{ "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": 0.10308968275785446, "fcm_dpo/delta": 0.29979780316352844, "fcm_dpo/margin": -0.02287006378173828, "fcm_dpo/q_t": 0.5005706548690796, "grad_norm": 86.183349609375, "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": 0.1061793640255928, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.06572261452674866, "fcm_dpo/q_t": 0.5017440319061279, "grad_norm": 76.63970184326172, "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": 1.3936, "margin_dpo/margin_mean": -0.06572240591049194, "margin_dpo/margin_std": 0.35048407316207886, "step": 2 }, { "epoch": 0.004405286343612335, "fcm_dpo/beta": 0.10943304747343063, "fcm_dpo/delta": 0.29740938544273376, "fcm_dpo/margin": 0.009920269250869751, "fcm_dpo/q_t": 0.4997372329235077, "grad_norm": 77.7017822265625, "learning_rate": 1.4492753623188406e-08, "logits/chosen": -0.4817405045032501, "logits/rejected": -0.4422529935836792, "logps/chosen": -60.98210144042969, "logps/ref_chosen": -60.981597900390625, "logps/ref_rejected": -68.67259216308594, "logps/rejected": -68.68302154541016, "loss": 1.3855, "margin_dpo/margin_mean": 0.009920388460159302, "margin_dpo/margin_std": 0.3543340563774109, "step": 3 }, { "epoch": 0.005873715124816446, "fcm_dpo/beta": 0.1161462590098381, "fcm_dpo/delta": 0.29794785380363464, "fcm_dpo/margin": -0.01907077431678772, "fcm_dpo/q_t": 0.5005367994308472, "grad_norm": 83.44183349609375, "learning_rate": 2.1739130434782606e-08, "logits/chosen": -0.46823394298553467, "logits/rejected": -0.44061601161956787, "logps/chosen": -56.77912139892578, "logps/ref_chosen": -56.7677116394043, "logps/ref_rejected": -86.64710998535156, "logps/rejected": -86.63945007324219, "loss": 1.3887, "margin_dpo/margin_mean": -0.01907128095626831, "margin_dpo/margin_std": 0.34393125772476196, "step": 4 }, { "epoch": 0.007342143906020558, "fcm_dpo/beta": 0.13080811500549316, "fcm_dpo/delta": 0.5949947834014893, "fcm_dpo/margin": 0.04144471883773804, "fcm_dpo/q_t": 0.4987483024597168, "grad_norm": 116.84333038330078, "learning_rate": 2.898550724637681e-08, "logits/chosen": -0.5144734978675842, "logits/rejected": -0.4707837700843811, "logps/chosen": -53.85641860961914, "logps/ref_chosen": -53.859375, "logps/ref_rejected": -84.14918518066406, "logps/rejected": -84.18767547607422, "loss": 1.3815, "margin_dpo/margin_mean": 0.041444987058639526, "margin_dpo/margin_std": 0.32837140560150146, "step": 5 }, { "epoch": 0.00881057268722467, "fcm_dpo/beta": 0.13887986540794373, "fcm_dpo/delta": 0.2996029257774353, "fcm_dpo/margin": -0.049517929553985596, "fcm_dpo/q_t": 0.5016666650772095, "grad_norm": 127.5752182006836, "learning_rate": 3.6231884057971014e-08, "logits/chosen": -0.5032555460929871, "logits/rejected": -0.4607737958431244, "logps/chosen": -63.0675048828125, "logps/ref_chosen": -63.007484436035156, "logps/ref_rejected": -92.64534759521484, "logps/rejected": -92.65585327148438, "loss": 1.3938, "margin_dpo/margin_mean": -0.04951813817024231, "margin_dpo/margin_std": 