{ "best_global_step": null, "best_metric": null, "best_model_checkpoint": null, "epoch": 1.0, "eval_steps": 200, "global_step": 681, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0014684287812041115, "fcm_dpo/beta": 0.10443365573883057, "fcm_dpo/delta": 0.42479780316352844, "fcm_dpo/margin": -0.02287006378173828, "fcm_dpo/q_t": 0.5005706548690796, "grad_norm": 87.35734558105469, "learning_rate": 0.0, "logits/chosen": -0.4974287748336792, "logits/rejected": -0.43299180269241333, "logps/chosen": -50.1435661315918, "logps/ref_chosen": -50.14883804321289, "logps/ref_rejected": -74.1280517578125, "logps/rejected": -74.09991455078125, "loss": 1.389, "margin_dpo/margin_mean": -0.02287048101425171, "margin_dpo/margin_std": 0.41920793056488037, "step": 1 }, { "epoch": 0.002936857562408223, "fcm_dpo/beta": 0.10886730253696442, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.06572261452674866, "fcm_dpo/q_t": 0.5017880797386169, "grad_norm": 78.55793762207031, "learning_rate": 7.246376811594203e-09, "logits/chosen": -0.49536412954330444, "logits/rejected": -0.4594460427761078, "logps/chosen": -52.65568923950195, "logps/ref_chosen": -52.620704650878906, "logps/ref_rejected": -75.30413818359375, "logps/rejected": -75.27340698242188, "loss": 1.3938, "margin_dpo/margin_mean": -0.06572240591049194, "margin_dpo/margin_std": 0.35048407316207886, "step": 2 }, { "epoch": 0.004405286343612335, "fcm_dpo/beta": 0.11366599053144455, "fcm_dpo/delta": 0.4224254786968231, "fcm_dpo/margin": 0.009388208389282227, "fcm_dpo/q_t": 0.4997447431087494, "grad_norm": 80.87034606933594, "learning_rate": 1.4492753623188406e-08, "logits/chosen": -0.48167288303375244, "logits/rejected": -0.44224029779434204, "logps/chosen": -60.9686279296875, "logps/ref_chosen": -60.981597900390625, "logps/ref_rejected": -68.67259216308594, "logps/rejected": -68.66902160644531, "loss": 1.3853, "margin_dpo/margin_mean": 0.009388208389282227, "margin_dpo/margin_std": 0.282347708940506, "step": 3 }, { "epoch": 0.005873715124816446, "fcm_dpo/beta": 0.11846467852592468, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.04822826385498047, "fcm_dpo/q_t": 0.5014264583587646, "grad_norm": 85.6677474975586, "learning_rate": 2.1739130434782606e-08, "logits/chosen": -0.471701055765152, "logits/rejected": -0.44420742988586426, "logps/chosen": -56.783233642578125, "logps/ref_chosen": -56.7677116394043, "logps/ref_rejected": -86.64710998535156, "logps/rejected": -86.61439514160156, "loss": 1.3927, "margin_dpo/margin_mean": -0.04822826385498047, "margin_dpo/margin_std": 0.4498218595981598, "step": 4 }, { "epoch": 0.007342143906020558, "fcm_dpo/beta": 0.12882092595100403, "fcm_dpo/delta": 0.41904187202453613, "fcm_dpo/margin": 0.016975760459899902, "fcm_dpo/q_t": 0.4995834231376648, "grad_norm": 115.2800064086914, "learning_rate": 2.898550724637681e-08, "logits/chosen": -0.5144484043121338, "logits/rejected": -0.47074341773986816, "logps/chosen": -53.837677001953125, "logps/ref_chosen": -53.859375, "logps/ref_rejected": -84.14918518066406, "logps/rejected": -84.14445495605469, "loss": 1.3846, "margin_dpo/margin_mean": 0.01697593927383423, "margin_dpo/margin_std": 0.3396869897842407, "step": 5 }, { "epoch": 0.00881057268722467, "fcm_dpo/beta": 0.13447463512420654, "fcm_dpo/delta": 0.42067742347717285, "fcm_dpo/margin": 0.010400176048278809, "fcm_dpo/q_t": 0.49966543912887573, "grad_norm": 123.6097640991211, "learning_rate": 3.6231884057971014e-08, "logits/chosen": -0.5179563760757446, "logits/rejected": -0.4767148494720459, "logps/chosen": -63.031429290771484, "logps/ref_chosen": -63.007484436035156, "logps/ref_rejected": -92.64534759521484, "logps/rejected": -92.6796875, "loss": 1.3852, "margin_dpo/margin_mean": 0.010399848222732544, "margin_dpo/margin_std": 0.3767673969268799, "step": 6 }, { "epoch": 0.010279001468428781, "fcm_dpo/beta": 0.1462460160255432, "fcm_dpo/delta": 0.418557733297348, "fcm_dpo/margin": 0.0423530638217926, "fcm_dpo/q_t": 0.4985184967517853, "grad_norm": 120.26276397705078, "learning_rate": 4.347826086956521e-08, "logits/chosen": -0.5139331817626953, "logits/rejected": -0.4799633026123047, "logps/chosen": -57.72328186035156, "logps/ref_chosen": -57.774818420410156, "logps/ref_rejected": -103.92059326171875, "logps/rejected": -103.91140747070312, "loss": 1.3807, "margin_dpo/margin_mean": 0.042353659868240356, "margin_dpo/margin_std": 0.41376739740371704, "step": 7 }, { "epoch": 0.011747430249632892, "fcm_dpo/beta": 0.16579358279705048, "fcm_dpo/delta": 0.42236655950546265, "fcm_dpo/margin": -0.051773250102996826, "fcm_dpo/q_t": 0.5021986961364746, "grad_norm": 130.60256958007812, "learning_rate": 5.0724637681159424e-08, "logits/chosen": -0.502518892288208, "logits/rejected": -0.47607821226119995, "logps/chosen": -58.71660614013672, "logps/ref_chosen": -58.716033935546875, "logps/ref_rejected": -79.3114242553711, "logps/rejected": -79.26022338867188, "loss": 1.396, "margin_dpo/margin_mean": -0.05177316069602966, "margin_dpo/margin_std": 0.37757202982902527, "step": 8 }, { "epoch": 0.013215859030837005, "fcm_dpo/beta": 0.18036609888076782, "fcm_dpo/delta": 0.42122572660446167, "fcm_dpo/margin": -0.041811347007751465, "fcm_dpo/q_t": 0.5019670724868774, "grad_norm": 154.1094970703125, "learning_rate": 5.797101449275362e-08, "logits/chosen": -0.4861254096031189, "logits/rejected": -0.4390403628349304, "logps/chosen": -69.8838882446289, "logps/ref_chosen": -69.8668441772461, "logps/ref_rejected": -99.6026611328125, "logps/rejected": -99.57789611816406, "loss": 1.3954, "margin_dpo/margin_mean": -0.041812002658843994, "margin_dpo/margin_std": 0.42343536019325256, "step": 9 }, { "epoch": 0.014684287812041116, "fcm_dpo/beta": 0.20495112240314484, "fcm_dpo/delta": 0.8467953205108643, "fcm_dpo/margin": 0.017669588327407837, "fcm_dpo/q_t": 0.4991891384124756, "grad_norm": 147.3089141845703, "learning_rate": 6.521739130434782e-08, "logits/chosen": -0.4706907868385315, "logits/rejected": -0.4263506531715393, "logps/chosen": -48.373626708984375, "logps/ref_chosen": -48.35768508911133, "logps/ref_rejected": -80.37206268310547, "logps/rejected": -80.40567016601562, "loss": 1.3847, "margin_dpo/margin_mean": 0.017669767141342163, "margin_dpo/margin_std": 0.4244072437286377, "step": 10 }, { "epoch": 0.016152716593245228, "fcm_dpo/beta": 0.24274222552776337, "fcm_dpo/delta": 0.8439900875091553, "fcm_dpo/margin": 0.026336967945098877, "fcm_dpo/q_t": 0.49850064516067505, "grad_norm": 164.48699951171875, "learning_rate": 7.246376811594203e-08, "logits/chosen": -0.4563661217689514, "logits/rejected": -0.4311213493347168, "logps/chosen": -53.00157928466797, "logps/ref_chosen": -53.01685333251953, "logps/ref_rejected": -87.78038024902344, "logps/rejected": -87.79144287109375, "loss": 1.3814, "margin_dpo/margin_mean": 0.026336759328842163, "margin_dpo/margin_std": 0.3369394540786743, "step": 11 }, { "epoch": 0.01762114537444934, "fcm_dpo/beta": 0.26402783393859863, "fcm_dpo/delta": 0.42025870084762573, "fcm_dpo/margin": 0.01722872257232666, "fcm_dpo/q_t": 0.4989096224308014, "grad_norm": 249.8050079345703, "learning_rate": 7.971014492753623e-08, "logits/chosen": -0.5214487314224243, "logits/rejected": -0.48471778631210327, "logps/chosen": -61.81374740600586, "logps/ref_chosen": -61.80543518066406, "logps/ref_rejected": -104.8582763671875, "logps/rejected": -104.8838119506836, "loss": 1.3842, "margin_dpo/margin_mean": 0.01722833514213562, "margin_dpo/margin_std": 0.39332982897758484, "step": 12 }, { "epoch": 0.01908957415565345, "fcm_dpo/beta": 0.31265515089035034, "fcm_dpo/delta": 0.8470354080200195, "fcm_dpo/margin": 0.010488659143447876, "fcm_dpo/q_t": 0.499286413192749, "grad_norm": 244.24612426757812, "learning_rate": 8.695652173913042e-08, "logits/chosen": -0.4612182378768921, "logits/rejected": -0.4315667748451233, "logps/chosen": -64.26567840576172, "logps/ref_chosen": -64.2603530883789, "logps/ref_rejected": -87.20307922363281, "logps/rejected": -87.21888732910156, "loss": 1.3867, "margin_dpo/margin_mean": 0.010488033294677734, "margin_dpo/margin_std": 0.37841933965682983, "step": 13 }, { "epoch": 0.020558002936857563, "fcm_dpo/beta": 0.3400227725505829, "fcm_dpo/delta": 0.4154511094093323, "fcm_dpo/margin": 0.02040991187095642, "fcm_dpo/q_t": 0.4983346462249756, "grad_norm": 287.9231262207031, "learning_rate": 9.420289855072464e-08, "logits/chosen": -0.4573858082294464, "logits/rejected": -0.41736847162246704, "logps/chosen": -58.149322509765625, "logps/ref_chosen": -58.11021041870117, "logps/ref_rejected": -104.04708099365234, "logps/rejected": -104.10659790039062, "loss": 1.3835, "margin_dpo/margin_mean": 0.020410001277923584, "margin_dpo/margin_std": 0.3933468461036682, "step": 14 }, { "epoch": 0.022026431718061675, "fcm_dpo/beta": 0.3845078647136688, "fcm_dpo/delta": 0.4113464057445526, "fcm_dpo/margin": 0.021445751190185547, "fcm_dpo/q_t": 0.4981946051120758, "grad_norm": 246.7677764892578, "learning_rate": 1.0144927536231885e-07, "logits/chosen": -0.4819180965423584, "logits/rejected": -0.46261194348335266, "logps/chosen": -56.989723205566406, "logps/ref_chosen": -56.96691131591797, "logps/ref_rejected": -80.80863952636719, "logps/rejected": -80.8529052734375, "loss": 1.3822, "margin_dpo/margin_mean": 0.02144584059715271, "margin_dpo/margin_std": 0.3318941593170166, "step": 15 }, { "epoch": 0.023494860499265784, "fcm_dpo/beta": 0.4179382026195526, "fcm_dpo/delta": 0.4168465733528137, "fcm_dpo/margin": 0.020205020904541016, "fcm_dpo/q_t": 0.49806952476501465, "grad_norm": 354.9513244628906, "learning_rate": 1.0869565217391303e-07, "logits/chosen": -0.5375505089759827, "logits/rejected": -0.49767178297042847, "logps/chosen": -61.75188446044922, "logps/ref_chosen": -61.739891052246094, "logps/ref_rejected": -84.36947631835938, "logps/rejected": -84.40167236328125, "loss": 1.384, "margin_dpo/margin_mean": 0.020204484462738037, "margin_dpo/margin_std": 0.374006986618042, "step": 16 }, { "epoch": 0.024963289280469897, "fcm_dpo/beta": 0.47306138277053833, "fcm_dpo/delta": 0.8088477849960327, "fcm_dpo/margin": 0.0909477174282074, "fcm_dpo/q_t": 0.4897420406341553, "grad_norm": 359.2418212890625, "learning_rate": 1.1594202898550725e-07, "logits/chosen": -0.49428272247314453, "logits/rejected": -0.4564746022224426, "logps/chosen": -67.70065307617188, "logps/ref_chosen": -67.71033477783203, "logps/ref_rejected": -85.37865447998047, "logps/rejected": -85.45992279052734, "loss": 1.3526, "margin_dpo/margin_mean": 0.09094801545143127, "margin_dpo/margin_std": 0.41880565881729126, "step": 17 }, { "epoch": 0.02643171806167401, "fcm_dpo/beta": 0.49132367968559265, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.024234741926193237, "fcm_dpo/q_t": 0.5029305219650269, "grad_norm": 411.5271911621094, "learning_rate": 1.2318840579710146e-07, "logits/chosen": -0.5078892707824707, "logits/rejected": -0.45399364829063416, "logps/chosen": -47.75193786621094, "logps/ref_chosen": -47.7394905090332, "logps/ref_rejected": -75.4722900390625, "logps/rejected": -75.46050262451172, "loss": 1.4031, "margin_dpo/margin_mean": -0.02423509955406189, "margin_dpo/margin_std": 0.28412166237831116, "step": 18 }, { "epoch": 0.027900146842878122, "fcm_dpo/beta": 0.5553874969482422, "fcm_dpo/delta": 0.800998330116272, "fcm_dpo/margin": 0.09300807118415833, "fcm_dpo/q_t": 0.48788678646087646, "grad_norm": 400.787109375, "learning_rate": 1.3043478260869563e-07, "logits/chosen": -0.48504626750946045, "logits/rejected": -0.43480148911476135, "logps/chosen": -70.17445373535156, "logps/ref_chosen": -70.20536041259766, "logps/ref_rejected": -89.7575912475586, "logps/rejected": -89.81969451904297, "loss": 1.3413, "margin_dpo/margin_mean": 0.09300819039344788, "margin_dpo/margin_std": 0.2757360339164734, "step": 19 }, { "epoch": 0.02936857562408223, "fcm_dpo/beta": 0.649980366230011, "fcm_dpo/delta": 0.7827930450439453, "fcm_dpo/margin": 0.10994091629981995, "fcm_dpo/q_t": 0.48339736461639404, "grad_norm": 451.47674560546875, "learning_rate": 1.3768115942028986e-07, "logits/chosen": -0.5489249229431152, "logits/rejected": -0.49235957860946655, "logps/chosen": -50.79475784301758, "logps/ref_chosen": -50.80324172973633, "logps/ref_rejected": -78.82334899902344, "logps/rejected": -78.9248046875, "loss": 1.3281, "margin_dpo/margin_mean": 0.10994073748588562, "margin_dpo/margin_std": 0.3477732241153717, "step": 20 }, { "epoch": 0.030837004405286344, "fcm_dpo/beta": 0.7579153776168823, "fcm_dpo/delta": 0.7742947340011597, "fcm_dpo/margin": 0.1082988828420639, "fcm_dpo/q_t": 0.48166587948799133, "grad_norm": 542.7781982421875, "learning_rate": 1.4492753623188405e-07, "logits/chosen": -0.48926520347595215, "logits/rejected": -0.4650164246559143, "logps/chosen": -50.03901672363281, "logps/ref_chosen": -50.063018798828125, "logps/ref_rejected": -77.86878967285156, "logps/rejected": -77.95309448242188, "loss": 1.3293, "margin_dpo/margin_mean": 0.10829953849315643, "margin_dpo/margin_std": 0.39389097690582275, "step": 21 }, { "epoch": 0.032305433186490456, "fcm_dpo/beta": 0.8798083066940308, "fcm_dpo/delta": 0.7186980247497559, "fcm_dpo/margin": 0.15809443593025208, "fcm_dpo/q_t": 0.46808773279190063, "grad_norm": 664.1295776367188, "learning_rate": 1.5217391304347825e-07, "logits/chosen": -0.5107871294021606, "logits/rejected": -0.4682098627090454, "logps/chosen": -59.02461242675781, "logps/ref_chosen": -59.05763626098633, "logps/ref_rejected": -97.50466918945312, "logps/rejected": -97.62974548339844, "loss": 1.2817, "margin_dpo/margin_mean": 0.15809422731399536, "margin_dpo/margin_std": 0.4001656174659729, "step": 22 }, { "epoch": 0.033773861967694566, "fcm_dpo/beta": 1.0026525259017944, "fcm_dpo/delta": 0.6618263721466064, "fcm_dpo/margin": 0.20202824473381042, "fcm_dpo/q_t": 0.4552893042564392, "grad_norm": 672.5951538085938, "learning_rate": 1.5942028985507245e-07, "logits/chosen": -0.5025913119316101, "logits/rejected": -0.4807313084602356, "logps/chosen": -60.07830810546875, "logps/ref_chosen": -60.07769775390625, "logps/ref_rejected": -81.13955688476562, "logps/rejected": -81.34219360351562, "loss": 1.231, "margin_dpo/margin_mean": 0.20202842354774475, "margin_dpo/margin_std": 0.3648604154586792, "step": 23 }, { "epoch": 0.03524229074889868, "fcm_dpo/beta": 1.1451796293258667, "fcm_dpo/delta": 0.6144854426383972, "fcm_dpo/margin": 0.21480031311511993, "fcm_dpo/q_t": 0.4435741901397705, "grad_norm": 806.4075317382812, "learning_rate": 1.6666666666666665e-07, "logits/chosen": -0.4962637424468994, "logits/rejected": -0.4799681305885315, "logps/chosen": -44.290122985839844, "logps/ref_chosen": -44.29103469848633, "logps/ref_rejected": -99.12521362304688, "logps/rejected": -99.33910369873047, "loss": 1.1955, "margin_dpo/margin_mean": 0.21480104327201843, "margin_dpo/margin_std": 0.34698599576950073, "step": 24 }, { "epoch": 0.03671071953010279, "fcm_dpo/beta": 1.2869091033935547, "fcm_dpo/delta": 0.5991071462631226, "fcm_dpo/margin": 0.20611488819122314, "fcm_dpo/q_t": 0.44144344329833984, "grad_norm": 790.001953125, "learning_rate": 1.7391304347826085e-07, "logits/chosen": -0.5114257335662842, "logits/rejected": -0.48204663395881653, "logps/chosen": -52.55467987060547, "logps/ref_chosen": -52.537052154541016, "logps/ref_rejected": -89.34219360351562, "logps/rejected": -89.56593322753906, "loss": 1.2231, "margin_dpo/margin_mean": 0.2061150074005127, "margin_dpo/margin_std": 0.4720209240913391, "step": 25 }, { "epoch": 0.0381791483113069, "fcm_dpo/beta": 1.4049187898635864, "fcm_dpo/delta": 0.3693993091583252, "fcm_dpo/margin": 0.3534828722476959, "fcm_dpo/q_t": 0.4011384844779968, "grad_norm": 799.3506469726562, "learning_rate": 1.8115942028985507e-07, "logits/chosen": -0.5231348872184753, "logits/rejected": -0.4912256896495819, "logps/chosen": -53.86709976196289, "logps/ref_chosen": -53.92280578613281, "logps/ref_rejected": -103.35971069335938, "logps/rejected": -103.65748596191406, "loss": 1.0901, "margin_dpo/margin_mean": 0.3534834384918213, "margin_dpo/margin_std": 0.576106071472168, "step": 26 }, { "epoch": 0.039647577092511016, "fcm_dpo/beta": 1.4509608745574951, "fcm_dpo/delta": 0.16326822340488434, "fcm_dpo/margin": 0.4813149869441986, "fcm_dpo/q_t": 0.3541679382324219, "grad_norm": 837.9793701171875, "learning_rate": 1.8840579710144927e-07, "logits/chosen": -0.5251492857933044, "logits/rejected": -0.4889242649078369, "logps/chosen": -42.80194854736328, "logps/ref_chosen": -42.898529052734375, "logps/ref_rejected": -98.72419738769531, "logps/rejected": -99.10894012451172, "loss": 0.9227, "margin_dpo/margin_mean": 0.4813147783279419, "margin_dpo/margin_std": 0.5135878324508667, "step": 27 }, { "epoch": 0.041116005873715125, "fcm_dpo/beta": 1.5709292888641357, "fcm_dpo/delta": 0.3916556239128113, "fcm_dpo/margin": 0.3012993633747101, "fcm_dpo/q_t": 0.4064401686191559, "grad_norm": 838.3126831054688, "learning_rate": 1.9565217391304347e-07, "logits/chosen": -0.5148231983184814, "logits/rejected": -0.4603886008262634, "logps/chosen": -60.546630859375, "logps/ref_chosen": -60.55650329589844, "logps/ref_rejected": -91.40111541748047, "logps/rejected": -91.69254302978516, "loss": 1.1735, "margin_dpo/margin_mean": 0.301299124956131, "margin_dpo/margin_std": 0.6170388460159302, "step": 28 }, { "epoch": 0.042584434654919234, "fcm_dpo/beta": 1.6120238304138184, "fcm_dpo/delta": 0.06577476114034653, "fcm_dpo/margin": 0.4894725978374481, "fcm_dpo/q_t": 0.33954018354415894, "grad_norm": 836.0704345703125, "learning_rate": 2.028985507246377e-07, "logits/chosen": -0.5353908538818359, "logits/rejected": -0.4876295328140259, "logps/chosen": -57.748104095458984, "logps/ref_chosen": -57.80778503417969, "logps/ref_rejected": -97.39434814453125, "logps/rejected": -97.82414245605469, "loss": 0.8943, "margin_dpo/margin_mean": 0.48947349190711975, "margin_dpo/margin_std": 0.5322099328041077, "step": 29 }, { "epoch": 0.04405286343612335, "fcm_dpo/beta": 1.5611000061035156, "fcm_dpo/delta": -0.16747647523880005, "fcm_dpo/margin": 0.6395827531814575, "fcm_dpo/q_t": 0.2977033853530884, "grad_norm": 659.2262573242188, "learning_rate": 2.1014492753623187e-07, "logits/chosen": -0.47992774844169617, "logits/rejected": -0.44857877492904663, "logps/chosen": -52.428070068359375, "logps/ref_chosen": -52.577369689941406, "logps/ref_rejected": -98.48920440673828, "logps/rejected": -98.97948455810547, "loss": 0.7823, "margin_dpo/margin_mean": 0.6395825147628784, "margin_dpo/margin_std": 0.5751562118530273, "step": 30 }, { "epoch": 0.04552129221732746, "fcm_dpo/beta": 1.6095309257507324, "fcm_dpo/delta": 0.14502495527267456, "fcm_dpo/margin": 0.44233372807502747, "fcm_dpo/q_t": 0.36425676941871643, "grad_norm": 802.032470703125, "learning_rate": 2.1739130434782607e-07, "logits/chosen": -0.4975692331790924, "logits/rejected": -0.45189863443374634, "logps/chosen": -63.71034240722656, "logps/ref_chosen": -63.806922912597656, "logps/ref_rejected": -72.89400482177734, "logps/rejected": -73.23976135253906, "loss": 1.0302, "margin_dpo/margin_mean": 0.4423332214355469, "margin_dpo/margin_std": 0.6822764873504639, "step": 31 }, { "epoch": 0.04698972099853157, "fcm_dpo/beta": 1.5345242023468018, "fcm_dpo/delta": -0.2706429362297058, "fcm_dpo/margin": 0.7098908424377441, "fcm_dpo/q_t": 0.3092593550682068, "grad_norm": 571.6924438476562, "learning_rate": 2.2463768115942027e-07, "logits/chosen": -0.5193390250205994, "logits/rejected": -0.4783915877342224, "logps/chosen": -62.573360443115234, "logps/ref_chosen": -62.739524841308594, "logps/ref_rejected": -89.3175048828125, "logps/rejected": -89.86123657226562, "loss": 0.8387, "margin_dpo/margin_mean": 0.7098907232284546, "margin_dpo/margin_std": 0.8541468977928162, "step": 32 }, { "epoch": 0.048458149779735685, "fcm_dpo/beta": 1.5268961191177368, "fcm_dpo/delta": -0.031100787222385406, "fcm_dpo/margin": 0.5745024681091309, "fcm_dpo/q_t": 0.3211558759212494, "grad_norm": 612.5335083007812, "learning_rate": 2.318840579710145e-07, "logits/chosen": -0.4916543960571289, "logits/rejected": -0.46555328369140625, "logps/chosen": -53.1591682434082, "logps/ref_chosen": -53.26097106933594, "logps/ref_rejected": -87.8851318359375, "logps/rejected": -88.35783386230469, "loss": 0.8401, "margin_dpo/margin_mean": 0.5745028257369995, "margin_dpo/margin_std": 0.5594164133071899, "step": 33 }, { "epoch": 0.049926578560939794, "fcm_dpo/beta": 1.4713330268859863, "fcm_dpo/delta": -0.17497381567955017, "fcm_dpo/margin": 0.6845256090164185, "fcm_dpo/q_t": 0.3139664828777313, "grad_norm": 577.4857177734375, "learning_rate": 2.391304347826087e-07, "logits/chosen": -0.5157313346862793, "logits/rejected": -0.4982021450996399, "logps/chosen": -50.735015869140625, "logps/ref_chosen": -50.81732940673828, "logps/ref_rejected": -101.92184448242188, "logps/rejected": -102.52406311035156, "loss": 0.842, "margin_dpo/margin_mean": 0.6845252513885498, "margin_dpo/margin_std": 0.780299723148346, "step": 34 }, { "epoch": 0.0513950073421439, "fcm_dpo/beta": 1.321608066558838, "fcm_dpo/delta": -0.6159881353378296, "fcm_dpo/margin": 1.0402474403381348, "fcm_dpo/q_t": 0.25291207432746887, "grad_norm": 438.6875305175781, "learning_rate": 2.463768115942029e-07, "logits/chosen": -0.5221713185310364, "logits/rejected": -0.4853857457637787, "logps/chosen": -50.92929458618164, "logps/ref_chosen": -51.02449035644531, "logps/ref_rejected": -106.82443237304688, "logps/rejected": -107.76948547363281, "loss": 0.665, "margin_dpo/margin_mean": 1.0402469635009766, "margin_dpo/margin_std": 1.0385708808898926, "step": 35 }, { "epoch": 0.05286343612334802, "fcm_dpo/beta": 1.225428581237793, "fcm_dpo/delta": -0.465666800737381, "fcm_dpo/margin": 1.0272314548492432, "fcm_dpo/q_t": 0.2840074300765991, "grad_norm": 377.8875427246094, "learning_rate": 2.536231884057971e-07, "logits/chosen": -0.5505253672599792, "logits/rejected": -0.5139901041984558, "logps/chosen": -51.958282470703125, "logps/ref_chosen": -51.991493225097656, "logps/ref_rejected": -86.0406265258789, "logps/rejected": -87.03465270996094, "loss": 0.8189, "margin_dpo/margin_mean": 1.0272307395935059, "margin_dpo/margin_std": 1.1449289321899414, "step": 36 }, { "epoch": 0.05433186490455213, "fcm_dpo/beta": 1.1077017784118652, "fcm_dpo/delta": -0.27637869119644165, "fcm_dpo/margin": 0.9796682596206665, "fcm_dpo/q_t": 0.3130841851234436, "grad_norm": 408.67047119140625, "learning_rate": 2.6086956521739126e-07, "logits/chosen": -0.5258996486663818, "logits/rejected": -0.4838007390499115, "logps/chosen": -62.77622985839844, "logps/ref_chosen": -62.807106018066406, "logps/ref_rejected": -77.89507293701172, "logps/rejected": -78.84385681152344, "loss": 0.8676, "margin_dpo/margin_mean": 0.979668140411377, "margin_dpo/margin_std": 1.212325096130371, "step": 37 }, { "epoch": 0.055800293685756244, "fcm_dpo/beta": 1.0282557010650635, "fcm_dpo/delta": -0.5638899803161621, "fcm_dpo/margin": 1.3030188083648682, "fcm_dpo/q_t": 0.2778816223144531, "grad_norm": 352.4019470214844, "learning_rate": 2.681159420289855e-07, "logits/chosen": -0.5080125331878662, "logits/rejected": -0.4751642942428589, "logps/chosen": -48.19729995727539, "logps/ref_chosen": -48.39051818847656, "logps/ref_rejected": -97.91244506835938, "logps/rejected": -99.0222396850586, "loss": 0.7363, "margin_dpo/margin_mean": 1.3030191659927368, "margin_dpo/margin_std": 1.503594160079956, "step": 38 }, { "epoch": 0.05726872246696035, "fcm_dpo/beta": 0.9014551639556885, "fcm_dpo/delta": -0.5705999732017517, "fcm_dpo/margin": 1.47984778881073, "fcm_dpo/q_t": 0.24894997477531433, "grad_norm": 300.4083557128906, "learning_rate": 2.753623188405797e-07, "logits/chosen": -0.5485913753509521, "logits/rejected": -0.5085880756378174, "logps/chosen": -50.66078567504883, "logps/ref_chosen": -50.75047302246094, "logps/ref_rejected": -78.56951141357422, "logps/rejected": -79.95967102050781, "loss": 0.6737, "margin_dpo/margin_mean": 1.4798476696014404, "margin_dpo/margin_std": 1.253267765045166, "step": 39 }, { "epoch": 0.05873715124816446, "fcm_dpo/beta": 0.8415946960449219, "fcm_dpo/delta": -0.3833761215209961, "fcm_dpo/margin": 1.4126371145248413, "fcm_dpo/q_t": 0.28941503167152405, "grad_norm": 265.9231872558594, "learning_rate": 2.8260869565217386e-07, "logits/chosen": -0.5170412659645081, "logits/rejected": -0.48721107840538025, "logps/chosen": -57.76616668701172, "logps/ref_chosen": -57.985069274902344, "logps/ref_rejected": -74.3000717163086, "logps/rejected": -75.49380493164062, "loss": 0.7687, "margin_dpo/margin_mean": 1.4126379489898682, "margin_dpo/margin_std": 1.5114383697509766, "step": 40 }, { "epoch": 0.06020558002936858, "fcm_dpo/beta": 0.7668871879577637, "fcm_dpo/delta": -0.47359752655029297, "fcm_dpo/margin": 1.648345947265625, "fcm_dpo/q_t": 0.2783043086528778, "grad_norm": 277.421875, "learning_rate": 2.898550724637681e-07, "logits/chosen": -0.5229369401931763, "logits/rejected": -0.48551464080810547, "logps/chosen": -62.65602111816406, "logps/ref_chosen": -62.69581604003906, "logps/ref_rejected": -97.02352905273438, "logps/rejected": -98.632080078125, "loss": 0.7602, "margin_dpo/margin_mean": 1.6483460664749146, "margin_dpo/margin_std": 1.7497031688690186, "step": 41 }, { "epoch": 0.06167400881057269, "fcm_dpo/beta": 0.6656206846237183, "fcm_dpo/delta": -0.7854231595993042, "fcm_dpo/margin": 2.276841640472412, "fcm_dpo/q_t": 0.23248319327831268, "grad_norm": 194.76841735839844, "learning_rate": 2.971014492753623e-07, "logits/chosen": -0.5424157381057739, "logits/rejected": -0.4958062767982483, "logps/chosen": -58.732154846191406, "logps/ref_chosen": -58.966426849365234, "logps/ref_rejected": -109.90837097167969, "logps/rejected": -111.95094299316406, "loss": 0.6171, "margin_dpo/margin_mean": 2.276841640472412, "margin_dpo/margin_std": 2.000026226043701, "step": 42 }, { "epoch": 0.0631424375917768, "fcm_dpo/beta": 0.5898722410202026, "fcm_dpo/delta": -0.4751095175743103, "fcm_dpo/margin": 2.1357011795043945, "fcm_dpo/q_t": 0.25335824489593506, "grad_norm": 190.30642700195312, "learning_rate": 3.043478260869565e-07, "logits/chosen": -0.5324772596359253, "logits/rejected": -0.5075618028640747, "logps/chosen": -53.679229736328125, "logps/ref_chosen": -54.15599822998047, "logps/ref_rejected": -96.48019409179688, "logps/rejected": -98.13912963867188, "loss": 0.6504, "margin_dpo/margin_mean": 2.1357011795043945, "margin_dpo/margin_std": 1.690740942955017, "step": 43 }, { "epoch": 0.06461086637298091, "fcm_dpo/beta": 0.5374739170074463, "fcm_dpo/delta": -0.4890367388725281, "fcm_dpo/margin": 2.3702030181884766, "fcm_dpo/q_t": 0.2507287263870239, "grad_norm": 187.52468872070312, "learning_rate": 3.115942028985507e-07, "logits/chosen": -0.4761464595794678, "logits/rejected": -0.4557410180568695, "logps/chosen": -49.860252380371094, "logps/ref_chosen": -50.07849884033203, "logps/ref_rejected": -108.78376007080078, "logps/rejected": -110.93571472167969, "loss": 0.6411, "margin_dpo/margin_mean": 2.370201587677002, "margin_dpo/margin_std": 1.8710044622421265, "step": 44 }, { "epoch": 0.06607929515418502, "fcm_dpo/beta": 0.5190870761871338, "fcm_dpo/delta": -0.12950783967971802, "fcm_dpo/margin": 1.863038420677185, "fcm_dpo/q_t": 0.3222770094871521, "grad_norm": 173.381591796875, "learning_rate": 3.188405797101449e-07, "logits/chosen": -0.4617079496383667, "logits/rejected": -0.44868040084838867, "logps/chosen": -48.283241271972656, "logps/ref_chosen": -48.4149284362793, "logps/ref_rejected": -77.93643188476562, "logps/rejected": -79.66777038574219, "loss": 0.8516, "margin_dpo/margin_mean": 1.8630379438400269, "margin_dpo/margin_std": 2.1737613677978516, "step": 45 }, { "epoch": 0.06754772393538913, "fcm_dpo/beta": 0.4841097593307495, "fcm_dpo/delta": -0.41270509362220764, "fcm_dpo/margin": 2.5065417289733887, "fcm_dpo/q_t": 0.28057238459587097, "grad_norm": 169.7103271484375, "learning_rate": 3.260869565217391e-07, "logits/chosen": -0.5448036193847656, "logits/rejected": -0.49467170238494873, "logps/chosen": -55.76255798339844, "logps/ref_chosen": -55.999427795410156, "logps/ref_rejected": -95.652587890625, "logps/rejected": -97.92225646972656, "loss": 0.7446, "margin_dpo/margin_mean": 2.5065412521362305, "margin_dpo/margin_std": 2.575545072555542, "step": 46 }, { "epoch": 0.06901615271659324, "fcm_dpo/beta": 0.4579503536224365, "fcm_dpo/delta": -0.2642349898815155, "fcm_dpo/margin": 2.3719077110290527, "fcm_dpo/q_t": 0.28744781017303467, "grad_norm": 176.13291931152344, "learning_rate": 3.333333333333333e-07, "logits/chosen": -0.5746684074401855, "logits/rejected": -0.5221650004386902, "logps/chosen": -57.57725524902344, "logps/ref_chosen": -57.92607879638672, "logps/ref_rejected": -94.67920684814453, "logps/rejected": -96.7022933959961, "loss": 0.7478, "margin_dpo/margin_mean": 2.371908664703369, "margin_dpo/margin_std": 2.1393775939941406, "step": 47 }, { "epoch": 0.07048458149779736, "fcm_dpo/beta": 0.43097007274627686, "fcm_dpo/delta": -0.18657633662223816, "fcm_dpo/margin": 2.351602554321289, "fcm_dpo/q_t": 0.3005352318286896, "grad_norm": 185.70761108398438, "learning_rate": 3.4057971014492755e-07, "logits/chosen": -0.5795384645462036, "logits/rejected": -0.5214025378227234, "logps/chosen": -57.07599639892578, "logps/ref_chosen": -57.188072204589844, "logps/ref_rejected": -88.0166015625, "logps/rejected": -90.25611877441406, "loss": 0.7955, "margin_dpo/margin_mean": 2.3516030311584473, "margin_dpo/margin_std": 2.1867825984954834, "step": 48 }, { "epoch": 0.07195301027900147, "fcm_dpo/beta": 0.4181467294692993, "fcm_dpo/delta": -0.271043062210083, "fcm_dpo/margin": 2.6118998527526855, "fcm_dpo/q_t": 0.29078733921051025, "grad_norm": 488.25555419921875, "learning_rate": 3.478260869565217e-07, "logits/chosen": -0.5379482507705688, "logits/rejected": -0.4796925187110901, "logps/chosen": -61.225616455078125, "logps/ref_chosen": -61.685272216796875, "logps/ref_rejected": -83.76747131347656, "logps/rejected": -85.91972351074219, "loss": 0.7584, "margin_dpo/margin_mean": 2.6118998527526855, "margin_dpo/margin_std": 2.511692523956299, "step": 49 }, { "epoch": 0.07342143906020558, "fcm_dpo/beta": 0.3943514823913574, "fcm_dpo/delta": -0.21341389417648315, "fcm_dpo/margin": 2.6372203826904297, "fcm_dpo/q_t": 0.29721564054489136, "grad_norm": 156.1033935546875, "learning_rate": 3.5507246376811595e-07, "logits/chosen": -0.5170506834983826, "logits/rejected": -0.48113885521888733, "logps/chosen": -58.57736587524414, "logps/ref_chosen": -58.72413635253906, "logps/ref_rejected": -96.35814666748047, "logps/rejected": -98.84859466552734, "loss": 0.7786, "margin_dpo/margin_mean": 2.6372203826904297, "margin_dpo/margin_std": 2.533358097076416, "step": 50 }, { "epoch": 0.07488986784140969, "fcm_dpo/beta": 0.37727904319763184, "fcm_dpo/delta": -0.17445003986358643, "fcm_dpo/margin": 2.6572084426879883, "fcm_dpo/q_t": 0.317116379737854, "grad_norm": 127.68961334228516, "learning_rate": 3.6231884057971015e-07, "logits/chosen": -0.49076852202415466, "logits/rejected": -0.45782989263534546, "logps/chosen": -61.265625, "logps/ref_chosen": -61.3736686706543, "logps/ref_rejected": -76.00199890136719, "logps/rejected": -78.55116271972656, "loss": 0.847, "margin_dpo/margin_mean": 2.6572086811065674, "margin_dpo/margin_std": 3.0995936393737793, "step": 51 }, { "epoch": 0.0763582966226138, "fcm_dpo/beta": 0.34055250883102417, "fcm_dpo/delta": -0.6958788633346558, "fcm_dpo/margin": 4.237062931060791, "fcm_dpo/q_t": 0.2313612997531891, "grad_norm": 108.22584533691406, "learning_rate": 3.695652173913043e-07, "logits/chosen": -0.5450096130371094, "logits/rejected": -0.49051666259765625, "logps/chosen": -51.71245574951172, "logps/ref_chosen": -52.33735656738281, "logps/ref_rejected": -79.97391510009766, "logps/rejected": -83.58607482910156, "loss": 0.6041, "margin_dpo/margin_mean": 4.237063407897949, "margin_dpo/margin_std": 3.3426589965820312, "step": 52 }, { "epoch": 0.07782672540381791, "fcm_dpo/beta": 0.3099779486656189, "fcm_dpo/delta": -0.5153226256370544, "fcm_dpo/margin": 4.196630477905273, "fcm_dpo/q_t": 0.26962772011756897, "grad_norm": 118.41160583496094, "learning_rate": 3.7681159420289855e-07, "logits/chosen": -0.5767320394515991, "logits/rejected": -0.554436206817627, "logps/chosen": -53.15708541870117, "logps/ref_chosen": -53.31465148925781, "logps/ref_rejected": -91.78359985351562, "logps/rejected": -95.82264709472656, "loss": 0.6986, "margin_dpo/margin_mean": 4.196630477905273, "margin_dpo/margin_std": 3.9024457931518555, "step": 53 }, { "epoch": 0.07929515418502203, "fcm_dpo/beta": 0.284817636013031, "fcm_dpo/delta": -0.3066610097885132, "fcm_dpo/margin": 3.9426732063293457, "fcm_dpo/q_t": 0.2827852964401245, "grad_norm": 96.0743408203125, "learning_rate": 3.8405797101449274e-07, "logits/chosen": -0.5662685632705688, "logits/rejected": -0.511724591255188, "logps/chosen": -50.537540435791016, "logps/ref_chosen": -50.68865966796875, "logps/ref_rejected": -91.71539306640625, "logps/rejected": -95.50694274902344, "loss": 0.7202, "margin_dpo/margin_mean": 3.9426727294921875, "margin_dpo/margin_std": 3.563432216644287, "step": 54 }, { "epoch": 0.08076358296622614, "fcm_dpo/beta": 0.2634790539741516, "fcm_dpo/delta": -0.4253869354724884, "fcm_dpo/margin": 4.647494792938232, "fcm_dpo/q_t": 0.2821628153324127, "grad_norm": 104.46064758300781, "learning_rate": 3.9130434782608694e-07, "logits/chosen": -0.6163734197616577, "logits/rejected": -0.5525354146957397, "logps/chosen": -62.8038330078125, "logps/ref_chosen": -62.615234375, "logps/ref_rejected": -88.99349975585938, "logps/rejected": -93.82960510253906, "loss": 0.7725, "margin_dpo/margin_mean": 4.647494316101074, "margin_dpo/margin_std": 4.973275184631348, "step": 55 }, { "epoch": 0.08223201174743025, "fcm_dpo/beta": 0.24703602492809296, "fcm_dpo/delta": -0.26557958126068115, "fcm_dpo/margin": 4.4003424644470215, "fcm_dpo/q_t": 0.2998681366443634, "grad_norm": 82.96701049804688, "learning_rate": 3.9855072463768114e-07, "logits/chosen": -0.5948492288589478, "logits/rejected": -0.5514296293258667, "logps/chosen": -57.949623107910156, "logps/ref_chosen": -57.9327278137207, "logps/ref_rejected": -94.1744384765625, "logps/rejected": -98.5916748046875, "loss": 0.8196, "margin_dpo/margin_mean": 4.4003424644470215, "margin_dpo/margin_std": 4.814118385314941, "step": 56 }, { "epoch": 0.08370044052863436, "fcm_dpo/beta": 0.2332727015018463, "fcm_dpo/delta": -0.33255159854888916, "fcm_dpo/margin": 4.910983085632324, "fcm_dpo/q_t": 0.2743523120880127, "grad_norm": 88.67000579833984, "learning_rate": 4.057971014492754e-07, "logits/chosen": -0.5736434459686279, "logits/rejected": -0.5462496280670166, "logps/chosen": -70.63133239746094, "logps/ref_chosen": -70.49528503417969, "logps/ref_rejected": -95.56546020507812, "logps/rejected": -100.61249542236328, "loss": 0.7252, "margin_dpo/margin_mean": 4.910983085632324, "margin_dpo/margin_std": 4.256160259246826, "step": 57 }, { "epoch": 0.08516886930983847, "fcm_dpo/beta": 0.21495123207569122, "fcm_dpo/delta": -0.4436984658241272, "fcm_dpo/margin": 5.7713303565979, "fcm_dpo/q_t": 0.27471083402633667, "grad_norm": 93.05653381347656, "learning_rate": 4.1304347826086954e-07, "logits/chosen": -0.5842028856277466, "logits/rejected": -0.5084757804870605, "logps/chosen": -62.119361877441406, "logps/ref_chosen": -62.13294219970703, "logps/ref_rejected": -84.61729431152344, "logps/rejected": -90.37504577636719, "loss": 0.7351, "margin_dpo/margin_mean": 5.771330833435059, "margin_dpo/margin_std": 5.583444595336914, "step": 58 }, { "epoch": 0.08663729809104258, "fcm_dpo/beta": 0.19481170177459717, "fcm_dpo/delta": -0.34936845302581787, "fcm_dpo/margin": 5.904120445251465, "fcm_dpo/q_t": 0.2848263382911682, "grad_norm": 84.72233581542969, "learning_rate": 4.2028985507246374e-07, "logits/chosen": -0.608730673789978, "logits/rejected": -0.5659762620925903, "logps/chosen": -52.416080474853516, "logps/ref_chosen": -51.932525634765625, "logps/ref_rejected": -88.88520050048828, "logps/rejected": -95.27287292480469, "loss": 0.755, "margin_dpo/margin_mean": 5.904120922088623, "margin_dpo/margin_std": 5.746884822845459, "step": 59 }, { "epoch": 0.0881057268722467, "fcm_dpo/beta": 0.18991079926490784, "fcm_dpo/delta": -0.10931669175624847, "fcm_dpo/margin": 4.990189075469971, "fcm_dpo/q_t": 0.30916330218315125, "grad_norm": 100.4244384765625, "learning_rate": 4.2753623188405794e-07, "logits/chosen": -0.573798418045044, "logits/rejected": -0.51241534948349, "logps/chosen": -62.01226806640625, "logps/ref_chosen": -60.94218826293945, "logps/ref_rejected": -85.39340209960938, "logps/rejected": -91.45367431640625, "loss": 0.8378, "margin_dpo/margin_mean": 4.9901885986328125, "margin_dpo/margin_std": 5.157164096832275, "step": 60 }, { "epoch": 0.08957415565345081, "fcm_dpo/beta": 0.1817345768213272, "fcm_dpo/delta": -0.2543361485004425, "fcm_dpo/margin": 5.912901401519775, "fcm_dpo/q_t": 0.3096533417701721, "grad_norm": 76.46479797363281, "learning_rate": 4.3478260869565214e-07, "logits/chosen": -0.5715320110321045, "logits/rejected": -0.5351488590240479, "logps/chosen": -60.89082336425781, "logps/ref_chosen": -60.633522033691406, "logps/ref_rejected": -89.85249328613281, "logps/rejected": -96.02268981933594, "loss": 0.8382, "margin_dpo/margin_mean": 5.912901401519775, "margin_dpo/margin_std": 7.353564262390137, "step": 61 }, { "epoch": 0.09104258443465492, "fcm_dpo/beta": 0.18260258436203003, "fcm_dpo/delta": 0.0254787877202034, "fcm_dpo/margin": 4.519835948944092, "fcm_dpo/q_t": 0.33831268548965454, "grad_norm": 84.54349517822266, "learning_rate": 4.420289855072464e-07, "logits/chosen": -0.5869123935699463, "logits/rejected": -0.5528802871704102, "logps/chosen": -56.695011138916016, "logps/ref_chosen": -56.15077209472656, "logps/ref_rejected": -75.56619262695312, "logps/rejected": -80.6302719116211, "loss": 0.9, "margin_dpo/margin_mean": 4.51983642578125, "margin_dpo/margin_std": 5.208561420440674, "step": 62 }, { "epoch": 0.09251101321585903, "fcm_dpo/beta": 0.17352905869483948, "fcm_dpo/delta": -0.23942901194095612, "fcm_dpo/margin": 6.1104888916015625, "fcm_dpo/q_t": 0.29208725690841675, "grad_norm": 78.32554626464844, "learning_rate": 4.4927536231884053e-07, "logits/chosen": -0.5659127235412598, "logits/rejected": -0.5172122716903687, "logps/chosen": -74.04725646972656, "logps/ref_chosen": -73.14739227294922, "logps/ref_rejected": -97.61006164550781, "logps/rejected": -104.62040710449219, "loss": 0.7887, "margin_dpo/margin_mean": 6.1104888916015625, "margin_dpo/margin_std": 6.0076375007629395, "step": 63 }, { "epoch": 0.09397944199706314, "fcm_dpo/beta": 0.1625455915927887, "fcm_dpo/delta": -0.30817508697509766, "fcm_dpo/margin": 6.8527960777282715, "fcm_dpo/q_t": 0.29219383001327515, "grad_norm": 72.54730987548828, "learning_rate": 4.5652173913043473e-07, "logits/chosen": -0.540681004524231, "logits/rejected": -0.5078280568122864, "logps/chosen": -54.01070022583008, "logps/ref_chosen": -53.998600006103516, "logps/ref_rejected": -93.53019714355469, "logps/rejected": -100.39509582519531, "loss": 0.793, "margin_dpo/margin_mean": 6.8527960777282715, "margin_dpo/margin_std": 7.245349884033203, "step": 64 }, { "epoch": 0.09544787077826726, "fcm_dpo/beta": 0.15612056851387024, "fcm_dpo/delta": -0.23279529809951782, "fcm_dpo/margin": 6.764092445373535, "fcm_dpo/q_t": 0.2945548892021179, "grad_norm": 71.34774780273438, "learning_rate": 4.63768115942029e-07, "logits/chosen": -0.6489189267158508, "logits/rejected": -0.6330606937408447, "logps/chosen": -66.12747192382812, "logps/ref_chosen": -64.83599853515625, "logps/ref_rejected": -109.94645690917969, "logps/rejected": -118.00201416015625, "loss": 0.8093, "margin_dpo/margin_mean": 6.764091968536377, "margin_dpo/margin_std": 6.935907363891602, "step": 65 }, { "epoch": 0.09691629955947137, "fcm_dpo/beta": 0.15022742748260498, "fcm_dpo/delta": -0.14811818301677704, "fcm_dpo/margin": 6.522555828094482, "fcm_dpo/q_t": 0.31489285826683044, "grad_norm": 69.30376434326172, "learning_rate": 4.7101449275362313e-07, "logits/chosen": -0.6159607768058777, "logits/rejected": -0.5809246301651001, "logps/chosen": -52.61880874633789, "logps/ref_chosen": -51.44352722167969, "logps/ref_rejected": -75.63629913330078, "logps/rejected": -83.33413696289062, "loss": 0.834, "margin_dpo/margin_mean": 6.522555351257324, "margin_dpo/margin_std": 7.281524658203125, "step": 66 }, { "epoch": 0.09838472834067548, "fcm_dpo/beta": 0.14798378944396973, "fcm_dpo/delta": -0.15271174907684326, "fcm_dpo/margin": 6.6755170822143555, "fcm_dpo/q_t": 0.3114186227321625, "grad_norm": 61.069061279296875, "learning_rate": 4.782608695652174e-07, "logits/chosen": -0.5994589328765869, "logits/rejected": -0.5600971579551697, "logps/chosen": -60.29442596435547, "logps/ref_chosen": -59.34080505371094, "logps/ref_rejected": -72.78728485107422, "logps/rejected": -80.41642761230469, "loss": 0.8244, "margin_dpo/margin_mean": 6.6755170822143555, "margin_dpo/margin_std": 7.188636779785156, "step": 67 }, { "epoch": 0.09985315712187959, "fcm_dpo/beta": 0.1445780098438263, "fcm_dpo/delta": -0.11112671345472336, "fcm_dpo/margin": 6.57466983795166, "fcm_dpo/q_t": 0.3090569078922272, "grad_norm": 61.32022476196289, "learning_rate": 4.855072463768116e-07, "logits/chosen": -0.5993345975875854, "logits/rejected": -0.5429648160934448, "logps/chosen": -66.08113861083984, "logps/ref_chosen": -65.2058334350586, "logps/ref_rejected": -77.20724487304688, "logps/rejected": -84.6572265625, "loss": 0.795, "margin_dpo/margin_mean": 6.57466983795166, "margin_dpo/margin_std": 6.135951042175293, "step": 68 }, { "epoch": 0.1013215859030837, "fcm_dpo/beta": 0.1381378173828125, "fcm_dpo/delta": -0.18166013062000275, "fcm_dpo/margin": 7.318739891052246, "fcm_dpo/q_t": 0.2995903491973877, "grad_norm": 63.2430419921875, "learning_rate": 4.927536231884058e-07, "logits/chosen": -0.560828685760498, "logits/rejected": -0.5357776880264282, "logps/chosen": -61.07904052734375, "logps/ref_chosen": -59.81924057006836, "logps/ref_rejected": -103.38886260986328, "logps/rejected": -111.9674072265625, "loss": 0.7665, "margin_dpo/margin_mean": 7.3187408447265625, "margin_dpo/margin_std": 6.837162017822266, "step": 69 }, { "epoch": 0.1027900146842878, "fcm_dpo/beta": 0.13328900933265686, "fcm_dpo/delta": -0.2611111104488373, "fcm_dpo/margin": 8.128886222839355, "fcm_dpo/q_t": 0.295282244682312, "grad_norm": 66.4734878540039, "learning_rate": 5e-07, "logits/chosen": -0.5893011689186096, "logits/rejected": -0.5513046383857727, "logps/chosen": -63.8671875, "logps/ref_chosen": -61.930641174316406, "logps/ref_rejected": -91.06078338623047, "logps/rejected": -101.12621307373047, "loss": 0.7692, "margin_dpo/margin_mean": 8.128886222839355, "margin_dpo/margin_std": 8.00094223022461, "step": 70 }, { "epoch": 0.10425844346549193, "fcm_dpo/beta": 0.12505091726779938, "fcm_dpo/delta": -0.3048865795135498, "fcm_dpo/margin": 8.960905075073242, "fcm_dpo/q_t": 0.2825569808483124, "grad_norm": 55.4888916015625, "learning_rate": 4.999967061337492e-07, "logits/chosen": -0.6552280187606812, "logits/rejected": -0.612133800983429, "logps/chosen": -63.16753005981445, "logps/ref_chosen": -61.750335693359375, "logps/ref_rejected": -97.33662414550781, "logps/rejected": -107.71471405029297, "loss": 0.7315, "margin_dpo/margin_mean": 8.960905075073242, "margin_dpo/margin_std": 8.088345527648926, "step": 71 }, { "epoch": 0.10572687224669604, "fcm_dpo/beta": 0.11747831106185913, "fcm_dpo/delta": -0.285390704870224, "fcm_dpo/margin": 9.382757186889648, "fcm_dpo/q_t": 0.2830852270126343, "grad_norm": 57.36223220825195, "learning_rate": 4.999868246217933e-07, "logits/chosen": -0.5827009677886963, "logits/rejected": -0.5448130369186401, "logps/chosen": -67.41600036621094, "logps/ref_chosen": -66.05341339111328, "logps/ref_rejected": -95.2869873046875, "logps/rejected": -106.03233337402344, "loss": 0.7334, "margin_dpo/margin_mean": 9.382757186889648, "margin_dpo/margin_std": 8.36812686920166, "step": 72 }, { "epoch": 0.10719530102790015, "fcm_dpo/beta": 0.11263042688369751, "fcm_dpo/delta": -0.17705786228179932, "fcm_dpo/margin": 8.946794509887695, "fcm_dpo/q_t": 0.3216911852359772, "grad_norm": 58.959449768066406, "learning_rate": 4.999703557245192e-07, "logits/chosen": -0.6371798515319824, "logits/rejected": -0.5916515588760376, "logps/chosen": -68.6202163696289, "logps/ref_chosen": -66.25627136230469, "logps/ref_rejected": -90.45613098144531, "logps/rejected": -101.76687622070312, "loss": 0.8956, "margin_dpo/margin_mean": 8.946794509887695, "margin_dpo/margin_std": 11.44607162475586, "step": 73 }, { "epoch": 0.10866372980910426, "fcm_dpo/beta": 0.10893216729164124, "fcm_dpo/delta": -0.17043882608413696, "fcm_dpo/margin": 9.200440406799316, "fcm_dpo/q_t": 0.31555283069610596, "grad_norm": 60.90262985229492, "learning_rate": 4.999472998758977e-07, "logits/chosen": -0.5984748601913452, "logits/rejected": -0.5802183151245117, "logps/chosen": -56.02476501464844, "logps/ref_chosen": -53.42488098144531, "logps/ref_rejected": -95.94693756103516, "logps/rejected": -107.74725341796875, "loss": 0.8747, "margin_dpo/margin_mean": 9.200439453125, "margin_dpo/margin_std": 12.053499221801758, "step": 74 }, { "epoch": 0.11013215859030837, "fcm_dpo/beta": 0.10171889513731003, "fcm_dpo/delta": -0.4963496923446655, "fcm_dpo/margin": 12.63138484954834, "fcm_dpo/q_t": 0.26315972208976746, "grad_norm": 49.437259674072266, "learning_rate": 4.999176576834721e-07, "logits/chosen": -0.6258114576339722, "logits/rejected": -0.6118018627166748, "logps/chosen": -54.22317886352539, "logps/ref_chosen": -51.861663818359375, "logps/ref_rejected": -111.25398254394531, "logps/rejected": -126.24687957763672, "loss": 0.6895, "margin_dpo/margin_mean": 12.631382942199707, "margin_dpo/margin_std": 11.423229217529297, "step": 75 }, { "epoch": 0.11160058737151249, "fcm_dpo/beta": 0.09800372272729874, "fcm_dpo/delta": -0.010731710121035576, "fcm_dpo/margin": 8.773109436035156, "fcm_dpo/q_t": 0.32228267192840576, "grad_norm": 49.78368377685547, "learning_rate": 4.998814299283415e-07, "logits/chosen": -0.6634424924850464, "logits/rejected": -0.6184474229812622, "logps/chosen": -56.04176330566406, "logps/ref_chosen": -53.26603698730469, "logps/ref_rejected": -78.21662902832031, "logps/rejected": -89.76544952392578, "loss": 0.8528, "margin_dpo/margin_mean": 8.773109436035156, "margin_dpo/margin_std": 9.061115264892578, "step": 76 }, { "epoch": 0.1130690161527166, "fcm_dpo/beta": 0.09110400080680847, "fcm_dpo/delta": -0.42337319254875183, "fcm_dpo/margin": 13.34306526184082, "fcm_dpo/q_t": 0.2679826617240906, "grad_norm": 54.54016876220703, "learning_rate": 4.998386175651409e-07, "logits/chosen": -0.6144517660140991, "logits/rejected": -0.5647158026695251, "logps/chosen": -60.33265686035156, "logps/ref_chosen": -58.0966796875, "logps/ref_rejected": -93.77361297607422, "logps/rejected": -109.3526611328125, "loss": 0.7215, "margin_dpo/margin_mean": 13.34306526184082, "margin_dpo/margin_std": 12.149935722351074, "step": 77 }, { "epoch": 0.1145374449339207, "fcm_dpo/beta": 0.08875822275876999, "fcm_dpo/delta": -0.14152249693870544, "fcm_dpo/margin": 11.015653610229492, "fcm_dpo/q_t": 0.30442285537719727, "grad_norm": 50.21555709838867, "learning_rate": 4.997892217220159e-07, "logits/chosen": -0.6342002153396606, "logits/rejected": -0.6031632423400879, "logps/chosen": -57.952491760253906, "logps/ref_chosen": -55.61378479003906, "logps/ref_rejected": -84.93436431884766, "logps/rejected": -98.28872680664062, "loss": 0.7928, "margin_dpo/margin_mean": 11.015655517578125, "margin_dpo/margin_std": 10.317201614379883, "step": 78 }, { "epoch": 0.11600587371512482, "fcm_dpo/beta": 0.08457382023334503, "fcm_dpo/delta": -0.2228245735168457, "fcm_dpo/margin": 12.404184341430664, "fcm_dpo/q_t": 0.3089568018913269, "grad_norm": 40.83265686035156, "learning_rate": 4.997332437005931e-07, "logits/chosen": -0.619921088218689, "logits/rejected": -0.5860031247138977, "logps/chosen": -57.76416778564453, "logps/ref_chosen": -55.45048522949219, "logps/ref_rejected": -87.64756774902344, "logps/rejected": -102.36543273925781, "loss": 0.8187, "margin_dpo/margin_mean": 12.404184341430664, "margin_dpo/margin_std": 13.576568603515625, "step": 79 }, { "epoch": 0.11747430249632893, "fcm_dpo/beta": 0.08190235495567322, "fcm_dpo/delta": -0.06250029802322388, "fcm_dpo/margin": 11.025638580322266, "fcm_dpo/q_t": 0.3301146924495697, "grad_norm": 44.95631790161133, "learning_rate": 4.996706849759452e-07, "logits/chosen": -0.6934123039245605, "logits/rejected": -0.643015444278717, "logps/chosen": -62.62284469604492, "logps/ref_chosen": -58.519290924072266, "logps/ref_rejected": -87.54750061035156, "logps/rejected": -102.67669677734375, "loss": 0.8811, "margin_dpo/margin_mean": 11.025639533996582, "margin_dpo/margin_std": 12.54986572265625, "step": 80 }, { "epoch": 0.11894273127753303, "fcm_dpo/beta": 0.07723959535360336, "fcm_dpo/delta": -0.3877883851528168, "fcm_dpo/margin": 15.350555419921875, "fcm_dpo/q_t": 0.28344881534576416, "grad_norm": 48.92912292480469, "learning_rate": 4.996015471965529e-07, "logits/chosen": -0.7091237306594849, "logits/rejected": -0.6722517013549805, "logps/chosen": -69.73745727539062, "logps/ref_chosen": -66.44886779785156, "logps/ref_rejected": -129.66270446777344, "logps/rejected": -148.30184936523438, "loss": 0.7526, "margin_dpo/margin_mean": 15.350555419921875, "margin_dpo/margin_std": 15.928838729858398, "step": 81 }, { "epoch": 0.12041116005873716, "fcm_dpo/beta": 0.0763830617070198, "fcm_dpo/delta": -0.12376299500465393, "fcm_dpo/margin": 12.550471305847168, "fcm_dpo/q_t": 0.32355159521102905, "grad_norm": 56.68275451660156, "learning_rate": 4.995258321842611e-07, "logits/chosen": -0.6178134679794312, "logits/rejected": -0.5973782539367676, "logps/chosen": -57.490257263183594, "logps/ref_chosen": -52.232383728027344, "logps/ref_rejected": -90.74325561523438, "logps/rejected": -108.55159759521484, "loss": 0.9461, "margin_dpo/margin_mean": 12.550471305847168, "margin_dpo/margin_std": 17.051355361938477, "step": 82 }, { "epoch": 0.12187958883994127, "fcm_dpo/beta": 0.07206125557422638, "fcm_dpo/delta": -0.21825863420963287, "fcm_dpo/margin": 14.49725341796875, "fcm_dpo/q_t": 0.30366265773773193, "grad_norm": 49.30983352661133, "learning_rate": 4.994435419342304e-07, "logits/chosen": -0.6791856288909912, "logits/rejected": -0.6300394535064697, "logps/chosen": -61.20024108886719, "logps/ref_chosen": -55.82738494873047, "logps/ref_rejected": -103.71589660644531, "logps/rejected": -123.58601379394531, "loss": 0.8149, "margin_dpo/margin_mean": 14.497251510620117, "margin_dpo/margin_std": 15.496784210205078, "step": 83 }, { "epoch": 0.12334801762114538, "fcm_dpo/beta": 0.06941388547420502, "fcm_dpo/delta": -0.12336910516023636, "fcm_dpo/margin": 13.803592681884766, "fcm_dpo/q_t": 0.3031036853790283, "grad_norm": 42.30681228637695, "learning_rate": 4.993546786148857e-07, "logits/chosen": -0.6525845527648926, "logits/rejected": -0.6165010333061218, "logps/chosen": -71.6080093383789, "logps/ref_chosen": -67.1761703491211, "logps/ref_rejected": -87.29859924316406, "logps/rejected": -105.53402709960938, "loss": 0.7964, "margin_dpo/margin_mean": 13.803592681884766, "margin_dpo/margin_std": 12.420629501342773, "step": 84 }, { "epoch": 0.12481644640234948, "fcm_dpo/beta": 0.06950154155492783, "fcm_dpo/delta": -0.10908734798431396, "fcm_dpo/margin": 13.623420715332031, "fcm_dpo/q_t": 0.32009202241897583, "grad_norm": 48.14596176147461, "learning_rate": 4.992592445678582e-07, "logits/chosen": -0.6232889890670776, "logits/rejected": -0.591201663017273, "logps/chosen": -63.80423355102539, "logps/ref_chosen": -58.4066162109375, "logps/ref_rejected": -78.63880157470703, "logps/rejected": -97.65983581542969, "loss": 0.856, "margin_dpo/margin_mean": 13.623420715332031, "margin_dpo/margin_std": 14.695003509521484, "step": 85 }, { "epoch": 0.1262848751835536, "fcm_dpo/beta": 0.06705532968044281, "fcm_dpo/delta": -0.1516844481229782, "fcm_dpo/margin": 14.717806816101074, "fcm_dpo/q_t": 0.3345668911933899, "grad_norm": 52.9031867980957, "learning_rate": 