/rejected": -338.3353271484375, "loss": 4.9145, "margin_dpo/margin_mean": 32.00230026245117, "margin_dpo/margin_std": 72.6375961303711, "step": 176 }, { "epoch": 0.3706806282722513, "fcm_dpo/beta": 0.008641266264021397, "fcm_dpo/delta": -0.004209747537970543, "fcm_dpo/margin": 32.934574127197266, "fcm_dpo/q_t": 0.43586111068725586, "grad_norm": 73.62203216552734, "learning_rate": 3.979811618281705e-07, "logits/chosen": -0.9073963761329651, "logits/rejected": -0.8772807717323303, "logps/chosen": -370.17852783203125, "logps/ref_chosen": -298.6463928222656, "logps/ref_rejected": -295.66534423828125, "logps/rejected": -400.1320495605469, "loss": 4.931, "margin_dpo/margin_mean": 32.93457794189453, "margin_dpo/margin_std": 73.69795227050781, "step": 177 }, { "epoch": 0.37277486910994767, "fcm_dpo/beta": 0.0086296908557415, "fcm_dpo/delta": -0.004849469289183617, "fcm_dpo/margin": 46.83763885498047, "fcm_dpo/q_t": 0.4095942974090576, "grad_norm": 68.19680786132812, "learning_rate": 3.9650162282919654e-07, "logits/chosen": -0.8406542539596558, "logits/rejected": -0.8333706855773926, "logps/chosen": -347.32861328125, "logps/ref_chosen": -286.2576599121094, "logps/ref_rejected": -243.97491455078125, "logps/rejected": -351.88348388671875, "loss": 4.4695, "margin_dpo/margin_mean": 46.83763885498047, "margin_dpo/margin_std": 71.97041320800781, "step": 178 }, { "epoch": 0.374869109947644, "fcm_dpo/beta": 0.008434956893324852, "fcm_dpo/delta": -0.034470878541469574, "fcm_dpo/margin": 42.3710823059082, "fcm_dpo/q_t": 0.4181768000125885, "grad_norm": 85.9342041015625, "learning_rate": 3.9501422739279953e-07, "logits/chosen": -0.8106435537338257, "logits/rejected": -0.8113298416137695, "logps/chosen": -324.47760009765625, "logps/ref_chosen": -259.737060546875, "logps/ref_rejected": -277.8813171386719, "logps/rejected": -384.9929504394531, "loss": 4.5927, "margin_dpo/margin_mean": 42.3710823059082, "margin_dpo/margin_std": 68.53620910644531, "step": 179 }, { "epoch": 0.3769633507853403, "fcm_dpo/beta": 0.008497622795403004, "fcm_dpo/delta": 0.006464208010584116, "fcm_dpo/margin": 43.05263137817383, "fcm_dpo/q_t": 0.41862213611602783, "grad_norm": 76.22695922851562, "learning_rate": 3.935190552834828e-07, "logits/chosen": -0.8361107110977173, "logits/rejected": -0.8606466054916382, "logps/chosen": -335.56817626953125, "logps/ref_chosen": -267.30889892578125, "logps/ref_rejected": -230.4376983642578, "logps/rejected": -341.74957275390625, "loss": 4.5463, "margin_dpo/margin_mean": 43.05263137817383, "margin_dpo/margin_std": 68.31497192382812, "step": 180 }, { "epoch": 0.37905759162303665, "fcm_dpo/beta": 0.0083799809217453, "fcm_dpo/delta": 0.008126728236675262, "fcm_dpo/margin": 40.54942321777344, "fcm_dpo/q_t": 0.42300912737846375, "grad_norm": 69.96990203857422, "learning_rate": 3.920161866827889e-07, "logits/chosen": -0.850619375705719, "logits/rejected": -0.8536400198936462, "logps/chosen": -375.7586975097656, "logps/ref_chosen": -300.49139404296875, "logps/ref_rejected": -278.98284912109375, "logps/rejected": -394.7995910644531, "loss": 4.6624, "margin_dpo/margin_mean": 40.54942321777344, "margin_dpo/margin_std": 72.2303466796875, "step": 181 }, { "epoch": 0.381151832460733, "fcm_dpo/beta": 0.00836803950369358, "fcm_dpo/delta": -0.06589318066835403, "fcm_dpo/margin": 52.413841247558594, "fcm_dpo/q_t": 0.40031716227531433, "grad_norm": 72.44232940673828, "learning_rate": 3.90505702185e-07, "logits/chosen": -0.8301101326942444, "logits/rejected": -0.8559277653694153, "logps/chosen": -355.3763427734375, "logps/ref_chosen": -279.4981689453125, "logps/ref_rejected": -263.6926574707031, "logps/rejected": -391.984619140625, "loss": 4.3269, "margin_dpo/margin_mean": 52.413841247558594, "margin_dpo/margin_std": 70.04794311523438, "step": 182 }, { "epoch": 0.3832460732984293, "fcm_dpo/beta": 0.00805988721549511, "fcm_dpo/delta": 0.019830908626317978, "fcm_dpo/margin": 47.256717681884766, "fcm_dpo/q_t": 0.41498610377311707, "grad_norm": 78.21089935302734, "learning_rate": 3.889876827928156e-07, "logits/chosen": -0.8640351295471191, "logits/rejected": -0.8657877445220947, "logps/chosen": -344.5203857421875, "logps/ref_chosen": -271.2057189941406, "logps/ref_rejected": -243.91549682617188, "logps/rejected": -364.4868469238281, "loss": 4.5346, "margin_dpo/margin_mean": 47.256717681884766, "margin_dpo/margin_std": 76.79619598388672, "step": 183 }, { "epoch": 0.38534031413612563, "fcm_dpo/beta": 0.007635125890374184, "fcm_dpo/delta": -0.10393635928630829, "fcm_dpo/margin": 64.89164733886719, "fcm_dpo/q_t": 0.3870086967945099, "grad_norm": 56.21622085571289, "learning_rate": 3.874622099130087e-07, "logits/chosen": -0.9125643372535706, "logits/rejected": -0.8902286291122437, "logps/chosen": -386.58197021484375, "logps/ref_chosen": -318.4457702636719, "logps/ref_rejected": -266.640869140625, "logps/rejected": -399.66864013671875, "loss": 4.1687, "margin_dpo/margin_mean": 64.89163970947266, "margin_dpo/margin_std": 79.78623962402344, "step": 184 }, { "epoch": 0.387434554973822, "fcm_dpo/beta": 0.0075467368587851524, "fcm_dpo/delta": 0.013490152545273304, "fcm_dpo/margin": 47.568668365478516, "fcm_dpo/q_t": 0.41884249448776245, "grad_norm": 59.321022033691406, "learning_rate": 3.859293653520604e-07, "logits/chosen": -0.9010488986968994, "logits/rejected": -0.8967990279197693, "logps/chosen": -351.34808349609375, "logps/ref_chosen": -274.308837890625, "logps/ref_rejected": -260.7274169921875, "logps/rejected": -385.3353271484375, "loss": 4.5723, "margin_dpo/margin_mean": 47.568668365478516, "margin_dpo/margin_std": 76.99652862548828, "step": 185 }, { "epoch": 0.38952879581151834, "fcm_dpo/beta": 0.0076223099604249, "fcm_dpo/delta": 0.00433752965182066, "fcm_dpo/margin": 46.567955017089844, "fcm_dpo/q_t": 0.418188214302063, "grad_norm": 67.55210876464844, "learning_rate": 3.8438923131177237e-07, "logits/chosen": -0.8974340558052063, "logits/rejected": -0.8985111713409424, "logps/chosen": -374.4447937011719, "logps/ref_chosen": -299.00537109375, "logps/ref_rejected": -274.4014587402344, "logps/rejected": -396.4088134765625, "loss": 4.5636, "margin_dpo/margin_mean": 46.567955017089844, "margin_dpo/margin_std": 71.06839752197266, "step": 186 }, { "epoch": 0.39162303664921466, "fcm_dpo/beta": 0.007788485381752253, "fcm_dpo/delta": 0.023075303062796593, "fcm_dpo/margin": 42.33740997314453, "fcm_dpo/q_t": 0.4230867028236389, "grad_norm": 82.19164276123047, "learning_rate": 3.828418903848593e-07, "logits/chosen": -0.8443529605865479, "logits/rejected": -0.8297170400619507, "logps/chosen": -410.2223205566406, "logps/ref_chosen": -329.8253173828125, "logps/ref_rejected": -263.73175048828125, "logps/rejected": -386.4662170410156, "loss": 4.7667, "margin_dpo/margin_mean": 42.33740997314453, "margin_dpo/margin_std": 83.36969757080078, "step": 187 }, { "epoch": 0.393717277486911, "fcm_dpo/beta": 0.007786841131746769, "fcm_dpo/delta": 0.005178054794669151, "fcm_dpo/margin": 48.28009033203125, "fcm_dpo/q_t": 0.41652175784111023, "grad_norm": 68.4957504272461, "learning_rate": 3.812874255505191e-07, "logits/chosen": -0.8509972095489502, "logits/rejected": -0.8415097594261169, "logps/chosen": -337.3232727050781, "logps/ref_chosen": -263.005615234375, "logps/ref_rejected": -247.08668518066406, "logps/rejected": -369.68438720703125, "loss": 4.6187, "margin_dpo/margin_mean": 48.280086517333984, "margin_dpo/margin_std": 84.95806121826172, "step": 188 }, { "epoch": 0.3958115183246073, "fcm_dpo/beta": 0.0074915410950779915, "fcm_dpo/delta": -0.07813064754009247, "fcm_dpo/margin": 55.84880065917969, "fcm_dpo/q_t": 0.40445610880851746, "grad_norm": 74.38936614990234, "learning_rate": 3.797259201699833e-07, "logits/chosen": -0.8831048011779785, "logits/rejected": -0.8907921314239502, "logps/chosen": -337.052490234375, "logps/ref_chosen": -272.96038818359375, "logps/ref_rejected": -275.13238525390625, "logps/rejected": -395.07330322265625, "loss": 4.3464, "margin_dpo/margin_mean": 55.84880065917969, "margin_dpo/margin_std": 72.70423126220703, "step": 189 }, { "epoch": 0.39790575916230364, "fcm_dpo/beta": 0.007398220710456371, "fcm_dpo/delta": 0.0288606658577919, "fcm_dpo/margin": 50.26428985595703, "fcm_dpo/q_t": 0.4147360920906067, "grad_norm": 70.62735748291016, "learning_rate": 3.781574579820464e-07, "logits/chosen": -0.8793290853500366, "logits/rejected": -0.8391152024269104, "logps/chosen": -323.90863037109375, "logps/ref_chosen": -257.79754638671875, "logps/ref_rejected": -225.2164306640625, "logps/rejected": -341.591796875, "loss": 4.4976, "margin_dpo/margin_mean": 50.26428985595703, "margin_dpo/margin_std": 73.94573974609375, "step": 190 }, { "epoch": 0.4, "fcm_dpo/beta": 0.007520074490457773, "fcm_dpo/delta": -0.0024281367659568787, "fcm_dpo/margin": 50.59221649169922, "fcm_dpo/q_t": 0.41343408823013306, "grad_norm": 70.87568664550781, "learning_rate": 3.765821230985757e-07, "logits/chosen": -0.9008585214614868, "logits/rejected": -0.9023429155349731, "logps/chosen": -303.3919982910156, "logps/ref_chosen": -243.8585205078125, "logps/ref_rejected": -245.12136840820312, "logps/rejected": -355.2470703125, "loss": 4.5217, "margin_dpo/margin_mean": 50.59221267700195, "margin_dpo/margin_std": 78.84461212158203, "step": 191 }, { "epoch": 0.40209424083769635, "fcm_dpo/beta": 0.007562983315438032, "fcm_dpo/delta": 0.018396003171801567, "fcm_dpo/margin": 40.44701385498047, "fcm_dpo/q_t": 0.4299015998840332, "grad_norm": 53.1352653503418, "learning_rate": 3.75e-07, "logits/chosen": -0.8452181816101074, "logits/rejected": -0.8295493125915527, "logps/chosen": -338.7875671386719, "logps/ref_chosen": -266.9799499511719, "logps/ref_rejected": -260.1697082519531, "logps/rejected": -372.42437744140625, "loss": 4.7355, "margin_dpo/margin_mean": 40.447021484375, "margin_dpo/margin_std": 72.97218322753906, "step": 192 }, { "epoch": 0.4041884816753927, "fcm_dpo/beta": 0.007469853386282921, "fcm_dpo/delta": -0.0201749037951231, "fcm_dpo/margin": 49.940818786621094, "fcm_dpo/q_t": 0.4158126711845398, "grad_norm": 71.28538513183594, "learning_rate": 3.734111735307796e-07, "logits/chosen": -0.8989495038986206, "logits/rejected": -0.8711603879928589, "logps/chosen": -363.9617919921875, "logps/ref_chosen": -280.25323486328125, "logps/ref_rejected": -291.0348815917969, "logps/rejected": -424.6842956542969, "loss": 4.5614, "margin_dpo/margin_mean": 49.940818786621094, "margin_dpo/margin_std": 80.18729400634766, "step": 193 }, { "epoch": 0.406282722513089, "fcm_dpo/beta": 0.007623044308274984, "fcm_dpo/delta": 0.0396825447678566, "fcm_dpo/margin": 38.62468719482422, "fcm_dpo/q_t": 0.4337855279445648, "grad_norm": 79.49999237060547, "learning_rate": 3.7181572889485623e-07, "logits/chosen": -0.8728510737419128, "logits/rejected": -0.8639151453971863, "logps/chosen": -375.9040222167969, "logps/ref_chosen": -288.13946533203125, "logps/ref_rejected": -251.31529235839844, "logps/rejected": -377.70452880859375, "loss": 4.7918, "margin_dpo/margin_mean": 38.62468719482422, "margin_dpo/margin_std": 75.4052505493164, "step": 194 }, { "epoch": 0.4083769633507853, "fcm_dpo/beta": 0.007884586229920387, "fcm_dpo/delta": 0.040129128843545914, "fcm_dpo/margin": 37.130008697509766, "fcm_dpo/q_t": 0.43522733449935913, "grad_norm": 89.50051879882812, "learning_rate": 3.7021375165108377e-07, "logits/chosen": -0.859967827796936, "logits/rejected": -0.8635672926902771, "logps/chosen": -366.3628234863281, "logps/ref_chosen": -274.0006408691406, "logps/ref_rejected": -280.22723388671875, "logps/rejected": -409.71942138671875, "loss": 4.8251, "margin_dpo/margin_mean": 37.130008697509766, "margin_dpo/margin_std": 76.77651977539062, "step": 195 }, { "epoch": 0.41047120418848165, "fcm_dpo/beta": 0.007902967743575573, "fcm_dpo/delta": -0.017355965450406075, "fcm_dpo/margin": 52.63897705078125, "fcm_dpo/q_t": 0.40730372071266174, "grad_norm": 91.81265258789062, "learning_rate": 3.6860532770864005e-07, "logits/chosen": -0.8587782979011536, "logits/rejected": -0.8644378185272217, "logps/chosen": -354.6200256347656, "logps/ref_chosen": -274.90069580078125, "logps/ref_rejected": -248.7281951904297, "logps/rejected": -381.0865478515625, "loss": 4.4737, "margin_dpo/margin_mean": 52.63897705078125, "margin_dpo/margin_std": 84.32215881347656, "step": 196 }, { "epoch": 0.41256544502617803, "fcm_dpo/beta": 0.00753373745828867, "fcm_dpo/delta": -0.10619766265153885, "fcm_dpo/margin": 58.62234115600586, "fcm_dpo/q_t": 0.40042629837989807, "grad_norm": 108.23560333251953, "learning_rate": 3.6699054332241985e-07, "logits/chosen": -0.8893225193023682, "logits/rejected": -0.874314546585083, "logps/chosen": -397.1566162109375, "logps/ref_chosen": -309.5348205566406, "logps/ref_rejected": -264.3179931640625, "logps/rejected": -410.5621032714844, "loss": 4.3535, "margin_dpo/margin_mean": 58.62234115600586, "margin_dpo/margin_std": 79.52429962158203, "step": 197 }, { "epoch": 0.41465968586387436, "fcm_dpo/beta": 0.00704828929156065, "fcm_dpo/delta": -0.030336569994688034, "fcm_dpo/margin": 60.71779251098633, "fcm_dpo/q_t": 0.4027983844280243, "grad_norm": 77.68168640136719, "learning_rate": 3.653694850884091e-07, "logits/chosen": -0.8886722922325134, "logits/rejected": -0.8632485866546631, "logps/chosen": -382.4344482421875, "logps/ref_chosen": -301.0134582519531, "logps/ref_rejected": -292.84185791015625, "logps/rejected": -434.98065185546875, "loss": 4.3485, "margin_dpo/margin_mean": 60.71779251098633, "margin_dpo/margin_std": 82.84037780761719, "step": 198 }, { "epoch": 0.4167539267015707, "fcm_dpo/beta": 0.007051732391119003, "fcm_dpo/delta": -0.012272719293832779, "fcm_dpo/margin": 53.73326110839844, "fcm_dpo/q_t": 0.4135727882385254, "grad_norm": 73.13438415527344, "learning_rate": 3.6374223993904124e-07, "logits/chosen": -0.8650781512260437, "logits/rejected": -0.8219231367111206, "logps/chosen": -353.032470703125, "logps/ref_chosen": -264.6058654785156, "logps/ref_rejected": -214.9014892578125, "logps/rejected": -357.0614013671875, "loss": 4.4976, "margin_dpo/margin_mean": 53.73326110839844, "margin_dpo/margin_std": 81.83422088623047, "step": 199 }, { "epoch": 0.418848167539267, "fcm_dpo/beta": 0.007016919087618589, "fcm_dpo/delta": 0.007905922830104828, "fcm_dpo/margin": 44.08357238769531, "fcm_dpo/q_t": 0.43098166584968567, "grad_norm": 82.60724639892578, "learning_rate": 3.621088951385353e-07, "logits/chosen": -0.9180527925491333, "logits/rejected": -0.8939597010612488, "logps/chosen": -419.4546813964844, "logps/ref_chosen": -324.1588134765625, "logps/ref_rejected": -277.80218505859375, "logps/rejected": -417.181640625, "loss": 4.8634, "margin_dpo/margin_mean": 44.08357238769531, "margin_dpo/margin_std": 95.21434020996094, "step": 200 }, { "epoch": 0.418848167539267, "eval_fcm_dpo/beta": 0.007041628006845713, "eval_logits/chosen": -0.8935934901237488, "eval_logits/rejected": -0.8767105340957642, "eval_logps/chosen": -385.81689453125, "eval_logps/ref_chosen": -287.8268127441406, "eval_logps/ref_rejected": -266.9300231933594, "eval_logps/rejected": -415.6835021972656, "eval_loss": 0.5780735015869141, "eval_margin_dpo/margin_mean": 50.763328552246094, "eval_margin_dpo/margin_std": 82.26912689208984, "eval_runtime": 79.4075, "eval_samples_per_second": 25.187, "eval_steps_per_second": 3.148, "step": 200 }, { "epoch": 0.42094240837696334, "fcm_dpo/beta": 0.007051797583699226, "fcm_dpo/delta": -0.013565017841756344, "fcm_dpo/margin": 52.115760803222656, "fcm_dpo/q_t": 0.41692644357681274, "grad_norm": 118.92222595214844, "learning_rate": 3.604695382782159e-07, "logits/chosen": -0.8829444050788879, "logits/rejected": -0.8692610263824463, "logps/chosen": -370.8045349121094, "logps/ref_chosen": -271.49566650390625, "logps/ref_rejected": -245.71414184570312, "logps/rejected": -397.1387939453125, "loss": 4.6069, "margin_dpo/margin_mean": 52.115760803222656, "margin_dpo/margin_std": 88.78590393066406, "step": 201 }, { "epoch": 0.42303664921465967, "fcm_dpo/beta": 0.006848223973065615, "fcm_dpo/delta": -0.02687113918364048, "fcm_dpo/margin": 52.322425842285156, "fcm_dpo/q_t": 0.4185172915458679, "grad_norm": 76.39278411865234, "learning_rate": 3.588242572718162e-07, "logits/chosen": -0.8918448090553284, "logits/rejected": -0.8787520527839661, "logps/chosen": -366.652587890625, "logps/ref_chosen": -272.0979309082031, "logps/ref_rejected": -235.94805908203125, "logps/rejected": -382.82513427734375, "loss": 4.6214, "margin_dpo/margin_mean": 52.322425842285156, "margin_dpo/margin_std": 89.93933868408203, "step": 202 }, { "epoch": 0.42513089005235605, "fcm_dpo/beta": 0.0069344537332654, "fcm_dpo/delta": 0.04031352326273918, "fcm_dpo/margin": 40.129302978515625, "fcm_dpo/q_t": 0.43582794070243835, "grad_norm": 95.53482055664062, "learning_rate": 3.571731403507635e-07, "logits/chosen": -0.8573673963546753, "logits/rejected": -0.8636147975921631, "logps/chosen": -379.1627197265625, "logps/ref_chosen": -280.2221374511719, "logps/ref_rejected": -251.79798889160156, "logps/rejected": -390.86785888671875, "loss": 4.8197, "margin_dpo/margin_mean": 40.129302978515625, "margin_dpo/margin_std": 78.97547149658203, "step": 203 }, { "epoch": 0.4272251308900524, "fcm_dpo/beta": 0.006983473431318998, "fcm_dpo/delta": 0.0020564058795571327, "fcm_dpo/margin": 56.94648361206055, "fcm_dpo/q_t": 0.4088953733444214, "grad_norm": 83.27347564697266, "learning_rate": 3.5551627605944746e-07, "logits/chosen": -0.9110021591186523, "logits/rejected": -0.8834875226020813, "logps/chosen": -407.1368408203125, "logps/ref_chosen": -318.7960510253906, "logps/ref_rejected": -269.69921875, "logps/rejected": -414.98651123046875, "loss": 4.4147, "margin_dpo/margin_mean": 56.94648361206055, "margin_dpo/margin_std": 79.84234619140625, "step": 204 }, { "epoch": 0.4293193717277487, "fcm_dpo/beta": 0.006971952971071005, "fcm_dpo/delta": -0.04280940815806389, "fcm_dpo/margin": 63.21536636352539, "fcm_dpo/q_t": 0.402166485786438, "grad_norm": 75.73958587646484, "learning_rate": 3.5385375325047163e-07, "logits/chosen": -0.8463097810745239, "logits/rejected": -0.8146329522132874, "logps/chosen": -370.4803161621094, "logps/ref_chosen": -283.7620544433594, "logps/ref_rejected": -297.69439697265625, "logps/rejected": -447.6280212402344, "loss": 4.3512, "margin_dpo/margin_mean": 63.21536636352539, "margin_dpo/margin_std": 89.52297973632812, "step": 205 }, { "epoch": 0.431413612565445, "fcm_dpo/beta": 0.006694549694657326, "fcm_dpo/delta": -0.02712525613605976, "fcm_dpo/margin": 49.10927963256836, "fcm_dpo/q_t": 0.42545533180236816, "grad_norm": 70.90359497070312, "learning_rate": 3.5218566107988867e-07, "logits/chosen": -0.8882994055747986, "logits/rejected": -0.909928560256958, "logps/chosen": -385.359619140625, "logps/ref_chosen": -293.66387939453125, "logps/ref_rejected": -291.3056640625, "logps/rejected": -432.1107177734375, "loss": 4.7352, "margin_dpo/margin_mean": 49.10927963256836, "margin_dpo/margin_std": 92.05156707763672, "step": 206 }, { "epoch": 0.43350785340314135, "fcm_dpo/beta": 0.006603881251066923, "fcm_dpo/delta": -0.01149575226008892, "fcm_dpo/margin": 49.813499450683594, "fcm_dpo/q_t": 0.4254821538925171, "grad_norm": 85.93264770507812, "learning_rate": 3.505120890024195e-07, "logits/chosen": -0.8319991827011108, "logits/rejected": -0.8400675058364868, "logps/chosen": -354.9524230957031, "logps/ref_chosen": -270.5350646972656, "logps/ref_rejected": -278.7747497558594, "logps/rejected": -413.005615234375, "loss": 4.7927, "margin_dpo/margin_mean": 49.81349563598633, "margin_dpo/margin_std": 102.63397216796875, "step": 207 }, { "epoch": 0.4356020942408377, "fcm_dpo/beta": 0.006531773135066032, "fcm_dpo/delta": -0.045300256460905075, "fcm_dpo/margin": 64.10903930664062, "fcm_dpo/q_t": 0.4059593081474304, "grad_norm": 69.88282012939453, "learning_rate": 3.4883312676665534e-07, "logits/chosen": -0.8945930004119873, "logits/rejected": -0.8459950685501099, "logps/chosen": -368.4111633300781, "logps/ref_chosen": -279.582763671875, "logps/ref_rejected": -290.041015625, "logps/rejected": -442.978515625, "loss": 4.368, "margin_dpo/margin_mean": 64.10904693603516, "margin_dpo/margin_std": 88.38412475585938, "step": 208 }, { "epoch": 0.437696335078534, "fcm_dpo/beta": 0.006543098483234644, "fcm_dpo/delta": 0.0669432207942009, "fcm_dpo/margin": 40.375823974609375, "fcm_dpo/q_t": 0.4394206702709198, "grad_norm": 94.81318664550781, "learning_rate": 3.4714886441024573e-07, "logits/chosen": -0.8037338852882385, "logits/rejected": -0.8001971244812012, "logps/chosen": -417.86273193359375, "logps/ref_chosen": -318.8725280761719, "logps/ref_rejected": -270.64324951171875, "logps/rejected": -410.0093078613281, "loss": 4.9026, "margin_dpo/margin_mean": 40.37582778930664, "margin_dpo/margin_std": 88.71975708007812, "step": 209 }, { "epoch": 0.4397905759162304, "fcm_dpo/beta": 0.006677450146526098, "fcm_dpo/delta": 0.031221631914377213, "fcm_dpo/margin": 55.327545166015625, "fcm_dpo/q_t": 0.4165865480899811, "grad_norm": 75.42611694335938, "learning_rate": 3.454593922550693e-07, "logits/chosen": -0.8339765667915344, "logits/rejected": -0.8234052062034607, "logps/chosen": -369.9593505859375, "logps/ref_chosen": -283.14031982421875, "logps/ref_rejected": -287.2986755371094, "logps/rejected": -429.44525146484375, "loss": 4.5568, "margin_dpo/margin_mean": 55.327545166015625, "margin_dpo/margin_std": 88.78600311279297, "step": 210 }, { "epoch": 0.4418848167539267, "fcm_dpo/beta": 0.006575165782123804, "fcm_dpo/delta": -0.08240824937820435, "fcm_dpo/margin": 66.77010345458984, "fcm_dpo/q_t": 0.39799728989601135, "grad_norm": 68.85022735595703, "learning_rate": 3.4376480090239047e-07, "logits/chosen": -0.8611884117126465, "logits/rejected": -0.8390584588050842, "logps/chosen": -354.44036865234375, "logps/ref_chosen": -276.4228515625, "logps/ref_rejected": -252.40603637695312, "logps/rejected": -397.193603515625, "loss": 4.2563, "margin_dpo/margin_mean": 66.77011108398438, "margin_dpo/margin_std": 76.85353088378906, "step": 211 }, { "epoch": 0.44397905759162304, "fcm_dpo/beta": 0.006527946330606937, "fcm_dpo/delta": 0.03064761869609356, "fcm_dpo/margin": 41.976829528808594, "fcm_dpo/q_t": 0.43568912148475647, "grad_norm": 106.75749969482422, "learning_rate": 3.4206518122800055e-07, "logits/chosen": -0.8498879075050354, "logits/rejected": -0.8480284214019775, "logps/chosen": -362.7585144042969, "logps/ref_chosen": -271.7055358886719, "logps/ref_rejected": -241.18511962890625, "logps/rejected": -374.2149658203125, "loss": 4.8563, "margin_dpo/margin_mean": 41.976829528808594, "margin_dpo/margin_std": 84.9686279296875, "step": 212 }, { "epoch": 0.44607329842931936, "fcm_dpo/beta": 0.006641289219260216, "fcm_dpo/delta": 0.0012545166537165642, "fcm_dpo/margin": 50.06464385986328, "fcm_dpo/q_t": 0.425908625125885, "grad_norm": 103.77131652832031, "learning_rate": 3.403606243773448e-07, "logits/chosen": -0.8439345955848694, "logits/rejected": -0.8587431311607361, "logps/chosen": -396.8846435546875, "logps/ref_chosen": -302.2976379394531, "logps/ref_rejected": -303.6202087402344, "logps/rejected": -448.2718200683594, "loss": 