0.41555964946746826, "step": 6 }, { "epoch": 0.010279001468428781, "fcm_dpo/beta": 0.15183104574680328, "fcm_dpo/delta": 0.29823803901672363, "fcm_dpo/margin": -0.0031449496746063232, "fcm_dpo/q_t": 0.5001453161239624, "grad_norm": 124.75188446044922, "learning_rate": 4.347826086956521e-08, "logits/chosen": -0.4999721646308899, "logits/rejected": -0.466085284948349, "logps/chosen": -57.7862548828125, "logps/ref_chosen": -57.774818420410156, "logps/ref_rejected": -103.92059326171875, "logps/rejected": -103.92887878417969, "loss": 1.3877, "margin_dpo/margin_mean": -0.0031441152095794678, "margin_dpo/margin_std": 0.39584606885910034, "step": 7 }, { "epoch": 0.011747430249632892, "fcm_dpo/beta": 0.1610293984413147, "fcm_dpo/delta": 0.2940930426120758, "fcm_dpo/margin": 0.014331787824630737, "fcm_dpo/q_t": 0.4995083510875702, "grad_norm": 126.80061340332031, "learning_rate": 5.0724637681159424e-08, "logits/chosen": -0.5133894085884094, "logits/rejected": -0.4880775511264801, "logps/chosen": -58.66931915283203, "logps/ref_chosen": -58.716033935546875, "logps/ref_rejected": -79.3114242553711, "logps/rejected": -79.27903747558594, "loss": 1.385, "margin_dpo/margin_mean": 0.014332354068756104, "margin_dpo/margin_std": 0.381694495677948, "step": 8 }, { "epoch": 0.013215859030837005, "fcm_dpo/beta": 0.17621421813964844, "fcm_dpo/delta": 0.5977947115898132, "fcm_dpo/margin": 0.013306349515914917, "fcm_dpo/q_t": 0.4994489848613739, "grad_norm": 150.93214416503906, "learning_rate": 5.797101449275362e-08, "logits/chosen": -0.4874652028083801, "logits/rejected": -0.44039708375930786, "logps/chosen": -69.85830688476562, "logps/ref_chosen": -69.8668441772461, "logps/ref_rejected": -99.6026611328125, "logps/rejected": -99.60743713378906, "loss": 1.3852, "margin_dpo/margin_mean": 0.013305962085723877, "margin_dpo/margin_std": 0.3926519751548767, "step": 9 }, { "epoch": 0.014684287812041116, "fcm_dpo/beta": 0.18701426684856415, "fcm_dpo/delta": 0.2960095703601837, "fcm_dpo/margin": 0.019543275237083435, "fcm_dpo/q_t": 0.4991165101528168, "grad_norm": 132.65455627441406, "learning_rate": 6.521739130434782e-08, "logits/chosen": -0.4857853651046753, "logits/rejected": -0.4421781301498413, "logps/chosen": -48.35417938232422, "logps/ref_chosen": -48.35768508911133, "logps/ref_rejected": -80.37206268310547, "logps/rejected": -80.38810729980469, "loss": 1.3838, "margin_dpo/margin_mean": 0.01954329013824463, "margin_dpo/margin_std": 0.37247973680496216, "step": 10 }, { "epoch": 0.016152716593245228, "fcm_dpo/beta": 0.20438191294670105, "fcm_dpo/delta": 0.29821285605430603, "fcm_dpo/margin": -0.0019693076610565186, "fcm_dpo/q_t": 0.5001306533813477, "grad_norm": 140.15432739257812, "learning_rate": 7.246376811594203e-08, "logits/chosen": -0.4591175615787506, "logits/rejected": -0.43393969535827637, "logps/chosen": -53.05950927734375, "logps/ref_chosen": -53.01685333251953, "logps/ref_rejected": -87.78038024902344, "logps/rejected": -87.82106018066406, "loss": 