4.991572423079235e-07, "logits/chosen": -0.6772565841674805, "logits/rejected": -0.6597760915756226, "logps/chosen": -63.28825378417969, "logps/ref_chosen": -56.13746643066406, "logps/ref_rejected": -88.12165069580078, "logps/rejected": -109.99024200439453, "loss": 0.9488, "margin_dpo/margin_mean": 14.717806816101074, "margin_dpo/margin_std": 21.25366973876953, "step": 86 }, { "epoch": 0.1277533039647577, "fcm_dpo/beta": 0.06366388499736786, "fcm_dpo/delta": -0.20331230759620667, "fcm_dpo/margin": 16.169401168823242, "fcm_dpo/q_t": 0.30775928497314453, "grad_norm": 43.02919387817383, "learning_rate": 4.990486745229364e-07, "logits/chosen": -0.6946601867675781, "logits/rejected": -0.6573343276977539, "logps/chosen": -63.13631057739258, "logps/ref_chosen": -55.63609313964844, "logps/ref_rejected": -95.46757507324219, "logps/rejected": -119.13719940185547, "loss": 0.8612, "margin_dpo/margin_mean": 16.16939926147461, "margin_dpo/margin_std": 18.847665786743164, "step": 87 }, { "epoch": 0.12922173274596183, "fcm_dpo/beta": 0.0635436624288559, "fcm_dpo/delta": 0.05298266187310219, "fcm_dpo/margin": 12.59765625, "fcm_dpo/q_t": 0.34781643748283386, "grad_norm": 48.83399200439453, "learning_rate": 4.989335440737586e-07, "logits/chosen": -0.6725143194198608, "logits/rejected": -0.6565263271331787, "logps/chosen": -83.05477905273438, "logps/ref_chosen": -73.67115020751953, "logps/ref_rejected": -106.70849609375, "logps/rejected": -128.68978881835938, "loss": 0.979, "margin_dpo/margin_mean": 12.597654342651367, "margin_dpo/margin_std": 16.687740325927734, "step": 88 }, { "epoch": 0.13069016152716592, "fcm_dpo/beta": 0.06436465680599213, "fcm_dpo/delta": -0.04542829841375351, "fcm_dpo/margin": 13.830251693725586, "fcm_dpo/q_t": 0.32423245906829834, "grad_norm": 41.47353744506836, "learning_rate": 4.988118539941847e-07, "logits/chosen": -0.6947389841079712, "logits/rejected": -0.6542242765426636, "logps/chosen": -66.13823699951172, "logps/ref_chosen": -60.624916076660156, "logps/ref_rejected": -82.08354949951172, "logps/rejected": -101.4271240234375, "loss": 0.8568, "margin_dpo/margin_mean": 13.83025074005127, "margin_dpo/margin_std": 15.321671485900879, "step": 89 }, { "epoch": 0.13215859030837004, "fcm_dpo/beta": 0.06196070462465286, "fcm_dpo/delta": -0.2721518874168396, "fcm_dpo/margin": 17.629770278930664, "fcm_dpo/q_t": 0.31731653213500977, "grad_norm": 45.96269607543945, "learning_rate": 4.986836074908615e-07, "logits/chosen": -0.6709829568862915, "logits/rejected": -0.6653557419776917, "logps/chosen": -60.85920715332031, "logps/ref_chosen": -53.285308837890625, "logps/ref_rejected": -111.54470825195312, "logps/rejected": -136.74838256835938, "loss": 0.8919, "margin_dpo/margin_mean": 17.629772186279297, "margin_dpo/margin_std": 22.947355270385742, "step": 90 }, { "epoch": 0.13362701908957417, "fcm_dpo/beta": 0.0593516007065773, "fcm_dpo/delta": -0.15248210728168488, "fcm_dpo/margin": 16.636859893798828, "fcm_dpo/q_t": 0.3205872178077698, "grad_norm": 45.15717697143555, "learning_rate": 4.985488079432037e-07, "logits/chosen": -0.6791576147079468, "logits/rejected": -0.6390055418014526, "logps/chosen": -69.1212158203125, "logps/ref_chosen": -61.802955627441406, "logps/ref_rejected": -87.87395477294922, "logps/rejected": -111.82907104492188, "loss": 0.8799, "margin_dpo/margin_mean": 16.636859893798828, "margin_dpo/margin_std": 19.837177276611328, "step": 91 }, { "epoch": 0.13509544787077826, "fcm_dpo/beta": 0.05815107375383377, "fcm_dpo/delta": -0.09225773066282272, "fcm_dpo/margin": 16.04425811767578, "fcm_dpo/q_t": 0.3281884789466858, "grad_norm": 41.01671600341797, "learning_rate": 4.984074589033043e-07, "logits/chosen": -0.7218481302261353, "logits/rejected": -0.6939176321029663, "logps/chosen": -58.69712829589844, "logps/ref_chosen": -51.640769958496094, "logps/ref_rejected": -77.88117980957031, "logps/rejected": -100.98180389404297, "loss": 0.9017, "margin_dpo/margin_mean": 16.04425811767578, "margin_dpo/margin_std": 19.757722854614258, "step": 92 }, { "epoch": 0.13656387665198239, "fcm_dpo/beta": 0.0565192773938179, "fcm_dpo/delta": -0.09390807151794434, "fcm_dpo/margin": 16.542964935302734, "fcm_dpo/q_t": 0.3172317147254944, "grad_norm": 36.77476501464844, "learning_rate": 4.982595640958425e-07, "logits/chosen": -0.7294448614120483, "logits/rejected": -0.6686340570449829, "logps/chosen": -60.201622009277344, "logps/ref_chosen": -52.529239654541016, "logps/ref_rejected": -77.16075134277344, "logps/rejected": -101.37608337402344, "loss": 0.832, "margin_dpo/margin_mean": 16.5429630279541, "margin_dpo/margin_std": 17.85883903503418, "step": 93 }, { "epoch": 0.13803230543318648, "fcm_dpo/beta": 0.05466047674417496, "fcm_dpo/delta": -0.14541637897491455, "fcm_dpo/margin": 17.906766891479492, "fcm_dpo/q_t": 0.3086274564266205, "grad_norm": 38.767738342285156, "learning_rate": 4.98105127417984e-07, "logits/chosen": -0.689984917640686, "logits/rejected": -0.6643059253692627, "logps/chosen": -69.84527587890625, "logps/ref_chosen": -61.22261047363281, "logps/ref_rejected": -99.59902954101562, "logps/rejected": -126.12846374511719, "loss": 0.809, "margin_dpo/margin_mean": 17.906766891479492, "margin_dpo/margin_std": 17.686237335205078, "step": 94 }, { "epoch": 0.1395007342143906, "fcm_dpo/beta": 0.054808467626571655, "fcm_dpo/delta": 0.03201872110366821, "fcm_dpo/margin": 14.971220016479492, "fcm_dpo/q_t": 0.3315742611885071, "grad_norm": 39.22067642211914, "learning_rate": 4.979441529392784e-07, "logits/chosen": -0.7354916334152222, "logits/rejected": -0.7008249759674072, "logps/chosen": -59.91139221191406, "logps/ref_chosen": -52.523643493652344, "logps/ref_rejected": -75.8803482055664, "logps/rejected": -98.23931884765625, "loss": 0.8765, "margin_dpo/margin_mean": 14.971220016479492, "margin_dpo/margin_std": 15.818527221679688, "step": 95 }, { "epoch": 0.14096916299559473, "fcm_dpo/beta": 0.0527215413749218, "fcm_dpo/delta": -0.2318594753742218, "fcm_dpo/margin": 19.9919490814209, "fcm_dpo/q_t": 0.2983996868133545, "grad_norm": 36.0137939453125, "learning_rate": 4.977766449015534e-07, "logits/chosen": -0.6940563917160034, "logits/rejected": -0.6533340811729431, "logps/chosen": -68.70280456542969, "logps/ref_chosen": -62.15697479248047, "logps/ref_rejected": -96.59601593017578, "logps/rejected": -123.13380432128906, "loss": 0.7806, "margin_dpo/margin_mean": 19.9919490814209, "margin_dpo/margin_std": 20.55471420288086, "step": 96 }, { "epoch": 0.14243759177679882, "fcm_dpo/beta": 0.0531981959939003, "fcm_dpo/delta": -0.009700343012809753, "fcm_dpo/margin": 16.080947875976562, "fcm_dpo/q_t": 0.3233848214149475, "grad_norm": 39.30295944213867, "learning_rate": 4.976026077188012e-07, "logits/chosen": -0.658359169960022, "logits/rejected": -0.6041878461837769, "logps/chosen": -62.353187561035156, "logps/ref_chosen": -54.646366119384766, "logps/ref_rejected": -76.96475219726562, "logps/rejected": -100.75251770019531, "loss": 0.8526, "margin_dpo/margin_mean": 16.080947875976562, "margin_dpo/margin_std": 15.41889762878418, "step": 97 }, { "epoch": 0.14390602055800295, "fcm_dpo/beta": 0.051846109330654144, "fcm_dpo/delta": -0.06934039294719696, "fcm_dpo/margin": 17.607675552368164, "fcm_dpo/q_t": 0.31782758235931396, "grad_norm": 39.470882415771484, "learning_rate": 4.974220459770639e-07, "logits/chosen": -0.6812083125114441, "logits/rejected": -0.6582015752792358, "logps/chosen": -74.72683715820312, "logps/ref_chosen": -65.25862884521484, "logps/ref_rejected": -96.5274887084961, "logps/rejected": -123.60337829589844, "loss": 0.8784, "margin_dpo/margin_mean": 17.607677459716797, "margin_dpo/margin_std": 19.72555923461914, "step": 98 }, { "epoch": 0.14537444933920704, "fcm_dpo/beta": 0.05020611733198166, "fcm_dpo/delta": -0.19243989884853363, "fcm_dpo/margin": 20.36504364013672, "fcm_dpo/q_t": 0.3088953495025635, "grad_norm": 34.95133590698242, "learning_rate": 4.972349644343108e-07, "logits/chosen": -0.6733574271202087, "logits/rejected": -0.6646283864974976, "logps/chosen": -53.47751998901367, "logps/ref_chosen": -45.638484954833984, "logps/ref_rejected": -86.43793487548828, "logps/rejected": -114.64201354980469, "loss": 0.8047, "margin_dpo/margin_mean": 20.36504364013672, "margin_dpo/margin_std": 22.298368453979492, "step": 99 }, { "epoch": 0.14684287812041116, "fcm_dpo/beta": 0.05120418965816498, "fcm_dpo/delta": 0.1581817865371704, "fcm_dpo/margin": 13.70488166809082, "fcm_dpo/q_t": 0.3617163896560669, "grad_norm": 41.91087341308594, "learning_rate": 4.970413680203148e-07, "logits/chosen": -0.6871071457862854, "logits/rejected": -0.6425069570541382, "logps/chosen": -66.13288879394531, "logps/ref_chosen": -57.59397888183594, "logps/ref_rejected": -74.06021118164062, "logps/rejected": -96.30400085449219, "loss": 1.0066, "margin_dpo/margin_mean": 13.70488166809082, "margin_dpo/margin_std": 19.342105865478516, "step": 100 }, { "epoch": 0.14831130690161526, "fcm_dpo/beta": 0.051217399537563324, "fcm_dpo/delta": 0.037239253520965576, "fcm_dpo/margin": 15.905902862548828, "fcm_dpo/q_t": 0.3492443561553955, "grad_norm": 41.5073127746582, "learning_rate": 4.968412618365215e-07, "logits/chosen": -0.7009834051132202, "logits/rejected": -0.6622995734214783, "logps/chosen": -71.91978454589844, "logps/ref_chosen": -61.64885330200195, "logps/ref_rejected": -83.18968200683594, "logps/rejected": -109.36653137207031, "loss": 0.951, "margin_dpo/margin_mean": 15.905902862548828, "margin_dpo/margin_std": 21.403085708618164, "step": 101 }, { "epoch": 0.14977973568281938, "fcm_dpo/beta": 0.05395771190524101, "fcm_dpo/delta": 0.20888186991214752, "fcm_dpo/margin": 12.05476188659668, "fcm_dpo/q_t": 0.37573665380477905, "grad_norm": 50.882755279541016, "learning_rate": 4.966346511559149e-07, "logits/chosen": -0.7070550918579102, "logits/rejected": -0.6569492220878601, "logps/chosen": -76.05787658691406, "logps/ref_chosen": -64.0788803100586, "logps/ref_rejected": -68.18707275390625, "logps/rejected": -92.2208251953125, "loss": 1.0619, "margin_dpo/margin_mean": 12.05476188659668, "margin_dpo/margin_std": 19.289947509765625, "step": 102 }, { "epoch": 0.1512481644640235, "fcm_dpo/beta": 0.05126365274190903, "fcm_dpo/delta": -0.3250649571418762, "fcm_dpo/margin": 22.175342559814453, "fcm_dpo/q_t": 0.28338393568992615, "grad_norm": 35.19338607788086, "learning_rate": 4.964215414228785e-07, "logits/chosen": -0.7068288326263428, "logits/rejected": -0.6695182919502258, "logps/chosen": -69.05276489257812, "logps/ref_chosen": -61.299278259277344, "logps/ref_rejected": -93.57270812988281, "logps/rejected": -123.50154876708984, "loss": 0.7378, "margin_dpo/margin_mean": 22.175342559814453, "margin_dpo/margin_std": 20.998464584350586, "step": 103 }, { "epoch": 0.1527165932452276, "fcm_dpo/beta": 0.049339231103658676, "fcm_dpo/delta": -0.19892087578773499, "fcm_dpo/margin": 20.853076934814453, "fcm_dpo/q_t": 0.3135518431663513, "grad_norm": 38.35378646850586, "learning_rate": 4.96201938253052e-07, "logits/chosen": -0.713874101638794, "logits/rejected": -0.6700358390808105, "logps/chosen": -63.76753234863281, "logps/ref_chosen": -54.372772216796875, "logps/ref_rejected": -89.5647201538086, "logps/rejected": -119.81255340576172, "loss": 0.844, "margin_dpo/margin_mean": 20.853076934814453, "margin_dpo/margin_std": 23.93612289428711, "step": 104 }, { "epoch": 0.15418502202643172, "fcm_dpo/beta": 0.04497908800840378, "fcm_dpo/delta": -0.5303850769996643, "fcm_dpo/margin": 29.143280029296875, "fcm_dpo/q_t": 0.24480026960372925, "grad_norm": 31.636995315551758, "learning_rate": 4.959758474331832e-07, "logits/chosen": -0.7144509553909302, "logits/rejected": -0.678684651851654, "logps/chosen": -62.63848876953125, "logps/ref_chosen": -54.638946533203125, "logps/ref_rejected": -97.97351837158203, "logps/rejected": -135.11634826660156, "loss": 0.6311, "margin_dpo/margin_mean": 29.143281936645508, "margin_dpo/margin_std": 22.500640869140625, "step": 105 }, { "epoch": 0.15565345080763582, "fcm_dpo/beta": 0.04392065852880478, "fcm_dpo/delta": 0.03618063032627106, "fcm_dpo/margin": 18.596851348876953, "fcm_dpo/q_t": 0.33291202783584595, "grad_norm": 33.444847106933594, "learning_rate": 4.957432749209755e-07, "logits/chosen": -0.6787040829658508, "logits/rejected": -0.6291109323501587, "logps/chosen": -64.25106811523438, "logps/ref_chosen": -54.83289337158203, "logps/ref_rejected": -85.22461700439453, "logps/rejected": -113.2396469116211, "loss": 0.873, "margin_dpo/margin_mean": 18.596851348876953, "margin_dpo/margin_std": 19.459993362426758, "step": 106 }, { "epoch": 0.15712187958883994, "fcm_dpo/beta": 0.043349266052246094, "fcm_dpo/delta": -0.07450622320175171, "fcm_dpo/margin": 21.153640747070312, "fcm_dpo/q_t": 0.31638288497924805, "grad_norm": 32.62092208862305, "learning_rate": 4.955042268449307e-07, "logits/chosen": -0.737189531326294, "logits/rejected": -0.6847895383834839, "logps/chosen": -81.13639831542969, "logps/ref_chosen": -69.70780944824219, "logps/ref_rejected": -94.73950958251953, "logps/rejected": -127.32173156738281, "loss": 0.8405, "margin_dpo/margin_mean": 21.153640747070312, "margin_dpo/margin_std": 21.512563705444336, "step": 107 }, { "epoch": 0.15859030837004406, "fcm_dpo/beta": 0.042633771896362305, "fcm_dpo/delta": -0.11348213255405426, "fcm_dpo/margin": 22.344249725341797, "fcm_dpo/q_t": 0.32922905683517456, "grad_norm": 39.67808532714844, "learning_rate": 4.952587095041881e-07, "logits/chosen": -0.7291974425315857, "logits/rejected": -0.6880236864089966, "logps/chosen": -67.06340789794922, "logps/ref_chosen": -56.0098876953125, "logps/ref_rejected": -95.79601287841797, "logps/rejected": -129.19378662109375, "loss": 0.905, "margin_dpo/margin_mean": 22.344249725341797, "margin_dpo/margin_std": 28.051963806152344, "step": 108 }, { "epoch": 0.16005873715124816, "fcm_dpo/beta": 0.04097578674554825, "fcm_dpo/delta": -0.20452482998371124, "fcm_dpo/margin": 25.194259643554688, "fcm_dpo/q_t": 0.29648420214653015, "grad_norm": 33.704471588134766, "learning_rate": 4.95006729368358e-07, "logits/chosen": -0.643449604511261, "logits/rejected": -0.6156653165817261, "logps/chosen": -72.79986572265625, "logps/ref_chosen": -62.88549041748047, "logps/ref_rejected": -98.68573760986328, "logps/rejected": -133.79437255859375, "loss": 0.7753, "margin_dpo/margin_mean": 25.194259643554688, "margin_dpo/margin_std": 23.329872131347656, "step": 109 }, { "epoch": 0.16152716593245228, "fcm_dpo/beta": 0.03958093747496605, "fcm_dpo/delta": -0.08387066423892975, "fcm_dpo/margin": 23.25214958190918, "fcm_dpo/q_t": 0.3206620514392853, "grad_norm": 30.748167037963867, "learning_rate": 4.947482930773511e-07, "logits/chosen": -0.6758570671081543, "logits/rejected": -0.6276640295982361, "logps/chosen": -67.95677185058594, "logps/ref_chosen": -58.753684997558594, "logps/ref_rejected": -79.75001525878906, "logps/rejected": -112.20524597167969, "loss": 0.8678, "margin_dpo/margin_mean": 23.25214958190918, "margin_dpo/margin_std": 25.55893898010254, "step": 110 }, { "epoch": 0.16299559471365638, "fcm_dpo/beta": 0.038935232907533646, "fcm_dpo/delta": -0.1342032104730606, "fcm_dpo/margin": 24.890220642089844, "fcm_dpo/q_t": 0.3145398795604706, "grad_norm": 32.24641418457031, "learning_rate": 4.944834074412042e-07, "logits/chosen": -0.6879921555519104, "logits/rejected": -0.659111738204956, "logps/chosen": -80.65528869628906, "logps/ref_chosen": -68.62410736083984, "logps/ref_rejected": -98.42886352539062, "logps/rejected": -135.35025024414062, "loss": 0.8545, "margin_dpo/margin_mean": 24.890220642089844, "margin_dpo/margin_std": 27.572715759277344, "step": 111 }, { "epoch": 0.1644640234948605, "fcm_dpo/beta": 0.039760373532772064, "fcm_dpo/delta": 0.18148502707481384, "fcm_dpo/margin": 17.13097381591797, "fcm_dpo/q_t": 0.3622838854789734, "grad_norm": 33.34195327758789, "learning_rate": 4.942120794399002e-07, "logits/chosen": -0.7036277055740356, "logits/rejected": -0.6543942093849182, "logps/chosen": -61.78867721557617, "logps/ref_chosen": -50.24964141845703, "logps/ref_rejected": -64.77442932128906, "logps/rejected": -93.44444274902344, "loss": 0.9724, "margin_dpo/margin_mean": 17.130971908569336, "margin_dpo/margin_std": 21.436376571655273, "step": 112 }, { "epoch": 0.16593245227606462, "fcm_dpo/beta": 0.04075583070516586, "fcm_dpo/delta": 0.062437187880277634, "fcm_dpo/margin": 19.445987701416016, "fcm_dpo/q_t": 0.3351534307003021, "grad_norm": 36.214141845703125, "learning_rate": 4.939343162231841e-07, "logits/chosen": -0.6719074249267578, "logits/rejected": -0.6214045882225037, "logps/chosen": -79.45092010498047, "logps/ref_chosen": -66.71295166015625, "logps/ref_rejected": -77.96870422363281, "logps/rejected": -110.15266418457031, "loss": 0.8755, "margin_dpo/margin_mean": 19.445987701416016, "margin_dpo/margin_std": 19.488388061523438, "step": 113 }, { "epoch": 0.16740088105726872, "fcm_dpo/beta": 0.03883285075426102, "fcm_dpo/delta": -0.27601903676986694, "fcm_dpo/margin": 28.083696365356445, "fcm_dpo/q_t": 0.29987218976020813, "grad_norm": 30.65591812133789, "learning_rate": 4.936501251103751e-07, "logits/chosen": -0.6819378137588501, "logits/rejected": -0.6380999684333801, "logps/chosen": -69.460205078125, "logps/ref_chosen": -57.78507995605469, "logps/ref_rejected": -87.10966491699219, "logps/rejected": -126.86849212646484, "loss": 0.7933, "margin_dpo/margin_mean": 28.083698272705078, "margin_dpo/margin_std": 31.275924682617188, "step": 114 }, { "epoch": 0.16886930983847284, "fcm_dpo/beta": 0.039183393120765686, "fcm_dpo/delta": 0.049480877816677094, "fcm_dpo/margin": 20.51511001586914, "fcm_dpo/q_t": 0.35244330763816833, "grad_norm": 45.887428283691406, "learning_rate": 4.933595135901732e-07, "logits/chosen": -0.6870021820068359, "logits/rejected": -0.6467078924179077, "logps/chosen": -81.95085906982422, "logps/ref_chosen": -65.5826416015625, "logps/ref_rejected": -98.56552124023438, "logps/rejected": -135.4488525390625, "loss": 0.9938, "margin_dpo/margin_mean": 20.51511001586914, "margin_dpo/margin_std": 30.72865867614746, "step": 115 }, { "epoch": 0.17033773861967694, "fcm_dpo/beta": 0.039394035935401917, "fcm_dpo/delta": 0.01498110219836235, "fcm_dpo/margin": 21.212890625, "fcm_dpo/q_t": 0.33000385761260986, "grad_norm": 34.129329681396484, "learning_rate": 4.930624893204624e-07, "logits/chosen": -0.7020711898803711, "logits/rejected": -0.6741104125976562, "logps/chosen": -63.63768005371094, "logps/ref_chosen": -51.40031433105469, "logps/ref_rejected": -80.5218505859375, "logps/rejected": -113.97211456298828, "loss": 0.8508, "margin_dpo/margin_mean": 21.212890625, "margin_dpo/margin_std": 21.642147064208984, "step": 116 }, { "epoch": 0.17180616740088106, "fcm_dpo/beta": 0.03966808319091797, "fcm_dpo/delta": 0.07746738195419312, "fcm_dpo/margin": 19.62699317932129, "fcm_dpo/q_t": 0.345183789730072, "grad_norm": 41.55912780761719, "learning_rate": 4.927590601281083e-07, "logits/chosen": -0.6694446802139282, "logits/rejected": -0.6278887987136841, "logps/chosen": -83.67584228515625, "logps/ref_chosen": -69.29840850830078, "logps/ref_rejected": -66.583984375, "logps/rejected": -100.58841705322266, "loss": 0.9269, "margin_dpo/margin_mean": 19.62699317932129, "margin_dpo/margin_std": 24.160232543945312, "step": 117 }, { "epoch": 0.17327459618208516, "fcm_dpo/beta": 0.04006565362215042, "fcm_dpo/delta": 0.035930898040533066, "fcm_dpo/margin": 20.39095115661621, "fcm_dpo/q_t": 0.3354613184928894, "grad_norm": 34.89082336425781, "learning_rate": 4.924492340087524e-07, "logits/chosen": -0.7189027070999146, "logits/rejected": -0.6873916387557983, "logps/chosen": -68.96400451660156, "logps/ref_chosen": -55.6409797668457, "logps/ref_rejected": -75.66905975341797, "logps/rejected": -109.3830337524414, "loss": 0.8748, "margin_dpo/margin_mean": 20.390953063964844, "margin_dpo/margin_std": 21.79241943359375, "step": 118 }, { "epoch": 0.17474302496328928, "fcm_dpo/beta": 0.03989461809396744, "fcm_dpo/delta": 0.016687780618667603, "fcm_dpo/margin": 20.894739151000977, "fcm_dpo/q_t": 0.3389769196510315, "grad_norm": 36.64957046508789, "learning_rate": 4.92133019126601e-07, "logits/chosen": -0.6799700260162354, "logits/rejected": -0.6564502716064453, "logps/chosen": -89.88623046875, "logps/ref_chosen": -73.51019287109375, "logps/ref_rejected": -102.977294921875, "logps/rejected": -140.24806213378906, "loss": 0.9122, "margin_dpo/margin_mean": 20.894737243652344, "margin_dpo/margin_std": 25.40456199645996, "step": 119 }, { "epoch": 0.1762114537444934, "fcm_dpo/beta": 0.038859717547893524, "fcm_dpo/delta": -0.2279738485813141, "fcm_dpo/margin": 27.11212921142578, "fcm_dpo/q_t": 0.29994580149650574, "grad_norm": 35.182918548583984, "learning_rate": 4.918104238142103e-07, "logits/chosen": -0.6844409704208374, "logits/rejected": -0.6399496793746948, "logps/chosen": -94.61766052246094, "logps/ref_chosen": -76.78083801269531, "logps/ref_rejected": -108.02374267578125, "logps/rejected": -152.9727020263672, "loss": 0.8006, "margin_dpo/margin_mean": 27.11212921142578, "margin_dpo/margin_std": 28.035802841186523, "step": 120 }, { "epoch": 0.1776798825256975, "fcm_dpo/beta": 0.036688677966594696, "fcm_dpo/delta": -0.2756182849407196, "fcm_dpo/margin": 29.79346466064453, "fcm_dpo/q_t": 0.2999020218849182, "grad_norm": 33.34456253051758, "learning_rate": 4.91481456572267e-07, "logits/chosen": -0.6548373699188232, "logits/rejected": -0.6413829326629639, "logps/chosen": -78.59416961669922, "logps/ref_chosen": -61.789894104003906, "logps/ref_rejected": -109.99456787109375, "logps/rejected": -156.59231567382812, "loss": 0.803, "margin_dpo/margin_mean": 29.79346466064453, "margin_dpo/margin_std": 31.758745193481445, "step": 121 }, { "epoch": 0.17914831130690162, "fcm_dpo/beta": 0.0338733084499836, "fcm_dpo/delta": -0.47818467020988464, "fcm_dpo/margin": 37.396575927734375, "fcm_dpo/q_t": 0.25546765327453613, "grad_norm": 32.44083023071289, "learning_rate": 4.911461260693638e-07, "logits/chosen": -0.6689621806144714, "logits/rejected": -0.6719322204589844, "logps/chosen": -62.280059814453125, "logps/ref_chosen": -46.9022102355957, "logps/ref_rejected": -106.71418762207031, "logps/rejected": -159.48861694335938, "loss": 0.6623, "margin_dpo/margin_mean": 37.396568298339844, "margin_dpo/margin_std": 30.360065460205078, "step": 122 }, { "epoch": 0.18061674008810572, "fcm_dpo/beta": 0.033041030168533325, "fcm_dpo/delta": 0.02030833438038826, "fcm_dpo/margin": 25.154926300048828, "fcm_dpo/q_t": 0.34362679719924927, "grad_norm": 34.55007553100586, "learning_rate": 4.908044411417711e-07, "logits/chosen": -0.6678287982940674, "logits/rejected": -0.63847815990448, "logps/chosen": -77.76958465576172, "logps/ref_chosen": -61.33863830566406, "logps/ref_rejected": -87.775390625, "logps/rejected": -129.36126708984375, "loss": 0.9621, "margin_dpo/margin_mean": 25.154926300048828, "margin_dpo/margin_std": 34.486244201660156, "step": 123 }, { "epoch": 0.18208516886930984, "fcm_dpo/beta": 0.03184504061937332, "fcm_dpo/delta": -0.2603537142276764, "fcm_dpo/margin": 33.94874572753906, "fcm_dpo/q_t": 0.3141557276248932, "grad_norm": 35.607940673828125, "learning_rate": 4.904564107932048e-07, "logits/chosen": -0.6511549353599548, "logits/rejected": -0.6457709074020386, "logps/chosen": -89.34744262695312, "logps/ref_chosen": -71.44833374023438, "logps/ref_rejected": -117.58056640625, "logps/rejected": -169.42843627929688, "loss": 0.865, "margin_dpo/margin_mean": 33.94874954223633, "margin_dpo/margin_std": 43.01670455932617, "step": 124 }, { "epoch": 0.18355359765051396, "fcm_dpo/beta": 0.03098900616168976, "fcm_dpo/delta": -0.136797696352005, "fcm_dpo/margin": 31.42009735107422, "fcm_dpo/q_t": 0.31026577949523926, "grad_norm": 29.541427612304688, "learning_rate": 4.90102044194588e-07, "logits/chosen": -0.6053882241249084, "logits/rejected": -0.5895368456840515, "logps/chosen": -64.38691711425781, "logps/ref_chosen": -50.136940002441406, "logps/ref_rejected": -83.98861694335938, "logps/rejected": -129.65869140625, "loss": 0.8086, "margin_dpo/margin_mean": 31.42009735107422, "margin_dpo/margin_std": 31.38010025024414, "step": 125 }, { "epoch": 0.18502202643171806, "fcm_dpo/beta": 0.030372876673936844, "fcm_dpo/delta": -0.059251464903354645, "fcm_dpo/margin": 29.75729751586914, "fcm_dpo/q_t": 0.32057902216911316, "grad_norm": 31.237802505493164, "learning_rate": 4.897413506838102e-07, "logits/chosen": -0.6320816874504089, "logits/rejected": -0.6053131222724915, "logps/chosen": -71.67118072509766, "logps/ref_chosen": -55.66706848144531, "logps/ref_rejected": -98.1297607421875, "logps/rejected": -143.89117431640625, "loss": 0.8432, "margin_dpo/margin_mean": 29.75729751586914, "margin_dpo/margin_std": 31.18754768371582, "step": 126 }, { "epoch": 0.18649045521292218, "fcm_dpo/beta": 0.030883219093084335, "fcm_dpo/delta": 0.10523413866758347, "fcm_dpo/margin": 24.356143951416016, "fcm_dpo/q_t": 0.3438740372657776, "grad_norm": 31.248865127563477, "learning_rate": 4.89374339765481e-07, "logits/chosen": -0.64985591173172, "logits/rejected": -0.6203486919403076, "logps/chosen": -71.13253021240234, "logps/ref_chosen": -56.55467987060547, "logps/ref_rejected": -76.7957763671875, "logps/rejected": -115.72977447509766, "loss": 0.9098, "margin_dpo/margin_mean": 24.356143951416016, "margin_dpo/margin_std": 26.364276885986328, "step": 127 }, { "epoch": 0.18795888399412627, "fcm_dpo/beta": 0.03142295777797699, "fcm_dpo/delta": 0.059074968099594116, "fcm_dpo/margin": 25.285329818725586, "fcm_dpo/q_t": 0.34666118025779724, "grad_norm": 36.50014877319336, "learning_rate": 4.890010211106795e-07, "logits/chosen": -0.6584875583648682, "logits/rejected": -0.6175848245620728, "logps/chosen": -74.28919219970703, "logps/ref_chosen": -58.12095642089844, "logps/ref_rejected": -76.43896484375, "logps/rejected": -117.89253234863281, "loss": 0.9482, "margin_dpo/margin_mean": 25.285327911376953, "margin_dpo/margin_std": 32.572845458984375, "step": 128 }, { "epoch": 0.1894273127753304, "fcm_dpo/beta": 0.031564533710479736, "fcm_dpo/delta": 0.029005445539951324, "fcm_dpo/margin": 26.0760498046875, "fcm_dpo/q_t": 0.34760773181915283, "grad_norm": 37.66703414916992, "learning_rate": 4.88621404556699e-07, "logits/chosen": -0.6664289236068726, "logits/rejected": -0.642329216003418, "logps/chosen": -88.56261444091797, "logps/ref_chosen": -66.91637420654297, "logps/ref_rejected": -96.6422119140625, "logps/rejected": -144.364501953125, "loss": 0.9722, "margin_dpo/margin_mean": 26.0760498046875, "margin_dpo/margin_std": 36.57023620605469, "step": 129 }, { "epoch": 0.19089574155653452, "fcm_dpo/beta": 0.03058389388024807, "fcm_dpo/delta": -0.32343757152557373, "fcm_dpo/margin": 37.14537048339844, "fcm_dpo/q_t": 0.2883310616016388, "grad_norm": 29.162261962890625, "learning_rate": 4.882355001067891e-07, "logits/chosen": -0.6344603300094604, "logits/rejected": -0.6262820959091187, "logps/chosen": -60.690242767333984, "logps/ref_chosen": -44.66685104370117, "logps/ref_rejected": -82.78165435791016, "logps/rejected": -135.95042419433594, "loss": 0.7976, "margin_dpo/margin_mean": 37.14537048339844, "margin_dpo/margin_std": 36.782554626464844, "step": 130 }, { "epoch": 0.19236417033773862, "fcm_dpo/beta": 0.02876003086566925, "fcm_dpo/delta": -0.16914135217666626, "fcm_dpo/margin": 34.836463928222656, "fcm_dpo/q_t": 0.29792845249176025, "grad_norm": 29.51772117614746, "learning_rate": 4.878433179298909e-07, "logits/chosen": -0.6310759782791138, "logits/rejected": -0.627743661403656, "logps/chosen": -57.54106903076172, "logps/ref_chosen": -44.924591064453125, "logps/ref_rejected": -88.44401550292969, "logps/rejected": -135.89695739746094, "loss": 0.7855, "margin_dpo/margin_mean": 34.836463928222656, "margin_dpo/margin_std": 32.454898834228516, "step": 131 }, { "epoch": 0.19383259911894274, "fcm_dpo/beta": 0.028265126049518585, "fcm_dpo/delta": -0.045259568840265274, "fcm_dpo/margin": 31.518386840820312, "fcm_dpo/q_t": 0.3281964957714081, "grad_norm": 31.3057918548584, "learning_rate": 4.874448683603694e-07, "logits/chosen": -0.6486563682556152, "logits/rejected": -0.6290041208267212, "logps/chosen": -77.1704330444336, "logps/ref_chosen": -59.00108337402344, "logps/ref_rejected": -87.89215087890625, "logps/rejected": -137.57989501953125, "loss": 0.8694, "margin_dpo/margin_mean": 31.518390655517578, "margin_dpo/margin_std": 36.52690124511719, "step": 132 }, { "epoch": 0.19530102790014683, "fcm_dpo/beta": 0.028410259634256363, "fcm_dpo/delta": 0.037730347365140915, "fcm_dpo/margin": 28.69933319091797, "fcm_dpo/q_t": 0.3451124429702759, "grad_norm": 30.7010440826416, "learning_rate": 4.870401618977415e-07, "logits/chosen": -0.6556867361068726, "logits/rejected": -0.6370849609375, "logps/chosen": -87.88372802734375, "logps/ref_chosen": -66.60449981689453, "logps/ref_rejected": -96.33355712890625, "logps/rejected": -146.3121337890625, "loss": 0.9262, "margin_dpo/margin_mean": 28.69933319091797, "margin_dpo/margin_std": 36.48450469970703, "step": 133 }, { "epoch": 0.19676945668135096, "fcm_dpo/beta": 0.028403986245393753, "fcm_dpo/delta": -0.05411171913146973, "fcm_dpo/margin": 31.654129028320312, "fcm_dpo/q_t": 0.31937503814697266, "grad_norm": 30.17385482788086, "learning_rate": 4.866292092063986e-07, "logits/chosen": -0.6521985530853271, "logits/rejected": -0.6217156648635864, "logps/chosen": -68.50302124023438, "logps/ref_chosen": -52.06925582885742, "logps/ref_rejected": -87.6545181274414, "logps/rejected": -135.74241638183594, "loss": 0.8195, "margin_dpo/margin_mean": 31.654129028320312, "margin_dpo/margin_std": 30.899707794189453, "step": 134 }, { "epoch": 0.19823788546255505, "fcm_dpo/beta": 0.027399450540542603, "fcm_dpo/delta": -0.2010505199432373, "fcm_dpo/margin": 37.6085319519043, "fcm_dpo/q_t": 0.3078111410140991, "grad_norm": 30.048168182373047, "learning_rate": 4.862120211153265e-07, "logits/chosen": -0.6633049249649048, "logits/rejected": -0.6791324019432068, "logps/chosen": -70.54566192626953, "logps/ref_chosen": -50.353858947753906, "logps/ref_rejected": -115.97975158691406, "logps/rejected": -173.7800750732422, "loss": 0.8237, "margin_dpo/margin_mean": 37.6085319519043, "margin_dpo/margin_std": 40.70466613769531, "step": 135 }, { "epoch": 0.19970631424375918, "fcm_dpo/beta": 0.02726733312010765, "fcm_dpo/delta": 0.10746175795793533, "fcm_dpo/margin": 27.498376846313477, "fcm_dpo/q_t": 0.3627038598060608, "grad_norm": 32.92823028564453, "learning_rate": 4.857886086178193e-07, "logits/chosen": -0.6476492881774902, "logits/rejected": -0.6248095035552979, "logps/chosen": -88.18641662597656, "logps/ref_chosen": -65.072509765625, "logps/ref_rejected": -96.32122802734375, "logps/rejected": -146.9335174560547, "loss": 0.9683, "margin_dpo/margin_mean": 27.498376846313477, "margin_dpo/margin_std": 38.72801208496094, "step": 136 }, { "epoch": 0.2011747430249633, "fcm_dpo/beta": 0.026446528732776642, "fcm_dpo/delta": -0.28031742572784424, "fcm_dpo/margin": 41.52157211303711, "fcm_dpo/q_t": 0.30795353651046753, "grad_norm": 29.9466609954834, "learning_rate": 4.853589828711902e-07, "logits/chosen": -0.6132858991622925, "logits/rejected": -0.6249020099639893, "logps/chosen": -73.17112731933594, "logps/ref_chosen": -48.759117126464844, "logps/ref_rejected": -113.86376953125, "logps/rejected": -179.79736328125, "loss": 0.8465, "margin_dpo/margin_mean": 41.52157211303711, "margin_dpo/margin_std": 49.937278747558594, "step": 137 }, { "epoch": 0.2026431718061674, "fcm_dpo/beta": 0.02638918161392212, "fcm_dpo/delta": 0.0248483307659626, "fcm_dpo/margin": 31.32628631591797, "fcm_dpo/q_t": 0.33315446972846985, "grad_norm": 33.18700408935547, "learning_rate": 4.849231551964771e-07, "logits/chosen": -0.6230736374855042, "logits/rejected": -0.603124737739563, "logps/chosen": -86.28077697753906, "logps/ref_chosen": -60.519649505615234, "logps/ref_rejected": -93.19694519042969, "logps/rejected": -150.28436279296875, "loss": 0.8956, "margin_dpo/margin_mean": 31.32628631591797, "margin_dpo/margin_std": 35.16431427001953, "step": 138 }, { "epoch": 0.20411160058737152, "fcm_dpo/beta": 0.025706283748149872, "fcm_dpo/delta": -0.17333011329174042, "fcm_dpo/margin": 39.14105987548828, "fcm_dpo/q_t": 0.30358999967575073, "grad_norm": 28.363384246826172, "learning_rate": 4.844811370781446e-07, "logits/chosen": -0.6366190910339355, "logits/rejected": -0.6161980032920837, "logps/chosen": -66.5361099243164, "logps/ref_chosen": -46.89138412475586, "logps/ref_rejected": -79.72798156738281, "logps/rejected": -138.51376342773438, "loss": 0.7925, "margin_dpo/margin_mean": 39.14105987548828, "margin_dpo/margin_std": 38.770042419433594, "step": 139 }, { "epoch": 0.2055800293685756, "fcm_dpo/beta": 0.02510003000497818, "fcm_dpo/delta": -0.06664561480283737, "fcm_dpo/margin": 36.26915740966797, "fcm_dpo/q_t": 0.3291313052177429, "grad_norm": 30.79945945739746, "learning_rate": 4.840329401637809e-07, "logits/chosen": -0.6232326030731201, "logits/rejected": -0.6018444299697876, "logps/chosen": -82.54205322265625, "logps/ref_chosen": -58.97471618652344, "logps/ref_rejected": -83.28410339355469, "logps/rejected": -143.12060546875, "loss": 0.8953, "margin_dpo/margin_mean": 36.26915740966797, "margin_dpo/margin_std": 43.71975326538086, "step": 140 }, { "epoch": 0.20704845814977973, "fcm_dpo/beta": 0.02491670660674572, "fcm_dpo/delta": -0.04375480115413666, "fcm_dpo/margin": 35.71399688720703, "fcm_dpo/q_t": 0.3214990496635437, "grad_norm": 33.90923309326172, "learning_rate": 4.83578576263792e-07, "logits/chosen": -0.6241617202758789, "logits/rejected": -0.6002821326255798, "logps/chosen": -98.45579528808594, "logps/ref_chosen": -75.07566833496094, "logps/ref_rejected": -98.1922607421875, "logps/rejected": -157.286376953125, "loss": 0.8779, "margin_dpo/margin_mean": 35.71399688720703, "margin_dpo/margin_std": 39.75990295410156, "step": 141 }, { "epoch": 0.20851688693098386, "fcm_dpo/beta": 0.024470651522278786, "fcm_dpo/delta": -0.15550553798675537, "fcm_dpo/margin": 40.46052551269531, "fcm_dpo/q_t": 0.3215620219707489, "grad_norm": 33.35566329956055, "learning_rate": 4.83118057351089e-07, "logits/chosen": -0.5816446542739868, "logits/rejected": -0.5725831985473633, "logps/chosen": -86.0451889038086, "logps/ref_chosen": -58.027931213378906, "logps/ref_rejected": -94.58222961425781, "logps/rejected": -163.06002807617188, "loss": 0.9117, "margin_dpo/margin_mean": 40.46052551269531, "margin_dpo/margin_std": 50.63873291015625, "step": 142 }, { "epoch": 0.20998531571218795, "fcm_dpo/beta": 0.0247478224337101, "fcm_dpo/delta": 0.21386152505874634, "fcm_dpo/margin": 26.287752151489258, "fcm_dpo/q_t": 0.37550124526023865, "grad_norm": 38.79267120361328, "learning_rate": 4.826513955607734e-07, "logits/chosen": -0.6034103631973267, "logits/rejected": -0.5768376588821411, "logps/chosen": -87.43092346191406, "logps/ref_chosen": -57.59645080566406, "logps/ref_rejected": -78.99957275390625, "logps/rejected": -135.12179565429688, "loss": 1.0509, "margin_dpo/margin_mean": 26.287750244140625, "margin_dpo/margin_std": 41.523109436035156, "step": 143 }, { "epoch": 0.21145374449339208, "fcm_dpo/beta": 0.02499964088201523, "fcm_dpo/delta": -0.029541175812482834, "fcm_dpo/margin": 35.07600784301758, "fcm_dpo/q_t": 0.3219374418258667, "grad_norm": 29.825363159179688, "learning_rate": 4.821786031898176e-07, "logits/chosen": -0.5954188704490662, "logits/rejected": -0.5570698380470276, "logps/chosen": -84.0816650390625, "logps/ref_chosen": -59.90636444091797, "logps/ref_rejected": -82.00025939941406, "logps/rejected": -141.25157165527344, "loss": 0.8468, "margin_dpo/margin_mean": 35.076011657714844, "margin_dpo/margin_std": 35.606632232666016, "step": 144 }, { "epoch": 0.21292217327459617, "fcm_dpo/beta": 0.024706725031137466, "fcm_dpo/delta": -0.06008060276508331, "fcm_dpo/margin": 36.604347229003906, "fcm_dpo/q_t": 0.324362576007843, "grad_norm": 35.410675048828125, "learning_rate": 4.816996926967401e-07, "logits/chosen": -0.5916780233383179, "logits/rejected": -0.55748450756073, "logps/chosen": -83.62374877929688, "logps/ref_chosen": -56.60066604614258, "logps/ref_rejected": -77.86631774902344, "logps/rejected": -141.49374389648438, "loss": 0.8636, "margin_dpo/margin_mean": 36.604347229003906, "margin_dpo/margin_std": 40.430137634277344, "step": 145 }, { "epoch": 0.2143906020558003, "fcm_dpo/beta": 0.025057196617126465, "fcm_dpo/delta": 0.12556242942810059, "fcm_dpo/margin": 29.2852783203125, "fcm_dpo/q_t": 0.3575107455253601, "grad_norm": 46.84541702270508, "learning_rate": 4.812146767012779e-07, "logits/chosen": -0.5719867944717407, "logits/rejected": -0.5233458280563354, "logps/chosen": -103.90762329101562, "logps/ref_chosen": -66.00045013427734, "logps/ref_rejected": -81.70278930664062, "logps/rejected": -148.89523315429688, "loss": 1.0217, "margin_dpo/margin_mean": 29.2852783203125, "margin_dpo/margin_std": 43.19415283203125, "step": 146 }, { "epoch": 0.21585903083700442, "fcm_dpo/beta": 0.02465198189020157, "fcm_dpo/delta": -0.14972899854183197, "fcm_dpo/margin": 39.91291046142578, "fcm_dpo/q_t": 0.3194273114204407, "grad_norm": 39.194644927978516, "learning_rate": 4.807235679840536e-07, "logits/chosen": -0.5805500149726868, "logits/rejected": -0.5455455183982849, "logps/chosen": -82.01380157470703, "logps/ref_chosen": -53.405487060546875, "logps/ref_rejected": -71.39060974121094, "logps/rejected": -139.91183471679688, "loss": 0.8708, "margin_dpo/margin_mean": 39.91291046142578, "margin_dpo/margin_std": 47.64521789550781, "step": 147 }, { "epoch": 0.2173274596182085, "fcm_dpo/beta": 0.02513396367430687, "fcm_dpo/delta": 0.046579986810684204, "fcm_dpo/margin": 31.903329849243164, "fcm_dpo/q_t": 0.3463394045829773, "grad_norm": 36.274044036865234, "learning_rate": 4.802263794862384e-07, "logits/chosen": -0.6228268146514893, "logits/rejected": -0.6043528318405151, "logps/chosen": -91.84988403320312, "logps/ref_chosen": -64.93708038330078, "logps/ref_rejected": -103.09384155273438, "logps/rejected": -161.90997314453125, "loss": 0.9264, "margin_dpo/margin_mean": 31.903329849243164, "margin_dpo/margin_std": 37.86564636230469, "step": 148 }, { "epoch": 0.21879588839941264, "fcm_dpo/beta": 0.023908747360110283, "fcm_dpo/delta": -0.15772250294685364, "fcm_dpo/margin": 41.27139663696289, "fcm_dpo/q_t": 0.30605483055114746, "grad_norm": 32.16698455810547, "learning_rate": 4.797231243092118e-07, "logits/chosen": -0.6210156083106995, "logits/rejected": -0.5976389050483704, "logps/chosen": -85.54385375976562, "logps/ref_chosen": -58.47376251220703, "logps/ref_rejected": -99.31474304199219, "logps/rejected": -167.65623474121094, "loss": 0.8122, "margin_dpo/margin_mean": 41.271400451660156, "margin_dpo/margin_std": 41.392826080322266, "step": 149 }, { "epoch": 0.22026431718061673, "fcm_dpo/beta": 0.02343493327498436, "fcm_dpo/delta": -0.06840753555297852, "fcm_dpo/margin": 38.6958122253418, "fcm_dpo/q_t": 0.33513975143432617, "grad_norm": 32.42837142944336, "learning_rate": 4.792138157142157e-07, "logits/chosen": -0.5992769002914429, "logits/rejected": -0.5919263362884521, "logps/chosen": -71.05235290527344, "logps/ref_chosen": -45.705810546875, "logps/ref_rejected": -83.34759521484375, "logps/rejected": -147.38995361328125, "loss": 0.9001, "margin_dpo/margin_mean": 38.6958122253418, "margin_dpo/margin_std": 49.43400573730469, "step": 150 }, { "epoch": 0.22173274596182085, "fcm_dpo/beta": 0.023284468799829483, "fcm_dpo/delta": -0.12595298886299133, "fcm_dpo/margin": 41.395328521728516, "fcm_dpo/q_t": 0.30875810980796814, "grad_norm": 31.666311264038086, "learning_rate": 4.786984671220053e-07, "logits/chosen": -0.6710444688796997, "logits/rejected": -0.6334470510482788, "logps/chosen": -98.85562133789062, "logps/ref_chosen": -70.57083129882812, "logps/ref_rejected": -100.46382141113281, "logps/rejected": -170.14393615722656, "loss": 0.8177, "margin_dpo/margin_mean": 41.39532470703125, "margin_dpo/margin_std": 41.780487060546875, "step": 151 }, { "epoch": 0.22320117474302498, "fcm_dpo/beta": 0.022631559520959854, "fcm_dpo/delta": -0.1879671961069107, "fcm_dpo/margin": 45.03105926513672, "fcm_dpo/q_t": 0.3069019615650177, "grad_norm": 28.960357666015625, "learning_rate": 4.78177092112495e-07, "logits/chosen": -0.6386080980300903, "logits/rejected": -0.6261081099510193, "logps/chosen": -84.44389343261719, "logps/ref_chosen": -60.16438674926758, "logps/ref_rejected": -106.14045715332031, "logps/rejected": -175.45101928710938, "loss": 0.8142, "margin_dpo/margin_mean": 45.03105926513672, "margin_dpo/margin_std": 46.33378601074219, "step": 152 }, { "epoch": 0.22466960352422907, "fcm_dpo/beta": 0.022312451153993607, "fcm_dpo/delta": -0.0069070253521203995, "fcm_dpo/margin": 38.37080001831055, "fcm_dpo/q_t": 0.3373130261898041, "grad_norm": 27.920183181762695, "learning_rate": 4.776497044244016e-07, "logits/chosen": -0.6307264566421509, "logits/rejected": -0.6185749769210815, "logps/chosen": -80.81060028076172, "logps/ref_chosen": -56.315277099609375, "logps/ref_rejected": -85.65583801269531, "logps/rejected": -148.52195739746094, "loss": 0.9136, "margin_dpo/margin_mean": 38.37080001831055, "margin_dpo/margin_std": 48.248252868652344, "step": 153 }, { "epoch": 0.2261380323054332, "fcm_dpo/beta": 0.022488757967948914, "fcm_dpo/delta": 0.013789080083370209, "fcm_dpo/margin": 37.155738830566406, "fcm_dpo/q_t": 0.34336578845977783, "grad_norm": 33.870460510253906, "learning_rate": 4.771163179548808e-07, "logits/chosen": -0.613184928894043, "logits/rejected": -0.6004446744918823, "logps/chosen": -91.80575561523438, "logps/ref_chosen": -62.74256896972656, "logps/ref_rejected": -104.24420166015625, "logps/rejected": -170.463134765625, "loss": 0.958, "margin_dpo/margin_mean": 37.15574264526367, "margin_dpo/margin_std": 48.641082763671875, "step": 154 }, { "epoch": 0.2276064610866373, "fcm_dpo/beta": 0.0222769845277071, "fcm_dpo/delta": -0.047925353050231934, "fcm_dpo/margin": 40.09435272216797, "fcm_dpo/q_t": 0.3238127827644348, "grad_norm": 28.904926300048828, "learning_rate": 4.7657694675916247e-07, "logits/chosen": -0.6429010033607483, "logits/rejected": -0.615513265132904, "logps/chosen": -84.72270965576172, "logps/ref_chosen": -60.65318298339844, "logps/ref_rejected": -77.49220275878906, "logps/rejected": -141.6560821533203, "loss": 0.8665, "margin_dpo/margin_mean": 40.09435272216797, "margin_dpo/margin_std": 44.73542022705078, "step": 155 }, { "epoch": 0.2290748898678414, "fcm_dpo/beta": 0.02297946810722351, "fcm_dpo/delta": 0.26264724135398865, "fcm_dpo/margin": 26.240877151489258, "fcm_dpo/q_t": 0.382540762424469, "grad_norm": 47.13714599609375, "learning_rate": 4.7603160505017893e-07, "logits/chosen": -0.609096109867096, "logits/rejected": -0.5791852474212646, "logps/chosen": -104.53880310058594, "logps/ref_chosen": -69.49188232421875, "logps/ref_rejected": -77.16929626464844, "logps/rejected": -138.45709228515625, "loss": 1.1578, "margin_dpo/margin_mean": 26.240875244140625, "margin_dpo/margin_std": 51.68781280517578, "step": 156 }, { "epoch": 0.2305433186490455, "fcm_dpo/beta": 0.02237515337765217, "fcm_dpo/delta": -0.21659667789936066, "fcm_dpo/margin": 46.5479736328125, "fcm_dpo/q_t": 0.29945075511932373, "grad_norm": 34.71981430053711, "learning_rate": 4.7548030719819154e-07, "logits/chosen": -0.6424323320388794, "logits/rejected": -0.6342293620109558, "logps/chosen": -94.13262939453125, "logps/ref_chosen": -61.368438720703125, "logps/ref_rejected": -107.64636993408203, "logps/rejected": -186.9585418701172, "loss": 0.8133, "margin_dpo/margin_mean": 46.5479736328125, "margin_dpo/margin_std": 47.89894104003906, "step": 157 }, { "epoch": 0.23201174743024963, "fcm_dpo/beta": 0.0216847974807024, "fcm_dpo/delta": -0.17651230096817017, "fcm_dpo/margin": 46.50477600097656, "fcm_dpo/q_t": 0.32549408078193665, "grad_norm": 27.118051528930664, "learning_rate": 4.7492306773041136e-07, "logits/chosen": -0.6172913312911987, "logits/rejected": -0.6179910898208618, "logps/chosen": -87.16582489013672, "logps/ref_chosen": -57.612918853759766, "logps/ref_rejected": -113.6946792602539, "logps/rejected": -189.75234985351562, "loss": 0.8785, "margin_dpo/margin_mean": 46.50477600097656, "margin_dpo/margin_std": 59.50420379638672, "step": 158 }, { "epoch": 0.23348017621145375, "fcm_dpo/beta": 0.021733447909355164, "fcm_dpo/delta": 0.028508372604846954, "fcm_dpo/margin": 37.868499755859375, "fcm_dpo/q_t": 0.3453836739063263, "grad_norm": 54.266422271728516, "learning_rate": 4.743599013306165e-07, "logits/chosen": -0.6216092705726624, "logits/rejected": -0.5767144560813904, "logps/chosen": -115.16629028320312, "logps/ref_chosen": -81.56034851074219, "logps/ref_rejected": -88.89871215820312, "logps/rejected": -160.37315368652344, "loss": 0.946, "margin_dpo/margin_mean": 37.868499755859375, "margin_dpo/margin_std": 50.3989143371582, "step": 159 }, { "epoch": 0.23494860499265785, "fcm_dpo/beta": 0.02132713794708252, "fcm_dpo/delta": -0.06508233398199081, "fcm_dpo/margin": 42.595428466796875, "fcm_dpo/q_t": 0.3389905095100403, "grad_norm": 37.15032196044922, "learning_rate": 4.737908228387656e-07, "logits/chosen": -0.6360121965408325, "logits/rejected": -0.6134800910949707, "logps/chosen": -99.04662322998047, "logps/ref_chosen": -65.73088073730469, "logps/ref_rejected": -97.21781921386719, "logps/rejected": -173.12899780273438, "loss": 0.9272, "margin_dpo/margin_mean": 42.59543228149414, "margin_dpo/margin_std": 56.86052703857422, "step": 160 }, { "epoch": 0.23641703377386197, "fcm_dpo/beta": 0.021413713693618774, "fcm_dpo/delta": -0.0029074866324663162, "fcm_dpo/margin": 39.81043243408203, "fcm_dpo/q_t": 0.3397424519062042, "grad_norm": 32.6351432800293, "learning_rate": 4.7321584725060594e-07, "logits/chosen": -0.6320241689682007, "logits/rejected": -0.614162802696228, "logps/chosen": -82.55064392089844, "logps/ref_chosen": -52.43647003173828, "logps/ref_rejected": -83.43095397949219, "logps/rejected": -153.3555450439453, "loss": 0.9246, "margin_dpo/margin_mean": 39.81043243408203, "margin_dpo/margin_std": 50.22443389892578, "step": 161 }, { "epoch": 0.23788546255506607, "fcm_dpo/beta": 0.020959045737981796, "fcm_dpo/delta": -0.020670533180236816, "fcm_dpo/margin": 41.22890090942383, "fcm_dpo/q_t": 0.3389648199081421, "grad_norm": 26.661542892456055, "learning_rate": 4.7263498971727905e-07, "logits/chosen": -0.6230508685112, "logits/rejected": -0.5922760367393494, "logps/chosen": -90.76399993896484, "logps/ref_chosen": -62.6105842590332, "logps/ref_rejected": -89.39057922363281, "logps/rejected": -158.77288818359375, "loss": 0.9148, "margin_dpo/margin_mean": 41.22889709472656, "margin_dpo/margin_std": 52.275760650634766, "step": 162 }, { "epoch": 0.2393538913362702, "fcm_dpo/beta": 0.021323660388588905, "fcm_dpo/delta": 0.011209924705326557, "fcm_dpo/margin": 39.377601623535156, "fcm_dpo/q_t": 0.33567047119140625, "grad_norm": 28.12457847595215, "learning_rate": 4.720482655449212e-07, "logits/chosen": -0.6243945956230164, "logits/rejected": -0.5960164070129395, "logps/chosen": -83.83268737792969, "logps/ref_chosen": -55.021629333496094, "logps/ref_rejected": -75.418212890625, "logps/rejected": -143.60687255859375, "loss": 0.8906, "margin_dpo/margin_mean": 39.377601623535156, "margin_dpo/margin_std": 45.302734375, "step": 163 }, { "epoch": 0.24082232011747431, "fcm_dpo/beta": 0.020698610693216324, "fcm_dpo/delta": -0.11274396628141403, "fcm_dpo/margin": 45.797706604003906, "fcm_dpo/q_t": 0.311231791973114, "grad_norm": 29.976978302001953, "learning_rate": 4.714556901942599e-07, "logits/chosen": -0.6112655401229858, "logits/rejected": -0.5806326866149902, "logps/chosen": -83.19902038574219, "logps/ref_chosen": -55.64066696166992, "logps/ref_rejected": -79.66463470458984, "logps/rejected": -153.02069091796875, "loss": 0.8155, "margin_dpo/margin_mean": 45.79771041870117, "margin_dpo/margin_std": 45.62419128417969, "step": 164 }, { "epoch": 0.2422907488986784, "fcm_dpo/beta": 0.021473418921232224, "fcm_dpo/delta": 0.20934727787971497, "fcm_dpo/margin": 30.47848129272461, "fcm_dpo/q_t": 0.36805397272109985, "grad_norm": 38.93865203857422, "learning_rate": 4.708572792802069e-07, "logits/chosen": -0.6419720649719238, "logits/rejected": -0.5996274948120117, "logps/chosen": -92.79157257080078, "logps/ref_chosen": -61.310691833496094, "logps/ref_rejected": -73.67060852050781, "logps/rejected": -135.62997436523438, "loss": 1.0007, "margin_dpo/margin_mean": 30.47848129272461, "margin_dpo/margin_std": 41.282371520996094, "step": 165 }, { "epoch": 0.24375917767988253, "fcm_dpo/beta": 0.020592810586094856, "fcm_dpo/delta": -0.3078145384788513, "fcm_dpo/margin": 54.38330841064453, "fcm_dpo/q_t": 0.30514246225357056, "grad_norm": 27.210079193115234, "learning_rate": 4.702530485714461e-07, "logits/chosen": -0.5745774507522583, "logits/rejected": -0.580484926700592, "logps/chosen": -80.91297912597656, "logps/ref_chosen": -50.98360061645508, "logps/ref_rejected": -98.09512329101562, "logps/rejected": -182.40780639648438, "loss": 0.8056, "margin_dpo/margin_mean": 54.38330841064453, "margin_dpo/margin_std": 63.627532958984375, "step": 166 }, { "epoch": 0.24522760646108663, "fcm_dpo/beta": 0.0195004865527153, "fcm_dpo/delta": -0.295537531375885, "fcm_dpo/margin": 57.01265335083008, "fcm_dpo/q_t": 0.28371232748031616, "grad_norm": 26.915447235107422, "learning_rate": 4.6964301399001877e-07, "logits/chosen": -0.5345897674560547, "logits/rejected": -0.5327332615852356, "logps/chosen": -81.28956604003906, "logps/ref_chosen": -50.424095153808594, "logps/ref_rejected": -96.03042602539062, "logps/rejected": -183.90853881835938, "loss": 0.7552, "margin_dpo/margin_mean": 57.01264953613281, "margin_dpo/margin_std": 52.89997863769531, "step": 167 }, { "epoch": 0.24669603524229075, "fcm_dpo/beta": 0.019154991954565048, "fcm_dpo/delta": -0.011862130835652351, "fcm_dpo/margin": 44.928321838378906, "fcm_dpo/q_t": 0.3287070691585541, "grad_norm": 30.181053161621094, "learning_rate": 4.690271916109034e-07, "logits/chosen": -0.5706725716590881, "logits/rejected": -0.5486756563186646, "logps/chosen": -83.01795959472656, "logps/ref_chosen": -49.462825775146484, "logps/ref_rejected": -75.30855560302734, "logps/rejected": -153.79202270507812, "loss": 0.8619, "margin_dpo/margin_mean": 44.928321838378906, "margin_dpo/margin_std": 48.069007873535156, "step": 168 }, { "epoch": 0.24816446402349487, "fcm_dpo/beta": 0.019270140677690506, "fcm_dpo/delta": 0.09090002626180649, "fcm_dpo/margin": 39.69366455078125, "fcm_dpo/q_t": 0.36279886960983276, "grad_norm": 31.3607234954834, "learning_rate": 4.6840559766159235e-07, "logits/chosen": -0.5877891778945923, "logits/rejected": -0.564651370048523, "logps/chosen": -92.62950134277344, "logps/ref_chosen": -59.803443908691406, "logps/ref_rejected": -83.34574890136719, "logps/rejected": -155.86546325683594, "loss": 1.004, "margin_dpo/margin_mean": 39.69366455078125, "margin_dpo/margin_std": 61.303009033203125, "step": 169 }, { "epoch": 0.24963289280469897, "fcm_dpo/beta": 0.019043531268835068, "fcm_dpo/delta": -0.11414046585559845, "fcm_dpo/margin": 49.9086799621582, "fcm_dpo/q_t": 0.3118130564689636, "grad_norm": 29.482023239135742, "learning_rate": 4.6777824852166437e-07, "logits/chosen": -0.5438283681869507, "logits/rejected": -0.5250686407089233, "logps/chosen": -79.19264221191406, "logps/ref_chosen": -49.471771240234375, "logps/ref_rejected": -75.91734313964844, "logps/rejected": -155.546875, "loss": 0.8341, "margin_dpo/margin_mean": 49.9086799621582, "margin_dpo/margin_std": 50.66975402832031, "step": 170 }, { "epoch": 0.2511013215859031, "fcm_dpo/beta": 0.019304830580949783, "fcm_dpo/delta": 0.06730199605226517, "fcm_dpo/margin": 40.81938171386719, "fcm_dpo/q_t": 0.3554888367652893, "grad_norm": 40.410728454589844, "learning_rate": 4.6714516072235273e-07, "logits/chosen": -0.591011643409729, "logits/rejected": -0.5511142015457153, "logps/chosen": -129.15768432617188, "logps/ref_chosen": -84.49931335449219, "logps/ref_rejected": -109.38209533691406, "logps/rejected": -194.85986328125, "loss": 0.9864, "margin_dpo/margin_mean": 40.81937789916992, "margin_dpo/margin_std": 60.447486877441406, "step": 171 }, { "epoch": 