4.7044, "margin_dpo/margin_mean": 50.06464385986328, "margin_dpo/margin_std": 93.97140502929688, "step": 213 }, { "epoch": 0.4481675392670157, "fcm_dpo/beta": 0.006577081512659788, "fcm_dpo/delta": -0.015572590753436089, "fcm_dpo/margin": 62.8525390625, "fcm_dpo/q_t": 0.405367910861969, "grad_norm": 92.35052490234375, "learning_rate": 3.3865122176063385e-07, "logits/chosen": -0.8551336526870728, "logits/rejected": -0.8550840616226196, "logps/chosen": -380.4796447753906, "logps/ref_chosen": -272.13262939453125, "logps/ref_rejected": -294.82354736328125, "logps/rejected": -466.0230712890625, "loss": 4.3617, "margin_dpo/margin_mean": 62.8525390625, "margin_dpo/margin_std": 80.1742935180664, "step": 214 }, { "epoch": 0.450261780104712, "fcm_dpo/beta": 0.006535133346915245, "fcm_dpo/delta": -0.00995543785393238, "fcm_dpo/margin": 42.939754486083984, "fcm_dpo/q_t": 0.43834444880485535, "grad_norm": 89.51997375488281, "learning_rate": 3.3693706504794243e-07, "logits/chosen": -0.8979097604751587, "logits/rejected": -0.8771700263023376, "logps/chosen": -402.0087585449219, "logps/ref_chosen": -291.3782958984375, "logps/ref_rejected": -261.05792236328125, "logps/rejected": -414.62811279296875, "loss": 4.8962, "margin_dpo/margin_mean": 42.939754486083984, "margin_dpo/margin_std": 94.85189819335938, "step": 215 }, { "epoch": 0.4523560209424084, "fcm_dpo/beta": 0.006419910583645105, "fcm_dpo/delta": 0.00330403633415699, "fcm_dpo/margin": 56.20130920410156, "fcm_dpo/q_t": 0.418740838766098, "grad_norm": 73.42216491699219, "learning_rate": 3.3521824616429284e-07, "logits/chosen": -0.922217607498169, "logits/rejected": -0.9141356945037842, "logps/chosen": -444.15765380859375, "logps/ref_chosen": -338.50543212890625, "logps/ref_rejected": -305.76104736328125, "logps/rejected": -467.61456298828125, "loss": 4.6374, "margin_dpo/margin_mean": 56.20130920410156, "margin_dpo/margin_std": 99.16251373291016, "step": 216 }, { "epoch": 0.4544502617801047, "fcm_dpo/beta": 0.006329713389277458, "fcm_dpo/delta": -0.04338594153523445, "fcm_dpo/margin": 69.61051940917969, "fcm_dpo/q_t": 0.4017544090747833, "grad_norm": 74.04085540771484, "learning_rate": 3.334948572847253e-07, "logits/chosen": -0.8338743448257446, "logits/rejected": -0.7945307493209839, "logps/chosen": -406.1719665527344, "logps/ref_chosen": -293.5498046875, "logps/ref_rejected": -256.7830810546875, "logps/rejected": -439.01580810546875, "loss": 4.3716, "margin_dpo/margin_mean": 69.61051940917969, "margin_dpo/margin_std": 101.19683074951172, "step": 217 }, { "epoch": 0.45654450261780105, "fcm_dpo/beta": 0.006193040404468775, "fcm_dpo/delta": -0.006936352234333754, "fcm_dpo/margin": 65.53926086425781, "fcm_dpo/q_t": 0.40807247161865234, "grad_norm": 74.45264434814453, "learning_rate": 3.317669908293554e-07, "logits/chosen": -0.8653545379638672, "logits/rejected": -0.8810456991195679, "logps/chosen": -432.00384521484375, "logps/ref_chosen": -320.579345703125, "logps/ref_rejected": -294.0381164550781, "logps/rejected": -471.00189208984375, "loss": 4.421, "margin_dpo/margin_mean": 65.53926086425781, "margin_dpo/margin_std": 95.39942932128906, "step": 218 }, { "epoch": 0.4586387434554974, "fcm_dpo/beta": 0.006144728511571884, "fcm_dpo/delta": -0.03086962178349495, "fcm_dpo/margin": 63.95124816894531, "fcm_dpo/q_t": 0.41116613149642944, "grad_norm": 76.28233337402344, "learning_rate": 3.300347394584172e-07, "logits/chosen": -0.8417205214500427, "logits/rejected": -0.8665354251861572, "logps/chosen": -379.6574401855469, "logps/ref_chosen": -268.4186096191406, "logps/ref_rejected": -265.7808837890625, "logps/rejected": -440.970947265625, "loss": 4.5161, "margin_dpo/margin_mean": 63.95124816894531, "margin_dpo/margin_std": 100.1285629272461, "step": 219 }, { "epoch": 0.4607329842931937, "fcm_dpo/beta": 0.006020966451615095, "fcm_dpo/delta": -0.006018012762069702, "fcm_dpo/margin": 67.32181549072266, "fcm_dpo/q_t": 0.4066336750984192, "grad_norm": 67.24308776855469, "learning_rate": 3.2829819606729477e-07, "logits/chosen": -0.9152698516845703, "logits/rejected": -0.8838323354721069, "logps/chosen": -423.98956298828125, "logps/ref_chosen": -312.8864440917969, "logps/ref_rejected": -259.5191955566406, "logps/rejected": -437.9441223144531, "loss": 4.4003, "margin_dpo/margin_mean": 67.32180786132812, "margin_dpo/margin_std": 93.29956817626953, "step": 220 }, { "epoch": 0.46282722513089003, "fcm_dpo/beta": 0.006161336787045002, "fcm_dpo/delta": -0.010663069784641266, "fcm_dpo/margin": 55.39729309082031, "fcm_dpo/q_t": 0.4242016077041626, "grad_norm": 75.08380889892578, "learning_rate": 3.265574537815398e-07, "logits/chosen": -0.8223947286605835, "logits/rejected": -0.8319114446640015, "logps/chosen": -420.740966796875, "logps/ref_chosen": -300.32586669921875, "logps/ref_rejected": -286.312255859375, "logps/rejected": -462.12469482421875, "loss": 4.7053, "margin_dpo/margin_mean": 55.39729309082031, "margin_dpo/margin_std": 102.78092193603516, "step": 221 }, { "epoch": 0.4649214659685864, "fcm_dpo/beta": 0.006012233905494213, "fcm_dpo/delta": 0.020973335951566696, "fcm_dpo/margin": 58.58289337158203, "fcm_dpo/q_t": 0.4185751974582672, "grad_norm": 64.78793334960938, "learning_rate": 3.248126059518784e-07, "logits/chosen": -0.9194785952568054, "logits/rejected": -0.8871119022369385, "logps/chosen": -414.5599060058594, "logps/ref_chosen": -297.1113586425781, "logps/ref_rejected": -235.53146362304688, "logps/rejected": -411.56292724609375, "loss": 4.5659, "margin_dpo/margin_mean": 58.5828971862793, "margin_dpo/margin_std": 90.73377990722656, "step": 222 }, { "epoch": 0.46701570680628274, "fcm_dpo/beta": 0.0061044651083648205, "fcm_dpo/delta": 0.009785637259483337, "fcm_dpo/margin": 60.01387023925781, "fcm_dpo/q_t": 0.4157088100910187, "grad_norm": 87.03768157958984, "learning_rate": 3.230637461492043e-07, "logits/chosen": -0.8523328304290771, "logits/rejected": -0.816593587398529, "logps/chosen": -409.787841796875, "logps/ref_chosen": -286.41510009765625, "logps/ref_rejected": -241.1181640625, "logps/rejected": -424.5048522949219, "loss": 4.5608, "margin_dpo/margin_mean": 60.01387023925781, "margin_dpo/margin_std": 96.16004180908203, "step": 223 }, { "epoch": 0.46910994764397906, "fcm_dpo/beta": 0.006154531612992287, "fcm_dpo/delta": -0.03359140083193779, "fcm_dpo/margin": 64.03076171875, "fcm_dpo/q_t": 0.41083890199661255, "grad_norm": 90.42691040039062, "learning_rate": 3.213109681595612e-07, "logits/chosen": -0.8257457613945007, "logits/rejected": -0.8410820960998535, "logps/chosen": -365.0633544921875, "logps/ref_chosen": -249.49234008789062, "logps/ref_rejected": -233.10752868652344, "logps/rejected": -412.70928955078125, "loss": 4.4432, "margin_dpo/margin_mean": 64.03076171875, "margin_dpo/margin_std": 90.88447570800781, "step": 224 }, { "epoch": 0.4712041884816754, "fcm_dpo/beta": 0.005969415418803692, "fcm_dpo/delta": 0.012547025457024574, "fcm_dpo/margin": 59.763431549072266, "fcm_dpo/q_t": 0.4206285774707794, "grad_norm": 93.77128601074219, "learning_rate": 3.1955436597911315e-07, "logits/chosen": -0.8661923408508301, "logits/rejected": -0.839612603187561, "logps/chosen": -444.10955810546875, "logps/ref_chosen": -311.8583679199219, "logps/ref_rejected": -336.8523864746094, "logps/rejected": -528.8670043945312, "loss": 4.6451, "margin_dpo/margin_mean": 59.763431549072266, "margin_dpo/margin_std": 106.94706726074219, "step": 225 }, { "epoch": 0.4732984293193717, "fcm_dpo/beta": 0.006135667208582163, "fcm_dpo/delta": 0.021072357892990112, "fcm_dpo/margin": 50.27009582519531, "fcm_dpo/q_t": 0.4313587546348572, "grad_norm": 90.91165924072266, "learning_rate": 3.1779403380910425e-07, "logits/chosen": -0.8851912021636963, "logits/rejected": -0.8732432723045349, "logps/chosen": -371.9580993652344, "logps/ref_chosen": -252.20123291015625, "logps/ref_rejected": -254.41162109375, "logps/rejected": -424.4385681152344, "loss": 4.7914, "margin_dpo/margin_mean": 50.27009201049805, "margin_dpo/margin_std": 102.74089050292969, "step": 226 }, { "epoch": 0.47539267015706804, "fcm_dpo/beta": 0.006114313844591379, "fcm_dpo/delta": -0.011526936665177345, "fcm_dpo/margin": 67.09516143798828, "fcm_dpo/q_t": 0.4065951108932495, "grad_norm": 83.9355239868164, "learning_rate": 3.160300660508064e-07, "logits/chosen": -0.8550764918327332, "logits/rejected": -0.8507223129272461, "logps/chosen": -400.2503967285156, "logps/ref_chosen": -285.25946044921875, "logps/ref_rejected": -261.3220520019531, "logps/rejected": -443.40814208984375, "loss": 4.4568, "margin_dpo/margin_mean": 67.09516143798828, "margin_dpo/margin_std": 102.07121276855469, "step": 227 }, { "epoch": 0.4774869109947644, "fcm_dpo/beta": 0.006071330048143864, "fcm_dpo/delta": -0.03209439292550087, "fcm_dpo/margin": 67.11500549316406, "fcm_dpo/q_t": 0.407455712556839, "grad_norm": 74.55585479736328, "learning_rate": 3.1426255730045695e-07, "logits/chosen": -0.8666589856147766, "logits/rejected": -0.8316813707351685, "logps/chosen": -417.8719482421875, "logps/ref_chosen": -313.81878662109375, "logps/ref_rejected": -258.07061767578125, "logps/rejected": -429.23876953125, "loss": 4.41, "margin_dpo/margin_mean": 67.11499786376953, "margin_dpo/margin_std": 92.93700408935547, "step": 228 }, { "epoch": 0.47958115183246075, "fcm_dpo/beta": 0.005753038451075554, "fcm_dpo/delta": -0.07461543381214142, "fcm_dpo/margin": 77.0120849609375, "fcm_dpo/q_t": 0.3978939652442932, "grad_norm": 78.69764709472656, "learning_rate": 3.1249160234418644e-07, "logits/chosen": -0.8438678979873657, "logits/rejected": -0.8469147086143494, "logps/chosen": -410.0827331542969, "logps/ref_chosen": -291.9707946777344, "logps/ref_rejected": -263.42059326171875, "logps/rejected": -458.5446472167969, "loss": 4.2955, "margin_dpo/margin_mean": 77.0120849609375, "margin_dpo/margin_std": 98.03457641601562, "step": 229 }, { "epoch": 0.4816753926701571, "fcm_dpo/beta": 0.005542149767279625, "fcm_dpo/delta": -0.007225902751088142, "fcm_dpo/margin": 67.66149139404297, "fcm_dpo/q_t": 0.41533219814300537, "grad_norm": 66.38292694091797, "learning_rate": 3.1071729615293424e-07, "logits/chosen": -0.8894914984703064, "logits/rejected": -0.8882502913475037, "logps/chosen": -343.53985595703125, "logps/ref_chosen": -233.2601318359375, "logps/ref_rejected": -238.922119140625, "logps/rejected": -416.86334228515625, "loss": 4.5255, "margin_dpo/margin_mean": 67.66149139404297, "margin_dpo/margin_std": 105.4851303100586, "step": 230 }, { "epoch": 0.4837696335078534, "fcm_dpo/beta": 0.005650937091559172, "fcm_dpo/delta": 0.033986907452344894, "fcm_dpo/margin": 53.576927185058594, "fcm_dpo/q_t": 0.4296777844429016, "grad_norm": 108.66732788085938, "learning_rate": 3.0893973387735683e-07, "logits/chosen": -0.872734010219574, "logits/rejected": -0.8595795035362244, "logps/chosen": -443.4432067871094, "logps/ref_chosen": -322.1551818847656, "logps/ref_rejected": -280.97613525390625, "logps/rejected": -455.8411560058594, "loss": 4.7128, "margin_dpo/margin_mean": 53.57693099975586, "margin_dpo/margin_std": 93.04562377929688, "step": 231 }, { "epoch": 0.48586387434554973, "fcm_dpo/beta": 0.005723518319427967, "fcm_dpo/delta": -0.00418502651154995, "fcm_dpo/margin": 59.888526916503906, "fcm_dpo/q_t": 0.4218381643295288, "grad_norm": 60.41381072998047, "learning_rate": 3.071590108427243e-07, "logits/chosen": -0.8330118656158447, "logits/rejected": -0.8127278089523315, "logps/chosen": -386.4776306152344, "logps/ref_chosen": -271.7437744140625, "logps/ref_rejected": -249.94981384277344, "logps/rejected": -424.57220458984375, "loss": 4.6566, "margin_dpo/margin_mean": 59.888526916503906, "margin_dpo/margin_std": 101.21884155273438, "step": 232 }, { "epoch": 0.48795811518324606, "fcm_dpo/beta": 0.005629006307572126, "fcm_dpo/delta": -0.04465636610984802, "fcm_dpo/margin": 66.1546401977539, "fcm_dpo/q_t": 0.4142661988735199, "grad_norm": 70.80461120605469, "learning_rate": 3.05375222543809e-07, "logits/chosen": -0.888208270072937, "logits/rejected": -0.8782560229301453, "logps/chosen": -396.4659423828125, "logps/ref_chosen": -285.3423156738281, "logps/ref_rejected": -266.34320068359375, "logps/rejected": -443.6214599609375, "loss": 4.5031, "margin_dpo/margin_mean": 66.1546401977539, "margin_dpo/margin_std": 94.81285858154297, "step": 233 }, { "epoch": 0.4900523560209424, "fcm_dpo/beta": 0.005589499603956938, "fcm_dpo/delta": 0.039450354874134064, "fcm_dpo/margin": 60.44734191894531, "fcm_dpo/q_t": 0.42315778136253357, "grad_norm": 74.92431640625, "learning_rate": 3.035884646397637e-07, "logits/chosen": -0.8505942821502686, "logits/rejected": -0.8354034423828125, "logps/chosen": -406.40936279296875, "logps/ref_chosen": -294.9057312011719, "logps/ref_rejected": -299.37054443359375, "logps/rejected": -471.3215637207031, "loss": 4.673, "margin_dpo/margin_mean": 60.44734191894531, "margin_dpo/margin_std": 108.0342025756836, "step": 234 }, { "epoch": 0.49214659685863876, "fcm_dpo/beta": 0.005842198152095079, "fcm_dpo/delta": 0.028937824070453644, "fcm_dpo/margin": 63.570350646972656, "fcm_dpo/q_t": 0.41493111848831177, "grad_norm": 84.25675964355469, "learning_rate": 3.017988329489923e-07, "logits/chosen": -0.8750848770141602, "logits/rejected": -0.8671593070030212, "logps/chosen": -400.6502380371094, "logps/ref_chosen": -289.49755859375, "logps/ref_rejected": -247.55076599121094, "logps/rejected": -422.2738037109375, "loss": 4.5306, "margin_dpo/margin_mean": 63.57035827636719, "margin_dpo/margin_std": 100.27571105957031, "step": 235 }, { "epoch": 0.4942408376963351, "fcm_dpo/beta": 0.005945481359958649, "fcm_dpo/delta": 0.008337687700986862, "fcm_dpo/margin": 61.09648895263672, "fcm_dpo/q_t": 0.4176371097564697, "grad_norm": 99.53137969970703, "learning_rate": 3.000064234440111e-07, "logits/chosen": -0.8861789107322693, "logits/rejected": -0.886303722858429, "logps/chosen": -398.58074951171875, "logps/ref_chosen": -288.8846435546875, "logps/ref_rejected": -242.0452880859375, "logps/rejected": -412.837890625, "loss": 4.5815, "margin_dpo/margin_mean": 61.09648895263672, "margin_dpo/margin_std": 100.48046875, "step": 236 }, { "epoch": 0.4963350785340314, "fcm_dpo/beta": 0.005861309356987476, "fcm_dpo/delta": -0.013880519196391106, "fcm_dpo/margin": 64.32054901123047, "fcm_dpo/q_t": 0.4137817323207855, "grad_norm": 74.3822250366211, "learning_rate": 2.9821133224630223e-07, "logits/chosen": -0.8518946170806885, "logits/rejected": -0.8302955627441406, "logps/chosen": -378.9283447265625, "logps/ref_chosen": -265.47869873046875, "logps/ref_rejected": -267.9891357421875, "logps/rejected": -445.7593078613281, "loss": 4.4939, "margin_dpo/margin_mean": 64.320556640625, "margin_dpo/margin_std": 96.50254821777344, "step": 237 }, { "epoch": 0.49842931937172774, "fcm_dpo/beta": 0.005778386257588863, "fcm_dpo/delta": -0.012118183076381683, "fcm_dpo/margin": 63.46300506591797, "fcm_dpo/q_t": 0.4175938665866852, "grad_norm": 70.04883575439453, "learning_rate": 2.964136556211588e-07, "logits/chosen": -0.8639597296714783, "logits/rejected": -0.8326221704483032, "logps/chosen": -430.5649108886719, "logps/ref_chosen": -312.0026550292969, "logps/ref_rejected": -270.0257263183594, "logps/rejected": -452.0509338378906, "loss": 4.5471, "margin_dpo/margin_mean": 63.46300506591797, "margin_dpo/margin_std": 100.95140075683594, "step": 238 }, { "epoch": 0.5005235602094241, "fcm_dpo/beta": 0.005817565601319075, "fcm_dpo/delta": 0.051502831280231476, "fcm_dpo/margin": 54.329078674316406, "fcm_dpo/q_t": 0.4292038083076477, "grad_norm": 98.75479888916016, "learning_rate": 2.946134899725226e-07, "logits/chosen": -0.8470821380615234, "logits/rejected": -0.8864820003509521, "logps/chosen": -374.77459716796875, "logps/ref_chosen": -267.167236328125, "logps/ref_rejected": -275.99468994140625, "logps/rejected": -437.93115234375, "loss": 4.81, "margin_dpo/margin_mean": 54.32908248901367, "margin_dpo/margin_std": 114.33778381347656, "step": 239 }, { "epoch": 0.5026178010471204, "fcm_dpo/beta": 0.0058992416597902775, "fcm_dpo/delta": -0.005943944677710533, "fcm_dpo/margin": 68.67879486083984, "fcm_dpo/q_t": 0.40951305627822876, "grad_norm": 67.22561645507812, "learning_rate": 2.9281093183781403e-07, "logits/chosen": -0.9137483835220337, "logits/rejected": -0.907131552696228, "logps/chosen": -390.2958984375, "logps/ref_chosen": -285.9796142578125, "logps/ref_rejected": -256.8258056640625, "logps/rejected": -429.8209228515625, "loss": 4.4476, "margin_dpo/margin_mean": 68.67879486083984, "margin_dpo/margin_std": 104.33981323242188, "step": 240 }, { "epoch": 0.5047120418848168, "fcm_dpo/beta": 0.005862614139914513, "fcm_dpo/delta": -0.021802593022584915, "fcm_dpo/margin": 54.68340301513672, "fcm_dpo/q_t": 0.4280877113342285, "grad_norm": 72.86209106445312, "learning_rate": 2.910060778827554e-07, "logits/chosen": -0.8242596983909607, "logits/rejected": -0.8007845878601074, "logps/chosen": -376.6900634765625, "logps/ref_chosen": -261.516845703125, "logps/ref_rejected": -250.2250518798828, "logps/rejected": -420.0816650390625, "loss": 4.7379, "margin_dpo/margin_mean": 54.68340301513672, "margin_dpo/margin_std": 101.14353942871094, "step": 241 }, { "epoch": 0.506806282722513, "fcm_dpo/beta": 0.005866930354386568, "fcm_dpo/delta": 0.0024305330589413643, "fcm_dpo/margin": 62.140159606933594, "fcm_dpo/q_t": 0.4178650677204132, "grad_norm": 91.07866668701172, "learning_rate": 2.891990248961871e-07, "logits/chosen": -0.8852311968803406, "logits/rejected": -0.8625715374946594, "logps/chosen": -377.75555419921875, "logps/ref_chosen": -270.51397705078125, "logps/ref_rejected": -244.8560791015625, "logps/rejected": -414.23779296875, "loss": 4.5198, "margin_dpo/margin_mean": 62.140159606933594, "margin_dpo/margin_std": 93.75469970703125, "step": 242 }, { "epoch": 0.5089005235602094, "fcm_dpo/beta": 0.005890816915780306, "fcm_dpo/delta": -0.03509378433227539, "fcm_dpo/margin": 69.48873138427734, "fcm_dpo/q_t": 0.41002142429351807, "grad_norm": 100.6084213256836, "learning_rate": 2.873898697848762e-07, "logits/chosen": -0.8823165893554688, "logits/rejected": -0.8675477504730225, "logps/chosen": -425.2785949707031, "logps/ref_chosen": -324.68206787109375, "logps/ref_rejected": -307.1111755371094, "logps/rejected": -477.1964111328125, "loss": 4.4415, "margin_dpo/margin_mean": 69.48873138427734, "margin_dpo/margin_std": 104.60400390625, "step": 243 }, { "epoch": 0.5109947643979058, "fcm_dpo/beta": 0.005591516848653555, "fcm_dpo/delta": -0.020960122346878052, "fcm_dpo/margin": 70.47373962402344, "fcm_dpo/q_t": 0.4083375036716461, "grad_norm": 69.3757095336914, "learning_rate": 2.8557870956832133e-07, "logits/chosen": -0.8660691380500793, "logits/rejected": -0.8196066617965698, "logps/chosen": -420.827880859375, "logps/ref_chosen": -318.979248046875, "logps/ref_rejected": -269.67572021484375, "logps/rejected": -441.9980773925781, "loss": 4.421, "margin_dpo/margin_mean": 70.47373962402344, "margin_dpo/margin_std": 94.018310546875, "step": 244 }, { "epoch": 0.5130890052356021, "fcm_dpo/beta": 0.005576140247285366, "fcm_dpo/delta": -0.03441760316491127, "fcm_dpo/margin": 69.78692626953125, "fcm_dpo/q_t": 0.40951162576675415, "grad_norm": 68.46469116210938, "learning_rate": 2.837656413735479e-07, "logits/chosen": -0.8726673126220703, "logits/rejected": -0.8780596852302551, "logps/chosen": -396.1797790527344, "logps/ref_chosen": -294.8980712890625, "logps/ref_rejected": -239.8111114501953, "logps/rejected": -410.8797302246094, "loss": 4.4389, "margin_dpo/margin_mean": 69.78692626953125, "margin_dpo/margin_std": 94.53988647460938, "step": 245 }, { "epoch": 0.5151832460732985, "fcm_dpo/beta": 0.0054697575978934765, "fcm_dpo/delta": 0.019709015265107155, "fcm_dpo/margin": 50.34794616699219, "fcm_dpo/q_t": 0.4375728964805603, "grad_norm": 76.81562042236328, "learning_rate": 2.8195076242990116e-07, "logits/chosen": -0.8578512072563171, "logits/rejected": -0.8632481098175049, "logps/chosen": -399.8997497558594, "logps/ref_chosen": -280.6854248046875, "logps/ref_rejected": -253.65382385253906, "logps/rejected": -423.2160949707031, "loss": 4.8526, "margin_dpo/margin_mean": 50.34794616699219, "margin_dpo/margin_std": 104.99066925048828, "step": 246 }, { "epoch": 0.5172774869109947, "fcm_dpo/beta": 0.005524596199393272, "fcm_dpo/delta": -0.014701386913657188, "fcm_dpo/margin": 62.018592834472656, "fcm_dpo/q_t": 0.42184895277023315, "grad_norm": 54.281822204589844, "learning_rate": 2.801341700638307e-07, "logits/chosen": -0.8541011214256287, "logits/rejected": -0.8522176742553711, "logps/chosen": -391.5700988769531, "logps/ref_chosen": -281.1091003417969, "logps/ref_rejected": -260.3700866699219, "logps/rejected": -432.8497009277344, "loss": 4.594, "margin_dpo/margin_mean": 62.018592834472656, "margin_dpo/margin_std": 97.36813354492188, "step": 247 }, { "epoch": 0.5193717277486911, "fcm_dpo/beta": 0.0054681808687746525, "fcm_dpo/delta": 0.015473026782274246, "fcm_dpo/margin": 64.1961669921875, "fcm_dpo/q_t": 0.4186955392360687, "grad_norm": 63.69669723510742, "learning_rate": 2.7831596169367227e-07, "logits/chosen": -0.8172490000724792, "logits/rejected": -0.8243724703788757, "logps/chosen": -383.0730895996094, "logps/ref_chosen": -270.318359375, "logps/ref_rejected": -233.46778869628906, "logps/rejected": -410.4187316894531, "loss": 4.5768, "margin_dpo/margin_mean": 64.1961669921875, "margin_dpo/margin_std": 99.34171295166016, "step": 248 }, { "epoch": 0.5214659685863874, "fcm_dpo/beta": 0.005608071107417345, "fcm_dpo/delta": 0.03930206224322319, "fcm_dpo/margin": 53.0928955078125, "fcm_dpo/q_t": 0.43159425258636475, "grad_norm": 73.9835205078125, "learning_rate": 2.7649623482442274e-07, "logits/chosen": -0.8559899926185608, "logits/rejected": -0.8310372233390808, "logps/chosen": -408.27587890625, "logps/ref_chosen": -275.8088684082031, "logps/ref_rejected": -243.45138549804688, "logps/rejected": -429.0112609863281, "loss": 4.8237, "margin_dpo/margin_mean": 53.092891693115234, "margin_dpo/margin_std": 110.56341552734375, "step": 249 }, { "epoch": 0.5235602094240838, "fcm_dpo/beta": 0.005487566348165274, "fcm_dpo/delta": -0.07008317857980728, "fcm_dpo/margin": 72.46844482421875, "fcm_dpo/q_t": 0.4099717140197754, "grad_norm": 65.68206787109375, "learning_rate": 2.7467508704251135e-07, "logits/chosen": -0.8580400943756104, "logits/rejected": -0.8534699082374573, "logps/chosen": -428.248779296875, "logps/ref_chosen": -292.4945373535156, "logps/ref_rejected": -284.2869567871094, "logps/rejected": -492.5096740722656, "loss": 4.5088, "margin_dpo/margin_mean": 72.46844482421875, "margin_dpo/margin_std": 115.16688537597656, "step": 250 }, { "epoch": 0.5256544502617801, "fcm_dpo/beta": 0.005311680492013693, "fcm_dpo/delta": -0.06994134932756424, "fcm_dpo/margin": 68.04769134521484, "fcm_dpo/q_t": 0.4215785264968872, "grad_norm": 83.41246032714844, "learning_rate": 2.7285261601056697e-07, "logits/chosen": -0.8579660654067993, "logits/rejected": -0.8418058753013611, "logps/chosen": -408.4231872558594, "logps/ref_chosen": -281.736572265625, "logps/ref_rejected": -255.9419708251953, "logps/rejected": -450.6763000488281, "loss": 