1.388, "margin_dpo/margin_mean": -0.001968860626220703, "margin_dpo/margin_std": 0.34557127952575684, "step": 11 }, { "epoch": 0.01762114537444934, "fcm_dpo/beta": 0.2106446623802185, "fcm_dpo/delta": 0.29740071296691895, "fcm_dpo/margin": 0.00841417908668518, "fcm_dpo/q_t": 0.4995623826980591, "grad_norm": 188.97665405273438, "learning_rate": 7.971014492753623e-08, "logits/chosen": -0.5176132321357727, "logits/rejected": -0.4807056188583374, "logps/chosen": -61.83100128173828, "logps/ref_chosen": -61.80543518066406, "logps/ref_rejected": -104.8582763671875, "logps/rejected": -104.89225769042969, "loss": 1.3861, "margin_dpo/margin_mean": 0.008413076400756836, "margin_dpo/margin_std": 0.3932174742221832, "step": 12 }, { "epoch": 0.01908957415565345, "fcm_dpo/beta": 0.2233951985836029, "fcm_dpo/delta": 0.2904997766017914, "fcm_dpo/margin": 0.003662973642349243, "fcm_dpo/q_t": 0.4998163878917694, "grad_norm": 176.16429138183594, "learning_rate": 8.695652173913042e-08, "logits/chosen": -0.4700925946235657, "logits/rejected": -0.4411465525627136, "logps/chosen": -64.25465393066406, "logps/ref_chosen": -64.2603530883789, "logps/ref_rejected": -87.20307922363281, "logps/rejected": -87.20104217529297, "loss": 1.387, "margin_dpo/margin_mean": 0.0036628246307373047, "margin_dpo/margin_std": 0.39085596799850464, "step": 13 }, { "epoch": 0.020558002936857563, "fcm_dpo/beta": 0.24400313198566437, "fcm_dpo/delta": 0.29805275797843933, "fcm_dpo/margin": 0.005518615245819092, "fcm_dpo/q_t": 0.499683141708374, "grad_norm": 209.18350219726562, "learning_rate": 9.420289855072464e-08, "logits/chosen": -0.47941046953201294, "logits/rejected": -0.4401160478591919, "logps/chosen": -58.11064910888672, "logps/ref_chosen": -58.11021041870117, "logps/ref_rejected": -104.04708099365234, "logps/rejected": -104.05303192138672, "loss": 1.3883, "margin_dpo/margin_mean": 0.005518555641174316, "margin_dpo/margin_std": 0.47164684534072876, "step": 14 }, { "epoch": 0.022026431718061675, "fcm_dpo/beta": 0.25842905044555664, "fcm_dpo/delta": 0.2872004508972168, "fcm_dpo/margin": 0.03296121954917908, "fcm_dpo/q_t": 0.49806299805641174, "grad_norm": 164.04367065429688, "learning_rate": 1.0144927536231885e-07, "logits/chosen": -0.5029046535491943, "logits/rejected": -0.4847126603126526, "logps/chosen": -56.98225402832031, "logps/ref_chosen": -56.96691131591797, "logps/ref_rejected": -80.80863952636719, "logps/rejected": -80.85694885253906, "loss": 1.3795, "margin_dpo/margin_mean": 0.03296104073524475, "margin_dpo/margin_std": 0.3130000829696655, "step": 15 }, { "epoch": 0.023494860499265784, "fcm_dpo/beta": 0.2661767303943634, "fcm_dpo/delta": 0.2911548614501953, "fcm_dpo/margin": 0.006061911582946777, "fcm_dpo/q_t": 0.49960726499557495, "grad_norm": 225.11155700683594, "learning_rate": 1.0869565217391303e-07, "logits/chosen": -0.5254543423652649, "logits/rejected": -0.48440250754356384, "logps/chosen": -61.74738693237305, "logps/ref_chosen": -61.739891052246094, "logps/ref_rejected": -84.36947631835938, "logps/rejected": -84.38302612304688, "loss": 