0.2525697503671072, "fcm_dpo/beta": 0.01938115805387497, "fcm_dpo/delta": 0.0339435450732708, "fcm_dpo/margin": 42.23426055908203, "fcm_dpo/q_t": 0.34325116872787476, "grad_norm": 37.05482864379883, "learning_rate": 4.6650635094610966e-07, "logits/chosen": -0.5794566869735718, "logits/rejected": -0.5384765863418579, "logps/chosen": -112.9964599609375, "logps/ref_chosen": -68.65391540527344, "logps/ref_rejected": -85.43667602539062, "logps/rejected": -172.0134735107422, "loss": 0.9452, "margin_dpo/margin_mean": 42.23426055908203, "margin_dpo/margin_std": 55.385589599609375, "step": 172 }, { "epoch": 0.2540381791483113, "fcm_dpo/beta": 0.01962945982813835, "fcm_dpo/delta": 0.012805420905351639, "fcm_dpo/margin": 42.6837158203125, "fcm_dpo/q_t": 0.33354562520980835, "grad_norm": 35.34359359741211, "learning_rate": 4.6586183602616687e-07, "logits/chosen": -0.579889714717865, "logits/rejected": -0.5342808961868286, "logps/chosen": -101.8213119506836, "logps/ref_chosen": -63.050880432128906, "logps/ref_rejected": -78.68392181396484, "logps/rejected": -160.1380615234375, "loss": 0.8791, "margin_dpo/margin_mean": 42.6837158203125, "margin_dpo/margin_std": 47.47747039794922, "step": 173 }, { "epoch": 0.2555066079295154, "fcm_dpo/beta": 0.019577287137508392, "fcm_dpo/delta": -0.051362376660108566, "fcm_dpo/margin": 45.75571060180664, "fcm_dpo/q_t": 0.3332166075706482, "grad_norm": 41.42070770263672, "learning_rate": 4.652116329460919e-07, "logits/chosen": -0.5348306894302368, "logits/rejected": -0.5489102602005005, "logps/chosen": -93.36699676513672, "logps/ref_chosen": -53.36296844482422, "logps/ref_rejected": -101.91120910644531, "logps/rejected": -187.67095947265625, "loss": 0.9046, "margin_dpo/margin_mean": 45.755714416503906, "margin_dpo/margin_std": 55.97746276855469, "step": 174 }, { "epoch": 0.25697503671071953, "fcm_dpo/beta": 0.018269890919327736, "fcm_dpo/delta": -0.3679980933666229, "fcm_dpo/margin": 64.21873474121094, "fcm_dpo/q_t": 0.2768920063972473, "grad_norm": 34.53947830200195, "learning_rate": 4.645557588393406e-07, "logits/chosen": -0.45041948556900024, "logits/rejected": -0.4352826476097107, "logps/chosen": -84.18891906738281, "logps/ref_chosen": -45.417762756347656, "logps/ref_rejected": -89.50579833984375, "logps/rejected": -192.49569702148438, "loss": 0.7209, "margin_dpo/margin_mean": 64.21873474121094, "margin_dpo/margin_std": 58.910743713378906, "step": 175 }, { "epoch": 0.25844346549192365, "fcm_dpo/beta": 0.01778358779847622, "fcm_dpo/delta": -0.09820236265659332, "fcm_dpo/margin": 52.805381774902344, "fcm_dpo/q_t": 0.3222000002861023, "grad_norm": 47.036460876464844, "learning_rate": 4.638942309888058e-07, "logits/chosen": -0.5136569738388062, "logits/rejected": -0.5245624780654907, "logps/chosen": -90.69015502929688, "logps/ref_chosen": -50.452842712402344, "logps/ref_rejected": -95.5589599609375, "logps/rejected": -188.60165405273438, "loss": 0.8453, "margin_dpo/margin_mean": 52.80537796020508, "margin_dpo/margin_std": 58.76519012451172, "step": 176 }, { "epoch": 0.2599118942731278, "fcm_dpo/beta": 0.01743006333708763, "fcm_dpo/delta": -0.10580591857433319, "fcm_dpo/margin": 54.26726531982422, "fcm_dpo/q_t": 0.3224429488182068, "grad_norm": 32.768924713134766, "learning_rate": 4.6322706682636137e-07, "logits/chosen": -0.5340214967727661, "logits/rejected": -0.5136857032775879, "logps/chosen": -106.60789489746094, "logps/ref_chosen": -61.216468811035156, "logps/ref_rejected": -95.89378356933594, "logps/rejected": -195.55247497558594, "loss": 0.855, "margin_dpo/margin_mean": 54.26726150512695, "margin_dpo/margin_std": 61.632484436035156, "step": 177 }, { "epoch": 0.26138032305433184, "fcm_dpo/beta": 0.016440367326140404, "fcm_dpo/delta": -0.292092502117157, "fcm_dpo/margin": 67.36396789550781, "fcm_dpo/q_t": 0.29801633954048157, "grad_norm": 42.785648345947266, "learning_rate": 4.6255428393240354e-07, "logits/chosen": -0.4795859456062317, "logits/rejected": -0.48533114790916443, "logps/chosen": -108.70709228515625, "logps/ref_chosen": -58.26478958129883, "logps/ref_rejected": -105.3653335571289, "logps/rejected": -223.17160034179688, "loss": 0.8, "margin_dpo/margin_mean": 67.36396026611328, "margin_dpo/margin_std": 74.31996154785156, "step": 178 }, { "epoch": 0.26284875183553597, "fcm_dpo/beta": 0.01604972966015339, "fcm_dpo/delta": -0.04819926619529724, "fcm_dpo/margin": 55.60845184326172, "fcm_dpo/q_t": 0.33276423811912537, "grad_norm": 37.59048080444336, "learning_rate": 4.6187590003538724e-07, "logits/chosen": -0.49698591232299805, "logits/rejected": -0.48694515228271484, "logps/chosen": -108.184814453125, "logps/ref_chosen": -61.05832290649414, "logps/ref_rejected": -90.52782440185547, "logps/rejected": -193.26278686523438, "loss": 0.9377, "margin_dpo/margin_mean": 55.60845184326172, "margin_dpo/margin_std": 71.06735229492188, "step": 179 }, { "epoch": 0.2643171806167401, "fcm_dpo/beta": 0.01576114445924759, "fcm_dpo/delta": -0.13302375376224518, "fcm_dpo/margin": 61.51070785522461, "fcm_dpo/q_t": 0.30989372730255127, "grad_norm": 25.751888275146484, "learning_rate": 4.611919330113591e-07, "logits/chosen": -0.5168595314025879, "logits/rejected": -0.5018938183784485, "logps/chosen": -93.7746810913086, "logps/ref_chosen": -54.34272003173828, "logps/ref_rejected": -98.21183776855469, "logps/rejected": -199.15451049804688, "loss": 0.8218, "margin_dpo/margin_mean": 61.510704040527344, "margin_dpo/margin_std": 62.60074996948242, "step": 180 }, { "epoch": 0.2657856093979442, "fcm_dpo/beta": 0.016240477561950684, "fcm_dpo/delta": 0.2024461030960083, "fcm_dpo/margin": 40.64165496826172, "fcm_dpo/q_t": 0.3688216805458069, "grad_norm": 28.267780303955078, "learning_rate": 4.605024008834863e-07, "logits/chosen": -0.5522741079330444, "logits/rejected": -0.5164051055908203, "logps/chosen": -89.09597778320312, "logps/ref_chosen": -55.000457763671875, "logps/ref_rejected": -61.656166076660156, "logps/rejected": -136.39334106445312, "loss": 1.0021, "margin_dpo/margin_mean": 40.64165496826172, "margin_dpo/margin_std": 56.71002197265625, "step": 181 }, { "epoch": 0.26725403817914833, "fcm_dpo/beta": 0.015759017318487167, "fcm_dpo/delta": -0.20672228932380676, "fcm_dpo/margin": 65.56771850585938, "fcm_dpo/q_t": 0.30198216438293457, "grad_norm": 35.510986328125, "learning_rate": 4.598073218215817e-07, "logits/chosen": -0.5237354636192322, "logits/rejected": -0.5221885442733765, "logps/chosen": -75.54692077636719, "logps/ref_chosen": -41.107852935791016, "logps/ref_rejected": -89.5215835571289, "logps/rejected": -189.52838134765625, "loss": 0.8296, "margin_dpo/margin_mean": 65.56771850585938, "margin_dpo/margin_std": 69.53907775878906, "step": 182 }, { "epoch": 0.2687224669603524, "fcm_dpo/beta": 0.015870921313762665, "fcm_dpo/delta": 0.18142208456993103, "fcm_dpo/margin": 42.763145446777344, "fcm_dpo/q_t": 0.363423615694046, "grad_norm": 36.854557037353516, "learning_rate": 4.5910671414162484e-07, "logits/chosen": -0.5377739667892456, "logits/rejected": -0.5063446760177612, "logps/chosen": -108.02732849121094, "logps/ref_chosen": -57.52456283569336, "logps/ref_rejected": -75.97572326660156, "logps/rejected": -169.24163818359375, "loss": 0.9778, "margin_dpo/margin_mean": 42.763145446777344, "margin_dpo/margin_std": 50.845008850097656, "step": 183 }, { "epoch": 0.2701908957415565, "fcm_dpo/beta": 0.016876371577382088, "fcm_dpo/delta": 0.18617403507232666, "fcm_dpo/margin": 39.896549224853516, "fcm_dpo/q_t": 0.36848223209381104, "grad_norm": 28.983440399169922, "learning_rate": 4.5840059630527985e-07, "logits/chosen": -0.5860706567764282, "logits/rejected": -0.5714048743247986, "logps/chosen": -97.67625427246094, "logps/ref_chosen": -58.544952392578125, "logps/ref_rejected": -76.63406372070312, "logps/rejected": -155.66189575195312, "loss": 0.9873, "margin_dpo/margin_mean": 39.89654541015625, "margin_dpo/margin_std": 53.60821533203125, "step": 184 }, { "epoch": 0.27165932452276065, "fcm_dpo/beta": 0.017562009394168854, "fcm_dpo/delta": 0.273959755897522, "fcm_dpo/margin": 33.74722671508789, "fcm_dpo/q_t": 0.3876458406448364, "grad_norm": 37.04210662841797, "learning_rate": 4.5768898691940836e-07, "logits/chosen": -0.5385079383850098, "logits/rejected": -0.4919096827507019, "logps/chosen": -107.01687622070312, "logps/ref_chosen": -62.025848388671875, "logps/ref_rejected": -73.7625961303711, "logps/rejected": -152.5008544921875, "loss": 1.0766, "margin_dpo/margin_mean": 33.74722671508789, "margin_dpo/margin_std": 56.945701599121094, "step": 185 }, { "epoch": 0.27312775330396477, "fcm_dpo/beta": 0.017246492207050323, "fcm_dpo/delta": -0.20956075191497803, "fcm_dpo/margin": 60.13277816772461, "fcm_dpo/q_t": 0.29886382818222046, "grad_norm": 54.477352142333984, "learning_rate": 4.5697190473557947e-07, "logits/chosen": -0.5471476912498474, "logits/rejected": -0.5003129839897156, "logps/chosen": -112.1868667602539, "logps/ref_chosen": -69.35346984863281, "logps/ref_rejected": -88.07244873046875, "logps/rejected": -191.03863525390625, "loss": 0.7824, "margin_dpo/margin_mean": 60.132774353027344, "margin_dpo/margin_std": 59.29591369628906, "step": 186 }, { "epoch": 0.2745961820851689, "fcm_dpo/beta": 0.01733904704451561, "fcm_dpo/delta": 0.010633766651153564, "fcm_dpo/margin": 48.30116271972656, "fcm_dpo/q_t": 0.3396226167678833, "grad_norm": 53.08622741699219, "learning_rate": 4.5624936864957555e-07, "logits/chosen": -0.5102699398994446, "logits/rejected": -0.5049658417701721, "logps/chosen": -97.04034423828125, "logps/ref_chosen": -52.7564582824707, "logps/ref_rejected": -81.96910095214844, "logps/rejected": -174.5541534423828, "loss": 0.9272, "margin_dpo/margin_mean": 48.30116271972656, "margin_dpo/margin_std": 60.002471923828125, "step": 187 }, { "epoch": 0.27606461086637296, "fcm_dpo/beta": 0.016646115109324455, "fcm_dpo/delta": -0.19869595766067505, "fcm_dpo/margin": 61.76872253417969, "fcm_dpo/q_t": 0.3067232370376587, "grad_norm": 65.48516082763672, "learning_rate": 4.5552139770089454e-07, "logits/chosen": -0.4937124252319336, "logits/rejected": -0.4901246428489685, "logps/chosen": -89.49577331542969, "logps/ref_chosen": -49.415489196777344, "logps/ref_rejected": -89.54043579101562, "logps/rejected": -191.3894500732422, "loss": 0.817, "margin_dpo/margin_mean": 61.76872253417969, "margin_dpo/margin_std": 66.22196960449219, "step": 188 }, { "epoch": 0.2775330396475771, "fcm_dpo/beta": 0.016446605324745178, "fcm_dpo/delta": -0.023531386628746986, "fcm_dpo/margin": 52.98753356933594, "fcm_dpo/q_t": 0.33847445249557495, "grad_norm": 57.1566276550293, "learning_rate": 4.5478801107224794e-07, "logits/chosen": -0.4816792607307434, "logits/rejected": -0.4451908469200134, "logps/chosen": -101.60316467285156, "logps/ref_chosen": -52.39896011352539, "logps/ref_rejected": -72.16735076904297, "logps/rejected": -174.3590850830078, "loss": 0.9291, "margin_dpo/margin_mean": 52.98753356933594, "margin_dpo/margin_std": 70.50366973876953, "step": 189 }, { "epoch": 0.2790014684287812, "fcm_dpo/beta": 0.01647689938545227, "fcm_dpo/delta": -0.08300600945949554, "fcm_dpo/margin": 55.913509368896484, "fcm_dpo/q_t": 0.3302146792411804, "grad_norm": 43.29951477050781, "learning_rate": 4.5404922808905543e-07, "logits/chosen": -0.5044703483581543, "logits/rejected": -0.47580382227897644, "logps/chosen": -114.02642822265625, "logps/ref_chosen": -64.68305969238281, "logps/ref_rejected": -102.55052185058594, "logps/rejected": -207.80740356445312, "loss": 0.9189, "margin_dpo/margin_mean": 55.913509368896484, "margin_dpo/margin_std": 68.61431884765625, "step": 190 }, { "epoch": 0.28046989720998533, "fcm_dpo/beta": 0.015079159289598465, "fcm_dpo/delta": -0.40096384286880493, "fcm_dpo/margin": 79.48197174072266, "fcm_dpo/q_t": 0.2806900441646576, "grad_norm": 30.056840896606445, "learning_rate": 4.5330506821893565e-07, "logits/chosen": -0.4775508940219879, "logits/rejected": -0.46088314056396484, "logps/chosen": -111.59598541259766, "logps/ref_chosen": -68.65887451171875, "logps/ref_rejected": -110.1396713256836, "logps/rejected": -232.55877685546875, "loss": 0.7341, "margin_dpo/margin_mean": 79.48197174072266, "margin_dpo/margin_std": 80.22380065917969, "step": 191 }, { "epoch": 0.28193832599118945, "fcm_dpo/beta": 0.014918088912963867, "fcm_dpo/delta": -0.015953052788972855, "fcm_dpo/margin": 57.933658599853516, "fcm_dpo/q_t": 0.33742040395736694, "grad_norm": 35.163917541503906, "learning_rate": 4.5255555107119336e-07, "logits/chosen": -0.46824973821640015, "logits/rejected": -0.458457350730896, "logps/chosen": -126.23261260986328, "logps/ref_chosen": -69.72691345214844, "logps/ref_rejected": -103.32135009765625, "logps/rejected": -217.7606964111328, "loss": 0.9085, "margin_dpo/margin_mean": 57.93366241455078, "margin_dpo/margin_std": 71.61734008789062, "step": 192 }, { "epoch": 0.2834067547723935, "fcm_dpo/beta": 0.015718581154942513, "fcm_dpo/delta": 0.3292880654335022, "fcm_dpo/margin": 33.98414611816406, "fcm_dpo/q_t": 0.3883090615272522, "grad_norm": 56.01692199707031, "learning_rate": 4.5180069639630236e-07, "logits/chosen": -0.49984651803970337, "logits/rejected": -0.4666469395160675, "logps/chosen": -117.69091796875, "logps/ref_chosen": -60.19049835205078, "logps/ref_rejected": -76.40755462646484, "logps/rejected": -167.89212036132812, "loss": 1.1352, "margin_dpo/margin_mean": 33.98414611816406, "margin_dpo/margin_std": 63.35324478149414, "step": 193 }, { "epoch": 0.28487518355359764, "fcm_dpo/beta": 0.01578504592180252, "fcm_dpo/delta": -0.01547658909112215, "fcm_dpo/margin": 54.73874282836914, "fcm_dpo/q_t": 0.3235979378223419, "grad_norm": 30.953447341918945, "learning_rate": 4.510405240853854e-07, "logits/chosen": -0.4425453841686249, "logits/rejected": -0.4208832383155823, "logps/chosen": -73.02734375, "logps/ref_chosen": -37.84037399291992, "logps/ref_rejected": -60.684783935546875, "logps/rejected": -150.61050415039062, "loss": 0.8382, "margin_dpo/margin_mean": 54.73874282836914, "margin_dpo/margin_std": 53.040504455566406, "step": 194 }, { "epoch": 0.28634361233480177, "fcm_dpo/beta": 0.015666445717215538, "fcm_dpo/delta": -0.03931724652647972, "fcm_dpo/margin": 56.528350830078125, "fcm_dpo/q_t": 0.3216269016265869, "grad_norm": 33.33697509765625, "learning_rate": 4.5027505416968985e-07, "logits/chosen": -0.43572598695755005, "logits/rejected": -0.448942631483078, "logps/chosen": -111.7900390625, "logps/ref_chosen": -54.891571044921875, "logps/ref_rejected": -96.77095794677734, "logps/rejected": -210.19778442382812, "loss": 0.8393, "margin_dpo/margin_mean": 56.528350830078125, "margin_dpo/margin_std": 57.825653076171875, "step": 195 }, { "epoch": 0.2878120411160059, "fcm_dpo/beta": 0.015181340277194977, "fcm_dpo/delta": -0.17460601031780243, "fcm_dpo/margin": 66.21810913085938, "fcm_dpo/q_t": 0.3101344108581543, "grad_norm": 34.602691650390625, "learning_rate": 4.495043068200599e-07, "logits/chosen": -0.4735764265060425, "logits/rejected": -0.4467168152332306, "logps/chosen": -97.20225524902344, "logps/ref_chosen": -53.245243072509766, "logps/ref_rejected": -76.05294799804688, "logps/rejected": -186.2280731201172, "loss": 0.8288, "margin_dpo/margin_mean": 66.21810913085938, "margin_dpo/margin_std": 70.18364715576172, "step": 196 }, { "epoch": 0.28928046989721, "fcm_dpo/beta": 0.015351845882833004, "fcm_dpo/delta": 0.05590180307626724, "fcm_dpo/margin": 51.957035064697266, "fcm_dpo/q_t": 0.3400971293449402, "grad_norm": 36.760929107666016, "learning_rate": 4.4872830234640493e-07, "logits/chosen": -0.4230024814605713, "logits/rejected": -0.4091572165489197, "logps/chosen": -105.07384490966797, "logps/ref_chosen": -60.42033386230469, "logps/ref_rejected": -77.20890808105469, "logps/rejected": -173.8194580078125, "loss": 0.8925, "margin_dpo/margin_mean": 51.95703125, "margin_dpo/margin_std": 57.920127868652344, "step": 197 }, { "epoch": 0.2907488986784141, "fcm_dpo/beta": 0.015172794461250305, "fcm_dpo/delta": -0.09196484833955765, "fcm_dpo/margin": 61.515472412109375, "fcm_dpo/q_t": 0.3227071166038513, "grad_norm": 36.14360046386719, "learning_rate": 4.479470611971645e-07, "logits/chosen": -0.47960442304611206, "logits/rejected": -0.47948503494262695, "logps/chosen": -102.31593322753906, "logps/ref_chosen": -55.03618621826172, "logps/ref_rejected": -97.24325561523438, "logps/rejected": -206.03848266601562, "loss": 0.8429, "margin_dpo/margin_mean": 61.515472412109375, "margin_dpo/margin_std": 70.07635498046875, "step": 198 }, { "epoch": 0.2922173274596182, "fcm_dpo/beta": 0.014803415164351463, "fcm_dpo/delta": -0.04618511348962784, "fcm_dpo/margin": 60.14692687988281, "fcm_dpo/q_t": 0.32798993587493896, "grad_norm": 36.83704376220703, "learning_rate": 4.471606039587695e-07, "logits/chosen": -0.45688751339912415, "logits/rejected": -0.44097527861595154, "logps/chosen": -105.17231750488281, "logps/ref_chosen": -56.828826904296875, "logps/ref_rejected": -84.64820861816406, "logps/rejected": -193.1386260986328, "loss": 0.8872, "margin_dpo/margin_mean": 60.14692687988281, "margin_dpo/margin_std": 69.28252410888672, "step": 199 }, { "epoch": 0.2936857562408223, "fcm_dpo/beta": 0.014579104259610176, "fcm_dpo/delta": -0.10652126371860504, "fcm_dpo/margin": 64.84251403808594, "fcm_dpo/q_t": 0.3242376148700714, "grad_norm": 37.86602783203125, "learning_rate": 4.4636895135509966e-07, "logits/chosen": -0.4370264708995819, "logits/rejected": -0.41041016578674316, "logps/chosen": -102.37236022949219, "logps/ref_chosen": -53.06706237792969, "logps/ref_rejected": -80.60843658447266, "logps/rejected": -194.75624084472656, "loss": 0.8901, "margin_dpo/margin_mean": 64.84251403808594, "margin_dpo/margin_std": 79.6728515625, "step": 200 }, { "epoch": 0.2936857562408223, "eval_fcm_dpo/beta": 0.014577739872038364, "eval_logits/chosen": -0.48076131939888, "eval_logits/rejected": -0.46049416065216064, "eval_logps/chosen": -146.9453125, "eval_logps/ref_chosen": -79.05104064941406, "eval_logps/ref_rejected": -86.79793548583984, "eval_logps/rejected": -191.740234375, "eval_loss": 0.5793219804763794, "eval_margin_dpo/margin_mean": 37.048030853271484, "eval_margin_dpo/margin_std": 73.52235412597656, "eval_runtime": 39.2951, "eval_samples_per_second": 59.524, "eval_steps_per_second": 1.883, "step": 200 }, { "epoch": 0.29515418502202645, "fcm_dpo/beta": 0.014311077073216438, "fcm_dpo/delta": -0.11738879978656769, "fcm_dpo/margin": 66.80262756347656, "fcm_dpo/q_t": 0.31848424673080444, "grad_norm": 38.43661880493164, "learning_rate": 4.455721242469372e-07, "logits/chosen": -0.4631321132183075, "logits/rejected": -0.4501439929008484, "logps/chosen": -125.53224182128906, "logps/ref_chosen": -75.4022216796875, "logps/ref_rejected": -114.80821990966797, "logps/rejected": -231.74085998535156, "loss": 0.8657, "margin_dpo/margin_mean": 66.80262756347656, "margin_dpo/margin_std": 77.89523315429688, "step": 201 }, { "epoch": 0.2966226138032305, "fcm_dpo/beta": 0.014638571999967098, "fcm_dpo/delta": 0.17033323645591736, "fcm_dpo/margin": 47.20063018798828, "fcm_dpo/q_t": 0.3694015145301819, "grad_norm": 47.15175247192383, "learning_rate": 4.4477014363141755e-07, "logits/chosen": -0.41728901863098145, "logits/rejected": -0.4195215702056885, "logps/chosen": -106.94139099121094, "logps/ref_chosen": -50.101318359375, "logps/ref_rejected": -86.98503112792969, "logps/rejected": -191.0257568359375, "loss": 1.0404, "margin_dpo/margin_mean": 47.20063400268555, "margin_dpo/margin_std": 73.57470703125, "step": 202 }, { "epoch": 0.29809104258443464, "fcm_dpo/beta": 0.01473167072981596, "fcm_dpo/delta": 0.005034193862229586, "fcm_dpo/margin": 57.385345458984375, "fcm_dpo/q_t": 0.33291637897491455, "grad_norm": 36.741172790527344, "learning_rate": 4.439630306414758e-07, "logits/chosen": -0.4235112965106964, "logits/rejected": -0.4039536118507385, "logps/chosen": -114.47615051269531, "logps/ref_chosen": -60.60969543457031, "logps/ref_rejected": -85.89596557617188, "logps/rejected": -197.14776611328125, "loss": 0.8768, "margin_dpo/margin_mean": 57.385345458984375, "margin_dpo/margin_std": 63.92000961303711, "step": 203 }, { "epoch": 0.29955947136563876, "fcm_dpo/beta": 0.01482885517179966, "fcm_dpo/delta": 0.04627120867371559, "fcm_dpo/margin": 54.453792572021484, "fcm_dpo/q_t": 0.3538212776184082, "grad_norm": 52.60590744018555, "learning_rate": 4.431508065452897e-07, "logits/chosen": -0.4854148030281067, "logits/rejected": -0.44390296936035156, "logps/chosen": -144.6455078125, "logps/ref_chosen": -80.16496276855469, "logps/ref_rejected": -87.69590759277344, "logps/rejected": -206.6302490234375, "loss": 0.9524, "margin_dpo/margin_mean": 54.453792572021484, "margin_dpo/margin_std": 75.84136199951172, "step": 204 }, { "epoch": 0.3010279001468429, "fcm_dpo/beta": 0.014350036159157753, "fcm_dpo/delta": -0.1424178183078766, "fcm_dpo/margin": 67.729736328125, "fcm_dpo/q_t": 0.3128374218940735, "grad_norm": 36.14482116699219, "learning_rate": 4.4233349274571974e-07, "logits/chosen": -0.37768441438674927, "logits/rejected": -0.33807796239852905, "logps/chosen": -124.49932861328125, "logps/ref_chosen": -59.384735107421875, "logps/ref_rejected": -85.12505340576172, "logps/rejected": -217.96939086914062, "loss": 0.8763, "margin_dpo/margin_mean": 67.729736328125, "margin_dpo/margin_std": 74.6842269897461, "step": 205 }, { "epoch": 0.302496328928047, "fcm_dpo/beta": 0.013974905014038086, "fcm_dpo/delta": -0.1962062567472458, "fcm_dpo/margin": 73.30876159667969, "fcm_dpo/q_t": 0.3037898540496826, "grad_norm": 53.73223876953125, "learning_rate": 4.415111107797445e-07, "logits/chosen": -0.370413213968277, "logits/rejected": -0.37365084886550903, "logps/chosen": -109.06692504882812, "logps/ref_chosen": -46.964500427246094, "logps/ref_rejected": -98.9534912109375, "logps/rejected": -234.36468505859375, "loss": 0.8314, "margin_dpo/margin_mean": 73.30876159667969, "margin_dpo/margin_std": 78.49857330322266, "step": 206 }, { "epoch": 0.3039647577092511, "fcm_dpo/beta": 0.013306405395269394, "fcm_dpo/delta": -0.31155675649642944, "fcm_dpo/margin": 84.67915344238281, "fcm_dpo/q_t": 0.2934865355491638, "grad_norm": 29.86716651916504, "learning_rate": 4.4068368231789365e-07, "logits/chosen": -0.4337068796157837, "logits/rejected": -0.39963075518608093, "logps/chosen": -102.7784194946289, "logps/ref_chosen": -56.05625915527344, "logps/ref_rejected": -84.44779968261719, "logps/rejected": -215.84910583496094, "loss": 0.7677, "margin_dpo/margin_mean": 84.67914581298828, "margin_dpo/margin_std": 87.53607177734375, "step": 207 }, { "epoch": 0.3054331864904552, "fcm_dpo/beta": 0.01302929688245058, "fcm_dpo/delta": -0.000983063131570816, "fcm_dpo/margin": 65.29347229003906, "fcm_dpo/q_t": 0.33387383818626404, "grad_norm": 40.39805221557617, "learning_rate": 4.398512291636768e-07, "logits/chosen": -0.5328778028488159, "logits/rejected": -0.5141773819923401, "logps/chosen": -137.27557373046875, "logps/ref_chosen": -67.06761169433594, "logps/ref_rejected": -94.28689575195312, "logps/rejected": -229.78834533691406, "loss": 0.9107, "margin_dpo/margin_mean": 65.29347229003906, "margin_dpo/margin_std": 79.34541320800781, "step": 208 }, { "epoch": 0.3069016152716593, "fcm_dpo/beta": 0.013248629868030548, "fcm_dpo/delta": 0.10473860800266266, "fcm_dpo/margin": 56.855796813964844, "fcm_dpo/q_t": 0.3515823483467102, "grad_norm": 38.002723693847656, "learning_rate": 4.3901377325300857e-07, "logits/chosen": -0.43297621607780457, "logits/rejected": -0.41614365577697754, "logps/chosen": -114.31430053710938, "logps/ref_chosen": -56.18169403076172, "logps/ref_rejected": -80.94152069091797, "logps/rejected": -195.929931640625, "loss": 0.9608, "margin_dpo/margin_mean": 56.85579299926758, "margin_dpo/margin_std": 72.0236587524414, "step": 209 }, { "epoch": 0.30837004405286345, "fcm_dpo/beta": 0.013421540148556232, "fcm_dpo/delta": 0.023709196597337723, "fcm_dpo/margin": 61.6995735168457, "fcm_dpo/q_t": 0.3359639644622803, "grad_norm": 31.023426055908203, "learning_rate": 4.381713366536311e-07, "logits/chosen": -0.4762294888496399, "logits/rejected": -0.4587492346763611, "logps/chosen": -98.13655090332031, "logps/ref_chosen": -46.371822357177734, "logps/ref_rejected": -76.68162536621094, "logps/rejected": -190.1459197998047, "loss": 0.8813, "margin_dpo/margin_mean": 61.69956970214844, "margin_dpo/margin_std": 67.59247589111328, "step": 210 }, { "epoch": 0.30983847283406757, "fcm_dpo/beta": 0.013691660948097706, "fcm_dpo/delta": 0.142348051071167, "fcm_dpo/margin": 52.43247985839844, "fcm_dpo/q_t": 0.36083507537841797, "grad_norm": 44.8113899230957, "learning_rate": 4.373239415645323e-07, "logits/chosen": -0.49992769956588745, "logits/rejected": -0.4475579857826233, "logps/chosen": -151.08291625976562, "logps/ref_chosen": -78.93235778808594, "logps/ref_rejected": -86.82098388671875, "logps/rejected": -211.40402221679688, "loss": 0.9757, "margin_dpo/margin_mean": 52.432472229003906, "margin_dpo/margin_std": 69.52543640136719, "step": 211 }, { "epoch": 0.31130690161527164, "fcm_dpo/beta": 0.0132517721503973, "fcm_dpo/delta": -0.173972487449646, "fcm_dpo/margin": 75.48409271240234, "fcm_dpo/q_t": 0.31110864877700806, "grad_norm": 29.362991333007812, "learning_rate": 4.3647161031536086e-07, "logits/chosen": -0.5245535373687744, "logits/rejected": -0.51213139295578, "logps/chosen": -115.64869689941406, "logps/ref_chosen": -58.19701385498047, "logps/ref_rejected": -103.05785369873047, "logps/rejected": -235.99363708496094, "loss": 0.8399, "margin_dpo/margin_mean": 75.48409271240234, "margin_dpo/margin_std": 78.64497375488281, "step": 212 }, { "epoch": 0.31277533039647576, "fcm_dpo/beta": 0.012959420680999756, "fcm_dpo/delta": -0.1306145340204239, "fcm_dpo/margin": 74.51495361328125, "fcm_dpo/q_t": 0.3106541037559509, "grad_norm": 33.29969787597656, "learning_rate": 4.3561436536583774e-07, "logits/chosen": -0.49456068873405457, "logits/rejected": -0.4497288167476654, "logps/chosen": -117.433349609375, "logps/ref_chosen": -67.51271057128906, "logps/ref_rejected": -93.91471862792969, "logps/rejected": -218.35031127929688, "loss": 0.8229, "margin_dpo/margin_mean": 74.51496124267578, "margin_dpo/margin_std": 75.42293548583984, "step": 213 }, { "epoch": 0.3142437591776799, "fcm_dpo/beta": 0.013114094734191895, "fcm_dpo/delta": 0.05126992613077164, "fcm_dpo/margin": 61.19904327392578, "fcm_dpo/q_t": 0.3447904884815216, "grad_norm": 33.585716247558594, "learning_rate": 4.3475222930516473e-07, "logits/chosen": -0.4499906897544861, "logits/rejected": -0.4545586109161377, "logps/chosen": -91.10122680664062, "logps/ref_chosen": -41.604888916015625, "logps/ref_rejected": -77.51741027832031, "logps/rejected": -188.21279907226562, "loss": 0.9278, "margin_dpo/margin_mean": 61.19904327392578, "margin_dpo/margin_std": 76.41812896728516, "step": 214 }, { "epoch": 0.315712187958884, "fcm_dpo/beta": 0.013036966323852539, "fcm_dpo/delta": -0.00907570868730545, "fcm_dpo/margin": 65.80323791503906, "fcm_dpo/q_t": 0.3300132751464844, "grad_norm": 35.7015380859375, "learning_rate": 4.3388522485142885e-07, "logits/chosen": -0.5079599022865295, "logits/rejected": -0.49217790365219116, "logps/chosen": -107.33973693847656, "logps/ref_chosen": -53.279266357421875, "logps/ref_rejected": -89.96464538574219, "logps/rejected": -209.82835388183594, "loss": 0.8646, "margin_dpo/margin_mean": 65.80323791503906, "margin_dpo/margin_std": 71.56494140625, "step": 215 }, { "epoch": 0.31718061674008813, "fcm_dpo/beta": 0.013059217482805252, "fcm_dpo/delta": -0.06660346686840057, "fcm_dpo/margin": 69.67720794677734, "fcm_dpo/q_t": 0.32552677392959595, "grad_norm": 29.45123863220215, "learning_rate": 4.330133748510036e-07, "logits/chosen": -0.4530643820762634, "logits/rejected": -0.4277815818786621, "logps/chosen": -105.15579223632812, "logps/ref_chosen": -48.887794494628906, "logps/ref_rejected": -77.19892883300781, "logps/rejected": -203.14413452148438, "loss": 0.8853, "margin_dpo/margin_mean": 69.67721557617188, "margin_dpo/margin_std": 79.81889343261719, "step": 216 }, { "epoch": 0.3186490455212922, "fcm_dpo/beta": 0.012638933025300503, "fcm_dpo/delta": -0.13730451464653015, "fcm_dpo/margin": 77.040771484375, "fcm_dpo/q_t": 0.30401867628097534, "grad_norm": 30.868886947631836, "learning_rate": 4.3213670227794757e-07, "logits/chosen": -0.49128514528274536, "logits/rejected": -0.47628241777420044, "logps/chosen": -102.47686767578125, "logps/ref_chosen": -49.845306396484375, "logps/ref_rejected": -100.07832336425781, "logps/rejected": -229.75064086914062, "loss": 0.7844, "margin_dpo/margin_mean": 77.040771484375, "margin_dpo/margin_std": 70.88652801513672, "step": 217 }, { "epoch": 0.3201174743024963, "fcm_dpo/beta": 0.012668648734688759, "fcm_dpo/delta": 0.0585474818944931, "fcm_dpo/margin": 62.8431282043457, "fcm_dpo/q_t": 0.3426221013069153, "grad_norm": 26.22398567199707, "learning_rate": 4.3125523023339815e-07, "logits/chosen": -0.49315348267555237, "logits/rejected": -0.47417598962783813, "logps/chosen": -115.05643463134766, "logps/ref_chosen": -58.576683044433594, "logps/ref_rejected": -87.84639739990234, "logps/rejected": -207.16928100585938, "loss": 0.9079, "margin_dpo/margin_mean": 62.84312438964844, "margin_dpo/margin_std": 73.27041625976562, "step": 218 }, { "epoch": 0.32158590308370044, "fcm_dpo/beta": 0.01306487713009119, "fcm_dpo/delta": 0.1448213905096054, "fcm_dpo/margin": 54.6587028503418, "fcm_dpo/q_t": 0.3576211929321289, "grad_norm": 34.506874084472656, "learning_rate": 4.303689819449636e-07, "logits/chosen": -0.4658815264701843, "logits/rejected": -0.44217032194137573, "logps/chosen": -120.91069030761719, "logps/ref_chosen": -61.083858489990234, "logps/ref_rejected": -85.83042907714844, "logps/rejected": -200.3159637451172, "loss": 0.9701, "margin_dpo/margin_mean": 54.6587028503418, "margin_dpo/margin_std": 71.40689849853516, "step": 219 }, { "epoch": 0.32305433186490456, "fcm_dpo/beta": 0.013401372358202934, "fcm_dpo/delta": 0.20168578624725342, "fcm_dpo/margin": 49.403968811035156, "fcm_dpo/q_t": 0.3585790991783142, "grad_norm": 34.5196418762207, "learning_rate": 4.2947798076611047e-07, "logits/chosen": -0.4772907495498657, "logits/rejected": -0.44704103469848633, "logps/chosen": -136.77674865722656, "logps/ref_chosen": -70.03128051757812, "logps/ref_rejected": -87.68551635742188, "logps/rejected": -203.83493041992188, "loss": 0.9412, "margin_dpo/margin_mean": 49.403968811035156, "margin_dpo/margin_std": 51.8128662109375, "step": 220 }, { "epoch": 0.3245227606461087, "fcm_dpo/beta": 0.01300697959959507, "fcm_dpo/delta": -0.36966100335121155, "fcm_dpo/margin": 90.53009033203125, "fcm_dpo/q_t": 0.2712687849998474, "grad_norm": 39.37263107299805, "learning_rate": 4.285822501755485e-07, "logits/chosen": -0.41888752579689026, "logits/rejected": -0.42424139380455017, "logps/chosen": -122.67888641357422, "logps/ref_chosen": -52.15470886230469, "logps/ref_rejected": -106.46768188476562, "logps/rejected": -267.52197265625, "loss": 0.7145, "margin_dpo/margin_mean": 90.53008270263672, "margin_dpo/margin_std": 76.58544921875, "step": 221 }, { "epoch": 0.32599118942731276, "fcm_dpo/beta": 0.012651478871703148, "fcm_dpo/delta": -0.035454705357551575, "fcm_dpo/margin": 69.71733856201172, "fcm_dpo/q_t": 0.3251156210899353, "grad_norm": 42.33450698852539, "learning_rate": 4.276818137766118e-07, "logits/chosen": -0.4270946979522705, "logits/rejected": -0.40993452072143555, "logps/chosen": -134.14321899414062, "logps/ref_chosen": -60.971099853515625, "logps/ref_rejected": -100.00115203857422, "logps/rejected": -242.89060974121094, "loss": 0.8663, "margin_dpo/margin_mean": 69.71733856201172, "margin_dpo/margin_std": 75.56416320800781, "step": 222 }, { "epoch": 0.3274596182085169, "fcm_dpo/beta": 0.01269703172147274, "fcm_dpo/delta": 0.07957568764686584, "fcm_dpo/margin": 61.16942596435547, "fcm_dpo/q_t": 0.34925732016563416, "grad_norm": 45.7380485534668, "learning_rate": 4.2677669529663686e-07, "logits/chosen": -0.3831120729446411, "logits/rejected": -0.361503005027771, "logps/chosen": -134.04067993164062, "logps/ref_chosen": -52.64057540893555, "logps/ref_rejected": -82.82502746582031, "logps/rejected": -225.39456176757812, "loss": 1.0058, "margin_dpo/margin_mean": 61.16942596435547, "margin_dpo/margin_std": 87.82357025146484, "step": 223 }, { "epoch": 0.328928046989721, "fcm_dpo/beta": 0.012370316311717033, "fcm_dpo/delta": -0.12362442910671234, "fcm_dpo/margin": 77.3526611328125, "fcm_dpo/q_t": 0.3199881315231323, "grad_norm": 43.84239196777344, "learning_rate": 4.2586691858633747e-07, "logits/chosen": -0.4017111659049988, "logits/rejected": -0.3758423328399658, "logps/chosen": -117.4913330078125, "logps/ref_chosen": -48.59541320800781, "logps/ref_rejected": -77.11648559570312, "logps/rejected": -223.36508178710938, "loss": 0.863, "margin_dpo/margin_mean": 77.3526611328125, "margin_dpo/margin_std": 87.10986328125, "step": 224 }, { "epoch": 0.3303964757709251, "fcm_dpo/beta": 0.011899469420313835, "fcm_dpo/delta": -0.2886032462120056, "fcm_dpo/margin": 92.91471862792969, "fcm_dpo/q_t": 0.29158496856689453, "grad_norm": 33.034423828125, "learning_rate": 4.249525076191759e-07, "logits/chosen": -0.39564913511276245, "logits/rejected": -0.3767988085746765, "logps/chosen": -139.4900665283203, "logps/ref_chosen": -58.000465393066406, "logps/ref_rejected": -99.90291595458984, "logps/rejected": -274.3072204589844, "loss": 0.784, "margin_dpo/margin_mean": 92.91471862792969, "margin_dpo/margin_std": 94.30024719238281, "step": 225 }, { "epoch": 0.33186490455212925, "fcm_dpo/beta": 0.01162062305957079, "fcm_dpo/delta": 0.011304348707199097, "fcm_dpo/margin": 71.96773529052734, "fcm_dpo/q_t": 0.34078845381736755, "grad_norm": 36.67145538330078, "learning_rate": 4.2403348649073167e-07, "logits/chosen": -0.4350174069404602, "logits/rejected": -0.3871588110923767, "logps/chosen": -128.597900390625, "logps/ref_chosen": -58.898799896240234, "logps/ref_rejected": -78.68775939941406, "logps/rejected": -220.35458374023438, "loss": 0.9178, "margin_dpo/margin_mean": 71.96773529052734, "margin_dpo/margin_std": 87.11494445800781, "step": 226 }, { "epoch": 0.3333333333333333, "fcm_dpo/beta": 0.011324774473905563, "fcm_dpo/delta": -0.20764365792274475, "fcm_dpo/margin": 91.20738983154297, "fcm_dpo/q_t": 0.2983647286891937, "grad_norm": 33.768497467041016, "learning_rate": 4.2310987941806615e-07, "logits/chosen": -0.3924615979194641, "logits/rejected": -0.37529265880584717, "logps/chosen": -140.55740356445312, "logps/ref_chosen": -59.072181701660156, "logps/ref_rejected": -99.41236877441406, "logps/rejected": -272.10498046875, "loss": 0.8177, "margin_dpo/margin_mean": 91.20738220214844, "margin_dpo/margin_std": 95.2232666015625, "step": 227 }, { "epoch": 0.33480176211453744, "fcm_dpo/beta": 0.011514011770486832, "fcm_dpo/delta": 0.1022706851363182, "fcm_dpo/margin": 65.56757354736328, "fcm_dpo/q_t": 0.349528431892395, "grad_norm": 39.931148529052734, "learning_rate": 4.2218171073908463e-07, "logits/chosen": -0.3640768527984619, "logits/rejected": -0.340065598487854, "logps/chosen": -146.19509887695312, "logps/ref_chosen": -65.89128875732422, "logps/ref_rejected": -91.04875183105469, "logps/rejected": -236.92013549804688, "loss": 0.9489, "margin_dpo/margin_mean": 65.56758117675781, "margin_dpo/margin_std": 82.21015930175781, "step": 228 }, { "epoch": 0.33627019089574156, "fcm_dpo/beta": 0.011491687968373299, "fcm_dpo/delta": -0.03256909176707268, "fcm_dpo/margin": 76.55232238769531, "fcm_dpo/q_t": 0.3270722031593323, "grad_norm": 46.43395233154297, "learning_rate": 4.212490049118951e-07, "logits/chosen": -0.4288903772830963, "logits/rejected": -0.3912368714809418, "logps/chosen": -151.40249633789062, "logps/ref_chosen": -70.70637512207031, "logps/ref_rejected": -84.52741241455078, "logps/rejected": -241.77587890625, "loss": 0.8898, "margin_dpo/margin_mean": 76.55232238769531, "margin_dpo/margin_std": 88.84889221191406, "step": 229 }, { "epoch": 0.3377386196769457, "fcm_dpo/beta": 0.010963954031467438, "fcm_dpo/delta": -0.26099899411201477, "fcm_dpo/margin": 98.563720703125, "fcm_dpo/q_t": 0.28947991132736206, "grad_norm": 47.571136474609375, "learning_rate": 4.203117865141635e-07, "logits/chosen": -0.33261680603027344, "logits/rejected": -0.33441460132598877, "logps/chosen": -108.84280395507812, "logps/ref_chosen": -39.282005310058594, "logps/ref_rejected": -85.62191009521484, "logps/rejected": -253.74642944335938, "loss": 0.7737, "margin_dpo/margin_mean": 98.563720703125, "margin_dpo/margin_std": 94.94647216796875, "step": 230 }, { "epoch": 0.3392070484581498, "fcm_dpo/beta": 0.010946284979581833, "fcm_dpo/delta": 0.0592944361269474, "fcm_dpo/margin": 72.66645812988281, "fcm_dpo/q_t": 0.3454594314098358, "grad_norm": 39.91952896118164, "learning_rate": 4.1937008024246625e-07, "logits/chosen": -0.3895813822746277, "logits/rejected": -0.3568507432937622, "logps/chosen": -135.733642578125, "logps/ref_chosen": -63.27644348144531, "logps/ref_rejected": -74.1239013671875, "logps/rejected": -219.24757385253906, "loss": 0.9076, "margin_dpo/margin_mean": 72.66645812988281, "margin_dpo/margin_std": 87.05967712402344, "step": 231 }, { "epoch": 0.3406754772393539, "fcm_dpo/beta": 0.011238744482398033, "fcm_dpo/delta": 0.17818310856819153, "fcm_dpo/margin": 60.888694763183594, "fcm_dpo/q_t": 0.3727785050868988, "grad_norm": 45.982215881347656, "learning_rate": 4.1842391091163933e-07, "logits/chosen": -0.40042543411254883, "logits/rejected": -0.3696829676628113, "logps/chosen": -156.334228515625, "logps/ref_chosen": -70.74876403808594, "logps/ref_rejected": -83.97706604003906, "logps/rejected": -230.45123291015625, "loss": 1.0139, "margin_dpo/margin_mean": 60.888694763183594, "margin_dpo/margin_std": 91.31798553466797, "step": 232 }, { "epoch": 0.342143906020558, "fcm_dpo/beta": 0.011158278211951256, "fcm_dpo/delta": -0.15974467992782593, "fcm_dpo/margin": 89.08323669433594, "fcm_dpo/q_t": 0.3230384588241577, "grad_norm": 36.23651885986328, "learning_rate": 4.174733034541245e-07, "logits/chosen": -0.3902140259742737, "logits/rejected": -0.3841424584388733, "logps/chosen": -128.5865478515625, "logps/ref_chosen": -54.8829345703125, "logps/ref_rejected": -107.4800796508789, "logps/rejected": -270.26690673828125, "loss": 0.8862, "margin_dpo/margin_mean": 89.08323669433594, "margin_dpo/margin_std": 109.74813079833984, "step": 233 }, { "epoch": 0.3436123348017621, "fcm_dpo/beta": 0.010733449831604958, "fcm_dpo/delta": -0.12230968475341797, "fcm_dpo/margin": 89.16376495361328, "fcm_dpo/q_t": 0.3087769150733948, "grad_norm": 52.80620193481445, "learning_rate": 4.165182829193126e-07, "logits/chosen": -0.3361210823059082, "logits/rejected": -0.36225271224975586, "logps/chosen": -117.68397521972656, "logps/ref_chosen": -44.094520568847656, "logps/ref_rejected": -100.00663757324219, "logps/rejected": -262.7598571777344, "loss": 0.8077, "margin_dpo/margin_mean": 89.16377258300781, "margin_dpo/margin_std": 85.72807312011719, "step": 234 }, { "epoch": 0.34508076358296624, "fcm_dpo/beta": 0.011142631061375141, "fcm_dpo/delta": 0.19422681629657745, "fcm_dpo/margin": 59.925628662109375, "fcm_dpo/q_t": 0.366345077753067, "grad_norm": 39.07419204711914, "learning_rate": 4.1555887447288255e-07, "logits/chosen": -0.3953653573989868, "logits/rejected": -0.3707156777381897, "logps/chosen": -151.11203002929688, "logps/ref_chosen": -62.237911224365234, "logps/ref_rejected": -90.39506530761719, "logps/rejected": -239.19482421875, "loss": 0.9945, "margin_dpo/margin_mean": 59.925628662109375, "margin_dpo/margin_std": 81.4887466430664, "step": 235 }, { "epoch": 0.3465491923641703, "fcm_dpo/beta": 0.011004526168107986, "fcm_dpo/delta": -0.1253395676612854, "fcm_dpo/margin": 87.53258514404297, "fcm_dpo/q_t": 0.3034972548484802, "grad_norm": 45.433135986328125, "learning_rate": 4.1459510339613946e-07, "logits/chosen": -0.36071985960006714, "logits/rejected": -0.3635398745536804, "logps/chosen": -114.3785400390625, "logps/ref_chosen": -49.34136199951172, "logps/ref_rejected": -103.51162719726562, "logps/rejected": -256.0813903808594, "loss": 0.7776, "margin_dpo/margin_mean": 87.5325927734375, "margin_dpo/margin_std": 76.69570922851562, "step": 236 }, { "epoch": 0.34801762114537443, "fcm_dpo/beta": 0.010965939611196518, "fcm_dpo/delta": 0.021763239055871964, "fcm_dpo/margin": 75.69317626953125, "fcm_dpo/q_t": 0.3360444903373718, "grad_norm": 35.225894927978516, "learning_rate": 4.136269950853473e-07, "logits/chosen": -0.3720950484275818, "logits/rejected": -0.36715224385261536, "logps/chosen": -138.06900024414062, "logps/ref_chosen": -54.168121337890625, "logps/ref_rejected": -94.78036499023438, "logps/rejected": -254.3744354248047, "loss": 0.9023, "margin_dpo/margin_mean": 75.69317626953125, "margin_dpo/margin_std": 88.33975219726562, "step": 237 }, { "epoch": 0.34948604992657856, "fcm_dpo/beta": 0.011054374277591705, "fcm_dpo/delta": 0.08588720113039017, "fcm_dpo/margin": 69.70079040527344, "fcm_dpo/q_t": 0.34868574142456055, "grad_norm": 44.55629348754883, "learning_rate": 4.126545750510605e-07, "logits/chosen": -0.36067309975624084, "logits/rejected": -0.3768489360809326, "logps/chosen": -133.702880859375, "logps/ref_chosen": -53.973121643066406, "logps/ref_rejected": -89.41795349121094, "logps/rejected": -238.84848022460938, "loss": 0.917, "margin_dpo/margin_mean": 69.70079040527344, "margin_dpo/margin_std": 83.06000518798828, "step": 238 }, { "epoch": 0.3509544787077827, "fcm_dpo/beta": 0.010967254638671875, "fcm_dpo/delta": -0.034537769854068756, "fcm_dpo/margin": 80.02545166015625, "fcm_dpo/q_t": 0.32616370916366577, "grad_norm": 34.387508392333984, "learning_rate": 4.116778689174514e-07, "logits/chosen": -0.34563374519348145, "logits/rejected": -0.3208710551261902, "logps/chosen": -145.20004272460938, "logps/ref_chosen": -58.09782409667969, "logps/ref_rejected": -93.59294128417969, "logps/rejected": -260.7206115722656, "loss": 0.8603, "margin_dpo/margin_mean": 80.02545166015625, "margin_dpo/margin_std": 81.53392028808594, "step": 239 }, { "epoch": 0.3524229074889868, "fcm_dpo/beta": 0.011072592809796333, "fcm_dpo/delta": 0.026514161378145218, "fcm_dpo/margin": 74.52180480957031, "fcm_dpo/q_t": 0.3355349600315094, "grad_norm": 43.181861877441406, "learning_rate": 4.106969024216348e-07, "logits/chosen": -0.39122214913368225, "logits/rejected": -0.3691304326057434, "logps/chosen": -150.86891174316406, "logps/ref_chosen": -60.6144905090332, "logps/ref_rejected": -74.1185302734375, "logps/rejected": -238.894775390625, "loss": 0.9114, "margin_dpo/margin_mean": 74.52180480957031, "margin_dpo/margin_std": 85.87225341796875, "step": 240 }, { "epoch": 0.35389133627019087, "fcm_dpo/beta": 0.010732254013419151, "fcm_dpo/delta": -0.28711360692977905, "fcm_dpo/margin": 103.05683898925781, "fcm_dpo/q_t": 0.2914666533470154, "grad_norm": 35.48265838623047, "learning_rate": 4.097117014129903e-07, "logits/chosen": -0.4265785217285156, "logits/rejected": -0.39695611596107483, "logps/chosen": -144.91229248046875, "logps/ref_chosen": -66.091064453125, "logps/ref_rejected": -88.06088256835938, "logps/rejected": -269.93896484375, "loss": 0.7766, "margin_dpo/margin_mean": 103.05683898925781, "margin_dpo/margin_std": 102.60064697265625, "step": 241 }, { "epoch": 0.355359765051395, "fcm_dpo/beta": 0.01074141077697277, "fcm_dpo/delta": 0.087952621281147, "fcm_dpo/margin": 71.43954467773438, "fcm_dpo/q_t": 0.34943699836730957, "grad_norm": 43.884765625, "learning_rate": 4.087222918524807e-07, "logits/chosen": -0.31767386198043823, "logits/rejected": -0.2909330725669861, "logps/chosen": -160.47384643554688, "logps/ref_chosen": -67.86392974853516, "logps/ref_rejected": -83.36033630371094, "logps/rejected": -247.4097900390625, "loss": 0.9196, "margin_dpo/margin_mean": 71.43954467773438, "margin_dpo/margin_std": 86.29911804199219, "step": 242 }, { "epoch": 0.3568281938325991, "fcm_dpo/beta": 0.01043357141315937, "fcm_dpo/delta": -0.16880205273628235, "fcm_dpo/margin": 95.98949432373047, "fcm_dpo/q_t": 0.3014954924583435, "grad_norm": 38.88534927368164, "learning_rate": 4.07728699811968e-07, "logits/chosen": -0.33565038442611694, "logits/rejected": -0.28926849365234375, "logps/chosen": -157.85023498535156, "logps/ref_chosen": -63.0842399597168, "logps/ref_rejected": -76.33563232421875, "logps/rejected": -267.09112548828125, "loss": 0.7836, "margin_dpo/margin_mean": 95.98949432373047, "margin_dpo/margin_std": 91.47665405273438, "step": 243 }, { "epoch": 0.35829662261380324, "fcm_dpo/beta": 0.010181013494729996, "fcm_dpo/delta": -0.10228382050991058, "fcm_dpo/margin": 92.55367279052734, "fcm_dpo/q_t": 0.30694684386253357, "grad_norm": 43.90488052368164, "learning_rate": 4.067309514735267e-07, "logits/chosen": -0.39476436376571655, "logits/rejected": -0.3839525282382965, "logps/chosen": -143.63925170898438, "logps/ref_chosen": -61.140689849853516, "logps/ref_rejected": -94.89193725585938, "logps/rejected": -269.9441833496094, "loss": 0.7917, "margin_dpo/margin_mean": 92.55368041992188, "margin_dpo/margin_std": 81.99246215820312, "step": 244 }, { "epoch": 0.35976505139500736, "fcm_dpo/beta": 0.010019933804869652, "fcm_dpo/delta": 0.03025764226913452, "fcm_dpo/margin": 81.50281524658203, "fcm_dpo/q_t": 0.34738224744796753, "grad_norm": 61.226661682128906, "learning_rate": 4.057290731287531e-07, "logits/chosen": -0.40170228481292725, "logits/rejected": -0.3602461814880371, "logps/chosen": -163.54153442382812, "logps/ref_chosen": -67.26228332519531, "logps/ref_rejected": -87.64010620117188, "logps/rejected": -265.4221496582031, "loss": 0.947, "margin_dpo/margin_mean": 81.50281524658203, "margin_dpo/margin_std": 102.8226318359375, "step": 245 }, { "epoch": 0.36123348017621143, "fcm_dpo/beta": 0.010238472372293472, "fcm_dpo/delta": 0.03806059807538986, "fcm_dpo/margin": 79.59729766845703, "fcm_dpo/q_t": 0.3448247015476227, "grad_norm": 45.083065032958984, "learning_rate": 4.047230911780736e-07, "logits/chosen": -0.4234234392642975, "logits/rejected": -0.38395100831985474, "logps/chosen": -156.4099884033203, "logps/ref_chosen": -66.69696807861328, "logps/ref_rejected": -84.34634399414062, "logps/rejected": -253.65667724609375, "loss": 0.9226, "margin_dpo/margin_mean": 79.59729766845703, "margin_dpo/margin_std": 100.47013854980469, "step": 246 }, { "epoch": 0.36270190895741555, "fcm_dpo/beta": 0.009807577356696129, "fcm_dpo/delta": -0.2844681143760681, "fcm_dpo/margin": 112.23686218261719, "fcm_dpo/q_t": 0.29242298007011414, "grad_norm": 48.097572326660156, "learning_rate": 4.0371303213004814e-07, "logits/chosen": -0.33529412746429443, "logits/rejected": -0.3306733965873718, "logps/chosen": -160.50535583496094, "logps/ref_chosen": -56.6053466796875, "logps/ref_rejected": -106.29326629638672, "logps/rejected": -322.43017578125, "loss": 0.7889, "margin_dpo/margin_mean": 112.23686218261719, "margin_dpo/margin_std": 113.30484008789062, "step": 247 }, { "epoch": 0.3641703377386197, "fcm_dpo/beta": 0.00968283973634243, "fcm_dpo/delta": -0.006413431838154793, "fcm_dpo/margin": 88.3683090209961, "fcm_dpo/q_t": 0.319698691368103, "grad_norm": 42.922454833984375, "learning_rate": 4.0269892260067197e-07, "logits/chosen": -0.34435319900512695, "logits/rejected": -0.3552175462245941, "logps/chosen": -127.05135345458984, "logps/ref_chosen": -44.043216705322266, "logps/ref_rejected": -91.85687255859375, "logps/rejected": -263.23333740234375, "loss": 0.8132, "margin_dpo/margin_mean": 88.36830139160156, "margin_dpo/margin_std": 75.46261596679688, "step": 248 }, { "epoch": 0.3656387665198238, "fcm_dpo/beta": 0.01001214049756527, "fcm_dpo/delta": 0.22601774334907532, "fcm_dpo/margin": 63.81077194213867, "fcm_dpo/q_t": 0.37422841787338257, "grad_norm": 45.26923751831055, "learning_rate": 4.0168078931267426e-07, "logits/chosen": -0.3437703251838684, "logits/rejected": -0.31204476952552795, "logps/chosen": -162.9739227294922, "logps/ref_chosen": -62.442352294921875, "logps/ref_rejected": -80.46806335449219, "logps/rejected": -244.8104248046875, "loss": 1.0405, "margin_dpo/margin_mean": 63.81077575683594, "margin_dpo/margin_std": 95.67174530029297, "step": 249 }, { "epoch": 0.3671071953010279, "fcm_dpo/beta": 0.010107604786753654, "fcm_dpo/delta": -0.07880331575870514, "fcm_dpo/margin": 91.13977813720703, "fcm_dpo/q_t": 0.3046584725379944, "grad_norm": 43.14041519165039, "learning_rate": 4.006586590948141e-07, "logits/chosen": -0.3341864347457886, "logits/rejected": -0.2635290324687958, "logps/chosen": -150.46229553222656, "logps/ref_chosen": -65.63668823242188, "logps/ref_rejected": -73.87184143066406, "logps/rejected": -249.83721923828125, "loss": 0.7832, "margin_dpo/margin_mean": 91.13977813720703, "margin_dpo/margin_std": 72.58334350585938, "step": 250 }, { "epoch": 0.368575624082232, "fcm_dpo/beta": 0.010061986744403839, "fcm_dpo/delta": 0.08627409487962723, "fcm_dpo/margin": 76.52996826171875, "fcm_dpo/q_t": 0.3464193344116211, "grad_norm": 42.19496154785156, "learning_rate": 3.9963255888117325e-07, "logits/chosen": -0.34372633695602417, "logits/rejected": -0.307092547416687, "logps/chosen": -154.13720703125, "logps/ref_chosen": -57.182716369628906, "logps/ref_rejected": -77.66343688964844, "logps/rejected": -251.14788818359375, "loss": 0.9374, "margin_dpo/margin_mean": 76.52996826171875, "margin_dpo/margin_std": 91.52153015136719, "step": 251 }, { "epoch": 0.3700440528634361, "fcm_dpo/beta": 0.00997251458466053, "fcm_dpo/delta": -0.07944738864898682, "fcm_dpo/margin": 92.27542114257812, "fcm_dpo/q_t": 0.300639808177948, "grad_norm": 44.86310958862305, "learning_rate": 3.9860251571044666e-07, "logits/chosen": -0.3977884352207184, "logits/rejected": -0.3488880395889282, "logps/chosen": -165.05645751953125, "logps/ref_chosen": -71.68563842773438, "logps/ref_rejected": -84.75799560546875, "logps/rejected": -270.40423583984375, "loss": 0.7912, "margin_dpo/margin_mean": 92.27542114257812, "margin_dpo/margin_std": 75.70187377929688, "step": 252 }, { "epoch": 0.37151248164464024, "fcm_dpo/beta": 0.010129924863576889, "fcm_dpo/delta": 0.004893705248832703, "fcm_dpo/margin": 83.24456024169922, "fcm_dpo/q_t": 0.3302081227302551, "grad_norm": 47.6492805480957, "learning_rate": 3.9756855672522986e-07, "logits/chosen": -0.3927938640117645, "logits/rejected": -0.38195687532424927, "logps/chosen": -155.09674072265625, "logps/ref_chosen": -69.1339340209961, "logps/ref_rejected": -98.70252990722656, "logps/rejected": -267.909912109375, "loss": 0.9012, "margin_dpo/margin_mean": 83.24456024169922, "margin_dpo/margin_std": 93.68569946289062, "step": 253 }, { "epoch": 0.37298091042584436, "fcm_dpo/beta": 0.010117759928107262, "fcm_dpo/delta": 0.06923794001340866, "fcm_dpo/margin": 77.70970153808594, "fcm_dpo/q_t": 0.3566807806491852, "grad_norm": 47.6483154296875, "learning_rate": 3.965307091713037e-07, "logits/chosen": -0.3602520823478699, "logits/rejected": -0.33540549874305725, "logps/chosen": -154.7190704345703, "logps/ref_chosen": -54.154998779296875, "logps/ref_rejected": -90.30764770507812, "logps/rejected": -268.5814208984375, "loss": 0.9903, "margin_dpo/margin_mean": 77.70970153808594, "margin_dpo/margin_std": 112.6717758178711, "step": 254 }, { "epoch": 0.3744493392070485, "fcm_dpo/beta": 0.010053331032395363, "fcm_dpo/delta": -0.01586991548538208, "fcm_dpo/margin": 85.83883666992188, "fcm_dpo/q_t": 0.3369311988353729, "grad_norm": 55.98381423950195, "learning_rate": 3.954890003969163e-07, "logits/chosen": -0.2981446087360382, "logits/rejected": -0.28794023394584656, "logps/chosen": -166.04107666015625, "logps/ref_chosen": -57.14167022705078, "logps/ref_rejected": -90.2085952758789, "logps/rejected": -284.94683837890625, "loss": 0.9489, "margin_dpo/margin_mean": 85.8388442993164, "margin_dpo/margin_std": 111.47303771972656, "step": 255 }, { "epoch": 0.37591776798825255, "fcm_dpo/beta": 0.009983371943235397, "fcm_dpo/delta": -0.07811053097248077, "fcm_dpo/margin": 92.20353698730469, "fcm_dpo/q_t": 0.3229755163192749, "grad_norm": 58.64165115356445, "learning_rate": 