4.6009, "margin_dpo/margin_mean": 68.04769134521484, "margin_dpo/margin_std": 112.29561614990234, "step": 251 }, { "epoch": 0.5277486910994764, "fcm_dpo/beta": 0.005035778041929007, "fcm_dpo/delta": -0.015283557586371899, "fcm_dpo/margin": 65.64408874511719, "fcm_dpo/q_t": 0.422878623008728, "grad_norm": 59.68517303466797, "learning_rate": 2.7102891946217994e-07, "logits/chosen": -0.900702178478241, "logits/rejected": -0.8740664124488831, "logps/chosen": -434.17852783203125, "logps/ref_chosen": -295.9674072265625, "logps/ref_rejected": -280.111572265625, "logps/rejected": -483.966796875, "loss": 4.6299, "margin_dpo/margin_mean": 65.64408111572266, "margin_dpo/margin_std": 106.49784088134766, "step": 252 }, { "epoch": 0.5298429319371728, "fcm_dpo/beta": 0.005060987081378698, "fcm_dpo/delta": 0.01637944206595421, "fcm_dpo/margin": 60.1594123840332, "fcm_dpo/q_t": 0.4316590428352356, "grad_norm": 71.8646240234375, "learning_rate": 2.692040951966617e-07, "logits/chosen": -0.8741543292999268, "logits/rejected": -0.8654627799987793, "logps/chosen": -426.32373046875, "logps/ref_chosen": -277.072265625, "logps/ref_rejected": -247.31643676757812, "logps/rejected": -456.7273254394531, "loss": 4.7849, "margin_dpo/margin_mean": 60.15940856933594, "margin_dpo/margin_std": 116.84982299804688, "step": 253 }, { "epoch": 0.5319371727748691, "fcm_dpo/beta": 0.005230544600635767, "fcm_dpo/delta": 0.0011180378496646881, "fcm_dpo/margin": 66.23080444335938, "fcm_dpo/q_t": 0.4211539626121521, "grad_norm": 74.46015167236328, "learning_rate": 2.6737824107379947e-07, "logits/chosen": -0.8116303086280823, "logits/rejected": -0.7934176921844482, "logps/chosen": -410.2343444824219, "logps/ref_chosen": -269.9478454589844, "logps/ref_rejected": -249.45005798339844, "logps/rejected": -455.96734619140625, "loss": 4.5907, "margin_dpo/margin_mean": 66.23080444335938, "margin_dpo/margin_std": 105.38709259033203, "step": 254 }, { "epoch": 0.5340314136125655, "fcm_dpo/beta": 0.005045011173933744, "fcm_dpo/delta": -0.041239336133003235, "fcm_dpo/margin": 81.40687561035156, "fcm_dpo/q_t": 0.4063635468482971, "grad_norm": 63.57786560058594, "learning_rate": 2.655514550086086e-07, "logits/chosen": -0.8293032050132751, "logits/rejected": -0.7931342720985413, "logps/chosen": -439.56988525390625, "logps/ref_chosen": -306.6552734375, "logps/ref_rejected": -254.47528076171875, "logps/rejected": -468.7967529296875, "loss": 4.4478, "margin_dpo/margin_mean": 81.40687561035156, "margin_dpo/margin_std": 123.06184387207031, "step": 255 }, { "epoch": 0.5361256544502618, "fcm_dpo/beta": 0.004904068075120449, "fcm_dpo/delta": -0.02955251932144165, "fcm_dpo/margin": 66.60382843017578, "fcm_dpo/q_t": 0.4222066402435303, "grad_norm": 98.1750259399414, "learning_rate": 2.6372383496608186e-07, "logits/chosen": -0.8569645285606384, "logits/rejected": -0.8438127040863037, "logps/chosen": -463.7706604003906, "logps/ref_chosen": -323.7181701660156, "logps/ref_rejected": -254.1871337890625, "logps/rejected": -460.8434143066406, "loss": 4.7499, "margin_dpo/margin_mean": 66.60383605957031, "margin_dpo/margin_std": 121.21151733398438, "step": 256 }, { "epoch": 0.5382198952879581, "fcm_dpo/beta": 0.004877617582678795, "fcm_dpo/delta": 0.029186341911554337, "fcm_dpo/margin": 76.22327423095703, "fcm_dpo/q_t": 0.4146016538143158, "grad_norm": 75.77265167236328, "learning_rate": 2.618954789559356e-07, "logits/chosen": -0.8500054478645325, "logits/rejected": -0.8413119316101074, "logps/chosen": -405.16949462890625, "logps/ref_chosen": -267.21209716796875, "logps/ref_rejected": -249.12579345703125, "logps/rejected": -463.3064880371094, "loss": 4.485, "margin_dpo/margin_mean": 76.22327423095703, "margin_dpo/margin_std": 111.8670425415039, "step": 257 }, { "epoch": 0.5403141361256545, "fcm_dpo/beta": 0.004769755993038416, "fcm_dpo/delta": -0.0641399696469307, "fcm_dpo/margin": 77.75953674316406, "fcm_dpo/q_t": 0.4129870533943176, "grad_norm": 69.70844268798828, "learning_rate": 2.600664850273538e-07, "logits/chosen": -0.8756992816925049, "logits/rejected": -0.8453248143196106, "logps/chosen": -424.1638488769531, "logps/ref_chosen": -277.6827392578125, "logps/ref_rejected": -250.73385620117188, "logps/rejected": -474.9745178222656, "loss": 4.4652, "margin_dpo/margin_mean": 77.75953674316406, "margin_dpo/margin_std": 105.55914306640625, "step": 258 }, { "epoch": 0.5424083769633508, "fcm_dpo/beta": 0.004843279719352722, "fcm_dpo/delta": 0.04182092845439911, "fcm_dpo/margin": 74.08509826660156, "fcm_dpo/q_t": 0.4186086654663086, "grad_norm": 64.96932983398438, "learning_rate": 2.582369512637302e-07, "logits/chosen": -0.8835079073905945, "logits/rejected": -0.8777540326118469, "logps/chosen": -430.7030029296875, "logps/ref_chosen": -294.6099853515625, "logps/ref_rejected": -272.2725830078125, "logps/rejected": -482.45074462890625, "loss": 4.5272, "margin_dpo/margin_mean": 74.0851058959961, "margin_dpo/margin_std": 110.6324234008789, "step": 259 }, { "epoch": 0.5445026178010471, "fcm_dpo/beta": 0.005027398467063904, "fcm_dpo/delta": 0.04363645985722542, "fcm_dpo/margin": 30.95121192932129, "fcm_dpo/q_t": 0.464776873588562, "grad_norm": 92.6754379272461, "learning_rate": 2.5640697577740815e-07, "logits/chosen": -0.8735372424125671, "logits/rejected": -0.8678438663482666, "logps/chosen": -436.0877380371094, "logps/ref_chosen": -290.85711669921875, "logps/ref_rejected": -277.5970153808594, "logps/rejected": -453.77886962890625, "loss": 5.227, "margin_dpo/margin_mean": 30.951213836669922, "margin_dpo/margin_std": 104.45787048339844, "step": 260 }, { "epoch": 0.5465968586387434, "fcm_dpo/beta": 0.005115479230880737, "fcm_dpo/delta": 0.01615825854241848, "fcm_dpo/margin": 57.5634765625, "fcm_dpo/q_t": 0.4337337613105774, "grad_norm": 66.06670379638672, "learning_rate": 2.5457665670441937e-07, "logits/chosen": -0.7693032026290894, "logits/rejected": -0.779098391532898, "logps/chosen": -405.8778991699219, "logps/ref_chosen": -251.13223266601562, "logps/ref_rejected": -244.76016235351562, "logps/rejected": -457.0693054199219, "loss": 4.8364, "margin_dpo/margin_mean": 57.5634765625, "margin_dpo/margin_std": 120.9645767211914, "step": 261 }, { "epoch": 0.5486910994764398, "fcm_dpo/beta": 0.005165139213204384, "fcm_dpo/delta": -0.0015618130564689636, "fcm_dpo/margin": 77.5936279296875, "fcm_dpo/q_t": 0.4082508683204651, "grad_norm": 67.6512680053711, "learning_rate": 2.527460921992209e-07, "logits/chosen": -0.802527666091919, "logits/rejected": -0.7902448177337646, "logps/chosen": -441.3621520996094, "logps/ref_chosen": -299.7217712402344, "logps/ref_rejected": -277.0969543457031, "logps/rejected": -496.3309326171875, "loss": 4.4112, "margin_dpo/margin_mean": 77.5936279296875, "margin_dpo/margin_std": 107.57256317138672, "step": 262 }, { "epoch": 0.5507853403141362, "fcm_dpo/beta": 0.0051011075265705585, "fcm_dpo/delta": -0.00584130734205246, "fcm_dpo/margin": 68.14926147460938, "fcm_dpo/q_t": 0.4196454882621765, "grad_norm": 53.193485260009766, "learning_rate": 2.509153804294318e-07, "logits/chosen": -0.8066005706787109, "logits/rejected": -0.7877044677734375, "logps/chosen": -427.0610656738281, "logps/ref_chosen": -279.95257568359375, "logps/ref_rejected": -256.5327453613281, "logps/rejected": -471.7904357910156, "loss": 4.6222, "margin_dpo/margin_mean": 68.14926147460938, "margin_dpo/margin_std": 111.96087646484375, "step": 263 }, { "epoch": 0.5528795811518324, "fcm_dpo/beta": 0.004925370682030916, "fcm_dpo/delta": -0.04882174730300903, "fcm_dpo/margin": 83.56140899658203, "fcm_dpo/q_t": 0.40587475895881653, "grad_norm": 90.4492416381836, "learning_rate": 2.4908461957056825e-07, "logits/chosen": -0.833106517791748, "logits/rejected": -0.8265419602394104, "logps/chosen": -398.6038513183594, "logps/ref_chosen": -260.53509521484375, "logps/ref_rejected": -255.53799438476562, "logps/rejected": -477.16815185546875, "loss": 4.3584, "margin_dpo/margin_mean": 83.56140899658203, "margin_dpo/margin_std": 109.89978790283203, "step": 264 }, { "epoch": 0.5549738219895288, "fcm_dpo/beta": 0.0048619420267641544, "fcm_dpo/delta": -0.008534754626452923, "fcm_dpo/margin": 83.80038452148438, "fcm_dpo/q_t": 0.40826499462127686, "grad_norm": 55.20663833618164, "learning_rate": 2.4725390780077905e-07, "logits/chosen": -0.8974971771240234, "logits/rejected": -0.8996181488037109, "logps/chosen": -429.3133239746094, "logps/ref_chosen": -283.7130432128906, "logps/ref_rejected": -270.3209533691406, "logps/rejected": -499.7215881347656, "loss": 4.456, "margin_dpo/margin_mean": 83.80038452148438, "margin_dpo/margin_std": 126.92974853515625, "step": 265 }, { "epoch": 0.5570680628272251, "fcm_dpo/beta": 0.00486271595582366, "fcm_dpo/delta": -0.01186143234372139, "fcm_dpo/margin": 84.45220184326172, "fcm_dpo/q_t": 0.4058952331542969, "grad_norm": 56.53108596801758, "learning_rate": 2.454233432955807e-07, "logits/chosen": -0.9156976342201233, "logits/rejected": -0.8858764171600342, "logps/chosen": -408.8074645996094, "logps/ref_chosen": -278.09930419921875, "logps/ref_rejected": -260.6734619140625, "logps/rejected": -475.8338623046875, "loss": 4.3442, "margin_dpo/margin_mean": 84.45219421386719, "margin_dpo/margin_std": 107.75846862792969, "step": 266 }, { "epoch": 0.5591623036649215, "fcm_dpo/beta": 0.004877687431871891, "fcm_dpo/delta": 0.04254019632935524, "fcm_dpo/margin": 63.55180740356445, "fcm_dpo/q_t": 0.42915400862693787, "grad_norm": 75.33948516845703, "learning_rate": 2.435930242225919e-07, "logits/chosen": -0.8628742694854736, "logits/rejected": -0.8702367544174194, "logps/chosen": -437.90972900390625, "logps/ref_chosen": -280.33319091796875, "logps/ref_rejected": -247.78099060058594, "logps/rejected": -468.9093322753906, "loss": 4.7191, "margin_dpo/margin_mean": 63.55180740356445, "margin_dpo/margin_std": 111.55426025390625, "step": 267 }, { "epoch": 0.5612565445026177, "fcm_dpo/beta": 0.005009549204260111, "fcm_dpo/delta": -0.00161817017942667, "fcm_dpo/margin": 80.10919189453125, "fcm_dpo/q_t": 0.4091428220272064, "grad_norm": 85.9771728515625, "learning_rate": 2.4176304873626984e-07, "logits/chosen": -0.8233053684234619, "logits/rejected": -0.8017244338989258, "logps/chosen": -450.60400390625, "logps/ref_chosen": -304.1787109375, "logps/ref_rejected": -272.80316162109375, "logps/rejected": -499.337646484375, "loss": 4.4485, "margin_dpo/margin_mean": 80.10919189453125, "margin_dpo/margin_std": 118.68637084960938, "step": 268 }, { "epoch": 0.5633507853403141, "fcm_dpo/beta": 0.005120859481394291, "fcm_dpo/delta": 0.07810668647289276, "fcm_dpo/margin": 56.62480926513672, "fcm_dpo/q_t": 0.4334860146045685, "grad_norm": 95.90203857421875, "learning_rate": 2.399335149726463e-07, "logits/chosen": -0.8469225168228149, "logits/rejected": -0.8406363725662231, "logps/chosen": -397.451171875, "logps/ref_chosen": -249.84512329101562, "logps/ref_rejected": -223.37356567382812, "logps/rejected": -427.6044616699219, "loss": 4.8125, "margin_dpo/margin_mean": 56.62481689453125, "margin_dpo/margin_std": 115.53034973144531, "step": 269 }, { "epoch": 0.5654450261780105, "fcm_dpo/beta": 0.005242231767624617, "fcm_dpo/delta": -0.004636672325432301, "fcm_dpo/margin": 72.5596923828125, "fcm_dpo/q_t": 0.41470980644226074, "grad_norm": 76.87252807617188, "learning_rate": 2.381045210440644e-07, "logits/chosen": -0.9306774735450745, "logits/rejected": -0.9363957047462463, "logps/chosen": -470.66033935546875, "logps/ref_chosen": -318.5623779296875, "logps/ref_rejected": -281.1880798339844, "logps/rejected": -505.8457336425781, "loss": 4.5897, "margin_dpo/margin_mean": 72.5596923828125, "margin_dpo/margin_std": 127.58087158203125, "step": 270 }, { "epoch": 0.5675392670157068, "fcm_dpo/beta": 0.005433530081063509, "fcm_dpo/delta": 0.04082181677222252, "fcm_dpo/margin": 66.16189575195312, "fcm_dpo/q_t": 0.41947028040885925, "grad_norm": 74.05042266845703, "learning_rate": 2.3627616503391812e-07, "logits/chosen": -0.7789531946182251, "logits/rejected": -0.7725875377655029, "logps/chosen": -430.7342224121094, "logps/ref_chosen": -284.104736328125, "logps/ref_rejected": -253.9580535888672, "logps/rejected": -466.74945068359375, "loss": 4.5904, "margin_dpo/margin_mean": 66.16189575195312, "margin_dpo/margin_std": 110.10599517822266, "step": 271 }, { "epoch": 0.5696335078534032, "fcm_dpo/beta": 0.005346586927771568, "fcm_dpo/delta": -0.024095682427287102, "fcm_dpo/margin": 72.80128479003906, "fcm_dpo/q_t": 0.41350868344306946, "grad_norm": 74.77401733398438, "learning_rate": 2.344485449913914e-07, "logits/chosen": -0.8851820826530457, "logits/rejected": -0.8685218691825867, "logps/chosen": -435.2021789550781, "logps/ref_chosen": -297.3590087890625, "logps/ref_rejected": -279.20196533203125, "logps/rejected": -489.84637451171875, "loss": 4.5746, "margin_dpo/margin_mean": 72.80128479003906, "margin_dpo/margin_std": 126.85377502441406, "step": 272 }, { "epoch": 0.5717277486910994, "fcm_dpo/beta": 0.005236013326793909, "fcm_dpo/delta": -0.0287264883518219, "fcm_dpo/margin": 77.18766021728516, "fcm_dpo/q_t": 0.4100334346294403, "grad_norm": 94.14215850830078, "learning_rate": 2.3262175892620062e-07, "logits/chosen": -0.8636338710784912, "logits/rejected": -0.8754411935806274, "logps/chosen": -429.3293151855469, "logps/ref_chosen": -293.20574951171875, "logps/ref_rejected": -274.7646789550781, "logps/rejected": -488.0759582519531, "loss": 4.4805, "margin_dpo/margin_mean": 77.18766021728516, "margin_dpo/margin_std": 121.6794204711914, "step": 273 }, { "epoch": 0.5738219895287958, "fcm_dpo/beta": 0.005103556904941797, "fcm_dpo/delta": -0.03355909138917923, "fcm_dpo/margin": 84.57637023925781, "fcm_dpo/q_t": 0.4024480879306793, "grad_norm": 73.24378204345703, "learning_rate": 2.3079590480333827e-07, "logits/chosen": -0.8207792043685913, "logits/rejected": -0.7887733578681946, "logps/chosen": -406.8528747558594, "logps/ref_chosen": -270.55865478515625, "logps/ref_rejected": -239.47048950195312, "logps/rejected": -460.3410339355469, "loss": 4.3589, "margin_dpo/margin_mean": 84.57637023925781, "margin_dpo/margin_std": 119.9758529663086, "step": 274 }, { "epoch": 0.5759162303664922, "fcm_dpo/beta": 0.004932451993227005, "fcm_dpo/delta": -0.0627283900976181, "fcm_dpo/margin": 93.08427429199219, "fcm_dpo/q_t": 0.3963705599308014, "grad_norm": 54.85517883300781, "learning_rate": 2.2897108053782e-07, "logits/chosen": -0.8735396265983582, "logits/rejected": -0.8545541167259216, "logps/chosen": -372.95513916015625, "logps/ref_chosen": -250.31922912597656, "logps/ref_rejected": -249.3187255859375, "logps/rejected": -465.03887939453125, "loss": 4.2232, "margin_dpo/margin_mean": 93.08427429199219, "margin_dpo/margin_std": 114.38727569580078, "step": 275 }, { "epoch": 0.5780104712041885, "fcm_dpo/beta": 0.004693637136369944, "fcm_dpo/delta": -0.02087191678583622, "fcm_dpo/margin": 84.33247375488281, "fcm_dpo/q_t": 0.4087947607040405, "grad_norm": 56.812870025634766, "learning_rate": 2.2714738398943308e-07, "logits/chosen": -0.9385483264923096, "logits/rejected": -0.9158673286437988, "logps/chosen": -425.16796875, "logps/ref_chosen": -297.6310729980469, "logps/ref_rejected": -295.225830078125, "logps/rejected": -507.0952453613281, "loss": 4.4153, "margin_dpo/margin_mean": 84.33247375488281, "margin_dpo/margin_std": 113.92533111572266, "step": 276 }, { "epoch": 0.5801047120418849, "fcm_dpo/beta": 0.004713152069598436, "fcm_dpo/delta": -0.0023143887519836426, "fcm_dpo/margin": 60.579532623291016, "fcm_dpo/q_t": 0.435024619102478, "grad_norm": 88.57623291015625, "learning_rate": 2.2532491295748865e-07, "logits/chosen": -0.8644924163818359, "logits/rejected": -0.862896203994751, "logps/chosen": -413.9727478027344, "logps/ref_chosen": -266.3604736328125, "logps/ref_rejected": -253.36767578125, "logps/rejected": -461.5594787597656, "loss": 4.8088, "margin_dpo/margin_mean": 60.57952880859375, "margin_dpo/margin_std": 119.62329864501953, "step": 277 }, { "epoch": 0.5821989528795811, "fcm_dpo/beta": 0.004758753813803196, "fcm_dpo/delta": 0.0039313724264502525, "fcm_dpo/margin": 48.2127685546875, "fcm_dpo/q_t": 0.4520694613456726, "grad_norm": 71.64497375488281, "learning_rate": 2.2350376517557726e-07, "logits/chosen": -0.8992569446563721, "logits/rejected": -0.8604874610900879, "logps/chosen": -429.3371887207031, "logps/ref_chosen": -267.40728759765625, "logps/ref_rejected": -229.5758514404297, "logps/rejected": -439.71844482421875, "loss": 5.129, "margin_dpo/margin_mean": 48.2127685546875, "margin_dpo/margin_std": 137.95343017578125, "step": 278 }, { "epoch": 0.5842931937172775, "fcm_dpo/beta": 0.004630140028893948, "fcm_dpo/delta": -0.053754448890686035, "fcm_dpo/margin": 88.50892639160156, "fcm_dpo/q_t": 0.4081512689590454, "grad_norm": 72.25919342041016, "learning_rate": 2.2168403830632769e-07, "logits/chosen": -0.8258615732192993, "logits/rejected": -0.8100024461746216, "logps/chosen": -460.78607177734375, "logps/ref_chosen": -313.3677978515625, "logps/ref_rejected": -299.1744384765625, "logps/rejected": -535.1016235351562, "loss": 4.433, "margin_dpo/margin_mean": 88.50892639160156, "margin_dpo/margin_std": 131.84158325195312, "step": 279 }, { "epoch": 0.5863874345549738, "fcm_dpo/beta": 0.00471831439062953, "fcm_dpo/delta": 0.07752120494842529, "fcm_dpo/margin": 68.82743072509766, "fcm_dpo/q_t": 0.4256266951560974, "grad_norm": 99.79914855957031, "learning_rate": 2.1986582993616925e-07, "logits/chosen": -0.8818544745445251, "logits/rejected": -0.8929405212402344, "logps/chosen": -402.27252197265625, "logps/ref_chosen": -265.5558166503906, "logps/ref_rejected": -247.1573944091797, "logps/rejected": -452.7015075683594, "loss": 4.6549, "margin_dpo/margin_mean": 68.82743072509766, "margin_dpo/margin_std": 118.33948516845703, "step": 280 }, { "epoch": 0.5884816753926702, "fcm_dpo/beta": 0.004735568072646856, "fcm_dpo/delta": -0.01806649938225746, "fcm_dpo/margin": 82.60304260253906, "fcm_dpo/q_t": 0.4123595356941223, "grad_norm": 80.9848403930664, "learning_rate": 2.1804923757009882e-07, "logits/chosen": -0.8654816746711731, "logits/rejected": -0.8740391731262207, "logps/chosen": -450.2287292480469, "logps/ref_chosen": -295.2995910644531, "logps/ref_rejected": -293.80877685546875, "logps/rejected": -531.3409423828125, "loss": 4.449, "margin_dpo/margin_mean": 82.60304260253906, "margin_dpo/margin_std": 119.16482543945312, "step": 281 }, { "epoch": 0.5905759162303665, "fcm_dpo/beta": 0.004867776297032833, "fcm_dpo/delta": 0.05005919188261032, "fcm_dpo/margin": 72.23600769042969, "fcm_dpo/q_t": 0.42044195532798767, "grad_norm": 58.8224983215332, "learning_rate": 2.1623435862645205e-07, "logits/chosen": -0.8650054335594177, "logits/rejected": -0.8629118204116821, "logps/chosen": -474.85797119140625, "logps/ref_chosen": -318.63714599609375, "logps/ref_rejected": -273.5943603515625, "logps/rejected": -502.0511474609375, "loss": 4.6176, "margin_dpo/margin_mean": 72.23600769042969, "margin_dpo/margin_std": 123.56523895263672, "step": 282 }, { "epoch": 0.5926701570680628, "fcm_dpo/beta": 0.005018922034651041, "fcm_dpo/delta": -0.012126617133617401, "fcm_dpo/margin": 71.27740478515625, "fcm_dpo/q_t": 0.4213730990886688, "grad_norm": 69.60215759277344, "learning_rate": 2.1442129043167873e-07, "logits/chosen": -0.8856289386749268, "logits/rejected": -0.8754106163978577, "logps/chosen": -412.56719970703125, "logps/ref_chosen": -254.66053771972656, "logps/ref_rejected": -236.8627166748047, "logps/rejected": -466.04681396484375, "loss": 4.6072, "margin_dpo/margin_mean": 71.27740478515625, "margin_dpo/margin_std": 120.48731994628906, "step": 283 }, { "epoch": 0.5947643979057592, "fcm_dpo/beta": 0.004824446514248848, "fcm_dpo/delta": -0.02942648157477379, "fcm_dpo/margin": 76.38724517822266, "fcm_dpo/q_t": 0.417685866355896, "grad_norm": 83.93893432617188, "learning_rate": 2.1261013021512378e-07, "logits/chosen": -0.8511925935745239, "logits/rejected": -0.8199682235717773, "logps/chosen": -439.28143310546875, "logps/ref_chosen": -273.355224609375, "logps/ref_rejected": -259.84759521484375, "logps/rejected": -502.1610107421875, "loss": 4.6521, "margin_dpo/margin_mean": 76.38724517822266, "margin_dpo/margin_std": 137.0977020263672, "step": 284 }, { "epoch": 0.5968586387434555, "fcm_dpo/beta": 0.004868158604949713, "fcm_dpo/delta": 0.020393716171383858, "fcm_dpo/margin": 51.72495651245117, "fcm_dpo/q_t": 0.4441811740398407, "grad_norm": 100.43576049804688, "learning_rate": 2.1080097510381294e-07, "logits/chosen": -0.857195258140564, "logits/rejected": -0.8506641983985901, "logps/chosen": -479.5204162597656, "logps/ref_chosen": -309.8022155761719, "logps/ref_rejected": -279.11846923828125, "logps/rejected": -500.56158447265625, "loss": 5.0083, "margin_dpo/margin_mean": 51.72496032714844, "margin_dpo/margin_std": 129.5350341796875, "step": 285 }, { "epoch": 0.5989528795811518, "fcm_dpo/beta": 0.004937829915434122, "fcm_dpo/delta": 0.030296623706817627, "fcm_dpo/margin": 75.0483169555664, "fcm_dpo/q_t": 0.418967604637146, "grad_norm": 80.72938537597656, "learning_rate": 2.089939221172446e-07, "logits/chosen": -0.8420528769493103, "logits/rejected": -0.8334404230117798, "logps/chosen": -428.751708984375, "logps/ref_chosen": -271.4655456542969, "logps/ref_rejected": -279.531494140625, "logps/rejected": -511.86590576171875, "loss": 4.5885, "margin_dpo/margin_mean": 75.0483169555664, "margin_dpo/margin_std": 129.76441955566406, "step": 286 }, { "epoch": 0.6010471204188481, "fcm_dpo/beta": 0.005080940201878548, "fcm_dpo/delta": 0.021761678159236908, "fcm_dpo/margin": 69.6319808959961, "fcm_dpo/q_t": 0.4205819368362427, "grad_norm": 79.26040649414062, "learning_rate": 2.0718906816218595e-07, "logits/chosen": -0.862068772315979, "logits/rejected": -0.8453592658042908, "logps/chosen": -439.6429138183594, "logps/ref_chosen": -277.0932312011719, "logps/ref_rejected": -233.55599975585938, "logps/rejected": -465.7376708984375, "loss": 4.6931, "margin_dpo/margin_mean": 69.63197326660156, "margin_dpo/margin_std": 132.2081298828125, "step": 287 }, { "epoch": 0.6031413612565445, "fcm_dpo/beta": 0.0051753930747509, "fcm_dpo/delta": 0.0042931921780109406, "fcm_dpo/margin": 69.9853286743164, "fcm_dpo/q_t": 0.41905421018600464, "grad_norm": 87.4442367553711, "learning_rate": 2.053865100274774e-07, "logits/chosen": -0.8746588230133057, "logits/rejected": -0.8808341026306152, "logps/chosen": -454.9645690917969, "logps/ref_chosen": -293.1681823730469, "logps/ref_rejected": -263.4059143066406, "logps/rejected": -495.1876525878906, "loss": 4.6358, "margin_dpo/margin_mean": 69.98532104492188, "margin_dpo/margin_std": 125.97311401367188, "step": 288 }, { "epoch": 0.6052356020942409, "fcm_dpo/beta": 0.0051370663568377495, "fcm_dpo/delta": 0.02070062793791294, "fcm_dpo/margin": 44.1146240234375, "fcm_dpo/q_t": 0.44912248849868774, "grad_norm": 95.3306655883789, "learning_rate": 2.035863443788411e-07, "logits/chosen": -0.8595668077468872, "logits/rejected": -0.8414284586906433, "logps/chosen": -511.45074462890625, "logps/ref_chosen": -329.9574279785156, "logps/ref_rejected": -276.7565002441406, "logps/rejected": -502.36444091796875, "loss": 