{ "best_metric": null, "best_model_checkpoint": null, "epoch": 2.0, "eval_steps": 100, "global_step": 574, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.003484320557491289, "grad_norm": 572.2640100744908, "learning_rate": 1.2068965517241378e-08, "logits/chosen": -2.5345611572265625, "logits/rejected": -2.581700563430786, "logps/chosen": -60.002105712890625, "logps/rejected": -99.98374938964844, "loss": 0.6931, "rewards/accuracies": 0.0, "rewards/chosen": 0.0, "rewards/margins": 0.0, "rewards/rejected": 0.0, "step": 1 }, { "epoch": 0.03484320557491289, "grad_norm": 548.4940657606528, "learning_rate": 1.206896551724138e-07, "logits/chosen": -2.5637850761413574, "logits/rejected": -2.562532424926758, "logps/chosen": -59.66712188720703, "logps/rejected": -73.37606811523438, "loss": 0.7017, "rewards/accuracies": 0.1805555522441864, "rewards/chosen": -0.00738478871062398, "rewards/margins": -0.01218735333532095, "rewards/rejected": 0.00480256462469697, "step": 10 }, { "epoch": 0.06968641114982578, "grad_norm": 669.195296069125, "learning_rate": 2.413793103448276e-07, "logits/chosen": -2.606231927871704, "logits/rejected": -2.565000534057617, "logps/chosen": -104.05134582519531, "logps/rejected": -94.89227294921875, "loss": 0.6861, "rewards/accuracies": 0.3687500059604645, "rewards/chosen": 0.07496137917041779, "rewards/margins": 0.07313241064548492, "rewards/rejected": 0.001828978187404573, "step": 20 }, { "epoch": 0.10452961672473868, "grad_norm": 701.7098593476545, "learning_rate": 3.620689655172414e-07, "logits/chosen": -2.5953564643859863, "logits/rejected": -2.575517177581787, "logps/chosen": -82.11582946777344, "logps/rejected": -91.40339660644531, "loss": 0.6768, "rewards/accuracies": 0.36250001192092896, "rewards/chosen": 0.37693265080451965, "rewards/margins": 0.24151858687400818, "rewards/rejected": 0.1354140341281891, "step": 30 }, { "epoch": 0.13937282229965156, "grad_norm": 569.2211096554255, "learning_rate": 4.827586206896552e-07, "logits/chosen": -2.49927020072937, "logits/rejected": -2.497345209121704, "logps/chosen": -77.96027374267578, "logps/rejected": -73.55770111083984, "loss": 0.6699, "rewards/accuracies": 0.3187499940395355, "rewards/chosen": -0.04143521189689636, "rewards/margins": 0.5093806982040405, "rewards/rejected": -0.5508158802986145, "step": 40 }, { "epoch": 0.17421602787456447, "grad_norm": 446.25211605228685, "learning_rate": 6.034482758620689e-07, "logits/chosen": -2.5248429775238037, "logits/rejected": -2.529026508331299, "logps/chosen": -63.092262268066406, "logps/rejected": -75.61325073242188, "loss": 0.7196, "rewards/accuracies": 0.30000001192092896, "rewards/chosen": 0.838420033454895, "rewards/margins": 0.35312479734420776, "rewards/rejected": 0.48529529571533203, "step": 50 }, { "epoch": 0.20905923344947736, "grad_norm": 431.8823983807605, "learning_rate": 6.999740526496426e-07, "logits/chosen": -2.494865894317627, "logits/rejected": -2.4892258644104004, "logps/chosen": -70.98899841308594, "logps/rejected": -66.5857925415039, "loss": 0.7101, "rewards/accuracies": 0.32499998807907104, "rewards/chosen": 2.1953351497650146, "rewards/margins": 0.3907918632030487, "rewards/rejected": 1.8045432567596436, "step": 60 }, { "epoch": 0.24390243902439024, "grad_norm": 543.5545103197261, "learning_rate": 6.990662992822431e-07, "logits/chosen": -2.5150294303894043, "logits/rejected": -2.510296583175659, "logps/chosen": -61.68668746948242, "logps/rejected": -66.4362564086914, "loss": 0.7767, "rewards/accuracies": 0.3062500059604645, "rewards/chosen": 2.524014949798584, "rewards/margins": 0.5535615086555481, "rewards/rejected": 