1.3865, "margin_dpo/margin_mean": 0.0060617029666900635, "margin_dpo/margin_std": 0.35263463854789734, "step": 16 }, { "epoch": 0.024963289280469897, "fcm_dpo/beta": 0.29960495233535767, "fcm_dpo/delta": 0.5921929478645325, "fcm_dpo/margin": 0.02691623568534851, "fcm_dpo/q_t": 0.49804800748825073, "grad_norm": 234.62332153320312, "learning_rate": 1.1594202898550725e-07, "logits/chosen": -0.4995841979980469, "logits/rejected": -0.4622904658317566, "logps/chosen": -67.72857666015625, "logps/ref_chosen": -67.71033477783203, "logps/ref_rejected": -85.37865447998047, "logps/rejected": -85.42382049560547, "loss": 1.3808, "margin_dpo/margin_mean": 0.02691650390625, "margin_dpo/margin_std": 0.35206741094589233, "step": 17 }, { "epoch": 0.02643171806167401, "fcm_dpo/beta": 0.3366771340370178, "fcm_dpo/delta": 0.5766884088516235, "fcm_dpo/margin": 0.07164651155471802, "fcm_dpo/q_t": 0.4941858947277069, "grad_norm": 273.0143127441406, "learning_rate": 1.2318840579710146e-07, "logits/chosen": -0.5020445585250854, "logits/rejected": -0.44765499234199524, "logps/chosen": -47.74727249145508, "logps/ref_chosen": -47.7394905090332, "logps/ref_rejected": -75.4722900390625, "logps/rejected": -75.55171966552734, "loss": 1.3652, "margin_dpo/margin_mean": 0.07164624333381653, "margin_dpo/margin_std": 0.3414478302001953, "step": 18 }, { "epoch": 0.027900146842878122, "fcm_dpo/beta": 0.37680280208587646, "fcm_dpo/delta": 0.5566200017929077, "fcm_dpo/margin": 0.11977455019950867, "fcm_dpo/q_t": 0.4892091155052185, "grad_norm": 273.0553894042969, "learning_rate": 1.3043478260869563e-07, "logits/chosen": -0.4712415337562561, "logits/rejected": -0.41980695724487305, "logps/chosen": -70.16747283935547, "logps/ref_chosen": -70.20536041259766, "logps/ref_rejected": -89.7575912475586, "logps/rejected": -89.83948516845703, "loss": 1.3457, "margin_dpo/margin_mean": 0.11977550387382507, "margin_dpo/margin_std": 0.3553928732872009, "step": 19 }, { "epoch": 0.02936857562408223, "fcm_dpo/beta": 0.39811867475509644, "fcm_dpo/delta": 0.2839474380016327, "fcm_dpo/margin": 0.03722500801086426, "fcm_dpo/q_t": 0.49641382694244385, "grad_norm": 294.3759460449219, "learning_rate": 1.3768115942028986e-07, "logits/chosen": -0.5450824499130249, "logits/rejected": -0.48803430795669556, "logps/chosen": -50.81207275390625, "logps/ref_chosen": -50.80324172973633, "logps/ref_rejected": -78.82334899902344, "logps/rejected": -78.86940002441406, "loss": 1.3748, "margin_dpo/margin_mean": 0.037224918603897095, "margin_dpo/margin_std": 0.3072945773601532, "step": 20 }, { "epoch": 0.030837004405286344, "fcm_dpo/beta": 0.443250834941864, "fcm_dpo/delta": 0.5273383259773254, "fcm_dpo/margin": 0.1729957014322281, "fcm_dpo/q_t": 0.48192811012268066, "grad_norm": 327.8818054199219, "learning_rate": 1.4492753623188405e-07, "logits/chosen": -0.517810046672821, "logits/rejected": -0.49564769864082336, "logps/chosen": -50.032989501953125, "logps/ref_chosen": -50.063018798828125, "logps/ref_rejected": -77.86878967285156, "logps/rejected": -78.01175689697266, "loss": 