5.0894, "margin_dpo/margin_mean": 44.11461639404297, "margin_dpo/margin_std": 121.32159423828125, "step": 289 }, { "epoch": 0.6073298429319371, "fcm_dpo/beta": 0.005096154753118753, "fcm_dpo/delta": -7.194560021162033e-05, "fcm_dpo/margin": 67.14459991455078, "fcm_dpo/q_t": 0.42460641264915466, "grad_norm": 65.90011596679688, "learning_rate": 2.0178866775369774e-07, "logits/chosen": -0.870934247970581, "logits/rejected": -0.8062535524368286, "logps/chosen": -490.2457275390625, "logps/ref_chosen": -324.6690673828125, "logps/ref_rejected": -311.8439636230469, "logps/rejected": -544.565185546875, "loss": 4.7519, "margin_dpo/margin_mean": 67.14459991455078, "margin_dpo/margin_std": 129.93505859375, "step": 290 }, { "epoch": 0.6094240837696335, "fcm_dpo/beta": 0.00493369298055768, "fcm_dpo/delta": -0.10477574914693832, "fcm_dpo/margin": 93.4791259765625, "fcm_dpo/q_t": 0.3963235020637512, "grad_norm": 67.77316284179688, "learning_rate": 1.9999357655598891e-07, "logits/chosen": -0.843751847743988, "logits/rejected": -0.8341413736343384, "logps/chosen": -427.7283020019531, "logps/ref_chosen": -274.1440734863281, "logps/ref_rejected": -278.07208251953125, "logps/rejected": -525.1354370117188, "loss": 4.2918, "margin_dpo/margin_mean": 93.4791259765625, "margin_dpo/margin_std": 124.0925064086914, "step": 291 }, { "epoch": 0.6115183246073298, "fcm_dpo/beta": 0.004825775045901537, "fcm_dpo/delta": 0.0411670058965683, "fcm_dpo/margin": 69.62248229980469, "fcm_dpo/q_t": 0.4242509603500366, "grad_norm": 80.3649673461914, "learning_rate": 1.9820116705100775e-07, "logits/chosen": -0.8350319266319275, "logits/rejected": -0.8311899304389954, "logps/chosen": -408.04681396484375, "logps/ref_chosen": -259.3636779785156, "logps/ref_rejected": -279.30218505859375, "logps/rejected": -497.6078186035156, "loss": 4.6933, "margin_dpo/margin_mean": 69.62248229980469, "margin_dpo/margin_std": 126.6540756225586, "step": 292 }, { "epoch": 0.6136125654450262, "fcm_dpo/beta": 0.004759517032653093, "fcm_dpo/delta": -0.059455305337905884, "fcm_dpo/margin": 83.2960205078125, "fcm_dpo/q_t": 0.4100151062011719, "grad_norm": 87.87876892089844, "learning_rate": 1.9641153536023642e-07, "logits/chosen": -0.9336636066436768, "logits/rejected": -0.8930791020393372, "logps/chosen": -467.6336975097656, "logps/ref_chosen": -303.77081298828125, "logps/ref_rejected": -270.07513427734375, "logps/rejected": -517.2340087890625, "loss": 4.4479, "margin_dpo/margin_mean": 83.29600524902344, "margin_dpo/margin_std": 118.2719497680664, "step": 293 }, { "epoch": 0.6157068062827226, "fcm_dpo/beta": 0.004677574150264263, "fcm_dpo/delta": -0.016423970460891724, "fcm_dpo/margin": 69.96250915527344, "fcm_dpo/q_t": 0.4272857904434204, "grad_norm": 78.82388305664062, "learning_rate": 1.9462477745619106e-07, "logits/chosen": -0.8205356597900391, "logits/rejected": -0.8255838751792908, "logps/chosen": -395.3521423339844, "logps/ref_chosen": -240.23831176757812, "logps/ref_rejected": -229.187744140625, "logps/rejected": -454.2640686035156, "loss": 4.7076, "margin_dpo/margin_mean": 69.96250915527344, "margin_dpo/margin_std": 129.4488525390625, "step": 294 }, { "epoch": 0.6178010471204188, "fcm_dpo/beta": 0.0048176939599215984, "fcm_dpo/delta": 0.07201889902353287, "fcm_dpo/margin": 68.49911499023438, "fcm_dpo/q_t": 0.4257529377937317, "grad_norm": 70.18067169189453, "learning_rate": 1.928409891572757e-07, "logits/chosen": -0.8293942809104919, "logits/rejected": -0.8370305299758911, "logps/chosen": -405.40020751953125, "logps/ref_chosen": -251.00970458984375, "logps/ref_rejected": -244.15142822265625, "logps/rejected": -467.04107666015625, "loss": 4.6407, "margin_dpo/margin_mean": 68.49911499023438, "margin_dpo/margin_std": 115.05201721191406, "step": 295 }, { "epoch": 0.6198952879581152, "fcm_dpo/beta": 0.004698417615145445, "fcm_dpo/delta": -0.10293805599212646, "fcm_dpo/margin": 94.19628143310547, "fcm_dpo/q_t": 0.39925792813301086, "grad_norm": 77.33747100830078, "learning_rate": 1.9106026612264315e-07, "logits/chosen": -0.8193361163139343, "logits/rejected": -0.7859427332878113, "logps/chosen": -453.1386413574219, "logps/ref_chosen": -293.880615234375, "logps/ref_rejected": -283.4175720214844, "logps/rejected": -536.871826171875, "loss": 4.3803, "margin_dpo/margin_mean": 94.19628143310547, "margin_dpo/margin_std": 136.64402770996094, "step": 296 }, { "epoch": 0.6219895287958115, "fcm_dpo/beta": 0.004520534072071314, "fcm_dpo/delta": 0.020609168335795403, "fcm_dpo/margin": 70.18128967285156, "fcm_dpo/q_t": 0.4274723529815674, "grad_norm": 74.41461181640625, "learning_rate": 1.8928270384706582e-07, "logits/chosen": -0.9127456545829773, "logits/rejected": -0.9068336486816406, "logps/chosen": -445.99737548828125, "logps/ref_chosen": -289.4600830078125, "logps/ref_rejected": -283.69110107421875, "logps/rejected": -510.4096374511719, "loss": 4.6623, "margin_dpo/margin_mean": 70.18128204345703, "margin_dpo/margin_std": 116.06803894042969, "step": 297 }, { "epoch": 0.6240837696335079, "fcm_dpo/beta": 0.004513401072472334, "fcm_dpo/delta": -0.054587312042713165, "fcm_dpo/margin": 85.99492645263672, "fcm_dpo/q_t": 0.41201794147491455, "grad_norm": 96.58114624023438, "learning_rate": 1.875083976558136e-07, "logits/chosen": -0.8198727369308472, "logits/rejected": -0.806681752204895, "logps/chosen": -451.9382629394531, "logps/ref_chosen": -306.5150146484375, "logps/ref_rejected": -280.6969909667969, "logps/rejected": -512.1151123046875, "loss": 4.4997, "margin_dpo/margin_mean": 85.99492645263672, "margin_dpo/margin_std": 131.11622619628906, "step": 298 }, { "epoch": 0.6261780104712041, "fcm_dpo/beta": 0.004364268854260445, "fcm_dpo/delta": 0.0030471403151750565, "fcm_dpo/margin": 65.27977752685547, "fcm_dpo/q_t": 0.434037983417511, "grad_norm": 108.02179718017578, "learning_rate": 1.8573744269954297e-07, "logits/chosen": -0.8087353706359863, "logits/rejected": -0.7938067317008972, "logps/chosen": -452.2359619140625, "logps/ref_chosen": -281.36376953125, "logps/ref_rejected": -270.39508056640625, "logps/rejected": -506.5469970703125, "loss": 4.7327, "margin_dpo/margin_mean": 65.27976989746094, "margin_dpo/margin_std": 106.35151672363281, "step": 299 }, { "epoch": 0.6282722513089005, "fcm_dpo/beta": 0.004509190563112497, "fcm_dpo/delta": 0.03436897695064545, "fcm_dpo/margin": 64.97026062011719, "fcm_dpo/q_t": 0.4333421587944031, "grad_norm": 89.92118072509766, "learning_rate": 1.839699339491937e-07, "logits/chosen": -0.8405961394309998, "logits/rejected": -0.8179261684417725, "logps/chosen": -488.3516845703125, "logps/ref_chosen": -314.83575439453125, "logps/ref_rejected": -269.1154479980469, "logps/rejected": -507.60162353515625, "loss": 4.8012, "margin_dpo/margin_mean": 64.97026062011719, "margin_dpo/margin_std": 130.7782745361328, "step": 300 }, { "epoch": 0.6303664921465969, "fcm_dpo/beta": 0.004641157109290361, "fcm_dpo/delta": 0.02008698135614395, "fcm_dpo/margin": 65.45753479003906, "fcm_dpo/q_t": 0.4307628273963928, "grad_norm": 92.52803039550781, "learning_rate": 1.8220596619089573e-07, "logits/chosen": -0.8634936809539795, "logits/rejected": -0.8615702986717224, "logps/chosen": -440.056396484375, "logps/ref_chosen": -279.89453125, "logps/ref_rejected": -271.6694641113281, "logps/rejected": -497.28887939453125, "loss": 4.7366, "margin_dpo/margin_mean": 65.45753479003906, "margin_dpo/margin_std": 119.31248474121094, "step": 301 }, { "epoch": 0.6324607329842932, "fcm_dpo/beta": 0.004585019312798977, "fcm_dpo/delta": -0.02878117375075817, "fcm_dpo/margin": 86.58937072753906, "fcm_dpo/q_t": 0.41024476289749146, "grad_norm": 90.13982391357422, "learning_rate": 1.8044563402088682e-07, "logits/chosen": -0.8196691274642944, "logits/rejected": -0.8060354590415955, "logps/chosen": -425.8458557128906, "logps/ref_chosen": -271.3318176269531, "logps/ref_rejected": -256.5587158203125, "logps/rejected": -497.662109375, "loss": 4.4549, "margin_dpo/margin_mean": 86.58937072753906, "margin_dpo/margin_std": 128.49935913085938, "step": 302 }, { "epoch": 0.6345549738219896, "fcm_dpo/beta": 0.004555375315248966, "fcm_dpo/delta": 0.001330256462097168, "fcm_dpo/margin": 75.77513122558594, "fcm_dpo/q_t": 0.4200977385044098, "grad_norm": 67.80012512207031, "learning_rate": 1.7868903184043885e-07, "logits/chosen": -0.8044949769973755, "logits/rejected": -0.7818641662597656, "logps/chosen": -465.8041687011719, "logps/ref_chosen": -304.88104248046875, "logps/ref_rejected": -269.063720703125, "logps/rejected": -505.7619934082031, "loss": 4.5828, "margin_dpo/margin_mean": 75.77513122558594, "margin_dpo/margin_std": 117.83062744140625, "step": 303 }, { "epoch": 0.6366492146596858, "fcm_dpo/beta": 0.00458510359749198, "fcm_dpo/delta": 0.032261330634355545, "fcm_dpo/margin": 80.3938980102539, "fcm_dpo/q_t": 0.4192785322666168, "grad_norm": 106.86233520507812, "learning_rate": 1.7693625385079574e-07, "logits/chosen": -0.8235783576965332, "logits/rejected": -0.8320932984352112, "logps/chosen": -467.0224914550781, "logps/ref_chosen": -290.7109680175781, "logps/ref_rejected": -237.6885986328125, "logps/rejected": -494.3940124511719, "loss": 4.578, "margin_dpo/margin_mean": 80.39390563964844, "margin_dpo/margin_std": 138.0482940673828, "step": 304 }, { "epoch": 0.6387434554973822, "fcm_dpo/beta": 0.004341554827988148, "fcm_dpo/delta": -0.15256354212760925, "fcm_dpo/margin": 125.14611053466797, "fcm_dpo/q_t": 0.376392126083374, "grad_norm": 80.46356964111328, "learning_rate": 1.7518739404812155e-07, "logits/chosen": -0.8713183999061584, "logits/rejected": -0.8404445052146912, "logps/chosen": -404.1719970703125, "logps/ref_chosen": -256.4839782714844, "logps/ref_rejected": -266.4063415527344, "logps/rejected": -539.2404174804688, "loss": 4.0074, "margin_dpo/margin_mean": 125.14610290527344, "margin_dpo/margin_std": 134.68260192871094, "step": 305 }, { "epoch": 0.6408376963350786, "fcm_dpo/beta": 0.004074703436344862, "fcm_dpo/delta": -0.005278175696730614, "fcm_dpo/margin": 75.07460021972656, "fcm_dpo/q_t": 0.42913058400154114, "grad_norm": 89.13674926757812, "learning_rate": 1.7344254621846017e-07, "logits/chosen": -0.8748356103897095, "logits/rejected": -0.85565185546875, "logps/chosen": -481.8687438964844, "logps/ref_chosen": -320.6492004394531, "logps/ref_rejected": -273.36773681640625, "logps/rejected": -509.661865234375, "loss": 4.6633, "margin_dpo/margin_mean": 75.07460021972656, "margin_dpo/margin_std": 118.12226867675781, "step": 306 }, { "epoch": 0.6429319371727749, "fcm_dpo/beta": 0.00412240345031023, "fcm_dpo/delta": 0.012996518984436989, "fcm_dpo/margin": 80.91729736328125, "fcm_dpo/q_t": 0.4214695692062378, "grad_norm": 82.34172821044922, "learning_rate": 1.717018039327053e-07, "logits/chosen": -0.797187328338623, "logits/rejected": -0.8197383880615234, "logps/chosen": -467.67266845703125, "logps/ref_chosen": -279.4541931152344, "logps/ref_rejected": -240.3796844482422, "logps/rejected": -509.5154724121094, "loss": 4.5522, "margin_dpo/margin_mean": 80.91729736328125, "margin_dpo/margin_std": 112.15681457519531, "step": 307 }, { "epoch": 0.6450261780104712, "fcm_dpo/beta": 0.004232144448906183, "fcm_dpo/delta": 0.035529762506484985, "fcm_dpo/margin": 54.43116760253906, "fcm_dpo/q_t": 0.44700539112091064, "grad_norm": 83.30457305908203, "learning_rate": 1.699652605415828e-07, "logits/chosen": -0.8497411608695984, "logits/rejected": -0.861391007900238, "logps/chosen": -498.0782775878906, "logps/ref_chosen": -296.598388671875, "logps/ref_rejected": -258.6953430175781, "logps/rejected": -514.6063232421875, "loss": 4.9775, "margin_dpo/margin_mean": 54.43116760253906, "margin_dpo/margin_std": 127.53193664550781, "step": 308 }, { "epoch": 0.6471204188481675, "fcm_dpo/beta": 0.004292917437851429, "fcm_dpo/delta": -0.0011705029755830765, "fcm_dpo/margin": 93.36857604980469, "fcm_dpo/q_t": 0.4098639190196991, "grad_norm": 85.22145080566406, "learning_rate": 1.6823300917064458e-07, "logits/chosen": -0.8477848768234253, "logits/rejected": -0.8559066653251648, "logps/chosen": -473.5204772949219, "logps/ref_chosen": -281.3881530761719, "logps/ref_rejected": -262.458740234375, "logps/rejected": -547.9596557617188, "loss": 4.4233, "margin_dpo/margin_mean": 93.36858367919922, "margin_dpo/margin_std": 136.71563720703125, "step": 309 }, { "epoch": 0.6492146596858639, "fcm_dpo/beta": 0.004227439407259226, "fcm_dpo/delta": -0.016845714300870895, "fcm_dpo/margin": 92.73661041259766, "fcm_dpo/q_t": 0.4104473292827606, "grad_norm": 60.97142791748047, "learning_rate": 1.6650514271527465e-07, "logits/chosen": -0.8435422778129578, "logits/rejected": -0.8177253603935242, "logps/chosen": -460.1365966796875, "logps/ref_chosen": -279.1872863769531, "logps/ref_rejected": -261.8279724121094, "logps/rejected": -535.5139770507812, "loss": 4.4209, "margin_dpo/margin_mean": 92.73661804199219, "margin_dpo/margin_std": 128.3031768798828, "step": 310 }, { "epoch": 0.6513089005235602, "fcm_dpo/beta": 0.004221054259687662, "fcm_dpo/delta": 0.017837712541222572, "fcm_dpo/margin": 90.44427490234375, "fcm_dpo/q_t": 0.4134417176246643, "grad_norm": 96.84386444091797, "learning_rate": 1.647817538357072e-07, "logits/chosen": -0.8366580009460449, "logits/rejected": -0.8164286017417908, "logps/chosen": -461.29248046875, "logps/ref_chosen": -271.39813232421875, "logps/ref_rejected": -266.12701416015625, "logps/rejected": -546.4656372070312, "loss": 4.533, "margin_dpo/margin_mean": 90.44427490234375, "margin_dpo/margin_std": 142.28948974609375, "step": 311 }, { "epoch": 0.6534031413612565, "fcm_dpo/beta": 0.004198572598397732, "fcm_dpo/delta": -0.027397790923714638, "fcm_dpo/margin": 80.31361389160156, "fcm_dpo/q_t": 0.42590564489364624, "grad_norm": 79.09515380859375, "learning_rate": 1.6306293495205755e-07, "logits/chosen": -0.854910135269165, "logits/rejected": -0.8275889158248901, "logps/chosen": -471.61541748046875, "logps/ref_chosen": -282.3850402832031, "logps/ref_rejected": -246.35389709472656, "logps/rejected": -515.89794921875, "loss": 4.7836, "margin_dpo/margin_mean": 80.31360626220703, "margin_dpo/margin_std": 162.17025756835938, "step": 312 }, { "epoch": 0.6554973821989529, "fcm_dpo/beta": 0.00422461424022913, "fcm_dpo/delta": -0.010405285283923149, "fcm_dpo/margin": 79.79673767089844, "fcm_dpo/q_t": 0.42779484391212463, "grad_norm": 71.66043090820312, "learning_rate": 1.6134877823936607e-07, "logits/chosen": -0.8918551802635193, "logits/rejected": -0.8759722113609314, "logps/chosen": -499.627685546875, "logps/ref_chosen": -303.630859375, "logps/ref_rejected": -273.1156921386719, "logps/rejected": -548.9091796875, "loss": 4.7486, "margin_dpo/margin_mean": 79.79674530029297, "margin_dpo/margin_std": 153.3138427734375, "step": 313 }, { "epoch": 0.6575916230366492, "fcm_dpo/beta": 0.004204154014587402, "fcm_dpo/delta": 0.028585905209183693, "fcm_dpo/margin": 82.11723327636719, "fcm_dpo/q_t": 0.4200282096862793, "grad_norm": 86.5287094116211, "learning_rate": 1.5963937562265522e-07, "logits/chosen": -0.9091111421585083, "logits/rejected": -0.8985500931739807, "logps/chosen": -488.7541809082031, "logps/ref_chosen": -302.3042907714844, "logps/ref_rejected": -273.6416015625, "logps/rejected": -542.208740234375, "loss": 4.5815, "margin_dpo/margin_mean": 82.11723327636719, "margin_dpo/margin_std": 129.3647918701172, "step": 314 }, { "epoch": 0.6596858638743456, "fcm_dpo/beta": 0.004223410505801439, "fcm_dpo/delta": -0.0216530729085207, "fcm_dpo/margin": 99.56622314453125, "fcm_dpo/q_t": 0.4038830101490021, "grad_norm": 83.03425598144531, "learning_rate": 1.5793481877199943e-07, "logits/chosen": -0.8853978514671326, "logits/rejected": -0.8687517642974854, "logps/chosen": -488.8519287109375, "logps/ref_chosen": -302.729248046875, "logps/ref_rejected": -270.26910400390625, "logps/rejected": -555.9579467773438, "loss": 4.3519, "margin_dpo/margin_mean": 99.56622314453125, "margin_dpo/margin_std": 133.00856018066406, "step": 315 }, { "epoch": 0.6617801047120419, "fcm_dpo/beta": 0.00421603349968791, "fcm_dpo/delta": 0.030252397060394287, "fcm_dpo/margin": 87.81375885009766, "fcm_dpo/q_t": 0.41594982147216797, "grad_norm": 83.00886535644531, "learning_rate": 1.562351990976095e-07, "logits/chosen": -0.8991329669952393, "logits/rejected": -0.8829232454299927, "logps/chosen": -503.4185791015625, "logps/ref_chosen": -310.5706481933594, "logps/ref_rejected": -272.9354553222656, "logps/rejected": -553.5971069335938, "loss": 4.5283, "margin_dpo/margin_mean": 87.81375885009766, "margin_dpo/margin_std": 137.16415405273438, "step": 316 }, { "epoch": 0.6638743455497382, "fcm_dpo/beta": 0.0043156505562365055, "fcm_dpo/delta": 0.03792955353856087, "fcm_dpo/margin": 84.17947387695312, "fcm_dpo/q_t": 0.41570180654525757, "grad_norm": 84.58760070800781, "learning_rate": 1.5454060774493065e-07, "logits/chosen": -0.8874872326850891, "logits/rejected": -0.856697142124176, "logps/chosen": -421.66094970703125, "logps/ref_chosen": -253.90036010742188, "logps/ref_rejected": -218.74078369140625, "logps/rejected": -470.6808166503906, "loss": 4.4808, "margin_dpo/margin_mean": 84.17947387695312, "margin_dpo/margin_std": 114.98719787597656, "step": 317 }, { "epoch": 0.6659685863874345, "fcm_dpo/beta": 0.004230810329318047, "fcm_dpo/delta": -0.027981776744127274, "fcm_dpo/margin": 100.45878601074219, "fcm_dpo/q_t": 0.40261608362197876, "grad_norm": 77.26625061035156, "learning_rate": 1.5285113558975427e-07, "logits/chosen": -0.8962690234184265, "logits/rejected": -0.8660792112350464, "logps/chosen": -450.1083068847656, "logps/ref_chosen": -270.8228759765625, "logps/ref_rejected": -255.30972290039062, "logps/rejected": -535.053955078125, "loss": 4.3339, "margin_dpo/margin_mean": 100.45878601074219, "margin_dpo/margin_std": 129.95045471191406, "step": 318 }, { "epoch": 0.6680628272251309, "fcm_dpo/beta": 0.004204513505101204, "fcm_dpo/delta": -0.003102468326687813, "fcm_dpo/margin": 95.72822570800781, "fcm_dpo/q_t": 0.4059513807296753, "grad_norm": 78.08094787597656, "learning_rate": 1.5116687323334464e-07, "logits/chosen": -0.8939340710639954, "logits/rejected": -0.8638850450515747, "logps/chosen": -490.3048095703125, "logps/ref_chosen": -301.0028076171875, "logps/ref_rejected": -242.39002990722656, "logps/rejected": -527.4202270507812, "loss": 4.3517, "margin_dpo/margin_mean": 95.72822570800781, "margin_dpo/margin_std": 119.91444396972656, "step": 319 }, { "epoch": 0.6701570680628273, "fcm_dpo/beta": 0.004289536736905575, "fcm_dpo/delta": 0.027375973761081696, "fcm_dpo/margin": 86.9979248046875, "fcm_dpo/q_t": 0.4167506992816925, "grad_norm": 90.90167999267578, "learning_rate": 1.4948791099758052e-07, "logits/chosen": -0.8552721738815308, "logits/rejected": -0.847703218460083, "logps/chosen": -487.06640625, "logps/ref_chosen": -303.6225891113281, "logps/ref_rejected": -280.85174560546875, "logps/rejected": -551.29345703125, "loss": 4.5643, "margin_dpo/margin_mean": 86.99790954589844, "margin_dpo/margin_std": 145.2067108154297, "step": 320 }, { "epoch": 0.6722513089005235, "fcm_dpo/beta": 0.004387743771076202, "fcm_dpo/delta": 0.007859650999307632, "fcm_dpo/margin": 63.733009338378906, "fcm_dpo/q_t": 0.43765002489089966, "grad_norm": 98.84996795654297, "learning_rate": 1.478143389201113e-07, "logits/chosen": -0.8751094341278076, "logits/rejected": -0.839246392250061, "logps/chosen": -492.1627197265625, "logps/ref_chosen": -288.98583984375, "logps/ref_rejected": -241.1822052001953, "logps/rejected": -508.09210205078125, "loss": 4.8733, "margin_dpo/margin_mean": 63.733009338378906, "margin_dpo/margin_std": 136.8006134033203, "step": 321 }, { "epoch": 0.6743455497382199, "fcm_dpo/beta": 0.004408278036862612, "fcm_dpo/delta": 0.0005717193707823753, "fcm_dpo/margin": 78.75841522216797, "fcm_dpo/q_t": 0.4216492176055908, "grad_norm": 109.37677001953125, "learning_rate": 1.461462467495284e-07, "logits/chosen": -0.9387675523757935, "logits/rejected": -0.9010107517242432, "logps/chosen": -514.61083984375, "logps/ref_chosen": -308.54345703125, "logps/ref_rejected": -269.7995910644531, "logps/rejected": -554.6253662109375, "loss": 4.6688, "margin_dpo/margin_mean": 78.75841522216797, "margin_dpo/margin_std": 143.8833770751953, "step": 322 }, { "epoch": 0.6764397905759162, "fcm_dpo/beta": 0.00443174596875906, "fcm_dpo/delta": 0.02427692338824272, "fcm_dpo/margin": 56.93408203125, "fcm_dpo/q_t": 0.4440670907497406, "grad_norm": 74.94538879394531, "learning_rate": 1.4448372394055246e-07, "logits/chosen": -0.8896764516830444, "logits/rejected": -0.8818888664245605, "logps/chosen": -475.91375732421875, "logps/ref_chosen": -282.49365234375, "logps/ref_rejected": -227.7105255126953, "logps/rejected": -478.0647277832031, "loss": 4.9551, "margin_dpo/margin_mean": 56.934078216552734, "margin_dpo/margin_std": 135.64027404785156, "step": 323 }, { "epoch": 0.6785340314136126, "fcm_dpo/beta": 0.004333406686782837, "fcm_dpo/delta": -0.036704014986753464, "fcm_dpo/margin": 94.2815170288086, "fcm_dpo/q_t": 0.4057313799858093, "grad_norm": 85.12026977539062, "learning_rate": 1.428268596492364e-07, "logits/chosen": -0.8165361881256104, "logits/rejected": -0.8111166954040527, "logps/chosen": -414.974609375, "logps/ref_chosen": -239.33836364746094, "logps/ref_rejected": -230.53775024414062, "logps/rejected": -500.4554748535156, "loss": 4.3573, "margin_dpo/margin_mean": 94.2815170288086, "margin_dpo/margin_std": 120.80596923828125, "step": 324 }, { "epoch": 0.680628272251309, "fcm_dpo/beta": 0.0042426129803061485, "fcm_dpo/delta": -0.04353080317378044, "fcm_dpo/margin": 80.39572143554688, "fcm_dpo/q_t": 0.42427390813827515, "grad_norm": 99.24666595458984, "learning_rate": 1.4117574272818386e-07, "logits/chosen": -0.8341665267944336, "logits/rejected": -0.8182295560836792, "logps/chosen": -474.089599609375, "logps/ref_chosen": -280.62896728515625, "logps/ref_rejected": -270.5085754394531, "logps/rejected": -544.3649291992188, "loss": 4.7072, "margin_dpo/margin_mean": 80.39572143554688, "margin_dpo/margin_std": 150.9606475830078, "step": 325 }, { "epoch": 0.6827225130890052, "fcm_dpo/beta": 0.004248365759849548, "fcm_dpo/delta": 0.007518507540225983, "fcm_dpo/margin": 86.27885437011719, "fcm_dpo/q_t": 0.41677290201187134, "grad_norm": 89.02911376953125, "learning_rate": 1.3953046172178413e-07, "logits/chosen": -0.9490206837654114, "logits/rejected": -0.9398881793022156, "logps/chosen": -419.48565673828125, "logps/ref_chosen": -240.9871368408203, "logps/ref_rejected": -261.0238342285156, "logps/rejected": -525.8012084960938, "loss": 4.5545, "margin_dpo/margin_mean": 86.27885437011719, "margin_dpo/margin_std": 138.67755126953125, "step": 326 }, { "epoch": 0.6848167539267016, "fcm_dpo/beta": 0.0041707661002874374, "fcm_dpo/delta": -0.025365371257066727, "fcm_dpo/margin": 95.02404022216797, "fcm_dpo/q_t": 0.4098770022392273, "grad_norm": 87.0390625, "learning_rate": 1.3789110486146468e-07, "logits/chosen": -0.8910970091819763, "logits/rejected": -0.8668678402900696, "logps/chosen": -449.1541748046875, "logps/ref_chosen": -279.52001953125, "logps/ref_rejected": -269.51824951171875, "logps/rejected": -534.1763916015625, "loss": 4.4215, "margin_dpo/margin_mean": 