1.9704535007476807, "step": 70 }, { "epoch": 0.2787456445993031, "grad_norm": 602.8845854238214, "learning_rate": 6.96865023062192e-07, "logits/chosen": -2.461080551147461, "logits/rejected": -2.452082395553589, "logps/chosen": -73.27593231201172, "logps/rejected": -75.97522735595703, "loss": 0.8282, "rewards/accuracies": 0.3125, "rewards/chosen": 2.365041732788086, "rewards/margins": 0.8407068252563477, "rewards/rejected": 1.5243349075317383, "step": 80 }, { "epoch": 0.313588850174216, "grad_norm": 645.4166740504302, "learning_rate": 6.93378381182268e-07, "logits/chosen": -2.5054688453674316, "logits/rejected": -2.5200397968292236, "logps/chosen": -63.54323196411133, "logps/rejected": -68.08236694335938, "loss": 0.9178, "rewards/accuracies": 0.29374998807907104, "rewards/chosen": 2.4146523475646973, "rewards/margins": 0.40329408645629883, "rewards/rejected": 2.0113582611083984, "step": 90 }, { "epoch": 0.34843205574912894, "grad_norm": 625.9456943514227, "learning_rate": 6.886192939700987e-07, "logits/chosen": -2.495720863342285, "logits/rejected": -2.494643449783325, "logps/chosen": -72.901611328125, "logps/rejected": -79.84459686279297, "loss": 0.9118, "rewards/accuracies": 0.32499998807907104, "rewards/chosen": 2.345865488052368, "rewards/margins": 1.1508872509002686, "rewards/rejected": 1.1949782371520996, "step": 100 }, { "epoch": 0.34843205574912894, "eval_logits/chosen": -2.572685718536377, "eval_logits/rejected": -2.556438684463501, "eval_logps/chosen": -72.59263610839844, "eval_logps/rejected": -79.98140716552734, "eval_loss": 0.8951926827430725, "eval_rewards/accuracies": 0.335317462682724, "eval_rewards/chosen": 1.8796159029006958, "eval_rewards/margins": 0.7441409230232239, "eval_rewards/rejected": 1.1354750394821167, "eval_runtime": 114.849, "eval_samples_per_second": 17.414, "eval_steps_per_second": 0.549, "step": 100 }, { "epoch": 0.3832752613240418, "grad_norm": 633.8111608364293, "learning_rate": 6.826053970097538e-07, "logits/chosen": -2.5071868896484375, "logits/rejected": -2.4732370376586914, "logps/chosen": -72.26206970214844, "logps/rejected": -62.43015670776367, "loss": 0.934, "rewards/accuracies": 0.2562499940395355, "rewards/chosen": 0.8988760709762573, "rewards/margins": 0.3242531418800354, "rewards/rejected": 0.5746229887008667, "step": 110 }, { "epoch": 0.4181184668989547, "grad_norm": 481.08142593629555, "learning_rate": 6.753589757901721e-07, "logits/chosen": -2.5381650924682617, "logits/rejected": -2.50740385055542, "logps/chosen": -76.40122985839844, "logps/rejected": -66.65751647949219, "loss": 0.857, "rewards/accuracies": 0.29374998807907104, "rewards/chosen": 1.0497384071350098, "rewards/margins": 1.0358079671859741, "rewards/rejected": 0.013930544257164001, "step": 120 }, { "epoch": 0.4529616724738676, "grad_norm": 807.6631711479599, "learning_rate": 6.669068831226014e-07, "logits/chosen": -2.5767252445220947, "logits/rejected": -2.557955265045166, "logps/chosen": -83.18122863769531, "logps/rejected": -88.09697723388672, "loss": 1.0372, "rewards/accuracies": 0.33125001192092896, "rewards/chosen": 0.6925557255744934, "rewards/margins": 1.275132417678833, "rewards/rejected": -0.5825767517089844, "step": 130 }, { "epoch": 0.4878048780487805, "grad_norm": 443.55201800181277, "learning_rate": 6.572804396330676e-07, "logits/chosen": -2.4845736026763916, "logits/rejected": -2.4748897552490234, "logps/chosen": -79.59736633300781, "logps/rejected": -70.54890441894531, "loss": 0.9077, "rewards/accuracies": 0.33125001192092896, "rewards/chosen": 1.5325819253921509, "rewards/margins": 