1.3156, "margin_dpo/margin_mean": 0.1729956418275833, "margin_dpo/margin_std": 0.3008769452571869, "step": 21 }, { "epoch": 0.032305433186490456, "fcm_dpo/beta": 0.49217575788497925, "fcm_dpo/delta": 0.4990707039833069, "fcm_dpo/margin": 0.21139326691627502, "fcm_dpo/q_t": 0.4752381443977356, "grad_norm": 389.9123229980469, "learning_rate": 1.5217391304347825e-07, "logits/chosen": -0.48431238532066345, "logits/rejected": -0.4406638741493225, "logps/chosen": -58.97994613647461, "logps/ref_chosen": -59.05763626098633, "logps/ref_rejected": -97.50466918945312, "logps/rejected": -97.63837432861328, "loss": 1.2977, "margin_dpo/margin_mean": 0.21139344573020935, "margin_dpo/margin_std": 0.46386557817459106, "step": 22 }, { "epoch": 0.033773861967694566, "fcm_dpo/beta": 0.5401284694671631, "fcm_dpo/delta": 0.4757743179798126, "fcm_dpo/margin": 0.24120861291885376, "fcm_dpo/q_t": 0.469452828168869, "grad_norm": 389.5323791503906, "learning_rate": 1.5942028985507245e-07, "logits/chosen": -0.4874040186405182, "logits/rejected": -0.4647292494773865, "logps/chosen": -60.02528381347656, "logps/ref_chosen": -60.07769775390625, "logps/ref_rejected": -81.13955688476562, "logps/rejected": -81.3283462524414, "loss": 1.2741, "margin_dpo/margin_mean": 0.24120870232582092, "margin_dpo/margin_std": 0.418229341506958, "step": 23 }, { "epoch": 0.03524229074889868, "fcm_dpo/beta": 0.5957721471786499, "fcm_dpo/delta": 0.4795808792114258, "fcm_dpo/margin": 0.20983095467090607, "fcm_dpo/q_t": 0.47032594680786133, "grad_norm": 471.896728515625, "learning_rate": 1.6666666666666665e-07, "logits/chosen": -0.5221496820449829, "logits/rejected": -0.5054002404212952, "logps/chosen": -44.25273895263672, "logps/ref_chosen": -44.29103469848633, "logps/ref_rejected": -99.12521362304688, "logps/rejected": -99.2967529296875, "loss": 1.2789, "margin_dpo/margin_mean": 0.20983177423477173, "margin_dpo/margin_std": 0.3987899422645569, "step": 24 }, { "epoch": 0.03671071953010279, "fcm_dpo/beta": 0.6493321657180786, "fcm_dpo/delta": 0.4309549033641815, "fcm_dpo/margin": 0.27194589376449585, "fcm_dpo/q_t": 0.4587538540363312, "grad_norm": 428.5508117675781, "learning_rate": 1.7391304347826085e-07, "logits/chosen": -0.4805753827095032, "logits/rejected": -0.4500219225883484, "logps/chosen": -52.49953842163086, "logps/ref_chosen": -52.537052154541016, "logps/ref_rejected": -89.34219360351562, "logps/rejected": -89.57662200927734, "loss": 1.2366, "margin_dpo/margin_mean": 0.27194616198539734, "margin_dpo/margin_std": 0.4171033203601837, "step": 25 }, { "epoch": 0.0381791483113069, "fcm_dpo/beta": 0.6972817182540894, "fcm_dpo/delta": 0.3234151601791382, "fcm_dpo/margin": 0.4091455340385437, "fcm_dpo/q_t": 0.4343709349632263, "grad_norm": 487.4859619140625, "learning_rate": 