95.02404022216797, "margin_dpo/margin_std": 131.46246337890625, "step": 327 }, { "epoch": 0.6869109947643979, "fcm_dpo/beta": 0.00401464756578207, "fcm_dpo/delta": -0.02622944489121437, "fcm_dpo/margin": 85.4066162109375, "fcm_dpo/q_t": 0.4205254912376404, "grad_norm": 76.4749984741211, "learning_rate": 1.362577600609588e-07, "logits/chosen": -0.8552443981170654, "logits/rejected": -0.8537087440490723, "logps/chosen": -487.69732666015625, "logps/ref_chosen": -301.033447265625, "logps/ref_rejected": -284.2101135253906, "logps/rejected": -556.2805786132812, "loss": 4.5527, "margin_dpo/margin_mean": 85.4066162109375, "margin_dpo/margin_std": 124.48458099365234, "step": 328 }, { "epoch": 0.6890052356020943, "fcm_dpo/beta": 0.003990216180682182, "fcm_dpo/delta": -0.024558117613196373, "fcm_dpo/margin": 92.10006713867188, "fcm_dpo/q_t": 0.42077887058258057, "grad_norm": 75.36544799804688, "learning_rate": 1.3463051491159093e-07, "logits/chosen": -0.8669880628585815, "logits/rejected": -0.8379704356193542, "logps/chosen": -517.63623046875, "logps/ref_chosen": -319.9888610839844, "logps/ref_rejected": -307.5588684082031, "logps/rejected": -597.3063354492188, "loss": 4.6236, "margin_dpo/margin_mean": 92.10005950927734, "margin_dpo/margin_std": 161.97157287597656, "step": 329 }, { "epoch": 0.6910994764397905, "fcm_dpo/beta": 0.003911246545612812, "fcm_dpo/delta": -0.030769605189561844, "fcm_dpo/margin": 88.48474884033203, "fcm_dpo/q_t": 0.42029914259910583, "grad_norm": 71.30492401123047, "learning_rate": 1.3300945667758012e-07, "logits/chosen": -0.8565258979797363, "logits/rejected": -0.8653894066810608, "logps/chosen": -496.8501281738281, "logps/ref_chosen": -301.11474609375, "logps/ref_rejected": -299.673095703125, "logps/rejected": -583.8932495117188, "loss": 4.5507, "margin_dpo/margin_mean": 88.48474884033203, "margin_dpo/margin_std": 128.97657775878906, "step": 330 }, { "epoch": 0.6931937172774869, "fcm_dpo/beta": 0.003975651692599058, "fcm_dpo/delta": 0.05870046094059944, "fcm_dpo/margin": 86.33970642089844, "fcm_dpo/q_t": 0.4234520494937897, "grad_norm": 80.85090637207031, "learning_rate": 1.3139467229135998e-07, "logits/chosen": -0.8877269625663757, "logits/rejected": -0.8683705925941467, "logps/chosen": -460.4813537597656, "logps/ref_chosen": -277.59149169921875, "logps/ref_rejected": -256.025634765625, "logps/rejected": -525.2552490234375, "loss": 4.6499, "margin_dpo/margin_mean": 86.33970642089844, "margin_dpo/margin_std": 154.41885375976562, "step": 331 }, { "epoch": 0.6952879581151833, "fcm_dpo/beta": 0.004023251123726368, "fcm_dpo/delta": 0.006525871343910694, "fcm_dpo/margin": 86.42638397216797, "fcm_dpo/q_t": 0.4220367670059204, "grad_norm": 101.5770492553711, "learning_rate": 1.2978624834891626e-07, "logits/chosen": -0.8897654414176941, "logits/rejected": -0.8640623688697815, "logps/chosen": -462.71600341796875, "logps/ref_chosen": -269.97369384765625, "logps/ref_rejected": -235.03164672851562, "logps/rejected": -514.2003173828125, "loss": 4.6419, "margin_dpo/margin_mean": 86.4263916015625, "margin_dpo/margin_std": 151.5164794921875, "step": 332 }, { "epoch": 0.6973821989528796, "fcm_dpo/beta": 0.004221638664603233, "fcm_dpo/delta": 0.05486953258514404, "fcm_dpo/margin": 73.84952545166016, "fcm_dpo/q_t": 0.4293982982635498, "grad_norm": 72.1107177734375, "learning_rate": 1.281842711051438e-07, "logits/chosen": -0.962734043598175, "logits/rejected": -0.9257638454437256, "logps/chosen": -493.1941833496094, "logps/ref_chosen": -296.76300048828125, "logps/ref_rejected": -265.97991943359375, "logps/rejected": -536.2606201171875, "loss": 4.7055, "margin_dpo/margin_mean": 73.84952545166016, "margin_dpo/margin_std": 133.64422607421875, "step": 333 }, { "epoch": 0.6994764397905759, "fcm_dpo/beta": 0.0043148864060640335, "fcm_dpo/delta": -0.00039426563307642937, "fcm_dpo/margin": 82.25745391845703, "fcm_dpo/q_t": 0.4192145764827728, "grad_norm": 142.5056610107422, "learning_rate": 1.2658882646922033e-07, "logits/chosen": -0.8579068183898926, "logits/rejected": -0.8308162093162537, "logps/chosen": -490.0419616699219, "logps/ref_chosen": -301.0367431640625, "logps/ref_rejected": -268.87652587890625, "logps/rejected": -540.1392211914062, "loss": 4.6052, "margin_dpo/margin_mean": 82.2574462890625, "margin_dpo/margin_std": 134.9955596923828, "step": 334 }, { "epoch": 0.7015706806282722, "fcm_dpo/beta": 0.0042117261327803135, "fcm_dpo/delta": -0.005903298035264015, "fcm_dpo/margin": 88.66685485839844, "fcm_dpo/q_t": 0.4157736599445343, "grad_norm": 82.71585845947266, "learning_rate": 1.2500000000000005e-07, "logits/chosen": -0.8564417958259583, "logits/rejected": -0.850709080696106, "logps/chosen": -481.53729248046875, "logps/ref_chosen": -276.13275146484375, "logps/ref_rejected": -243.44203186035156, "logps/rejected": -537.5133666992188, "loss": 4.551, "margin_dpo/margin_mean": 88.66685485839844, "margin_dpo/margin_std": 140.4600372314453, "step": 335 }, { "epoch": 0.7036649214659686, "fcm_dpo/beta": 0.004215937573462725, "fcm_dpo/delta": -0.017655007541179657, "fcm_dpo/margin": 89.4686508178711, "fcm_dpo/q_t": 0.4191427230834961, "grad_norm": 69.75729370117188, "learning_rate": 1.2341787690142435e-07, "logits/chosen": -0.8736502528190613, "logits/rejected": -0.812055230140686, "logps/chosen": -450.0522155761719, "logps/ref_chosen": -246.2626495361328, "logps/ref_rejected": -261.0617980957031, "logps/rejected": -554.3200073242188, "loss": 4.574, "margin_dpo/margin_mean": 89.4686508178711, "margin_dpo/margin_std": 151.58094787597656, "step": 336 }, { "epoch": 0.7057591623036649, "fcm_dpo/beta": 0.0042139277793467045, "fcm_dpo/delta": -0.02299882099032402, "fcm_dpo/margin": 93.4886474609375, "fcm_dpo/q_t": 0.41079553961753845, "grad_norm": 79.24539184570312, "learning_rate": 1.2184254201795363e-07, "logits/chosen": -0.8735207915306091, "logits/rejected": -0.8454840779304504, "logps/chosen": -458.2684631347656, "logps/ref_chosen": -266.9937744140625, "logps/ref_rejected": -253.015625, "logps/rejected": -537.7789916992188, "loss": 4.4457, "margin_dpo/margin_mean": 93.48865509033203, "margin_dpo/margin_std": 134.67636108398438, "step": 337 }, { "epoch": 0.7078534031413612, "fcm_dpo/beta": 0.0041390592232346535, "fcm_dpo/delta": 0.013191011734306812, "fcm_dpo/margin": 93.37725830078125, "fcm_dpo/q_t": 0.4126090407371521, "grad_norm": 120.32525634765625, "learning_rate": 1.202740798300168e-07, "logits/chosen": -0.8974190354347229, "logits/rejected": -0.8797302842140198, "logps/chosen": -459.82012939453125, "logps/ref_chosen": -276.5925598144531, "logps/ref_rejected": -233.979248046875, "logps/rejected": -510.5840759277344, "loss": 4.4602, "margin_dpo/margin_mean": 93.37725830078125, "margin_dpo/margin_std": 135.0716552734375, "step": 338 }, { "epoch": 0.7099476439790576, "fcm_dpo/beta": 0.0041606188751757145, "fcm_dpo/delta": -0.020034300163388252, "fcm_dpo/margin": 100.59808349609375, "fcm_dpo/q_t": 0.40568268299102783, "grad_norm": 83.7204818725586, "learning_rate": 1.1871257444948096e-07, "logits/chosen": -0.9190041422843933, "logits/rejected": -0.906645655632019, "logps/chosen": -496.95233154296875, "logps/ref_chosen": -303.5277404785156, "logps/ref_rejected": -283.11676025390625, "logps/rejected": -577.1394653320312, "loss": 4.4169, "margin_dpo/margin_mean": 100.59807586669922, "margin_dpo/margin_std": 148.40968322753906, "step": 339 }, { "epoch": 0.7120418848167539, "fcm_dpo/beta": 0.004161160439252853, "fcm_dpo/delta": 0.032229986041784286, "fcm_dpo/margin": 82.14170837402344, "fcm_dpo/q_t": 0.42351680994033813, "grad_norm": 130.79080200195312, "learning_rate": 1.1715810961514072e-07, "logits/chosen": -0.8522177338600159, "logits/rejected": -0.8474718928337097, "logps/chosen": -462.0082092285156, "logps/ref_chosen": -261.5257568359375, "logps/ref_rejected": -259.39862060546875, "logps/rejected": -542.0227661132812, "loss": 4.7617, "margin_dpo/margin_mean": 82.14170837402344, "margin_dpo/margin_std": 163.79534912109375, "step": 340 }, { "epoch": 0.7141361256544503, "fcm_dpo/beta": 0.004220739006996155, "fcm_dpo/delta": 0.013660956174135208, "fcm_dpo/margin": 72.86450958251953, "fcm_dpo/q_t": 0.4320552945137024, "grad_norm": 74.50018310546875, "learning_rate": 1.1561076868822755e-07, "logits/chosen": -0.8715128302574158, "logits/rejected": -0.8439801335334778, "logps/chosen": -547.073974609375, "logps/ref_chosen": -315.903564453125, "logps/ref_rejected": -308.02392578125, "logps/rejected": -612.058837890625, "loss": 4.9256, "margin_dpo/margin_mean": 72.86450958251953, "margin_dpo/margin_std": 168.4367218017578, "step": 341 }, { "epoch": 0.7162303664921466, "fcm_dpo/beta": 0.0043666851706802845, "fcm_dpo/delta": 0.02423066645860672, "fcm_dpo/margin": 86.04833984375, "fcm_dpo/q_t": 0.4149947166442871, "grad_norm": 124.93977355957031, "learning_rate": 1.1407063464793965e-07, "logits/chosen": -0.872001051902771, "logits/rejected": -0.8675014972686768, "logps/chosen": -462.8236389160156, "logps/ref_chosen": -269.17864990234375, "logps/ref_rejected": -260.8977355957031, "logps/rejected": -540.591064453125, "loss": 4.5223, "margin_dpo/margin_mean": 86.04833984375, "margin_dpo/margin_std": 132.913330078125, "step": 342 }, { "epoch": 0.7183246073298429, "fcm_dpo/beta": 0.004478834103792906, "fcm_dpo/delta": 0.07679037004709244, "fcm_dpo/margin": 72.70214080810547, "fcm_dpo/q_t": 0.4263514280319214, "grad_norm": 76.69132232666016, "learning_rate": 1.125377900869913e-07, "logits/chosen": -0.8687958121299744, "logits/rejected": -0.8440370559692383, "logps/chosen": -512.5940551757812, "logps/ref_chosen": -310.719970703125, "logps/ref_rejected": -263.5224914550781, "logps/rejected": -538.0987548828125, "loss": 4.6727, "margin_dpo/margin_mean": 72.70214080810547, "margin_dpo/margin_std": 129.76275634765625, "step": 343 }, { "epoch": 0.7204188481675393, "fcm_dpo/beta": 0.004755112808197737, "fcm_dpo/delta": 0.020119212567806244, "fcm_dpo/margin": 74.89362335205078, "fcm_dpo/q_t": 0.41980746388435364, "grad_norm": 83.10517883300781, "learning_rate": 1.110123172071844e-07, "logits/chosen": -0.8663875460624695, "logits/rejected": -0.84798264503479, "logps/chosen": -502.5235595703125, "logps/ref_chosen": -301.7999267578125, "logps/ref_rejected": -257.9061584472656, "logps/rejected": -533.5234375, "loss": 4.6603, "margin_dpo/margin_mean": 74.89362335205078, "margin_dpo/margin_std": 134.99688720703125, "step": 344 }, { "epoch": 0.7225130890052356, "fcm_dpo/beta": 0.0046288310550153255, "fcm_dpo/delta": -0.04487127810716629, "fcm_dpo/margin": 70.49630737304688, "fcm_dpo/q_t": 0.4266197681427002, "grad_norm": 96.98583221435547, "learning_rate": 1.09494297815e-07, "logits/chosen": -0.8665734529495239, "logits/rejected": -0.8673686385154724, "logps/chosen": -476.5813293457031, "logps/ref_chosen": -283.0184326171875, "logps/ref_rejected": -266.8457336425781, "logps/rejected": -530.9049072265625, "loss": 4.6982, "margin_dpo/margin_mean": 70.49630737304688, "margin_dpo/margin_std": 119.51622009277344, "step": 345 }, { "epoch": 0.724607329842932, "fcm_dpo/beta": 0.004324314650148153, "fcm_dpo/delta": -0.08264245837926865, "fcm_dpo/margin": 94.54742431640625, "fcm_dpo/q_t": 0.4078022241592407, "grad_norm": 105.37242889404297, "learning_rate": 1.0798381331721107e-07, "logits/chosen": -0.9467877149581909, "logits/rejected": -0.8949044942855835, "logps/chosen": -467.9733581542969, "logps/ref_chosen": -268.44122314453125, "logps/ref_rejected": -227.8225860595703, "logps/rejected": -521.9021606445312, "loss": 4.4796, "margin_dpo/margin_mean": 94.54740905761719, "margin_dpo/margin_std": 140.02549743652344, "step": 346 }, { "epoch": 0.7267015706806282, "fcm_dpo/beta": 0.004258990287780762, "fcm_dpo/delta": 0.015542431734502316, "fcm_dpo/margin": 73.12641906738281, "fcm_dpo/q_t": 0.4280158281326294, "grad_norm": 129.21615600585938, "learning_rate": 1.0648094471651722e-07, "logits/chosen": -0.7985144853591919, "logits/rejected": -0.8250570297241211, "logps/chosen": -464.026123046875, "logps/ref_chosen": -273.70355224609375, "logps/ref_rejected": -243.65521240234375, "logps/rejected": -507.10418701171875, "loss": 4.6989, "margin_dpo/margin_mean": 73.12641906738281, "margin_dpo/margin_std": 124.44351196289062, "step": 347 }, { "epoch": 0.7287958115183246, "fcm_dpo/beta": 0.004324248060584068, "fcm_dpo/delta": 0.028727039694786072, "fcm_dpo/margin": 64.96218872070312, "fcm_dpo/q_t": 0.4387781620025635, "grad_norm": 88.1794204711914, "learning_rate": 1.0498577260720048e-07, "logits/chosen": -0.9022852182388306, "logits/rejected": -0.8873810172080994, "logps/chosen": -476.2867431640625, "logps/ref_chosen": -285.64141845703125, "logps/ref_rejected": -265.6270446777344, "logps/rejected": -521.2345581054688, "loss": 4.856, "margin_dpo/margin_mean": 64.96218872070312, "margin_dpo/margin_std": 137.919677734375, "step": 348 }, { "epoch": 0.7308900523560209, "fcm_dpo/beta": 0.004297360777854919, "fcm_dpo/delta": -0.05822340026497841, "fcm_dpo/margin": 99.03752899169922, "fcm_dpo/q_t": 0.40331995487213135, "grad_norm": 105.6890869140625, "learning_rate": 1.0349837717080347e-07, "logits/chosen": -0.8452585935592651, "logits/rejected": -0.8333654999732971, "logps/chosen": -520.8684692382812, "logps/ref_chosen": -328.3175048828125, "logps/ref_rejected": -292.37872314453125, "logps/rejected": -583.96728515625, "loss": 4.4394, "margin_dpo/margin_mean": 99.03752899169922, "margin_dpo/margin_std": 149.55874633789062, "step": 349 }, { "epoch": 0.7329842931937173, "fcm_dpo/beta": 0.004243547562509775, "fcm_dpo/delta": 0.00749508710578084, "fcm_dpo/margin": 84.4142837524414, "fcm_dpo/q_t": 0.41955479979515076, "grad_norm": 84.40665435791016, "learning_rate": 1.0201883817182949e-07, "logits/chosen": -0.837745189666748, "logits/rejected": -0.8492467403411865, "logps/chosen": -495.3204345703125, "logps/ref_chosen": -292.8046569824219, "logps/ref_rejected": -250.35504150390625, "logps/rejected": -537.2850341796875, "loss": 4.6258, "margin_dpo/margin_mean": 84.41429138183594, "margin_dpo/margin_std": 147.89923095703125, "step": 350 }, { "epoch": 0.7350785340314137, "fcm_dpo/beta": 0.0042998939752578735, "fcm_dpo/delta": 0.04794508218765259, "fcm_dpo/margin": 55.46940231323242, "fcm_dpo/q_t": 0.4465586841106415, "grad_norm": 104.56925964355469, "learning_rate": 1.0054723495346482e-07, "logits/chosen": -0.9081228971481323, "logits/rejected": -0.890868604183197, "logps/chosen": -506.27545166015625, "logps/ref_chosen": -311.8890380859375, "logps/ref_rejected": -263.59033203125, "logps/rejected": -513.4461669921875, "loss": 5.0328, "margin_dpo/margin_mean": 55.46940231323242, "margin_dpo/margin_std": 144.48912048339844, "step": 351 }, { "epoch": 0.7371727748691099, "fcm_dpo/beta": 0.004258439410477877, "fcm_dpo/delta": -0.04971547797322273, "fcm_dpo/margin": 104.7568359375, "fcm_dpo/q_t": 0.3977693021297455, "grad_norm": 80.12105560302734, "learning_rate": 9.908364643332398e-08, "logits/chosen": -0.8247711658477783, "logits/rejected": -0.7942376136779785, "logps/chosen": -434.2840270996094, "logps/ref_chosen": -254.9078826904297, "logps/ref_rejected": -257.1688232421875, "logps/rejected": -541.3017578125, "loss": 4.3164, "margin_dpo/margin_mean": 104.75682830810547, "margin_dpo/margin_std": 138.1365203857422, "step": 352 }, { "epoch": 0.7392670157068063, "fcm_dpo/beta": 0.0041551426984369755, "fcm_dpo/delta": -0.007299358025193214, "fcm_dpo/margin": 88.50493621826172, "fcm_dpo/q_t": 0.41657987236976624, "grad_norm": 94.97777557373047, "learning_rate": 9.76281510992176e-08, "logits/chosen": -0.8463056087493896, "logits/rejected": -0.8403018712997437, "logps/chosen": -462.415771484375, "logps/ref_chosen": -270.3760681152344, "logps/ref_rejected": -264.65234375, "logps/rejected": -545.197021484375, "loss": 4.5474, "margin_dpo/margin_mean": 88.50493621826172, "margin_dpo/margin_std": 137.2867431640625, "step": 353 }, { "epoch": 0.7413612565445026, "fcm_dpo/beta": 0.0042567066848278046, "fcm_dpo/delta": 0.03705097734928131, "fcm_dpo/margin": 65.5347671508789, "fcm_dpo/q_t": 0.43766477704048157, "grad_norm": 72.41242980957031, "learning_rate": 9.618082700494318e-08, "logits/chosen": -0.8418868780136108, "logits/rejected": -0.8751367330551147, "logps/chosen": -450.8121337890625, "logps/ref_chosen": -257.6485595703125, "logps/ref_rejected": -246.94203186035156, "logps/rejected": -505.6404113769531, "loss": 4.8655, "margin_dpo/margin_mean": 65.53477478027344, "margin_dpo/margin_std": 140.56112670898438, "step": 354 }, { "epoch": 0.743455497382199, "fcm_dpo/beta": 0.004172952845692635, "fcm_dpo/delta": -0.03083830326795578, "fcm_dpo/margin": 95.28348541259766, "fcm_dpo/q_t": 0.41133761405944824, "grad_norm": 79.00312042236328, "learning_rate": 9.474175176609956e-08, "logits/chosen": -0.8952876925468445, "logits/rejected": -0.8920347690582275, "logps/chosen": -486.91064453125, "logps/ref_chosen": -293.35333251953125, "logps/ref_rejected": -275.6051940917969, "logps/rejected": -564.4459838867188, "loss": 4.5619, "margin_dpo/margin_mean": 95.28348541259766, "margin_dpo/margin_std": 158.88365173339844, "step": 355 }, { "epoch": 0.7455497382198953, "fcm_dpo/beta": 0.004256485030055046, "fcm_dpo/delta": 0.039695486426353455, "fcm_dpo/margin": 67.3565444946289, "fcm_dpo/q_t": 0.43414920568466187, "grad_norm": 107.32857513427734, "learning_rate": 9.331100255592436e-08, "logits/chosen": -0.8071290850639343, "logits/rejected": -0.834446132183075, "logps/chosen": -386.90557861328125, "logps/ref_chosen": -204.25550842285156, "logps/ref_rejected": -213.467529296875, "logps/rejected": -463.4741516113281, "loss": 4.7445, "margin_dpo/margin_mean": 67.35653686523438, "margin_dpo/margin_std": 116.29612731933594, "step": 356 }, { "epoch": 0.7476439790575916, "fcm_dpo/beta": 0.004231618717312813, "fcm_dpo/delta": -0.04037679359316826, "fcm_dpo/margin": 89.51058959960938, "fcm_dpo/q_t": 0.4154285788536072, "grad_norm": 89.29653930664062, "learning_rate": 9.18886561011557e-08, "logits/chosen": -0.7836157083511353, "logits/rejected": -0.7811770439147949, "logps/chosen": -463.0835876464844, "logps/ref_chosen": -266.3705749511719, "logps/ref_rejected": -239.04490661621094, "logps/rejected": -525.2685546875, "loss": 4.5513, "margin_dpo/margin_mean": 89.51058197021484, "margin_dpo/margin_std": 143.0915985107422, "step": 357 }, { "epoch": 0.749738219895288, "fcm_dpo/beta": 0.004202042240649462, "fcm_dpo/delta": -0.018734265118837357, "fcm_dpo/margin": 99.36197662353516, "fcm_dpo/q_t": 0.40445786714553833, "grad_norm": 67.8469467163086, "learning_rate": 9.047478867791731e-08, "logits/chosen": -0.8765335083007812, "logits/rejected": -0.8540104627609253, "logps/chosen": -477.45489501953125, "logps/ref_chosen": -299.1474609375, "logps/ref_rejected": -257.2531433105469, "logps/rejected": -534.9225463867188, "loss": 4.4071, "margin_dpo/margin_mean": 99.36197662353516, "margin_dpo/margin_std": 142.21287536621094, "step": 358 }, { "epoch": 0.7518324607329843, "fcm_dpo/beta": 0.004150255583226681, "fcm_dpo/delta": -0.015420145355165005, "fcm_dpo/margin": 84.46359252929688, "fcm_dpo/q_t": 0.42003896832466125, "grad_norm": 105.28828430175781, "learning_rate": 8.906947610762825e-08, "logits/chosen": -0.8406848907470703, "logits/rejected": -0.8504139184951782, "logps/chosen": -489.37860107421875, "logps/ref_chosen": -302.99786376953125, "logps/ref_rejected": -260.4137268066406, "logps/rejected": -531.258056640625, "loss": 4.5874, "margin_dpo/margin_mean": 84.46359252929688, "margin_dpo/margin_std": 134.04608154296875, "step": 359 }, { "epoch": 0.7539267015706806, "fcm_dpo/beta": 0.004011007957160473, "fcm_dpo/delta": -0.024799039587378502, "fcm_dpo/margin": 78.62114715576172, "fcm_dpo/q_t": 0.4280991852283478, "grad_norm": 78.52674102783203, "learning_rate": 8.76727937529367e-08, "logits/chosen": -0.8800378441810608, "logits/rejected": -0.8672605752944946, "logps/chosen": -503.25250244140625, "logps/ref_chosen": -309.6114501953125, "logps/ref_rejected": -256.64031982421875, "logps/rejected": -528.9025268554688, "loss": 4.7802, "margin_dpo/margin_mean": 78.62113952636719, "margin_dpo/margin_std": 147.0281982421875, "step": 360 }, { "epoch": 0.7560209424083769, "fcm_dpo/beta": 0.004003939218819141, "fcm_dpo/delta": -0.013433972373604774, "fcm_dpo/margin": 103.06330871582031, "fcm_dpo/q_t": 0.4082551598548889, "grad_norm": 89.6968765258789, "learning_rate": 8.628481651367875e-08, "logits/chosen": -0.8125383853912354, "logits/rejected": -0.7936655282974243, "logps/chosen": -427.3260498046875, "logps/ref_chosen": -263.3797607421875, "logps/ref_rejected": -271.18157958984375, "logps/rejected": -538.191162109375, "loss": 4.4476, "margin_dpo/margin_mean": 103.06332397460938, "margin_dpo/margin_std": 158.7577362060547, "step": 361 }, { "epoch": 0.7581151832460733, "fcm_dpo/beta": 0.004012004937976599, "fcm_dpo/delta": 0.03470272570848465, "fcm_dpo/margin": 84.82909393310547, "fcm_dpo/q_t": 0.4218509793281555, "grad_norm": 105.21653747558594, "learning_rate": 8.490561882286135e-08, "logits/chosen": -0.8308958411216736, "logits/rejected": -0.8169477581977844, "logps/chosen": -480.77783203125, "logps/ref_chosen": -303.2583923339844, "logps/ref_rejected": -243.22891235351562, "logps/rejected": -505.57745361328125, "loss": 4.5377, "margin_dpo/margin_mean": 84.82909393310547, "margin_dpo/margin_std": 120.54029083251953, "step": 362 }, { "epoch": 0.7602094240837697, "fcm_dpo/beta": 0.004087929613888264, "fcm_dpo/delta": 0.005159515887498856, "fcm_dpo/margin": 88.3600082397461, "fcm_dpo/q_t": 0.419447124004364, "grad_norm": 70.80838012695312, "learning_rate": 8.353527464267104e-08, "logits/chosen": -0.8576685190200806, "logits/rejected": -0.8124222755432129, "logps/chosen": -484.9991760253906, "logps/ref_chosen": -303.34722900390625, "logps/ref_rejected": -262.05419921875, "logps/rejected": -532.066162109375, "loss": 4.6004, "margin_dpo/margin_mean": 88.36001586914062, "margin_dpo/margin_std": 147.58880615234375, "step": 363 }, { "epoch": 0.762303664921466, "fcm_dpo/beta": 0.0041374522261321545, "fcm_dpo/delta": 0.045890823006629944, "fcm_dpo/margin": 81.01228332519531, "fcm_dpo/q_t": 0.42662695050239563, "grad_norm": 78.44388580322266, "learning_rate": 8.217385746050742e-08, "logits/chosen": -0.829826831817627, "logits/rejected": -0.8344470858573914, "logps/chosen": -491.9500427246094, "logps/ref_chosen": -285.54376220703125, "logps/ref_rejected": -284.84619140625, "logps/rejected": -572.2648315429688, "loss": 4.727, "margin_dpo/margin_mean": 81.01229095458984, "margin_dpo/margin_std": 156.3982391357422, "step": 364 }, { "epoch": 0.7643979057591623, "fcm_dpo/beta": 0.004236162174493074, "fcm_dpo/delta": -0.0016637099906802177, "fcm_dpo/margin": 78.55353546142578, "fcm_dpo/q_t": 0.42524176836013794, "grad_norm": 84.3416748046875, "learning_rate": 8.082144028504231e-08, "logits/chosen": -0.8499488830566406, "logits/rejected": -0.8522325158119202, "logps/chosen": -469.8583984375, "logps/ref_chosen": -274.7878112792969, "logps/ref_rejected": -256.5738220214844, "logps/rejected": -530.1979370117188, "loss": 4.6929, "margin_dpo/margin_mean": 