1.3193892240524292, "rewards/rejected": 0.21319285035133362, "step": 140 }, { "epoch": 0.5226480836236934, "grad_norm": 574.2676129428694, "learning_rate": 6.465153176986211e-07, "logits/chosen": -2.579017162322998, "logits/rejected": -2.5382275581359863, "logps/chosen": -77.61248016357422, "logps/rejected": -78.82749938964844, "loss": 1.0493, "rewards/accuracies": 0.2750000059604645, "rewards/chosen": 1.853137731552124, "rewards/margins": 1.137075662612915, "rewards/rejected": 0.7160621881484985, "step": 150 }, { "epoch": 0.5574912891986062, "grad_norm": 575.7733117089178, "learning_rate": 6.346514092574479e-07, "logits/chosen": -2.573914051055908, "logits/rejected": -2.5928311347961426, "logps/chosen": -62.48719024658203, "logps/rejected": -70.96321105957031, "loss": 0.9487, "rewards/accuracies": 0.3062500059604645, "rewards/chosen": 2.017285108566284, "rewards/margins": 1.0176817178726196, "rewards/rejected": 0.9996035695075989, "step": 160 }, { "epoch": 0.5923344947735192, "grad_norm": 636.1057741626599, "learning_rate": 6.21732677982701e-07, "logits/chosen": -2.5915422439575195, "logits/rejected": -2.5760223865509033, "logps/chosen": -67.14725494384766, "logps/rejected": -75.51747131347656, "loss": 0.9285, "rewards/accuracies": 0.26875001192092896, "rewards/chosen": 1.8019440174102783, "rewards/margins": 1.0133386850357056, "rewards/rejected": 0.7886053919792175, "step": 170 }, { "epoch": 0.627177700348432, "grad_norm": 634.0125212661374, "learning_rate": 6.078069963678453e-07, "logits/chosen": -2.6139795780181885, "logits/rejected": -2.60251522064209, "logps/chosen": -89.45849609375, "logps/rejected": -84.91694641113281, "loss": 1.1004, "rewards/accuracies": 0.3375000059604645, "rewards/chosen": 2.3746800422668457, "rewards/margins": 0.797105073928833, "rewards/rejected": 1.577574610710144, "step": 180 }, { "epoch": 0.662020905923345, "grad_norm": 494.843810287881, "learning_rate": 5.929259683272219e-07, "logits/chosen": -2.602254629135132, "logits/rejected": -2.594606399536133, "logps/chosen": -68.91534423828125, "logps/rejected": -80.00617218017578, "loss": 0.9944, "rewards/accuracies": 0.28125, "rewards/chosen": 2.114778518676758, "rewards/margins": 0.607982873916626, "rewards/rejected": 1.5067954063415527, "step": 190 }, { "epoch": 0.6968641114982579, "grad_norm": 826.9190409978122, "learning_rate": 5.771447379692167e-07, "logits/chosen": -2.625168800354004, "logits/rejected": -2.6301324367523193, "logps/chosen": -87.10128784179688, "logps/rejected": -90.02182006835938, "loss": 0.9553, "rewards/accuracies": 0.3687500059604645, "rewards/chosen": 3.519533157348633, "rewards/margins": 1.5109440088272095, "rewards/rejected": 2.008589029312134, "step": 200 }, { "epoch": 0.6968641114982579, "eval_logits/chosen": -2.58929705619812, "eval_logits/rejected": -2.5725698471069336, "eval_logps/chosen": -71.86607360839844, "eval_logps/rejected": -79.71355438232422, "eval_loss": 1.070008635520935, "eval_rewards/accuracies": 0.341269850730896, "eval_rewards/chosen": 2.5989110469818115, "eval_rewards/margins": 1.1982702016830444, "eval_rewards/rejected": 1.4006409645080566, "eval_runtime": 113.335, "eval_samples_per_second": 17.647, "eval_steps_per_second": 0.556, "step": 200 }, { "epoch": 0.7317073170731707, "grad_norm": 847.0633281542522, "learning_rate": 5.605217852506545e-07, "logits/chosen": -2.6012160778045654, "logits/rejected": -2.57619047164917, "logps/chosen": -67.00978088378906, "logps/rejected": -62.69084548950195, "loss": 0.9355, "rewards/accuracies": 0.3687500059604645, "rewards/chosen": 2.286196231842041, "rewards/margins": 