3.944434578520628e-07, "logits/chosen": -0.29079824686050415, "logits/rejected": -0.2855415940284729, "logps/chosen": -154.9361572265625, "logps/ref_chosen": -55.163490295410156, "logps/ref_rejected": -92.56291961669922, "logps/rejected": -284.53912353515625, "loss": 0.8674, "margin_dpo/margin_mean": 92.20353698730469, "margin_dpo/margin_std": 104.3232192993164, "step": 256 }, { "epoch": 0.37738619676945667, "fcm_dpo/beta": 0.009662920609116554, "fcm_dpo/delta": -0.09995823353528976, "fcm_dpo/margin": 96.6123275756836, "fcm_dpo/q_t": 0.3234117329120636, "grad_norm": 42.73604202270508, "learning_rate": 3.933941090877615e-07, "logits/chosen": -0.29401636123657227, "logits/rejected": -0.2769302725791931, "logps/chosen": -144.01602172851562, "logps/ref_chosen": -49.42369842529297, "logps/ref_rejected": -79.53791809082031, "logps/rejected": -270.7425537109375, "loss": 0.8724, "margin_dpo/margin_mean": 96.6123275756836, "margin_dpo/margin_std": 106.85414123535156, "step": 257 }, { "epoch": 0.3788546255506608, "fcm_dpo/beta": 0.00979782734066248, "fcm_dpo/delta": -0.00526980496942997, "fcm_dpo/margin": 87.2305908203125, "fcm_dpo/q_t": 0.32909661531448364, "grad_norm": 49.11513900756836, "learning_rate": 3.923409817553284e-07, "logits/chosen": -0.2832171320915222, "logits/rejected": -0.2739581763744354, "logps/chosen": -169.74600219726562, "logps/ref_chosen": -59.384124755859375, "logps/ref_rejected": -95.99010467529297, "logps/rejected": -293.58258056640625, "loss": 0.9349, "margin_dpo/margin_mean": 87.23057556152344, "margin_dpo/margin_std": 109.7698974609375, "step": 258 }, { "epoch": 0.3803230543318649, "fcm_dpo/beta": 0.009825894609093666, "fcm_dpo/delta": 0.08960148692131042, "fcm_dpo/margin": 78.04417419433594, "fcm_dpo/q_t": 0.34981006383895874, "grad_norm": 45.765663146972656, "learning_rate": 3.9128410360564793e-07, "logits/chosen": -0.3537431061267853, "logits/rejected": -0.3431965112686157, "logps/chosen": -151.65065002441406, "logps/ref_chosen": -52.828346252441406, "logps/ref_rejected": -89.191650390625, "logps/rejected": -266.05816650390625, "loss": 0.9498, "margin_dpo/margin_mean": 78.04417419433594, "margin_dpo/margin_std": 97.46719360351562, "step": 259 }, { "epoch": 0.38179148311306904, "fcm_dpo/beta": 0.009930366650223732, "fcm_dpo/delta": -0.03357648104429245, "fcm_dpo/margin": 88.66500091552734, "fcm_dpo/q_t": 0.3270108103752136, "grad_norm": 43.459815979003906, "learning_rate": 3.9022350248844246e-07, "logits/chosen": -0.3347843885421753, "logits/rejected": -0.35157203674316406, "logps/chosen": -141.62643432617188, "logps/ref_chosen": -47.41767501831055, "logps/ref_rejected": -95.08978271484375, "logps/rejected": -277.96356201171875, "loss": 0.8549, "margin_dpo/margin_mean": 88.66499328613281, "margin_dpo/margin_std": 95.96063995361328, "step": 260 }, { "epoch": 0.3832599118942731, "fcm_dpo/beta": 0.009731737896800041, "fcm_dpo/delta": -0.07620503008365631, "fcm_dpo/margin": 94.40481567382812, "fcm_dpo/q_t": 0.3174518346786499, "grad_norm": 46.79330825805664, "learning_rate": 3.891592063515376e-07, "logits/chosen": -0.31741079688072205, "logits/rejected": -0.30679404735565186, "logps/chosen": -132.945068359375, "logps/ref_chosen": -53.03137969970703, "logps/ref_rejected": -88.51494598388672, "logps/rejected": -262.8334655761719, "loss": 0.8336, "margin_dpo/margin_mean": 94.40481567382812, "margin_dpo/margin_std": 97.66380310058594, "step": 261 }, { "epoch": 0.38472834067547723, "fcm_dpo/beta": 0.009914107620716095, "fcm_dpo/delta": 0.10215713828802109, "fcm_dpo/margin": 76.14309692382812, "fcm_dpo/q_t": 0.33958059549331665, "grad_norm": 38.953277587890625, "learning_rate": 3.880912432401264e-07, "logits/chosen": -0.28771814703941345, "logits/rejected": -0.2547649145126343, "logps/chosen": -152.20550537109375, "logps/ref_chosen": -59.620140075683594, "logps/ref_rejected": -86.41853332519531, "logps/rejected": -255.14697265625, "loss": 0.8977, "margin_dpo/margin_mean": 76.14309692382812, "margin_dpo/margin_std": 78.52725219726562, "step": 262 }, { "epoch": 0.38619676945668135, "fcm_dpo/beta": 0.009619778022170067, "fcm_dpo/delta": -0.17337682843208313, "fcm_dpo/margin": 104.40455627441406, "fcm_dpo/q_t": 0.2986921966075897, "grad_norm": 40.7275390625, "learning_rate": 3.870196412960302e-07, "logits/chosen": -0.38451990485191345, "logits/rejected": -0.35206347703933716, "logps/chosen": -142.514892578125, "logps/ref_chosen": -59.42094421386719, "logps/ref_rejected": -96.85720825195312, "logps/rejected": -284.355712890625, "loss": 0.7864, "margin_dpo/margin_mean": 104.40455627441406, "margin_dpo/margin_std": 96.11820983886719, "step": 263 }, { "epoch": 0.3876651982378855, "fcm_dpo/beta": 0.009586556814610958, "fcm_dpo/delta": 0.04767933860421181, "fcm_dpo/margin": 84.01654052734375, "fcm_dpo/q_t": 0.33650076389312744, "grad_norm": 62.09348678588867, "learning_rate": 3.8594442875695665e-07, "logits/chosen": -0.3532578945159912, "logits/rejected": -0.337126761674881, "logps/chosen": -148.93775939941406, "logps/ref_chosen": -62.722084045410156, "logps/ref_rejected": -93.85620880126953, "logps/rejected": -264.0884094238281, "loss": 0.8848, "margin_dpo/margin_mean": 84.01654052734375, "margin_dpo/margin_std": 89.25030517578125, "step": 264 }, { "epoch": 0.3891336270190896, "fcm_dpo/beta": 0.009736664593219757, "fcm_dpo/delta": 0.019567368552088737, "fcm_dpo/margin": 85.43596649169922, "fcm_dpo/q_t": 0.34142905473709106, "grad_norm": 50.95734786987305, "learning_rate": 3.848656339557562e-07, "logits/chosen": -0.27792888879776, "logits/rejected": -0.2561708092689514, "logps/chosen": -161.2020721435547, "logps/ref_chosen": -61.971466064453125, "logps/ref_rejected": -88.02059936523438, "logps/rejected": -272.6871643066406, "loss": 0.9227, "margin_dpo/margin_mean": 85.43595886230469, "margin_dpo/margin_std": 107.73658752441406, "step": 265 }, { "epoch": 0.39060205580029367, "fcm_dpo/beta": 0.009648610837757587, "fcm_dpo/delta": -0.01776065304875374, "fcm_dpo/margin": 89.72529602050781, "fcm_dpo/q_t": 0.3276539444923401, "grad_norm": 39.50182342529297, "learning_rate": 3.8378328531967507e-07, "logits/chosen": -0.33635252714157104, "logits/rejected": -0.27758198976516724, "logps/chosen": -164.45074462890625, "logps/ref_chosen": -67.09967041015625, "logps/ref_rejected": -67.97122192382812, "logps/rejected": -255.047607421875, "loss": 0.8672, "margin_dpo/margin_mean": 89.72528839111328, "margin_dpo/margin_std": 96.26741027832031, "step": 266 }, { "epoch": 0.3920704845814978, "fcm_dpo/beta": 0.009558433666825294, "fcm_dpo/delta": -0.05971603840589523, "fcm_dpo/margin": 94.54733276367188, "fcm_dpo/q_t": 0.31543660163879395, "grad_norm": 42.246028900146484, "learning_rate": 3.8269741136960646e-07, "logits/chosen": -0.28289055824279785, "logits/rejected": -0.2381281852722168, "logps/chosen": -171.30865478515625, "logps/ref_chosen": -68.97075653076172, "logps/ref_rejected": -90.16844940185547, "logps/rejected": -287.0536804199219, "loss": 0.8204, "margin_dpo/margin_mean": 94.54733276367188, "margin_dpo/margin_std": 91.51085662841797, "step": 267 }, { "epoch": 0.3935389133627019, "fcm_dpo/beta": 0.00957027729600668, "fcm_dpo/delta": 0.02075892500579357, "fcm_dpo/margin": 86.81520080566406, "fcm_dpo/q_t": 0.33144694566726685, "grad_norm": 69.94318389892578, "learning_rate": 3.8160804071933894e-07, "logits/chosen": -0.23065903782844543, "logits/rejected": -0.23640581965446472, "logps/chosen": -172.29135131835938, "logps/ref_chosen": -55.90031051635742, "logps/ref_rejected": -101.64763641357422, "logps/rejected": -304.8538818359375, "loss": 0.88, "margin_dpo/margin_mean": 86.81520080566406, "margin_dpo/margin_std": 94.50149536132812, "step": 268 }, { "epoch": 0.39500734214390604, "fcm_dpo/beta": 0.009249732829630375, "fcm_dpo/delta": -0.2376570701599121, "fcm_dpo/margin": 114.76429748535156, "fcm_dpo/q_t": 0.29645225405693054, "grad_norm": 83.19200134277344, "learning_rate": 3.8051520207480204e-07, "logits/chosen": -0.2872861623764038, "logits/rejected": -0.2640247941017151, "logps/chosen": -198.46896362304688, "logps/ref_chosen": -70.03955841064453, "logps/ref_rejected": -107.34937286376953, "logps/rejected": -350.5430908203125, "loss": 0.7972, "margin_dpo/margin_mean": 114.7643051147461, "margin_dpo/margin_std": 115.25680541992188, "step": 269 }, { "epoch": 0.3964757709251101, "fcm_dpo/beta": 0.009293366223573685, "fcm_dpo/delta": 0.07112763822078705, "fcm_dpo/margin": 84.35902404785156, "fcm_dpo/q_t": 0.34950098395347595, "grad_norm": 53.65988540649414, "learning_rate": 3.794189242333106e-07, "logits/chosen": -0.28146451711654663, "logits/rejected": -0.2732328176498413, "logps/chosen": -189.49839782714844, "logps/ref_chosen": -69.53347778320312, "logps/ref_rejected": -109.92864990234375, "logps/rejected": -314.25262451171875, "loss": 0.948, "margin_dpo/margin_mean": 84.35902404785156, "margin_dpo/margin_std": 109.4448471069336, "step": 270 }, { "epoch": 0.39794419970631423, "fcm_dpo/beta": 0.009072428569197655, "fcm_dpo/delta": -0.13254715502262115, "fcm_dpo/margin": 106.77680969238281, "fcm_dpo/q_t": 0.31320929527282715, "grad_norm": 39.882137298583984, "learning_rate": 3.7831923608280514e-07, "logits/chosen": -0.21552950143814087, "logits/rejected": -0.18930025398731232, "logps/chosen": -177.8402099609375, "logps/ref_chosen": -56.76456832885742, "logps/ref_rejected": -92.51383972167969, "logps/rejected": -320.36627197265625, "loss": 0.8185, "margin_dpo/margin_mean": 106.77680969238281, "margin_dpo/margin_std": 110.38920593261719, "step": 271 }, { "epoch": 0.39941262848751835, "fcm_dpo/beta": 0.008728357963263988, "fcm_dpo/delta": -0.29591071605682373, "fcm_dpo/margin": 127.61502075195312, "fcm_dpo/q_t": 0.28334927558898926, "grad_norm": 33.21427917480469, "learning_rate": 3.772161666010912e-07, "logits/chosen": -0.18609288334846497, "logits/rejected": -0.19840866327285767, "logps/chosen": -163.57485961914062, "logps/ref_chosen": -49.497154235839844, "logps/ref_rejected": -105.54279327392578, "logps/rejected": -347.23553466796875, "loss": 0.755, "margin_dpo/margin_mean": 127.61502075195312, "margin_dpo/margin_std": 117.34019470214844, "step": 272 }, { "epoch": 0.4008810572687225, "fcm_dpo/beta": 0.008227063342928886, "fcm_dpo/delta": -0.1768997311592102, "fcm_dpo/margin": 122.21438598632812, "fcm_dpo/q_t": 0.30270469188690186, "grad_norm": 33.36330032348633, "learning_rate": 3.761097448550755e-07, "logits/chosen": -0.22537121176719666, "logits/rejected": -0.2015794813632965, "logps/chosen": -183.01629638671875, "logps/ref_chosen": -62.97539520263672, "logps/ref_rejected": -92.49858093261719, "logps/rejected": -334.75384521484375, "loss": 0.7921, "margin_dpo/margin_mean": 122.21438598632812, "margin_dpo/margin_std": 115.96615600585938, "step": 273 }, { "epoch": 0.4023494860499266, "fcm_dpo/beta": 0.008325198665261269, "fcm_dpo/delta": 0.05208263173699379, "fcm_dpo/margin": 96.30369567871094, "fcm_dpo/q_t": 0.3377155065536499, "grad_norm": 50.85749435424805, "learning_rate": 3.75e-07, "logits/chosen": -0.19594232738018036, "logits/rejected": -0.17438998818397522, "logps/chosen": -197.13754272460938, "logps/ref_chosen": -55.66770935058594, "logps/ref_rejected": -77.33308410644531, "logps/rejected": -315.10662841796875, "loss": 0.9015, "margin_dpo/margin_mean": 96.30369567871094, "margin_dpo/margin_std": 107.05192565917969, "step": 274 }, { "epoch": 0.40381791483113066, "fcm_dpo/beta": 0.008365076035261154, "fcm_dpo/delta": -0.002092186361551285, "fcm_dpo/margin": 101.77294158935547, "fcm_dpo/q_t": 0.3336763381958008, "grad_norm": 45.55134201049805, "learning_rate": 3.738869612786737e-07, "logits/chosen": -0.22732514142990112, "logits/rejected": -0.2275867611169815, "logps/chosen": -155.5432891845703, "logps/ref_chosen": -48.594703674316406, "logps/ref_rejected": -93.30369567871094, "logps/rejected": -302.02520751953125, "loss": 0.9015, "margin_dpo/margin_mean": 101.77294158935547, "margin_dpo/margin_std": 118.06379699707031, "step": 275 }, { "epoch": 0.4052863436123348, "fcm_dpo/beta": 0.008263555355370045, "fcm_dpo/delta": -0.0206641536206007, "fcm_dpo/margin": 105.10835266113281, "fcm_dpo/q_t": 0.33232614398002625, "grad_norm": 45.553375244140625, "learning_rate": 3.7277065802070204e-07, "logits/chosen": -0.2473652958869934, "logits/rejected": -0.21824447810649872, "logps/chosen": -171.64883422851562, "logps/ref_chosen": -56.57740783691406, "logps/ref_rejected": -70.36566925048828, "logps/rejected": -290.5454406738281, "loss": 0.8898, "margin_dpo/margin_mean": 105.10834503173828, "margin_dpo/margin_std": 122.83244323730469, "step": 276 }, { "epoch": 0.4067547723935389, "fcm_dpo/beta": 0.0082924272865057, "fcm_dpo/delta": -0.02089064195752144, "fcm_dpo/margin": 104.77349090576172, "fcm_dpo/q_t": 0.32611536979675293, "grad_norm": 36.01552963256836, "learning_rate": 3.71651119641714e-07, "logits/chosen": -0.20939847826957703, "logits/rejected": -0.19383195042610168, "logps/chosen": -180.56362915039062, "logps/ref_chosen": -56.27156066894531, "logps/ref_rejected": -92.88127136230469, "logps/rejected": -321.94683837890625, "loss": 0.8521, "margin_dpo/margin_mean": 104.77348327636719, "margin_dpo/margin_std": 107.48456573486328, "step": 277 }, { "epoch": 0.40822320117474303, "fcm_dpo/beta": 0.008100062608718872, "fcm_dpo/delta": -0.09823326766490936, "fcm_dpo/margin": 115.85311126708984, "fcm_dpo/q_t": 0.3185454308986664, "grad_norm": 34.42502212524414, "learning_rate": 3.705283756425872e-07, "logits/chosen": -0.20616397261619568, "logits/rejected": -0.20798000693321228, "logps/chosen": -170.531982421875, "logps/ref_chosen": -52.94194030761719, "logps/ref_rejected": -91.25357818603516, "logps/rejected": -324.69671630859375, "loss": 0.8401, "margin_dpo/margin_mean": 115.85311126708984, "margin_dpo/margin_std": 122.75349426269531, "step": 278 }, { "epoch": 0.40969162995594716, "fcm_dpo/beta": 0.007804838474839926, "fcm_dpo/delta": -0.16851356625556946, "fcm_dpo/margin": 127.84455871582031, "fcm_dpo/q_t": 0.31053388118743896, "grad_norm": 41.10043716430664, "learning_rate": 3.6940245560867e-07, "logits/chosen": -0.04199061542749405, "logits/rejected": -0.032916110008955, "logps/chosen": -192.05714416503906, "logps/ref_chosen": -48.641319274902344, "logps/ref_rejected": -87.8514404296875, "logps/rejected": -359.11181640625, "loss": 0.8453, "margin_dpo/margin_mean": 127.84455871582031, "margin_dpo/margin_std": 138.51828002929688, "step": 279 }, { "epoch": 0.4111600587371512, "fcm_dpo/beta": 0.0076360213570296764, "fcm_dpo/delta": -0.13522008061408997, "fcm_dpo/margin": 127.201171875, "fcm_dpo/q_t": 0.2989714741706848, "grad_norm": 48.5097541809082, "learning_rate": 3.6827338920900253e-07, "logits/chosen": -0.10075664520263672, "logits/rejected": -0.09791351854801178, "logps/chosen": -199.2039031982422, "logps/ref_chosen": -58.797122955322266, "logps/ref_rejected": -98.61885070800781, "logps/rejected": -366.226806640625, "loss": 0.7771, "margin_dpo/margin_mean": 127.201171875, "margin_dpo/margin_std": 109.58746337890625, "step": 280 }, { "epoch": 0.41262848751835535, "fcm_dpo/beta": 0.007574355695396662, "fcm_dpo/delta": -0.06523635238409042, "fcm_dpo/margin": 120.03223419189453, "fcm_dpo/q_t": 0.3079676628112793, "grad_norm": 41.980621337890625, "learning_rate": 3.6714120619553435e-07, "logits/chosen": -0.1654478907585144, "logits/rejected": -0.13742399215698242, "logps/chosen": -192.7757568359375, "logps/ref_chosen": -55.488521575927734, "logps/ref_rejected": -80.88258361816406, "logps/rejected": -338.2020568847656, "loss": 0.8205, "margin_dpo/margin_mean": 120.03223419189453, "margin_dpo/margin_std": 112.45529174804688, "step": 281 }, { "epoch": 0.41409691629955947, "fcm_dpo/beta": 0.007696120534092188, "fcm_dpo/delta": 0.11213569343090057, "fcm_dpo/margin": 96.7197036743164, "fcm_dpo/q_t": 0.3515746593475342, "grad_norm": 70.64493560791016, "learning_rate": 3.660059364023408e-07, "logits/chosen": -0.18230430781841278, "logits/rejected": -0.14879056811332703, "logps/chosen": -227.57904052734375, "logps/ref_chosen": -73.07014465332031, "logps/ref_rejected": -95.35098266601562, "logps/rejected": -346.57958984375, "loss": 0.9293, "margin_dpo/margin_mean": 96.71971130371094, "margin_dpo/margin_std": 116.51522827148438, "step": 282 }, { "epoch": 0.4155653450807636, "fcm_dpo/beta": 0.007496010046452284, "fcm_dpo/delta": -0.20360887050628662, "fcm_dpo/margin": 137.80856323242188, "fcm_dpo/q_t": 0.294839084148407, "grad_norm": 55.21201705932617, "learning_rate": 3.6486760974483685e-07, "logits/chosen": -0.0822724774479866, "logits/rejected": -0.07609227299690247, "logps/chosen": -223.06536865234375, "logps/ref_chosen": -61.89844512939453, "logps/ref_rejected": -96.98655700683594, "logps/rejected": -395.9620361328125, "loss": 0.784, "margin_dpo/margin_mean": 137.80856323242188, "margin_dpo/margin_std": 129.84677124023438, "step": 283 }, { "epoch": 0.4170337738619677, "fcm_dpo/beta": 0.007216787431389093, "fcm_dpo/delta": -0.14114126563072205, "fcm_dpo/margin": 135.40316772460938, "fcm_dpo/q_t": 0.3051418662071228, "grad_norm": 40.07844924926758, "learning_rate": 3.6372625621898863e-07, "logits/chosen": -0.1315004974603653, "logits/rejected": -0.11729149520397186, "logps/chosen": -218.79432678222656, "logps/ref_chosen": -58.4355354309082, "logps/ref_rejected": -93.46926879882812, "logps/rejected": -389.2312316894531, "loss": 0.784, "margin_dpo/margin_mean": 135.40316772460938, "margin_dpo/margin_std": 128.7941131591797, "step": 284 }, { "epoch": 0.4185022026431718, "fcm_dpo/beta": 0.007242846768349409, "fcm_dpo/delta": 0.008408665657043457, "fcm_dpo/margin": 116.0887680053711, "fcm_dpo/q_t": 0.323802649974823, "grad_norm": 40.012691497802734, "learning_rate": 3.625819059005228e-07, "logits/chosen": -0.14234209060668945, "logits/rejected": -0.12387385219335556, "logps/chosen": -239.96295166015625, "logps/ref_chosen": -66.23219299316406, "logps/ref_rejected": -99.1268310546875, "logps/rejected": -388.94635009765625, "loss": 0.8585, "margin_dpo/margin_mean": 116.08877563476562, "margin_dpo/margin_std": 113.36100769042969, "step": 285 }, { "epoch": 0.4199706314243759, "fcm_dpo/beta": 0.007111635059118271, "fcm_dpo/delta": -0.06945645064115524, "fcm_dpo/margin": 128.3986053466797, "fcm_dpo/q_t": 0.3180208206176758, "grad_norm": 50.79453659057617, "learning_rate": 3.614345889441346e-07, "logits/chosen": -0.20035997033119202, "logits/rejected": -0.1742473840713501, "logps/chosen": -238.18199157714844, "logps/ref_chosen": -72.95100402832031, "logps/ref_rejected": -88.58845520019531, "logps/rejected": -382.218017578125, "loss": 0.8479, "margin_dpo/margin_mean": 128.3986053466797, "margin_dpo/margin_std": 134.0906982421875, "step": 286 }, { "epoch": 0.42143906020558003, "fcm_dpo/beta": 0.007145530544221401, "fcm_dpo/delta": 0.0370517298579216, "fcm_dpo/margin": 114.15289306640625, "fcm_dpo/q_t": 0.3385983407497406, "grad_norm": 34.708343505859375, "learning_rate": 3.6028433558269275e-07, "logits/chosen": -0.11621429771184921, "logits/rejected": -0.09012611210346222, "logps/chosen": -222.60751342773438, "logps/ref_chosen": -61.54115295410156, "logps/ref_rejected": -77.69607543945312, "logps/rejected": -352.91534423828125, "loss": 0.9046, "margin_dpo/margin_mean": 114.15290069580078, "margin_dpo/margin_std": 134.21795654296875, "step": 287 }, { "epoch": 0.42290748898678415, "fcm_dpo/beta": 0.00695967860519886, "fcm_dpo/delta": -0.07556534558534622, "fcm_dpo/margin": 131.31358337402344, "fcm_dpo/q_t": 0.3097105920314789, "grad_norm": 35.59653091430664, "learning_rate": 3.5913117612644327e-07, "logits/chosen": -0.08861184120178223, "logits/rejected": -0.06481438130140305, "logps/chosen": -225.13528442382812, "logps/ref_chosen": -56.661224365234375, "logps/ref_rejected": -87.33570098876953, "logps/rejected": -387.1233215332031, "loss": 0.8115, "margin_dpo/margin_mean": 131.31358337402344, "margin_dpo/margin_std": 114.73504638671875, "step": 288 }, { "epoch": 0.4243759177679883, "fcm_dpo/beta": 0.0068605877459049225, "fcm_dpo/delta": -0.19726723432540894, "fcm_dpo/margin": 149.75587463378906, "fcm_dpo/q_t": 0.3031596541404724, "grad_norm": 43.76935958862305, "learning_rate": 3.5797514096221024e-07, "logits/chosen": -0.011087626218795776, "logits/rejected": -0.0034407819621264935, "logps/chosen": -216.83993530273438, "logps/ref_chosen": -45.23039245605469, "logps/ref_rejected": -87.64266967773438, "logps/rejected": -409.008056640625, "loss": 0.7988, "margin_dpo/margin_mean": 149.75587463378906, "margin_dpo/margin_std": 149.0826873779297, "step": 289 }, { "epoch": 0.42584434654919234, "fcm_dpo/beta": 0.006617412902414799, "fcm_dpo/delta": -0.13556578755378723, "fcm_dpo/margin": 146.91989135742188, "fcm_dpo/q_t": 0.31431177258491516, "grad_norm": 37.676513671875, "learning_rate": 3.568162605525952e-07, "logits/chosen": -0.017872124910354614, "logits/rejected": -0.041775476187467575, "logps/chosen": -242.56007385253906, "logps/ref_chosen": -55.47149658203125, "logps/ref_rejected": -116.70857238769531, "logps/rejected": -450.717041015625, "loss": 0.8339, "margin_dpo/margin_mean": 146.91989135742188, "margin_dpo/margin_std": 160.46890258789062, "step": 290 }, { "epoch": 0.42731277533039647, "fcm_dpo/beta": 0.006590387783944607, "fcm_dpo/delta": -0.04984688758850098, "fcm_dpo/margin": 135.67874145507812, "fcm_dpo/q_t": 0.3153119683265686, "grad_norm": 41.912757873535156, "learning_rate": 3.5565456543517485e-07, "logits/chosen": -0.08200214803218842, "logits/rejected": -0.06535626202821732, "logps/chosen": -219.4686279296875, "logps/ref_chosen": -63.26036834716797, "logps/ref_rejected": -89.29708862304688, "logps/rejected": -381.1841125488281, "loss": 0.8297, "margin_dpo/margin_mean": 135.67874145507812, "margin_dpo/margin_std": 126.33541870117188, "step": 291 }, { "epoch": 0.4287812041116006, "fcm_dpo/beta": 0.006525726057589054, "fcm_dpo/delta": -0.022788774222135544, "fcm_dpo/margin": 133.373291015625, "fcm_dpo/q_t": 0.3237302601337433, "grad_norm": 60.781497955322266, "learning_rate": 3.5449008622169583e-07, "logits/chosen": -0.07108621299266815, "logits/rejected": -0.05422316491603851, "logps/chosen": -232.05926513671875, "logps/ref_chosen": -53.91852951049805, "logps/ref_rejected": -89.96138000488281, "logps/rejected": -401.47540283203125, "loss": 0.9028, "margin_dpo/margin_mean": 133.37327575683594, "margin_dpo/margin_std": 155.2228240966797, "step": 292 }, { "epoch": 0.4302496328928047, "fcm_dpo/beta": 0.00664651021361351, "fcm_dpo/delta": 0.1351955235004425, "fcm_dpo/margin": 108.85354614257812, "fcm_dpo/q_t": 0.3587031066417694, "grad_norm": 39.997802734375, "learning_rate": 3.5332285359726846e-07, "logits/chosen": -0.08526704460382462, "logits/rejected": -0.07327552884817123, "logps/chosen": -225.83401489257812, "logps/ref_chosen": -60.376033782958984, "logps/ref_rejected": -77.85244750976562, "logps/rejected": -352.1639709472656, "loss": 0.9827, "margin_dpo/margin_mean": 108.85354614257812, "margin_dpo/margin_std": 145.9722442626953, "step": 293 }, { "epoch": 0.43171806167400884, "fcm_dpo/beta": 0.006676151417195797, "fcm_dpo/delta": 0.04422283172607422, "fcm_dpo/margin": 121.20598602294922, "fcm_dpo/q_t": 0.3434596061706543, "grad_norm": 39.22120666503906, "learning_rate": 3.5215289831955786e-07, "logits/chosen": -0.07479946315288544, "logits/rejected": -0.07693535089492798, "logps/chosen": -197.52316284179688, "logps/ref_chosen": -48.0875358581543, "logps/ref_rejected": -81.89698791503906, "logps/rejected": -352.53857421875, "loss": 0.9206, "margin_dpo/margin_mean": 121.20599365234375, "margin_dpo/margin_std": 148.40878295898438, "step": 294 }, { "epoch": 0.4331864904552129, "fcm_dpo/beta": 0.0067435759119689465, "fcm_dpo/delta": 0.031169120222330093, "fcm_dpo/margin": 121.80484008789062, "fcm_dpo/q_t": 0.3364499807357788, "grad_norm": 40.891448974609375, "learning_rate": 3.509802512179737e-07, "logits/chosen": -0.11838773638010025, "logits/rejected": -0.11896628886461258, "logps/chosen": -223.58511352539062, "logps/ref_chosen": -49.92467498779297, "logps/ref_rejected": -87.45632934570312, "logps/rejected": -382.9216003417969, "loss": 0.8978, "margin_dpo/margin_mean": 121.80484008789062, "margin_dpo/margin_std": 137.68096923828125, "step": 295 }, { "epoch": 0.434654919236417, "fcm_dpo/beta": 0.006936186458915472, "fcm_dpo/delta": 0.2200748324394226, "fcm_dpo/margin": 92.93003845214844, "fcm_dpo/q_t": 0.3718491196632385, "grad_norm": 62.75762939453125, "learning_rate": 3.498049431928577e-07, "logits/chosen": -0.12827003002166748, "logits/rejected": -0.10901685059070587, "logps/chosen": -262.9681396484375, "logps/ref_chosen": -65.49124145507812, "logps/ref_rejected": -93.08908081054688, "logps/rejected": -383.49603271484375, "loss": 1.0533, "margin_dpo/margin_mean": 92.93003845214844, "margin_dpo/margin_std": 143.3027801513672, "step": 296 }, { "epoch": 0.43612334801762115, "fcm_dpo/beta": 0.007118170149624348, "fcm_dpo/delta": 0.05691620334982872, "fcm_dpo/margin": 112.05933380126953, "fcm_dpo/q_t": 0.33294984698295593, "grad_norm": 44.155250549316406, "learning_rate": 3.486270052146694e-07, "logits/chosen": -0.14343738555908203, "logits/rejected": -0.14404615759849548, "logps/chosen": -233.68067932128906, "logps/ref_chosen": -56.476951599121094, "logps/ref_rejected": -95.1385498046875, "logps/rejected": -384.401611328125, "loss": 0.8644, "margin_dpo/margin_mean": 112.05933380126953, "margin_dpo/margin_std": 108.87353515625, "step": 297 }, { "epoch": 0.43759177679882527, "fcm_dpo/beta": 0.007042343728244305, "fcm_dpo/delta": -0.10975862294435501, "fcm_dpo/margin": 134.819091796875, "fcm_dpo/q_t": 0.32128387689590454, "grad_norm": 55.54197692871094, "learning_rate": 3.474464683231698e-07, "logits/chosen": -0.12262261658906937, "logits/rejected": -0.14369015395641327, "logps/chosen": -240.33999633789062, "logps/ref_chosen": -67.32516479492188, "logps/ref_rejected": -116.66217041015625, "logps/rejected": -424.49609375, "loss": 0.8518, "margin_dpo/margin_mean": 134.819091796875, "margin_dpo/margin_std": 153.47982788085938, "step": 298 }, { "epoch": 0.4390602055800294, "fcm_dpo/beta": 0.007004514802247286, "fcm_dpo/delta": -0.062288179993629456, "fcm_dpo/margin": 129.19944763183594, "fcm_dpo/q_t": 0.32569801807403564, "grad_norm": 44.929222106933594, "learning_rate": 3.462633636266041e-07, "logits/chosen": -0.1638999879360199, "logits/rejected": -0.15911725163459778, "logps/chosen": -198.703369140625, "logps/ref_chosen": -48.96209716796875, "logps/ref_rejected": -84.32823944091797, "logps/rejected": -363.26898193359375, "loss": 0.8819, "margin_dpo/margin_mean": 129.19944763183594, "margin_dpo/margin_std": 146.23391723632812, "step": 299 }, { "epoch": 0.44052863436123346, "fcm_dpo/beta": 0.006798220798373222, "fcm_dpo/delta": -0.12738555669784546, "fcm_dpo/margin": 141.98765563964844, "fcm_dpo/q_t": 0.31477490067481995, "grad_norm": 67.45763397216797, "learning_rate": 3.4507772230088147e-07, "logits/chosen": -0.1142905056476593, "logits/rejected": -0.10768699645996094, "logps/chosen": -250.97042846679688, "logps/ref_chosen": -59.07371139526367, "logps/ref_rejected": -95.9664535522461, "logps/rejected": -429.85076904296875, "loss": 0.8901, "margin_dpo/margin_mean": 141.98765563964844, "margin_dpo/margin_std": 166.83888244628906, "step": 300 }, { "epoch": 0.4419970631424376, "fcm_dpo/beta": 0.006596191320568323, "fcm_dpo/delta": -0.09547975659370422, "fcm_dpo/margin": 141.79217529296875, "fcm_dpo/q_t": 0.31292590498924255, "grad_norm": 43.63452911376953, "learning_rate": 3.4388957558875316e-07, "logits/chosen": -0.1395409107208252, "logits/rejected": -0.1367347240447998, "logps/chosen": -224.599853515625, "logps/ref_chosen": -57.249366760253906, "logps/ref_rejected": -92.35354614257812, "logps/rejected": -401.4962158203125, "loss": 0.8159, "margin_dpo/margin_mean": 141.79217529296875, "margin_dpo/margin_std": 137.7713165283203, "step": 301 }, { "epoch": 0.4434654919236417, "fcm_dpo/beta": 0.00664362870156765, "fcm_dpo/delta": 0.07332435995340347, "fcm_dpo/margin": 117.74930572509766, "fcm_dpo/q_t": 0.34151118993759155, "grad_norm": 47.059879302978516, "learning_rate": 3.426989547989902e-07, "logits/chosen": -0.14914287626743317, "logits/rejected": -0.15093174576759338, "logps/chosen": -200.8439483642578, "logps/ref_chosen": -51.197994232177734, "logps/ref_rejected": -97.22636413574219, "logps/rejected": -364.62164306640625, "loss": 0.904, "margin_dpo/margin_mean": 117.74931335449219, "margin_dpo/margin_std": 131.7711639404297, "step": 302 }, { "epoch": 0.44493392070484583, "fcm_dpo/beta": 0.006741734221577644, "fcm_dpo/delta": 0.05900689214468002, "fcm_dpo/margin": 118.02484893798828, "fcm_dpo/q_t": 0.3473062813282013, "grad_norm": 72.38304901123047, "learning_rate": 3.4150589130555773e-07, "logits/chosen": -0.1549404263496399, "logits/rejected": -0.13192573189735413, "logps/chosen": -220.18551635742188, "logps/ref_chosen": -66.71394348144531, "logps/ref_rejected": -86.94542694091797, "logps/rejected": -358.44183349609375, "loss": 0.9342, "margin_dpo/margin_mean": 118.02484893798828, "margin_dpo/margin_std": 147.0444793701172, "step": 303 }, { "epoch": 0.44640234948604995, "fcm_dpo/beta": 0.006808813661336899, "fcm_dpo/delta": -0.0030750595033168793, "fcm_dpo/margin": 125.19535827636719, "fcm_dpo/q_t": 0.3163961172103882, "grad_norm": 48.08484649658203, "learning_rate": 3.403104165467883e-07, "logits/chosen": -0.16453275084495544, "logits/rejected": -0.14551252126693726, "logps/chosen": -214.55648803710938, "logps/ref_chosen": -71.95069885253906, "logps/ref_rejected": -90.47203063964844, "logps/rejected": -358.273193359375, "loss": 0.811, "margin_dpo/margin_mean": 125.19536590576172, "margin_dpo/margin_std": 101.456298828125, "step": 304 }, { "epoch": 0.447870778267254, "fcm_dpo/beta": 0.006762784440070391, "fcm_dpo/delta": 0.04657618701457977, "fcm_dpo/margin": 119.18153381347656, "fcm_dpo/q_t": 0.3422441780567169, "grad_norm": 46.43743896484375, "learning_rate": 3.391125620245535e-07, "logits/chosen": -0.18260565400123596, "logits/rejected": -0.15914301574230194, "logps/chosen": -220.98190307617188, "logps/ref_chosen": -66.79523468017578, "logps/ref_rejected": -92.75459289550781, "logps/rejected": -366.122802734375, "loss": 0.9052, "margin_dpo/margin_mean": 119.18153381347656, "margin_dpo/margin_std": 139.260986328125, "step": 305 }, { "epoch": 0.44933920704845814, "fcm_dpo/beta": 0.006842019967734814, "fcm_dpo/delta": 0.010429871268570423, "fcm_dpo/margin": 122.83320617675781, "fcm_dpo/q_t": 0.32789498567581177, "grad_norm": 54.641883850097656, "learning_rate": 3.3791235930343417e-07, "logits/chosen": -0.14581012725830078, "logits/rejected": -0.11724609136581421, "logps/chosen": -221.95294189453125, "logps/ref_chosen": -69.68389892578125, "logps/ref_rejected": -85.15919494628906, "logps/rejected": -360.2614440917969, "loss": 0.8568, "margin_dpo/margin_mean": 122.83321380615234, "margin_dpo/margin_std": 124.98741149902344, "step": 306 }, { "epoch": 0.45080763582966227, "fcm_dpo/beta": 0.00686724903061986, "fcm_dpo/delta": 0.04468066990375519, "fcm_dpo/margin": 117.76041412353516, "fcm_dpo/q_t": 0.3355296850204468, "grad_norm": 43.83578109741211, "learning_rate": 3.367098400098881e-07, "logits/chosen": -0.12014691531658173, "logits/rejected": -0.09900471568107605, "logps/chosen": -214.15711975097656, "logps/ref_chosen": -70.16542053222656, "logps/ref_rejected": -86.97230529785156, "logps/rejected": -348.7243957519531, "loss": 0.9003, "margin_dpo/margin_mean": 117.76041412353516, "margin_dpo/margin_std": 131.61854553222656, "step": 307 }, { "epoch": 0.4522760646108664, "fcm_dpo/beta": 0.00685135880485177, "fcm_dpo/delta": -0.05890195071697235, "fcm_dpo/margin": 131.8763885498047, "fcm_dpo/q_t": 0.3151218891143799, "grad_norm": 77.11981201171875, "learning_rate": 3.355050358314172e-07, "logits/chosen": -0.07158000022172928, "logits/rejected": -0.048265837132930756, "logps/chosen": -199.24583435058594, "logps/ref_chosen": -55.2449951171875, "logps/ref_rejected": -79.37226104736328, "logps/rejected": -355.2494812011719, "loss": 0.824, "margin_dpo/margin_mean": 131.8763885498047, "margin_dpo/margin_std": 125.1200942993164, "step": 308 }, { "epoch": 0.45374449339207046, "fcm_dpo/beta": 0.006839036010205746, "fcm_dpo/delta": -0.05211016535758972, "fcm_dpo/margin": 131.0915069580078, "fcm_dpo/q_t": 0.3178969919681549, "grad_norm": 51.315860748291016, "learning_rate": 3.3429797851573183e-07, "logits/chosen": -0.036357562988996506, "logits/rejected": -0.011226684786379337, "logps/chosen": -194.12411499023438, "logps/ref_chosen": -48.959083557128906, "logps/ref_rejected": -82.34072875976562, "logps/rejected": -358.5972900390625, "loss": 0.8367, "margin_dpo/margin_mean": 131.0915069580078, "margin_dpo/margin_std": 127.46534729003906, "step": 309 }, { "epoch": 0.4552129221732746, "fcm_dpo/beta": 0.006814408116042614, "fcm_dpo/delta": 0.012445982545614243, "fcm_dpo/margin": 122.96669006347656, "fcm_dpo/q_t": 0.321883887052536, "grad_norm": 40.05405044555664, "learning_rate": 3.3308869986991487e-07, "logits/chosen": -0.0659695714712143, "logits/rejected": -0.039075713604688644, "logps/chosen": -229.92140197753906, "logps/ref_chosen": -62.74177932739258, "logps/ref_rejected": -79.9302978515625, "logps/rejected": -370.07659912109375, "loss": 0.8222, "margin_dpo/margin_mean": 122.96668243408203, "margin_dpo/margin_std": 104.34278106689453, "step": 310 }, { "epoch": 0.4566813509544787, "fcm_dpo/beta": 0.006636098027229309, "fcm_dpo/delta": -0.12000095099210739, "fcm_dpo/margin": 144.33042907714844, "fcm_dpo/q_t": 0.31533536314964294, "grad_norm": 39.31062698364258, "learning_rate": 3.3187723175958346e-07, "logits/chosen": -0.06826113164424896, "logits/rejected": -0.042183272540569305, "logps/chosen": -235.5238494873047, "logps/ref_chosen": -53.02798080444336, "logps/ref_rejected": -77.43820190429688, "logps/rejected": -404.2644958496094, "loss": 0.8325, "margin_dpo/margin_mean": 144.33042907714844, "margin_dpo/margin_std": 152.86956787109375, "step": 311 }, { "epoch": 0.4581497797356828, "fcm_dpo/beta": 0.006625012028962374, "fcm_dpo/delta": -0.02862781658768654, "fcm_dpo/margin": 132.19378662109375, "fcm_dpo/q_t": 0.32340696454048157, "grad_norm": 56.44401550292969, "learning_rate": 3.306636061080487e-07, "logits/chosen": 0.06051143631339073, "logits/rejected": 0.07712406665086746, "logps/chosen": -231.40371704101562, "logps/ref_chosen": -49.39221954345703, "logps/ref_rejected": -75.79280853271484, "logps/rejected": -389.9980773925781, "loss": 0.8657, "margin_dpo/margin_mean": 132.19378662109375, "margin_dpo/margin_std": 143.23089599609375, "step": 312 }, { "epoch": 0.45961820851688695, "fcm_dpo/beta": 0.006699627265334129, "fcm_dpo/delta": 0.05388227105140686, "fcm_dpo/margin": 119.13378143310547, "fcm_dpo/q_t": 0.34248650074005127, "grad_norm": 41.009490966796875, "learning_rate": 3.2944785489547537e-07, "logits/chosen": 0.02956494688987732, "logits/rejected": 0.03531932458281517, "logps/chosen": -225.14703369140625, "logps/ref_chosen": -50.152740478515625, "logps/ref_rejected": -86.40620422363281, "logps/rejected": -380.5343017578125, "loss": 0.9434, "margin_dpo/margin_mean": 119.13378143310547, "margin_dpo/margin_std": 149.03701782226562, "step": 313 }, { "epoch": 0.461086637298091, "fcm_dpo/beta": 0.006667881738394499, "fcm_dpo/delta": 0.05602387711405754, "fcm_dpo/margin": 119.68208312988281, "fcm_dpo/q_t": 0.3448660671710968, "grad_norm": 40.57001495361328, "learning_rate": 3.2823001015803857e-07, "logits/chosen": -0.01754029095172882, "logits/rejected": -0.010302603244781494, "logps/chosen": -255.24281311035156, "logps/ref_chosen": -57.237579345703125, "logps/ref_rejected": -97.5965347290039, "logps/rejected": -415.2838439941406, "loss": 0.9455, "margin_dpo/margin_mean": 119.68208312988281, "margin_dpo/margin_std": 152.344482421875, "step": 314 }, { "epoch": 0.46255506607929514, "fcm_dpo/beta": 0.006808501668274403, "fcm_dpo/delta": 0.08605809509754181, "fcm_dpo/margin": 113.18682861328125, "fcm_dpo/q_t": 0.3449942171573639, "grad_norm": 55.45918273925781, "learning_rate": 3.270101039870797e-07, "logits/chosen": 0.13534711301326752, "logits/rejected": 0.13563254475593567, "logps/chosen": -255.06546020507812, "logps/ref_chosen": -49.06958770751953, "logps/ref_rejected": -85.68087768554688, "logps/rejected": -404.86358642578125, "loss": 0.9118, "margin_dpo/margin_mean": 113.18682861328125, "margin_dpo/margin_std": 130.35165405273438, "step": 315 }, { "epoch": 0.46402349486049926, "fcm_dpo/beta": 0.006600411608815193, "fcm_dpo/delta": -0.19463670253753662, "fcm_dpo/margin": 154.8727569580078, "fcm_dpo/q_t": 0.3011382222175598, "grad_norm": 46.55292892456055, "learning_rate": 3.2578816852826086e-07, "logits/chosen": 0.06141500920057297, "logits/rejected": 0.06320170313119888, "logps/chosen": -257.7305908203125, "logps/ref_chosen": -54.26074981689453, "logps/ref_rejected": -101.2814712524414, "logps/rejected": -459.6240539550781, "loss": 0.7816, "margin_dpo/margin_mean": 154.87277221679688, "margin_dpo/margin_std": 149.25509643554688, "step": 316 }, { "epoch": 0.4654919236417034, "fcm_dpo/beta": 0.006331765092909336, "fcm_dpo/delta": -0.276128351688385, "fcm_dpo/margin": 173.1220703125, "fcm_dpo/q_t": 0.27989763021469116, "grad_norm": 38.22316360473633, "learning_rate": 3.2456423598071783e-07, "logits/chosen": 0.021085377782583237, "logits/rejected": 0.03478589653968811, "logps/chosen": -270.232666015625, "logps/ref_chosen": -56.094207763671875, "logps/ref_rejected": -100.69905090332031, "logps/rejected": -487.9595947265625, "loss": 0.726, "margin_dpo/margin_mean": 173.1220703125, "margin_dpo/margin_std": 143.30935668945312, "step": 317 }, { "epoch": 0.4669603524229075, "fcm_dpo/beta": 0.0061859143897891045, "fcm_dpo/delta": -0.0931069478392601, "fcm_dpo/margin": 151.012939453125, "fcm_dpo/q_t": 0.30796754360198975, "grad_norm": 49.025367736816406, "learning_rate": 3.233383385962115e-07, "logits/chosen": -0.04002639278769493, "logits/rejected": -0.00955579336732626, "logps/chosen": -275.85675048828125, "logps/ref_chosen": -64.64569854736328, "logps/ref_rejected": -82.76425170898438, "logps/rejected": -444.98822021484375, "loss": 0.801, "margin_dpo/margin_mean": 151.012939453125, "margin_dpo/margin_std": 136.198486328125, "step": 318 }, { "epoch": 0.4684287812041116, "fcm_dpo/beta": 0.006093823350965977, "fcm_dpo/delta": -0.023943927139043808, "fcm_dpo/margin": 143.07208251953125, "fcm_dpo/q_t": 0.3199493885040283, "grad_norm": 36.52496337890625, "learning_rate": 3.2211050867827805e-07, "logits/chosen": 0.015176716260612011, "logits/rejected": 0.0014225305058062077, "logps/chosen": -254.9478302001953, "logps/ref_chosen": -49.383758544921875, "logps/ref_rejected": -113.90650939941406, "logps/rejected": -462.54266357421875, "loss": 0.8406, "margin_dpo/margin_mean": 143.07208251953125, "margin_dpo/margin_std": 138.97381591796875, "step": 319 }, { "epoch": 0.4698972099853157, "fcm_dpo/beta": 0.005952038802206516, "fcm_dpo/delta": -0.11902501434087753, "fcm_dpo/margin": 160.73431396484375, "fcm_dpo/q_t": 0.30818331241607666, "grad_norm": 37.61782455444336, "learning_rate": 3.208807785813777e-07, "logits/chosen": -0.004372608382254839, "logits/rejected": -0.003199207130819559, "logps/chosen": -259.04400634765625, "logps/ref_chosen": -59.50489044189453, "logps/ref_rejected": -97.66717529296875, "logps/rejected": -457.94061279296875, "loss": 0.7985, "margin_dpo/margin_mean": 160.73431396484375, "margin_dpo/margin_std": 153.39141845703125, "step": 320 }, { "epoch": 0.4713656387665198, "fcm_dpo/beta": 0.005892496090382338, "fcm_dpo/delta": 0.027914032340049744, "fcm_dpo/margin": 139.42694091796875, "fcm_dpo/q_t": 0.3354644179344177, "grad_norm": 50.34774398803711, "learning_rate": 3.1964918071004217e-07, "logits/chosen": 0.11003893613815308, "logits/rejected": 0.12183515727519989, "logps/chosen": -296.3280029296875, "logps/ref_chosen": -61.548683166503906, "logps/ref_rejected": -91.64103698730469, "logps/rejected": -465.8472900390625, "loss": 0.9028, "margin_dpo/margin_mean": 139.42694091796875, "margin_dpo/margin_std": 153.59918212890625, "step": 321 }, { "epoch": 0.47283406754772395, "fcm_dpo/beta": 0.005919340066611767, "fcm_dpo/delta": -0.02661764994263649, "fcm_dpo/margin": 147.57391357421875, "fcm_dpo/q_t": 0.3203071355819702, "grad_norm": 56.95943069458008, "learning_rate": 3.184157475180207e-07, "logits/chosen": 0.08124169707298279, "logits/rejected": 0.08749310672283173, "logps/chosen": -270.20855712890625, "logps/ref_chosen": -57.29003143310547, "logps/ref_rejected": -95.74992370605469, "logps/rejected": -456.24237060546875, "loss": 0.841, "margin_dpo/margin_mean": 147.5738983154297, "margin_dpo/margin_std": 142.39199829101562, "step": 322 }, { "epoch": 0.47430249632892807, "fcm_dpo/beta": 0.005923746153712273, "fcm_dpo/delta": 0.007014013826847076, "fcm_dpo/margin": 142.34339904785156, "fcm_dpo/q_t": 0.32728075981140137, "grad_norm": 70.69721984863281, "learning_rate": 3.171805115074251e-07, "logits/chosen": 0.04329624027013779, "logits/rejected": 0.046803638339042664, "logps/chosen": -281.20391845703125, "logps/ref_chosen": -51.23395919799805, "logps/ref_rejected": -75.06192016601562, "logps/rejected": -447.37530517578125, "loss": 0.8642, "margin_dpo/margin_mean": 142.34339904785156, "margin_dpo/margin_std": 142.3323211669922, "step": 323 }, { "epoch": 0.47577092511013214, "fcm_dpo/beta": 0.00593388918787241, "fcm_dpo/delta": -0.08985871076583862, "fcm_dpo/margin": 156.79869079589844, "fcm_dpo/q_t": 0.31534451246261597, "grad_norm": 74.83609008789062, "learning_rate": 3.1594350522787295e-07, "logits/chosen": 0.022530967369675636, "logits/rejected": 0.047397322952747345, "logps/chosen": -317.4718933105469, "logps/ref_chosen": -65.13516998291016, "logps/ref_rejected": -86.47750854492188, "logps/rejected": -495.6129150390625, "loss": 0.8564, "margin_dpo/margin_mean": 156.79869079589844, "margin_dpo/margin_std": 162.0347900390625, "step": 324 }, { "epoch": 0.47723935389133626, "fcm_dpo/beta": 0.005914532113820314, "fcm_dpo/delta": 0.08749772608280182, "fcm_dpo/margin": 130.07130432128906, "fcm_dpo/q_t": 0.340700626373291, "grad_norm": 51.47658157348633, "learning_rate": 3.147047612756302e-07, "logits/chosen": -0.0764886736869812, "logits/rejected": -0.04481898248195648, "logps/chosen": -261.2364807128906, "logps/ref_chosen": -56.215599060058594, "logps/ref_rejected": -70.08592987060547, "logps/rejected": -405.1781311035156, "loss": 0.9054, "margin_dpo/margin_mean": 130.07130432128906, "margin_dpo/margin_std": 138.21401977539062, "step": 325 }, { "epoch": 0.4787077826725404, "fcm_dpo/beta": 0.006035645958036184, "fcm_dpo/delta": 0.10336153209209442, "fcm_dpo/margin": 125.01016235351562, "fcm_dpo/q_t": 0.34127509593963623, "grad_norm": 60.38080978393555, "learning_rate": 3.134643122927519e-07, "logits/chosen": -0.05668081343173981, "logits/rejected": -0.019624464213848114, "logps/chosen": -288.8016357421875, "logps/ref_chosen": -72.72496032714844, "logps/ref_rejected": -79.8467788696289, "logps/rejected": -420.93359375, "loss": 0.8786, "margin_dpo/margin_mean": 125.0101547241211, "margin_dpo/margin_std": 122.0008773803711, "step": 326 }, { "epoch": 0.4801762114537445, "fcm_dpo/beta": 0.0058260331861674786, "fcm_dpo/delta": -0.22583037614822388, "fcm_dpo/margin": 180.06192016601562, "fcm_dpo/q_t": 0.2888588607311249, "grad_norm": 49.314125061035156, "learning_rate": 3.1222219096622264e-07, "logits/chosen": -0.04947715997695923, "logits/rejected": -0.02368859574198723, "logps/chosen": -264.2978515625, "logps/ref_chosen": -69.13441467285156, "logps/ref_rejected": -111.93377685546875, "logps/rejected": -487.15911865234375, "loss": 0.7669, "margin_dpo/margin_mean": 180.06192016601562, "margin_dpo/margin_std": 160.97418212890625, "step": 327 }, { "epoch": 0.48164464023494863, "fcm_dpo/beta": 0.005800084210932255, "fcm_dpo/delta": -0.018652496859431267, "fcm_dpo/margin": 149.482177734375, "fcm_dpo/q_t": 0.3247663974761963, "grad_norm": 56.6755485534668, "learning_rate": 3.1097843002709427e-07, "logits/chosen": -0.004470609128475189, "logits/rejected": 0.00039456598460674286, "logps/chosen": -270.512451171875, "logps/ref_chosen": -59.68719482421875, "logps/ref_rejected": -90.85499572753906, "logps/rejected": -451.16241455078125, "loss": 0.8476, "margin_dpo/margin_mean": 149.482177734375, "margin_dpo/margin_std": 151.96820068359375, "step": 328 }, { "epoch": 0.4831130690161527, "fcm_dpo/beta": 0.00566395279020071, "fcm_dpo/delta": -0.09518752992153168, "fcm_dpo/margin": 165.03314208984375, "fcm_dpo/q_t": 0.31437113881111145, "grad_norm": 45.39422607421875, "learning_rate": 3.0973306224962437e-07, "logits/chosen": 0.044825874269008636, "logits/rejected": 0.059470441192388535, "logps/chosen": -300.3798828125, "logps/ref_chosen": -65.2461929321289, "logps/ref_rejected": -100.69770812988281, "logps/rejected": -500.8645324707031, "loss": 0.8432, "margin_dpo/margin_mean": 165.03314208984375, "margin_dpo/margin_std": 168.14520263671875, "step": 329 }, { "epoch": 0.4845814977973568, "fcm_dpo/beta": 0.005626134108752012, "fcm_dpo/delta": -0.04790551960468292, "fcm_dpo/margin": 158.8123321533203, "fcm_dpo/q_t": 0.31789863109588623, "grad_norm": 45.673545837402344, "learning_rate": 3.084861204504122e-07, "logits/chosen": 0.09089761972427368, "logits/rejected": 0.08711989969015121, "logps/chosen": -265.07464599609375, "logps/ref_chosen": -46.998348236083984, "logps/ref_rejected": -86.87684631347656, "logps/rejected": -463.76544189453125, "loss": 0.8397, "margin_dpo/margin_mean": 158.8123321533203, "margin_dpo/margin_std": 159.0528106689453, "step": 330 }, { "epoch": 0.48604992657856094, "fcm_dpo/beta": 0.005541889928281307, "fcm_dpo/delta": -0.0821513682603836, "fcm_dpo/margin": 166.78671264648438, "fcm_dpo/q_t": 0.30144935846328735, "grad_norm": 50.89043426513672, "learning_rate": 3.072376374875335e-07, "logits/chosen": 0.1645202934741974, "logits/rejected": 0.18161094188690186, "logps/chosen": -286.89776611328125, "logps/ref_chosen": -50.52424621582031, "logps/ref_rejected": -89.01544189453125, "logps/rejected": -492.1756896972656, "loss": 0.7535, "margin_dpo/margin_mean": 166.78672790527344, "margin_dpo/margin_std": 120.23434448242188, "step": 331 }, { "epoch": 0.48751835535976507, "fcm_dpo/beta": 0.005678609013557434, "fcm_dpo/delta": 0.16144299507141113, "fcm_dpo/margin": 123.15026092529297, "fcm_dpo/q_t": 0.35300952196121216, "grad_norm": 50.86631393432617, "learning_rate": 3.059876462596758e-07, "logits/chosen": 0.11794328689575195, "logits/rejected": 0.1290109008550644, "logps/chosen": -287.29925537109375, "logps/ref_chosen": -49.18028259277344, "logps/ref_rejected": -76.48515319824219, "logps/rejected": -437.75439453125, "loss": 0.9257, "margin_dpo/margin_mean": 123.15026092529297, "margin_dpo/margin_std": 133.58187866210938, "step": 332 }, { "epoch": 0.4889867841409692, "fcm_dpo/beta": 0.005601478740572929, "fcm_dpo/delta": -0.06622818112373352, "fcm_dpo/margin": 162.07449340820312, "fcm_dpo/q_t": 0.321658194065094, "grad_norm": 54.75845718383789, "learning_rate": 3.0473617970527015e-07, "logits/chosen": 0.036008186638355255, "logits/rejected": 0.044931698590517044, "logps/chosen": -301.5813903808594, "logps/ref_chosen": -63.75574493408203, "logps/ref_rejected": -95.04411315917969, "logps/rejected": -494.9442443847656, "loss": 0.8579, "margin_dpo/margin_mean": 162.07449340820312, "margin_dpo/margin_std": 169.41943359375, "step": 333 }, { "epoch": 0.49045521292217326, "fcm_dpo/beta": 0.00569639727473259, "fcm_dpo/delta": 0.07207376509904861, "fcm_dpo/margin": 137.55638122558594, "fcm_dpo/q_t": 0.3387569487094879, "grad_norm": 55.08982467651367, "learning_rate": 3.034832708016243e-07, "logits/chosen": -0.03774498403072357, "logits/rejected": -0.032692015171051025, "logps/chosen": -284.002685546875, "logps/ref_chosen": -66.97975158691406, "logps/ref_rejected": -95.31692504882812, "logps/rejected": -449.896240234375, "loss": 0.9061, "margin_dpo/margin_mean": 137.55636596679688, "margin_dpo/margin_std": 151.70651245117188, "step": 334 }, { "epoch": 0.4919236417033774, "fcm_dpo/beta": 0.00587943522259593, "fcm_dpo/delta": 0.1511404812335968, "fcm_dpo/margin": 120.45762634277344, "fcm_dpo/q_t": 0.3560691773891449, "grad_norm": 64.96064758300781, "learning_rate": 3.022289525640531e-07, "logits/chosen": 0.07412372529506683, "logits/rejected": 0.09628117084503174, "logps/chosen": -290.11248779296875, "logps/ref_chosen": -62.54248046875, "logps/ref_rejected": -87.61770629882812, "logps/rejected": -435.64532470703125, "loss": 0.9656, "margin_dpo/margin_mean": 120.4576187133789, "margin_dpo/margin_std": 151.005615234375, "step": 335 }, { "epoch": 0.4933920704845815, "fcm_dpo/beta": 0.005829405039548874, "fcm_dpo/delta": -0.06654589623212814, "fcm_dpo/margin": 156.17337036132812, "fcm_dpo/q_t": 0.3145473599433899, "grad_norm": 35.54839324951172, "learning_rate": 3.009732580450086e-07, "logits/chosen": 0.029192904010415077, "logits/rejected": 0.03370156139135361, "logps/chosen": -260.6049499511719, "logps/ref_chosen": -54.53115463256836, "logps/ref_rejected": -104.40424346923828, "logps/rejected": -466.65142822265625, "loss": 0.8382, "margin_dpo/margin_mean": 156.17337036132812, "margin_dpo/margin_std": 156.29110717773438, "step": 336 }, { "epoch": 0.4948604992657856, "fcm_dpo/beta": 0.0057077473029494286, "fcm_dpo/delta": -0.0889483243227005, "fcm_dpo/margin": 162.87271118164062, "fcm_dpo/q_t": 0.3076227307319641, "grad_norm": 49.6478385925293, "learning_rate": 2.9971622033320914e-07, "logits/chosen": -0.03223178908228874, "logits/rejected": -0.024181274697184563, "logps/chosen": -233.42138671875, "logps/ref_chosen": -65.12869262695312, "logps/ref_rejected": -101.72701263427734, "logps/rejected": -432.89239501953125, "loss": 0.7977, "margin_dpo/margin_mean": 162.87271118164062, "margin_dpo/margin_std": 144.59527587890625, "step": 337 }, { "epoch": 0.49632892804698975, "fcm_dpo/beta": 0.005662037990987301, "fcm_dpo/delta": -0.05141569674015045, "fcm_dpo/margin": 158.37158203125, "fcm_dpo/q_t": 0.31238317489624023, "grad_norm": 50.83082580566406, "learning_rate": 2.984578725527675e-07, "logits/chosen": -0.012493651360273361, "logits/rejected": 0.010093292221426964, "logps/chosen": -225.07749938964844, "logps/ref_chosen": -58.422706604003906, "logps/ref_rejected": -89.06854248046875, "logps/rejected": -414.09490966796875, "loss": 0.8022, "margin_dpo/margin_mean": 158.37158203125, "margin_dpo/margin_std": 136.95310974121094, "step": 338 }, { "epoch": 0.4977973568281938, "fcm_dpo/beta": 0.005685793701559305, "fcm_dpo/delta": -0.012543223798274994, "fcm_dpo/margin": 151.34521484375, "fcm_dpo/q_t": 0.3188544511795044, "grad_norm": 43.71391677856445, "learning_rate": 2.9719824786231796e-07, "logits/chosen": -0.0183430016040802, "logits/rejected": -0.007623875513672829, "logps/chosen": -240.96888732910156, "logps/ref_chosen": -59.99531555175781, "logps/ref_rejected": -103.9109115600586, "logps/rejected": -436.2297058105469, "loss": 0.8153, "margin_dpo/margin_mean": 151.34521484375, "margin_dpo/margin_std": 130.1288604736328, "step": 339 }, { "epoch": 0.49926578560939794, "fcm_dpo/beta": 0.0056188758462667465, "fcm_dpo/delta": 0.01368454098701477, "fcm_dpo/margin": 148.94017028808594, "fcm_dpo/q_t": 0.3267067074775696, "grad_norm": 61.01662063598633, "learning_rate": 2.959373794541426e-07, "logits/chosen": 0.02150595560669899, "logits/rejected": 0.04815651848912239, "logps/chosen": -254.2559356689453, "logps/ref_chosen": -52.83022689819336, "logps/ref_rejected": -73.10723114013672, "logps/rejected": -423.4731140136719, "loss": 0.856, "margin_dpo/margin_mean": 148.940185546875, "margin_dpo/margin_std": 147.6375732421875, "step": 340 }, { "epoch": 0.5007342143906021, "fcm_dpo/beta": 0.0055615827441215515, "fcm_dpo/delta": -0.07392589747905731, "fcm_dpo/margin": 164.7185821533203, "fcm_dpo/q_t": 0.3134787678718567, "grad_norm": 54.293521881103516, "learning_rate": 2.946753005532965e-07, "logits/chosen": 0.03781311213970184, "logits/rejected": 0.034403085708618164, "logps/chosen": -265.510986328125, "logps/ref_chosen": -47.899803161621094, "logps/ref_rejected": -101.80987548828125, "logps/rejected": -484.1396484375, "loss": 