1.8115942028985507e-07, "logits/chosen": -0.5440469980239868, "logits/rejected": -0.5126461386680603, "logps/chosen": -53.82080078125, "logps/ref_chosen": -53.92280578613281, "logps/ref_rejected": -103.35971069335938, "logps/rejected": -103.66685485839844, "loss": 1.1593, "margin_dpo/margin_mean": 0.4091459810733795, "margin_dpo/margin_std": 0.5674811601638794, "step": 26 }, { "epoch": 0.039647577092511016, "fcm_dpo/beta": 0.7315391302108765, "fcm_dpo/delta": 0.22802981734275818, "fcm_dpo/margin": 0.5205895304679871, "fcm_dpo/q_t": 0.41203832626342773, "grad_norm": 530.424560546875, "learning_rate": 1.8840579710144927e-07, "logits/chosen": -0.5132520198822021, "logits/rejected": -0.47679802775382996, "logps/chosen": -42.7576904296875, "logps/ref_chosen": -42.898529052734375, "logps/ref_rejected": -98.72419738769531, "logps/rejected": -99.10395812988281, "loss": 1.0773, "margin_dpo/margin_mean": 0.5205891132354736, "margin_dpo/margin_std": 0.5357187986373901, "step": 27 }, { "epoch": 0.041116005873715125, "fcm_dpo/beta": 0.7871224284172058, "fcm_dpo/delta": 0.3837660551071167, "fcm_dpo/margin": 0.28373512625694275, "fcm_dpo/q_t": 0.44818979501724243, "grad_norm": 503.4446105957031, "learning_rate": 1.9565217391304347e-07, "logits/chosen": -0.5058661699295044, "logits/rejected": -0.4506160616874695, "logps/chosen": -60.58196258544922, "logps/ref_chosen": -60.55650329589844, "logps/ref_rejected": -91.40111541748047, "logps/rejected": -91.71031188964844, "loss": 1.2172, "margin_dpo/margin_mean": 0.2837350368499756, "margin_dpo/margin_std": 0.5103665590286255, "step": 28 }, { "epoch": 0.042584434654919234, "fcm_dpo/beta": 0.8140842914581299, "fcm_dpo/delta": 0.12859481573104858, "fcm_dpo/margin": 0.5867732763290405, "fcm_dpo/q_t": 0.39105674624443054, "grad_norm": 521.9832763671875, "learning_rate": 2.028985507246377e-07, "logits/chosen": -0.5453035831451416, "logits/rejected": -0.498327374458313, "logps/chosen": -57.687744140625, "logps/ref_chosen": -57.80778503417969, "logps/ref_rejected": -97.39434814453125, "logps/rejected": -97.86107635498047, "loss": 1.0177, "margin_dpo/margin_mean": 0.5867743492126465, "margin_dpo/margin_std": 0.5909589529037476, "step": 29 }, { "epoch": 0.04405286343612335, "fcm_dpo/beta": 0.8216707706451416, "fcm_dpo/delta": 0.0340820774435997, "fcm_dpo/margin": 0.690904438495636, "fcm_dpo/q_t": 0.3710797429084778, "grad_norm": 468.0931396484375, "learning_rate": 2.1014492753623187e-07, "logits/chosen": -0.4391070306301117, "logits/rejected": -0.40617626905441284, "logps/chosen": -52.45497131347656, "logps/ref_chosen": -52.577369689941406, "logps/ref_rejected": -98.48920440673828, "logps/rejected": -99.05770874023438, "loss": 0.9603, "margin_dpo/margin_mean": 0.6909043788909912, "margin_dpo/margin_std": 0.6301661729812622, "step": 30 }, { "epoch": 0.04552129221732746, "fcm_dpo/beta": 0.8536430597305298, "fcm_dpo/delta": 0.21446718275547028, "fcm_dpo/margin": 0.46136218309402466, "fcm_dpo/q_t": 0.4132009744644165, "grad_norm": 441.92120361328125, "learning_rate": 