1.579746961593628, "rewards/rejected": 0.7064491510391235, "step": 210 }, { "epoch": 0.7665505226480837, "grad_norm": 558.5058882421968, "learning_rate": 5.43118709269656e-07, "logits/chosen": -2.652629852294922, "logits/rejected": -2.6338040828704834, "logps/chosen": -71.07757568359375, "logps/rejected": -69.96295928955078, "loss": 1.0861, "rewards/accuracies": 0.2562499940395355, "rewards/chosen": 2.3567943572998047, "rewards/margins": 0.6735709309577942, "rewards/rejected": 1.6832237243652344, "step": 220 }, { "epoch": 0.8013937282229965, "grad_norm": 874.1019734865492, "learning_rate": 5.25e-07, "logits/chosen": -2.675318956375122, "logits/rejected": -2.6557412147521973, "logps/chosen": -86.0287857055664, "logps/rejected": -86.36713409423828, "loss": 1.1663, "rewards/accuracies": 0.3375000059604645, "rewards/chosen": 3.9321389198303223, "rewards/margins": 2.168241024017334, "rewards/rejected": 1.7638976573944092, "step": 230 }, { "epoch": 0.8362369337979094, "grad_norm": 612.8944076163772, "learning_rate": 5.062327993128697e-07, "logits/chosen": -2.6777350902557373, "logits/rejected": -2.6439225673675537, "logps/chosen": -82.02351379394531, "logps/rejected": -76.44059753417969, "loss": 1.1645, "rewards/accuracies": 0.33125001192092896, "rewards/chosen": 4.670910835266113, "rewards/margins": 1.0123672485351562, "rewards/rejected": 3.658543825149536, "step": 240 }, { "epoch": 0.8710801393728222, "grad_norm": 661.3902849256391, "learning_rate": 4.868866521715546e-07, "logits/chosen": -2.6785645484924316, "logits/rejected": -2.642632007598877, "logps/chosen": -91.0850601196289, "logps/rejected": -86.79383850097656, "loss": 0.8849, "rewards/accuracies": 0.3687500059604645, "rewards/chosen": 4.320570468902588, "rewards/margins": 1.1997525691986084, "rewards/rejected": 3.1208178997039795, "step": 250 }, { "epoch": 0.9059233449477352, "grad_norm": 559.1976202405948, "learning_rate": 4.6703324892109645e-07, "logits/chosen": -2.567991256713867, "logits/rejected": -2.581616163253784, "logps/chosen": -56.56006622314453, "logps/rejected": -64.0355224609375, "loss": 1.1214, "rewards/accuracies": 0.2874999940395355, "rewards/chosen": 2.46440052986145, "rewards/margins": 0.8875689506530762, "rewards/rejected": 1.5768316984176636, "step": 260 }, { "epoch": 0.9407665505226481, "grad_norm": 923.7220287977552, "learning_rate": 4.4674615962787004e-07, "logits/chosen": -2.647984504699707, "logits/rejected": -2.646916151046753, "logps/chosen": -66.90876770019531, "logps/rejected": -81.64918518066406, "loss": 1.0676, "rewards/accuracies": 0.3125, "rewards/chosen": 1.9516960382461548, "rewards/margins": 1.0779374837875366, "rewards/rejected": 0.8737583160400391, "step": 270 }, { "epoch": 0.975609756097561, "grad_norm": 647.9814126875335, "learning_rate": 4.2610056145354496e-07, "logits/chosen": -2.538573741912842, "logits/rejected": -2.5176804065704346, "logps/chosen": -65.74501037597656, "logps/rejected": -70.30267333984375, "loss": 0.862, "rewards/accuracies": 0.3187499940395355, "rewards/chosen": 2.068479299545288, "rewards/margins": 1.5420581102371216, "rewards/rejected": 0.5264211893081665, "step": 280 }, { "epoch": 1.0104529616724738, "grad_norm": 56.46750622929306, "learning_rate": 4.051729600736907e-07, "logits/chosen": -2.5648231506347656, "logits/rejected": -2.536726474761963, "logps/chosen": -68.04057312011719, "logps/rejected": -65.41937255859375, "loss": 0.7918, "rewards/accuracies": 0.38749998807907104, "rewards/chosen": 3.9868416786193848, "rewards/margins": 4.487275123596191, "rewards/rejected": -0.5004340410232544, "step": 290 }, { "epoch": 