0.8249, "margin_dpo/margin_mean": 164.71856689453125, "margin_dpo/margin_std": 157.4544677734375, "step": 341 }, { "epoch": 0.5022026431718062, "fcm_dpo/beta": 0.00556070264428854, "fcm_dpo/delta": -0.04328972473740578, "fcm_dpo/margin": 159.92022705078125, "fcm_dpo/q_t": 0.3194407820701599, "grad_norm": 41.392662048339844, "learning_rate": 2.934120444167326e-07, "logits/chosen": -0.04628030210733414, "logits/rejected": -0.02110264264047146, "logps/chosen": -283.7849426269531, "logps/ref_chosen": -71.99664306640625, "logps/ref_rejected": -92.58959197998047, "logps/rejected": -464.2981262207031, "loss": 0.8333, "margin_dpo/margin_mean": 159.92022705078125, "margin_dpo/margin_std": 158.2020263671875, "step": 342 }, { "epoch": 0.5036710719530103, "fcm_dpo/beta": 0.005446239374577999, "fcm_dpo/delta": -0.08976535499095917, "fcm_dpo/margin": 170.99798583984375, "fcm_dpo/q_t": 0.3034150302410126, "grad_norm": 50.54157257080078, "learning_rate": 2.9214764433242476e-07, "logits/chosen": 0.11037549376487732, "logits/rejected": 0.11976627260446548, "logps/chosen": -277.7355041503906, "logps/ref_chosen": -54.405616760253906, "logps/ref_rejected": -111.04142761230469, "logps/rejected": -505.3692626953125, "loss": 0.7733, "margin_dpo/margin_mean": 170.99798583984375, "margin_dpo/margin_std": 136.3294219970703, "step": 343 }, { "epoch": 0.5051395007342144, "fcm_dpo/beta": 0.005502544809132814, "fcm_dpo/delta": 0.005939602851867676, "fcm_dpo/margin": 152.88720703125, "fcm_dpo/q_t": 0.3289456069469452, "grad_norm": 63.03850173950195, "learning_rate": 2.9088213361849126e-07, "logits/chosen": 0.09415749460458755, "logits/rejected": 0.09810211509466171, "logps/chosen": -292.4935302734375, "logps/ref_chosen": -53.96466827392578, "logps/ref_rejected": -90.62336730957031, "logps/rejected": -482.0394287109375, "loss": 0.859, "margin_dpo/margin_mean": 152.88720703125, "margin_dpo/margin_std": 154.0537109375, "step": 344 }, { "epoch": 0.5066079295154186, "fcm_dpo/beta": 0.005438483320176601, "fcm_dpo/delta": -0.030558232218027115, "fcm_dpo/margin": 161.34080505371094, "fcm_dpo/q_t": 0.3203575611114502, "grad_norm": 54.166324615478516, "learning_rate": 2.896155456223163e-07, "logits/chosen": 0.0032112393528223038, "logits/rejected": 0.020324693992733955, "logps/chosen": -325.7508544921875, "logps/ref_chosen": -61.685699462890625, "logps/ref_rejected": -99.49041748046875, "logps/rejected": -524.8964233398438, "loss": 0.8401, "margin_dpo/margin_mean": 161.3408203125, "margin_dpo/margin_std": 158.29058837890625, "step": 345 }, { "epoch": 0.5080763582966226, "fcm_dpo/beta": 0.0054172007367014885, "fcm_dpo/delta": 0.03349316492676735, "fcm_dpo/margin": 151.2305145263672, "fcm_dpo/q_t": 0.3277038335800171, "grad_norm": 49.29082489013672, "learning_rate": 2.883479137196714e-07, "logits/chosen": 0.19713056087493896, "logits/rejected": 0.2153664231300354, "logps/chosen": -339.09173583984375, "logps/ref_chosen": -55.256263732910156, "logps/ref_rejected": -77.41532135009766, "logps/rejected": -512.4812622070312, "loss": 0.8599, "margin_dpo/margin_mean": 151.2305145263672, "margin_dpo/margin_std": 147.90621948242188, "step": 346 }, { "epoch": 0.5095447870778267, "fcm_dpo/beta": 0.00541126262396574, "fcm_dpo/delta": -0.013398189097642899, "fcm_dpo/margin": 159.3125, "fcm_dpo/q_t": 0.32410484552383423, "grad_norm": 47.86494827270508, "learning_rate": 2.8707927131383614e-07, "logits/chosen": 0.14842018485069275, "logits/rejected": 0.16279613971710205, "logps/chosen": -331.56488037109375, "logps/ref_chosen": -57.56623840332031, "logps/ref_rejected": -92.35509490966797, "logps/rejected": -525.6661987304688, "loss": 0.8662, "margin_dpo/margin_mean": 159.3125, "margin_dpo/margin_std": 166.56910705566406, "step": 347 }, { "epoch": 0.5110132158590308, "fcm_dpo/beta": 0.005429641343653202, "fcm_dpo/delta": 0.07389134168624878, "fcm_dpo/margin": 143.79965209960938, "fcm_dpo/q_t": 0.3401951193809509, "grad_norm": 47.6584587097168, "learning_rate": 2.858096518347179e-07, "logits/chosen": 0.09554639458656311, "logits/rejected": 0.10207704454660416, "logps/chosen": -311.5487060546875, "logps/ref_chosen": -56.31770324707031, "logps/ref_rejected": -89.13836669921875, "logps/rejected": -488.16900634765625, "loss": 0.8958, "margin_dpo/margin_mean": 143.79965209960938, "margin_dpo/margin_std": 152.26095581054688, "step": 348 }, { "epoch": 0.5124816446402349, "fcm_dpo/beta": 0.005589696578681469, "fcm_dpo/delta": 0.036192864179611206, "fcm_dpo/margin": 145.68653869628906, "fcm_dpo/q_t": 0.3400714099407196, "grad_norm": 44.34540939331055, "learning_rate": 2.845390887379706e-07, "logits/chosen": 0.1389956921339035, "logits/rejected": 0.1430295705795288, "logps/chosen": -307.2108154296875, "logps/ref_chosen": -58.025516510009766, "logps/ref_rejected": -97.50515747070312, "logps/rejected": -492.3769836425781, "loss": 0.9291, "margin_dpo/margin_mean": 145.68653869628906, "margin_dpo/margin_std": 177.18255615234375, "step": 349 }, { "epoch": 0.5139500734214391, "fcm_dpo/beta": 0.005505557172000408, "fcm_dpo/delta": 0.0005776733160018921, "fcm_dpo/margin": 154.13563537597656, "fcm_dpo/q_t": 0.32515567541122437, "grad_norm": 74.10888671875, "learning_rate": 2.8326761550411346e-07, "logits/chosen": 0.05175316706299782, "logits/rejected": 0.061344295740127563, "logps/chosen": -309.2527770996094, "logps/ref_chosen": -64.33049011230469, "logps/ref_rejected": -89.87164306640625, "logps/rejected": -488.9295654296875, "loss": 0.869, "margin_dpo/margin_mean": 154.13565063476562, "margin_dpo/margin_std": 157.90634155273438, "step": 350 }, { "epoch": 0.5154185022026432, "fcm_dpo/beta": 0.005446532741189003, "fcm_dpo/delta": -0.13678933680057526, "fcm_dpo/margin": 178.76748657226562, "fcm_dpo/q_t": 0.305521160364151, "grad_norm": 50.48676681518555, "learning_rate": 2.819952656376487e-07, "logits/chosen": -0.02746611461043358, "logits/rejected": -0.01481974683701992, "logps/chosen": -298.5670166015625, "logps/ref_chosen": -60.6721305847168, "logps/ref_rejected": -101.5654296875, "logps/rejected": -518.227783203125, "loss": 0.8216, "margin_dpo/margin_mean": 178.76748657226562, "margin_dpo/margin_std": 179.7300567626953, "step": 351 }, { "epoch": 0.5168869309838473, "fcm_dpo/beta": 0.005514339543879032, "fcm_dpo/delta": 0.19816651940345764, "fcm_dpo/margin": 120.65501403808594, "fcm_dpo/q_t": 0.36773788928985596, "grad_norm": 75.69171905517578, "learning_rate": 2.8072207266617854e-07, "logits/chosen": 0.0787711963057518, "logits/rejected": 0.11258897185325623, "logps/chosen": -345.6042175292969, "logps/ref_chosen": -70.9434585571289, "logps/ref_rejected": -76.6419677734375, "logps/rejected": -471.957763671875, "loss": 1.0103, "margin_dpo/margin_mean": 120.65501403808594, "margin_dpo/margin_std": 167.7642364501953, "step": 352 }, { "epoch": 0.5183553597650514, "fcm_dpo/beta": 0.005604561883956194, "fcm_dpo/delta": 0.042658913880586624, "fcm_dpo/margin": 144.52413940429688, "fcm_dpo/q_t": 0.3393586277961731, "grad_norm": 46.874942779541016, "learning_rate": 2.794480701395219e-07, "logits/chosen": 0.08481930941343307, "logits/rejected": 0.10338996350765228, "logps/chosen": -321.98089599609375, "logps/ref_chosen": -58.39533996582031, "logps/ref_rejected": -80.33553314208984, "logps/rejected": -488.4452209472656, "loss": 0.9148, "margin_dpo/margin_mean": 144.52413940429688, "margin_dpo/margin_std": 167.46084594726562, "step": 353 }, { "epoch": 0.5198237885462555, "fcm_dpo/beta": 0.005544104613363743, "fcm_dpo/delta": -0.1198074221611023, "fcm_dpo/margin": 172.76959228515625, "fcm_dpo/q_t": 0.29841989278793335, "grad_norm": 36.41807556152344, "learning_rate": 2.781732916288303e-07, "logits/chosen": 0.0665828287601471, "logits/rejected": 0.08835647255182266, "logps/chosen": -291.6112365722656, "logps/ref_chosen": -59.80299377441406, "logps/ref_rejected": -88.75750732421875, "logps/rejected": -493.3353271484375, "loss": 0.7587, "margin_dpo/margin_mean": 172.76959228515625, "margin_dpo/margin_std": 136.2689208984375, "step": 354 }, { "epoch": 0.5212922173274597, "fcm_dpo/beta": 0.005405407398939133, "fcm_dpo/delta": -0.11095987260341644, "fcm_dpo/margin": 175.61483764648438, "fcm_dpo/q_t": 0.30244117975234985, "grad_norm": 57.12250518798828, "learning_rate": 2.7689777072570284e-07, "logits/chosen": 0.06327119469642639, "logits/rejected": 0.09310467541217804, "logps/chosen": -288.7947692871094, "logps/ref_chosen": -54.12849807739258, "logps/ref_rejected": -82.40606689453125, "logps/rejected": -492.68719482421875, "loss": 0.7733, "margin_dpo/margin_mean": 175.61483764648438, "margin_dpo/margin_std": 145.6189422607422, "step": 355 }, { "epoch": 0.5227606461086637, "fcm_dpo/beta": 0.0056666238233447075, "fcm_dpo/delta": 0.23844945430755615, "fcm_dpo/margin": 109.56277465820312, "fcm_dpo/q_t": 0.38272541761398315, "grad_norm": 58.59632110595703, "learning_rate": 2.7562154104130176e-07, "logits/chosen": 0.0763608068227768, "logits/rejected": 0.09287572652101517, "logps/chosen": -345.8939208984375, "logps/ref_chosen": -64.6738052368164, "logps/ref_rejected": -75.89926147460938, "logps/rejected": -466.6821594238281, "loss": 1.1128, "margin_dpo/margin_mean": 109.56277465820312, "margin_dpo/margin_std": 190.751708984375, "step": 356 }, { "epoch": 0.5242290748898678, "fcm_dpo/beta": 0.005567336454987526, "fcm_dpo/delta": -0.07430887222290039, "fcm_dpo/margin": 164.54934692382812, "fcm_dpo/q_t": 0.31005075573921204, "grad_norm": 52.44282531738281, "learning_rate": 2.7434463620546594e-07, "logits/chosen": 0.09520265460014343, "logits/rejected": 0.11559751629829407, "logps/chosen": -318.9443359375, "logps/ref_chosen": -52.725799560546875, "logps/ref_rejected": -86.84115600585938, "logps/rejected": -517.6090087890625, "loss": 0.8028, "margin_dpo/margin_mean": 164.54934692382812, "margin_dpo/margin_std": 146.5643768310547, "step": 357 }, { "epoch": 0.5256975036710719, "fcm_dpo/beta": 0.005649616941809654, "fcm_dpo/delta": 0.07324951887130737, "fcm_dpo/margin": 138.49087524414062, "fcm_dpo/q_t": 0.34356987476348877, "grad_norm": 37.71842956542969, "learning_rate": 2.730670898658255e-07, "logits/chosen": 0.0035172095522284508, "logits/rejected": 0.023625530302524567, "logps/chosen": -319.94830322265625, "logps/ref_chosen": -63.20543670654297, "logps/ref_rejected": -88.373291015625, "logps/rejected": -483.60699462890625, "loss": 0.9045, "margin_dpo/margin_mean": 138.49087524414062, "margin_dpo/margin_std": 159.35696411132812, "step": 358 }, { "epoch": 0.527165932452276, "fcm_dpo/beta": 0.005610806867480278, "fcm_dpo/delta": -0.041874960064888, "fcm_dpo/margin": 158.2200927734375, "fcm_dpo/q_t": 0.3280462622642517, "grad_norm": 69.10176086425781, "learning_rate": 2.717889356869146e-07, "logits/chosen": 0.1252935528755188, "logits/rejected": 0.1476270705461502, "logps/chosen": -356.76495361328125, "logps/ref_chosen": -56.370216369628906, "logps/ref_rejected": -82.17375183105469, "logps/rejected": -540.78857421875, "loss": 0.8751, "margin_dpo/margin_mean": 158.2200927734375, "margin_dpo/margin_std": 177.02398681640625, "step": 359 }, { "epoch": 0.5286343612334802, "fcm_dpo/beta": 0.005699576810002327, "fcm_dpo/delta": 0.0714188739657402, "fcm_dpo/margin": 137.50169372558594, "fcm_dpo/q_t": 0.33583831787109375, "grad_norm": 87.95269775390625, "learning_rate": 2.7051020734928443e-07, "logits/chosen": 0.1796594262123108, "logits/rejected": 0.19369468092918396, "logps/chosen": -336.75543212890625, "logps/ref_chosen": -51.460384368896484, "logps/ref_rejected": -69.83892059326172, "logps/rejected": -492.6356506347656, "loss": 0.8694, "margin_dpo/margin_mean": 137.50169372558594, "margin_dpo/margin_std": 132.57077026367188, "step": 360 }, { "epoch": 0.5301027900146843, "fcm_dpo/beta": 0.005806830711662769, "fcm_dpo/delta": 0.09413307905197144, "fcm_dpo/margin": 131.26763916015625, "fcm_dpo/q_t": 0.34591707587242126, "grad_norm": 43.504676818847656, "learning_rate": 2.6923093854861593e-07, "logits/chosen": 0.10329671949148178, "logits/rejected": 0.1067652627825737, "logps/chosen": -333.27484130859375, "logps/ref_chosen": -53.86951446533203, "logps/ref_rejected": -90.7692642211914, "logps/rejected": -501.4422607421875, "loss": 0.9286, "margin_dpo/margin_mean": 131.2676239013672, "margin_dpo/margin_std": 154.85382080078125, "step": 361 }, { "epoch": 0.5315712187958884, "fcm_dpo/beta": 0.005605396814644337, "fcm_dpo/delta": -0.21299204230308533, "fcm_dpo/margin": 185.47732543945312, "fcm_dpo/q_t": 0.28802040219306946, "grad_norm": 42.840492248535156, "learning_rate": 2.679511629948319e-07, "logits/chosen": 0.09067463874816895, "logits/rejected": 0.08635248243808746, "logps/chosen": -302.639404296875, "logps/ref_chosen": -58.639060974121094, "logps/ref_rejected": -105.58195495605469, "logps/rejected": -535.0595703125, "loss": 0.7337, "margin_dpo/margin_mean": 185.47732543945312, "margin_dpo/margin_std": 153.58267211914062, "step": 362 }, { "epoch": 0.5330396475770925, "fcm_dpo/beta": 0.005374937318265438, "fcm_dpo/delta": -0.2472880780696869, "fcm_dpo/margin": 199.2676239013672, "fcm_dpo/q_t": 0.28192996978759766, "grad_norm": 39.75864028930664, "learning_rate": 2.6667091441120816e-07, "logits/chosen": 0.1431807279586792, "logits/rejected": 0.1652299463748932, "logps/chosen": -266.96466064453125, "logps/ref_chosen": -44.558380126953125, "logps/ref_rejected": -74.69496154785156, "logps/rejected": -496.3688659667969, "loss": 0.73, "margin_dpo/margin_mean": 199.26763916015625, "margin_dpo/margin_std": 164.5372772216797, "step": 363 }, { "epoch": 0.5345080763582967, "fcm_dpo/beta": 0.005400581285357475, "fcm_dpo/delta": 0.07649517804384232, "fcm_dpo/margin": 144.00555419921875, "fcm_dpo/q_t": 0.33972108364105225, "grad_norm": 45.31873321533203, "learning_rate": 2.6539022653348575e-07, "logits/chosen": 0.071674644947052, "logits/rejected": 0.07084225863218307, "logps/chosen": -291.6654052734375, "logps/ref_chosen": -48.894622802734375, "logps/ref_rejected": -91.395751953125, "logps/rejected": -478.17205810546875, "loss": 0.9032, "margin_dpo/margin_mean": 144.00555419921875, "margin_dpo/margin_std": 157.2249755859375, "step": 364 }, { "epoch": 0.5359765051395007, "fcm_dpo/beta": 0.005438569001853466, "fcm_dpo/delta": 0.050673648715019226, "fcm_dpo/margin": 147.65203857421875, "fcm_dpo/q_t": 0.3388054668903351, "grad_norm": 49.04033279418945, "learning_rate": 2.641091331089811e-07, "logits/chosen": 0.05171611160039902, "logits/rejected": 0.04313598573207855, "logps/chosen": -280.9577331542969, "logps/ref_chosen": -51.49274444580078, "logps/ref_rejected": -92.70166778564453, "logps/rejected": -469.8186950683594, "loss": 0.8784, "margin_dpo/margin_mean": 147.65203857421875, "margin_dpo/margin_std": 159.82666015625, "step": 365 }, { "epoch": 0.5374449339207048, "fcm_dpo/beta": 0.0054289670661091805, "fcm_dpo/delta": 0.009159095585346222, "fcm_dpo/margin": 155.0063018798828, "fcm_dpo/q_t": 0.3273472189903259, "grad_norm": 38.92621994018555, "learning_rate": 2.6282766789569736e-07, "logits/chosen": 0.14211732149124146, "logits/rejected": 0.14219552278518677, "logps/chosen": -267.73809814453125, "logps/ref_chosen": -44.7205696105957, "logps/ref_rejected": -83.31040954589844, "logps/rejected": -461.3342590332031, "loss": 0.8501, "margin_dpo/margin_mean": 155.00631713867188, "margin_dpo/margin_std": 154.13015747070312, "step": 366 }, { "epoch": 0.5389133627019089, "fcm_dpo/beta": 0.005595469381660223, "fcm_dpo/delta": 0.18162932991981506, "fcm_dpo/margin": 121.64521789550781, "fcm_dpo/q_t": 0.355856716632843, "grad_norm": 38.51764678955078, "learning_rate": 2.615458646614349e-07, "logits/chosen": 0.06456591933965683, "logits/rejected": 0.0903320163488388, "logps/chosen": -279.4564208984375, "logps/ref_chosen": -58.405418395996094, "logps/ref_rejected": -76.75132751464844, "logps/rejected": -419.447509765625, "loss": 0.9315, "margin_dpo/margin_mean": 121.64521789550781, "margin_dpo/margin_std": 129.7523956298828, "step": 367 }, { "epoch": 0.540381791483113, "fcm_dpo/beta": 0.005565423984080553, "fcm_dpo/delta": -0.12588441371917725, "fcm_dpo/margin": 173.19337463378906, "fcm_dpo/q_t": 0.2921599745750427, "grad_norm": 65.85102081298828, "learning_rate": 2.6026375718290083e-07, "logits/chosen": 0.03500935807824135, "logits/rejected": 0.04446982964873314, "logps/chosen": -250.3231964111328, "logps/ref_chosen": -44.452518463134766, "logps/ref_rejected": -98.55526733398438, "logps/rejected": -477.61932373046875, "loss": 0.7302, "margin_dpo/margin_mean": 173.19337463378906, "margin_dpo/margin_std": 117.90557861328125, "step": 368 }, { "epoch": 0.5418502202643172, "fcm_dpo/beta": 0.005647559650242329, "fcm_dpo/delta": 0.13559746742248535, "fcm_dpo/margin": 128.01370239257812, "fcm_dpo/q_t": 0.3497123420238495, "grad_norm": 69.40166473388672, "learning_rate": 2.589813792448196e-07, "logits/chosen": 0.015510164201259613, "logits/rejected": 0.03292276710271835, "logps/chosen": -309.7448425292969, "logps/ref_chosen": -71.38150024414062, "logps/ref_rejected": -91.29582214355469, "logps/rejected": -457.6728515625, "loss": 0.9356, "margin_dpo/margin_mean": 128.01370239257812, "margin_dpo/margin_std": 145.52603149414062, "step": 369 }, { "epoch": 0.5433186490455213, "fcm_dpo/beta": 0.005865996703505516, "fcm_dpo/delta": 0.22589412331581116, "fcm_dpo/margin": 108.72503662109375, "fcm_dpo/q_t": 0.36666351556777954, "grad_norm": 48.482791900634766, "learning_rate": 2.5769876463904263e-07, "logits/chosen": 0.0964912623167038, "logits/rejected": 0.10018481314182281, "logps/chosen": -338.19866943359375, "logps/ref_chosen": -71.60749816894531, "logps/ref_rejected": -97.25978088378906, "logps/rejected": -472.57598876953125, "loss": 0.9842, "margin_dpo/margin_mean": 108.72503662109375, "margin_dpo/margin_std": 135.08151245117188, "step": 370 }, { "epoch": 0.5447870778267254, "fcm_dpo/beta": 0.005965778138488531, "fcm_dpo/delta": 0.03497808426618576, "fcm_dpo/margin": 137.05462646484375, "fcm_dpo/q_t": 0.3335404396057129, "grad_norm": 47.60246658325195, "learning_rate": 2.5641594716365744e-07, "logits/chosen": 0.08590900897979736, "logits/rejected": 0.10128141194581985, "logps/chosen": -339.6130065917969, "logps/ref_chosen": -69.41448974609375, "logps/ref_rejected": -99.17217254638672, "logps/rejected": -506.42529296875, "loss": 0.8969, "margin_dpo/margin_mean": 137.05462646484375, "margin_dpo/margin_std": 152.06521606445312, "step": 371 }, { "epoch": 0.5462555066079295, "fcm_dpo/beta": 0.005722453817725182, "fcm_dpo/delta": -0.25100451707839966, "fcm_dpo/margin": 187.50184631347656, "fcm_dpo/q_t": 0.29311686754226685, "grad_norm": 39.004390716552734, "learning_rate": 2.551329606220976e-07, "logits/chosen": 0.16508790850639343, "logits/rejected": 0.20037144422531128, "logps/chosen": -342.9236755371094, "logps/ref_chosen": -61.8179931640625, "logps/ref_rejected": -78.53948974609375, "logps/rejected": -547.1470336914062, "loss": 0.775, "margin_dpo/margin_mean": 187.50184631347656, "margin_dpo/margin_std": 180.50228881835938, "step": 372 }, { "epoch": 0.5477239353891337, "fcm_dpo/beta": 0.005600764416158199, "fcm_dpo/delta": -0.13974212110042572, "fcm_dpo/margin": 174.2254180908203, "fcm_dpo/q_t": 0.3003678619861603, "grad_norm": 49.6671257019043, "learning_rate": 2.538498388222517e-07, "logits/chosen": 0.20778214931488037, "logits/rejected": 0.24689488112926483, "logps/chosen": -361.0768127441406, "logps/ref_chosen": -64.21713256835938, "logps/ref_rejected": -85.95960998535156, "logps/rejected": -557.044677734375, "loss": 0.792, "margin_dpo/margin_mean": 174.22540283203125, "margin_dpo/margin_std": 154.6001434326172, "step": 373 }, { "epoch": 0.5491923641703378, "fcm_dpo/beta": 0.005461122840642929, "fcm_dpo/delta": -0.006661958992481232, "fcm_dpo/margin": 156.51187133789062, "fcm_dpo/q_t": 0.3341420888900757, "grad_norm": 39.634254455566406, "learning_rate": 2.525666155755725e-07, "logits/chosen": 0.042114850133657455, "logits/rejected": 0.07254066318273544, "logps/chosen": -347.50177001953125, "logps/ref_chosen": -70.65018463134766, "logps/ref_rejected": -93.64016723632812, "logps/rejected": -527.0036010742188, "loss": 0.8785, "margin_dpo/margin_mean": 156.51187133789062, "margin_dpo/margin_std": 173.5503692626953, "step": 374 }, { "epoch": 0.5506607929515418, "fcm_dpo/beta": 0.005447746254503727, "fcm_dpo/delta": 0.027071624994277954, "fcm_dpo/margin": 150.794677734375, "fcm_dpo/q_t": 0.33665597438812256, "grad_norm": 71.83676147460938, "learning_rate": 2.512833246961859e-07, "logits/chosen": 0.09779814630746841, "logits/rejected": 0.10543471574783325, "logps/chosen": -369.1651611328125, "logps/ref_chosen": -60.080223083496094, "logps/ref_rejected": -88.93830871582031, "logps/rejected": -548.81787109375, "loss": 0.9248, "margin_dpo/margin_mean": 150.794677734375, "margin_dpo/margin_std": 173.79681396484375, "step": 375 }, { "epoch": 0.5521292217327459, "fcm_dpo/beta": 0.005424397066235542, "fcm_dpo/delta": -0.13509052991867065, "fcm_dpo/margin": 179.1793670654297, "fcm_dpo/q_t": 0.3073885142803192, "grad_norm": 41.7952880859375, "learning_rate": 2.5e-07, "logits/chosen": 0.09893445670604706, "logits/rejected": 0.11599358916282654, "logps/chosen": -363.12640380859375, "logps/ref_chosen": -62.660308837890625, "logps/ref_rejected": -105.52660369873047, "logps/rejected": -585.1720581054688, "loss": 0.8182, "margin_dpo/margin_mean": 179.1793670654297, "margin_dpo/margin_std": 180.31005859375, "step": 376 }, { "epoch": 0.55359765051395, "fcm_dpo/beta": 0.005333360284566879, "fcm_dpo/delta": -0.07344111055135727, "fcm_dpo/margin": 171.89007568359375, "fcm_dpo/q_t": 0.3136495351791382, "grad_norm": 46.397911071777344, "learning_rate": 2.487166753038141e-07, "logits/chosen": 0.19385088980197906, "logits/rejected": 0.20475682616233826, "logps/chosen": -379.18011474609375, "logps/ref_chosen": -54.478736877441406, "logps/ref_rejected": -98.70335388183594, "logps/rejected": -595.2947998046875, "loss": 0.8312, "margin_dpo/margin_mean": 171.89007568359375, "margin_dpo/margin_std": 168.72779846191406, "step": 377 }, { "epoch": 0.5550660792951542, "fcm_dpo/beta": 0.005202522035688162, "fcm_dpo/delta": -0.10398915410041809, "fcm_dpo/margin": 181.31509399414062, "fcm_dpo/q_t": 0.30345821380615234, "grad_norm": 75.51764678955078, "learning_rate": 2.4743338442442754e-07, "logits/chosen": 0.3088299334049225, "logits/rejected": 0.31884270906448364, "logps/chosen": -373.27655029296875, "logps/ref_chosen": -45.02053451538086, "logps/ref_rejected": -88.0469741821289, "logps/rejected": -597.6180419921875, "loss": 0.7906, "margin_dpo/margin_mean": 181.31509399414062, "margin_dpo/margin_std": 157.98651123046875, "step": 378 }, { "epoch": 0.5565345080763583, "fcm_dpo/beta": 0.005089180544018745, "fcm_dpo/delta": -0.08599420636892319, "fcm_dpo/margin": 181.9398651123047, "fcm_dpo/q_t": 0.31637871265411377, "grad_norm": 59.954917907714844, "learning_rate": 2.461501611777483e-07, "logits/chosen": 0.23598089814186096, "logits/rejected": 0.2249039113521576, "logps/chosen": -385.88226318359375, "logps/ref_chosen": -53.182098388671875, "logps/ref_rejected": -114.3001708984375, "logps/rejected": -628.940185546875, "loss": 0.8413, "margin_dpo/margin_mean": 181.93988037109375, "margin_dpo/margin_std": 186.93975830078125, "step": 379 }, { "epoch": 0.5580029368575624, "fcm_dpo/beta": 0.005020034499466419, "fcm_dpo/delta": -0.168114572763443, "fcm_dpo/margin": 199.48626708984375, "fcm_dpo/q_t": 0.298037052154541, "grad_norm": 41.02925491333008, "learning_rate": 2.4486703937790243e-07, "logits/chosen": 0.17431041598320007, "logits/rejected": 0.15981586277484894, "logps/chosen": -365.4420471191406, "logps/ref_chosen": -51.3530387878418, "logps/ref_rejected": -104.19169616699219, "logps/rejected": -617.7669677734375, "loss": 0.8056, "margin_dpo/margin_mean": 199.48626708984375, "margin_dpo/margin_std": 189.1451873779297, "step": 380 }, { "epoch": 0.5594713656387665, "fcm_dpo/beta": 0.004953205585479736, "fcm_dpo/delta": 0.021930061280727386, "fcm_dpo/margin": 167.54666137695312, "fcm_dpo/q_t": 0.3372423052787781, "grad_norm": 67.10491180419922, "learning_rate": 2.435840528363426e-07, "logits/chosen": 0.19572949409484863, "logits/rejected": 0.21503984928131104, "logps/chosen": -375.8944091796875, "logps/ref_chosen": -57.80306625366211, "logps/ref_rejected": -79.21940612792969, "logps/rejected": -564.857421875, "loss": 0.9236, "margin_dpo/margin_mean": 167.54666137695312, "margin_dpo/margin_std": 204.34890747070312, "step": 381 }, { "epoch": 0.5609397944199707, "fcm_dpo/beta": 0.004994013346731663, "fcm_dpo/delta": 0.05573827028274536, "fcm_dpo/margin": 159.93917846679688, "fcm_dpo/q_t": 0.3334108591079712, "grad_norm": 50.57058334350586, "learning_rate": 2.4230123536095745e-07, "logits/chosen": 0.12441153824329376, "logits/rejected": 0.13224869966506958, "logps/chosen": -371.9028625488281, "logps/ref_chosen": -66.02030181884766, "logps/ref_rejected": -110.71016693115234, "logps/rejected": -576.5319213867188, "loss": 0.8654, "margin_dpo/margin_mean": 159.93917846679688, "margin_dpo/margin_std": 155.31463623046875, "step": 382 }, { "epoch": 0.5624082232011748, "fcm_dpo/beta": 0.004986027255654335, "fcm_dpo/delta": -0.02076072059571743, "fcm_dpo/margin": 174.21910095214844, "fcm_dpo/q_t": 0.3251308500766754, "grad_norm": 40.4781494140625, "learning_rate": 2.4101862075518037e-07, "logits/chosen": 0.18596576154232025, "logits/rejected": 0.1823549121618271, "logps/chosen": -351.5233154296875, "logps/ref_chosen": -50.39148712158203, "logps/ref_rejected": -93.71589660644531, "logps/rejected": -569.06689453125, "loss": 0.8845, "margin_dpo/margin_mean": 174.2191162109375, "margin_dpo/margin_std": 194.69053649902344, "step": 383 }, { "epoch": 0.5638766519823789, "fcm_dpo/beta": 0.005112666171044111, "fcm_dpo/delta": 0.09205277264118195, "fcm_dpo/margin": 149.1619873046875, "fcm_dpo/q_t": 0.3353465795516968, "grad_norm": 38.69187927246094, "learning_rate": 2.397362428170992e-07, "logits/chosen": 0.04655902460217476, "logits/rejected": 0.05491742864251137, "logps/chosen": -335.37225341796875, "logps/ref_chosen": -52.046104431152344, "logps/ref_rejected": -85.76089477539062, "logps/rejected": -518.2490234375, "loss": 0.8667, "margin_dpo/margin_mean": 149.1619873046875, "margin_dpo/margin_std": 133.10096740722656, "step": 384 }, { "epoch": 0.5653450807635829, "fcm_dpo/beta": 0.005061786156147718, "fcm_dpo/delta": -0.047479018568992615, "fcm_dpo/margin": 176.4716796875, "fcm_dpo/q_t": 0.30865800380706787, "grad_norm": 36.45566177368164, "learning_rate": 2.3845413533856514e-07, "logits/chosen": 0.12907695770263672, "logits/rejected": 0.16897349059581757, "logps/chosen": -322.817626953125, "logps/ref_chosen": -65.55215454101562, "logps/ref_rejected": -77.82792663574219, "logps/rejected": -511.5650634765625, "loss": 0.785, "margin_dpo/margin_mean": 176.47166442871094, "margin_dpo/margin_std": 139.1245574951172, "step": 385 }, { "epoch": 0.566813509544787, "fcm_dpo/beta": 0.005007788073271513, "fcm_dpo/delta": -0.07901398092508316, "fcm_dpo/margin": 184.04327392578125, "fcm_dpo/q_t": 0.3119524121284485, "grad_norm": 47.43567657470703, "learning_rate": 2.3717233210430254e-07, "logits/chosen": 0.04530417546629906, "logits/rejected": 0.06010700762271881, "logps/chosen": -312.85601806640625, "logps/ref_chosen": -58.22185516357422, "logps/ref_rejected": -92.32742309570312, "logps/rejected": -531.0048828125, "loss": 0.8184, "margin_dpo/margin_mean": 184.04327392578125, "margin_dpo/margin_std": 175.82864379882812, "step": 386 }, { "epoch": 0.5682819383259912, "fcm_dpo/beta": 0.005064930766820908, "fcm_dpo/delta": 0.16430482268333435, "fcm_dpo/margin": 137.70285034179688, "fcm_dpo/q_t": 0.34966596961021423, "grad_norm": 46.66494369506836, "learning_rate": 2.3589086689101889e-07, "logits/chosen": 0.02450702153146267, "logits/rejected": 0.05837446078658104, "logps/chosen": -343.4558410644531, "logps/ref_chosen": -66.41944885253906, "logps/ref_rejected": -92.16915893554688, "logps/rejected": -506.90838623046875, "loss": 0.9043, "margin_dpo/margin_mean": 137.70285034179688, "margin_dpo/margin_std": 132.16102600097656, "step": 387 }, { "epoch": 0.5697503671071953, "fcm_dpo/beta": 0.005017186980694532, "fcm_dpo/delta": -0.10396112501621246, "fcm_dpo/margin": 187.8830108642578, "fcm_dpo/q_t": 0.30914634466171265, "grad_norm": 44.683433532714844, "learning_rate": 2.3460977346651428e-07, "logits/chosen": 0.11675406992435455, "logits/rejected": 0.10898333787918091, "logps/chosen": -318.9471740722656, "logps/ref_chosen": -50.129459381103516, "logps/ref_rejected": -104.43305969238281, "logps/rejected": -561.1337890625, "loss": 0.8084, "margin_dpo/margin_mean": 187.8830108642578, "margin_dpo/margin_std": 178.2823486328125, "step": 388 }, { "epoch": 0.5712187958883994, "fcm_dpo/beta": 0.0049585141241550446, "fcm_dpo/delta": -0.07406939566135406, "fcm_dpo/margin": 184.94287109375, "fcm_dpo/q_t": 0.3153570890426636, "grad_norm": 38.42765808105469, "learning_rate": 2.3332908558879177e-07, "logits/chosen": 0.13295167684555054, "logits/rejected": 0.1595715433359146, "logps/chosen": -336.7831726074219, "logps/ref_chosen": -57.906593322753906, "logps/ref_rejected": -77.91454315185547, "logps/rejected": -541.7340087890625, "loss": 0.8249, "margin_dpo/margin_mean": 184.94285583496094, "margin_dpo/margin_std": 180.59262084960938, "step": 389 }, { "epoch": 0.5726872246696035, "fcm_dpo/beta": 0.004955414682626724, "fcm_dpo/delta": 0.04264906048774719, "fcm_dpo/margin": 163.55552673339844, "fcm_dpo/q_t": 0.3355684280395508, "grad_norm": 64.1507339477539, "learning_rate": 2.320488370051681e-07, "logits/chosen": 0.10355956852436066, "logits/rejected": 0.11937595903873444, "logps/chosen": -331.3822326660156, "logps/ref_chosen": -49.22591781616211, "logps/ref_rejected": -85.5281982421875, "logps/rejected": -531.239990234375, "loss": 0.9176, "margin_dpo/margin_mean": 163.55551147460938, "margin_dpo/margin_std": 192.38583374023438, "step": 390 }, { "epoch": 0.5741556534508077, "fcm_dpo/beta": 0.005186218768358231, "fcm_dpo/delta": 0.2533876895904541, "fcm_dpo/margin": 118.00922393798828, "fcm_dpo/q_t": 0.3784768879413605, "grad_norm": 48.2935905456543, "learning_rate": 2.3076906145138405e-07, "logits/chosen": 0.08619362860918045, "logits/rejected": 0.08983877301216125, "logps/chosen": -342.7335205078125, "logps/ref_chosen": -64.32965087890625, "logps/ref_rejected": -86.73820495605469, "logps/rejected": -483.15130615234375, "loss": 1.0202, "margin_dpo/margin_mean": 118.00923156738281, "margin_dpo/margin_std": 167.4725341796875, "step": 391 }, { "epoch": 0.5756240822320118, "fcm_dpo/beta": 0.005176296457648277, "fcm_dpo/delta": -0.12520991265773773, "fcm_dpo/margin": 186.098388671875, "fcm_dpo/q_t": 0.3001660406589508, "grad_norm": 46.61800003051758, "learning_rate": 2.294897926507156e-07, "logits/chosen": 0.054711125791072845, "logits/rejected": 0.06594612449407578, "logps/chosen": -293.49102783203125, "logps/ref_chosen": -53.50397872924805, "logps/ref_rejected": -102.34584045410156, "logps/rejected": -528.4312744140625, "loss": 0.7649, "margin_dpo/margin_mean": 186.098388671875, "margin_dpo/margin_std": 153.24871826171875, "step": 392 }, { "epoch": 0.5770925110132159, "fcm_dpo/beta": 0.005192288197577, "fcm_dpo/delta": 0.10276921838521957, "fcm_dpo/margin": 145.42605590820312, "fcm_dpo/q_t": 0.35600078105926514, "grad_norm": 46.10240936279297, "learning_rate": 2.2821106431308543e-07, "logits/chosen": 0.13289561867713928, "logits/rejected": 0.14119097590446472, "logps/chosen": -310.223388671875, "logps/ref_chosen": -46.473915100097656, "logps/ref_rejected": -71.96885681152344, "logps/rejected": -481.1444091796875, "loss": 0.9668, "margin_dpo/margin_mean": 145.42605590820312, "margin_dpo/margin_std": 197.1509552001953, "step": 393 }, { "epoch": 0.57856093979442, "fcm_dpo/beta": 0.00520074088126421, "fcm_dpo/delta": -0.019946470856666565, "fcm_dpo/margin": 166.911865234375, "fcm_dpo/q_t": 0.3283039927482605, "grad_norm": 37.210731506347656, "learning_rate": 2.2693291013417452e-07, "logits/chosen": 0.10127435624599457, "logits/rejected": 0.10776053369045258, "logps/chosen": -332.1357421875, "logps/ref_chosen": -52.91154861450195, "logps/ref_rejected": -90.8226318359375, "logps/rejected": -536.958740234375, "loss": 0.8841, "margin_dpo/margin_mean": 166.911865234375, "margin_dpo/margin_std": 186.99267578125, "step": 394 }, { "epoch": 0.580029368575624, "fcm_dpo/beta": 0.005109791178256273, "fcm_dpo/delta": -0.09205502271652222, "fcm_dpo/margin": 182.4871826171875, "fcm_dpo/q_t": 0.3203425407409668, "grad_norm": 76.74991607666016, "learning_rate": 2.2565536379453404e-07, "logits/chosen": 0.12516123056411743, "logits/rejected": 0.13688474893569946, "logps/chosen": -341.91693115234375, "logps/ref_chosen": -62.546112060546875, "logps/ref_rejected": -83.78262329101562, "logps/rejected": -545.640625, "loss": 0.8685, "margin_dpo/margin_mean": 182.4871826171875, "margin_dpo/margin_std": 205.46337890625, "step": 395 }, { "epoch": 0.5814977973568282, "fcm_dpo/beta": 0.00510101392865181, "fcm_dpo/delta": 0.0003374125808477402, "fcm_dpo/margin": 166.55291748046875, "fcm_dpo/q_t": 0.32344305515289307, "grad_norm": 42.463134765625, "learning_rate": 2.2437845895869825e-07, "logits/chosen": 0.07252983748912811, "logits/rejected": 0.09801161289215088, "logps/chosen": -347.77801513671875, "logps/ref_chosen": -68.99594116210938, "logps/ref_rejected": -88.64665985107422, "logps/rejected": -533.981689453125, "loss": 0.8382, "margin_dpo/margin_mean": 166.55291748046875, "margin_dpo/margin_std": 159.85546875, "step": 396 }, { "epoch": 0.5829662261380323, "fcm_dpo/beta": 0.004994560033082962, "fcm_dpo/delta": -0.10879573971033096, "fcm_dpo/margin": 189.552734375, "fcm_dpo/q_t": 0.30495208501815796, "grad_norm": 57.0134391784668, "learning_rate": 2.2310222927429716e-07, "logits/chosen": 0.1951860934495926, "logits/rejected": 0.20993033051490784, "logps/chosen": -337.2643737792969, "logps/ref_chosen": -61.27716827392578, "logps/ref_rejected": -103.11612701416016, "logps/rejected": -568.656005859375, "loss": 0.7919, "margin_dpo/margin_mean": 189.552734375, "margin_dpo/margin_std": 167.1475372314453, "step": 397 }, { "epoch": 0.5844346549192364, "fcm_dpo/beta": 0.0049682073295116425, "fcm_dpo/delta": -0.03307997062802315, "fcm_dpo/margin": 177.1311798095703, "fcm_dpo/q_t": 0.3241208791732788, "grad_norm": 74.6159896850586, "learning_rate": 2.2182670837116972e-07, "logits/chosen": 0.004022831097245216, "logits/rejected": 0.014983203262090683, "logps/chosen": -374.131103515625, "logps/ref_chosen": -68.15155029296875, "logps/ref_rejected": -108.52360534667969, "logps/rejected": -591.6343994140625, "loss": 0.853, "margin_dpo/margin_mean": 177.13116455078125, "margin_dpo/margin_std": 186.1813201904297, "step": 398 }, { "epoch": 0.5859030837004405, "fcm_dpo/beta": 0.004931504372507334, "fcm_dpo/delta": -0.030671503394842148, "fcm_dpo/margin": 177.97613525390625, "fcm_dpo/q_t": 0.3222649097442627, "grad_norm": 54.23942947387695, "learning_rate": 2.2055192986047804e-07, "logits/chosen": 0.11953572928905487, "logits/rejected": 0.16603019833564758, "logps/chosen": -341.5679931640625, "logps/ref_chosen": -60.889801025390625, "logps/ref_rejected": -77.965576171875, "logps/rejected": -536.619873046875, "loss": 0.8604, "margin_dpo/margin_mean": 177.97613525390625, "margin_dpo/margin_std": 184.6171112060547, "step": 399 }, { "epoch": 0.5873715124816447, "fcm_dpo/beta": 0.004671948961913586, "fcm_dpo/delta": -0.326324462890625, "fcm_dpo/margin": 243.12725830078125, "fcm_dpo/q_t": 0.27325916290283203, "grad_norm": 30.489665985107422, "learning_rate": 2.192779273338215e-07, "logits/chosen": 0.15955214202404022, "logits/rejected": 0.17605489492416382, "logps/chosen": -354.65338134765625, "logps/ref_chosen": -63.64359664916992, "logps/ref_rejected": -105.252685546875, "logps/rejected": -639.3897094726562, "loss": 0.7153, "margin_dpo/margin_mean": 243.12725830078125, "margin_dpo/margin_std": 202.89080810546875, "step": 400 }, { "epoch": 0.5873715124816447, "eval_fcm_dpo/beta": 0.0046274252235889435, "eval_logits/chosen": 0.12993372976779938, "eval_logits/rejected": 0.15252064168453217, "eval_logps/chosen": -440.1346435546875, "eval_logps/ref_chosen": -79.05104064941406, "eval_logps/ref_rejected": -86.79793548583984, "eval_logps/rejected": -563.7314453125, "eval_loss": 0.544421374797821, "eval_margin_dpo/margin_mean": 115.849853515625, "eval_margin_dpo/margin_std": 187.16053771972656, "eval_runtime": 39.2673, "eval_samples_per_second": 59.566, "eval_steps_per_second": 1.885, "step": 400 }, { "epoch": 0.5888399412628488, "fcm_dpo/beta": 0.004744245205074549, "fcm_dpo/delta": 0.1801946461200714, "fcm_dpo/margin": 143.82565307617188, "fcm_dpo/q_t": 0.3702501058578491, "grad_norm": 62.11568069458008, "learning_rate": 2.1800473436235136e-07, "logits/chosen": 0.17814716696739197, "logits/rejected": 0.18520209193229675, "logps/chosen": -379.033447265625, "logps/ref_chosen": -57.16303253173828, "logps/ref_rejected": -83.79249572753906, "logps/rejected": -549.488525390625, "loss": 1.0428, "margin_dpo/margin_mean": 143.8256378173828, "margin_dpo/margin_std": 222.36459350585938, "step": 401 }, { "epoch": 0.5903083700440529, "fcm_dpo/beta": 0.004563698545098305, "fcm_dpo/delta": -0.3387846052646637, "fcm_dpo/margin": 251.93418884277344, "fcm_dpo/q_t": 0.2641059160232544, "grad_norm": 33.36614990234375, "learning_rate": 2.1673238449588665e-07, "logits/chosen": 0.15705952048301697, "logits/rejected": 0.17883501946926117, "logps/chosen": -324.14556884765625, "logps/ref_chosen": -50.74037170410156, "logps/ref_rejected": -81.0460433959961, "logps/rejected": -606.385498046875, "loss": 0.6684, "margin_dpo/margin_mean": 251.9342041015625, "margin_dpo/margin_std": 182.95848083496094, "step": 402 }, { "epoch": 0.591776798825257, "fcm_dpo/beta": 0.0044688875786960125, "fcm_dpo/delta": 0.0052043236792087555, "fcm_dpo/margin": 189.0760498046875, "fcm_dpo/q_t": 0.3240288197994232, "grad_norm": 42.3422966003418, "learning_rate": 2.154609112620295e-07, "logits/chosen": 0.2680968940258026, "logits/rejected": 0.27872592210769653, "logps/chosen": -372.03997802734375, "logps/ref_chosen": -47.14731216430664, "logps/ref_rejected": -77.2666015625, "logps/rejected": -591.2353515625, "loss": 0.8459, "margin_dpo/margin_mean": 189.0760498046875, "margin_dpo/margin_std": 177.94534301757812, "step": 403 }, { "epoch": 0.593245227606461, "fcm_dpo/beta": 0.004463577643036842, "fcm_dpo/delta": -0.029174722731113434, "fcm_dpo/margin": 196.38226318359375, "fcm_dpo/q_t": 0.32246384024620056, "grad_norm": 48.92930603027344, "learning_rate": 2.1419034816528218e-07, "logits/chosen": 0.3021692931652069, "logits/rejected": 0.3231460452079773, "logps/chosen": -387.607666015625, "logps/ref_chosen": -47.875274658203125, "logps/ref_rejected": -77.15499877929688, "logps/rejected": -613.2696533203125, "loss": 0.8722, "margin_dpo/margin_mean": 196.38226318359375, "margin_dpo/margin_std": 209.30274963378906, "step": 404 }, { "epoch": 0.5947136563876652, "fcm_dpo/beta": 0.0045960224233567715, "fcm_dpo/delta": 0.11465553939342499, "fcm_dpo/margin": 160.8705596923828, "fcm_dpo/q_t": 0.3535424470901489, "grad_norm": 59.457088470458984, "learning_rate": 2.129207286861638e-07, "logits/chosen": 0.16498160362243652, "logits/rejected": 0.17874008417129517, "logps/chosen": -429.4807434082031, "logps/ref_chosen": -65.16290283203125, "logps/ref_rejected": -87.18678283691406, "logps/rejected": -612.375244140625, "loss": 0.9583, "margin_dpo/margin_mean": 160.8705596923828, "margin_dpo/margin_std": 197.54196166992188, "step": 405 }, { "epoch": 0.5961820851688693, "fcm_dpo/beta": 0.004600249230861664, "fcm_dpo/delta": -0.017502881586551666, "fcm_dpo/margin": 187.85006713867188, "fcm_dpo/q_t": 0.3287266790866852, "grad_norm": 47.70918655395508, "learning_rate": 2.1165208628032861e-07, "logits/chosen": 0.23692449927330017, "logits/rejected": 0.25910329818725586, "logps/chosen": -408.302001953125, "logps/ref_chosen": -49.740814208984375, "logps/ref_rejected": -92.07862854003906, "logps/rejected": -638.4898681640625, "loss": 0.8766, "margin_dpo/margin_mean": 187.85006713867188, "margin_dpo/margin_std": 203.72189331054688, "step": 406 }, { "epoch": 0.5976505139500734, "fcm_dpo/beta": 0.004692500457167625, "fcm_dpo/delta": 0.1510598063468933, "fcm_dpo/margin": 150.65615844726562, "fcm_dpo/q_t": 0.3553357720375061, "grad_norm": 61.100006103515625, "learning_rate": 2.1038445437768375e-07, "logits/chosen": 0.3105884790420532, "logits/rejected": 0.3434818387031555, "logps/chosen": -443.04608154296875, "logps/ref_chosen": -56.33069610595703, "logps/ref_rejected": -77.51209259033203, "logps/rejected": -614.8836669921875, "loss": 0.979, "margin_dpo/margin_mean": 150.65615844726562, "margin_dpo/margin_std": 192.8316650390625, "step": 407 }, { "epoch": 0.5991189427312775, "fcm_dpo/beta": 0.00478686299175024, "fcm_dpo/delta": 0.14895686507225037, "fcm_dpo/margin": 148.69833374023438, "fcm_dpo/q_t": 0.3522757291793823, "grad_norm": 55.596675872802734, "learning_rate": 2.0911786638150872e-07, "logits/chosen": 0.2758781313896179, "logits/rejected": 0.3151041269302368, "logps/chosen": -463.2239990234375, "logps/ref_chosen": -69.789306640625, "logps/ref_rejected": -90.09693908691406, "logps/rejected": -632.2299194335938, "loss": 0.9397, "margin_dpo/margin_mean": 148.69833374023438, "margin_dpo/margin_std": 172.4322967529297, "step": 408 }, { "epoch": 0.6005873715124816, "fcm_dpo/beta": 0.004894952289760113, "fcm_dpo/delta": 0.08498449623584747, "fcm_dpo/margin": 157.61297607421875, "fcm_dpo/q_t": 0.3446779251098633, "grad_norm": 94.63024139404297, "learning_rate": 2.0785235566757517e-07, "logits/chosen": 0.1406092494726181, "logits/rejected": 0.15482883155345917, "logps/chosen": -447.18670654296875, "logps/ref_chosen": -67.31744384765625, "logps/ref_rejected": -84.904296875, "logps/rejected": -622.386474609375, "loss": 0.907, "margin_dpo/margin_mean": 157.61297607421875, "margin_dpo/margin_std": 178.1060333251953, "step": 409 }, { "epoch": 0.6020558002936858, "fcm_dpo/beta": 0.004944339860230684, "fcm_dpo/delta": 0.0544782392680645, "fcm_dpo/margin": 161.7826690673828, "fcm_dpo/q_t": 0.3345832824707031, "grad_norm": 47.36720657348633, "learning_rate": 2.065879555832674e-07, "logits/chosen": 0.228413388133049, "logits/rejected": 0.23824666440486908, "logps/chosen": -425.039306640625, "logps/ref_chosen": -51.465354919433594, "logps/ref_rejected": -83.198974609375, "logps/rejected": -618.5556030273438, "loss": 0.9026, "margin_dpo/margin_mean": 161.78265380859375, "margin_dpo/margin_std": 179.0279541015625, "step": 410 }, { "epoch": 0.6035242290748899, "fcm_dpo/beta": 0.004880721680819988, "fcm_dpo/delta": -0.02468109130859375, "fcm_dpo/margin": 177.94451904296875, "fcm_dpo/q_t": 0.32889145612716675, "grad_norm": 57.21861267089844, "learning_rate": 2.0532469944670343e-07, "logits/chosen": 0.35708197951316833, "logits/rejected": 0.3739537000656128, "logps/chosen": -432.9054870605469, "logps/ref_chosen": -52.30727005004883, "logps/ref_rejected": -80.69495391845703, "logps/rejected": -639.2376708984375, "loss": 0.8709, "margin_dpo/margin_mean": 177.9445037841797, "margin_dpo/margin_std": 190.1000213623047, "step": 411 }, { "epoch": 0.604992657856094, "fcm_dpo/beta": 0.004977867007255554, "fcm_dpo/delta": 0.04423771798610687, "fcm_dpo/margin": 162.58218383789062, "fcm_dpo/q_t": 0.335659384727478, "grad_norm": 44.3292236328125, "learning_rate": 2.0406262054585738e-07, "logits/chosen": 0.23362818360328674, "logits/rejected": 0.22613993287086487, "logps/chosen": -432.71868896484375, "logps/ref_chosen": -53.144126892089844, "logps/ref_rejected": -100.0608139038086, "logps/rejected": -642.2175903320312, "loss": 0.9013, "margin_dpo/margin_mean": 162.5821990966797, "margin_dpo/margin_std": 181.373779296875, "step": 412 }, { "epoch": 0.6064610866372981, "fcm_dpo/beta": 0.005048343911767006, "fcm_dpo/delta": 0.06283044815063477, "fcm_dpo/margin": 156.84176635742188, "fcm_dpo/q_t": 0.3339266777038574, "grad_norm": 74.36189270019531, "learning_rate": 2.0280175213768205e-07, "logits/chosen": 0.10526064038276672, "logits/rejected": 0.12061142921447754, "logps/chosen": -434.7674255371094, "logps/ref_chosen": -61.58196258544922, "logps/ref_rejected": -99.47340393066406, "logps/rejected": -629.5006103515625, "loss": 0.8827, "margin_dpo/margin_mean": 156.84176635742188, "margin_dpo/margin_std": 160.830322265625, "step": 413 }, { "epoch": 0.6079295154185022, "fcm_dpo/beta": 0.0049922592006623745, "fcm_dpo/delta": -0.13861587643623352, "fcm_dpo/margin": 195.24468994140625, "fcm_dpo/q_t": 0.2993001341819763, "grad_norm": 50.09534454345703, "learning_rate": 2.0154212744723247e-07, "logits/chosen": 0.2176135778427124, "logits/rejected": 0.23482248187065125, "logps/chosen": -374.04327392578125, "logps/ref_chosen": -46.63148498535156, "logps/ref_rejected": -87.64653015136719, "logps/rejected": -610.302978515625, "loss": 0.7763, "margin_dpo/margin_mean": 195.24468994140625, "margin_dpo/margin_std": 165.15785217285156, "step": 414 }, { "epoch": 0.6093979441997063, "fcm_dpo/beta": 0.004968525841832161, "fcm_dpo/delta": 0.11446189880371094, "fcm_dpo/margin": 149.75445556640625, "fcm_dpo/q_t": 0.34472692012786865, "grad_norm": 49.72021484375, "learning_rate": 2.002837796667909e-07, "logits/chosen": 0.085420161485672, "logits/rejected": 0.1004190742969513, "logps/chosen": -415.4371337890625, "logps/ref_chosen": -78.6182861328125, "logps/ref_rejected": -100.47752380371094, "logps/rejected": -587.05078125, "loss": 0.914, "margin_dpo/margin_mean": 149.75445556640625, "margin_dpo/margin_std": 160.1766357421875, "step": 415 }, { "epoch": 0.6108663729809104, "fcm_dpo/beta": 0.004934309050440788, "fcm_dpo/delta": -0.17341846227645874, "fcm_dpo/margin": 203.90164184570312, "fcm_dpo/q_t": 0.28993839025497437, "grad_norm": 51.852142333984375, "learning_rate": 1.990267419549914e-07, "logits/chosen": 0.10152135044336319, "logits/rejected": 0.1114257276058197, "logps/chosen": -357.18414306640625, "logps/ref_chosen": -58.27912521362305, "logps/ref_rejected": -90.56871795654297, "logps/rejected": -593.3753662109375, "loss": 0.7466, "margin_dpo/margin_mean": 203.90164184570312, "margin_dpo/margin_std": 158.02981567382812, "step": 416 }, { "epoch": 0.6123348017621145, "fcm_dpo/beta": 0.004816145170480013, "fcm_dpo/delta": -0.07927640527486801, "fcm_dpo/margin": 191.4312744140625, "fcm_dpo/q_t": 0.3030052185058594, "grad_norm": 42.43336486816406, "learning_rate": 1.9777104743594686e-07, "logits/chosen": 0.23988962173461914, "logits/rejected": 0.2814761996269226, "logps/chosen": -353.5125732421875, "logps/ref_chosen": -50.1987190246582, "logps/ref_rejected": -68.15184020996094, "logps/rejected": -562.89697265625, "loss": 0.7679, "margin_dpo/margin_mean": 191.4312744140625, "margin_dpo/margin_std": 146.1686553955078, "step": 417 }, { "epoch": 0.6138032305433186, "fcm_dpo/beta": 0.004821361042559147, "fcm_dpo/delta": -0.0207485631108284, "fcm_dpo/margin": 179.69488525390625, "fcm_dpo/q_t": 0.32698148488998413, "grad_norm": 37.9013786315918, "learning_rate": 1.965167291983757e-07, "logits/chosen": 0.0310177244246006, "logits/rejected": 0.06632434576749802, "logps/chosen": -372.91217041015625, "logps/ref_chosen": -81.97846984863281, "logps/ref_rejected": -104.69148254394531, "logps/rejected": -575.320068359375, "loss": 0.8727, "margin_dpo/margin_mean": 179.69488525390625, "margin_dpo/margin_std": 190.57337951660156, "step": 418 }, { "epoch": 0.6152716593245228, "fcm_dpo/beta": 0.004708138294517994, "fcm_dpo/delta": -0.051440898329019547, "fcm_dpo/margin": 190.4224090576172, "fcm_dpo/q_t": 0.3137531876564026, "grad_norm": 40.84404373168945, "learning_rate": 1.9526382029472988e-07, "logits/chosen": -0.034985870122909546, "logits/rejected": -0.023904774338006973, "logps/chosen": -312.14996337890625, "logps/ref_chosen": -52.948646545410156, "logps/ref_rejected": -91.58309936523438, "logps/rejected": -541.2068481445312, "loss": 0.8186, "margin_dpo/margin_mean": 190.4224090576172, "margin_dpo/margin_std": 174.2207794189453, "step": 419 }, { "epoch": 0.6167400881057269, "fcm_dpo/beta": 0.004817042499780655, "fcm_dpo/delta": 0.1355183720588684, "fcm_dpo/margin": 150.2655029296875, "fcm_dpo/q_t": 0.3537171483039856, "grad_norm": 53.41609191894531, "learning_rate": 1.9401235374032425e-07, "logits/chosen": 0.02854771912097931, "logits/rejected": 0.07809595763683319, "logps/chosen": -342.8812255859375, "logps/ref_chosen": -77.7699203491211, "logps/ref_rejected": -69.31985473632812, "logps/rejected": -484.6966857910156, "loss": 0.9521, "margin_dpo/margin_mean": 150.2655029296875, "margin_dpo/margin_std": 187.6026611328125, "step": 420 }, { "epoch": 0.618208516886931, "fcm_dpo/beta": 0.004980705678462982, "fcm_dpo/delta": 0.17126131057739258, "fcm_dpo/margin": 138.48049926757812, "fcm_dpo/q_t": 0.35512322187423706, "grad_norm": 50.58280563354492, "learning_rate": 1.9276236251246653e-07, "logits/chosen": -0.004294605925679207, "logits/rejected": 0.012868620455265045, "logps/chosen": -289.5254211425781, "logps/ref_chosen": -53.765865325927734, "logps/ref_rejected": -89.28144836425781, "logps/rejected": -463.521484375, "loss": 0.9536, "margin_dpo/margin_mean": 138.48049926757812, "margin_dpo/margin_std": 160.81411743164062, "step": 421 }, { "epoch": 0.6196769456681351, "fcm_dpo/beta": 0.005074765998870134, "fcm_dpo/delta": 0.07042841613292694, "fcm_dpo/margin": 154.63316345214844, "fcm_dpo/q_t": 0.3365428149700165, "grad_norm": 43.70001983642578, "learning_rate": 1.9151387954958792e-07, "logits/chosen": 0.05924776941537857, "logits/rejected": 0.0849265307188034, "logps/chosen": -328.7698974609375, "logps/ref_chosen": -68.6337661743164, "logps/ref_rejected": -87.86351013183594, "logps/rejected": -502.6328125, "loss": 0.9044, "margin_dpo/margin_mean": 154.6331787109375, "margin_dpo/margin_std": 169.25784301757812, "step": 422 }, { "epoch": 0.6211453744493393, "fcm_dpo/beta": 0.005103256553411484, "fcm_dpo/delta": 0.04598752781748772, "fcm_dpo/margin": 158.27920532226562, "fcm_dpo/q_t": 0.3329399824142456, "grad_norm": 52.14820098876953, "learning_rate": 1.902669377503756e-07, "logits/chosen": 0.04696730524301529, "logits/rejected": 0.05033382400870323, "logps/chosen": -299.9007873535156, "logps/ref_chosen": -54.99030303955078, "logps/ref_rejected": -86.30654907226562, "logps/rejected": -489.4962463378906, "loss": 0.8829, "margin_dpo/margin_mean": 158.27920532226562, "margin_dpo/margin_std": 166.1439971923828, "step": 423 }, { "epoch": 0.6226138032305433, "fcm_dpo/beta": 0.005140416789799929, "fcm_dpo/delta": 0.04995416849851608, "fcm_dpo/margin": 156.386474609375, "fcm_dpo/q_t": 0.3389360010623932, "grad_norm": 62.656551361083984, "learning_rate": 1.890215699729057e-07, "logits/chosen": -0.046080753207206726, "logits/rejected": -0.017214350402355194, "logps/chosen": -261.4268798828125, "logps/ref_chosen": -56.01192092895508, "logps/ref_rejected": -66.47896575927734, "logps/rejected": -428.2803955078125, "loss": 0.9007, "margin_dpo/margin_mean": 156.38648986816406, "margin_dpo/margin_std": 177.43490600585938, "step": 424 }, { "epoch": 0.6240822320117474, "fcm_dpo/beta": 0.005301492288708687, "fcm_dpo/delta": 0.1069888174533844, "fcm_dpo/margin": 141.27867126464844, "fcm_dpo/q_t": 0.3437122702598572, "grad_norm": 44.34437942504883, "learning_rate": 1.8777780903377732e-07, "logits/chosen": 0.0495096892118454, "logits/rejected": 0.04999320209026337, "logps/chosen": -264.53546142578125, "logps/ref_chosen": -46.86899948120117, "logps/ref_rejected": -95.92545318603516, "logps/rejected": -454.87060546875, "loss": 0.9226, "margin_dpo/margin_mean": 141.2786865234375, "margin_dpo/margin_std": 155.26663208007812, "step": 425 }, { "epoch": 0.6255506607929515, "fcm_dpo/beta": 0.005260363221168518, "fcm_dpo/delta": -0.09204407781362534, "fcm_dpo/margin": 177.3383331298828, "fcm_dpo/q_t": 0.3091350197792053, "grad_norm": 76.20099639892578, "learning_rate": 1.8653568770724803e-07, "logits/chosen": -0.04422163963317871, "logits/rejected": 