78.55352783203125, "margin_dpo/margin_std": 141.90133666992188, "step": 365 }, { "epoch": 0.7664921465968586, "fcm_dpo/beta": 0.0043019927106797695, "fcm_dpo/delta": 0.025872234255075455, "fcm_dpo/margin": 87.0827407836914, "fcm_dpo/q_t": 0.4149934649467468, "grad_norm": 73.31763458251953, "learning_rate": 7.947809564230445e-08, "logits/chosen": -0.802319347858429, "logits/rejected": -0.8103886842727661, "logps/chosen": -465.9633483886719, "logps/ref_chosen": -286.6496276855469, "logps/ref_rejected": -251.97140502929688, "logps/rejected": -518.367919921875, "loss": 4.548, "margin_dpo/margin_mean": 87.08274841308594, "margin_dpo/margin_std": 142.1220703125, "step": 366 }, { "epoch": 0.768586387434555, "fcm_dpo/beta": 0.004179948940873146, "fcm_dpo/delta": -0.04910705238580704, "fcm_dpo/margin": 94.82267761230469, "fcm_dpo/q_t": 0.41003602743148804, "grad_norm": 79.67829895019531, "learning_rate": 7.814389557179016e-08, "logits/chosen": -0.8363237977027893, "logits/rejected": -0.812557578086853, "logps/chosen": -481.86407470703125, "logps/ref_chosen": -301.9449768066406, "logps/ref_rejected": -265.5677185058594, "logps/rejected": -540.3095092773438, "loss": 4.4263, "margin_dpo/margin_mean": 94.82267761230469, "margin_dpo/margin_std": 134.9529266357422, "step": 367 }, { "epoch": 0.7706806282722513, "fcm_dpo/beta": 0.004044831730425358, "fcm_dpo/delta": -0.05517401173710823, "fcm_dpo/margin": 111.77228546142578, "fcm_dpo/q_t": 0.39542829990386963, "grad_norm": 81.13077545166016, "learning_rate": 7.681891162260015e-08, "logits/chosen": -0.8041467666625977, "logits/rejected": -0.8103511333465576, "logps/chosen": -464.62017822265625, "logps/ref_chosen": -294.62652587890625, "logps/ref_rejected": -258.7628479003906, "logps/rejected": -540.5286865234375, "loss": 4.2009, "margin_dpo/margin_mean": 111.77228546142578, "margin_dpo/margin_std": 126.04940795898438, "step": 368 }, { "epoch": 0.7727748691099476, "fcm_dpo/beta": 0.004040499217808247, "fcm_dpo/delta": 0.0448799803853035, "fcm_dpo/margin": 80.99122619628906, "fcm_dpo/q_t": 0.4241292178630829, "grad_norm": 75.39326477050781, "learning_rate": 7.550321484960251e-08, "logits/chosen": -0.8849146962165833, "logits/rejected": -0.8683555126190186, "logps/chosen": -463.06689453125, "logps/ref_chosen": -282.5057373046875, "logps/ref_rejected": -266.41607666015625, "logps/rejected": -527.968505859375, "loss": 4.5871, "margin_dpo/margin_mean": 80.9912338256836, "margin_dpo/margin_std": 121.13188171386719, "step": 369 }, { "epoch": 0.774869109947644, "fcm_dpo/beta": 0.004057829733937979, "fcm_dpo/delta": -0.008466588333249092, "fcm_dpo/margin": 100.35714721679688, "fcm_dpo/q_t": 0.4076663553714752, "grad_norm": 73.64390563964844, "learning_rate": 7.419687580962222e-08, "logits/chosen": -0.8662645816802979, "logits/rejected": -0.8904479146003723, "logps/chosen": -421.3550720214844, "logps/ref_chosen": -251.00640869140625, "logps/ref_rejected": -238.12542724609375, "logps/rejected": -508.8311767578125, "loss": 4.453, "margin_dpo/margin_mean": 100.3571548461914, "margin_dpo/margin_std": 152.1621856689453, "step": 370 }, { "epoch": 0.7769633507853403, "fcm_dpo/beta": 0.004208661150187254, "fcm_dpo/delta": 0.04080788791179657, "fcm_dpo/margin": 66.90896606445312, "fcm_dpo/q_t": 0.4354203939437866, "grad_norm": 99.27238464355469, "learning_rate": 7.289996455765748e-08, "logits/chosen": -0.8164863586425781, "logits/rejected": -0.8000683188438416, "logps/chosen": -491.5625915527344, "logps/ref_chosen": -296.6591491699219, "logps/ref_rejected": -251.14675903320312, "logps/rejected": -512.9591064453125, "loss": 4.8201, "margin_dpo/margin_mean": 66.90897369384766, "margin_dpo/margin_std": 131.07626342773438, "step": 371 }, { "epoch": 0.7790575916230367, "fcm_dpo/beta": 0.004100184887647629, "fcm_dpo/delta": -0.043359022587537766, "fcm_dpo/margin": 100.76567840576172, "fcm_dpo/q_t": 0.40735262632369995, "grad_norm": 83.19894409179688, "learning_rate": 7.161255064312283e-08, "logits/chosen": -0.7977643013000488, "logits/rejected": -0.7871427536010742, "logps/chosen": -520.5551147460938, "logps/ref_chosen": -331.3714599609375, "logps/ref_rejected": -285.56805419921875, "logps/rejected": -575.5173950195312, "loss": 4.449, "margin_dpo/margin_mean": 100.76567840576172, "margin_dpo/margin_std": 149.07765197753906, "step": 372 }, { "epoch": 0.7811518324607329, "fcm_dpo/beta": 0.004079765640199184, "fcm_dpo/delta": 0.007463835179805756, "fcm_dpo/margin": 90.10908508300781, "fcm_dpo/q_t": 0.41459009051322937, "grad_norm": 77.88961791992188, "learning_rate": 7.033470310611945e-08, "logits/chosen": -0.8931885361671448, "logits/rejected": -0.8614373207092285, "logps/chosen": -496.9786376953125, "logps/ref_chosen": -321.9429931640625, "logps/ref_rejected": -271.2288513183594, "logps/rejected": -536.37353515625, "loss": 4.4999, "margin_dpo/margin_mean": 90.10908508300781, "margin_dpo/margin_std": 128.7992401123047, "step": 373 }, { "epoch": 0.7832460732984293, "fcm_dpo/beta": 0.0040543461218476295, "fcm_dpo/delta": -0.009488560259342194, "fcm_dpo/margin": 81.82614135742188, "fcm_dpo/q_t": 0.4247649908065796, "grad_norm": 60.28491973876953, "learning_rate": 6.906649047373245e-08, "logits/chosen": -0.8894190192222595, "logits/rejected": -0.8857640027999878, "logps/chosen": -503.9471435546875, "logps/ref_chosen": -319.1685485839844, "logps/ref_rejected": -284.6263732910156, "logps/rejected": -551.2310791015625, "loss": 4.6405, "margin_dpo/margin_mean": 81.82613372802734, "margin_dpo/margin_std": 135.69937133789062, "step": 374 }, { "epoch": 0.7853403141361257, "fcm_dpo/beta": 0.0040982929058372974, "fcm_dpo/delta": 0.019265731796622276, "fcm_dpo/margin": 69.89092254638672, "fcm_dpo/q_t": 0.4342796802520752, "grad_norm": 83.22428894042969, "learning_rate": 6.780798075635675e-08, "logits/chosen": -0.8811032176017761, "logits/rejected": -0.860395073890686, "logps/chosen": -508.0401611328125, "logps/ref_chosen": -314.87579345703125, "logps/ref_rejected": -259.1965026855469, "logps/rejected": -522.2518310546875, "loss": 4.7725, "margin_dpo/margin_mean": 69.89092254638672, "margin_dpo/margin_std": 129.87033081054688, "step": 375 }, { "epoch": 0.787434554973822, "fcm_dpo/beta": 0.0042695943266153336, "fcm_dpo/delta": 0.045676298439502716, "fcm_dpo/margin": 83.08172607421875, "fcm_dpo/q_t": 0.42078420519828796, "grad_norm": 97.34649658203125, "learning_rate": 6.655924144404906e-08, "logits/chosen": -0.8564417958259583, "logits/rejected": -0.8596733212471008, "logps/chosen": -483.13177490234375, "logps/ref_chosen": -287.6732482910156, "logps/ref_rejected": -256.6697082519531, "logps/rejected": -535.2098999023438, "loss": 4.6391, "margin_dpo/margin_mean": 83.08173370361328, "margin_dpo/margin_std": 145.4075164794922, "step": 376 }, { "epoch": 0.7895287958115184, "fcm_dpo/beta": 0.004419329576194286, "fcm_dpo/delta": 0.04526427388191223, "fcm_dpo/margin": 62.52194595336914, "fcm_dpo/q_t": 0.43734949827194214, "grad_norm": 94.04608154296875, "learning_rate": 6.532033950290885e-08, "logits/chosen": -0.8489782214164734, "logits/rejected": -0.846108615398407, "logps/chosen": -511.7943115234375, "logps/ref_chosen": -305.261474609375, "logps/ref_rejected": -271.8887023925781, "logps/rejected": -540.9434814453125, "loss": 4.9284, "margin_dpo/margin_mean": 62.52194595336914, "margin_dpo/margin_std": 146.0769500732422, "step": 377 }, { "epoch": 0.7916230366492146, "fcm_dpo/beta": 0.004514284431934357, "fcm_dpo/delta": 0.037094950675964355, "fcm_dpo/margin": 74.94515991210938, "fcm_dpo/q_t": 0.4244396984577179, "grad_norm": 101.25901794433594, "learning_rate": 6.409134137148736e-08, "logits/chosen": -0.8392292857170105, "logits/rejected": -0.82718425989151, "logps/chosen": -479.11871337890625, "logps/ref_chosen": -281.5295715332031, "logps/ref_rejected": -296.980224609375, "logps/rejected": -569.5145874023438, "loss": 4.6926, "margin_dpo/margin_mean": 74.94515991210938, "margin_dpo/margin_std": 139.2892608642578, "step": 378 }, { "epoch": 0.793717277486911, "fcm_dpo/beta": 0.004631263203918934, "fcm_dpo/delta": 0.02148812636733055, "fcm_dpo/margin": 75.75040435791016, "fcm_dpo/q_t": 0.4205975830554962, "grad_norm": 86.93419647216797, "learning_rate": 6.28723129572247e-08, "logits/chosen": -0.9042958617210388, "logits/rejected": -0.8856334686279297, "logps/chosen": -448.8710632324219, "logps/ref_chosen": -265.0807800292969, "logps/ref_rejected": -230.58932495117188, "logps/rejected": -490.1300048828125, "loss": 4.6479, "margin_dpo/margin_mean": 75.75040435791016, "margin_dpo/margin_std": 135.121337890625, "step": 379 }, { "epoch": 0.7958115183246073, "fcm_dpo/beta": 0.004544341005384922, "fcm_dpo/delta": -0.061347946524620056, "fcm_dpo/margin": 84.75240325927734, "fcm_dpo/q_t": 0.41487401723861694, "grad_norm": 89.89930725097656, "learning_rate": 6.166331963291519e-08, "logits/chosen": -0.8618828058242798, "logits/rejected": -0.839131772518158, "logps/chosen": -501.7915954589844, "logps/ref_chosen": -305.90838623046875, "logps/ref_rejected": -286.5906677246094, "logps/rejected": -567.226318359375, "loss": 4.5946, "margin_dpo/margin_mean": 84.75240325927734, "margin_dpo/margin_std": 142.7577667236328, "step": 380 }, { "epoch": 0.7979057591623037, "fcm_dpo/beta": 0.00437317555770278, "fcm_dpo/delta": -0.05076461285352707, "fcm_dpo/margin": 95.79244995117188, "fcm_dpo/q_t": 0.4083157181739807, "grad_norm": 88.44004821777344, "learning_rate": 6.046442623320145e-08, "logits/chosen": -0.8316866159439087, "logits/rejected": -0.7976770997047424, "logps/chosen": -444.54351806640625, "logps/ref_chosen": -252.87066650390625, "logps/ref_rejected": -261.1927490234375, "logps/rejected": -548.6580200195312, "loss": 4.461, "margin_dpo/margin_mean": 95.79244995117188, "margin_dpo/margin_std": 149.4072723388672, "step": 381 }, { "epoch": 0.8, "fcm_dpo/beta": 0.004201515112072229, "fcm_dpo/delta": -0.05036533996462822, "fcm_dpo/margin": 106.53213500976562, "fcm_dpo/q_t": 0.39790698885917664, "grad_norm": 85.8750991821289, "learning_rate": 5.9275697051098275e-08, "logits/chosen": -0.8765906095504761, "logits/rejected": -0.8718726634979248, "logps/chosen": -472.893310546875, "logps/ref_chosen": -289.2114562988281, "logps/ref_rejected": -278.45751953125, "logps/rejected": -568.6714477539062, "loss": 4.2905, "margin_dpo/margin_mean": 106.53213500976562, "margin_dpo/margin_std": 138.6427001953125, "step": 382 }, { "epoch": 0.8020942408376963, "fcm_dpo/beta": 0.0041129072196781635, "fcm_dpo/delta": -0.022012613713741302, "fcm_dpo/margin": 95.4480972290039, "fcm_dpo/q_t": 0.4109508991241455, "grad_norm": 95.30079650878906, "learning_rate": 5.809719583454414e-08, "logits/chosen": -0.8609268665313721, "logits/rejected": -0.8443728685379028, "logps/chosen": -453.16119384765625, "logps/ref_chosen": -273.630859375, "logps/ref_rejected": -261.44024658203125, "logps/rejected": -536.418701171875, "loss": 4.4769, "margin_dpo/margin_mean": 95.4480972290039, "margin_dpo/margin_std": 144.06008911132812, "step": 383 }, { "epoch": 0.8041884816753927, "fcm_dpo/beta": 0.004073744174093008, "fcm_dpo/delta": 0.0247461199760437, "fcm_dpo/margin": 83.81874084472656, "fcm_dpo/q_t": 0.4218423664569855, "grad_norm": 79.8589859008789, "learning_rate": 5.6928985782982524e-08, "logits/chosen": -0.8856214880943298, "logits/rejected": -0.8819814324378967, "logps/chosen": -463.6858825683594, "logps/ref_chosen": -274.5699462890625, "logps/ref_rejected": -285.8253479003906, "logps/rejected": -558.7599487304688, "loss": 4.5903, "margin_dpo/margin_mean": 83.81874084472656, "margin_dpo/margin_std": 133.73779296875, "step": 384 }, { "epoch": 0.806282722513089, "fcm_dpo/beta": 0.00418306328356266, "fcm_dpo/delta": 0.019975785166025162, "fcm_dpo/margin": 74.35530853271484, "fcm_dpo/q_t": 0.42889341711997986, "grad_norm": 95.49700164794922, "learning_rate": 5.57711295439732e-08, "logits/chosen": -0.8372036218643188, "logits/rejected": -0.8347649574279785, "logps/chosen": -479.953857421875, "logps/ref_chosen": -284.150634765625, "logps/ref_rejected": -244.87921142578125, "logps/rejected": -515.037841796875, "loss": 4.6754, "margin_dpo/margin_mean": 74.35530853271484, "margin_dpo/margin_std": 125.24661254882812, "step": 385 }, { "epoch": 0.8083769633507853, "fcm_dpo/beta": 0.003999017644673586, "fcm_dpo/delta": -0.0784250870347023, "fcm_dpo/margin": 106.15267944335938, "fcm_dpo/q_t": 0.403393030166626, "grad_norm": 86.98873138427734, "learning_rate": 5.4623689209832484e-08, "logits/chosen": -0.8149560689926147, "logits/rejected": -0.813643217086792, "logps/chosen": -503.624755859375, "logps/ref_chosen": -320.1762390136719, "logps/ref_rejected": -302.05023193359375, "logps/rejected": -591.6513671875, "loss": 4.3475, "margin_dpo/margin_mean": 106.15266418457031, "margin_dpo/margin_std": 134.6205596923828, "step": 386 }, { "epoch": 0.8104712041884817, "fcm_dpo/beta": 0.0039867255836725235, "fcm_dpo/delta": 0.02052573300898075, "fcm_dpo/margin": 82.58099365234375, "fcm_dpo/q_t": 0.4234527051448822, "grad_norm": 85.00679016113281, "learning_rate": 5.3486726314303175e-08, "logits/chosen": -0.8490534424781799, "logits/rejected": -0.8536210656166077, "logps/chosen": -472.10137939453125, "logps/ref_chosen": -272.2801513671875, "logps/ref_rejected": -265.1615905761719, "logps/rejected": -547.5638427734375, "loss": 4.6743, "margin_dpo/margin_mean": 82.58099365234375, "margin_dpo/margin_std": 141.04251098632812, "step": 387 }, { "epoch": 0.812565445026178, "fcm_dpo/beta": 0.00392739474773407, "fcm_dpo/delta": -0.014465522021055222, "fcm_dpo/margin": 58.15325927734375, "fcm_dpo/q_t": 0.4480987787246704, "grad_norm": 101.83181762695312, "learning_rate": 5.2360301829254745e-08, "logits/chosen": -0.8298582434654236, "logits/rejected": -0.8175155520439148, "logps/chosen": -483.71356201171875, "logps/ref_chosen": -272.5313415527344, "logps/ref_rejected": -239.55735778808594, "logps/rejected": -508.89288330078125, "loss": 5.0353, "margin_dpo/margin_mean": 58.153263092041016, "margin_dpo/margin_std": 146.24984741210938, "step": 388 }, { "epoch": 0.8146596858638744, "fcm_dpo/beta": 0.003913066349923611, "fcm_dpo/delta": -0.007896999828517437, "fcm_dpo/margin": 82.79248809814453, "fcm_dpo/q_t": 0.42776596546173096, "grad_norm": 78.87464141845703, "learning_rate": 5.1244476161413806e-08, "logits/chosen": -0.8589950203895569, "logits/rejected": -0.8571079969406128, "logps/chosen": -489.0984802246094, "logps/ref_chosen": -281.0892639160156, "logps/ref_rejected": -246.50045776367188, "logps/rejected": -537.3021240234375, "loss": 4.7148, "margin_dpo/margin_mean": 82.79248809814453, "margin_dpo/margin_std": 149.3322296142578, "step": 389 }, { "epoch": 0.8167539267015707, "fcm_dpo/beta": 0.003945834934711456, "fcm_dpo/delta": -0.002285042777657509, "fcm_dpo/margin": 89.30833435058594, "fcm_dpo/q_t": 0.41917407512664795, "grad_norm": 66.9520034790039, "learning_rate": 5.013930914912476e-08, "logits/chosen": -0.8890639543533325, "logits/rejected": -0.8983543515205383, "logps/chosen": -483.90277099609375, "logps/ref_chosen": -283.98748779296875, "logps/ref_rejected": -283.465087890625, "logps/rejected": -572.688720703125, "loss": 4.5633, "margin_dpo/margin_mean": 89.3083267211914, "margin_dpo/margin_std": 138.32273864746094, "step": 390 }, { "epoch": 0.818848167539267, "fcm_dpo/beta": 0.0038222242146730423, "fcm_dpo/delta": -0.038259271532297134, "fcm_dpo/margin": 95.59449768066406, "fcm_dpo/q_t": 0.4170438051223755, "grad_norm": 66.74435424804688, "learning_rate": 4.904486005914027e-08, "logits/chosen": -0.8340939879417419, "logits/rejected": -0.81984543800354, "logps/chosen": -498.55877685546875, "logps/ref_chosen": -283.86138916015625, "logps/ref_rejected": -263.5093688964844, "logps/rejected": -573.8013305664062, "loss": 4.5437, "margin_dpo/margin_mean": 95.59449768066406, "margin_dpo/margin_std": 147.83688354492188, "step": 391 }, { "epoch": 0.8209424083769633, "fcm_dpo/beta": 0.0036374391056597233, "fcm_dpo/delta": -0.05550744757056236, "fcm_dpo/margin": 108.41232299804688, "fcm_dpo/q_t": 0.4071521759033203, "grad_norm": 60.498085021972656, "learning_rate": 4.796118758344353e-08, "logits/chosen": -0.8120825886726379, "logits/rejected": -0.8258223533630371, "logps/chosen": -504.80535888671875, "logps/ref_chosen": -310.070068359375, "logps/ref_rejected": -252.89817810058594, "logps/rejected": -556.0457153320312, "loss": 4.3798, "margin_dpo/margin_mean": 108.41232299804688, "margin_dpo/margin_std": 132.855712890625, "step": 392 }, { "epoch": 0.8230366492146597, "fcm_dpo/beta": 0.003737508552148938, "fcm_dpo/delta": 0.06319691240787506, "fcm_dpo/margin": 80.97676849365234, "fcm_dpo/q_t": 0.43021535873413086, "grad_norm": 58.454811096191406, "learning_rate": 4.688834983610082e-08, "logits/chosen": -0.8596187829971313, "logits/rejected": -0.8425779342651367, "logps/chosen": -482.13433837890625, "logps/ref_chosen": -286.7156677246094, "logps/ref_rejected": -230.00357055664062, "logps/rejected": -506.39910888671875, "loss": 4.6788, "margin_dpo/margin_mean": 80.97676849365234, "margin_dpo/margin_std": 135.61444091796875, "step": 393 }, { "epoch": 0.8251308900523561, "fcm_dpo/beta": 0.0038439815398305655, "fcm_dpo/delta": 0.020402822643518448, "fcm_dpo/margin": 80.80021667480469, "fcm_dpo/q_t": 0.4315696358680725, "grad_norm": 80.42355346679688, "learning_rate": 4.582640435014459e-08, "logits/chosen": -0.8997487425804138, "logits/rejected": -0.8925558924674988, "logps/chosen": -524.9955444335938, "logps/ref_chosen": -325.9934387207031, "logps/ref_rejected": -317.42706298828125, "logps/rejected": -597.2293701171875, "loss": 4.7637, "margin_dpo/margin_mean": 80.80021667480469, "margin_dpo/margin_std": 155.62696838378906, "step": 394 }, { "epoch": 0.8272251308900523, "fcm_dpo/beta": 0.0038858535699546337, "fcm_dpo/delta": -0.007227412424981594, "fcm_dpo/margin": 104.61137390136719, "fcm_dpo/q_t": 0.41017356514930725, "grad_norm": 86.83416748046875, "learning_rate": 4.477540807448832e-08, "logits/chosen": -0.8377059698104858, "logits/rejected": -0.8466890454292297, "logps/chosen": -462.9100036621094, "logps/ref_chosen": -268.90081787109375, "logps/ref_rejected": -272.85809326171875, "logps/rejected": -571.4786376953125, "loss": 4.4153, "margin_dpo/margin_mean": 104.61136627197266, "margin_dpo/margin_std": 150.77389526367188, "step": 395 }, { "epoch": 0.8293193717277487, "fcm_dpo/beta": 0.0039118435233831406, "fcm_dpo/delta": 0.03104759007692337, "fcm_dpo/margin": 82.23450469970703, "fcm_dpo/q_t": 0.42661815881729126, "grad_norm": 90.65218353271484, "learning_rate": 4.373541737087263e-08, "logits/chosen": -0.8444753885269165, "logits/rejected": -0.8268455862998962, "logps/chosen": -496.99774169921875, "logps/ref_chosen": -291.19830322265625, "logps/ref_rejected": -253.2803955078125, "logps/rejected": -541.3143310546875, "loss": 4.6647, "margin_dpo/margin_mean": 82.2344970703125, "margin_dpo/margin_std": 138.35423278808594, "step": 396 }, { "epoch": 0.831413612565445, "fcm_dpo/beta": 0.0039418223313987255, "fcm_dpo/delta": -0.02278411015868187, "fcm_dpo/margin": 79.29293823242188, "fcm_dpo/q_t": 0.4295212924480438, "grad_norm": 88.5928726196289, "learning_rate": 4.270648801084295e-08, "logits/chosen": -0.8564522862434387, "logits/rejected": -0.8358593583106995, "logps/chosen": -506.43353271484375, "logps/ref_chosen": -309.8224182128906, "logps/ref_rejected": -291.9057922363281, "logps/rejected": -567.809814453125, "loss": 4.7629, "margin_dpo/margin_mean": 79.29293823242188, "margin_dpo/margin_std": 146.21200561523438, "step": 397 }, { "epoch": 0.8335078534031414, "fcm_dpo/beta": 0.0038612037897109985, "fcm_dpo/delta": 0.01422278955578804, "fcm_dpo/margin": 67.4949722290039, "fcm_dpo/q_t": 0.4390268921852112, "grad_norm": 99.15848541259766, "learning_rate": 4.168867517275806e-08, "logits/chosen": -0.76848304271698, "logits/rejected": -0.8057135343551636, "logps/chosen": -519.5203247070312, "logps/ref_chosen": -297.8135070800781, "logps/ref_rejected": -270.5025634765625, "logps/rejected": -559.704345703125, "loss": 5.0489, "margin_dpo/margin_mean": 67.4949722290039, "margin_dpo/margin_std": 172.52947998046875, "step": 398 }, { "epoch": 0.8356020942408376, "fcm_dpo/beta": 0.0039389231242239475, "fcm_dpo/delta": 0.028249990195035934, "fcm_dpo/margin": 84.23013305664062, "fcm_dpo/q_t": 0.4258180558681488, "grad_norm": 118.06265258789062, "learning_rate": 4.0682033438831584e-08, "logits/chosen": -0.8657823204994202, "logits/rejected": -0.8276122808456421, "logps/chosen": -506.72381591796875, "logps/ref_chosen": -292.8467712402344, "logps/ref_rejected": -268.3638916015625, "logps/rejected": -566.4711303710938, "loss": 4.6836, "margin_dpo/margin_mean": 84.23013305664062, "margin_dpo/margin_std": 150.7490234375, "step": 399 }, { "epoch": 0.837696335078534, "fcm_dpo/beta": 0.0039605312049388885, "fcm_dpo/delta": -0.0016172388568520546, "fcm_dpo/margin": 82.38078308105469, "fcm_dpo/q_t": 0.4250349998474121, "grad_norm": 81.15113830566406, "learning_rate": 3.968661679220467e-08, "logits/chosen": -0.9057132601737976, "logits/rejected": -0.9061620831489563, "logps/chosen": -469.662841796875, "logps/ref_chosen": -263.6763916015625, "logps/ref_rejected": -258.67266845703125, "logps/rejected": -547.0399169921875, "loss": 4.6885, "margin_dpo/margin_mean": 82.38077545166016, "margin_dpo/margin_std": 143.50645446777344, "step": 400 }, { "epoch": 0.837696335078534, "eval_fcm_dpo/beta": 0.004001657944172621, "eval_logits/chosen": -0.8691577911376953, "eval_logits/rejected": -0.8528784513473511, "eval_logps/chosen": -496.1020812988281, "eval_logps/ref_chosen": -287.8268127441406, "eval_logps/ref_rejected": -266.9300231933594, "eval_logps/rejected": -563.8033447265625, "eval_loss": 0.5783530473709106, "eval_margin_dpo/margin_mean": 88.59801483154297, "eval_margin_dpo/margin_std": 143.06631469726562, "eval_runtime": 78.6395, "eval_samples_per_second": 25.433, "eval_steps_per_second": 3.179, "step": 400 }, { "epoch": 0.8397905759162304, "fcm_dpo/beta": 0.003983091097325087, "fcm_dpo/delta": -0.011052620597183704, "fcm_dpo/margin": 89.36437225341797, "fcm_dpo/q_t": 0.419382244348526, "grad_norm": 93.5476303100586, "learning_rate": 3.8702478614051345e-08, "logits/chosen": -0.8399940729141235, "logits/rejected": -0.8391348719596863, "logps/chosen": -523.5870361328125, "logps/ref_chosen": -318.2853088378906, "logps/ref_rejected": -293.75225830078125, "logps/rejected": -588.4183349609375, "loss": 4.5995, "margin_dpo/margin_mean": 89.36437225341797, "margin_dpo/margin_std": 148.24929809570312, "step": 401 }, { "epoch": 0.8418848167539267, "fcm_dpo/beta": 0.004000155255198479, "fcm_dpo/delta": 0.028147714212536812, "fcm_dpo/margin": 87.35748291015625, "fcm_dpo/q_t": 0.4228289723396301, "grad_norm": 73.04254150390625, "learning_rate": 3.772967168071517e-08, "logits/chosen": -0.8962114453315735, "logits/rejected": -0.8734457492828369, "logps/chosen": -506.75689697265625, "logps/ref_chosen": -309.4278564453125, "logps/ref_rejected": -282.0279846191406, "logps/rejected": -566.7145385742188, "loss": 4.6107, "margin_dpo/margin_mean": 87.35748291015625, "margin_dpo/margin_std": 151.18124389648438, "step": 