1.0452961672473868, "grad_norm": 20.388910811138857, "learning_rate": 3.8404090617335413e-07, "logits/chosen": -2.5940048694610596, "logits/rejected": -2.579204559326172, "logps/chosen": -59.690216064453125, "logps/rejected": -76.28305053710938, "loss": 0.4066, "rewards/accuracies": 0.41874998807907104, "rewards/chosen": 6.790108680725098, "rewards/margins": 13.831901550292969, "rewards/rejected": -7.0417938232421875, "step": 300 }, { "epoch": 1.0452961672473868, "eval_logits/chosen": -2.612644672393799, "eval_logits/rejected": -2.5961754322052, "eval_logps/chosen": -72.1514663696289, "eval_logps/rejected": -80.20658111572266, "eval_loss": 1.0729384422302246, "eval_rewards/accuracies": 0.3432539701461792, "eval_rewards/chosen": 2.3163692951202393, "eval_rewards/margins": 1.4038203954696655, "eval_rewards/rejected": 0.9125491380691528, "eval_runtime": 113.475, "eval_samples_per_second": 17.625, "eval_steps_per_second": 0.555, "step": 300 }, { "epoch": 1.0801393728222997, "grad_norm": 3.5419086603813197, "learning_rate": 3.6278270807018065e-07, "logits/chosen": -2.558955430984497, "logits/rejected": -2.5596249103546143, "logps/chosen": -59.938392639160156, "logps/rejected": -84.96260833740234, "loss": 0.416, "rewards/accuracies": 0.4375, "rewards/chosen": 7.621278285980225, "rewards/margins": 17.239816665649414, "rewards/rejected": -9.618535995483398, "step": 310 }, { "epoch": 1.1149825783972125, "grad_norm": 201.12315966640935, "learning_rate": 3.414771415300036e-07, "logits/chosen": -2.5916085243225098, "logits/rejected": -2.5778439044952393, "logps/chosen": -66.45283508300781, "logps/rejected": -87.47081756591797, "loss": 0.3907, "rewards/accuracies": 0.4625000059604645, "rewards/chosen": 7.431988716125488, "rewards/margins": 16.59490203857422, "rewards/rejected": -9.16291332244873, "step": 320 }, { "epoch": 1.1498257839721253, "grad_norm": 403.8100928464834, "learning_rate": 3.2020315785022746e-07, "logits/chosen": -2.5842907428741455, "logits/rejected": -2.5574951171875, "logps/chosen": -75.70329284667969, "logps/rejected": -86.71253967285156, "loss": 0.3965, "rewards/accuracies": 0.518750011920929, "rewards/chosen": 8.334002494812012, "rewards/margins": 16.338943481445312, "rewards/rejected": -8.004940032958984, "step": 330 }, { "epoch": 1.1846689895470384, "grad_norm": 184.77353798917835, "learning_rate": 2.9903959129274836e-07, "logits/chosen": -2.561433792114258, "logits/rejected": -2.5655293464660645, "logps/chosen": -74.55548095703125, "logps/rejected": -104.60441589355469, "loss": 0.4148, "rewards/accuracies": 0.53125, "rewards/chosen": 9.100366592407227, "rewards/margins": 18.328720092773438, "rewards/rejected": -9.228353500366211, "step": 340 }, { "epoch": 1.2195121951219512, "grad_norm": 188.43635904114095, "learning_rate": 2.7806486695056977e-07, "logits/chosen": -2.5871026515960693, "logits/rejected": -2.55499529838562, "logps/chosen": -58.55460739135742, "logps/rejected": -74.59812927246094, "loss": 0.4111, "rewards/accuracies": 0.48750001192092896, "rewards/chosen": 9.941993713378906, "rewards/margins": 18.771812438964844, "rewards/rejected": -8.829817771911621, "step": 350 }, { "epoch": 1.254355400696864, "grad_norm": 154.26226580174057, "learning_rate": 2.573567101306622e-07, "logits/chosen": -2.5829975605010986, "logits/rejected": -2.5526790618896484, "logps/chosen": -62.894317626953125, "logps/rejected": -71.01216888427734, "loss": 0.4026, "rewards/accuracies": 0.4437499940395355, "rewards/chosen": 8.866216659545898, "rewards/margins": 14.160728454589844, "rewards/rejected": -5.294511795043945, "step": 360 }, { "epoch": 