0.00882166437804699, "logps/chosen": -279.64617919921875, "logps/ref_chosen": -76.58354187011719, "logps/ref_rejected": -81.26658630371094, "logps/rejected": -461.66754150390625, "loss": 0.8229, "margin_dpo/margin_mean": 177.33831787109375, "margin_dpo/margin_std": 168.63809204101562, "step": 426 }, { "epoch": 0.6270190895741556, "fcm_dpo/beta": 0.005375740118324757, "fcm_dpo/delta": 0.23773564398288727, "fcm_dpo/margin": 116.69024658203125, "fcm_dpo/q_t": 0.36856818199157715, "grad_norm": 53.63347244262695, "learning_rate": 1.8529523872436977e-07, "logits/chosen": -0.021857187151908875, "logits/rejected": 0.0070018284022808075, "logps/chosen": -265.6365966796875, "logps/ref_chosen": -64.8538818359375, "logps/ref_rejected": -78.5660171508789, "logps/rejected": -396.03900146484375, "loss": 0.9804, "margin_dpo/margin_mean": 116.69024658203125, "margin_dpo/margin_std": 143.55154418945312, "step": 427 }, { "epoch": 0.6284875183553598, "fcm_dpo/beta": 0.00542007852345705, "fcm_dpo/delta": -0.032713260501623154, "fcm_dpo/margin": 162.3301239013672, "fcm_dpo/q_t": 0.3234860599040985, "grad_norm": 43.0552978515625, "learning_rate": 1.8405649477212697e-07, "logits/chosen": -0.028629587963223457, "logits/rejected": -0.013756831176578999, "logps/chosen": -317.6119384765625, "logps/ref_chosen": -62.63666534423828, "logps/ref_rejected": -103.28181457519531, "logps/rejected": -520.5872192382812, "loss": 0.8812, "margin_dpo/margin_mean": 162.3301239013672, "margin_dpo/margin_std": 179.6146240234375, "step": 428 }, { "epoch": 0.6299559471365639, "fcm_dpo/beta": 0.005543345585465431, "fcm_dpo/delta": 0.10069851577281952, "fcm_dpo/margin": 136.0623779296875, "fcm_dpo/q_t": 0.34946388006210327, "grad_norm": 80.5544662475586, "learning_rate": 1.828194884925749e-07, "logits/chosen": 0.011677625589072704, "logits/rejected": 0.06275086104869843, "logps/chosen": -339.78070068359375, "logps/ref_chosen": -81.23401641845703, "logps/ref_rejected": -91.79493713378906, "logps/rejected": -486.40399169921875, "loss": 0.9663, "margin_dpo/margin_mean": 136.0623779296875, "margin_dpo/margin_std": 171.47412109375, "step": 429 }, { "epoch": 0.631424375917768, "fcm_dpo/beta": 0.005538782104849815, "fcm_dpo/delta": -0.005278989672660828, "fcm_dpo/margin": 154.27493286132812, "fcm_dpo/q_t": 0.3279253840446472, "grad_norm": 38.59894561767578, "learning_rate": 1.8158425248197928e-07, "logits/chosen": -0.07191786915063858, "logits/rejected": -0.06347661465406418, "logps/chosen": -284.88165283203125, "logps/ref_chosen": -60.920326232910156, "logps/ref_rejected": -104.42280578613281, "logps/rejected": -482.6590576171875, "loss": 0.8471, "margin_dpo/margin_mean": 154.27493286132812, "margin_dpo/margin_std": 157.42654418945312, "step": 430 }, { "epoch": 0.6328928046989721, "fcm_dpo/beta": 0.005310466047376394, "fcm_dpo/delta": -0.21516066789627075, "fcm_dpo/margin": 196.02615356445312, "fcm_dpo/q_t": 0.28844383358955383, "grad_norm": 46.57950210571289, "learning_rate": 1.8035081928995788e-07, "logits/chosen": 0.06265595555305481, "logits/rejected": 0.0806708112359047, "logps/chosen": -286.49560546875, "logps/ref_chosen": -57.34874725341797, "logps/ref_rejected": -92.84022521972656, "logps/rejected": -518.0132446289062, "loss": 0.7499, "margin_dpo/margin_mean": 196.0261688232422, "margin_dpo/margin_std": 165.00961303710938, "step": 431 }, { "epoch": 0.6343612334801763, "fcm_dpo/beta": 0.005262886174023151, "fcm_dpo/delta": -0.11242011934518814, "fcm_dpo/margin": 180.4857635498047, "fcm_dpo/q_t": 0.3041759133338928, "grad_norm": 33.68878173828125, "learning_rate": 1.791192214186223e-07, "logits/chosen": -0.06168440729379654, "logits/rejected": -0.031240500509738922, "logps/chosen": -296.9165954589844, "logps/ref_chosen": -71.07479095458984, "logps/ref_rejected": -98.57952880859375, "logps/rejected": -504.9071044921875, "loss": 0.7839, "margin_dpo/margin_mean": 180.4857635498047, "margin_dpo/margin_std": 153.55856323242188, "step": 432 }, { "epoch": 0.6358296622613803, "fcm_dpo/beta": 0.0052696773782372475, "fcm_dpo/delta": 0.12958739697933197, "fcm_dpo/margin": 138.4744415283203, "fcm_dpo/q_t": 0.35025665163993835, "grad_norm": 49.32050323486328, "learning_rate": 1.7788949132172193e-07, "logits/chosen": 0.043396446853876114, "logits/rejected": 0.06199165806174278, "logps/chosen": -338.40191650390625, "logps/ref_chosen": -58.273193359375, "logps/ref_rejected": -95.95089721679688, "logps/rejected": -514.5540771484375, "loss": 0.9753, "margin_dpo/margin_mean": 138.4744415283203, "margin_dpo/margin_std": 176.79898071289062, "step": 433 }, { "epoch": 0.6372980910425844, "fcm_dpo/beta": 0.005316519178450108, "fcm_dpo/delta": 0.03797819837927818, "fcm_dpo/margin": 153.3128204345703, "fcm_dpo/q_t": 0.3362678289413452, "grad_norm": 42.75183868408203, "learning_rate": 1.7666166140378853e-07, "logits/chosen": 0.09770476818084717, "logits/rejected": 0.12834317982196808, "logps/chosen": -315.55572509765625, "logps/ref_chosen": -61.97370147705078, "logps/ref_rejected": -78.49861145019531, "logps/rejected": -485.3934631347656, "loss": 0.8842, "margin_dpo/margin_mean": 153.31283569335938, "margin_dpo/margin_std": 169.85568237304688, "step": 434 }, { "epoch": 0.6387665198237885, "fcm_dpo/beta": 0.00526207871735096, "fcm_dpo/delta": -0.09070765972137451, "fcm_dpo/margin": 177.1383056640625, "fcm_dpo/q_t": 0.3105207681655884, "grad_norm": 38.45564270019531, "learning_rate": 1.7543576401928218e-07, "logits/chosen": 0.17958560585975647, "logits/rejected": 0.21033020317554474, "logps/chosen": -298.01806640625, "logps/ref_chosen": -51.502052307128906, "logps/ref_rejected": -87.56689453125, "logps/rejected": -511.22119140625, "loss": 0.8444, "margin_dpo/margin_mean": 177.13829040527344, "margin_dpo/margin_std": 178.4004669189453, "step": 435 }, { "epoch": 0.6402349486049926, "fcm_dpo/beta": 0.005174779333174229, "fcm_dpo/delta": -0.07778394967317581, "fcm_dpo/margin": 177.8702392578125, "fcm_dpo/q_t": 0.31033855676651, "grad_norm": 66.37271118164062, "learning_rate": 1.742118314717391e-07, "logits/chosen": -0.01764148473739624, "logits/rejected": 0.029303671792149544, "logps/chosen": -323.1204833984375, "logps/ref_chosen": -71.40371704101562, "logps/ref_rejected": -82.72775268554688, "logps/rejected": -512.3147583007812, "loss": 0.8009, "margin_dpo/margin_mean": 177.8702392578125, "margin_dpo/margin_std": 161.47940063476562, "step": 436 }, { "epoch": 0.6417033773861968, "fcm_dpo/beta": 0.005187171511352062, "fcm_dpo/delta": 0.008135082200169563, "fcm_dpo/margin": 162.37954711914062, "fcm_dpo/q_t": 0.32683050632476807, "grad_norm": 44.145938873291016, "learning_rate": 1.7298989601292036e-07, "logits/chosen": 0.0989217460155487, "logits/rejected": 0.1424962878227234, "logps/chosen": -336.16766357421875, "logps/ref_chosen": -64.7442626953125, "logps/ref_rejected": -82.04356384277344, "logps/rejected": -515.8464965820312, "loss": 0.8751, "margin_dpo/margin_mean": 162.37954711914062, "margin_dpo/margin_std": 169.3526611328125, "step": 437 }, { "epoch": 0.6431718061674009, "fcm_dpo/beta": 0.005006927065551281, "fcm_dpo/delta": -0.15296681225299835, "fcm_dpo/margin": 196.67837524414062, "fcm_dpo/q_t": 0.29662394523620605, "grad_norm": 44.88563537597656, "learning_rate": 1.7176998984196144e-07, "logits/chosen": 0.14420339465141296, "logits/rejected": 0.1939929574728012, "logps/chosen": -323.58074951171875, "logps/ref_chosen": -59.0186653137207, "logps/ref_rejected": -83.07682800292969, "logps/rejected": -544.3172607421875, "loss": 0.7626, "margin_dpo/margin_mean": 196.67837524414062, "margin_dpo/margin_std": 161.43214416503906, "step": 438 }, { "epoch": 0.644640234948605, "fcm_dpo/beta": 0.005011453293263912, "fcm_dpo/delta": 0.08284294605255127, "fcm_dpo/margin": 153.88717651367188, "fcm_dpo/q_t": 0.34648653864860535, "grad_norm": 50.955806732177734, "learning_rate": 1.7055214510452458e-07, "logits/chosen": 0.1353827714920044, "logits/rejected": 0.14328114688396454, "logps/chosen": -347.3162536621094, "logps/ref_chosen": -53.78407669067383, "logps/ref_rejected": -83.98545837402344, "logps/rejected": -531.40478515625, "loss": 0.9245, "margin_dpo/margin_mean": 153.88717651367188, "margin_dpo/margin_std": 174.40078735351562, "step": 439 }, { "epoch": 0.6461086637298091, "fcm_dpo/beta": 0.0051966458559036255, "fcm_dpo/delta": 0.10422302782535553, "fcm_dpo/margin": 144.91168212890625, "fcm_dpo/q_t": 0.3527141511440277, "grad_norm": 75.77989196777344, "learning_rate": 1.6933639389195134e-07, "logits/chosen": 0.07095863670110703, "logits/rejected": 0.11175039410591125, "logps/chosen": -385.72686767578125, "logps/ref_chosen": -78.56671905517578, "logps/ref_rejected": -96.49775695800781, "logps/rejected": -548.569580078125, "loss": 0.9497, "margin_dpo/margin_mean": 144.91168212890625, "margin_dpo/margin_std": 183.58206176757812, "step": 440 }, { "epoch": 0.6475770925110133, "fcm_dpo/beta": 0.005188990384340286, "fcm_dpo/delta": -0.04949396103620529, "fcm_dpo/margin": 172.47109985351562, "fcm_dpo/q_t": 0.3268754184246063, "grad_norm": 39.17652893066406, "learning_rate": 1.681227682404166e-07, "logits/chosen": 0.097470723092556, "logits/rejected": 0.1150805875658989, "logps/chosen": -390.5365295410156, "logps/ref_chosen": -60.824440002441406, "logps/ref_rejected": -96.47080993652344, "logps/rejected": -598.6539916992188, "loss": 0.9063, "margin_dpo/margin_mean": 172.47109985351562, "margin_dpo/margin_std": 205.70474243164062, "step": 441 }, { "epoch": 0.6490455212922174, "fcm_dpo/beta": 0.005070436745882034, "fcm_dpo/delta": -0.18530398607254028, "fcm_dpo/margin": 200.40408325195312, "fcm_dpo/q_t": 0.3012937307357788, "grad_norm": 38.1731071472168, "learning_rate": 1.669113001300851e-07, "logits/chosen": 0.060047049075365067, "logits/rejected": 0.07558459043502808, "logps/chosen": -342.12982177734375, "logps/ref_chosen": -47.01121520996094, "logps/ref_rejected": -76.53926086425781, "logps/rejected": -572.0619506835938, "loss": 0.788, "margin_dpo/margin_mean": 200.40408325195312, "margin_dpo/margin_std": 188.1474151611328, "step": 442 }, { "epoch": 0.6505139500734214, "fcm_dpo/beta": 0.0050366222858428955, "fcm_dpo/delta": 0.1421085149049759, "fcm_dpo/margin": 142.5586700439453, "fcm_dpo/q_t": 0.35665544867515564, "grad_norm": 53.045677185058594, "learning_rate": 1.6570202148426815e-07, "logits/chosen": 0.09138190746307373, "logits/rejected": 0.10863150656223297, "logps/chosen": -402.77294921875, "logps/ref_chosen": -71.27301788330078, "logps/ref_rejected": -86.679931640625, "logps/rejected": -560.738525390625, "loss": 0.9873, "margin_dpo/margin_mean": 142.55868530273438, "margin_dpo/margin_std": 193.21839904785156, "step": 443 }, { "epoch": 0.6519823788546255, "fcm_dpo/beta": 0.004996387287974358, "fcm_dpo/delta": -0.14721286296844482, "fcm_dpo/margin": 196.6611785888672, "fcm_dpo/q_t": 0.3072533905506134, "grad_norm": 48.08612823486328, "learning_rate": 1.6449496416858282e-07, "logits/chosen": 0.18914419412612915, "logits/rejected": 0.20519858598709106, "logps/chosen": -396.1791076660156, "logps/ref_chosen": -57.213706970214844, "logps/ref_rejected": -97.25489807128906, "logps/rejected": -632.8814697265625, "loss": 0.8182, "margin_dpo/margin_mean": 196.66119384765625, "margin_dpo/margin_std": 203.6089630126953, "step": 444 }, { "epoch": 0.6534508076358296, "fcm_dpo/beta": 0.0048555051907896996, "fcm_dpo/delta": -0.1426890343427658, "fcm_dpo/margin": 201.54562377929688, "fcm_dpo/q_t": 0.3039454221725464, "grad_norm": 74.20765686035156, "learning_rate": 1.6329015999011182e-07, "logits/chosen": 0.11190081387758255, "logits/rejected": 0.13852503895759583, "logps/chosen": -388.1588134765625, "logps/ref_chosen": -67.29979705810547, "logps/ref_rejected": -92.68267059326172, "logps/rejected": -615.0872802734375, "loss": 0.7866, "margin_dpo/margin_mean": 201.54562377929688, "margin_dpo/margin_std": 183.888427734375, "step": 445 }, { "epoch": 0.6549192364170338, "fcm_dpo/beta": 0.004782508127391338, "fcm_dpo/delta": -0.11369301378726959, "fcm_dpo/margin": 199.1475830078125, "fcm_dpo/q_t": 0.30260688066482544, "grad_norm": 52.3062858581543, "learning_rate": 1.6208764069656578e-07, "logits/chosen": 0.16620711982250214, "logits/rejected": 0.1732136458158493, "logps/chosen": -367.4357604980469, "logps/ref_chosen": -59.098487854003906, "logps/ref_rejected": -101.26419067382812, "logps/rejected": -608.7490844726562, "loss": 0.7721, "margin_dpo/margin_mean": 199.1475830078125, "margin_dpo/margin_std": 163.44342041015625, "step": 446 }, { "epoch": 0.6563876651982379, "fcm_dpo/beta": 0.004636620171368122, "fcm_dpo/delta": -0.07215973734855652, "fcm_dpo/margin": 197.22711181640625, "fcm_dpo/q_t": 0.3173784613609314, "grad_norm": 45.591400146484375, "learning_rate": 1.608874379754465e-07, "logits/chosen": 0.1511530876159668, "logits/rejected": 0.15351176261901855, "logps/chosen": -388.9348449707031, "logps/ref_chosen": -56.07533264160156, "logps/ref_rejected": -98.69475555419922, "logps/rejected": -628.7813720703125, "loss": 0.8258, "margin_dpo/margin_mean": 197.22711181640625, "margin_dpo/margin_std": 203.49270629882812, "step": 447 }, { "epoch": 0.657856093979442, "fcm_dpo/beta": 0.004568365402519703, "fcm_dpo/delta": -0.14031019806861877, "fcm_dpo/margin": 213.8214874267578, "fcm_dpo/q_t": 0.3027946352958679, "grad_norm": 43.747013092041016, "learning_rate": 1.5968958345321177e-07, "logits/chosen": 0.1592235565185547, "logits/rejected": 0.17409782111644745, "logps/chosen": -417.6317138671875, "logps/ref_chosen": -60.00384521484375, "logps/ref_rejected": -102.26465606689453, "logps/rejected": -673.7140502929688, "loss": 0.7801, "margin_dpo/margin_mean": 213.82147216796875, "margin_dpo/margin_std": 191.90713500976562, "step": 448 }, { "epoch": 0.6593245227606461, "fcm_dpo/beta": 0.00449769850820303, "fcm_dpo/delta": -0.014355389401316643, "fcm_dpo/margin": 191.867919921875, "fcm_dpo/q_t": 0.3324083089828491, "grad_norm": 48.77326202392578, "learning_rate": 1.584941086944423e-07, "logits/chosen": 0.10339124500751495, "logits/rejected": 0.13415296375751495, "logps/chosen": -414.3675537109375, "logps/ref_chosen": -67.52661895751953, "logps/ref_rejected": -88.59690856933594, "logps/rejected": -627.3057861328125, "loss": 0.9054, "margin_dpo/margin_mean": 191.867919921875, "margin_dpo/margin_std": 233.46246337890625, "step": 449 }, { "epoch": 0.6607929515418502, "fcm_dpo/beta": 0.004431567154824734, "fcm_dpo/delta": -0.14691829681396484, "fcm_dpo/margin": 221.7427520751953, "fcm_dpo/q_t": 0.2959328293800354, "grad_norm": 35.59035873413086, "learning_rate": 1.573010452010098e-07, "logits/chosen": 0.13303320109844208, "logits/rejected": 0.14206843078136444, "logps/chosen": -363.13812255859375, "logps/ref_chosen": -57.10811996459961, "logps/ref_rejected": -102.75494384765625, "logps/rejected": -630.5277099609375, "loss": 0.7723, "margin_dpo/margin_mean": 221.74276733398438, "margin_dpo/margin_std": 192.90106201171875, "step": 450 }, { "epoch": 0.6622613803230544, "fcm_dpo/beta": 0.004432865884155035, "fcm_dpo/delta": 0.047808244824409485, "fcm_dpo/margin": 181.61512756347656, "fcm_dpo/q_t": 0.3292296528816223, "grad_norm": 51.70342254638672, "learning_rate": 1.5611042441124687e-07, "logits/chosen": 0.31042128801345825, "logits/rejected": 0.3536532521247864, "logps/chosen": -417.8612365722656, "logps/ref_chosen": -58.46883010864258, "logps/ref_rejected": -72.92941284179688, "logps/rejected": -613.9369506835938, "loss": 0.9141, "margin_dpo/margin_mean": 181.61514282226562, "margin_dpo/margin_std": 208.4745330810547, "step": 451 }, { "epoch": 0.6637298091042585, "fcm_dpo/beta": 0.004379277117550373, "fcm_dpo/delta": -0.04874721169471741, "fcm_dpo/margin": 204.20883178710938, "fcm_dpo/q_t": 0.3148379921913147, "grad_norm": 42.03379440307617, "learning_rate": 1.549222776991186e-07, "logits/chosen": 0.17657610774040222, "logits/rejected": 0.17502948641777039, "logps/chosen": -340.4559326171875, "logps/ref_chosen": -50.39055252075195, "logps/ref_rejected": -97.77142333984375, "logps/rejected": -592.045654296875, "loss": 0.8076, "margin_dpo/margin_mean": 204.20883178710938, "margin_dpo/margin_std": 183.80862426757812, "step": 452 }, { "epoch": 0.6651982378854625, "fcm_dpo/beta": 0.004408560693264008, "fcm_dpo/delta": 0.09181352704763412, "fcm_dpo/margin": 173.53721618652344, "fcm_dpo/q_t": 0.34526383876800537, "grad_norm": 42.1302490234375, "learning_rate": 1.5373663637339584e-07, "logits/chosen": 0.09675667434930801, "logits/rejected": 0.1360025852918625, "logps/chosen": -382.1043701171875, "logps/ref_chosen": -57.71485137939453, "logps/ref_rejected": -82.20741271972656, "logps/rejected": -580.1341552734375, "loss": 0.9098, "margin_dpo/margin_mean": 173.53721618652344, "margin_dpo/margin_std": 197.82093811035156, "step": 453 }, { "epoch": 0.6666666666666666, "fcm_dpo/beta": 0.004366679582744837, "fcm_dpo/delta": -0.10246938467025757, "fcm_dpo/margin": 215.6974334716797, "fcm_dpo/q_t": 0.3123183846473694, "grad_norm": 42.43370819091797, "learning_rate": 1.5255353167683017e-07, "logits/chosen": 0.2050754278898239, "logits/rejected": 0.22980359196662903, "logps/chosen": -414.22125244140625, "logps/ref_chosen": -60.945648193359375, "logps/ref_rejected": -84.95079040527344, "logps/rejected": -653.923828125, "loss": 0.8234, "margin_dpo/margin_mean": 215.6974334716797, "margin_dpo/margin_std": 221.3548583984375, "step": 454 }, { "epoch": 0.6681350954478708, "fcm_dpo/beta": 0.0042896876111626625, "fcm_dpo/delta": -0.17381757497787476, "fcm_dpo/margin": 234.59808349609375, "fcm_dpo/q_t": 0.29876023530960083, "grad_norm": 34.73609924316406, "learning_rate": 1.5137299478533064e-07, "logits/chosen": 0.09607726335525513, "logits/rejected": 0.10077625513076782, "logps/chosen": -362.557861328125, "logps/ref_chosen": -44.88671112060547, "logps/ref_rejected": -115.30147552490234, "logps/rejected": -667.5707397460938, "loss": 0.778, "margin_dpo/margin_mean": 234.59808349609375, "margin_dpo/margin_std": 209.17662048339844, "step": 455 }, { "epoch": 0.6696035242290749, "fcm_dpo/beta": 0.0041549326851964, "fcm_dpo/delta": -0.10536178946495056, "fcm_dpo/margin": 227.55477905273438, "fcm_dpo/q_t": 0.3024177551269531, "grad_norm": 29.933616638183594, "learning_rate": 1.5019505680714232e-07, "logits/chosen": 0.11028112471103668, "logits/rejected": 0.1089896708726883, "logps/chosen": -378.89727783203125, "logps/ref_chosen": -57.036781311035156, "logps/ref_rejected": -105.21784210205078, "logps/rejected": -654.633056640625, "loss": 0.7617, "margin_dpo/margin_mean": 227.55477905273438, "margin_dpo/margin_std": 182.1309051513672, "step": 456 }, { "epoch": 0.671071953010279, "fcm_dpo/beta": 0.0040961261838674545, "fcm_dpo/delta": -0.016104087233543396, "fcm_dpo/margin": 210.951171875, "fcm_dpo/q_t": 0.31706666946411133, "grad_norm": 50.435768127441406, "learning_rate": 1.4901974878202627e-07, "logits/chosen": 0.11340844631195068, "logits/rejected": 0.15247830748558044, "logps/chosen": -358.2364501953125, "logps/ref_chosen": -54.24253845214844, "logps/ref_rejected": -85.10956573486328, "logps/rejected": -600.0546264648438, "loss": 0.8288, "margin_dpo/margin_mean": 210.951171875, "margin_dpo/margin_std": 189.23641967773438, "step": 457 }, { "epoch": 0.6725403817914831, "fcm_dpo/beta": 0.004061573650687933, "fcm_dpo/delta": -0.05085879936814308, "fcm_dpo/margin": 220.47543334960938, "fcm_dpo/q_t": 0.3158029317855835, "grad_norm": 33.672279357910156, "learning_rate": 1.4784710168044212e-07, "logits/chosen": 0.1747414469718933, "logits/rejected": 0.1942700743675232, "logps/chosen": -371.38616943359375, "logps/ref_chosen": -55.40888214111328, "logps/ref_rejected": -97.68325805664062, "logps/rejected": -634.135986328125, "loss": 0.819, "margin_dpo/margin_mean": 220.47544860839844, "margin_dpo/margin_std": 210.21676635742188, "step": 458 }, { "epoch": 0.6740088105726872, "fcm_dpo/beta": 0.004079152829945087, "fcm_dpo/delta": 0.02714114636182785, "fcm_dpo/margin": 202.2575225830078, "fcm_dpo/q_t": 0.328235924243927, "grad_norm": 34.5672607421875, "learning_rate": 1.466771464027316e-07, "logits/chosen": 0.1831570863723755, "logits/rejected": 0.19782592356204987, "logps/chosen": -396.4834289550781, "logps/ref_chosen": -46.55748748779297, "logps/ref_rejected": -86.16854095458984, "logps/rejected": -638.3519897460938, "loss": 0.8566, "margin_dpo/margin_mean": 202.2575225830078, "margin_dpo/margin_std": 194.767578125, "step": 459 }, { "epoch": 0.6754772393538914, "fcm_dpo/beta": 0.004061030223965645, "fcm_dpo/delta": -0.057284481823444366, "fcm_dpo/margin": 222.14083862304688, "fcm_dpo/q_t": 0.3124066889286041, "grad_norm": 40.84368133544922, "learning_rate": 1.4550991377830423e-07, "logits/chosen": 0.12749359011650085, "logits/rejected": 0.12204212695360184, "logps/chosen": -421.6571044921875, "logps/ref_chosen": -51.63489532470703, "logps/ref_rejected": -104.11935424804688, "logps/rejected": -696.2824096679688, "loss": 0.809, "margin_dpo/margin_mean": 222.14083862304688, "margin_dpo/margin_std": 199.95516967773438, "step": 460 }, { "epoch": 0.6769456681350955, "fcm_dpo/beta": 0.004167584702372551, "fcm_dpo/delta": 0.1559598743915558, "fcm_dpo/margin": 168.80178833007812, "fcm_dpo/q_t": 0.35687270760536194, "grad_norm": 46.724700927734375, "learning_rate": 1.4434543456482518e-07, "logits/chosen": 0.15208810567855835, "logits/rejected": 0.16312140226364136, "logps/chosen": -449.5841369628906, "logps/ref_chosen": -55.18195724487305, "logps/ref_rejected": -86.47689819335938, "logps/rejected": -649.680908203125, "loss": 0.9531, "margin_dpo/margin_mean": 168.80178833007812, "margin_dpo/margin_std": 204.61175537109375, "step": 461 }, { "epoch": 0.6784140969162996, "fcm_dpo/beta": 0.004273770377039909, "fcm_dpo/delta": 0.17702484130859375, "fcm_dpo/margin": 160.3841552734375, "fcm_dpo/q_t": 0.35806363821029663, "grad_norm": 62.64442825317383, "learning_rate": 1.4318373944740484e-07, "logits/chosen": 0.17464204132556915, "logits/rejected": 0.1923956274986267, "logps/chosen": -459.37158203125, "logps/ref_chosen": -69.92803192138672, "logps/ref_rejected": -78.84111022949219, "logps/rejected": -628.6688232421875, "loss": 0.9393, "margin_dpo/margin_mean": 160.3841552734375, "margin_dpo/margin_std": 182.3723602294922, "step": 462 }, { "epoch": 0.6798825256975036, "fcm_dpo/beta": 0.004382624290883541, "fcm_dpo/delta": 0.026647619903087616, "fcm_dpo/margin": 187.98629760742188, "fcm_dpo/q_t": 0.3361142873764038, "grad_norm": 109.82402801513672, "learning_rate": 1.4202485903778976e-07, "logits/chosen": 0.15215685963630676, "logits/rejected": 0.16740122437477112, "logps/chosen": -449.4383850097656, "logps/ref_chosen": -55.27437210083008, "logps/ref_rejected": -89.02497863769531, "logps/rejected": -671.17529296875, "loss": 0.8995, "margin_dpo/margin_mean": 187.98629760742188, "margin_dpo/margin_std": 210.312255859375, "step": 463 }, { "epoch": 0.6813509544787077, "fcm_dpo/beta": 0.004141380079090595, "fcm_dpo/delta": -0.27502915263175964, "fcm_dpo/margin": 263.5333557128906, "fcm_dpo/q_t": 0.2828797698020935, "grad_norm": 56.10831832885742, "learning_rate": 1.4086882387355658e-07, "logits/chosen": 0.16000524163246155, "logits/rejected": 0.16301248967647552, "logps/chosen": -435.59814453125, "logps/ref_chosen": -50.91230010986328, "logps/ref_rejected": -102.4893798828125, "logps/rejected": -750.7085571289062, "loss": 0.7456, "margin_dpo/margin_mean": 263.5333557128906, "margin_dpo/margin_std": 231.73806762695312, "step": 464 }, { "epoch": 0.6828193832599119, "fcm_dpo/beta": 0.0040384442545473576, "fcm_dpo/delta": -0.1352299153804779, "fcm_dpo/margin": 240.72128295898438, "fcm_dpo/q_t": 0.2990262508392334, "grad_norm": 49.06411361694336, "learning_rate": 1.3971566441730714e-07, "logits/chosen": 0.1766095906496048, "logits/rejected": 0.183375284075737, "logps/chosen": -446.39813232421875, "logps/ref_chosen": -60.116851806640625, "logps/ref_rejected": -113.94602966308594, "logps/rejected": -740.9486083984375, "loss": 0.7733, "margin_dpo/margin_mean": 240.72129821777344, "margin_dpo/margin_std": 206.07766723632812, "step": 465 }, { "epoch": 0.684287812041116, "fcm_dpo/beta": 0.004008342977613211, "fcm_dpo/delta": 0.03794410452246666, "fcm_dpo/margin": 203.26609802246094, "fcm_dpo/q_t": 0.33106571435928345, "grad_norm": 64.88581848144531, "learning_rate": 1.3856541105586545e-07, "logits/chosen": 0.15662892162799835, "logits/rejected": 0.16130861639976501, "logps/chosen": -469.90032958984375, "logps/ref_chosen": -52.920921325683594, "logps/ref_rejected": -90.3154296875, "logps/rejected": -710.5609130859375, "loss": 0.8857, "margin_dpo/margin_mean": 203.26611328125, "margin_dpo/margin_std": 214.98695373535156, "step": 466 }, { "epoch": 0.6857562408223201, "fcm_dpo/beta": 0.003955831751227379, "fcm_dpo/delta": -0.06096991151571274, "fcm_dpo/margin": 228.0748291015625, "fcm_dpo/q_t": 0.32432740926742554, "grad_norm": 67.74291229248047, "learning_rate": 1.3741809409947729e-07, "logits/chosen": 0.20276635885238647, "logits/rejected": 0.22354087233543396, "logps/chosen": -561.0784912109375, "logps/ref_chosen": -78.7158203125, "logps/ref_rejected": -102.86019897460938, "logps/rejected": -813.2977294921875, "loss": 0.8992, "margin_dpo/margin_mean": 228.0748291015625, "margin_dpo/margin_std": 262.29058837890625, "step": 467 }, { "epoch": 0.6872246696035242, "fcm_dpo/beta": 0.003934096544981003, "fcm_dpo/delta": -0.05663018673658371, "fcm_dpo/margin": 228.93344116210938, "fcm_dpo/q_t": 0.3182372748851776, "grad_norm": 49.3443717956543, "learning_rate": 1.362737437810114e-07, "logits/chosen": 0.13819798827171326, "logits/rejected": 0.17619337141513824, "logps/chosen": -467.938720703125, "logps/ref_chosen": -69.93536376953125, "logps/ref_rejected": -101.02880859375, "logps/rejected": -727.965576171875, "loss": 0.8409, "margin_dpo/margin_mean": 228.93344116210938, "margin_dpo/margin_std": 230.77337646484375, "step": 468 }, { "epoch": 0.6886930983847284, "fcm_dpo/beta": 0.0038513573817908764, "fcm_dpo/delta": -0.11148861795663834, "fcm_dpo/margin": 246.43438720703125, "fcm_dpo/q_t": 0.30608806014060974, "grad_norm": 49.65244674682617, "learning_rate": 1.351323902551631e-07, "logits/chosen": 0.12890158593654633, "logits/rejected": 0.16653740406036377, "logps/chosen": -514.1097412109375, "logps/ref_chosen": -68.12469482421875, "logps/ref_rejected": -104.78640747070312, "logps/rejected": -797.205810546875, "loss": 0.7908, "margin_dpo/margin_mean": 246.43438720703125, "margin_dpo/margin_std": 220.98977661132812, "step": 469 }, { "epoch": 0.6901615271659325, "fcm_dpo/beta": 0.0038081035017967224, "fcm_dpo/delta": -0.07959811389446259, "fcm_dpo/margin": 242.1552734375, "fcm_dpo/q_t": 0.3072653114795685, "grad_norm": 37.78285217285156, "learning_rate": 1.339940635976592e-07, "logits/chosen": 0.19035938382148743, "logits/rejected": 0.20564529299736023, "logps/chosen": -424.4998779296875, "logps/ref_chosen": -43.791927337646484, "logps/ref_rejected": -82.70285034179688, "logps/rejected": -705.5660400390625, "loss": 0.8099, "margin_dpo/margin_mean": 242.15525817871094, "margin_dpo/margin_std": 222.6582489013672, "step": 470 }, { "epoch": 0.6916299559471366, "fcm_dpo/beta": 0.003821744117885828, "fcm_dpo/delta": 0.09349526464939117, "fcm_dpo/margin": 199.71197509765625, "fcm_dpo/q_t": 0.3398565649986267, "grad_norm": 52.308223724365234, "learning_rate": 1.3285879380446563e-07, "logits/chosen": 0.16461865603923798, "logits/rejected": 0.1858828216791153, "logps/chosen": -481.19781494140625, "logps/ref_chosen": -63.33952331542969, "logps/ref_rejected": -83.61048126220703, "logps/rejected": -701.1807250976562, "loss": 0.8859, "margin_dpo/margin_mean": 199.71197509765625, "margin_dpo/margin_std": 196.94680786132812, "step": 471 }, { "epoch": 0.6930983847283406, "fcm_dpo/beta": 0.003807269735261798, "fcm_dpo/delta": -0.10628817975521088, "fcm_dpo/margin": 248.5273895263672, "fcm_dpo/q_t": 0.3150752782821655, "grad_norm": 34.88620376586914, "learning_rate": 1.317266107909975e-07, "logits/chosen": -0.0007737856358289719, "logits/rejected": 0.03774175047874451, "logps/chosen": -490.617919921875, "logps/ref_chosen": -83.66610717773438, "logps/ref_rejected": -117.20919799804688, "logps/rejected": -772.6884155273438, "loss": 0.8165, "margin_dpo/margin_mean": 248.5273895263672, "margin_dpo/margin_std": 252.5367431640625, "step": 472 }, { "epoch": 0.6945668135095447, "fcm_dpo/beta": 0.0038944047410041094, "fcm_dpo/delta": 0.18922805786132812, "fcm_dpo/margin": 172.9485321044922, "fcm_dpo/q_t": 0.362824410200119, "grad_norm": 64.09027099609375, "learning_rate": 1.3059754439133002e-07, "logits/chosen": 0.20194607973098755, "logits/rejected": 0.22875045239925385, "logps/chosen": -531.016845703125, "logps/ref_chosen": -63.49696731567383, "logps/ref_rejected": -81.14657592773438, "logps/rejected": -721.614990234375, "loss": 1.0217, "margin_dpo/margin_mean": 172.9485321044922, "margin_dpo/margin_std": 254.22259521484375, "step": 473 }, { "epoch": 0.6960352422907489, "fcm_dpo/beta": 0.0040451884269714355, "fcm_dpo/delta": 0.13111956417560577, "fcm_dpo/margin": 179.43878173828125, "fcm_dpo/q_t": 0.3505546450614929, "grad_norm": 61.069889068603516, "learning_rate": 1.2947162435741277e-07, "logits/chosen": 0.12781840562820435, "logits/rejected": 0.14249897003173828, "logps/chosen": -453.209228515625, "logps/ref_chosen": -52.6119384765625, "logps/ref_rejected": -90.08041381835938, "logps/rejected": -670.116455078125, "loss": 0.9636, "margin_dpo/margin_mean": 179.43878173828125, "margin_dpo/margin_std": 223.81166076660156, "step": 474 }, { "epoch": 0.697503671071953, "fcm_dpo/beta": 0.003963591530919075, "fcm_dpo/delta": -0.11174008250236511, "fcm_dpo/margin": 239.89425659179688, "fcm_dpo/q_t": 0.303996205329895, "grad_norm": 83.10292053222656, "learning_rate": 1.2834888035828596e-07, "logits/chosen": 0.24623063206672668, "logits/rejected": 0.2509077191352844, "logps/chosen": -415.0943298339844, "logps/ref_chosen": -42.49519348144531, "logps/ref_rejected": -90.06294250488281, "logps/rejected": -702.556396484375, "loss": 0.7845, "margin_dpo/margin_mean": 239.89425659179688, "margin_dpo/margin_std": 210.8565673828125, "step": 475 }, { "epoch": 0.6989720998531571, "fcm_dpo/beta": 0.003959137946367264, "fcm_dpo/delta": 0.025470474734902382, "fcm_dpo/margin": 208.79278564453125, "fcm_dpo/q_t": 0.32757434248924255, "grad_norm": 55.3691520690918, "learning_rate": 1.2722934197929802e-07, "logits/chosen": 0.18537373840808868, "logits/rejected": 0.20475903153419495, "logps/chosen": -407.0469970703125, "logps/ref_chosen": -42.94938278198242, "logps/ref_rejected": -73.71023559570312, "logps/rejected": -646.6005859375, "loss": 0.847, "margin_dpo/margin_mean": 208.7927703857422, "margin_dpo/margin_std": 197.02679443359375, "step": 476 }, { "epoch": 0.7004405286343612, "fcm_dpo/beta": 0.004016070161014795, "fcm_dpo/delta": 0.07374633848667145, "fcm_dpo/margin": 194.69967651367188, "fcm_dpo/q_t": 0.3365083634853363, "grad_norm": 47.38435363769531, "learning_rate": 1.2611303872132631e-07, "logits/chosen": 0.12200789898633957, "logits/rejected": 0.1759755164384842, "logps/chosen": -456.8595275878906, "logps/ref_chosen": -70.77261352539062, "logps/ref_rejected": -76.13737487792969, "logps/rejected": -656.9239501953125, "loss": 0.903, "margin_dpo/margin_mean": 194.69967651367188, "margin_dpo/margin_std": 213.22210693359375, "step": 477 }, { "epoch": 0.7019089574155654, "fcm_dpo/beta": 0.004027483984827995, "fcm_dpo/delta": -0.03790191560983658, "fcm_dpo/margin": 219.52687072753906, "fcm_dpo/q_t": 0.31887826323509216, "grad_norm": 45.00698471069336, "learning_rate": 1.2500000000000005e-07, "logits/chosen": 0.14754655957221985, "logits/rejected": 0.16361096501350403, "logps/chosen": -366.9229736328125, "logps/ref_chosen": -41.440513610839844, "logps/ref_rejected": -85.36196899414062, "logps/rejected": -630.3712768554688, "loss": 0.8133, "margin_dpo/margin_mean": 219.52688598632812, "margin_dpo/margin_std": 199.22171020507812, "step": 478 }, { "epoch": 0.7033773861967695, "fcm_dpo/beta": 0.004113093018531799, "fcm_dpo/delta": 0.10552472621202469, "fcm_dpo/margin": 181.81784057617188, "fcm_dpo/q_t": 0.339580774307251, "grad_norm": 50.76175308227539, "learning_rate": 1.2389025514492456e-07, "logits/chosen": 0.15951776504516602, "logits/rejected": 0.168975830078125, "logps/chosen": -449.15814208984375, "logps/ref_chosen": -53.907920837402344, "logps/ref_rejected": -95.1163330078125, "logps/rejected": -672.1844482421875, "loss": 0.9043, "margin_dpo/margin_mean": 181.81784057617188, "margin_dpo/margin_std": 180.60443115234375, "step": 479 }, { "epoch": 0.7048458149779736, "fcm_dpo/beta": 0.0042091310024261475, "fcm_dpo/delta": 0.14673718810081482, "fcm_dpo/margin": 168.90711975097656, "fcm_dpo/q_t": 0.35500940680503845, "grad_norm": 72.191650390625, "learning_rate": 1.227838333989088e-07, "logits/chosen": 0.20184411108493805, "logits/rejected": 0.23821671307086945, "logps/chosen": -465.4715576171875, "logps/ref_chosen": -58.682701110839844, "logps/ref_rejected": -82.93248748779297, "logps/rejected": -658.6284790039062, "loss": 0.9739, "margin_dpo/margin_mean": 168.90713500976562, "margin_dpo/margin_std": 211.50201416015625, "step": 480 }, { "epoch": 0.7063142437591777, "fcm_dpo/beta": 0.0041699884459376335, "fcm_dpo/delta": -0.032555945217609406, "fcm_dpo/margin": 210.88694763183594, "fcm_dpo/q_t": 0.3185991644859314, "grad_norm": 53.58448028564453, "learning_rate": 1.2168076391719489e-07, "logits/chosen": 0.14901934564113617, "logits/rejected": 0.1714628040790558, "logps/chosen": -417.12506103515625, "logps/ref_chosen": -54.964271545410156, "logps/ref_rejected": -92.42044067382812, "logps/rejected": -665.4682006835938, "loss": 0.8391, "margin_dpo/margin_mean": 210.886962890625, "margin_dpo/margin_std": 205.45460510253906, "step": 481 }, { "epoch": 0.7077826725403817, "fcm_dpo/beta": 0.004300164058804512, "fcm_dpo/delta": 0.2011958360671997, "fcm_dpo/margin": 154.00543212890625, "fcm_dpo/q_t": 0.3627600073814392, "grad_norm": 81.67280578613281, "learning_rate": 1.2058107576668938e-07, "logits/chosen": 0.04958938807249069, "logits/rejected": 0.0661940723657608, "logps/chosen": -464.51654052734375, "logps/ref_chosen": -67.553466796875, "logps/ref_rejected": -87.58953857421875, "logps/rejected": -638.5580444335938, "loss": 1.0015, "margin_dpo/margin_mean": 154.0054473876953, "margin_dpo/margin_std": 211.11434936523438, "step": 482 }, { "epoch": 0.7092511013215859, "fcm_dpo/beta": 0.004210878163576126, "fcm_dpo/delta": -0.2048397660255432, "fcm_dpo/margin": 245.37893676757812, "fcm_dpo/q_t": 0.2938799262046814, "grad_norm": 45.401424407958984, "learning_rate": 1.194847979251979e-07, "logits/chosen": 0.1060781329870224, "logits/rejected": 0.14212235808372498, "logps/chosen": -432.8665771484375, "logps/ref_chosen": -63.32981872558594, "logps/ref_rejected": -95.78697204589844, "logps/rejected": -710.70263671875, "loss": 0.7583, "margin_dpo/margin_mean": 245.37890625, "margin_dpo/margin_std": 219.05149841308594, "step": 483 }, { "epoch": 0.71071953010279, "fcm_dpo/beta": 0.004176275804638863, "fcm_dpo/delta": -0.05583025887608528, "fcm_dpo/margin": 215.44418334960938, "fcm_dpo/q_t": 0.313819944858551, "grad_norm": 44.620079040527344, "learning_rate": 1.1839195928066101e-07, "logits/chosen": 0.1926443874835968, "logits/rejected": 0.22298608720302582, "logps/chosen": -402.5946350097656, "logps/ref_chosen": -59.13812255859375, "logps/ref_rejected": -84.37144470214844, "logps/rejected": -643.2721557617188, "loss": 0.7975, "margin_dpo/margin_mean": 215.44418334960938, "margin_dpo/margin_std": 184.650146484375, "step": 484 }, { "epoch": 0.7121879588839941, "fcm_dpo/beta": 0.0040985457599163055, "fcm_dpo/delta": -0.07049298286437988, "fcm_dpo/margin": 223.0164031982422, "fcm_dpo/q_t": 0.31292468309402466, "grad_norm": 58.52872848510742, "learning_rate": 1.1730258863039347e-07, "logits/chosen": 0.12118280678987503, "logits/rejected": 0.13599839806556702, "logps/chosen": -386.78875732421875, "logps/ref_chosen": -58.849571228027344, "logps/ref_rejected": -103.36408233642578, "logps/rejected": -654.3197021484375, "loss": 0.825, "margin_dpo/margin_mean": 223.0164031982422, "margin_dpo/margin_std": 212.99270629882812, "step": 485 }, { "epoch": 0.7136563876651982, "fcm_dpo/beta": 0.004043485503643751, "fcm_dpo/delta": -0.08242425322532654, "fcm_dpo/margin": 228.66433715820312, "fcm_dpo/q_t": 0.31584662199020386, "grad_norm": 74.97518157958984, "learning_rate": 1.1621671468032493e-07, "logits/chosen": 0.17448630928993225, "logits/rejected": 0.20201058685779572, "logps/chosen": -424.857177734375, "logps/ref_chosen": -55.25966262817383, "logps/ref_rejected": -92.13936614990234, "logps/rejected": -690.4011840820312, "loss": 0.8548, "margin_dpo/margin_mean": 228.66432189941406, "margin_dpo/margin_std": 240.98989868164062, "step": 486 }, { "epoch": 0.7151248164464024, "fcm_dpo/beta": 0.004021212458610535, "fcm_dpo/delta": 0.06276712566614151, "fcm_dpo/margin": 196.9682159423828, "fcm_dpo/q_t": 0.32838380336761475, "grad_norm": 51.10014724731445, "learning_rate": 1.1513436604424378e-07, "logits/chosen": 0.13462239503860474, "logits/rejected": 0.15736320614814758, "logps/chosen": -429.49041748046875, "logps/ref_chosen": -53.06330871582031, "logps/ref_rejected": -92.41883087158203, "logps/rejected": -665.8141479492188, "loss": 0.857, "margin_dpo/margin_mean": 196.9682159423828, "margin_dpo/margin_std": 176.92190551757812, "step": 487 }, { "epoch": 0.7165932452276065, "fcm_dpo/beta": 0.004063923843204975, "fcm_dpo/delta": 0.014564893208444118, "fcm_dpo/margin": 205.87074279785156, "fcm_dpo/q_t": 0.31959283351898193, "grad_norm": 45.07450485229492, "learning_rate": 1.1405557124304335e-07, "logits/chosen": 0.1483457088470459, "logits/rejected": 0.18104881048202515, "logps/chosen": -406.4317932128906, "logps/ref_chosen": -52.22815704345703, "logps/ref_rejected": -84.00656127929688, "logps/rejected": -644.0809326171875, "loss": 0.8086, "margin_dpo/margin_mean": 205.87075805664062, "margin_dpo/margin_std": 163.66893005371094, "step": 488 }, { "epoch": 0.7180616740088106, "fcm_dpo/beta": 0.00414654053747654, "fcm_dpo/delta": 0.13501515984535217, "fcm_dpo/margin": 174.794677734375, "fcm_dpo/q_t": 0.35005754232406616, "grad_norm": 48.02519607543945, "learning_rate": 1.1298035870396985e-07, "logits/chosen": 0.07817474007606506, "logits/rejected": 0.10687027126550674, "logps/chosen": -400.2724914550781, "logps/ref_chosen": -55.989627838134766, "logps/ref_rejected": -79.39812469482422, "logps/rejected": -598.4757080078125, "loss": 0.9059, "margin_dpo/margin_mean": 174.794677734375, "margin_dpo/margin_std": 185.54888916015625, "step": 489 }, { "epoch": 0.7195301027900147, "fcm_dpo/beta": 0.004186190664768219, "fcm_dpo/delta": -0.00793827697634697, "fcm_dpo/margin": 204.7318878173828, "fcm_dpo/q_t": 0.3310641646385193, "grad_norm": 46.17793273925781, "learning_rate": 1.1190875675987355e-07, "logits/chosen": 0.06553199887275696, "logits/rejected": 0.07176955789327621, "logps/chosen": -421.15704345703125, "logps/ref_chosen": -52.36639404296875, "logps/ref_rejected": -110.4090576171875, "logps/rejected": -683.9315795898438, "loss": 0.8888, "margin_dpo/margin_mean": 204.73190307617188, "margin_dpo/margin_std": 232.12014770507812, "step": 490 }, { "epoch": 0.7209985315712188, "fcm_dpo/beta": 0.00430173147469759, "fcm_dpo/delta": 0.19360697269439697, "fcm_dpo/margin": 155.4693603515625, "fcm_dpo/q_t": 0.3650928735733032, "grad_norm": 54.773380279541016, "learning_rate": 1.1084079364846241e-07, "logits/chosen": 0.20436616241931915, "logits/rejected": 0.23191238939762115, "logps/chosen": -448.61712646484375, "logps/ref_chosen": -60.11626434326172, "logps/ref_rejected": -73.27278900146484, "logps/rejected": -617.2430419921875, "loss": 0.9805, "margin_dpo/margin_mean": 155.46937561035156, "margin_dpo/margin_std": 202.46267700195312, "step": 491 }, { "epoch": 0.7224669603524229, "fcm_dpo/beta": 0.00451195752248168, "fcm_dpo/delta": 0.2712585926055908, "fcm_dpo/margin": 131.85690307617188, "fcm_dpo/q_t": 0.3848981261253357, "grad_norm": 64.95230102539062, "learning_rate": 1.097764975115576e-07, "logits/chosen": 0.2151469886302948, "logits/rejected": 0.23940692842006683, "logps/chosen": -462.7021789550781, "logps/ref_chosen": -53.994178771972656, "logps/ref_rejected": -72.65962219238281, "logps/rejected": -613.2244873046875, "loss": 1.12, "margin_dpo/margin_mean": 131.85690307617188, "margin_dpo/margin_std": 238.486328125, "step": 492 }, { "epoch": 0.723935389133627, "fcm_dpo/beta": 0.004592692479491234, "fcm_dpo/delta": 0.06759007275104523, "fcm_dpo/margin": 171.4335174560547, "fcm_dpo/q_t": 0.33537688851356506, "grad_norm": 69.35541534423828, "learning_rate": 1.0871589639435203e-07, "logits/chosen": 0.1804664134979248, "logits/rejected": 0.21824055910110474, "logps/chosen": -479.1470642089844, "logps/ref_chosen": -75.49723815917969, "logps/ref_rejected": -87.32301330566406, "logps/rejected": -662.4063720703125, "loss": 0.8862, "margin_dpo/margin_mean": 171.4335174560547, "margin_dpo/margin_std": 175.6283416748047, "step": 493 }, { "epoch": 0.7254038179148311, "fcm_dpo/beta": 0.004511960782110691, "fcm_dpo/delta": -0.20742294192314148, "fcm_dpo/margin": 229.67926025390625, "fcm_dpo/q_t": 0.285893052816391, "grad_norm": 53.599117279052734, "learning_rate": 1.0765901824467166e-07, "logits/chosen": 0.17816877365112305, "logits/rejected": 0.18146733939647675, "logps/chosen": -398.7049865722656, "logps/ref_chosen": -41.35926818847656, "logps/ref_rejected": -86.09136962890625, "logps/rejected": -673.1163330078125, "loss": 0.753, "margin_dpo/margin_mean": 229.6792449951172, "margin_dpo/margin_std": 192.52398681640625, "step": 494 }, { "epoch": 0.7268722466960352, "fcm_dpo/beta": 0.004417981021106243, "fcm_dpo/delta": -0.08530034869909286, "fcm_dpo/margin": 209.88265991210938, "fcm_dpo/q_t": 0.31514835357666016, "grad_norm": 48.33843994140625, "learning_rate": 1.0660589091223854e-07, "logits/chosen": 0.09450630843639374, "logits/rejected": 0.11175034940242767, "logps/chosen": -458.7616271972656, "logps/ref_chosen": -63.53507995605469, "logps/ref_rejected": -91.42443084716797, "logps/rejected": -696.5336303710938, "loss": 0.8294, "margin_dpo/margin_mean": 209.88265991210938, "margin_dpo/margin_std": 212.19480895996094, "step": 495 }, { "epoch": 0.7283406754772394, "fcm_dpo/beta": 0.004461340140551329, "fcm_dpo/delta": 0.14957168698310852, "fcm_dpo/margin": 159.43557739257812, "fcm_dpo/q_t": 0.3518192172050476, "grad_norm": 72.45964813232422, "learning_rate": 1.0555654214793722e-07, "logits/chosen": 0.12275470048189163, "logits/rejected": 0.15248063206672668, "logps/chosen": -506.01953125, "logps/ref_chosen": -72.5919189453125, "logps/ref_rejected": -84.32933807373047, "logps/rejected": -677.1925048828125, "loss": 0.9246, "margin_dpo/margin_mean": 159.43557739257812, "margin_dpo/margin_std": 172.95962524414062, "step": 496 }, { "epoch": 0.7298091042584435, "fcm_dpo/beta": 0.004558074288070202, "fcm_dpo/delta": 0.1612219661474228, "fcm_dpo/margin": 153.30006408691406, "fcm_dpo/q_t": 0.35529017448425293, "grad_norm": 54.61373519897461, "learning_rate": 1.0451099960308374e-07, "logits/chosen": 0.1395212858915329, "logits/rejected": 0.1674782931804657, "logps/chosen": -488.2263488769531, "logps/ref_chosen": -58.59397506713867, "logps/ref_rejected": -76.28836822509766, "logps/rejected": -659.2208251953125, "loss": 0.9637, "margin_dpo/margin_mean": 153.30006408691406, "margin_dpo/margin_std": 177.5799560546875, "step": 497 }, { "epoch": 0.7312775330396476, "fcm_dpo/beta": 0.004678555764257908, "fcm_dpo/delta": 0.03473177179694176, "fcm_dpo/margin": 174.86441040039062, "fcm_dpo/q_t": 0.3393765687942505, "grad_norm": 51.891902923583984, "learning_rate": 1.0346929082869641e-07, "logits/chosen": 0.16660353541374207, "logits/rejected": 0.19526565074920654, "logps/chosen": -492.501220703125, "logps/ref_chosen": -71.20565795898438, "logps/ref_rejected": -83.95803833007812, "logps/rejected": -680.1179809570312, "loss": 0.9475, "margin_dpo/margin_mean": 174.86439514160156, "margin_dpo/margin_std": 226.5364990234375, "step": 498 }, { "epoch": 0.7327459618208517, "fcm_dpo/beta": 0.004574180580675602, "fcm_dpo/delta": -0.16882994771003723, "fcm_dpo/margin": 219.0482940673828, "fcm_dpo/q_t": 0.2998501658439636, "grad_norm": 52.29130935668945, "learning_rate": 1.0243144327477013e-07, "logits/chosen": 0.23509922623634338, "logits/rejected": 0.2455160915851593, "logps/chosen": -449.7971496582031, "logps/ref_chosen": -51.25519561767578, "logps/ref_rejected": -101.07870483398438, "logps/rejected": -718.6689453125, "loss": 0.8221, "margin_dpo/margin_mean": 219.04830932617188, "margin_dpo/margin_std": 219.20545959472656, "step": 499 }, { "epoch": 0.7342143906020558, "fcm_dpo/beta": 0.004520269110798836, "fcm_dpo/delta": -0.008800679817795753, "fcm_dpo/margin": 189.812744140625, "fcm_dpo/q_t": 0.3284400701522827, "grad_norm": 70.3599853515625, "learning_rate": 1.0139748428955333e-07, "logits/chosen": 0.20507007837295532, "logits/rejected": 0.20711365342140198, "logps/chosen": -493.9044189453125, "logps/ref_chosen": -57.027442932128906, "logps/ref_rejected": -93.93421173095703, "logps/rejected": -720.6239013671875, "loss": 0.8791, "margin_dpo/margin_mean": 189.812744140625, "margin_dpo/margin_std": 210.09674072265625, "step": 500 }, { "epoch": 0.73568281938326, "fcm_dpo/beta": 0.004478934220969677, "fcm_dpo/delta": -0.10455437749624252, "fcm_dpo/margin": 210.9066162109375, "fcm_dpo/q_t": 0.30981889367103577, "grad_norm": 52.43331527709961, "learning_rate": 1.0036744111882672e-07, "logits/chosen": 0.19089315831661224, "logits/rejected": 0.22942286729812622, "logps/chosen": -450.37811279296875, "logps/ref_chosen": -54.359527587890625, "logps/ref_rejected": -80.15670013427734, "logps/rejected": -687.0819091796875, "loss": 0.8183, "margin_dpo/margin_mean": 210.9066162109375, "margin_dpo/margin_std": 205.54580688476562, "step": 501 }, { "epoch": 0.737151248164464, "fcm_dpo/beta": 0.004374990239739418, "fcm_dpo/delta": -0.14604619145393372, "fcm_dpo/margin": 224.31927490234375, "fcm_dpo/q_t": 0.28903883695602417, "grad_norm": 47.08027267456055, "learning_rate": 9.934134090518592e-08, "logits/chosen": 0.09788864850997925, "logits/rejected": 0.13583651185035706, "logps/chosen": -435.0950012207031, "logps/ref_chosen": -67.60050964355469, "logps/ref_rejected": -82.94876098632812, "logps/rejected": -674.7625122070312, "loss": 0.7296, "margin_dpo/margin_mean": 224.31927490234375, "margin_dpo/margin_std": 149.10888671875, "step": 502 }, { "epoch": 0.7386196769456681, "fcm_dpo/beta": 0.0042848363518714905, "fcm_dpo/delta": 0.005043823271989822, "fcm_dpo/margin": 197.19277954101562, "fcm_dpo/q_t": 0.3233737349510193, "grad_norm": 42.94404983520508, "learning_rate": 9.831921068732571e-08, "logits/chosen": 0.20290768146514893, "logits/rejected": 0.2336656153202057, "logps/chosen": -442.6397399902344, "logps/ref_chosen": -55.078407287597656, "logps/ref_rejected": -82.50544738769531, "logps/rejected": -667.259521484375, "loss": 0.8323, "margin_dpo/margin_mean": 197.19277954101562, "margin_dpo/margin_std": 179.80767822265625, "step": 503 }, { "epoch": 0.7400881057268722, "fcm_dpo/beta": 0.0042025502771139145, "fcm_dpo/delta": -0.16033101081848145, "fcm_dpo/margin": 236.57833862304688, "fcm_dpo/q_t": 0.2951487600803375, "grad_norm": 39.35810470581055, "learning_rate": 9.730107739932805e-08, "logits/chosen": 0.18868711590766907, "logits/rejected": 0.19342690706253052, "logps/chosen": -456.39227294921875, "logps/ref_chosen": -59.96575164794922, "logps/ref_rejected": -103.76212310791016, "logps/rejected": -736.7669677734375, "loss": 0.761, "margin_dpo/margin_mean": 236.57835388183594, "margin_dpo/margin_std": 194.76629638671875, "step": 504 }, { "epoch": 0.7415565345080763, "fcm_dpo/beta": 0.004324901849031448, "fcm_dpo/delta": 0.23218274116516113, "fcm_dpo/margin": 146.14663696289062, "fcm_dpo/q_t": 0.37902575731277466, "grad_norm": 94.34353637695312, "learning_rate": 9.628696786995188e-08, "logits/chosen": 0.12490086257457733, "logits/rejected": 0.15079885721206665, "logps/chosen": -492.49853515625, "logps/ref_chosen": -76.1549072265625, "logps/ref_rejected": -88.58537292480469, "logps/rejected": -651.0756225585938, "loss": 1.0587, "margin_dpo/margin_mean": 146.14663696289062, "margin_dpo/margin_std": 232.16616821289062, "step": 505 }, { "epoch": 0.7430249632892805, "fcm_dpo/beta": 0.004307071678340435, "fcm_dpo/delta": -0.050437767058610916, "fcm_dpo/margin": 207.72500610351562, "fcm_dpo/q_t": 0.3202528655529022, "grad_norm": 46.971092224121094, "learning_rate": 9.527690882192635e-08, "logits/chosen": 0.22537869215011597, "logits/rejected": 0.24729499220848083, "logps/chosen": -421.39666748046875, "logps/ref_chosen": -48.96050262451172, "logps/ref_rejected": -78.41505432128906, "logps/rejected": -658.5762329101562, "loss": 0.8507, "margin_dpo/margin_mean": 207.72500610351562, "margin_dpo/margin_std": 212.94607543945312, "step": 506 }, { "epoch": 0.7444933920704846, "fcm_dpo/beta": 0.004366847686469555, "fcm_dpo/delta": 0.05252450332045555, "fcm_dpo/margin": 183.55238342285156, "fcm_dpo/q_t": 0.33828064799308777, "grad_norm": 39.45782470703125, "learning_rate": 9.427092687124691e-08, "logits/chosen": 0.2182435542345047, "logits/rejected": 0.23851092159748077, "logps/chosen": -477.6820068359375, "logps/ref_chosen": -66.80149841308594, "logps/ref_rejected": -95.37289428710938, "logps/rejected": -689.8057861328125, "loss": 0.8826, "margin_dpo/margin_mean": 183.55239868164062, "margin_dpo/margin_std": 199.75308227539062, "step": 507 }, { "epoch": 0.7459618208516887, "fcm_dpo/beta": 0.004443483427166939, "fcm_dpo/delta": 0.08964800834655762, "fcm_dpo/margin": 172.5631561279297, "fcm_dpo/q_t": 0.34734398126602173, "grad_norm": 52.880027770996094, "learning_rate": 9.326904852647344e-08, "logits/chosen": 0.19013988971710205, "logits/rejected": 0.20654146373271942, "logps/chosen": -461.3097229003906, "logps/ref_chosen": -71.303466796875, "logps/ref_rejected": -95.6275405883789, "logps/rejected": -658.1969604492188, "loss": 0.9348, "margin_dpo/margin_mean": 172.5631561279297, "margin_dpo/margin_std": 208.7895965576172, "step": 508 }, { "epoch": 0.7474302496328928, "fcm_dpo/beta": 0.004504289943724871, "fcm_dpo/delta": 0.054536715149879456, "fcm_dpo/margin": 177.45721435546875, "fcm_dpo/q_t": 0.3389536142349243, "grad_norm": 45.22426223754883, "learning_rate": 9.227130018803195e-08, "logits/chosen": 0.03568051755428314, "logits/rejected": 0.04966065660119057, "logps/chosen": -405.90570068359375, "logps/ref_chosen": -63.81895065307617, "logps/ref_rejected": -83.25643920898438, "logps/rejected": -602.8004150390625, "loss": 0.8893, "margin_dpo/margin_mean": 177.45721435546875, "margin_dpo/margin_std": 193.94020080566406, "step": 509 }, { "epoch": 0.748898678414097, "fcm_dpo/beta": 0.0044273133389651775, "fcm_dpo/delta": -0.12143722176551819, "fcm_dpo/margin": 216.8166961669922, "fcm_dpo/q_t": 0.3008081316947937, "grad_norm": 44.7673225402832, "learning_rate": 9.127770814751932e-08, "logits/chosen": 0.1561957597732544, "logits/rejected": 0.15737219154834747, "logps/chosen": -436.9364013671875, "logps/ref_chosen": -51.878448486328125, "logps/ref_rejected": -102.7651596069336, "logps/rejected": -704.6397705078125, "loss": 0.7805, "margin_dpo/margin_mean": 216.8166961669922, "margin_dpo/margin_std": 185.16363525390625, "step": 510 }, { "epoch": 0.750367107195301, "fcm_dpo/beta": 0.004378630314022303, "fcm_dpo/delta": -0.055393002927303314, "fcm_dpo/margin": 205.57594299316406, "fcm_dpo/q_t": 0.32261648774147034, "grad_norm": 46.155799865722656, "learning_rate": 9.028829858700973e-08, "logits/chosen": 0.18099631369113922, "logits/rejected": 0.1909976303577423, "logps/chosen": -429.7141418457031, "logps/ref_chosen": -60.23811721801758, "logps/ref_rejected": -92.85676574707031, "logps/rejected": -667.9087524414062, "loss": 0.8824, "margin_dpo/margin_mean": 205.57594299316406, "margin_dpo/margin_std": 232.60919189453125, "step": 511 }, { "epoch": 0.7518355359765051, "fcm_dpo/beta": 0.00426853634417057, "fcm_dpo/delta": -0.1081642284989357, "fcm_dpo/margin": 