402 }, { "epoch": 0.8439790575916231, "fcm_dpo/beta": 0.003969723358750343, "fcm_dpo/delta": -0.08291508257389069, "fcm_dpo/margin": 120.73661804199219, "fcm_dpo/q_t": 0.39133724570274353, "grad_norm": 61.04887390136719, "learning_rate": 3.676824816087978e-08, "logits/chosen": -0.8727239966392517, "logits/rejected": -0.8497612476348877, "logps/chosen": -505.31695556640625, "logps/ref_chosen": -309.0284729003906, "logps/ref_rejected": -272.9622497558594, "logps/rejected": -589.9873046875, "loss": 4.1603, "margin_dpo/margin_mean": 120.73661804199219, "margin_dpo/margin_std": 138.56192016601562, "step": 403 }, { "epoch": 0.8460732984293193, "fcm_dpo/beta": 0.0038393645081669092, "fcm_dpo/delta": 0.025593645870685577, "fcm_dpo/margin": 90.4571533203125, "fcm_dpo/q_t": 0.42148569226264954, "grad_norm": 75.76798248291016, "learning_rate": 3.581825961277074e-08, "logits/chosen": -0.9005187749862671, "logits/rejected": -0.876252293586731, "logps/chosen": -509.41229248046875, "logps/ref_chosen": -297.2837219238281, "logps/ref_rejected": -256.99041748046875, "logps/rejected": -559.5761108398438, "loss": 4.6459, "margin_dpo/margin_mean": 90.45716094970703, "margin_dpo/margin_std": 158.7395477294922, "step": 404 }, { "epoch": 0.8481675392670157, "fcm_dpo/beta": 0.003894682740792632, "fcm_dpo/delta": 0.0062955510802567005, "fcm_dpo/margin": 95.108154296875, "fcm_dpo/q_t": 0.41595664620399475, "grad_norm": 55.920188903808594, "learning_rate": 3.487975698139084e-08, "logits/chosen": -0.8308718800544739, "logits/rejected": -0.839844822883606, "logps/chosen": -462.46856689453125, "logps/ref_chosen": -257.96533203125, "logps/ref_rejected": -255.811279296875, "logps/rejected": -555.4227294921875, "loss": 4.5156, "margin_dpo/margin_mean": 95.108154296875, "margin_dpo/margin_std": 146.5327911376953, "step": 405 }, { "epoch": 0.8502617801047121, "fcm_dpo/beta": 0.003938999027013779, "fcm_dpo/delta": 0.024059396237134933, "fcm_dpo/margin": 70.01969909667969, "fcm_dpo/q_t": 0.43556931614875793, "grad_norm": 102.48121643066406, "learning_rate": 3.3952790595787986e-08, "logits/chosen": -0.8572216033935547, "logits/rejected": -0.8341256380081177, "logps/chosen": -500.6681823730469, "logps/ref_chosen": -285.1810607910156, "logps/ref_rejected": -264.41351318359375, "logps/rejected": -549.9203491210938, "loss": 4.7904, "margin_dpo/margin_mean": 70.01969909667969, "margin_dpo/margin_std": 130.00796508789062, "step": 406 }, { "epoch": 0.8523560209424084, "fcm_dpo/beta": 0.003916039131581783, "fcm_dpo/delta": -0.02865220047533512, "fcm_dpo/margin": 95.51104736328125, "fcm_dpo/q_t": 0.4164567291736603, "grad_norm": 60.025779724121094, "learning_rate": 3.303741016635614e-08, "logits/chosen": -0.8459261655807495, "logits/rejected": -0.8647856712341309, "logps/chosen": -484.513427734375, "logps/ref_chosen": -265.23809814453125, "logps/ref_rejected": -219.0631561279297, "logps/rejected": -533.849609375, "loss": 4.563, "margin_dpo/margin_mean": 95.51104736328125, "margin_dpo/margin_std": 157.53103637695312, "step": 407 }, { "epoch": 0.8544502617801047, "fcm_dpo/beta": 0.003969256300479174, "fcm_dpo/delta": -0.00904204323887825, "fcm_dpo/margin": 96.21162414550781, "fcm_dpo/q_t": 0.41344723105430603, "grad_norm": 95.24205017089844, "learning_rate": 3.2133664782169944e-08, "logits/chosen": -0.8817875981330872, "logits/rejected": -0.8740147352218628, "logps/chosen": -494.42425537109375, "logps/ref_chosen": -296.9726257324219, "logps/ref_rejected": -295.4786376953125, "logps/rejected": -589.1419067382812, "loss": 4.5008, "margin_dpo/margin_mean": 96.21162414550781, "margin_dpo/margin_std": 141.05984497070312, "step": 408 }, { "epoch": 0.856544502617801, "fcm_dpo/beta": 0.0038298959843814373, "fcm_dpo/delta": -0.018867127597332, "fcm_dpo/margin": 99.35574340820312, "fcm_dpo/q_t": 0.4180399775505066, "grad_norm": 59.48664093017578, "learning_rate": 3.12416029083514e-08, "logits/chosen": -0.8571524024009705, "logits/rejected": -0.8414179086685181, "logps/chosen": -496.5021057128906, "logps/ref_chosen": -287.37933349609375, "logps/ref_rejected": -275.80291748046875, "logps/rejected": -584.2814331054688, "loss": 4.6107, "margin_dpo/margin_mean": 99.35574340820312, "margin_dpo/margin_std": 180.15744018554688, "step": 409 }, { "epoch": 0.8586387434554974, "fcm_dpo/beta": 0.003783642081543803, "fcm_dpo/delta": 0.0071961102075874805, "fcm_dpo/margin": 87.63327026367188, "fcm_dpo/q_t": 0.4247705340385437, "grad_norm": 72.14511108398438, "learning_rate": 3.036127238347164e-08, "logits/chosen": -0.8766495585441589, "logits/rejected": -0.8835284113883972, "logps/chosen": -481.9991455078125, "logps/ref_chosen": -281.7801818847656, "logps/ref_rejected": -266.7550354003906, "logps/rejected": -554.6072998046875, "loss": 4.6469, "margin_dpo/margin_mean": 87.63327026367188, "margin_dpo/margin_std": 149.70281982421875, "step": 410 }, { "epoch": 0.8607329842931937, "fcm_dpo/beta": 0.00382496346719563, "fcm_dpo/delta": -0.014396967366337776, "fcm_dpo/margin": 101.3914794921875, "fcm_dpo/q_t": 0.41269832849502563, "grad_norm": 72.0369644165039, "learning_rate": 2.9492720416985e-08, "logits/chosen": -0.8845562934875488, "logits/rejected": -0.8408842086791992, "logps/chosen": -481.53240966796875, "logps/ref_chosen": -281.5872497558594, "logps/ref_rejected": -254.78916931152344, "logps/rejected": -556.1257934570312, "loss": 4.4243, "margin_dpo/margin_mean": 101.3914794921875, "margin_dpo/margin_std": 137.44744873046875, "step": 411 }, { "epoch": 0.86282722513089, "fcm_dpo/beta": 0.003669111290946603, "fcm_dpo/delta": -0.013924513012170792, "fcm_dpo/margin": 79.34294128417969, "fcm_dpo/q_t": 0.43426990509033203, "grad_norm": 120.3460922241211, "learning_rate": 2.863599358669755e-08, "logits/chosen": -0.8453624248504639, "logits/rejected": -0.85740065574646, "logps/chosen": -497.730224609375, "logps/ref_chosen": -276.5341796875, "logps/ref_rejected": -273.8751220703125, "logps/rejected": -574.4141235351562, "loss": 4.7748, "margin_dpo/margin_mean": 79.34294128417969, "margin_dpo/margin_std": 146.27236938476562, "step": 412 }, { "epoch": 0.8649214659685864, "fcm_dpo/beta": 0.0038752357941120863, "fcm_dpo/delta": 0.08048185706138611, "fcm_dpo/margin": 76.79857635498047, "fcm_dpo/q_t": 0.43248826265335083, "grad_norm": 86.02536010742188, "learning_rate": 2.7791137836269158e-08, "logits/chosen": -0.8663272261619568, "logits/rejected": -0.8638010621070862, "logps/chosen": -479.69171142578125, "logps/ref_chosen": -271.2745666503906, "logps/ref_rejected": -270.16912841796875, "logps/rejected": -555.3848876953125, "loss": 4.7365, "margin_dpo/margin_mean": 76.79857635498047, "margin_dpo/margin_std": 141.86328125, "step": 413 }, { "epoch": 0.8670157068062827, "fcm_dpo/beta": 0.004013517405837774, "fcm_dpo/delta": -0.019187791272997856, "fcm_dpo/margin": 97.65512084960938, "fcm_dpo/q_t": 0.412454217672348, "grad_norm": 71.61355590820312, "learning_rate": 2.6958198472749717e-08, "logits/chosen": -0.9010047316551208, "logits/rejected": -0.8984125852584839, "logps/chosen": -504.4794616699219, "logps/ref_chosen": -297.11505126953125, "logps/ref_rejected": -271.7034606933594, "logps/rejected": -576.7229614257812, "loss": 4.4714, "margin_dpo/margin_mean": 97.65512084960938, "margin_dpo/margin_std": 143.94935607910156, "step": 414 }, { "epoch": 0.8691099476439791, "fcm_dpo/beta": 0.003955254331231117, "fcm_dpo/delta": 0.011070379987359047, "fcm_dpo/margin": 91.44115447998047, "fcm_dpo/q_t": 0.4168679714202881, "grad_norm": 71.55652618408203, "learning_rate": 2.613722016414943e-08, "logits/chosen": -0.8896200060844421, "logits/rejected": -0.875502347946167, "logps/chosen": -498.7607116699219, "logps/ref_chosen": -297.6926574707031, "logps/ref_rejected": -279.0503234863281, "logps/rejected": -571.5595703125, "loss": 4.5043, "margin_dpo/margin_mean": 91.44115447998047, "margin_dpo/margin_std": 134.12338256835938, "step": 415 }, { "epoch": 0.8712041884816754, "fcm_dpo/beta": 0.0038398101460188627, "fcm_dpo/delta": -0.04415631666779518, "fcm_dpo/margin": 108.07146453857422, "fcm_dpo/q_t": 0.404560923576355, "grad_norm": 62.76100158691406, "learning_rate": 2.5328246937043525e-08, "logits/chosen": -0.8924229145050049, "logits/rejected": -0.8925044536590576, "logps/chosen": -505.1080627441406, "logps/ref_chosen": -311.8255615234375, "logps/ref_rejected": -268.6170654296875, "logps/rejected": -569.9710693359375, "loss": 4.3709, "margin_dpo/margin_mean": 108.07145690917969, "margin_dpo/margin_std": 146.09751892089844, "step": 416 }, { "epoch": 0.8732984293193717, "fcm_dpo/beta": 0.0037508481182157993, "fcm_dpo/delta": -0.03793646767735481, "fcm_dpo/margin": 87.92816162109375, "fcm_dpo/q_t": 0.42599597573280334, "grad_norm": 88.90306091308594, "learning_rate": 2.4531322174210973e-08, "logits/chosen": -0.8363179564476013, "logits/rejected": -0.8270238041877747, "logps/chosen": -521.0562744140625, "logps/ref_chosen": -310.43682861328125, "logps/ref_rejected": -277.15283203125, "logps/rejected": -575.700439453125, "loss": 4.67, "margin_dpo/margin_mean": 87.92816162109375, "margin_dpo/margin_std": 148.5767822265625, "step": 417 }, { "epoch": 0.875392670157068, "fcm_dpo/beta": 0.003646118799224496, "fcm_dpo/delta": -0.036676209419965744, "fcm_dpo/margin": 90.21328735351562, "fcm_dpo/q_t": 0.425502210855484, "grad_norm": 65.71238708496094, "learning_rate": 2.3746488612308295e-08, "logits/chosen": -0.8376297354698181, "logits/rejected": -0.8152227401733398, "logps/chosen": -498.4665222167969, "logps/ref_chosen": -278.49591064453125, "logps/ref_rejected": -276.56671142578125, "logps/rejected": -586.7506103515625, "loss": 4.674, "margin_dpo/margin_mean": 90.21328735351562, "margin_dpo/margin_std": 156.35293579101562, "step": 418 }, { "epoch": 0.8774869109947644, "fcm_dpo/beta": 0.003632680047303438, "fcm_dpo/delta": 0.0225251205265522, "fcm_dpo/margin": 93.54913330078125, "fcm_dpo/q_t": 0.4209168255329132, "grad_norm": 73.43122100830078, "learning_rate": 2.297378833957761e-08, "logits/chosen": -0.879563570022583, "logits/rejected": -0.8511139750480652, "logps/chosen": -518.3619384765625, "logps/ref_chosen": -298.9002380371094, "logps/ref_rejected": -246.1540985107422, "logps/rejected": -559.1648559570312, "loss": 4.6436, "margin_dpo/margin_mean": 93.54913330078125, "margin_dpo/margin_std": 161.06155395507812, "step": 419 }, { "epoch": 0.8795811518324608, "fcm_dpo/beta": 0.0036652141716331244, "fcm_dpo/delta": 0.0008689370006322861, "fcm_dpo/margin": 108.68309020996094, "fcm_dpo/q_t": 0.4112725853919983, "grad_norm": 93.50495147705078, "learning_rate": 2.2213262793589482e-08, "logits/chosen": -0.8324740529060364, "logits/rejected": -0.8063269853591919, "logps/chosen": -477.1441345214844, "logps/ref_chosen": -264.5608825683594, "logps/ref_rejected": -245.67031860351562, "logps/rejected": -566.9367065429688, "loss": 4.5493, "margin_dpo/margin_mean": 108.68309020996094, "margin_dpo/margin_std": 184.1924285888672, "step": 420 }, { "epoch": 0.881675392670157, "fcm_dpo/beta": 0.0037031802348792553, "fcm_dpo/delta": 0.020971816033124924, "fcm_dpo/margin": 92.56175994873047, "fcm_dpo/q_t": 0.41996240615844727, "grad_norm": 95.28728485107422, "learning_rate": 2.1464952759020856e-08, "logits/chosen": -0.9097946882247925, "logits/rejected": -0.8875160217285156, "logps/chosen": -492.8577880859375, "logps/ref_chosen": -297.70501708984375, "logps/ref_rejected": -243.74771118164062, "logps/rejected": -531.4622192382812, "loss": 4.5551, "margin_dpo/margin_mean": 92.56175994873047, "margin_dpo/margin_std": 136.10009765625, "step": 421 }, { "epoch": 0.8837696335078534, "fcm_dpo/beta": 0.0038025444373488426, "fcm_dpo/delta": 0.02667798101902008, "fcm_dpo/margin": 92.04246520996094, "fcm_dpo/q_t": 0.42156127095222473, "grad_norm": 65.87984466552734, "learning_rate": 2.07288983654679e-08, "logits/chosen": -0.7622085809707642, "logits/rejected": -0.8034301996231079, "logps/chosen": -495.5381164550781, "logps/ref_chosen": -288.3587646484375, "logps/ref_rejected": -256.4377746582031, "logps/rejected": -555.6596069335938, "loss": 4.6091, "margin_dpo/margin_mean": 92.04246520996094, "margin_dpo/margin_std": 156.10035705566406, "step": 422 }, { "epoch": 0.8858638743455497, "fcm_dpo/beta": 0.0037769617047160864, "fcm_dpo/delta": -0.0050736647099256516, "fcm_dpo/margin": 99.29490661621094, "fcm_dpo/q_t": 0.41419917345046997, "grad_norm": 78.68079376220703, "learning_rate": 2.0005139085293942e-08, "logits/chosen": -0.8863624334335327, "logits/rejected": -0.8650214076042175, "logps/chosen": -506.8820495605469, "logps/ref_chosen": -296.00701904296875, "logps/ref_rejected": -261.3480529785156, "logps/rejected": -571.5179443359375, "loss": 4.5105, "margin_dpo/margin_mean": 99.29490661621094, "margin_dpo/margin_std": 152.7718963623047, "step": 423 }, { "epoch": 0.8879581151832461, "fcm_dpo/beta": 0.0038537802174687386, "fcm_dpo/delta": 0.03373270109295845, "fcm_dpo/margin": 95.16132354736328, "fcm_dpo/q_t": 0.4156450033187866, "grad_norm": 85.12425231933594, "learning_rate": 1.9293713731512673e-08, "logits/chosen": -0.8723393678665161, "logits/rejected": -0.86158686876297, "logps/chosen": -502.749755859375, "logps/ref_chosen": -309.421875, "logps/ref_rejected": -249.14886474609375, "logps/rejected": -537.6380615234375, "loss": 4.4869, "margin_dpo/margin_mean": 95.16132354736328, "margin_dpo/margin_std": 134.9690704345703, "step": 424 }, { "epoch": 0.8900523560209425, "fcm_dpo/beta": 0.0038695167750120163, "fcm_dpo/delta": -0.025678057223558426, "fcm_dpo/margin": 84.49606323242188, "fcm_dpo/q_t": 0.42579007148742676, "grad_norm": 82.99080657958984, "learning_rate": 1.8594660455706763e-08, "logits/chosen": -0.849087655544281, "logits/rejected": -0.8525911569595337, "logps/chosen": -484.4912109375, "logps/ref_chosen": -280.50909423828125, "logps/ref_rejected": -276.8252258300781, "logps/rejected": -565.3033447265625, "loss": 4.6773, "margin_dpo/margin_mean": 84.49606323242188, "margin_dpo/margin_std": 146.6011199951172, "step": 425 }, { "epoch": 0.8921465968586387, "fcm_dpo/beta": 0.003897757502272725, "fcm_dpo/delta": 0.036198608577251434, "fcm_dpo/margin": 93.49591064453125, "fcm_dpo/q_t": 0.41721320152282715, "grad_norm": 73.07560729980469, "learning_rate": 1.7908016745981856e-08, "logits/chosen": -0.8812398314476013, "logits/rejected": -0.8626936674118042, "logps/chosen": -500.7666931152344, "logps/ref_chosen": -292.78521728515625, "logps/ref_rejected": -255.62698364257812, "logps/rejected": -557.1043701171875, "loss": 4.5514, "margin_dpo/margin_mean": 93.49591064453125, "margin_dpo/margin_std": 146.1980743408203, "step": 426 }, { "epoch": 0.8942408376963351, "fcm_dpo/beta": 0.003795379539951682, "fcm_dpo/delta": -0.13308368623256683, "fcm_dpo/margin": 130.365966796875, "fcm_dpo/q_t": 0.38931721448898315, "grad_norm": 111.64500427246094, "learning_rate": 1.7233819424956247e-08, "logits/chosen": -0.8529506921768188, "logits/rejected": -0.817730724811554, "logps/chosen": -492.5892333984375, "logps/ref_chosen": -288.7687072753906, "logps/ref_rejected": -268.4986572265625, "logps/rejected": -602.6851806640625, "loss": 4.2511, "margin_dpo/margin_mean": 130.36595153808594, "margin_dpo/margin_std": 172.0804901123047, "step": 427 }, { "epoch": 0.8963350785340314, "fcm_dpo/beta": 0.0035148044116795063, "fcm_dpo/delta": -0.02863123267889023, "fcm_dpo/margin": 112.14768981933594, "fcm_dpo/q_t": 0.40907180309295654, "grad_norm": 72.58002471923828, "learning_rate": 1.6572104647786245e-08, "logits/chosen": -0.8079947233200073, "logits/rejected": -0.837670624256134, "logps/chosen": -522.7776489257812, "logps/ref_chosen": -295.5209655761719, "logps/ref_rejected": -275.71026611328125, "logps/rejected": -615.1146850585938, "loss": 4.4223, "margin_dpo/margin_mean": 112.14769744873047, "margin_dpo/margin_std": 153.65895080566406, "step": 428 }, { "epoch": 0.8984293193717278, "fcm_dpo/beta": 0.0034794604871422052, "fcm_dpo/delta": 0.012085347436368465, "fcm_dpo/margin": 96.80742645263672, "fcm_dpo/q_t": 0.4216752350330353, "grad_norm": 73.65662384033203, "learning_rate": 1.5922907900227017e-08, "logits/chosen": -0.8177259564399719, "logits/rejected": -0.8231886625289917, "logps/chosen": -485.8576965332031, "logps/ref_chosen": -274.392333984375, "logps/ref_rejected": -258.574462890625, "logps/rejected": -566.8472900390625, "loss": 4.6748, "margin_dpo/margin_mean": 96.80743408203125, "margin_dpo/margin_std": 170.624267578125, "step": 429 }, { "epoch": 0.900523560209424, "fcm_dpo/beta": 0.003550901310518384, "fcm_dpo/delta": 0.013876799494028091, "fcm_dpo/margin": 88.0494613647461, "fcm_dpo/q_t": 0.4287679195404053, "grad_norm": 69.73419952392578, "learning_rate": 1.5286263996730026e-08, "logits/chosen": -0.8986079096794128, "logits/rejected": -0.869090735912323, "logps/chosen": -497.7119140625, "logps/ref_chosen": -288.7391357421875, "logps/ref_rejected": -268.6106262207031, "logps/rejected": -565.6328125, "loss": 4.6721, "margin_dpo/margin_mean": 88.0494613647461, "margin_dpo/margin_std": 147.50392150878906, "step": 430 }, { "epoch": 0.9026178010471204, "fcm_dpo/beta": 0.0035896864719688892, "fcm_dpo/delta": 0.020418308675289154, "fcm_dpo/margin": 72.26271057128906, "fcm_dpo/q_t": 0.4400396943092346, "grad_norm": 72.33751678466797, "learning_rate": 1.4662207078575684e-08, "logits/chosen": -0.8635554909706116, "logits/rejected": -0.8318578004837036, "logps/chosen": -481.857177734375, "logps/ref_chosen": -275.7247314453125, "logps/ref_rejected": -268.91729736328125, "logps/rejected": -547.3124389648438, "loss": 4.8212, "margin_dpo/margin_mean": 72.26271057128906, "margin_dpo/margin_std": 134.55499267578125, "step": 431 }, { "epoch": 0.9047120418848168, "fcm_dpo/beta": 0.003594242036342621, "fcm_dpo/delta": 0.006677491124719381, "fcm_dpo/margin": 99.71732330322266, "fcm_dpo/q_t": 0.4194951355457306, "grad_norm": 72.6844482421875, "learning_rate": 1.40507706120426e-08, "logits/chosen": -0.9026849269866943, "logits/rejected": -0.8854382038116455, "logps/chosen": -486.9830017089844, "logps/ref_chosen": -291.42010498046875, "logps/ref_rejected": -255.48202514648438, "logps/rejected": -550.7622680664062, "loss": 4.558, "margin_dpo/margin_mean": 99.71731567382812, "margin_dpo/margin_std": 157.44149780273438, "step": 432 }, { "epoch": 0.9068062827225131, "fcm_dpo/beta": 0.003687588730826974, "fcm_dpo/delta": 0.06109008565545082, "fcm_dpo/margin": 85.34246063232422, "fcm_dpo/q_t": 0.42769885063171387, "grad_norm": 72.68798065185547, "learning_rate": 1.345198738661285e-08, "logits/chosen": -0.8471591472625732, "logits/rejected": -0.8419995307922363, "logps/chosen": -460.63702392578125, "logps/ref_chosen": -246.2268829345703, "logps/ref_rejected": -253.65924072265625, "logps/rejected": -553.4118041992188, "loss": 4.6434, "margin_dpo/margin_mean": 85.34246063232422, "margin_dpo/margin_std": 140.0343017578125, "step": 433 }, { "epoch": 0.9089005235602095, "fcm_dpo/beta": 0.0038499152287840843, "fcm_dpo/delta": 0.029952505603432655, "fcm_dpo/margin": 85.55077362060547, "fcm_dpo/q_t": 0.4248325228691101, "grad_norm": 72.78632354736328, "learning_rate": 1.2865889513213628e-08, "logits/chosen": -0.8568861484527588, "logits/rejected": -0.8612566590309143, "logps/chosen": -520.3125, "logps/ref_chosen": -295.4618225097656, "logps/ref_rejected": -256.2254333496094, "logps/rejected": -566.6268920898438, "loss": 4.6448, "margin_dpo/margin_mean": 85.55077362060547, "margin_dpo/margin_std": 144.971435546875, "step": 434 }, { "epoch": 0.9109947643979057, "fcm_dpo/beta": 0.0038717712741345167, "fcm_dpo/delta": -0.01573629304766655, "fcm_dpo/margin": 99.90605163574219, "fcm_dpo/q_t": 0.4140922725200653, "grad_norm": 112.00686645507812, "learning_rate": 1.2292508422495157e-08, "logits/chosen": -0.8434274196624756, "logits/rejected": -0.831263542175293, "logps/chosen": -464.3940124511719, "logps/ref_chosen": -260.7384033203125, "logps/ref_rejected": -248.5688018798828, "logps/rejected": -552.1304321289062, "loss": 4.4539, "margin_dpo/margin_mean": 99.90605163574219, "margin_dpo/margin_std": 141.65713500976562, "step": 435 }, { "epoch": 0.9130890052356021, "fcm_dpo/beta": 0.0039107538759708405, "fcm_dpo/delta": 0.03949038311839104, "fcm_dpo/margin": 78.44125366210938, "fcm_dpo/q_t": 0.4342585504055023, "grad_norm": 115.6869888305664, "learning_rate": 1.1731874863145142e-08, "logits/chosen": -0.8294715285301208, "logits/rejected": -0.8246615529060364, "logps/chosen": -540.929443359375, "logps/ref_chosen": -319.3224792480469, "logps/ref_rejected": -299.30322265625, "logps/rejected": -599.3515014648438, "loss": 4.8039, "margin_dpo/margin_mean": 78.44124603271484, "margin_dpo/margin_std": 165.59027099609375, "step": 436 }, { "epoch": 0.9151832460732985, "fcm_dpo/beta": 0.0038340180180966854, "fcm_dpo/delta": -0.055059827864170074, "fcm_dpo/margin": 108.45451354980469, "fcm_dpo/q_t": 0.4066217243671417, "grad_norm": 66.90118408203125, "learning_rate": 1.118401890024001e-08, "logits/chosen": -0.8663026094436646, "logits/rejected": -0.8491877913475037, "logps/chosen": -483.2367248535156, "logps/ref_chosen": -278.82879638671875, "logps/ref_rejected": -272.55303955078125, "logps/rejected": -585.41552734375, "loss": 4.4401, "margin_dpo/margin_mean": 108.45451354980469, "margin_dpo/margin_std": 160.95172119140625, "step": 437 }, { "epoch": 0.9172774869109948, "fcm_dpo/beta": 0.003806520253419876, "fcm_dpo/delta": 0.009153133258223534, "fcm_dpo/margin": 50.68528747558594, "fcm_dpo/q_t": 0.4563140869140625, "grad_norm": 79.08209991455078, "learning_rate": 1.06489699136324e-08, "logits/chosen": -0.8532021045684814, "logits/rejected": -0.8665780425071716, "logps/chosen": -475.7437438964844, "logps/ref_chosen": -259.31903076171875, "logps/ref_rejected": -240.99581909179688, "logps/rejected": -508.10577392578125, "loss": 5.1451, "margin_dpo/margin_mean": 50.6852912902832, "margin_dpo/margin_std": 148.322021484375, "step": 438 }, { "epoch": 0.9193717277486911, "fcm_dpo/beta": 0.0038379242178052664, "fcm_dpo/delta": 0.033137980848550797, "fcm_dpo/margin": 95.86003112792969, "fcm_dpo/q_t": 0.41530516743659973, "grad_norm": 98.0456771850586, "learning_rate": 1.0126756596375685e-08, "logits/chosen": -0.831727147102356, "logits/rejected": -0.8308047652244568, "logps/chosen": -466.54345703125, "logps/ref_chosen": -257.1243896484375, "logps/ref_rejected": -243.20416259765625, "logps/rejected": -548.4832763671875, "loss": 4.4882, "margin_dpo/margin_mean": 95.86003112792969, "margin_dpo/margin_std": 138.56790161132812, "step": 439 }, { "epoch": 0.9214659685863874, "fcm_dpo/beta": 0.00401148060336709, "fcm_dpo/delta": 0.05375323444604874, "fcm_dpo/margin": 69.56246185302734, "fcm_dpo/q_t": 0.4353058934211731, "grad_norm": 70.47123718261719, "learning_rate": 9.617406953185136e-09, "logits/chosen": -0.8858019113540649, "logits/rejected": -0.876878023147583, "logps/chosen": -533.7391967773438, "logps/ref_chosen": -307.5315246582031, "logps/ref_rejected": -264.3540954589844, "logps/rejected": -560.124267578125, "loss": 4.796, "margin_dpo/margin_mean": 69.56246185302734, "margin_dpo/margin_std": 130.65516662597656, "step": 440 }, { "epoch": 0.9235602094240838, "fcm_dpo/beta": 0.004070944152772427, "fcm_dpo/delta": -0.028871532529592514, "fcm_dpo/margin": 104.92424774169922, "fcm_dpo/q_t": 0.402651846408844, "grad_norm": 99.01258850097656, "learning_rate": 9.12094829893642e-09, "logits/chosen": -0.8418979048728943, "logits/rejected": -0.8321986794471741, "logps/chosen": -520.474365234375, "logps/ref_chosen": -309.9819641113281, "logps/ref_rejected": -297.4968566894531, "logps/rejected": -612.9135131835938, "loss": 4.3406, "margin_dpo/margin_mean": 104.92424774169922, "margin_dpo/margin_std": 137.64004516601562, "step": 441 }, { "epoch": 0.9256544502617801, "fcm_dpo/beta": 0.004000660963356495, "fcm_dpo/delta": 0.02609052136540413, "fcm_dpo/margin": 78.4242172241211, "fcm_dpo/q_t": 0.42918646335601807, "grad_norm": 61.02368927001953, "learning_rate": 8.637407257200496e-09, "logits/chosen": -0.9113872647285461, "logits/rejected": -0.8679369688034058, "logps/chosen": -509.2906494140625, "logps/ref_chosen": -278.9791564941406, "logps/ref_rejected": -242.87310791015625, "logps/rejected": -551.6088256835938, "loss": 4.7697, "margin_dpo/margin_mean": 78.42420959472656, "margin_dpo/margin_std": 155.16122436523438, "step": 442 }, { "epoch": 0.9277486910994764, "fcm_dpo/beta": 0.0041396538726985455, "fcm_dpo/delta": 0.014572596177458763, "fcm_dpo/margin": 85.65485382080078, "fcm_dpo/q_t": 0.41839903593063354, "grad_norm": 95.5350570678711, "learning_rate": 8.166809758815895e-09, "logits/chosen": -0.8070948123931885, "logits/rejected": -0.827476441860199, "logps/chosen": -486.9132080078125, "logps/ref_chosen": -273.5590515136719, "logps/ref_rejected": -264.0199279785156, "logps/rejected": -563.0289306640625, "loss": 4.6045, "margin_dpo/margin_mean": 85.65485382080078, "margin_dpo/margin_std": 139.71743774414062, "step": 443 }, { "epoch": 0.9298429319371728, "fcm_dpo/beta": 0.0040481919422745705, "fcm_dpo/delta": -0.02957877703011036, "fcm_dpo/margin": 99.84343719482422, "fcm_dpo/q_t": 0.41018837690353394, "grad_norm": 85.15692901611328, "learning_rate": 7.709181040498253e-09, "logits/chosen": -0.8221174478530884, "logits/rejected": -0.8085671663284302, "logps/chosen": -508.9268798828125, "logps/ref_chosen": -298.1441955566406, "logps/ref_rejected": -268.0572814941406, "logps/rejected": -578.6834106445312, "loss": 4.487, "margin_dpo/margin_mean": 99.84344482421875, "margin_dpo/margin_std": 157.79037475585938, "step": 444 }, { "epoch": 0.9319371727748691, "fcm_dpo/beta": 0.003955287858843803, "fcm_dpo/delta": -0.06290034204721451, "fcm_dpo/margin": 85.2093276977539, "fcm_dpo/q_t": 0.4250834882259369, "grad_norm": 72.96340942382812, "learning_rate": 7.2645456434869965e-09, "logits/chosen": -0.8938105702400208, "logits/rejected": -0.9017377495765686, "logps/chosen": -467.8274230957031, "logps/ref_chosen": -254.54067993164062, "logps/ref_rejected": -264.2445983886719, "logps/rejected": -562.7406616210938, "loss": 4.7053, "margin_dpo/margin_mean": 85.2093276977539, "margin_dpo/margin_std": 150.60330200195312, "step": 445 }, { "epoch": 0.9340314136125655, "fcm_dpo/beta": 0.003790568793192506, "fcm_dpo/delta": 0.013937651179730892, "fcm_dpo/margin": 94.15351867675781, "fcm_dpo/q_t": 0.4192274510860443, "grad_norm": 73.7974853515625, "learning_rate": 6.832927412229017e-09, "logits/chosen": -0.8159938454627991, "logits/rejected": -0.818867564201355, "logps/chosen": -512.8572387695312, "logps/ref_chosen": -306.72247314453125, "logps/ref_rejected": -266.3735656738281, "logps/rejected": -566.661865234375, "loss": 4.5708, "margin_dpo/margin_mean": 94.15351867675781, "margin_dpo/margin_std": 150.21368408203125, "step": 446 }, { "epoch": 0.9361256544502617, "fcm_dpo/beta": 0.0036977354902774096, "fcm_dpo/delta": -0.06611842662096024, "fcm_dpo/margin": 105.1883316040039, "fcm_dpo/q_t": 0.4103267788887024, "grad_norm": 62.978233337402344, "learning_rate": 6.414349493100129e-09, "logits/chosen": -0.8234304189682007, "logits/rejected": -0.8186841607093811, "logps/chosen": -459.4782409667969, "logps/ref_chosen": -260.51727294921875, "logps/ref_rejected": -236.47061157226562, "logps/rejected": -540.619873046875, "loss": 4.4225, "margin_dpo/margin_mean": 105.18832397460938, "margin_dpo/margin_std": 141.39520263671875, "step": 447 }, { "epoch": 0.9382198952879581, "fcm_dpo/beta": 0.003641946241259575, "fcm_dpo/delta": 0.012333293445408344, "fcm_dpo/margin": 97.73287963867188, "fcm_dpo/q_t": 0.4186139404773712, "grad_norm": 81.14934539794922, "learning_rate": 6.0088343331638756e-09, "logits/chosen": -0.8353531360626221, "logits/rejected": -0.8291007280349731, "logps/chosen": -484.96905517578125, "logps/ref_chosen": -268.78704833984375, "logps/ref_rejected": -262.1703796386719, "logps/rejected": -576.0852661132812, "loss": 4.5206, "margin_dpo/margin_mean": 97.73287963867188, "margin_dpo/margin_std": 139.15956115722656, "step": 448 }, { "epoch": 0.9403141361256544, "fcm_dpo/beta": 0.0036854774225503206, "fcm_dpo/delta": -0.009766047820448875, "fcm_dpo/margin": 96.53168487548828, "fcm_dpo/q_t": 0.4160364866256714, "grad_norm": 130.65623474121094, "learning_rate": 5.616403678967624e-09, "logits/chosen": -0.9169028997421265, "logits/rejected": -0.9006680250167847, "logps/chosen": -533.5242919921875, "logps/ref_chosen": -330.9514465332031, "logps/ref_rejected": -239.76974487304688, "logps/rejected": -538.8742065429688, "loss": 4.5593, "margin_dpo/margin_mean": 96.53167724609375, "margin_dpo/margin_std": 148.29200744628906, "step": 449 }, { "epoch": 0.9424083769633508, "fcm_dpo/beta": 0.0036835160572081804, "fcm_dpo/delta": 0.015206391923129559, "fcm_dpo/margin": 87.07230377197266, "fcm_dpo/q_t": 0.42582279443740845, "grad_norm": 80.95256042480469, "learning_rate": 5.2370785753763356e-09, "logits/chosen": -0.8152453303337097, "logits/rejected": -0.8101465702056885, "logps/chosen": -507.47833251953125, "logps/ref_chosen": -284.26544189453125, "logps/ref_rejected": -250.5401611328125, "logps/rejected": -560.825439453125, "loss": 4.6198, "margin_dpo/margin_mean": 87.07230377197266, "margin_dpo/margin_std": 132.29653930664062, "step": 450 }, { "epoch": 0.9445026178010472, "fcm_dpo/beta": 0.0036616751458495855, "fcm_dpo/delta": -0.018945183604955673, "fcm_dpo/margin": 79.61962890625, "fcm_dpo/q_t": 0.43234091997146606, "grad_norm": 58.6148567199707, "learning_rate": 4.8708793644441086e-09, "logits/chosen": -0.8456954956054688, "logits/rejected": -0.8164753913879395, "logps/chosen": -531.6286010742188, "logps/ref_chosen": -302.3209228515625, "logps/ref_rejected": -254.09747314453125, "logps/rejected": -563.0247802734375, "loss": 4.7455, "margin_dpo/margin_mean": 79.61962890625, "margin_dpo/margin_std": 140.28172302246094, "step": 451 }, { "epoch": 0.9465968586387434, "fcm_dpo/beta": 0.003525385633111, "fcm_dpo/delta": -0.05114733427762985, "fcm_dpo/margin": 109.65728759765625, "fcm_dpo/q_t": 0.41492602229118347, "grad_norm": 70.34674835205078, "learning_rate": 4.517825684323323e-09, "logits/chosen": -0.8856893181800842, "logits/rejected": -0.8606870770454407, "logps/chosen": -508.6758117675781, "logps/ref_chosen": -299.39215087890625, "logps/ref_rejected": -284.3475036621094, "logps/rejected": -603.2885131835938, "loss": 4.5145, "margin_dpo/margin_mean": 109.65728759765625, "margin_dpo/margin_std": 172.53543090820312, "step": 452 }, { "epoch": 0.9486910994764398, "fcm_dpo/beta": 0.0035962536931037903, "fcm_dpo/delta": 0.05969720333814621, "fcm_dpo/margin": 95.231201171875, "fcm_dpo/q_t": 0.42202088236808777, "grad_norm": 62.857933044433594, "learning_rate": 4.1779364682113794e-09, "logits/chosen": -0.8256725072860718, "logits/rejected": -0.8185856938362122, "logps/chosen": -548.9927978515625, "logps/ref_chosen": -324.6517028808594, "logps/ref_rejected": -304.1527099609375, "logps/rejected": -623.7249755859375, "loss": 4.5858, "margin_dpo/margin_mean": 95.23121643066406, "margin_dpo/margin_std": 154.75979614257812, "step": 453 }, { "epoch": 0.9507853403141361, "fcm_dpo/beta": 0.0036729713901877403, "fcm_dpo/delta": -0.00931104552000761, "fcm_dpo/margin": 111.07328796386719, "fcm_dpo/q_t": 0.407396137714386, "grad_norm": 63.60319137573242, "learning_rate": 3.851229943335393e-09, "logits/chosen": -0.8640636801719666, "logits/rejected": -0.8737033009529114, "logps/chosen": -515.0318603515625, "logps/ref_chosen": -299.6117248535156, "logps/ref_rejected": -303.74224853515625, "logps/rejected": -630.2357177734375, "loss": 4.4341, "margin_dpo/margin_mean": 111.07328796386719, "margin_dpo/margin_std": 162.49261474609375, "step": 454 }, { "epoch": 0.9528795811518325, "fcm_dpo/beta": 0.003721519373357296, "fcm_dpo/delta": 0.018941761925816536, "fcm_dpo/margin": 76.07414245605469, "fcm_dpo/q_t": 0.4374096989631653, "grad_norm": 75.65412139892578, "learning_rate": 3.5377236299748147e-09, "logits/chosen": -0.8347422480583191, "logits/rejected": -0.8438669443130493, "logps/chosen": -480.7205810546875, "logps/ref_chosen": -273.6116943359375, "logps/ref_rejected": -274.4293518066406, "logps/rejected": -557.6124267578125, "loss": 4.8186, "margin_dpo/margin_mean": 76.07414245605469, "margin_dpo/margin_std": 152.4740753173828, "step": 455 }, { "epoch": 0.9549738219895288, "fcm_dpo/beta": 0.0037363064475357533, "fcm_dpo/delta": 0.0011149700731039047, "fcm_dpo/margin": 100.02578735351562, "fcm_dpo/q_t": 0.419757217168808, "grad_norm": 85.6431884765625, "learning_rate": 3.2374343405217884e-09, "logits/chosen": -0.7592871785163879, "logits/rejected": -0.7727906107902527, "logps/chosen": -556.2762451171875, "logps/ref_chosen": -322.17193603515625, "logps/ref_rejected": -294.54461669921875, "logps/rejected": -628.6747436523438, "loss": 4.6092, "margin_dpo/margin_mean": 100.02578735351562, "margin_dpo/margin_std": 177.65438842773438, "step": 456 }, { "epoch": 0.9570680628272251, "fcm_dpo/beta": 0.0036805581767112017, "fcm_dpo/delta": -0.012985389679670334, "fcm_dpo/margin": 104.80935668945312, "fcm_dpo/q_t": 0.4097025394439697, "grad_norm": 78.89384460449219, "learning_rate": 2.9503781785795713e-09, "logits/chosen": -0.8332885503768921, "logits/rejected": -0.8361387848854065, "logps/chosen": -535.7222290039062, "logps/ref_chosen": -307.7962341308594, "logps/ref_rejected": -274.5501403808594, "logps/rejected": -607.2855224609375, "loss": 4.549, "margin_dpo/margin_mean": 104.80935668945312, "margin_dpo/margin_std": 170.84185791015625, "step": 457 }, { "epoch": 0.9591623036649215, "fcm_dpo/beta": 0.0037839512806385756, "fcm_dpo/delta": 0.04983433336019516, "fcm_dpo/margin": 92.8138427734375, "fcm_dpo/q_t": 0.4212765693664551, "grad_norm": 72.37201690673828, "learning_rate": 2.6765705380989432e-09, "logits/chosen": -0.8502821922302246, "logits/rejected": -0.8296154737472534, "logps/chosen": -515.9027709960938, "logps/ref_chosen": -297.0316467285156, "logps/ref_rejected": -276.1112365722656, "logps/rejected": -587.796142578125, "loss": 4.6273, "margin_dpo/margin_mean": 92.8138427734375, "margin_dpo/margin_std": 158.17807006835938, "step": 458 }, { "epoch": 0.9612565445026178, "fcm_dpo/beta": 0.003876004833728075, "fcm_dpo/delta": 0.018467195332050323, "fcm_dpo/margin": 78.23777770996094, "fcm_dpo/q_t": 0.4319803714752197, "grad_norm": 74.26271057128906, "learning_rate": 2.416026102552732e-09, "logits/chosen": -0.9194495677947998, "logits/rejected": -0.9117699861526489, "logps/chosen": -500.1540222167969, "logps/ref_chosen": -293.5252990722656, "logps/ref_rejected": -289.30126953125, "logps/rejected": -574.167724609375, "loss": 4.7467, "margin_dpo/margin_mean": 78.23778533935547, "margin_dpo/margin_std": 145.30636596679688, "step": 459 }, { "epoch": 0.9633507853403142, "fcm_dpo/beta": 0.003939047455787659, "fcm_dpo/delta": 0.0354711078107357, "fcm_dpo/margin": 78.7183837890625, "fcm_dpo/q_t": 0.42859572172164917, "grad_norm": 85.72238159179688, "learning_rate": 2.168758844148272e-09, "logits/chosen": -0.8774435520172119, "logits/rejected": -0.8768860101699829, "logps/chosen": -532.5010986328125, "logps/ref_chosen": -318.7803649902344, "logps/ref_rejected": -258.7906799316406, "logps/rejected": -551.229736328125, "loss": 4.7105, "margin_dpo/margin_mean": 78.71837615966797, "margin_dpo/margin_std": 142.5164794921875, "step": 460 }, { "epoch": 0.9654450261780104, "fcm_dpo/beta": 0.004013788886368275, "fcm_dpo/delta": -0.008083513006567955, "fcm_dpo/margin": 87.99292755126953, "fcm_dpo/q_t": 0.4219253361225128, "grad_norm": 99.12641143798828, "learning_rate": 1.9347820230782295e-09, "logits/chosen": -0.8341564536094666, "logits/rejected": -0.8578203320503235, "logps/chosen": -459.1229248046875, "logps/ref_chosen": -243.9099884033203, "logps/ref_rejected": -232.6382293701172, "logps/rejected": -535.8441162109375, "loss": 4.6733, "margin_dpo/margin_mean": 87.99293518066406, "margin_dpo/margin_std": 161.89053344726562, "step": 461 }, { "epoch": 0.9675392670157068, "fcm_dpo/beta": 0.0038696322590112686, "fcm_dpo/delta": -0.05201539024710655, "fcm_dpo/margin": 108.68580627441406, "fcm_dpo/q_t": 0.40523770451545715, "grad_norm": 98.13914489746094, "learning_rate": 1.7141081868094209e-09, "logits/chosen": -0.8697165250778198, "logits/rejected": -0.8183139562606812, "logps/chosen": -559.72705078125, "logps/ref_chosen": -344.09100341796875, "logps/ref_rejected": -252.45037841796875, "logps/rejected": -576.7722778320312, "loss": 4.3988, "margin_dpo/margin_mean": 108.68580627441406, "margin_dpo/margin_std": 156.15455627441406, "step": 462 }, { "epoch": 0.9696335078534032, "fcm_dpo/beta": 0.003800945356488228, "fcm_dpo/delta": -0.012423641048371792, "fcm_dpo/margin": 81.46054077148438, "fcm_dpo/q_t": 0.4303792119026184, "grad_norm": 74.05850219726562, "learning_rate": 1.5067491694100153e-09, "logits/chosen": -0.8569101095199585, "logits/rejected": -0.8149942755699158, "logps/chosen": -511.5487976074219, "logps/ref_chosen": -297.1424560546875, "logps/ref_rejected": -234.0208282470703, "logps/rejected": -529.8877563476562, "loss": 4.7769, "margin_dpo/margin_mean": 81.46053314208984, "margin_dpo/margin_std": 161.78460693359375, "step": 463 }, { "epoch": 0.9717277486910995, "fcm_dpo/beta": 0.003825580468401313, "fcm_dpo/delta": 0.033597130328416824, "fcm_dpo/margin": 81.79236602783203, "fcm_dpo/q_t": 0.4282078742980957, "grad_norm": 81.68612670898438, "learning_rate": 1.3127160909147672e-09, "logits/chosen": -0.8513908386230469, "logits/rejected": -0.8705244064331055, "logps/chosen": -494.5838623046875, "logps/ref_chosen": -265.71075439453125, "logps/ref_rejected": -256.4108581542969, "logps/rejected": -567.0762939453125, "loss": 4.7027, "margin_dpo/margin_mean": 81.7923583984375, "margin_dpo/margin_std": 145.36318969726562, "step": 464 }, { "epoch": 0.9738219895287958, "fcm_dpo/beta": 0.003732735523954034, "fcm_dpo/delta": -0.07069091498851776, "fcm_dpo/margin": 111.66655731201172, "fcm_dpo/q_t": 0.4052296280860901, "grad_norm": 71.67578125, "learning_rate": 1.1320193567288527e-09, "logits/chosen": -0.9181026816368103, "logits/rejected": -0.8817926049232483, "logps/chosen": -494.74371337890625, "logps/ref_chosen": -293.1527404785156, "logps/ref_rejected": -293.70947265625, "logps/rejected": -606.967041015625, "loss": 4.3887, "margin_dpo/margin_mean": 111.66655731201172, "margin_dpo/margin_std": 149.8941650390625, "step": 465 }, { "epoch": 0.9759162303664921, "fcm_dpo/beta": 0.0035918080247938633, "fcm_dpo/delta": -0.013611567206680775, "fcm_dpo/margin": 114.79224395751953, "fcm_dpo/q_t": 0.4061741828918457, "grad_norm": 64.40978240966797, "learning_rate": 9.64668657069706e-10, "logits/chosen": -0.8348801136016846, "logits/rejected": -0.7807033061981201, "logps/chosen": -456.18829345703125, "logps/ref_chosen": -261.4775695800781, "logps/ref_rejected": -248.36282348632812, "logps/rejected": -557.8658447265625, "loss": 4.3506, "margin_dpo/margin_mean": 114.79224395751953, "margin_dpo/margin_std": 149.470703125, "step": 466 }, { "epoch": 0.9780104712041885, "fcm_dpo/beta": 0.0036084747407585382, "fcm_dpo/delta": 0.014053348451852798, "fcm_dpo/margin": 83.56181335449219, "fcm_dpo/q_t": 0.43221747875213623, "grad_norm": 71.90433502197266, "learning_rate": 8.106729664475176e-10, "logits/chosen": -0.807646632194519, "logits/rejected": -0.8021789789199829, "logps/chosen": -482.59112548828125, "logps/ref_chosen": -266.354248046875, "logps/ref_rejected": -277.76324462890625, "logps/rejected": -577.5619506835938, "loss": 4.7741, "margin_dpo/margin_mean": 83.56181335449219, "margin_dpo/margin_std": 162.9016571044922, "step": 467 }, { "epoch": 0.9801047120418848, "fcm_dpo/beta": 0.0037187100388109684, "fcm_dpo/delta": 0.021228276193141937, "fcm_dpo/margin": 73.15608978271484, "fcm_dpo/q_t": 0.43574419617652893, "grad_norm": 87.2890853881836, "learning_rate": 6.700405431837585e-10, "logits/chosen": -0.9090844988822937, "logits/rejected": -0.8785009980201721, "logps/chosen": -532.4305419921875, "logps/ref_chosen": -317.9631652832031, "logps/ref_rejected": -261.8744201660156, "logps/rejected": -549.4978637695312, "loss": 4.877, "margin_dpo/margin_mean": 73.15608978271484, "margin_dpo/margin_std": 154.40957641601562, "step": 468 }, { "epoch": 0.9821989528795811, "fcm_dpo/beta": 0.0036812869366258383, "fcm_dpo/delta": 0.003829888068139553, "fcm_dpo/margin": 107.45279693603516, "fcm_dpo/q_t": 0.4098895788192749, "grad_norm": 60.7659912109375, "learning_rate": 5.427789289685347e-10, "logits/chosen": -0.8316172957420349, "logits/rejected": -0.8154798746109009, "logps/chosen": -528.8984985351562, "logps/ref_chosen": -324.8868103027344, "logps/ref_rejected": -264.0421447753906, "logps/rejected": -575.5067138671875, "loss": 4.4592, "margin_dpo/margin_mean": 107.45278930664062, "margin_dpo/margin_std": 159.00772094726562, "step": 469 }, { "epoch": 0.9842931937172775, "fcm_dpo/beta": 0.0037065560463815928, "fcm_dpo/delta": -0.03380197659134865, "fcm_dpo/margin": 101.50714111328125, "fcm_dpo/q_t": 0.4139646291732788, "grad_norm": 65.40029907226562, "learning_rate": 4.288949484559934e-10, "logits/chosen": -0.8253117799758911, "logits/rejected": -0.8232603669166565, "logps/chosen": -519.7571411132812, "logps/ref_chosen": -314.7042541503906, "logps/ref_rejected": -259.2276611328125, "logps/rejected": -565.7877197265625, "loss": 4.4905, "margin_dpo/margin_mean": 101.50714111328125, "margin_dpo/margin_std": 147.7560577392578, "step": 470 }, { "epoch": 0.9863874345549738, "fcm_dpo/beta": 0.0036814697086811066, "fcm_dpo/delta": -0.013820381835103035, "fcm_dpo/margin": 97.27108001708984, "fcm_dpo/q_t": 0.4229424297809601, "grad_norm": 81.09204864501953, "learning_rate": 3.2839470889836627e-10, "logits/chosen": -0.8735456466674805, "logits/rejected": -0.8611137270927429, "logps/chosen": -511.9947814941406, "logps/ref_chosen": -292.5748291015625, "logps/ref_rejected": -298.7525329589844, "logps/rejected": -615.4434814453125, "loss": 4.6116, "margin_dpo/margin_mean": 97.27108001708984, "margin_dpo/margin_std": 168.31797790527344, "step": 471 }, { "epoch": 0.9884816753926702, "fcm_dpo/beta": 0.0036161583848297596, "fcm_dpo/delta": 0.009807462804019451, "fcm_dpo/margin": 107.87297058105469, "fcm_dpo/q_t": 0.41221684217453003, "grad_norm": 71.03485870361328, "learning_rate": 2.412835998185092e-10, "logits/chosen": -0.8705576658248901, "logits/rejected": -0.8849227428436279, "logps/chosen": -442.5689697265625, "logps/ref_chosen": -243.37380981445312, "logps/ref_rejected": -251.12109375, "logps/rejected": -558.189208984375, "loss": 4.4378, "margin_dpo/margin_mean": 107.87297058105469, "margin_dpo/margin_std": 152.4186248779297, "step": 472 }, { "epoch": 0.9905759162303664, "fcm_dpo/beta": 0.0036868297029286623, "fcm_dpo/delta": 0.006859378889203072, "fcm_dpo/margin": 106.62423706054688, "fcm_dpo/q_t": 0.408770352602005, "grad_norm": 72.31365203857422, "learning_rate": 1.6756629272085544e-10, "logits/chosen": -0.8231157064437866, "logits/rejected": -0.822382390499115, "logps/chosen": -495.9771728515625, "logps/ref_chosen": -286.3286437988281, "logps/ref_rejected": -258.6535339355469, "logps/rejected": -574.9263916015625, "loss": 4.3838, "margin_dpo/margin_mean": 106.62422180175781, "margin_dpo/margin_std": 135.7769012451172, "step": 473 }, { "epoch": 0.9926701570680628, "fcm_dpo/beta": 0.003645974677056074, "fcm_dpo/delta": 0.02563265711069107, "fcm_dpo/margin": 90.80755615234375, "fcm_dpo/q_t": 0.42361754179000854, "grad_norm": 84.59915161132812, "learning_rate": 1.072467408408384e-10, "logits/chosen": -0.8701086640357971, "logits/rejected": -0.8666247129440308, "logps/chosen": -503.46722412109375, "logps/ref_chosen": -288.08966064453125, "logps/ref_rejected": -266.69696044921875, "logps/rejected": -572.8821411132812, "loss": 4.5952, "margin_dpo/margin_mean": 90.80755615234375, "margin_dpo/margin_std": 136.625, "step": 474 }, { "epoch": 0.9947643979057592, "fcm_dpo/beta": 0.003674778388813138, "fcm_dpo/delta": -0.003959197551012039, "fcm_dpo/margin": 81.83253479003906, "fcm_dpo/q_t": 0.43231987953186035, "grad_norm": 84.51021575927734, "learning_rate": 6.032817893297793e-11, "logits/chosen": -0.8477621078491211, "logits/rejected": -0.860824704170227, "logps/chosen": -456.145751953125, "logps/ref_chosen": -256.0030517578125, "logps/ref_rejected": -244.50660705566406, "logps/rejected": -526.4818115234375, "loss": 4.7206, "margin_dpo/margin_mean": 81.83253479003906, "margin_dpo/margin_std": 143.8907012939453, "step": 475 }, { "epoch": 0.9968586387434555, "fcm_dpo/beta": 0.00364405638538301, "fcm_dpo/delta": -0.0139269158244133, "fcm_dpo/margin": 90.71614074707031, "fcm_dpo/q_t": 0.4268645644187927, "grad_norm": 66.6181640625, "learning_rate": 2.6813123097352287e-11, "logits/chosen": -0.9170237183570862, "logits/rejected": -0.8723958730697632, "logps/chosen": -522.1641845703125, "logps/ref_chosen": -321.467529296875, "logps/ref_rejected": -295.0592956542969, "logps/rejected": -586.4722290039062, "loss": 4.6752, "margin_dpo/margin_mean": 90.71612548828125, "margin_dpo/margin_std": 153.20860290527344, "step": 476 }, { "epoch": 0.9989528795811519, "fcm_dpo/beta": 0.0037430531810969114, "fcm_dpo/delta": -0.004374518990516663, "fcm_dpo/margin": 91.00071716308594, "fcm_dpo/q_t": 0.42345336079597473, "grad_norm": 73.29693603515625, "learning_rate": 6.7033706447061635e-12, "logits/chosen": -0.795288622379303, "logits/rejected": -0.8026962876319885, "logps/chosen": -497.5155029296875, "logps/ref_chosen": -276.7939758300781, "logps/ref_rejected": -244.82919311523438, "logps/rejected": -556.551513671875, "loss": 4.7313, "margin_dpo/margin_mean": 91.00071716308594, "margin_dpo/margin_std": 175.25076293945312, "step": 477 }, { "epoch": 0.9989528795811519, "step": 477, "total_flos": 0.0, "train_loss": 4.781781088631108, "train_runtime": 6847.9581, "train_samples_per_second": 8.927, "train_steps_per_second": 0.07 } ], "logging_steps": 1, "max_steps": 477, "num_input_tokens_seen": 0, "num_train_epochs": 1, "save_steps": 200, "stateful_callbacks": { "TrainerControl": { "args": { "should_epoch_stop": false, "should_evaluate": false, "should_log": false, "should_save": true, "should_training_stop": true }, "attributes": {} } }, "total_flos": 0.0, "train_batch_size": 4, "trial_name": null, "trial_params": null }