1.289198606271777, "grad_norm": 16.651956659184666, "learning_rate": 2.369918583299939e-07, "logits/chosen": -2.5568833351135254, "logits/rejected": -2.5733718872070312, "logps/chosen": -62.53009033203125, "logps/rejected": -82.81775665283203, "loss": 0.4444, "rewards/accuracies": 0.46875, "rewards/chosen": 8.450610160827637, "rewards/margins": 16.84958839416504, "rewards/rejected": -8.398977279663086, "step": 370 }, { "epoch": 1.32404181184669, "grad_norm": 210.98546985458898, "learning_rate": 2.1704577687205507e-07, "logits/chosen": -2.5712497234344482, "logits/rejected": -2.567377805709839, "logps/chosen": -76.60011291503906, "logps/rejected": -98.51103210449219, "loss": 0.4409, "rewards/accuracies": 0.53125, "rewards/chosen": 10.171670913696289, "rewards/margins": 22.095985412597656, "rewards/rejected": -11.924314498901367, "step": 380 }, { "epoch": 1.3588850174216027, "grad_norm": 125.98956132873472, "learning_rate": 1.975923792576331e-07, "logits/chosen": -2.6548500061035156, "logits/rejected": -2.6388492584228516, "logps/chosen": -57.238861083984375, "logps/rejected": -80.28170776367188, "loss": 0.3853, "rewards/accuracies": 0.4375, "rewards/chosen": 9.658943176269531, "rewards/margins": 18.589771270751953, "rewards/rejected": -8.930827140808105, "step": 390 }, { "epoch": 1.3937282229965158, "grad_norm": 30.8789509954886, "learning_rate": 1.7870375326612014e-07, "logits/chosen": -2.638012647628784, "logits/rejected": -2.609297275543213, "logps/chosen": -76.772705078125, "logps/rejected": -106.2076644897461, "loss": 0.3805, "rewards/accuracies": 0.518750011920929, "rewards/chosen": 9.280598640441895, "rewards/margins": 17.006460189819336, "rewards/rejected": -7.7258620262146, "step": 400 }, { "epoch": 1.3937282229965158, "eval_logits/chosen": -2.6412570476531982, "eval_logits/rejected": -2.6246657371520996, "eval_logps/chosen": -71.48371124267578, "eval_logps/rejected": -79.92253875732422, "eval_loss": 1.1545917987823486, "eval_rewards/accuracies": 0.3373015820980072, "eval_rewards/chosen": 2.977447986602783, "eval_rewards/margins": 1.783706545829773, "eval_rewards/rejected": 1.1937415599822998, "eval_runtime": 113.3185, "eval_samples_per_second": 17.649, "eval_steps_per_second": 0.556, "step": 400 }, { "epoch": 1.4285714285714286, "grad_norm": 134.60970325056522, "learning_rate": 1.604498938223354e-07, "logits/chosen": -2.6223552227020264, "logits/rejected": -2.611177444458008, "logps/chosen": -71.12557220458984, "logps/rejected": -88.45091247558594, "loss": 0.6988, "rewards/accuracies": 0.48124998807907104, "rewards/chosen": 11.372858047485352, "rewards/margins": 19.366708755493164, "rewards/rejected": -7.993849754333496, "step": 410 }, { "epoch": 1.4634146341463414, "grad_norm": 144.0362669522128, "learning_rate": 1.4289844361876528e-07, "logits/chosen": -2.665538787841797, "logits/rejected": -2.665926933288574, "logps/chosen": -67.73007202148438, "logps/rejected": -91.9347915649414, "loss": 0.394, "rewards/accuracies": 0.42500001192092896, "rewards/chosen": 8.27946662902832, "rewards/margins": 15.272089958190918, "rewards/rejected": -6.9926252365112305, "step": 420 }, { "epoch": 1.4982578397212545, "grad_norm": 8.116844029360541, "learning_rate": 1.2611444245438813e-07, "logits/chosen": -2.629135847091675, "logits/rejected": -2.6135053634643555, "logps/chosen": -59.4090576171875, "logps/rejected": -79.82469177246094, "loss": 0.3924, "rewards/accuracies": 0.46875, "rewards/chosen": 9.160750389099121, "rewards/margins": 16.42301368713379, "rewards/rejected": -7.262263298034668, "step": 430 }, { "epoch": 1.533101045296167, "grad_norm": 