222.03216552734375, "fcm_dpo/q_t": 0.3019694983959198, "grad_norm": 37.78676223754883, "learning_rate": 8.930309757836516e-08, "logits/chosen": 0.26449891924858093, "logits/rejected": 0.2904709279537201, "logps/chosen": -410.3562927246094, "logps/ref_chosen": -54.905494689941406, "logps/ref_rejected": -81.87586975097656, "logps/rejected": -659.35888671875, "loss": 0.7736, "margin_dpo/margin_mean": 222.03216552734375, "margin_dpo/margin_std": 183.77151489257812, "step": 512 }, { "epoch": 0.7533039647577092, "fcm_dpo/beta": 0.004307272844016552, "fcm_dpo/delta": 0.13749617338180542, "fcm_dpo/margin": 167.71078491210938, "fcm_dpo/q_t": 0.35119718313217163, "grad_norm": 74.993896484375, "learning_rate": 8.832213108254863e-08, "logits/chosen": 0.2815062999725342, "logits/rejected": 0.31944650411605835, "logps/chosen": -453.9299621582031, "logps/ref_chosen": -64.91644287109375, "logps/ref_rejected": -76.06245422363281, "logps/rejected": -632.7867431640625, "loss": 0.9587, "margin_dpo/margin_mean": 167.71078491210938, "margin_dpo/margin_std": 203.91494750976562, "step": 513 }, { "epoch": 0.7547723935389133, "fcm_dpo/beta": 0.004421452060341835, "fcm_dpo/delta": 0.08648448437452316, "fcm_dpo/margin": 174.20184326171875, "fcm_dpo/q_t": 0.3494170308113098, "grad_norm": 63.60978698730469, "learning_rate": 8.734542494893954e-08, "logits/chosen": 0.15184932947158813, "logits/rejected": 0.18162909150123596, "logps/chosen": -454.2314453125, "logps/ref_chosen": -74.22957611083984, "logps/ref_rejected": -78.945556640625, "logps/rejected": -633.1492919921875, "loss": 0.9468, "margin_dpo/margin_mean": 174.20184326171875, "margin_dpo/margin_std": 221.69790649414062, "step": 514 }, { "epoch": 0.7562408223201175, "fcm_dpo/beta": 0.004536722786724567, "fcm_dpo/delta": 0.15707728266716003, "fcm_dpo/margin": 155.23159790039062, "fcm_dpo/q_t": 0.3546082079410553, "grad_norm": 68.81385803222656, "learning_rate": 8.637300491465272e-08, "logits/chosen": 0.09542754292488098, "logits/rejected": 0.10104284435510635, "logps/chosen": -383.7822570800781, "logps/ref_chosen": -50.40156555175781, "logps/ref_rejected": -87.09774780273438, "logps/rejected": -575.7100219726562, "loss": 0.9591, "margin_dpo/margin_mean": 155.23159790039062, "margin_dpo/margin_std": 189.18768310546875, "step": 515 }, { "epoch": 0.7577092511013216, "fcm_dpo/beta": 0.004510688595473766, "fcm_dpo/delta": -0.14710889756679535, "fcm_dpo/margin": 217.89169311523438, "fcm_dpo/q_t": 0.2980697453022003, "grad_norm": 47.67902755737305, "learning_rate": 8.540489660386064e-08, "logits/chosen": 0.25269433856010437, "logits/rejected": 0.25404471158981323, "logps/chosen": -409.7538757324219, "logps/ref_chosen": -64.64956665039062, "logps/ref_rejected": -111.72237396240234, "logps/rejected": -674.7183837890625, "loss": 0.7673, "margin_dpo/margin_mean": 217.89169311523438, "margin_dpo/margin_std": 184.21299743652344, "step": 516 }, { "epoch": 0.7591776798825257, "fcm_dpo/beta": 0.004363768734037876, "fcm_dpo/delta": -0.09877770394086838, "fcm_dpo/margin": 215.03372192382812, "fcm_dpo/q_t": 0.31161239743232727, "grad_norm": 51.91146469116211, "learning_rate": 8.444112552711752e-08, "logits/chosen": 0.12666018307209015, "logits/rejected": 0.15497204661369324, "logps/chosen": -419.92864990234375, "logps/ref_chosen": -60.913551330566406, "logps/ref_rejected": -89.08308410644531, "logps/rejected": -663.1319580078125, "loss": 0.8144, "margin_dpo/margin_mean": 215.03372192382812, "margin_dpo/margin_std": 210.22381591796875, "step": 517 }, { "epoch": 0.7606461086637298, "fcm_dpo/beta": 0.004400432575494051, "fcm_dpo/delta": 0.06045955419540405, "fcm_dpo/margin": 180.52365112304688, "fcm_dpo/q_t": 0.3386451005935669, "grad_norm": 47.44013595581055, "learning_rate": 8.348171708068747e-08, "logits/chosen": 0.14773935079574585, "logits/rejected": 0.1560356616973877, "logps/chosen": -411.3404541015625, "logps/ref_chosen": -57.45589065551758, "logps/ref_rejected": -85.31269836425781, "logps/rejected": -619.7208862304688, "loss": 0.9154, "margin_dpo/margin_mean": 180.52365112304688, "margin_dpo/margin_std": 208.49659729003906, "step": 518 }, { "epoch": 0.762114537444934, "fcm_dpo/beta": 0.004505423828959465, "fcm_dpo/delta": 0.12294422835111618, "fcm_dpo/margin": 163.3507080078125, "fcm_dpo/q_t": 0.35014083981513977, "grad_norm": 41.567039489746094, "learning_rate": 8.25266965458755e-08, "logits/chosen": 0.07905039191246033, "logits/rejected": 0.09712234139442444, "logps/chosen": -403.34033203125, "logps/ref_chosen": -74.06331634521484, "logps/ref_rejected": -104.44416809082031, "logps/rejected": -597.0718994140625, "loss": 0.9354, "margin_dpo/margin_mean": 163.3507080078125, "margin_dpo/margin_std": 193.86541748046875, "step": 519 }, { "epoch": 0.7635829662261381, "fcm_dpo/beta": 0.00451300572603941, "fcm_dpo/delta": 0.04465455561876297, "fcm_dpo/margin": 178.88897705078125, "fcm_dpo/q_t": 0.3343246579170227, "grad_norm": 39.79988479614258, "learning_rate": 8.15760890883607e-08, "logits/chosen": 0.14485794305801392, "logits/rejected": 0.15455442667007446, "logps/chosen": -402.969482421875, "logps/ref_chosen": -70.2998275756836, "logps/ref_rejected": -99.98133850097656, "logps/rejected": -611.5400390625, "loss": 0.8643, "margin_dpo/margin_mean": 178.88897705078125, "margin_dpo/margin_std": 172.38330078125, "step": 520 }, { "epoch": 0.7650513950073421, "fcm_dpo/beta": 0.0046501983888447285, "fcm_dpo/delta": 0.020647704601287842, "fcm_dpo/margin": 177.8216552734375, "fcm_dpo/q_t": 0.3344711661338806, "grad_norm": 54.50603103637695, "learning_rate": 8.062991975753378e-08, "logits/chosen": 0.19291797280311584, "logits/rejected": 0.2164836972951889, "logps/chosen": -377.03448486328125, "logps/ref_chosen": -58.14292526245117, "logps/ref_rejected": -83.28060913085938, "logps/rejected": -579.9937744140625, "loss": 0.8861, "margin_dpo/margin_mean": 177.82164001464844, "margin_dpo/margin_std": 186.47982788085938, "step": 521 }, { "epoch": 0.7665198237885462, "fcm_dpo/beta": 0.004662246443331242, "fcm_dpo/delta": 0.09469278156757355, "fcm_dpo/margin": 163.50579833984375, "fcm_dpo/q_t": 0.3395306468009949, "grad_norm": 50.51918029785156, "learning_rate": 7.968821348583643e-08, "logits/chosen": 0.21042799949645996, "logits/rejected": 0.22489967942237854, "logps/chosen": -402.761474609375, "logps/ref_chosen": -46.54766845703125, "logps/ref_rejected": -66.01388549804688, "logps/rejected": -585.7335205078125, "loss": 0.9021, "margin_dpo/margin_mean": 163.5058135986328, "margin_dpo/margin_std": 173.2716522216797, "step": 522 }, { "epoch": 0.7679882525697503, "fcm_dpo/beta": 0.0046957386657595634, "fcm_dpo/delta": 0.042239535599946976, "fcm_dpo/margin": 172.7479248046875, "fcm_dpo/q_t": 0.32918986678123474, "grad_norm": 57.94752502441406, "learning_rate": 7.875099508810484e-08, "logits/chosen": 0.1128598004579544, "logits/rejected": 0.1486237496137619, "logps/chosen": -405.51531982421875, "logps/ref_chosen": -61.76960372924805, "logps/ref_rejected": -83.76141357421875, "logps/rejected": -600.2550048828125, "loss": 0.8846, "margin_dpo/margin_mean": 172.74790954589844, "margin_dpo/margin_std": 180.76382446289062, "step": 523 }, { "epoch": 0.7694566813509545, "fcm_dpo/beta": 0.004721796605736017, "fcm_dpo/delta": 0.032916050404310226, "fcm_dpo/margin": 173.58346557617188, "fcm_dpo/q_t": 0.32910919189453125, "grad_norm": 65.28778076171875, "learning_rate": 7.781828926091535e-08, "logits/chosen": 0.13729023933410645, "logits/rejected": 0.17816482484340668, "logps/chosen": -419.248779296875, "logps/ref_chosen": -78.0720443725586, "logps/ref_rejected": -81.30198669433594, "logps/rejected": -596.0621948242188, "loss": 0.8804, "margin_dpo/margin_mean": 173.58346557617188, "margin_dpo/margin_std": 179.30389404296875, "step": 524 }, { "epoch": 0.7709251101321586, "fcm_dpo/beta": 0.0045774709433317184, "fcm_dpo/delta": -0.18320885300636292, "fcm_dpo/margin": 220.95278930664062, "fcm_dpo/q_t": 0.2898215353488922, "grad_norm": 56.061824798583984, "learning_rate": 7.689012058193384e-08, "logits/chosen": 0.2006542831659317, "logits/rejected": 0.2043372541666031, "logps/chosen": -392.4875183105469, "logps/ref_chosen": -50.827857971191406, "logps/ref_rejected": -100.05294036865234, "logps/rejected": -662.6654052734375, "loss": 0.747, "margin_dpo/margin_mean": 220.95277404785156, "margin_dpo/margin_std": 174.82464599609375, "step": 525 }, { "epoch": 0.7723935389133627, "fcm_dpo/beta": 0.0045228018425405025, "fcm_dpo/delta": -0.1400771588087082, "fcm_dpo/margin": 215.81175231933594, "fcm_dpo/q_t": 0.30008208751678467, "grad_norm": 49.047428131103516, "learning_rate": 7.596651350926836e-08, "logits/chosen": 0.07574430853128433, "logits/rejected": 0.10800427943468094, "logps/chosen": -406.01776123046875, "logps/ref_chosen": -63.167236328125, "logps/ref_rejected": -86.30934143066406, "logps/rejected": -644.9716186523438, "loss": 0.7778, "margin_dpo/margin_mean": 215.81175231933594, "margin_dpo/margin_std": 185.41110229492188, "step": 526 }, { "epoch": 0.7738619676945668, "fcm_dpo/beta": 0.0044424207881093025, "fcm_dpo/delta": 0.03708699345588684, "fcm_dpo/margin": 183.49468994140625, "fcm_dpo/q_t": 0.33004334568977356, "grad_norm": 53.18086624145508, "learning_rate": 7.504749238082414e-08, "logits/chosen": 0.27492955327033997, "logits/rejected": 0.317468523979187, "logps/chosen": -429.421142578125, "logps/ref_chosen": -71.12867736816406, "logps/ref_rejected": -78.3425521850586, "logps/rejected": -620.1297607421875, "loss": 0.854, "margin_dpo/margin_mean": 183.49468994140625, "margin_dpo/margin_std": 172.89886474609375, "step": 527 }, { "epoch": 0.775330396475771, "fcm_dpo/beta": 0.0044939653016626835, "fcm_dpo/delta": 0.03126390278339386, "fcm_dpo/margin": 182.74923706054688, "fcm_dpo/q_t": 0.33725494146347046, "grad_norm": 43.81291198730469, "learning_rate": 7.413308141366254e-08, "logits/chosen": 0.19857966899871826, "logits/rejected": 0.22464889287948608, "logps/chosen": -438.86474609375, "logps/ref_chosen": -68.0894546508789, "logps/ref_rejected": -93.91006469726562, "logps/rejected": -647.4345703125, "loss": 0.8966, "margin_dpo/margin_mean": 182.74923706054688, "margin_dpo/margin_std": 207.980224609375, "step": 528 }, { "epoch": 0.7767988252569751, "fcm_dpo/beta": 0.004637697245925665, "fcm_dpo/delta": 0.18158438801765442, "fcm_dpo/margin": 146.76773071289062, "fcm_dpo/q_t": 0.3623617887496948, "grad_norm": 63.04121017456055, "learning_rate": 7.322330470336313e-08, "logits/chosen": 0.2721368074417114, "logits/rejected": 0.27319759130477905, "logps/chosen": -482.9656982421875, "logps/ref_chosen": -55.57495880126953, "logps/ref_rejected": -89.20909118652344, "logps/rejected": -663.3675537109375, "loss": 0.9968, "margin_dpo/margin_mean": 146.76773071289062, "margin_dpo/margin_std": 192.93203735351562, "step": 529 }, { "epoch": 0.7782672540381792, "fcm_dpo/beta": 0.0045324950478971004, "fcm_dpo/delta": -0.1806645691394806, "fcm_dpo/margin": 223.07403564453125, "fcm_dpo/q_t": 0.29591965675354004, "grad_norm": 33.03998565673828, "learning_rate": 7.231818622338822e-08, "logits/chosen": 0.19959846138954163, "logits/rejected": 0.21465358138084412, "logps/chosen": -394.14892578125, "logps/ref_chosen": -47.601417541503906, "logps/ref_rejected": -87.2845230102539, "logps/rejected": -656.906005859375, "loss": 0.7904, "margin_dpo/margin_mean": 223.07403564453125, "margin_dpo/margin_std": 207.3885955810547, "step": 530 }, { "epoch": 0.7797356828193832, "fcm_dpo/beta": 0.004460107535123825, "fcm_dpo/delta": -0.07478072494268417, "fcm_dpo/margin": 205.7947235107422, "fcm_dpo/q_t": 0.31518983840942383, "grad_norm": 38.86375045776367, "learning_rate": 7.141774982445147e-08, "logits/chosen": 0.20167914032936096, "logits/rejected": 0.2316320836544037, "logps/chosen": -419.4433898925781, "logps/ref_chosen": -55.246063232421875, "logps/ref_rejected": -70.60598754882812, "logps/rejected": -640.5980224609375, "loss": 0.8204, "margin_dpo/margin_mean": 205.79470825195312, "margin_dpo/margin_std": 199.72259521484375, "step": 531 }, { "epoch": 0.7812041116005873, "fcm_dpo/beta": 0.004327258560806513, "fcm_dpo/delta": -0.08652237057685852, "fcm_dpo/margin": 213.4281005859375, "fcm_dpo/q_t": 0.3172354996204376, "grad_norm": 60.72142791748047, "learning_rate": 7.052201923388953e-08, "logits/chosen": 0.1827022284269333, "logits/rejected": 0.21410122513771057, "logps/chosen": -462.40985107421875, "logps/ref_chosen": -70.28601837158203, "logps/ref_rejected": -86.5913314819336, "logps/rejected": -692.1432495117188, "loss": 0.8703, "margin_dpo/margin_mean": 213.4281005859375, "margin_dpo/margin_std": 228.88232421875, "step": 532 }, { "epoch": 0.7826725403817915, "fcm_dpo/beta": 0.004470104351639748, "fcm_dpo/delta": 0.0903887152671814, "fcm_dpo/margin": 170.91561889648438, "fcm_dpo/q_t": 0.34557852149009705, "grad_norm": 68.10647583007812, "learning_rate": 6.963101805503646e-08, "logits/chosen": 0.28478050231933594, "logits/rejected": 0.3137226700782776, "logps/chosen": -441.4239196777344, "logps/ref_chosen": -64.8551025390625, "logps/ref_rejected": -76.58805847167969, "logps/rejected": -624.072509765625, "loss": 0.9315, "margin_dpo/margin_mean": 170.91561889648438, "margin_dpo/margin_std": 198.55360412597656, "step": 533 }, { "epoch": 0.7841409691629956, "fcm_dpo/beta": 0.00448148837313056, "fcm_dpo/delta": 0.044943343847990036, "fcm_dpo/margin": 180.44354248046875, "fcm_dpo/q_t": 0.32874345779418945, "grad_norm": 65.2886962890625, "learning_rate": 6.874476976660184e-08, "logits/chosen": 0.23872573673725128, "logits/rejected": 0.2578517496585846, "logps/chosen": -446.6659240722656, "logps/ref_chosen": -60.119388580322266, "logps/ref_rejected": -78.54347229003906, "logps/rejected": -645.5335083007812, "loss": 0.8705, "margin_dpo/margin_mean": 180.4435577392578, "margin_dpo/margin_std": 180.7930450439453, "step": 534 }, { "epoch": 0.7856093979441997, "fcm_dpo/beta": 0.004449739120900631, "fcm_dpo/delta": -0.12627378106117249, "fcm_dpo/margin": 216.5131072998047, "fcm_dpo/q_t": 0.3024979829788208, "grad_norm": 58.6319694519043, "learning_rate": 6.786329772205246e-08, "logits/chosen": 0.19479429721832275, "logits/rejected": 0.21016496419906616, "logps/chosen": -411.903076171875, "logps/ref_chosen": -54.330238342285156, "logps/ref_rejected": -96.30763244628906, "logps/rejected": -670.3935546875, "loss": 0.7944, "margin_dpo/margin_mean": 216.51309204101562, "margin_dpo/margin_std": 189.78607177734375, "step": 535 }, { "epoch": 0.7870778267254038, "fcm_dpo/beta": 0.004233510233461857, "fcm_dpo/delta": -0.2273169904947281, "fcm_dpo/margin": 248.81976318359375, "fcm_dpo/q_t": 0.29112428426742554, "grad_norm": 39.82522964477539, "learning_rate": 6.698662514899638e-08, "logits/chosen": 0.24402159452438354, "logits/rejected": 0.2668285071849823, "logps/chosen": -386.976806640625, "logps/ref_chosen": -47.08053207397461, "logps/ref_rejected": -89.09783935546875, "logps/rejected": -677.8138427734375, "loss": 0.7483, "margin_dpo/margin_mean": 248.81976318359375, "margin_dpo/margin_std": 229.91015625, "step": 536 }, { "epoch": 0.788546255506608, "fcm_dpo/beta": 0.0041783712804317474, "fcm_dpo/delta": -0.026768144220113754, "fcm_dpo/margin": 209.25230407714844, "fcm_dpo/q_t": 0.3225961923599243, "grad_norm": 42.43117904663086, "learning_rate": 6.611477514857114e-08, "logits/chosen": 0.30811983346939087, "logits/rejected": 0.3353317379951477, "logps/chosen": -426.273681640625, "logps/ref_chosen": -57.747467041015625, "logps/ref_rejected": -70.43838500976562, "logps/rejected": -648.2169189453125, "loss": 0.8527, "margin_dpo/margin_mean": 209.25230407714844, "margin_dpo/margin_std": 213.77560424804688, "step": 537 }, { "epoch": 0.7900146842878121, "fcm_dpo/beta": 0.004137400537729263, "fcm_dpo/delta": -0.022371456027030945, "fcm_dpo/margin": 210.346435546875, "fcm_dpo/q_t": 0.31828272342681885, "grad_norm": 72.39550018310547, "learning_rate": 6.524777069483525e-08, "logits/chosen": 0.17623773217201233, "logits/rejected": 0.20072564482688904, "logps/chosen": -463.2252197265625, "logps/ref_chosen": -66.41594696044922, "logps/ref_rejected": -84.22808837890625, "logps/rejected": -691.3837890625, "loss": 0.8202, "margin_dpo/margin_mean": 210.346435546875, "margin_dpo/margin_std": 190.68777465820312, "step": 538 }, { "epoch": 0.7914831130690162, "fcm_dpo/beta": 0.004174415022134781, "fcm_dpo/delta": 0.07315509021282196, "fcm_dpo/margin": 187.47557067871094, "fcm_dpo/q_t": 0.3407885730266571, "grad_norm": 45.78681564331055, "learning_rate": 6.438563463416221e-08, "logits/chosen": 0.27691328525543213, "logits/rejected": 0.316150963306427, "logps/chosen": -448.6224670410156, "logps/ref_chosen": -58.492855072021484, "logps/ref_rejected": -91.85395050048828, "logps/rejected": -669.4591064453125, "loss": 0.9127, "margin_dpo/margin_mean": 187.4755859375, "margin_dpo/margin_std": 210.38111877441406, "step": 539 }, { "epoch": 0.7929515418502202, "fcm_dpo/beta": 0.004112423397600651, "fcm_dpo/delta": -0.17697307467460632, "fcm_dpo/margin": 245.46176147460938, "fcm_dpo/q_t": 0.29645735025405884, "grad_norm": 56.68818664550781, "learning_rate": 6.352838968463919e-08, "logits/chosen": 0.1551690399646759, "logits/rejected": 0.17202973365783691, "logps/chosen": -429.8592834472656, "logps/ref_chosen": -63.482513427734375, "logps/ref_rejected": -116.42999267578125, "logps/rejected": -728.2685546875, "loss": 0.7798, "margin_dpo/margin_mean": 245.46176147460938, "margin_dpo/margin_std": 219.45751953125, "step": 540 }, { "epoch": 0.7944199706314243, "fcm_dpo/beta": 0.004118548706173897, "fcm_dpo/delta": 0.16333335638046265, "fcm_dpo/margin": 169.4357452392578, "fcm_dpo/q_t": 0.3581434488296509, "grad_norm": 41.786102294921875, "learning_rate": 6.267605843546767e-08, "logits/chosen": 0.24253280460834503, "logits/rejected": 0.25958988070487976, "logps/chosen": -488.85546875, "logps/ref_chosen": -78.28036499023438, "logps/ref_rejected": -103.273681640625, "logps/rejected": -683.2845458984375, "loss": 0.9505, "margin_dpo/margin_mean": 169.43572998046875, "margin_dpo/margin_std": 202.7386474609375, "step": 541 }, { "epoch": 0.7958883994126285, "fcm_dpo/beta": 0.004145963117480278, "fcm_dpo/delta": -0.03485635668039322, "fcm_dpo/margin": 212.4905548095703, "fcm_dpo/q_t": 0.3207229971885681, "grad_norm": 45.65351867675781, "learning_rate": 6.182866334636888e-08, "logits/chosen": 0.2975878119468689, "logits/rejected": 0.3112480044364929, "logps/chosen": -468.6758728027344, "logps/ref_chosen": -57.48497009277344, "logps/ref_rejected": -96.47506713867188, "logps/rejected": -720.156494140625, "loss": 0.855, "margin_dpo/margin_mean": 212.4905548095703, "margin_dpo/margin_std": 215.06793212890625, "step": 542 }, { "epoch": 0.7973568281938326, "fcm_dpo/beta": 0.004305819049477577, "fcm_dpo/delta": 0.2336502969264984, "fcm_dpo/margin": 146.6068878173828, "fcm_dpo/q_t": 0.37988683581352234, "grad_norm": 68.22380065917969, "learning_rate": 6.098622674699147e-08, "logits/chosen": 0.2173738032579422, "logits/rejected": 0.22196441888809204, "logps/chosen": -473.72698974609375, "logps/ref_chosen": -60.61750793457031, "logps/ref_rejected": -105.59896850585938, "logps/rejected": -665.3153076171875, "loss": 1.053, "margin_dpo/margin_mean": 146.6068878173828, "margin_dpo/margin_std": 232.6273956298828, "step": 543 }, { "epoch": 0.7988252569750367, "fcm_dpo/beta": 0.004282426554709673, "fcm_dpo/delta": -0.11429713666439056, "fcm_dpo/margin": 222.5892333984375, "fcm_dpo/q_t": 0.3054772913455963, "grad_norm": 75.6998062133789, "learning_rate": 6.01487708363232e-08, "logits/chosen": 0.2191922664642334, "logits/rejected": 0.2349478006362915, "logps/chosen": -479.7855224609375, "logps/ref_chosen": -59.642303466796875, "logps/ref_rejected": -100.95469665527344, "logps/rejected": -743.6871337890625, "loss": 0.7888, "margin_dpo/margin_mean": 222.5892333984375, "margin_dpo/margin_std": 208.547119140625, "step": 544 }, { "epoch": 0.8002936857562408, "fcm_dpo/beta": 0.0042346809059381485, "fcm_dpo/delta": -0.06286168098449707, "fcm_dpo/margin": 214.2252197265625, "fcm_dpo/q_t": 0.3142136335372925, "grad_norm": 44.67675018310547, "learning_rate": 5.9316317682106294e-08, "logits/chosen": 0.19240760803222656, "logits/rejected": 0.21735535562038422, "logps/chosen": -476.2900390625, "logps/ref_chosen": -67.64859771728516, "logps/ref_rejected": -95.90800476074219, "logps/rejected": -718.774658203125, "loss": 0.8253, "margin_dpo/margin_mean": 214.2252197265625, "margin_dpo/margin_std": 203.46847534179688, "step": 545 }, { "epoch": 0.801762114537445, "fcm_dpo/beta": 0.0042076813988387585, "fcm_dpo/delta": 0.02801503613591194, "fcm_dpo/margin": 195.83470153808594, "fcm_dpo/q_t": 0.323789119720459, "grad_norm": 63.804405212402344, "learning_rate": 5.848888922025552e-08, "logits/chosen": 0.2083980143070221, "logits/rejected": 0.21705862879753113, "logps/chosen": -414.2108154296875, "logps/ref_chosen": -50.744232177734375, "logps/ref_rejected": -81.86622619628906, "logps/rejected": -641.16748046875, "loss": 0.8426, "margin_dpo/margin_mean": 195.83468627929688, "margin_dpo/margin_std": 175.96954345703125, "step": 546 }, { "epoch": 0.8032305433186491, "fcm_dpo/beta": 0.004326602444052696, "fcm_dpo/delta": 0.07258103787899017, "fcm_dpo/margin": 180.27406311035156, "fcm_dpo/q_t": 0.34270453453063965, "grad_norm": 57.07929992675781, "learning_rate": 5.7666507254280265e-08, "logits/chosen": 0.15857240557670593, "logits/rejected": 0.18794408440589905, "logps/chosen": -483.2398376464844, "logps/ref_chosen": -73.6877212524414, "logps/ref_rejected": -90.76136779785156, "logps/rejected": -680.5875244140625, "loss": 0.9307, "margin_dpo/margin_mean": 180.2740478515625, "margin_dpo/margin_std": 210.82923889160156, "step": 547 }, { "epoch": 0.8046989720998532, "fcm_dpo/beta": 0.004292323254048824, "fcm_dpo/delta": -0.011337485164403915, "fcm_dpo/margin": 200.43960571289062, "fcm_dpo/q_t": 0.32825782895088196, "grad_norm": 38.38809585571289, "learning_rate": 5.684919345471029e-08, "logits/chosen": 0.2047964632511139, "logits/rejected": 0.23223000764846802, "logps/chosen": -470.2217102050781, "logps/ref_chosen": -65.24634552001953, "logps/ref_rejected": -94.11807250976562, "logps/rejected": -699.5330810546875, "loss": 0.8781, "margin_dpo/margin_mean": 200.43960571289062, "margin_dpo/margin_std": 224.59344482421875, "step": 548 }, { "epoch": 0.8061674008810573, "fcm_dpo/beta": 0.004320558160543442, "fcm_dpo/delta": 0.03653667867183685, "fcm_dpo/margin": 188.9108428955078, "fcm_dpo/q_t": 0.33160555362701416, "grad_norm": 45.373291015625, "learning_rate": 5.603696935852426e-08, "logits/chosen": 0.3711293637752533, "logits/rejected": 0.3961469829082489, "logps/chosen": -453.705322265625, "logps/ref_chosen": -49.21235656738281, "logps/ref_rejected": -73.91031646728516, "logps/rejected": -667.3141479492188, "loss": 0.8724, "margin_dpo/margin_mean": 188.9108428955078, "margin_dpo/margin_std": 194.69717407226562, "step": 549 }, { "epoch": 0.8076358296622613, "fcm_dpo/beta": 0.004383495077490807, "fcm_dpo/delta": 0.12022919952869415, "fcm_dpo/margin": 168.54380798339844, "fcm_dpo/q_t": 0.3434358239173889, "grad_norm": 69.41998291015625, "learning_rate": 5.5229856368582376e-08, "logits/chosen": 0.2264268547296524, "logits/rejected": 0.23777663707733154, "logps/chosen": -461.681640625, "logps/ref_chosen": -56.80695343017578, "logps/ref_rejected": -95.12580871582031, "logps/rejected": -668.5443115234375, "loss": 0.9071, "margin_dpo/margin_mean": 168.5438232421875, "margin_dpo/margin_std": 176.12771606445312, "step": 550 }, { "epoch": 0.8091042584434655, "fcm_dpo/beta": 0.004263910930603743, "fcm_dpo/delta": -0.23163707554340363, "fcm_dpo/margin": 247.79818725585938, "fcm_dpo/q_t": 0.282936692237854, "grad_norm": 62.00813674926758, "learning_rate": 5.4427875753062734e-08, "logits/chosen": 0.2144840955734253, "logits/rejected": 0.22916750609874725, "logps/chosen": -457.0981750488281, "logps/ref_chosen": -59.10633087158203, "logps/ref_rejected": -111.67280578613281, "logps/rejected": -757.4628295898438, "loss": 0.7384, "margin_dpo/margin_mean": 247.79818725585938, "margin_dpo/margin_std": 201.80621337890625, "step": 551 }, { "epoch": 0.8105726872246696, "fcm_dpo/beta": 0.0039656441658735275, "fcm_dpo/delta": -0.325361430644989, "fcm_dpo/margin": 285.0164794921875, "fcm_dpo/q_t": 0.2748231887817383, "grad_norm": 48.39642333984375, "learning_rate": 5.363104864490034e-08, "logits/chosen": 0.27102088928222656, "logits/rejected": 0.3110225796699524, "logps/chosen": -439.73016357421875, "logps/ref_chosen": -62.35459899902344, "logps/ref_rejected": -104.56210327148438, "logps/rejected": -766.9541625976562, "loss": 0.7144, "margin_dpo/margin_mean": 285.0164794921875, "margin_dpo/margin_std": 232.68673706054688, "step": 552 }, { "epoch": 0.8120411160058737, "fcm_dpo/beta": 0.004009841941297054, "fcm_dpo/delta": 0.11692027747631073, "fcm_dpo/margin": 185.01705932617188, "fcm_dpo/q_t": 0.349956750869751, "grad_norm": 53.67270278930664, "learning_rate": 5.2839396041230415e-08, "logits/chosen": 0.2673987150192261, "logits/rejected": 0.27861133217811584, "logps/chosen": -479.7618408203125, "logps/ref_chosen": -68.25881958007812, "logps/ref_rejected": -98.0971450805664, "logps/rejected": -694.6171875, "loss": 0.9325, "margin_dpo/margin_mean": 185.01705932617188, "margin_dpo/margin_std": 215.61569213867188, "step": 553 }, { "epoch": 0.8135095447870778, "fcm_dpo/beta": 0.004004964604973793, "fcm_dpo/delta": -0.13882309198379517, "fcm_dpo/margin": 243.39794921875, "fcm_dpo/q_t": 0.3023296296596527, "grad_norm": 40.74170684814453, "learning_rate": 5.205293880283551e-08, "logits/chosen": 0.21930161118507385, "logits/rejected": 0.2681744694709778, "logps/chosen": -468.79296875, "logps/ref_chosen": -67.94767761230469, "logps/ref_rejected": -89.78272247314453, "logps/rejected": -734.0260009765625, "loss": 0.8201, "margin_dpo/margin_mean": 243.39793395996094, "margin_dpo/margin_std": 235.294677734375, "step": 554 }, { "epoch": 0.8149779735682819, "fcm_dpo/beta": 0.0038568670861423016, "fcm_dpo/delta": -0.09782232344150543, "fcm_dpo/margin": 243.01727294921875, "fcm_dpo/q_t": 0.31605011224746704, "grad_norm": 46.52155685424805, "learning_rate": 5.127169765359515e-08, "logits/chosen": 0.1832244098186493, "logits/rejected": 0.19693729281425476, "logps/chosen": -459.95538330078125, "logps/ref_chosen": -53.33049011230469, "logps/ref_rejected": -108.47937774658203, "logps/rejected": -758.1215209960938, "loss": 0.8346, "margin_dpo/margin_mean": 243.01727294921875, "margin_dpo/margin_std": 251.5319366455078, "step": 555 }, { "epoch": 0.8164464023494861, "fcm_dpo/beta": 0.003947093617171049, "fcm_dpo/delta": 0.1339261531829834, "fcm_dpo/margin": 183.7998046875, "fcm_dpo/q_t": 0.3540470004081726, "grad_norm": 52.72172927856445, "learning_rate": 5.049569317994012e-08, "logits/chosen": 0.33074843883514404, "logits/rejected": 0.3498724699020386, "logps/chosen": -489.8416748046875, "logps/ref_chosen": -58.64447021484375, "logps/ref_rejected": -101.34040832519531, "logps/rejected": -716.33740234375, "loss": 0.9402, "margin_dpo/margin_mean": 183.7998046875, "margin_dpo/margin_std": 221.2926025390625, "step": 556 }, { "epoch": 0.8179148311306902, "fcm_dpo/beta": 0.0039650630205869675, "fcm_dpo/delta": 0.025837931782007217, "fcm_dpo/margin": 208.3311309814453, "fcm_dpo/q_t": 0.3323240578174591, "grad_norm": 60.84648895263672, "learning_rate": 4.9724945830310144e-08, "logits/chosen": 0.28956037759780884, "logits/rejected": 0.3096484839916229, "logps/chosen": -509.36187744140625, "logps/ref_chosen": -67.84066009521484, "logps/ref_rejected": -109.93965911865234, "logps/rejected": -759.7919921875, "loss": 0.8753, "margin_dpo/margin_mean": 208.3311309814453, "margin_dpo/margin_std": 217.27667236328125, "step": 557 }, { "epoch": 0.8193832599118943, "fcm_dpo/beta": 0.0037960303016006947, "fcm_dpo/delta": -0.2908962368965149, "fcm_dpo/margin": 291.4281005859375, "fcm_dpo/q_t": 0.2752188742160797, "grad_norm": 40.26952362060547, "learning_rate": 4.8959475914614554e-08, "logits/chosen": 0.3146205544471741, "logits/rejected": 0.34363412857055664, "logps/chosen": -474.06671142578125, "logps/ref_chosen": -62.36824035644531, "logps/ref_rejected": -102.16102600097656, "logps/rejected": -805.28759765625, "loss": 0.7158, "margin_dpo/margin_mean": 291.4281005859375, "margin_dpo/margin_std": 233.851806640625, "step": 558 }, { "epoch": 0.8208516886930984, "fcm_dpo/beta": 0.0037619564682245255, "fcm_dpo/delta": 0.016046149656176567, "fcm_dpo/margin": 222.0172576904297, "fcm_dpo/q_t": 0.32790541648864746, "grad_norm": 42.88578414916992, "learning_rate": 4.8199303603697614e-08, "logits/chosen": 0.36911362409591675, "logits/rejected": 0.41761040687561035, "logps/chosen": -506.20428466796875, "logps/ref_chosen": -60.752323150634766, "logps/ref_rejected": -93.44229125976562, "logps/rejected": -760.9114990234375, "loss": 0.8556, "margin_dpo/margin_mean": 222.01727294921875, "margin_dpo/margin_std": 218.80276489257812, "step": 559 }, { "epoch": 0.8223201174743024, "fcm_dpo/beta": 0.003879477269947529, "fcm_dpo/delta": 0.20861107110977173, "fcm_dpo/margin": 168.96829223632812, "fcm_dpo/q_t": 0.3588427007198334, "grad_norm": 42.48268127441406, "learning_rate": 4.7444448928806615e-08, "logits/chosen": 0.1362684667110443, "logits/rejected": 0.17207416892051697, "logps/chosen": -461.98980712890625, "logps/ref_chosen": -58.10382080078125, "logps/ref_rejected": -79.99122619628906, "logps/rejected": -652.8455200195312, "loss": 0.9334, "margin_dpo/margin_mean": 168.96829223632812, "margin_dpo/margin_std": 171.30262756347656, "step": 560 }, { "epoch": 0.8237885462555066, "fcm_dpo/beta": 0.004035362042486668, "fcm_dpo/delta": 0.13644452393054962, "fcm_dpo/margin": 178.99203491210938, "fcm_dpo/q_t": 0.34845227003097534, "grad_norm": 50.79757308959961, "learning_rate": 4.669493178106432e-08, "logits/chosen": 0.3586878478527069, "logits/rejected": 0.3700917065143585, "logps/chosen": -518.5108032226562, "logps/ref_chosen": -50.912879943847656, "logps/ref_rejected": -99.06856536865234, "logps/rejected": -745.6585083007812, "loss": 0.919, "margin_dpo/margin_mean": 178.99203491210938, "margin_dpo/margin_std": 194.42037963867188, "step": 561 }, { "epoch": 0.8252569750367107, "fcm_dpo/beta": 0.003967948257923126, "fcm_dpo/delta": -0.0841178148984909, "fcm_dpo/margin": 233.1253204345703, "fcm_dpo/q_t": 0.31131601333618164, "grad_norm": 41.71147155761719, "learning_rate": 4.5950771910944596e-08, "logits/chosen": 0.29285359382629395, "logits/rejected": 0.3221287131309509, "logps/chosen": -477.37384033203125, "logps/ref_chosen": -59.46440124511719, "logps/ref_rejected": -96.54266357421875, "logps/rejected": -747.577392578125, "loss": 0.7998, "margin_dpo/margin_mean": 233.1253204345703, "margin_dpo/margin_std": 215.35162353515625, "step": 562 }, { "epoch": 0.8267254038179148, "fcm_dpo/beta": 0.004087609238922596, "fcm_dpo/delta": 0.11210109293460846, "fcm_dpo/margin": 181.70925903320312, "fcm_dpo/q_t": 0.3462071716785431, "grad_norm": 52.469322204589844, "learning_rate": 4.521198892775202e-08, "logits/chosen": 0.39729082584381104, "logits/rejected": 0.40474557876586914, "logps/chosen": -518.0103149414062, "logps/ref_chosen": -60.60819625854492, "logps/ref_rejected": -94.56770324707031, "logps/rejected": -733.6790771484375, "loss": 0.9472, "margin_dpo/margin_mean": 181.70925903320312, "margin_dpo/margin_std": 211.15899658203125, "step": 563 }, { "epoch": 0.8281938325991189, "fcm_dpo/beta": 0.004054870456457138, "fcm_dpo/delta": -0.04862257093191147, "fcm_dpo/margin": 220.50332641601562, "fcm_dpo/q_t": 0.30795860290527344, "grad_norm": 54.09901428222656, "learning_rate": 4.447860229910544e-08, "logits/chosen": 0.2921653985977173, "logits/rejected": 0.34295654296875, "logps/chosen": -511.1554870605469, "logps/ref_chosen": -74.26837921142578, "logps/ref_rejected": -93.23818969726562, "logps/rejected": -750.628662109375, "loss": 0.7805, "margin_dpo/margin_mean": 220.50332641601562, "margin_dpo/margin_std": 169.0012664794922, "step": 564 }, { "epoch": 0.8296622613803231, "fcm_dpo/beta": 0.003968285396695137, "fcm_dpo/delta": -0.06328395754098892, "fcm_dpo/margin": 228.4828338623047, "fcm_dpo/q_t": 0.32925814390182495, "grad_norm": 44.953067779541016, "learning_rate": 4.375063135042445e-08, "logits/chosen": 0.26385927200317383, "logits/rejected": 0.29121464490890503, "logps/chosen": -519.4959716796875, "logps/ref_chosen": -69.0199203491211, "logps/ref_rejected": -85.7789306640625, "logps/rejected": -764.73779296875, "loss": 0.8788, "margin_dpo/margin_mean": 228.48284912109375, "margin_dpo/margin_std": 268.53692626953125, "step": 565 }, { "epoch": 0.8311306901615272, "fcm_dpo/beta": 0.0039923894219100475, "fcm_dpo/delta": -0.0426069051027298, "fcm_dpo/margin": 222.23052978515625, "fcm_dpo/q_t": 0.3280484676361084, "grad_norm": 52.76844024658203, "learning_rate": 4.3028095264420525e-08, "logits/chosen": 0.29818111658096313, "logits/rejected": 0.3079987168312073, "logps/chosen": -506.3726806640625, "logps/ref_chosen": -66.5453109741211, "logps/ref_rejected": -103.86932373046875, "logps/rejected": -765.9271850585938, "loss": 0.9018, "margin_dpo/margin_mean": 222.2305145263672, "margin_dpo/margin_std": 253.78787231445312, "step": 566 }, { "epoch": 0.8325991189427313, "fcm_dpo/beta": 0.00396195612847805, "fcm_dpo/delta": 0.037429843097925186, "fcm_dpo/margin": 205.86737060546875, "fcm_dpo/q_t": 0.3339134156703949, "grad_norm": 55.71341323852539, "learning_rate": 4.231101308059165e-08, "logits/chosen": 0.41299670934677124, "logits/rejected": 0.4278637766838074, "logps/chosen": -497.1105651855469, "logps/ref_chosen": -52.85829544067383, "logps/ref_rejected": -85.37095642089844, "logps/rejected": -735.4906005859375, "loss": 0.891, "margin_dpo/margin_mean": 205.86737060546875, "margin_dpo/margin_std": 222.82290649414062, "step": 567 }, { "epoch": 0.8340675477239354, "fcm_dpo/beta": 0.003875305177643895, "fcm_dpo/delta": -0.12368164956569672, "fcm_dpo/margin": 247.7938232421875, "fcm_dpo/q_t": 0.2994388937950134, "grad_norm": 42.331607818603516, "learning_rate": 4.1599403694720145e-08, "logits/chosen": 0.39330175518989563, "logits/rejected": 0.40852218866348267, "logps/chosen": -468.718505859375, "logps/ref_chosen": -45.1923828125, "logps/ref_rejected": -89.09236907958984, "logps/rejected": -760.412353515625, "loss": 0.7856, "margin_dpo/margin_mean": 247.79380798339844, "margin_dpo/margin_std": 211.7144012451172, "step": 568 }, { "epoch": 0.8355359765051396, "fcm_dpo/beta": 0.003913109190762043, "fcm_dpo/delta": 0.01810070499777794, "fcm_dpo/margin": 212.70703125, "fcm_dpo/q_t": 0.3307640552520752, "grad_norm": 61.88147735595703, "learning_rate": 4.089328585837512e-08, "logits/chosen": 0.3647564649581909, "logits/rejected": 0.3922520875930786, "logps/chosen": -534.0196533203125, "logps/ref_chosen": -63.72056198120117, "logps/ref_rejected": -79.10325622558594, "logps/rejected": -762.109375, "loss": 0.8971, "margin_dpo/margin_mean": 212.70703125, "margin_dpo/margin_std": 238.76895141601562, "step": 569 }, { "epoch": 0.8370044052863436, "fcm_dpo/beta": 0.0038719356525689363, "fcm_dpo/delta": -0.008450102061033249, "fcm_dpo/margin": 221.46524047851562, "fcm_dpo/q_t": 0.3211510479450226, "grad_norm": 38.58411407470703, "learning_rate": 4.019267817841834e-08, "logits/chosen": 0.3776249885559082, "logits/rejected": 0.4106351137161255, "logps/chosen": -510.22760009765625, "logps/ref_chosen": -61.61454391479492, "logps/ref_rejected": -82.14186096191406, "logps/rejected": -752.2201538085938, "loss": 0.8246, "margin_dpo/margin_mean": 221.46524047851562, "margin_dpo/margin_std": 198.52322387695312, "step": 570 }, { "epoch": 0.8384728340675477, "fcm_dpo/beta": 0.0038212304934859276, "fcm_dpo/delta": -0.06161798536777496, "fcm_dpo/margin": 236.76951599121094, "fcm_dpo/q_t": 0.31742697954177856, "grad_norm": 43.7412109375, "learning_rate": 3.9497599116513705e-08, "logits/chosen": 0.35367658734321594, "logits/rejected": 0.3578590452671051, "logps/chosen": -511.2191162109375, "logps/ref_chosen": -53.05406188964844, "logps/ref_rejected": -91.33682250976562, "logps/rejected": -786.2713623046875, "loss": 0.8589, "margin_dpo/margin_mean": 236.76953125, "margin_dpo/margin_std": 248.20687866210938, "step": 571 }, { "epoch": 0.8399412628487518, "fcm_dpo/beta": 0.003810744732618332, "fcm_dpo/delta": -0.016613949090242386, "fcm_dpo/margin": 226.94198608398438, "fcm_dpo/q_t": 0.33660879731178284, "grad_norm": 51.98419952392578, "learning_rate": 3.880806698864086e-08, "logits/chosen": 0.4193460941314697, "logits/rejected": 0.4366481304168701, "logps/chosen": -525.885986328125, "logps/ref_chosen": -48.45928955078125, "logps/ref_rejected": -83.55703735351562, "logps/rejected": -787.92578125, "loss": 0.9044, "margin_dpo/margin_mean": 226.9420166015625, "margin_dpo/margin_std": 278.49652099609375, "step": 572 }, { "epoch": 0.8414096916299559, "fcm_dpo/beta": 0.0038648974150419235, "fcm_dpo/delta": -0.006211414933204651, "fcm_dpo/margin": 220.71739196777344, "fcm_dpo/q_t": 0.325981080532074, "grad_norm": 35.73110580444336, "learning_rate": 3.812409996461275e-08, "logits/chosen": 0.40158382058143616, "logits/rejected": 0.42190420627593994, "logps/chosen": -503.1387634277344, "logps/ref_chosen": -51.62262725830078, "logps/ref_rejected": -85.32499694824219, "logps/rejected": -757.55859375, "loss": 0.859, "margin_dpo/margin_mean": 220.7174072265625, "margin_dpo/margin_std": 216.8104248046875, "step": 573 }, { "epoch": 0.8428781204111601, "fcm_dpo/beta": 0.0037943366914987564, "fcm_dpo/delta": -0.05984480306506157, "fcm_dpo/margin": 238.36622619628906, "fcm_dpo/q_t": 0.3168014883995056, "grad_norm": 39.142181396484375, "learning_rate": 3.74457160675965e-08, "logits/chosen": 0.27881181240081787, "logits/rejected": 0.29145705699920654, "logps/chosen": -482.63153076171875, "logps/ref_chosen": -51.04446029663086, "logps/ref_rejected": -92.80640411376953, "logps/rejected": -762.759765625, "loss": 0.8247, "margin_dpo/margin_mean": 238.36622619628906, "margin_dpo/margin_std": 230.02125549316406, "step": 574 }, { "epoch": 0.8443465491923642, "fcm_dpo/beta": 0.0037506907247006893, "fcm_dpo/delta": 0.04211752861738205, "fcm_dpo/margin": 215.6813507080078, "fcm_dpo/q_t": 0.3343145251274109, "grad_norm": 64.44995880126953, "learning_rate": 3.677293317363864e-08, "logits/chosen": 0.3192945718765259, "logits/rejected": 0.3290221691131592, "logps/chosen": -558.0654296875, "logps/ref_chosen": -71.7901382446289, "logps/ref_rejected": -95.38619995117188, "logps/rejected": -797.3428955078125, "loss": 0.9433, "margin_dpo/margin_mean": 215.6813507080078, "margin_dpo/margin_std": 254.57664489746094, "step": 575 }, { "epoch": 0.8458149779735683, "fcm_dpo/beta": 0.0038080730009824038, "fcm_dpo/delta": 0.0639888197183609, "fcm_dpo/margin": 207.49903869628906, "fcm_dpo/q_t": 0.34286561608314514, "grad_norm": 62.07902908325195, "learning_rate": 3.6105769011194224e-08, "logits/chosen": 0.39104163646698, "logits/rejected": 0.39094260334968567, "logps/chosen": -515.055419921875, "logps/ref_chosen": -54.262962341308594, "logps/ref_rejected": -100.75428009033203, "logps/rejected": -769.0457763671875, "loss": 0.94, "margin_dpo/margin_mean": 207.49903869628906, "margin_dpo/margin_std": 248.49119567871094, "step": 576 }, { "epoch": 0.8472834067547724, "fcm_dpo/beta": 0.0038475426845252514, "fcm_dpo/delta": -0.0008432501927018166, "fcm_dpo/margin": 221.11129760742188, "fcm_dpo/q_t": 0.32097095251083374, "grad_norm": 62.19895935058594, "learning_rate": 3.5444241160659304e-08, "logits/chosen": 0.3040631413459778, "logits/rejected": 0.3278041481971741, "logps/chosen": -475.20794677734375, "logps/ref_chosen": -61.909706115722656, "logps/ref_rejected": -84.07069396972656, "logps/rejected": -718.4802856445312, "loss": 0.8523, "margin_dpo/margin_mean": 221.11129760742188, "margin_dpo/margin_std": 212.4851837158203, "step": 577 }, { "epoch": 0.8487518355359766, "fcm_dpo/beta": 0.0038197203539311886, "fcm_dpo/delta": 0.012510977685451508, "fcm_dpo/margin": 219.00831604003906, "fcm_dpo/q_t": 0.32471764087677, "grad_norm": 44.588985443115234, "learning_rate": 3.478836705390808e-08, "logits/chosen": 0.26734110713005066, "logits/rejected": 0.2847135066986084, "logps/chosen": -472.98272705078125, "logps/ref_chosen": -49.26368713378906, "logps/ref_rejected": -83.4362564086914, "logps/rejected": -726.1636352539062, "loss": 0.8381, "margin_dpo/margin_mean": 219.00831604003906, "margin_dpo/margin_std": 195.05392456054688, "step": 578 }, { "epoch": 0.8502202643171806, "fcm_dpo/beta": 0.003959665074944496, "fcm_dpo/delta": 0.16658331453800201, "fcm_dpo/margin": 175.51071166992188, "fcm_dpo/q_t": 0.35233074426651, "grad_norm": 41.42998123168945, "learning_rate": 3.41381639738331e-08, "logits/chosen": 0.30626869201660156, "logits/rejected": 0.3083363175392151, "logps/chosen": -489.62347412109375, "logps/ref_chosen": -58.88581848144531, "logps/ref_rejected": -94.78762817382812, "logps/rejected": -701.0360107421875, "loss": 0.9361, "margin_dpo/margin_mean": 175.51071166992188, "margin_dpo/margin_std": 193.71348571777344, "step": 579 }, { "epoch": 0.8516886930983847, "fcm_dpo/beta": 0.0038715798873454332, "fcm_dpo/delta": -0.14521247148513794, "fcm_dpo/margin": 252.62350463867188, "fcm_dpo/q_t": 0.3019111752510071, "grad_norm": 31.885099411010742, "learning_rate": 3.349364905389032e-08, "logits/chosen": 0.17919771373271942, "logits/rejected": 0.19062325358390808, "logps/chosen": -409.65533447265625, "logps/ref_chosen": -48.70683670043945, "logps/ref_rejected": -81.7583999633789, "logps/rejected": -695.3304443359375, "loss": 0.8033, "margin_dpo/margin_mean": 252.62350463867188, "margin_dpo/margin_std": 242.10226440429688, "step": 580 }, { "epoch": 0.8531571218795888, "fcm_dpo/beta": 0.003961851820349693, "fcm_dpo/delta": 0.13293930888175964, "fcm_dpo/margin": 183.37216186523438, "fcm_dpo/q_t": 0.351911723613739, "grad_norm": 58.27201843261719, "learning_rate": 3.285483927764726e-08, "logits/chosen": 0.3543952703475952, "logits/rejected": 0.37118589878082275, "logps/chosen": -507.3199462890625, "logps/ref_chosen": -62.22235107421875, "logps/ref_rejected": -91.73568725585938, "logps/rejected": -720.2054443359375, "loss": 0.9357, "margin_dpo/margin_mean": 183.37216186523438, "margin_dpo/margin_std": 216.66363525390625, "step": 581 }, { "epoch": 0.8546255506607929, "fcm_dpo/beta": 0.003983854316174984, "fcm_dpo/delta": -0.07189857959747314, "fcm_dpo/margin": 229.39111328125, "fcm_dpo/q_t": 0.31426823139190674, "grad_norm": 43.69565963745117, "learning_rate": 3.222175147833556e-08, "logits/chosen": 0.32797807455062866, "logits/rejected": 0.3196754455566406, "logps/chosen": -459.9640197753906, "logps/ref_chosen": -58.228660583496094, "logps/ref_rejected": -110.06959533691406, "logps/rejected": -741.196044921875, "loss": 0.8473, "margin_dpo/margin_mean": 229.39111328125, "margin_dpo/margin_std": 227.16668701171875, "step": 582 }, { "epoch": 0.856093979441997, "fcm_dpo/beta": 0.004001200199127197, "fcm_dpo/delta": 0.17536485195159912, "fcm_dpo/margin": 171.66009521484375, "fcm_dpo/q_t": 0.36176609992980957, "grad_norm": 77.05546569824219, "learning_rate": 3.159440233840763e-08, "logits/chosen": 0.2569863796234131, "logits/rejected": 0.27533864974975586, "logps/chosen": -492.8332824707031, "logps/ref_chosen": -56.86286163330078, "logps/ref_rejected": -88.4039306640625, "logps/rejected": -696.034423828125, "loss": 0.9734, "margin_dpo/margin_mean": 171.66009521484375, "margin_dpo/margin_std": 217.009033203125, "step": 583 }, { "epoch": 0.8575624082232012, "fcm_dpo/beta": 0.0039225309155881405, "fcm_dpo/delta": -0.24615153670310974, "fcm_dpo/margin": 272.740966796875, "fcm_dpo/q_t": 0.2808777689933777, "grad_norm": 36.94783020019531, "learning_rate": 3.0972808389096635e-08, "logits/chosen": 0.3280525207519531, "logits/rejected": 0.35432863235473633, "logps/chosen": -447.52764892578125, "logps/ref_chosen": -56.90068054199219, "logps/ref_rejected": -97.63606262207031, "logps/rejected": -761.0040283203125, "loss": 0.7226, "margin_dpo/margin_mean": 272.740966796875, "margin_dpo/margin_std": 216.50282287597656, "step": 584 }, { "epoch": 0.8590308370044053, "fcm_dpo/beta": 0.0038455924950540066, "fcm_dpo/delta": -0.048364780843257904, "fcm_dpo/margin": 232.48892211914062, "fcm_dpo/q_t": 0.32074883580207825, "grad_norm": 63.80207443237305, "learning_rate": 3.035698600998121e-08, "logits/chosen": 0.3095451295375824, "logits/rejected": 0.34348541498184204, "logps/chosen": -490.47076416015625, "logps/ref_chosen": -60.973968505859375, "logps/ref_rejected": -84.16952514648438, "logps/rejected": -746.1552734375, "loss": 0.8478, "margin_dpo/margin_mean": 232.4889373779297, "margin_dpo/margin_std": 235.10165405273438, "step": 585 }, { "epoch": 0.8604992657856094, "fcm_dpo/beta": 0.003913631662726402, "fcm_dpo/delta": 0.12130466103553772, "fcm_dpo/margin": 188.37155151367188, "fcm_dpo/q_t": 0.3504549264907837, "grad_norm": 61.881404876708984, "learning_rate": 2.974695142855388e-08, "logits/chosen": 0.37644636631011963, "logits/rejected": 0.3861965537071228, "logps/chosen": -507.32269287109375, "logps/ref_chosen": -56.85559844970703, "logps/ref_rejected": -91.80261993408203, "logps/rejected": -730.6412353515625, "loss": 0.9639, "margin_dpo/margin_mean": 188.37155151367188, "margin_dpo/margin_std": 239.3880157470703, "step": 586 }, { "epoch": 0.8619676945668135, "fcm_dpo/beta": 0.003928901627659798, "fcm_dpo/delta": -0.005717615596950054, "fcm_dpo/margin": 217.67236328125, "fcm_dpo/q_t": 0.3260841369628906, "grad_norm": 38.709537506103516, "learning_rate": 2.9142720719793122e-08, "logits/chosen": 0.42554670572280884, "logits/rejected": 0.42748117446899414, "logps/chosen": -425.38934326171875, "logps/ref_chosen": -44.69159698486328, "logps/ref_rejected": -82.62385559082031, "logps/rejected": -680.9939575195312, "loss": 0.8644, "margin_dpo/margin_mean": 217.67236328125, "margin_dpo/margin_std": 225.70684814453125, "step": 587 }, { "epoch": 0.8634361233480177, "fcm_dpo/beta": 0.003943789284676313, "fcm_dpo/delta": 0.06734154373407364, "fcm_dpo/margin": 199.75949096679688, "fcm_dpo/q_t": 0.33906006813049316, "grad_norm": 51.019248962402344, "learning_rate": 2.8544309805740018e-08, "logits/chosen": 0.32632678747177124, "logits/rejected": 0.32503804564476013, "logps/chosen": -493.21112060546875, "logps/ref_chosen": -50.29494857788086, "logps/ref_rejected": -107.36988067626953, "logps/rejected": -750.0455322265625, "loss": 0.8986, "margin_dpo/margin_mean": 199.75949096679688, "margin_dpo/margin_std": 217.92599487304688, "step": 588 }, { "epoch": 0.8649045521292217, "fcm_dpo/beta": 0.0038381319027394056, "fcm_dpo/delta": -0.22184929251670837, "fcm_dpo/margin": 273.01702880859375, "fcm_dpo/q_t": 0.2866641879081726, "grad_norm": 51.86333084106445, "learning_rate": 2.7951734455078786e-08, "logits/chosen": 0.3192276358604431, "logits/rejected": 0.3345522880554199, "logps/chosen": -472.3736572265625, "logps/ref_chosen": -59.929908752441406, "logps/ref_rejected": -111.65534973144531, "logps/rejected": -797.1161499023438, "loss": 0.7354, "margin_dpo/margin_mean": 273.01702880859375, "margin_dpo/margin_std": 222.23690795898438, "step": 589 }, { "epoch": 0.8663729809104258, "fcm_dpo/beta": 0.003773223143070936, "fcm_dpo/delta": -0.03967685252428055, "fcm_dpo/margin": 234.80557250976562, "fcm_dpo/q_t": 0.3133876919746399, "grad_norm": 35.085304260253906, "learning_rate": 2.736501028272095e-08, "logits/chosen": 0.2434614598751068, "logits/rejected": 0.23982518911361694, "logps/chosen": -463.9925842285156, "logps/ref_chosen": -55.80979537963867, "logps/ref_rejected": -106.06282043457031, "logps/rejected": -749.0511474609375, "loss": 0.8136, "margin_dpo/margin_mean": 234.80555725097656, "margin_dpo/margin_std": 206.89405822753906, "step": 590 }, { "epoch": 0.8678414096916299, "fcm_dpo/beta": 0.003768111113458872, "fcm_dpo/delta": -0.02017318457365036, "fcm_dpo/margin": 230.463134765625, "fcm_dpo/q_t": 0.3158026337623596, "grad_norm": 53.08750534057617, "learning_rate": 2.678415274939408e-08, "logits/chosen": 0.4338451623916626, "logits/rejected": 0.47315138578414917, "logps/chosen": -498.3539733886719, "logps/ref_chosen": -56.24061965942383, "logps/ref_rejected": -83.78629302978516, "logps/rejected": -756.36279296875, "loss": 0.8215, "margin_dpo/margin_mean": 230.463134765625, "margin_dpo/margin_std": 205.03244018554688, "step": 591 }, { "epoch": 0.869309838472834, "fcm_dpo/beta": 0.0037732375785708427, "fcm_dpo/delta": 0.052767086774110794, "fcm_dpo/margin": 212.37567138671875, "fcm_dpo/q_t": 0.3374939262866974, "grad_norm": 74.47343444824219, "learning_rate": 2.6209177161234442e-08, "logits/chosen": 0.47971880435943604, "logits/rejected": 0.4983103573322296, "logps/chosen": -526.7408447265625, "logps/ref_chosen": -47.94025421142578, "logps/ref_rejected": -75.73287963867188, "logps/rejected": -766.9091186523438, "loss": 0.9714, "margin_dpo/margin_mean": 212.3756561279297, "margin_dpo/margin_std": 278.6368713378906, "step": 592 }, { "epoch": 0.8707782672540382, "fcm_dpo/beta": 0.003901700722053647, "fcm_dpo/delta": 0.1894446462392807, "fcm_dpo/margin": 172.6243896484375, "fcm_dpo/q_t": 0.36253821849823, "grad_norm": 71.90731811523438, "learning_rate": 2.564009866938349e-08, "logits/chosen": 0.2871604561805725, "logits/rejected": 0.30711573362350464, "logps/chosen": -473.5516357421875, "logps/ref_chosen": -48.690757751464844, "logps/ref_rejected": -60.90800094604492, "logps/rejected": -658.393310546875, "loss": 0.9842, "margin_dpo/margin_mean": 172.6243896484375, "margin_dpo/margin_std": 221.938232421875, "step": 593 }, { "epoch": 0.8722466960352423, "fcm_dpo/beta": 0.0039725289680063725, "fcm_dpo/delta": -0.07375653088092804, "fcm_dpo/margin": 229.670654296875, "fcm_dpo/q_t": 0.3251326084136963, "grad_norm": 57.134742736816406, "learning_rate": 2.5076932269588708e-08, "logits/chosen": 0.32449811697006226, "logits/rejected": 0.34925204515457153, "logps/chosen": -467.6834411621094, "logps/ref_chosen": -54.93488693237305, "logps/ref_rejected": -86.09967803955078, "logps/rejected": -728.5189208984375, "loss": 0.8996, "margin_dpo/margin_mean": 229.67066955566406, "margin_dpo/margin_std": 259.094482421875, "step": 594 }, { "epoch": 0.8737151248164464, "fcm_dpo/beta": 0.0038999663665890694, "fcm_dpo/delta": 0.03644701838493347, "fcm_dpo/margin": 209.3623504638672, "fcm_dpo/q_t": 0.33476635813713074, "grad_norm": 41.2265510559082, "learning_rate": 2.451969280180849e-08, "logits/chosen": 0.40078675746917725, "logits/rejected": 0.42494046688079834, "logps/chosen": -484.77764892578125, "logps/ref_chosen": -49.4204216003418, "logps/ref_rejected": -80.62731170654297, "logps/rejected": -725.346923828125, "loss": 0.8717, "margin_dpo/margin_mean": 209.3623504638672, "margin_dpo/margin_std": 223.2598419189453, "step": 595 }, { "epoch": 0.8751835535976505, "fcm_dpo/beta": 0.003989151678979397, "fcm_dpo/delta": 0.09338672459125519, "fcm_dpo/margin": 191.2503204345703, "fcm_dpo/q_t": 0.3463008403778076, "grad_norm": 54.039161682128906, "learning_rate": 2.396839494982103e-08, "logits/chosen": 0.3313291668891907, "logits/rejected": 0.35179513692855835, "logps/chosen": -501.94403076171875, "logps/ref_chosen": -59.791683197021484, "logps/ref_rejected": -80.09111785888672, "logps/rejected": -713.4937744140625, "loss": 0.9414, "margin_dpo/margin_mean": 191.2503204345703, "margin_dpo/margin_std": 234.5662841796875, "step": 596 }, { "epoch": 0.8766519823788547, "fcm_dpo/beta": 0.0038905441761016846, "fcm_dpo/delta": -0.08888904005289078, "fcm_dpo/margin": 238.23074340820312, "fcm_dpo/q_t": 0.3173202574253082, "grad_norm": 49.979923248291016, "learning_rate": 2.3423053240837514e-08, "logits/chosen": 0.3219120502471924, "logits/rejected": 0.3208301365375519, "logps/chosen": -515.6228637695312, "logps/ref_chosen": -57.26078796386719, "logps/ref_rejected": -100.6937255859375, "logps/rejected": -797.2865600585938, "loss": 0.849, "margin_dpo/margin_mean": 238.23074340820312, "margin_dpo/margin_std": 249.84375, "step": 597 }, { "epoch": 0.8781204111600588, "fcm_dpo/beta": 0.003907772246748209, "fcm_dpo/delta": 0.07058212906122208, "fcm_dpo/margin": 200.17642211914062, "fcm_dpo/q_t": 0.3359318971633911, "grad_norm": 63.312461853027344, "learning_rate": 2.2883682045119062e-08, "logits/chosen": 0.4510478973388672, "logits/rejected": 0.47181278467178345, "logps/chosen": -533.74072265625, "logps/ref_chosen": -52.51850509643555, "logps/ref_rejected": -89.44385528564453, "logps/rejected": -770.842529296875, "loss": 0.9574, "margin_dpo/margin_mean": 200.17642211914062, "margin_dpo/margin_std": 237.43214416503906, "step": 598 }, { "epoch": 0.8795888399412628, "fcm_dpo/beta": 0.003973391838371754, "fcm_dpo/delta": 0.02901943027973175, "fcm_dpo/margin": 207.14028930664062, "fcm_dpo/q_t": 0.32341066002845764, "grad_norm": 49.349613189697266, "learning_rate": 2.2350295575598367e-08, "logits/chosen": 0.3739485740661621, "logits/rejected": 0.3845040798187256, "logps/chosen": -481.701171875, "logps/ref_chosen": -49.802677154541016, "logps/ref_rejected": -82.978515625, "logps/rejected": -722.017333984375, "loss": 0.8359, "margin_dpo/margin_mean": 207.14028930664062, "margin_dpo/margin_std": 179.18177795410156, "step": 599 }, { "epoch": 0.8810572687224669, "fcm_dpo/beta": 0.00399825070053339, "fcm_dpo/delta": 0.015949290245771408, "fcm_dpo/margin": 208.917236328125, "fcm_dpo/q_t": 0.33174365758895874, "grad_norm": 52.51283645629883, "learning_rate": 2.1822907887504932e-08, "logits/chosen": 0.4077937602996826, "logits/rejected": 0.42742469906806946, "logps/chosen": -556.25830078125, "logps/ref_chosen": -66.43487548828125, "logps/ref_rejected": -85.45649719238281, "logps/rejected": -784.1971435546875, "loss": 0.8902, "margin_dpo/margin_mean": 208.917236328125, "margin_dpo/margin_std": 230.8261260986328, "step": 600 }, { "epoch": 0.8810572687224669, "eval_fcm_dpo/beta": 0.004011388868093491, "eval_logits/chosen": 0.3258829414844513, "eval_logits/rejected": 0.3477090895175934, "eval_logps/chosen": -579.7042236328125, "eval_logps/ref_chosen": -79.05104064941406, "eval_logps/ref_rejected": -86.79793548583984, "eval_logps/rejected": -736.6627807617188, "eval_loss": 0.531234085559845, "eval_margin_dpo/margin_mean": 149.2117156982422, "eval_margin_dpo/margin_std": 231.603759765625, "eval_runtime": 39.2746, "eval_samples_per_second": 59.555, "eval_steps_per_second": 1.884, "step": 600 }, { "epoch": 0.882525697503671, "fcm_dpo/beta": 0.003966460935771465, "fcm_dpo/delta": -0.039181776344776154, "fcm_dpo/margin": 223.1067352294922, "fcm_dpo/q_t": 0.32117369771003723, "grad_norm": 48.35001754760742, "learning_rate": 2.1301532877994742e-08, "logits/chosen": 0.3656480312347412, "logits/rejected": 0.38585516810417175, "logps/chosen": -557.62841796875, "logps/ref_chosen": -59.13361358642578, "logps/ref_rejected": -94.69093322753906, "logps/rejected": -816.29248046875, "loss": 0.8571, "margin_dpo/margin_mean": 223.10675048828125, "margin_dpo/margin_std": 232.555908203125, "step": 601 }, { "epoch": 0.8839941262848752, "fcm_dpo/beta": 0.003887756960466504, "fcm_dpo/delta": -0.1840885877609253, "fcm_dpo/margin": 261.26611328125, "fcm_dpo/q_t": 0.29467278718948364, "grad_norm": 46.679325103759766, "learning_rate": 2.0786184285784298e-08, "logits/chosen": 0.35937434434890747, "logits/rejected": 0.3779703676700592, "logps/chosen": -445.2738037109375, "logps/ref_chosen": -48.59352111816406, "logps/ref_rejected": -87.6685562133789, "logps/rejected": -745.614990234375, "loss": 0.7682, "margin_dpo/margin_mean": 261.26611328125, "margin_dpo/margin_std": 232.40969848632812, "step": 602 }, { "epoch": 0.8854625550660793, "fcm_dpo/beta": 0.003834263887256384, "fcm_dpo/delta": -0.023736726492643356, "fcm_dpo/margin": 227.3087158203125, "fcm_dpo/q_t": 0.3306729197502136, "grad_norm": 67.39020538330078, "learning_rate": 2.0276875690788204e-08, "logits/chosen": 0.29140499234199524, "logits/rejected": 0.31848928332328796, "logps/chosen": -495.6649169921875, "logps/ref_chosen": -70.41461944580078, "logps/ref_rejected": -100.32559967041016, "logps/rejected": -752.8846435546875, "loss": 0.8786, "margin_dpo/margin_mean": 227.3087158203125, "margin_dpo/margin_std": 258.9732971191406, "step": 603 }, { "epoch": 0.8869309838472834, "fcm_dpo/beta": 0.0037519652396440506, "fcm_dpo/delta": -0.11160653829574585, "fcm_dpo/margin": 253.43344116210938, "fcm_dpo/q_t": 0.3031109571456909, "grad_norm": 40.32583236694336, "learning_rate": 1.977362051376158e-08, "logits/chosen": 0.343853235244751, "logits/rejected": 0.3555789291858673, "logps/chosen": -452.9412841796875, "logps/ref_chosen": -46.45808029174805, "logps/ref_rejected": -91.8544921875, "logps/rejected": -751.7711181640625, "loss": 0.8107, "margin_dpo/margin_mean": 253.43344116210938, "margin_dpo/margin_std": 235.12753295898438, "step": 604 }, { "epoch": 0.8883994126284875, "fcm_dpo/beta": 0.003726169466972351, "fcm_dpo/delta": 0.026752449572086334, "fcm_dpo/margin": 221.3822479248047, "fcm_dpo/q_t": 0.3330296277999878, "grad_norm": 40.421016693115234, "learning_rate": 1.9276432015946446e-08, "logits/chosen": 0.3358033299446106, "logits/rejected": 0.3515666723251343, "logps/chosen": -529.6724853515625, "logps/ref_chosen": -66.24933624267578, "logps/ref_rejected": -102.30496978759766, "logps/rejected": -787.1104125976562, "loss": 0.8714, "margin_dpo/margin_mean": 221.3822479248047, "margin_dpo/margin_std": 240.65045166015625, "step": 605 }, { "epoch": 0.8898678414096917, "fcm_dpo/beta": 0.003759493585675955, "fcm_dpo/delta": -0.06783300638198853, "fcm_dpo/margin": 242.02098083496094, "fcm_dpo/q_t": 0.3137454688549042, "grad_norm": 37.28118896484375, "learning_rate": 1.8785323298722093e-08, "logits/chosen": 0.39390993118286133, "logits/rejected": 0.40036922693252563, "logps/chosen": -500.85211181640625, "logps/ref_chosen": -54.819122314453125, "logps/ref_rejected": -98.37146759033203, "logps/rejected": -786.4254760742188, "loss": 0.8139, "margin_dpo/margin_mean": 242.02099609375, "margin_dpo/margin_std": 221.83334350585938, "step": 606 }, { "epoch": 0.8913362701908958, "fcm_dpo/beta": 0.0037815175019204617, "fcm_dpo/delta": 0.13359740376472473, "fcm_dpo/margin": 191.9698944091797, "fcm_dpo/q_t": 0.34541046619415283, "grad_norm": 50.32596206665039, "learning_rate": 1.8300307303259904e-08, "logits/chosen": 0.2882560193538666, "logits/rejected": 0.30261147022247314, "logps/chosen": -519.4744262695312, "logps/ref_chosen": -58.08403778076172, "logps/ref_rejected": -79.777099609375, "logps/rejected": -733.137451171875, "loss": 0.9034, "margin_dpo/margin_mean": 191.96990966796875, "margin_dpo/margin_std": 189.93313598632812, "step": 607 }, { "epoch": 0.8928046989720999, "fcm_dpo/beta": 0.0038187121972441673, "fcm_dpo/delta": 0.02122838981449604, "fcm_dpo/margin": 217.48269653320312, "fcm_dpo/q_t": 0.32343119382858276, "grad_norm": 38.28920364379883, "learning_rate": 1.7821396810182437e-08, "logits/chosen": 0.33918851613998413, "logits/rejected": 0.37244075536727905, "logps/chosen": -508.8272399902344, "logps/ref_chosen": -57.450836181640625, "logps/ref_rejected": -94.77339172363281, "logps/rejected": -763.6325073242188, "loss": 0.8291, "margin_dpo/margin_mean": 217.4827117919922, "margin_dpo/margin_std": 190.2075958251953, "step": 608 }, { "epoch": 0.8942731277533039, "fcm_dpo/beta": 0.0037732236087322235, "fcm_dpo/delta": -0.0848221629858017, "fcm_dpo/margin": 245.67294311523438, "fcm_dpo/q_t": 0.3045893907546997, "grad_norm": 58.272987365722656, "learning_rate": 1.7348604439226617e-08, "logits/chosen": 0.3716045022010803, "logits/rejected": 0.3981201648712158, "logps/chosen": -489.30975341796875, "logps/ref_chosen": -58.805355072021484, "logps/ref_rejected": -88.81600952148438, "logps/rejected": -764.9933471679688, "loss": 0.7859, "margin_dpo/margin_mean": 245.67294311523438, "margin_dpo/margin_std": 212.65243530273438, "step": 609 }, { "epoch": 0.895741556534508, "fcm_dpo/beta": 0.003800982143729925, "fcm_dpo/delta": 0.07397213578224182, "fcm_dpo/margin": 205.66021728515625, "fcm_dpo/q_t": 0.3356453776359558, "grad_norm": 50.508609771728516, "learning_rate": 1.6881942648911074e-08, "logits/chosen": 0.3072008490562439, "logits/rejected": 0.3390934467315674, "logps/chosen": -503.4935302734375, "logps/ref_chosen": -65.69503784179688, "logps/ref_rejected": -83.40538787841797, "logps/rejected": -726.8641357421875, "loss": 0.8856, "margin_dpo/margin_mean": 205.66021728515625, "margin_dpo/margin_std": 207.76187133789062, "step": 610 }, { "epoch": 0.8972099853157122, "fcm_dpo/beta": 0.003831944428384304, "fcm_dpo/delta": -0.020623497664928436, "fcm_dpo/margin": 226.4250946044922, "fcm_dpo/q_t": 0.3246605396270752, "grad_norm": 47.413475036621094, "learning_rate": 1.6421423736208e-08, "logits/chosen": 0.37874239683151245, "logits/rejected": 0.39478302001953125, "logps/chosen": -521.7601318359375, "logps/ref_chosen": -52.59946823120117, "logps/ref_rejected": -86.33099365234375, "logps/rejected": -781.916748046875, "loss": 0.864, "margin_dpo/margin_mean": 226.42507934570312, "margin_dpo/margin_std": 232.66787719726562, "step": 611 }, { "epoch": 0.8986784140969163, "fcm_dpo/beta": 0.0037551377899944782, "fcm_dpo/delta": -0.0866103395819664, "fcm_dpo/margin": 247.30072021484375, "fcm_dpo/q_t": 0.3015454411506653, "grad_norm": 33.60853958129883, "learning_rate": 1.5967059836219042e-08, "logits/chosen": 0.4164635241031647, "logits/rejected": 0.4415166974067688, "logps/chosen": -524.9508056640625, "logps/ref_chosen": -59.32372283935547, "logps/ref_rejected": -88.31239318847656, "logps/rejected": -801.240234375, "loss": 0.7631, "margin_dpo/margin_mean": 247.30072021484375, "margin_dpo/margin_std": 186.49124145507812, "step": 612 }, { "epoch": 0.9001468428781204, "fcm_dpo/beta": 0.0037209526635706425, "fcm_dpo/delta": -0.020325224846601486, "fcm_dpo/margin": 233.4252471923828, "fcm_dpo/q_t": 0.3151775300502777, "grad_norm": 41.69904708862305, "learning_rate": 1.551886292185553e-08, "logits/chosen": 0.2626866400241852, "logits/rejected": 0.2640881836414337, "logps/chosen": -483.8834228515625, "logps/ref_chosen": -59.72996520996094, "logps/ref_rejected": -105.10752868652344, "logps/rejected": -762.6862182617188, "loss": 0.7969, "margin_dpo/margin_mean": 233.42526245117188, "margin_dpo/margin_std": 188.75119018554688, "step": 613 }, { "epoch": 0.9016152716593245, "fcm_dpo/beta": 0.003673199564218521, "fcm_dpo/delta": -0.06576398015022278, "fcm_dpo/margin": 247.63795471191406, "fcm_dpo/q_t": 0.3121992349624634, "grad_norm": 54.678871154785156, "learning_rate": 1.507684480352292e-08, "logits/chosen": 0.45456773042678833, "logits/rejected": 0.461800754070282, "logps/chosen": -518.4677124023438, "logps/ref_chosen": -52.93898010253906, "logps/ref_rejected": -104.67938232421875, "logps/rejected": -817.8460693359375, "loss": 0.8123, "margin_dpo/margin_mean": 247.63795471191406, "margin_dpo/margin_std": 225.6822052001953, "step": 614 }, { "epoch": 0.9030837004405287, "fcm_dpo/beta": 0.0037553110159933567, "fcm_dpo/delta": 0.12973493337631226, "fcm_dpo/margin": 194.0587158203125, "fcm_dpo/q_t": 0.34725838899612427, "grad_norm": 37.38533020019531, "learning_rate": 1.4641017128809801e-08, "logits/chosen": 0.2697402536869049, "logits/rejected": 0.28964436054229736, "logps/chosen": -507.75042724609375, "logps/ref_chosen": -65.81727600097656, "logps/ref_rejected": -95.17749786376953, "logps/rejected": -731.1693115234375, "loss": 0.8965, "margin_dpo/margin_mean": 194.0587158203125, "margin_dpo/margin_std": 199.11935424804688, "step": 615 }, { "epoch": 0.9045521292217328, "fcm_dpo/beta": 0.003917084541171789, "fcm_dpo/delta": 0.2312663495540619, "fcm_dpo/margin": 161.20529174804688, "fcm_dpo/q_t": 0.3703739643096924, "grad_norm": 56.471641540527344, "learning_rate": 1.4211391382180637e-08, "logits/chosen": 0.3983857035636902, "logits/rejected": 0.4274812340736389, "logps/chosen": -579.7181396484375, "logps/ref_chosen": -65.13285827636719, "logps/ref_rejected": -74.70050048828125, "logps/rejected": -750.4910888671875, "loss": 0.9979, "margin_dpo/margin_mean": 161.20529174804688, "margin_dpo/margin_std": 210.05697631835938, "step": 616 }, { "epoch": 0.9060205580029369, "fcm_dpo/beta": 0.004059099592268467, "fcm_dpo/delta": 0.15998278558254242, "fcm_dpo/margin": 172.38507080078125, "fcm_dpo/q_t": 0.3537973463535309, "grad_norm": 66.96424865722656, "learning_rate": 1.378797888467345e-08, "logits/chosen": 0.3297281563282013, "logits/rejected": 0.3546701967716217, "logps/chosen": -535.5614013671875, "logps/ref_chosen": -63.005550384521484, "logps/ref_rejected": -64.234130859375, "logps/rejected": -709.175048828125, "loss": 0.932, "margin_dpo/margin_mean": 172.38510131835938, "margin_dpo/margin_std": 191.4333953857422, "step": 617 }, { "epoch": 0.9074889867841409, "fcm_dpo/beta": 0.004026922397315502, "fcm_dpo/delta": -0.12715700268745422, "fcm_dpo/margin": 239.53054809570312, "fcm_dpo/q_t": 0.3070314824581146, "grad_norm": 62.99699020385742, "learning_rate": 1.3370790793601371e-08, "logits/chosen": 0.33015531301498413, "logits/rejected": 0.37518373131752014, "logps/chosen": -553.628173828125, "logps/ref_chosen": -67.10134887695312, "logps/ref_rejected": -92.15340423583984, "logps/rejected": -818.2107543945312, "loss": 0.823, "margin_dpo/margin_mean": 239.53053283691406, "margin_dpo/margin_std": 232.60379028320312, "step": 618 }, { "epoch": 0.908957415565345, "fcm_dpo/beta": 0.00403250940144062, "fcm_dpo/delta": 0.08001011610031128, "fcm_dpo/margin": 192.30813598632812, "fcm_dpo/q_t": 0.35367268323898315, "grad_norm": 59.05690383911133, "learning_rate": 1.2959838102258535e-08, "logits/chosen": 0.38170328736305237, "logits/rejected": 0.4015880227088928, "logps/chosen": -535.88916015625, "logps/ref_chosen": -55.978233337402344, "logps/ref_rejected": -93.1854019165039, "logps/rejected": -765.404541015625, "loss": 0.982, "margin_dpo/margin_mean": 192.30813598632812, "margin_dpo/margin_std": 265.18170166015625, "step": 619 }, { "epoch": 0.9104258443465492, "fcm_dpo/beta": 0.004014415666460991, "fcm_dpo/delta": 0.02012522518634796, "fcm_dpo/margin": 206.92474365234375, "fcm_dpo/q_t": 0.3321415185928345, "grad_norm": 58.67335891723633, "learning_rate": 1.2555131639630567e-08, "logits/chosen": 0.38881373405456543, "logits/rejected": 0.41763806343078613, "logps/chosen": -512.077880859375, "logps/ref_chosen": -59.79750061035156, "logps/ref_rejected": -78.41075134277344, "logps/rejected": -737.6159057617188, "loss": 0.8791, "margin_dpo/margin_mean": 206.92474365234375, "margin_dpo/margin_std": 222.4522705078125, "step": 620 }, { "epoch": 0.9118942731277533, "fcm_dpo/beta": 0.003981994464993477, "fcm_dpo/delta": -0.167287215590477, "fcm_dpo/margin": 251.25936889648438, "fcm_dpo/q_t": 0.2956671714782715, "grad_norm": 49.28880310058594, "learning_rate": 1.2156682070109086e-08, "logits/chosen": 0.4875348210334778, "logits/rejected": 0.504463255405426, "logps/chosen": -502.497802734375, "logps/ref_chosen": -53.93375778198242, "logps/ref_rejected": -88.36951446533203, "logps/rejected": -788.1929321289062, "loss": 0.7763, "margin_dpo/margin_mean": 251.25936889648438, "margin_dpo/margin_std": 220.83163452148438, "step": 621 }, { "epoch": 0.9133627019089574, "fcm_dpo/beta": 0.0038708075881004333, "fcm_dpo/delta": -0.01944926381111145, "fcm_dpo/margin": 223.79840087890625, "fcm_dpo/q_t": 0.3231251835823059, "grad_norm": 44.725032806396484, "learning_rate": 1.1764499893210878e-08, "logits/chosen": 0.297042578458786, "logits/rejected": 0.3068755269050598, "logps/chosen": -501.72003173828125, "logps/ref_chosen": -60.28582000732422, "logps/ref_rejected": -85.51873779296875, "logps/rejected": -750.7514038085938, "loss": 0.8368, "margin_dpo/margin_mean": 223.79840087890625, "margin_dpo/margin_std": 217.63043212890625, "step": 622 }, { "epoch": 0.9148311306901615, "fcm_dpo/beta": 0.003980166278779507, "fcm_dpo/delta": 0.1460477113723755, "fcm_dpo/margin": 179.51904296875, "fcm_dpo/q_t": 0.3587492108345032, "grad_norm": 75.94093322753906, "learning_rate": 1.1378595443300998e-08, "logits/chosen": 0.4989446997642517, "logits/rejected": 0.5313464403152466, "logps/chosen": -542.784423828125, "logps/ref_chosen": -64.1569595336914, "logps/ref_rejected": -85.08304595947266, "logps/rejected": -743.2296142578125, "loss": 0.9751, "margin_dpo/margin_mean": 179.51904296875, "margin_dpo/margin_std": 242.48092651367188, "step": 623 }, { "epoch": 0.9162995594713657, "fcm_dpo/beta": 0.0039426954463124275, "fcm_dpo/delta": -0.11975458264350891, "fcm_dpo/margin": 243.0377655029297, "fcm_dpo/q_t": 0.30364617705345154, "grad_norm": 53.13152313232422, "learning_rate": 1.0998978889320582e-08, "logits/chosen": 0.29805538058280945, "logits/rejected": 0.33957502245903015, "logps/chosen": -545.13232421875, "logps/ref_chosen": -71.91862487792969, "logps/ref_rejected": -97.13203430175781, "logps/rejected": -813.383544921875, "loss": 0.8084, "margin_dpo/margin_mean": 243.03778076171875, "margin_dpo/margin_std": 226.8072509765625, "step": 624 }, { "epoch": 0.9177679882525698, "fcm_dpo/beta": 0.003897000104188919, "fcm_dpo/delta": -0.03251214325428009, "fcm_dpo/margin": 225.71258544921875, "fcm_dpo/q_t": 0.3202013373374939, "grad_norm": 51.93879699707031, "learning_rate": 1.0625660234518913e-08, "logits/chosen": 0.36836007237434387, "logits/rejected": 0.4131643772125244, "logps/chosen": -496.9722900390625, "logps/ref_chosen": -58.342071533203125, "logps/ref_rejected": -86.09038543701172, "logps/rejected": -750.4332275390625, "loss": 0.8224, "margin_dpo/margin_mean": 225.71258544921875, "margin_dpo/margin_std": 214.71905517578125, "step": 625 }, { "epoch": 0.9192364170337739, "fcm_dpo/beta": 0.004009204916656017, "fcm_dpo/delta": 0.19411587715148926, "fcm_dpo/margin": 166.82199096679688, "fcm_dpo/q_t": 0.36621803045272827, "grad_norm": 52.73582458496094, "learning_rate": 1.0258649316189721e-08, "logits/chosen": 0.229852557182312, "logits/rejected": 0.24690240621566772, "logps/chosen": -562.8541259765625, "logps/ref_chosen": -75.11260986328125, "logps/ref_rejected": -99.188720703125, "logps/rejected": -753.752197265625, "loss": 0.9865, "margin_dpo/margin_mean": 166.82199096679688, "margin_dpo/margin_std": 216.57113647460938, "step": 626 }, { "epoch": 0.920704845814978, "fcm_dpo/beta": 0.003970067948102951, "fcm_dpo/delta": -0.1275482475757599, "fcm_dpo/margin": 243.10336303710938, "fcm_dpo/q_t": 0.31875503063201904, "grad_norm": 54.12873077392578, "learning_rate": 9.897955805412e-09, "logits/chosen": 0.2737880349159241, "logits/rejected": 0.2716522216796875, "logps/chosen": -457.3752136230469, "logps/ref_chosen": -47.74314880371094, "logps/ref_rejected": -106.75448608398438, "logps/rejected": -759.4898681640625, "loss": 0.8554, "margin_dpo/margin_mean": 243.1033477783203, "margin_dpo/margin_std": 278.228759765625, "step": 627 }, { "epoch": 0.922173274596182, "fcm_dpo/beta": 0.003842130536213517, "fcm_dpo/delta": -0.12872882187366486, "fcm_dpo/margin": 250.9643096923828, "fcm_dpo/q_t": 0.30476510524749756, "grad_norm": 57.179996490478516, "learning_rate": 9.543589206795238e-09, "logits/chosen": 0.40602755546569824, "logits/rejected": 0.43877482414245605, "logps/chosen": -525.9154663085938, "logps/ref_chosen": -60.182945251464844, "logps/ref_rejected": -101.55467224121094, "logps/rejected": -818.25146484375, "loss": 0.7991, "margin_dpo/margin_mean": 250.9643096923828, "margin_dpo/margin_std": 231.17425537109375, "step": 628 }, { "epoch": 0.9236417033773862, "fcm_dpo/beta": 0.0038544987328350544, "fcm_dpo/delta": 0.02374306507408619, "fcm_dpo/margin": 214.87026977539062, "fcm_dpo/q_t": 0.3250948488712311, "grad_norm": 44.98953628540039, "learning_rate": 9.19555885822887e-09, "logits/chosen": 0.38332074880599976, "logits/rejected": 0.4074392020702362, "logps/chosen": -532.5030517578125, "logps/ref_chosen": -64.21354675292969, "logps/ref_rejected": -91.65367126464844, "logps/rejected": -774.8133544921875, "loss": 0.8342, "margin_dpo/margin_mean": 214.87026977539062, "margin_dpo/margin_std": 191.0750732421875, "step": 629 }, { "epoch": 0.9251101321585903, "fcm_dpo/beta": 0.004004511050879955, "fcm_dpo/delta": 0.21656584739685059, "fcm_dpo/margin": 161.44676208496094, "fcm_dpo/q_t": 0.37145158648490906, "grad_norm": 80.25105285644531, "learning_rate": 8.85387393063622e-09, "logits/chosen": 0.24440869688987732, "logits/rejected": 0.2650578022003174, "logps/chosen": -476.164306640625, "logps/ref_chosen": -59.29100036621094, "logps/ref_rejected": -83.59829711914062, "logps/rejected": -661.9183349609375, "loss": 1.0023, "margin_dpo/margin_mean": 161.44677734375, "margin_dpo/margin_std": 225.14974975585938, "step": 630 }, { "epoch": 0.9265785609397944, "fcm_dpo/beta": 0.004093495197594166, "fcm_dpo/delta": 0.07495088130235672, "fcm_dpo/margin": 190.66612243652344, "fcm_dpo/q_t": 0.345422625541687, "grad_norm": 56.53142166137695, "learning_rate": 8.518543427732949e-09, "logits/chosen": 0.4773872494697571, "logits/rejected": 0.5107876658439636, "logps/chosen": -547.016845703125, "logps/ref_chosen": -59.45360565185547, "logps/ref_rejected": -80.95156860351562, "logps/rejected": -759.1809692382812, "loss": 0.9594, "margin_dpo/margin_mean": 190.6661376953125, "margin_dpo/margin_std": 245.64004516601562, "step": 631 }, { "epoch": 0.9280469897209985, "fcm_dpo/beta": 0.004062248859554529, "fcm_dpo/delta": -0.014684807509183884, "fcm_dpo/margin": 212.34161376953125, "fcm_dpo/q_t": 0.3262500762939453, "grad_norm": 50.996089935302734, "learning_rate": 8.189576185789637e-09, "logits/chosen": 0.4748516082763672, "logits/rejected": 0.5130486488342285, "logps/chosen": -536.0592041015625, "logps/ref_chosen": -61.35155487060547, "logps/ref_rejected": -86.16017150878906, "logps/rejected": -773.20947265625, "loss": 0.9309, "margin_dpo/margin_mean": 212.34161376953125, "margin_dpo/margin_std": 257.6206359863281, "step": 632 }, { "epoch": 0.9295154185022027, "fcm_dpo/beta": 0.004250052385032177, "fcm_dpo/delta": 0.20158423483371735, "fcm_dpo/margin": 154.8126678466797, "fcm_dpo/q_t": 0.3648519217967987, "grad_norm": 59.050899505615234, "learning_rate": 7.866980873399015e-09, "logits/chosen": 0.40178602933883667, "logits/rejected": 0.40316498279571533, "logps/chosen": -537.0020751953125, "logps/ref_chosen": -57.27816390991211, "logps/ref_rejected": -91.58395385742188, "logps/rejected": -726.1204833984375, "loss": 1.0244, "margin_dpo/margin_mean": 154.8126678466797, "margin_dpo/margin_std": 218.823486328125, "step": 633 }, { "epoch": 0.9309838472834068, "fcm_dpo/beta": 0.004316710866987705, "fcm_dpo/delta": 0.12352225929498672, "fcm_dpo/margin": 170.43814086914062, "fcm_dpo/q_t": 0.35788172483444214, "grad_norm": 69.04047393798828, "learning_rate": 7.550765991247654e-09, "logits/chosen": 0.33094990253448486, "logits/rejected": 0.3353114426136017, "logps/chosen": -551.9486694335938, "logps/ref_chosen": -66.61896514892578, "logps/ref_rejected": -107.12564849853516, "logps/rejected": -762.8934936523438, "loss": 0.9679, "margin_dpo/margin_mean": 170.43814086914062, "margin_dpo/margin_std": 223.65484619140625, "step": 634 }, { "epoch": 0.9324522760646109, "fcm_dpo/beta": 0.00439867377281189, "fcm_dpo/delta": 0.05664564669132233, "fcm_dpo/margin": 181.37939453125, "fcm_dpo/q_t": 0.34000444412231445, "grad_norm": 53.52330017089844, "learning_rate": 7.240939871891699e-09, "logits/chosen": 0.2910565435886383, "logits/rejected": 0.3432241678237915, "logps/chosen": -531.345703125, "logps/ref_chosen": -73.95551300048828, "logps/ref_rejected": -82.50045776367188, "logps/rejected": -721.2700805664062, "loss": 0.8902, "margin_dpo/margin_mean": 181.37940979003906, "margin_dpo/margin_std": 200.13851928710938, "step": 635 }, { "epoch": 0.933920704845815, "fcm_dpo/beta": 0.004396742209792137, "fcm_dpo/delta": -0.09724476933479309, "fcm_dpo/margin": 213.02749633789062, "fcm_dpo/q_t": 0.31033921241760254, "grad_norm": 61.685367584228516, "learning_rate": 6.937510679537628e-09, "logits/chosen": 0.26561087369918823, "logits/rejected": 0.2920706868171692, "logps/chosen": -503.4400329589844, "logps/ref_chosen": -59.628910064697266, "logps/ref_rejected": -81.97883605957031, "logps/rejected": -738.8173828125, "loss": 0.8165, "margin_dpo/margin_mean": 213.02749633789062, "margin_dpo/margin_std": 203.03334045410156, "step": 636 }, { "epoch": 0.9353891336270191, "fcm_dpo/beta": 0.00424953643232584, "fcm_dpo/delta": -0.104090616106987, "fcm_dpo/margin": 222.14358520507812, "fcm_dpo/q_t": 0.31200093030929565, "grad_norm": 56.29432678222656, "learning_rate": 6.640486409826785e-09, "logits/chosen": 0.47558510303497314, "logits/rejected": 0.4967935085296631, "logps/chosen": -525.0879516601562, "logps/ref_chosen": -49.652687072753906, "logps/ref_rejected": -98.40513610839844, "logps/rejected": -795.9840087890625, "loss": 0.8226, "margin_dpo/margin_mean": 222.14358520507812, "margin_dpo/margin_std": 222.4608154296875, "step": 637 }, { "epoch": 0.9368575624082232, "fcm_dpo/beta": 0.004241650924086571, "fcm_dpo/delta": -0.08875752985477448, "fcm_dpo/margin": 218.70045471191406, "fcm_dpo/q_t": 0.30789196491241455, "grad_norm": 49.7493896484375, "learning_rate": 6.349874889624962e-09, "logits/chosen": 0.3231136202812195, "logits/rejected": 0.34823527932167053, "logps/chosen": -474.68023681640625, "logps/ref_chosen": -58.156639099121094, "logps/ref_rejected": -79.3014907836914, "logps/rejected": -714.5255737304688, "loss": 0.7903, "margin_dpo/margin_mean": 218.700439453125, "margin_dpo/margin_std": 176.45599365234375, "step": 638 }, { "epoch": 0.9383259911894273, "fcm_dpo/beta": 0.004274230450391769, "fcm_dpo/delta": 0.15763217210769653, "fcm_dpo/margin": 164.3897247314453, "fcm_dpo/q_t": 0.3588542342185974, "grad_norm": 71.79363250732422, "learning_rate": 6.065683776815933e-09, "logits/chosen": 0.2641649842262268, "logits/rejected": 0.2947526276111603, "logps/chosen": -589.9560546875, "logps/ref_chosen": -72.32319641113281, "logps/ref_rejected": -74.2749252319336, "logps/rejected": -756.2974853515625, "loss": 0.9909, "margin_dpo/margin_mean": 164.38970947265625, "margin_dpo/margin_std": 221.13119506835938, "step": 639 }, { "epoch": 0.9397944199706314, "fcm_dpo/beta": 0.004170107655227184, "fcm_dpo/delta": -0.2509539723396301, "fcm_dpo/margin": 257.6911926269531, "fcm_dpo/q_t": 0.2941976487636566, "grad_norm": 45.10141372680664, "learning_rate": 5.7879205600998296e-09, "logits/chosen": 0.2849113941192627, "logits/rejected": 0.3171135187149048, "logps/chosen": -502.7310485839844, "logps/ref_chosen": -56.13436508178711, "logps/ref_rejected": -108.60014343261719, "logps/rejected": -812.8880004882812, "loss": 0.773, "margin_dpo/margin_mean": 257.691162109375, "margin_dpo/margin_std": 253.909423828125, "step": 640 }, { "epoch": 0.9412628487518355, "fcm_dpo/beta": 0.004141503944993019, "fcm_dpo/delta": 0.13046443462371826, "fcm_dpo/margin": 176.06033325195312, "fcm_dpo/q_t": 0.3567558825016022, "grad_norm": 63.07670593261719, "learning_rate": 5.516592558795746e-09, "logits/chosen": 0.3889944553375244, "logits/rejected": 0.41144323348999023, "logps/chosen": -580.721923828125, "logps/ref_chosen": -64.99689483642578, "logps/ref_rejected": -86.99232482910156, "logps/rejected": -778.7777099609375, "loss": 0.975, "margin_dpo/margin_mean": 176.06033325195312, "margin_dpo/margin_std": 239.46539306640625, "step": 641 }, { "epoch": 0.9427312775330396, "fcm_dpo/beta": 0.004168546758592129, "fcm_dpo/delta": -0.10531463474035263, "fcm_dpo/margin": 226.55740356445312, "fcm_dpo/q_t": 0.3239438533782959, "grad_norm": 55.98142623901367, "learning_rate": 5.251706922648868e-09, "logits/chosen": 0.3034379482269287, "logits/rejected": 0.32045918703079224, "logps/chosen": -527.4705810546875, "logps/ref_chosen": -65.68924713134766, "logps/ref_rejected": -110.24205017089844, "logps/rejected": -798.580810546875, "loss": 0.8795, "margin_dpo/margin_mean": 226.5574188232422, "margin_dpo/margin_std": 269.38568115234375, "step": 642 }, { "epoch": 0.9441997063142438, "fcm_dpo/beta": 0.0040768347680568695, "fcm_dpo/delta": 0.026233568787574768, "fcm_dpo/margin": 202.13023376464844, "fcm_dpo/q_t": 0.3279660940170288, "grad_norm": 47.93162536621094, "learning_rate": 4.993270631642038e-09, "logits/chosen": 0.41052716970443726, "logits/rejected": 0.4136253595352173, "logps/chosen": -503.70599365234375, "logps/ref_chosen": -51.94999694824219, "logps/ref_rejected": -87.46833801269531, "logps/rejected": -741.3546142578125, "loss": 0.8582, "margin_dpo/margin_mean": 202.1302490234375, "margin_dpo/margin_std": 190.30067443847656, "step": 643 }, { "epoch": 0.9456681350954479, "fcm_dpo/beta": 0.004156759940087795, "fcm_dpo/delta": 0.03570423275232315, "fcm_dpo/margin": 196.5560302734375, "fcm_dpo/q_t": 0.33971697092056274, "grad_norm": 55.22520446777344, "learning_rate": 4.741290495811873e-09, "logits/chosen": 0.30671143531799316, "logits/rejected": 0.3282683193683624, "logps/chosen": -509.8187255859375, "logps/ref_chosen": -59.017662048339844, "logps/ref_rejected": -87.13668823242188, "logps/rejected": -734.4937744140625, "loss": 0.9303, "margin_dpo/margin_mean": 196.55601501464844, "margin_dpo/margin_std": 242.78518676757812, "step": 644 }, { "epoch": 0.947136563876652, "fcm_dpo/beta": 0.004250319674611092, "fcm_dpo/delta": 0.20495975017547607, "fcm_dpo/margin": 155.01222229003906, "fcm_dpo/q_t": 0.37516582012176514, "grad_norm": 72.1839828491211, "learning_rate": 4.495773155069299e-09, "logits/chosen": 0.2767280340194702, "logits/rejected": 0.2697181701660156, "logps/chosen": -518.177978515625, "logps/ref_chosen": -55.87602233886719, "logps/ref_rejected": -97.78080749511719, "logps/rejected": -715.094970703125, "loss": 1.0398, "margin_dpo/margin_mean": 155.01222229003906, "margin_dpo/margin_std": 236.4263153076172, "step": 645 }, { "epoch": 0.9486049926578561, "fcm_dpo/beta": 0.004299083724617958, "fcm_dpo/delta": 0.007297724485397339, "fcm_dpo/margin": 195.88906860351562, "fcm_dpo/q_t": 0.32827213406562805, "grad_norm": 65.47455596923828, "learning_rate": 4.256725079024553e-09, "logits/chosen": 0.431162565946579, "logits/rejected": 0.451549768447876, "logps/chosen": -525.3079223632812, "logps/ref_chosen": -61.275787353515625, "logps/ref_rejected": -77.50580596923828, "logps/rejected": -737.427001953125, "loss": 0.8659, "margin_dpo/margin_mean": 195.88906860351562, "margin_dpo/margin_std": 201.49288940429688, "step": 646 }, { "epoch": 0.9500734214390602, "fcm_dpo/beta": 0.004347026348114014, "fcm_dpo/delta": 0.009064847603440285, "fcm_dpo/margin": 193.61483764648438, "fcm_dpo/q_t": 0.33186376094818115, "grad_norm": 48.380271911621094, "learning_rate": 4.024152566816791e-09, "logits/chosen": 0.3102216124534607, "logits/rejected": 0.3185359835624695, "logps/chosen": -478.88702392578125, "logps/ref_chosen": -54.8524169921875, "logps/ref_rejected": -93.5194091796875, "logps/rejected": -711.1688232421875, "loss": 0.902, "margin_dpo/margin_mean": 193.61483764648438, "margin_dpo/margin_std": 220.6673583984375, "step": 647 }, { "epoch": 0.9515418502202643, "fcm_dpo/beta": 0.004186989739537239, "fcm_dpo/delta": -0.2561682164669037, "fcm_dpo/margin": 257.642578125, "fcm_dpo/q_t": 0.30367809534072876, "grad_norm": 43.26372528076172, "learning_rate": 3.798061746947995e-09, "logits/chosen": 0.40941837430000305, "logits/rejected": 0.4284679889678955, "logps/chosen": -514.1665649414062, "logps/ref_chosen": -54.17146682739258, "logps/ref_rejected": -98.7127914428711, "logps/rejected": -816.3504638671875, "loss": 0.8326, "margin_dpo/margin_mean": 257.642578125, "margin_dpo/margin_std": 296.09918212890625, "step": 648 }, { "epoch": 0.9530102790014684, "fcm_dpo/beta": 0.004244859330356121, "fcm_dpo/delta": 0.16276229918003082, "fcm_dpo/margin": 164.3397979736328, "fcm_dpo/q_t": 0.36515751481056213, "grad_norm": 71.0632553100586, "learning_rate": 3.5784585771215235e-09, "logits/chosen": 0.400471031665802, "logits/rejected": 0.4290485382080078, "logps/chosen": -526.2083740234375, "logps/ref_chosen": -62.480350494384766, "logps/ref_rejected": -80.07717895507812, "logps/rejected": -708.14501953125, "loss": 0.9944, "margin_dpo/margin_mean": 164.33978271484375, "margin_dpo/margin_std": 230.75076293945312, "step": 649 }, { "epoch": 0.9544787077826725, "fcm_dpo/beta": 0.004142679274082184, "fcm_dpo/delta": -0.1436581313610077, "fcm_dpo/margin": 235.9899139404297, "fcm_dpo/q_t": 0.31016072630882263, "grad_norm": 54.92049789428711, "learning_rate": 3.3653488440851253e-09, "logits/chosen": 0.4193887710571289, "logits/rejected": 0.43305301666259766, "logps/chosen": -511.2436828613281, "logps/ref_chosen": -56.09281921386719, "logps/ref_rejected": -98.26483917236328, "logps/rejected": -789.4055786132812, "loss": 0.8367, "margin_dpo/margin_mean": 235.98989868164062, "margin_dpo/margin_std": 247.71554565429688, "step": 650 }, { "epoch": 0.9559471365638766, "fcm_dpo/beta": 0.004029122181236744, "fcm_dpo/delta": -0.15634776651859283, "fcm_dpo/margin": 245.6698455810547, "fcm_dpo/q_t": 0.29818981885910034, "grad_norm": 70.62873077392578, "learning_rate": 3.158738163478475e-09, "logits/chosen": 0.38377705216407776, "logits/rejected": 0.3832590878009796, "logps/chosen": -448.92266845703125, "logps/ref_chosen": -43.42544937133789, "logps/ref_rejected": -99.95791625976562, "logps/rejected": -751.1248779296875, "loss": 0.7724, "margin_dpo/margin_mean": 245.66983032226562, "margin_dpo/margin_std": 212.33193969726562, "step": 651 }, { "epoch": 0.9574155653450808, "fcm_dpo/beta": 0.004000368528068066, "fcm_dpo/delta": -0.031687185168266296, "fcm_dpo/margin": 219.70718383789062, "fcm_dpo/q_t": 0.32615405321121216, "grad_norm": 50.73948287963867, "learning_rate": 2.9586319796851555e-09, "logits/chosen": 0.328296959400177, "logits/rejected": 0.3477246165275574, "logps/chosen": -495.43768310546875, "logps/ref_chosen": -62.57680892944336, "logps/ref_rejected": -111.76779174804688, "logps/rejected": -764.3358154296875, "loss": 0.8531, "margin_dpo/margin_mean": 219.70718383789062, "margin_dpo/margin_std": 232.84384155273438, "step": 652 }, { "epoch": 0.9588839941262849, "fcm_dpo/beta": 0.003977050073444843, "fcm_dpo/delta": -0.02044288069009781, "fcm_dpo/margin": 218.42001342773438, "fcm_dpo/q_t": 0.3240339756011963, "grad_norm": 41.00118637084961, "learning_rate": 2.7650355656892166e-09, "logits/chosen": 0.3624820113182068, "logits/rejected": 0.36829936504364014, "logps/chosen": -525.9445190429688, "logps/ref_chosen": -61.11295700073242, "logps/ref_rejected": -103.24960327148438, "logps/rejected": -786.501220703125, "loss": 0.8455, "margin_dpo/margin_mean": 218.42002868652344, "margin_dpo/margin_std": 219.990966796875, "step": 653 }, { "epoch": 0.960352422907489, "fcm_dpo/beta": 0.004003086127340794, "fcm_dpo/delta": 0.099447101354599, "fcm_dpo/margin": 189.29266357421875, "fcm_dpo/q_t": 0.35448741912841797, "grad_norm": 52.74611282348633, "learning_rate": 2.577954022936174e-09, "logits/chosen": 0.3803136944770813, "logits/rejected": 0.3972058892250061, "logps/chosen": -546.88037109375, "logps/ref_chosen": -61.7281379699707, "logps/ref_rejected": -98.7738037109375, "logps/rejected": -773.2186889648438, "loss": 0.9898, "margin_dpo/margin_mean": 189.29266357421875, "margin_dpo/margin_std": 264.4116516113281, "step": 654 }, { "epoch": 0.9618208516886931, "fcm_dpo/beta": 0.0040547847747802734, "fcm_dpo/delta": 0.04648479074239731, "fcm_dpo/margin": 198.96510314941406, "fcm_dpo/q_t": 0.34709632396698, "grad_norm": 67.65664672851562, "learning_rate": 2.397392281198729e-09, "logits/chosen": 0.45737969875335693, "logits/rejected": 0.46436363458633423, "logps/chosen": -495.4986572265625, "logps/ref_chosen": -49.576812744140625, "logps/ref_rejected": -98.29183197021484, "logps/rejected": -743.1787719726562, "loss": 0.9368, "margin_dpo/margin_mean": 198.96511840820312, "margin_dpo/margin_std": 253.21934509277344, "step": 655 }, { "epoch": 0.9632892804698973, "fcm_dpo/beta": 0.003890886902809143, "fcm_dpo/delta": -0.3531903326511383, "fcm_dpo/margin": 298.80157470703125, "fcm_dpo/q_t": 0.26101160049438477, "grad_norm": 74.69690704345703, "learning_rate": 2.223355098446622e-09, "logits/chosen": 0.3172380328178406, "logits/rejected": 0.3397536277770996, "logps/chosen": -491.11737060546875, "logps/ref_chosen": -52.54943084716797, "logps/ref_rejected": -113.67464447021484, "logps/rejected": -851.044189453125, "loss": 0.6594, "margin_dpo/margin_mean": 298.80157470703125, "margin_dpo/margin_std": 208.3951416015625, "step": 656 }, { "epoch": 0.9647577092511013, "fcm_dpo/beta": 0.0036947480402886868, "fcm_dpo/delta": -0.17029432952404022, "fcm_dpo/margin": 271.0969543457031, "fcm_dpo/q_t": 0.2951889634132385, "grad_norm": 44.1884880065918, "learning_rate": 2.055847060721566e-09, "logits/chosen": 0.4047412574291229, "logits/rejected": 0.41397354006767273, "logps/chosen": -476.5009460449219, "logps/ref_chosen": -46.700538635253906, "logps/ref_rejected": -97.91487121582031, "logps/rejected": -798.812255859375, "loss": 0.7586, "margin_dpo/margin_mean": 271.0969543457031, "margin_dpo/margin_std": 232.99600219726562, "step": 657 }, { "epoch": 0.9662261380323054, "fcm_dpo/beta": 0.0036933058872818947, "fcm_dpo/delta": 0.026351526379585266, "fcm_dpo/margin": 223.59996032714844, "fcm_dpo/q_t": 0.32264280319213867, "grad_norm": 44.01875305175781, "learning_rate": 1.8948725820160662e-09, "logits/chosen": 0.4098268747329712, "logits/rejected": 0.4359194040298462, "logps/chosen": -537.1385498046875, "logps/ref_chosen": -60.95820999145508, "logps/ref_rejected": -95.93949127197266, "logps/rejected": -795.7197875976562, "loss": 0.8373, "margin_dpo/margin_mean": 223.59996032714844, "margin_dpo/margin_std": 197.29067993164062, "step": 658 }, { "epoch": 0.9676945668135095, "fcm_dpo/beta": 0.00366408284753561, "fcm_dpo/delta": -0.029130063951015472, "fcm_dpo/margin": 238.9795379638672, "fcm_dpo/q_t": 0.31922781467437744, "grad_norm": 37.774574279785156, "learning_rate": 1.7404359041573723e-09, "logits/chosen": 0.2552376389503479, "logits/rejected": 0.3021022081375122, "logps/chosen": -518.0164794921875, "logps/ref_chosen": -76.74298095703125, "logps/ref_rejected": -87.4709701538086, "logps/rejected": -767.7239990234375, "loss": 0.8311, "margin_dpo/margin_mean": 238.97952270507812, "margin_dpo/margin_std": 225.7298583984375, "step": 659 }, { "epoch": 0.9691629955947136, "fcm_dpo/beta": 0.0036529982462525368, "fcm_dpo/delta": -0.053744181990623474, "fcm_dpo/margin": 246.07659912109375, "fcm_dpo/q_t": 0.3094879984855652, "grad_norm": 46.95237731933594, "learning_rate": 1.592541096695571e-09, "logits/chosen": 0.4749673008918762, "logits/rejected": 0.4998527765274048, "logps/chosen": -520.5059814453125, "logps/ref_chosen": -59.04788589477539, "logps/ref_rejected": -75.96005249023438, "logps/rejected": -783.4947509765625, "loss": 0.7949, "margin_dpo/margin_mean": 246.07659912109375, "margin_dpo/margin_std": 207.9267578125, "step": 660 }, { "epoch": 0.9706314243759178, "fcm_dpo/beta": 0.0036703585647046566, "fcm_dpo/delta": -0.015187568962574005, "fcm_dpo/margin": 234.8988037109375, "fcm_dpo/q_t": 0.3215470314025879, "grad_norm": 45.32716751098633, "learning_rate": 1.4511920567963908e-09, "logits/chosen": 0.3362319767475128, "logits/rejected": 0.3641771674156189, "logps/chosen": -475.05731201171875, "logps/ref_chosen": -50.673973083496094, "logps/ref_rejected": -86.00569152832031, "logps/rejected": -745.287841796875, "loss": 0.8449, "margin_dpo/margin_mean": 234.89878845214844, "margin_dpo/margin_std": 223.83566284179688, "step": 661 }, { "epoch": 0.9720998531571219, "fcm_dpo/beta": 0.0037050643004477024, "fcm_dpo/delta": 0.12543530762195587, "fcm_dpo/margin": 197.98480224609375, "fcm_dpo/q_t": 0.34611234068870544, "grad_norm": 41.252899169921875, "learning_rate": 1.3163925091384532e-09, "logits/chosen": 0.2565455436706543, "logits/rejected": 0.2876202464103699, "logps/chosen": -525.746337890625, "logps/ref_chosen": -69.26106262207031, "logps/ref_rejected": -89.05593872070312, "logps/rejected": -743.5260009765625, "loss": 0.9211, "margin_dpo/margin_mean": 197.98480224609375, "margin_dpo/margin_std": 217.1225128173828, "step": 662 }, { "epoch": 0.973568281938326, "fcm_dpo/beta": 0.0036978046409785748, "fcm_dpo/delta": -0.027755947783589363, "fcm_dpo/margin": 236.65377807617188, "fcm_dpo/q_t": 0.325141966342926, "grad_norm": 35.571556091308594, "learning_rate": 1.1881460058152382e-09, "logits/chosen": 0.26951783895492554, "logits/rejected": 0.2777261435985565, "logps/chosen": -485.85986328125, "logps/ref_chosen": -64.87890625, "logps/ref_rejected": -113.92536926269531, "logps/rejected": -771.56005859375, "loss": 0.8532, "margin_dpo/margin_mean": 236.65377807617188, "margin_dpo/margin_std": 255.65853881835938, "step": 663 }, { "epoch": 0.9750367107195301, "fcm_dpo/beta": 0.0036702845245599747, "fcm_dpo/delta": -0.040541913360357285, "fcm_dpo/margin": 241.583984375, "fcm_dpo/q_t": 0.31657952070236206, "grad_norm": 57.33827209472656, "learning_rate": 1.066455926241383e-09, "logits/chosen": 0.39595216512680054, "logits/rejected": 0.41890090703964233, "logps/chosen": -528.522705078125, "logps/ref_chosen": -60.88847351074219, "logps/ref_rejected": -105.521728515625, "logps/rejected": -814.739990234375, "loss": 0.856, "margin_dpo/margin_mean": 241.583984375, "margin_dpo/margin_std": 250.49679565429688, "step": 664 }, { "epoch": 0.9765051395007343, "fcm_dpo/beta": 0.0036611277610063553, "fcm_dpo/delta": -0.006559427827596664, "fcm_dpo/margin": 233.7760009765625, "fcm_dpo/q_t": 0.321765661239624, "grad_norm": 42.60844039916992, "learning_rate": 9.513254770636137e-10, "logits/chosen": 0.516608715057373, "logits/rejected": 0.5561769604682922, "logps/chosen": -504.7494201660156, "logps/ref_chosen": -60.56413269042969, "logps/ref_rejected": -84.80882263183594, "logps/rejected": -762.77001953125, "loss": 0.8261, "margin_dpo/margin_mean": 233.77598571777344, "margin_dpo/margin_std": 211.45077514648438, "step": 665 }, { "epoch": 0.9779735682819384, "fcm_dpo/beta": 0.0036832215264439583, "fcm_dpo/delta": 0.041882507503032684, "fcm_dpo/margin": 220.32644653320312, "fcm_dpo/q_t": 0.3285210132598877, "grad_norm": 48.15813064575195, "learning_rate": 8.427576920763956e-10, "logits/chosen": 0.287090539932251, "logits/rejected": 0.313105046749115, "logps/chosen": -517.513671875, "logps/ref_chosen": -64.41996002197266, "logps/ref_rejected": -95.8916244506836, "logps/rejected": -769.311767578125, "loss": 0.8545, "margin_dpo/margin_mean": 220.32644653320312, "margin_dpo/margin_std": 204.0716094970703, "step": 666 }, { "epoch": 0.9794419970631424, "fcm_dpo/beta": 0.0037040242459625006, "fcm_dpo/delta": 0.003546607680618763, "fcm_dpo/margin": 228.5996856689453, "fcm_dpo/q_t": 0.32051968574523926, "grad_norm": 49.475914001464844, "learning_rate": 7.407554321417764e-10, "logits/chosen": 0.30403217673301697, "logits/rejected": 0.3480105996131897, "logps/chosen": -546.994873046875, "logps/ref_chosen": -69.27702331542969, "logps/ref_rejected": -87.83549499511719, "logps/rejected": -794.1530151367188, "loss": 0.8222, "margin_dpo/margin_mean": 228.5996856689453, "margin_dpo/margin_std": 200.08045959472656, "step": 667 }, { "epoch": 0.9809104258443465, "fcm_dpo/beta": 0.0037916412111371756, "fcm_dpo/delta": 0.11322961747646332, "fcm_dpo/margin": 196.06912231445312, "fcm_dpo/q_t": 0.35045015811920166, "grad_norm": 81.19744110107422, "learning_rate": 6.453213851142225e-10, "logits/chosen": 0.3269076347351074, "logits/rejected": 0.3389139771461487, "logps/chosen": -557.981689453125, "logps/ref_chosen": -72.60400390625, "logps/ref_rejected": -103.73905944824219, "logps/rejected": -785.1858520507812, "loss": 0.9587, "margin_dpo/margin_mean": 196.06912231445312, "margin_dpo/margin_std": 250.09097290039062, "step": 668 }, { "epoch": 0.9823788546255506, "fcm_dpo/beta": 0.003718376625329256, "fcm_dpo/delta": -0.11621291935443878, "fcm_dpo/margin": 256.8287048339844, "fcm_dpo/q_t": 0.30499178171157837, "grad_norm": 45.1661491394043, "learning_rate": 5.564580657695939e-10, "logits/chosen": 0.3824569880962372, "logits/rejected": 0.4118229150772095, "logps/chosen": -461.65716552734375, "logps/ref_chosen": -46.116416931152344, "logps/ref_rejected": -77.92434692382812, "logps/rejected": -750.2938232421875, "loss": 0.8171, "margin_dpo/margin_mean": 256.82867431640625, "margin_dpo/margin_std": 251.70738220214844, "step": 669 }, { "epoch": 0.9838472834067548, "fcm_dpo/beta": 0.0036435904912650585, "fcm_dpo/delta": -0.1457192599773407, "fcm_dpo/margin": 269.37896728515625, "fcm_dpo/q_t": 0.30129629373550415, "grad_norm": 36.26799774169922, "learning_rate": 4.741678157389739e-10, "logits/chosen": 0.3282250761985779, "logits/rejected": 0.3458458185195923, "logps/chosen": -467.46923828125, "logps/ref_chosen": -62.34575271606445, "logps/ref_rejected": -96.9405517578125, "logps/rejected": -771.4429931640625, "loss": 0.7877, "margin_dpo/margin_mean": 269.3789367675781, "margin_dpo/margin_std": 242.67819213867188, "step": 670 }, { "epoch": 0.9853157121879589, "fcm_dpo/beta": 0.0036506270989775658, "fcm_dpo/delta": 0.05353014916181564, "fcm_dpo/margin": 218.90988159179688, "fcm_dpo/q_t": 0.3327978253364563, "grad_norm": 48.93035888671875, "learning_rate": 3.9845280344705245e-10, "logits/chosen": 0.42166364192962646, "logits/rejected": 0.4403586983680725, "logps/chosen": -525.71435546875, "logps/ref_chosen": -48.00010681152344, "logps/ref_rejected": -83.81932067871094, "logps/rejected": -780.4434814453125, "loss": 0.8728, "margin_dpo/margin_mean": 218.90988159179688, "margin_dpo/margin_std": 214.47088623046875, "step": 671 }, { "epoch": 0.986784140969163, "fcm_dpo/beta": 0.003691216232255101, "fcm_dpo/delta": 0.11230534315109253, "fcm_dpo/margin": 202.13328552246094, "fcm_dpo/q_t": 0.3466678857803345, "grad_norm": 54.89302444458008, "learning_rate": 3.293150240547549e-10, "logits/chosen": 0.3996291756629944, "logits/rejected": 0.4267166554927826, "logps/chosen": -560.9033813476562, "logps/ref_chosen": -58.58328628540039, "logps/ref_rejected": -93.14015197753906, "logps/rejected": -797.593505859375, "loss": 0.926, "margin_dpo/margin_mean": 202.13328552246094, "margin_dpo/margin_std": 228.80947875976562, "step": 672 }, { "epoch": 0.9882525697503671, "fcm_dpo/beta": 0.0036685578525066376, "fcm_dpo/delta": -0.07623796164989471, "fcm_dpo/margin": 250.54046630859375, "fcm_dpo/q_t": 0.3029751181602478, "grad_norm": 41.40584945678711, "learning_rate": 2.6675629940689504e-10, "logits/chosen": 0.4694506525993347, "logits/rejected": 0.4862295389175415, "logps/chosen": -493.67596435546875, "logps/ref_chosen": -46.72320556640625, "logps/ref_rejected": -85.29623413085938, "logps/rejected": -782.7894287109375, "loss": 0.7762, "margin_dpo/margin_mean": 250.54046630859375, "margin_dpo/margin_std": 197.22250366210938, "step": 673 }, { "epoch": 0.9897209985315712, "fcm_dpo/beta": 0.003621612675487995, "fcm_dpo/delta": -0.09082603454589844, "fcm_dpo/margin": 257.4602355957031, "fcm_dpo/q_t": 0.30964556336402893, "grad_norm": 42.31052017211914, "learning_rate": 2.1077827798404725e-10, "logits/chosen": 0.41855841875076294, "logits/rejected": 0.4410872459411621, "logps/chosen": -469.55859375, "logps/ref_chosen": -45.445526123046875, "logps/ref_rejected": -70.04593658447266, "logps/rejected": -751.6192626953125, "loss": 0.7966, "margin_dpo/margin_mean": 257.46026611328125, "margin_dpo/margin_std": 237.116455078125, "step": 674 }, { "epoch": 0.9911894273127754, "fcm_dpo/beta": 0.0034906486980617046, "fcm_dpo/delta": -0.15879766643047333, "fcm_dpo/margin": 284.1156005859375, "fcm_dpo/q_t": 0.2941362261772156, "grad_norm": 45.08475875854492, "learning_rate": 1.6138243485910863e-10, "logits/chosen": 0.38219112157821655, "logits/rejected": 0.4025418162345886, "logps/chosen": -472.37396240234375, "logps/ref_chosen": -44.17628479003906, "logps/ref_rejected": -74.09197998046875, "logps/rejected": -786.4052734375, "loss": 0.7469, "margin_dpo/margin_mean": 284.1156005859375, "margin_dpo/margin_std": 227.85464477539062, "step": 675 }, { "epoch": 0.9926578560939795, "fcm_dpo/beta": 0.003486907808110118, "fcm_dpo/delta": 0.027279622852802277, "fcm_dpo/margin": 236.58204650878906, "fcm_dpo/q_t": 0.3229207992553711, "grad_norm": 32.578102111816406, "learning_rate": 1.1857007165852472e-10, "logits/chosen": 0.24134598672389984, "logits/rejected": 0.27230340242385864, "logps/chosen": -521.6865234375, "logps/ref_chosen": -71.39852905273438, "logps/ref_rejected": -88.3587646484375, "logps/rejected": -775.2288208007812, "loss": 0.8219, "margin_dpo/margin_mean": 236.58204650878906, "margin_dpo/margin_std": 195.95883178710938, "step": 676 }, { "epoch": 0.9941262848751835, "fcm_dpo/beta": 0.0034925828222185373, "fcm_dpo/delta": -0.01529073528945446, "fcm_dpo/margin": 247.37342834472656, "fcm_dpo/q_t": 0.3151392936706543, "grad_norm": 44.7586669921875, "learning_rate": 8.23423165278725e-11, "logits/chosen": 0.48236775398254395, "logits/rejected": 0.522483229637146, "logps/chosen": -513.5480346679688, "logps/ref_chosen": -56.527435302734375, "logps/ref_rejected": -78.22654724121094, "logps/rejected": -782.62060546875, "loss": 0.8038, "margin_dpo/margin_mean": 247.37344360351562, "margin_dpo/margin_std": 200.0, "step": 677 }, { "epoch": 0.9955947136563876, "fcm_dpo/beta": 0.0034883874468505383, "fcm_dpo/delta": -0.0267241969704628, "fcm_dpo/margin": 250.60003662109375, "fcm_dpo/q_t": 0.32206088304519653, "grad_norm": 45.41565704345703, "learning_rate": 5.270012410216185e-11, "logits/chosen": 0.40671655535697937, "logits/rejected": 0.4284959137439728, "logps/chosen": -485.13336181640625, "logps/ref_chosen": -46.13447570800781, "logps/ref_rejected": -80.60462951660156, "logps/rejected": -770.2034912109375, "loss": 0.8466, "margin_dpo/margin_mean": 250.60003662109375, "margin_dpo/margin_std": 250.03765869140625, "step": 678 }, { "epoch": 0.9970631424375918, "fcm_dpo/beta": 0.0035341887269169092, "fcm_dpo/delta": 0.12530364096164703, "fcm_dpo/margin": 207.66168212890625, "fcm_dpo/q_t": 0.3447020351886749, "grad_norm": 40.24350357055664, "learning_rate": 2.9644275480772416e-11, "logits/chosen": 0.3600369691848755, "logits/rejected": 0.37837547063827515, "logps/chosen": -492.26739501953125, "logps/ref_chosen": -50.294921875, "logps/ref_rejected": -76.59813690185547, "logps/rejected": -726.2322998046875, "loss": 0.8974, "margin_dpo/margin_mean": 207.66168212890625, "margin_dpo/margin_std": 209.05560302734375, "step": 679 }, { "epoch": 0.9985315712187959, "fcm_dpo/beta": 0.0035505383275449276, "fcm_dpo/delta": -0.02245294116437435, "fcm_dpo/margin": 245.17251586914062, "fcm_dpo/q_t": 0.3214852809906006, "grad_norm": 37.78384780883789, "learning_rate": 1.31753782067201e-11, "logits/chosen": 0.29616981744766235, "logits/rejected": 0.32709792256355286, "logps/chosen": -534.2918090820312, "logps/ref_chosen": -76.91569519042969, "logps/ref_rejected": -112.384765625, "logps/rejected": -814.933349609375, "loss": 0.8629, "margin_dpo/margin_mean": 245.17251586914062, "margin_dpo/margin_std": 256.5103759765625, "step": 680 }, { "epoch": 1.0, "fcm_dpo/beta": 0.003594727721065283, "fcm_dpo/delta": 0.04884641245007515, "fcm_dpo/margin": 223.56224060058594, "fcm_dpo/q_t": 0.334045946598053, "grad_norm": 73.19285583496094, "learning_rate": 3.2938662507808745e-12, "logits/chosen": 0.3444863557815552, "logits/rejected": 0.3690025806427002, "logps/chosen": -525.255615234375, "logps/ref_chosen": -60.957279205322266, "logps/ref_rejected": -88.55797576904297, "logps/rejected": -776.4185791015625, "loss": 0.8767, "margin_dpo/margin_mean": 223.562255859375, "margin_dpo/margin_std": 224.77215576171875, "step": 681 }, { "epoch": 1.0, "step": 681, "total_flos": 0.0, "train_loss": 0.8810622134572784, "train_runtime": 1867.9067, "train_samples_per_second": 23.341, "train_steps_per_second": 0.365 } ], "logging_steps": 1, "max_steps": 681, "num_input_tokens_seen": 0, "num_train_epochs": 1, "save_steps": 50, "stateful_callbacks": { "TrainerControl": { "args": { "should_epoch_stop": false, "should_evaluate": false, "should_log": false, "should_save": false, "should_training_stop": false }, "attributes": {} } }, "total_flos": 0.0, "train_batch_size": 8, "trial_name": null, "trial_params": null }