20.390949955841545, "learning_rate": 1.1016008621895228e-07, "logits/chosen": -2.6179585456848145, "logits/rejected": -2.6255526542663574, "logps/chosen": -59.94586944580078, "logps/rejected": -79.95707702636719, "loss": 0.385, "rewards/accuracies": 0.4312500059604645, "rewards/chosen": 9.932461738586426, "rewards/margins": 16.32512664794922, "rewards/rejected": -6.392666339874268, "step": 440 }, { "epoch": 1.5679442508710801, "grad_norm": 91.65774378764581, "learning_rate": 9.509449641582943e-08, "logits/chosen": -2.669267177581787, "logits/rejected": -2.6314785480499268, "logps/chosen": -78.78290557861328, "logps/rejected": -96.71044158935547, "loss": 0.4341, "rewards/accuracies": 0.518750011920929, "rewards/chosen": 11.938484191894531, "rewards/margins": 22.57853889465332, "rewards/rejected": -10.640054702758789, "step": 450 }, { "epoch": 1.6027874564459932, "grad_norm": 203.85164100757706, "learning_rate": 8.097350107751374e-08, "logits/chosen": -2.643629550933838, "logits/rejected": -2.621685743331909, "logps/chosen": -65.54093933105469, "logps/rejected": -86.33696746826172, "loss": 0.4012, "rewards/accuracies": 0.5, "rewards/chosen": 10.532175064086914, "rewards/margins": 19.771120071411133, "rewards/rejected": -9.238944053649902, "step": 460 }, { "epoch": 1.6376306620209058, "grad_norm": 0.18713678602063968, "learning_rate": 6.784942788562304e-08, "logits/chosen": -2.6330292224884033, "logits/rejected": -2.6224162578582764, "logps/chosen": -53.2932243347168, "logps/rejected": -78.58303833007812, "loss": 0.4188, "rewards/accuracies": 0.4000000059604645, "rewards/chosen": 7.929083347320557, "rewards/margins": 14.772134780883789, "rewards/rejected": -6.843050479888916, "step": 470 }, { "epoch": 1.6724738675958188, "grad_norm": 0.6147060015446919, "learning_rate": 5.5770910262027175e-08, "logits/chosen": -2.6437172889709473, "logits/rejected": -2.627516269683838, "logps/chosen": -48.605804443359375, "logps/rejected": -54.954551696777344, "loss": 0.3885, "rewards/accuracies": 0.36250001192092896, "rewards/chosen": 5.401150703430176, "rewards/margins": 8.344820022583008, "rewards/rejected": -2.943669319152832, "step": 480 }, { "epoch": 1.7073170731707317, "grad_norm": 77.39241567442491, "learning_rate": 4.47827071496673e-08, "logits/chosen": -2.630645275115967, "logits/rejected": -2.618098735809326, "logps/chosen": -65.59320831298828, "logps/rejected": -76.5845718383789, "loss": 0.4658, "rewards/accuracies": 0.40625, "rewards/chosen": 7.552302360534668, "rewards/margins": 14.009483337402344, "rewards/rejected": -6.457179069519043, "step": 490 }, { "epoch": 1.7421602787456445, "grad_norm": 9.754129842794342, "learning_rate": 3.492553715089692e-08, "logits/chosen": -2.5552730560302734, "logits/rejected": -2.546638011932373, "logps/chosen": -67.24018096923828, "logps/rejected": -92.03102111816406, "loss": 0.3975, "rewards/accuracies": 0.46875, "rewards/chosen": 8.957259178161621, "rewards/margins": 17.181909561157227, "rewards/rejected": -8.224650382995605, "step": 500 }, { "epoch": 1.7421602787456445, "eval_logits/chosen": -2.663222551345825, "eval_logits/rejected": -2.6463263034820557, "eval_logps/chosen": -71.91925048828125, "eval_logps/rejected": -80.58873748779297, "eval_loss": 1.182449460029602, "eval_rewards/accuracies": 0.3452380895614624, "eval_rewards/chosen": 2.54626727104187, "eval_rewards/margins": 2.0120491981506348, "eval_rewards/rejected": 0.5342182517051697, "eval_runtime": 113.333, "eval_samples_per_second": 17.647, "eval_steps_per_second": 0.556, "step": 500 }, { "epoch": 1.7770034843205575, "grad_norm": 9.598053942856874, "learning_rate": 2.6235927637971816e-08, "logits/chosen": -2.625732898712158, "logits/rejected": -2.6054017543792725, "logps/chosen": -61.70050048828125, "logps/rejected": -77.1646728515625, "loss": 0.3933, "rewards/accuracies": 0.4625000059604645, "rewards/chosen": 8.965360641479492, "rewards/margins": 16.470144271850586, "rewards/rejected": -7.504785060882568, "step": 510 }, { "epoch": 1.8118466898954704, "grad_norm": 19.555147588045187, "learning_rate": 1.8746079394836706e-08, "logits/chosen": -2.636565685272217, "logits/rejected": -2.6276707649230957, "logps/chosen": -65.6178207397461, "logps/rejected": -84.69874572753906, "loss": 0.4072, "rewards/accuracies": 0.4625000059604645, "rewards/chosen": 8.566861152648926, "rewards/margins": 15.977938652038574, "rewards/rejected": -7.411079406738281, "step": 520 }, { "epoch": 1.8466898954703832, "grad_norm": 45.12187365083394, "learning_rate": 1.2483747291799507e-08, "logits/chosen": -2.610792636871338, "logits/rejected": -2.603881597518921, "logps/chosen": -62.07392501831055, "logps/rejected": -80.65681457519531, "loss": 0.3989, "rewards/accuracies": 0.4625000059604645, "rewards/chosen": 9.169586181640625, "rewards/margins": 18.3698787689209, "rewards/rejected": -9.200292587280273, "step": 530 }, { "epoch": 1.8815331010452963, "grad_norm": 828.0040833812193, "learning_rate": 7.472137435272619e-09, "logits/chosen": -2.624800205230713, "logits/rejected": -2.638892889022827, "logps/chosen": -54.112327575683594, "logps/rejected": -85.64537048339844, "loss": 0.4706, "rewards/accuracies": 0.44999998807907104, "rewards/chosen": 9.522268295288086, "rewards/margins": 21.70328712463379, "rewards/rejected": -12.181015968322754, "step": 540 }, { "epoch": 1.916376306620209, "grad_norm": 6.099605163051641, "learning_rate": 3.729821173711411e-09, "logits/chosen": -2.6054625511169434, "logits/rejected": -2.585512161254883, "logps/chosen": -78.79656982421875, "logps/rejected": -100.56778717041016, "loss": 0.3741, "rewards/accuracies": 0.5, "rewards/chosen": 11.91783332824707, "rewards/margins": 25.462675094604492, "rewards/rejected": -13.544839859008789, "step": 550 }, { "epoch": 1.951219512195122, "grad_norm": 239.71729089566762, "learning_rate": 1.2706662784136513e-09, "logits/chosen": -2.601435422897339, "logits/rejected": -2.609788417816162, "logps/chosen": -56.35918045043945, "logps/rejected": -76.89906311035156, "loss": 0.4008, "rewards/accuracies": 0.4437499940395355, "rewards/chosen": 7.287221431732178, "rewards/margins": 13.824111938476562, "rewards/rejected": -6.536891937255859, "step": 560 }, { "epoch": 1.986062717770035, "grad_norm": 23.401979133014883, "learning_rate": 1.0378555420122448e-10, "logits/chosen": -2.6814956665039062, "logits/rejected": -2.6513266563415527, "logps/chosen": -57.64264678955078, "logps/rejected": -68.03236389160156, "loss": 0.3939, "rewards/accuracies": 0.4375, "rewards/chosen": 7.494875907897949, "rewards/margins": 12.909930229187012, "rewards/rejected": -5.4150543212890625, "step": 570 }, { "epoch": 2.0, "step": 574, "total_flos": 0.0, "train_loss": 0.6661292189920406, "train_runtime": 6507.4651, "train_samples_per_second": 5.637, "train_steps_per_second": 0.088 } ], "logging_steps": 10, "max_steps": 574, "num_input_tokens_seen": 0, "num_train_epochs": 2, "save_steps": 100, "stateful_callbacks": { "TrainerControl": { "args": { "should_epoch_stop": false, "should_evaluate": false, "should_log": false, "should_save": true, "should_training_stop": true }, "attributes": {} } }, "total_flos": 0.0, "train_batch_size": 8, "trial_name": null, "trial_params": null }