2.1739130434782607e-07, "logits/chosen": -0.5114755034446716, "logits/rejected": -0.46733975410461426, "logps/chosen": -63.724403381347656, "logps/ref_chosen": -63.806922912597656, "logps/ref_rejected": -72.89400482177734, "logps/rejected": -73.2728500366211, "loss": 1.0973, "margin_dpo/margin_mean": 0.46136197447776794, "margin_dpo/margin_std": 0.6410406827926636, "step": 31 }, { "epoch": 0.04698972099853157, "fcm_dpo/beta": 0.8558568954467773, "fcm_dpo/delta": -0.04837347939610481, "fcm_dpo/margin": 0.7538948655128479, "fcm_dpo/q_t": 0.363816499710083, "grad_norm": 431.97833251953125, "learning_rate": 2.2463768115942027e-07, "logits/chosen": -0.5344721078872681, "logits/rejected": -0.4945007562637329, "logps/chosen": -62.547889709472656, "logps/ref_chosen": -62.739524841308594, "logps/ref_rejected": -89.3175048828125, "logps/rejected": -89.8797607421875, "loss": 0.95, "margin_dpo/margin_mean": 0.7538947463035583, "margin_dpo/margin_std": 0.8735177516937256, "step": 32 }, { "epoch": 0.048458149779735685, "fcm_dpo/beta": 0.8686736822128296, "fcm_dpo/delta": 0.07413993775844574, "fcm_dpo/margin": 0.608627438545227, "fcm_dpo/q_t": 0.3796751797199249, "grad_norm": 434.5819091796875, "learning_rate": 2.318840579710145e-07, "logits/chosen": -0.48133665323257446, "logits/rejected": -0.4549505412578583, "logps/chosen": -53.15459442138672, "logps/ref_chosen": -53.26097106933594, "logps/ref_rejected": -87.8851318359375, "logps/rejected": -88.38739013671875, "loss": 0.9889, "margin_dpo/margin_mean": 0.6086278557777405, "margin_dpo/margin_std": 0.5917040109634399, "step": 33 }, { "epoch": 0.049926578560939794, "fcm_dpo/beta": 0.8607486486434937, "fcm_dpo/delta": -0.029908865690231323, "fcm_dpo/margin": 0.7293862104415894, "fcm_dpo/q_t": 0.3666630983352661, "grad_norm": 413.73388671875, "learning_rate": 2.391304347826087e-07, "logits/chosen": -0.5056343078613281, "logits/rejected": -0.4880971908569336, "logps/chosen": -50.74631118774414, "logps/ref_chosen": -50.81732940673828, "logps/ref_rejected": -101.92184448242188, "logps/rejected": -102.58021545410156, "loss": 0.9715, "margin_dpo/margin_mean": 0.7293854355812073, "margin_dpo/margin_std": 0.8646777272224426, "step": 34 }, { "epoch": 0.0513950073421439, "fcm_dpo/beta": 0.8129470348358154, "fcm_dpo/delta": -0.3546205759048462, "fcm_dpo/margin": 1.1309483051300049, "fcm_dpo/q_t": 0.3066979646682739, "grad_norm": 335.1808166503906, "learning_rate": 2.463768115942029e-07, "logits/chosen": -0.5237250328063965, "logits/rejected": -0.48680445551872253, "logps/chosen": -50.88426208496094, "logps/ref_chosen": -51.02449035644531, "logps/ref_rejected": -106.82443237304688, "logps/rejected": -107.81513977050781, "loss": 0.7836, "margin_dpo/margin_mean": 1.1309483051300049, "margin_dpo/margin_std": 1.0144567489624023, "step": 35 }, { "epoch": 0.05286343612334802, "fcm_dpo/beta": 0.7814772129058838, "fcm_dpo/delta": -0.24869973957538605, "fcm_dpo/margin": 1.0601816177368164, "fcm_dpo/q_t": 0.3314136862754822, "grad_norm": 286.6738586425781, "learning_rate": 2.536231884057971e-07, "logits/chosen": -0.5335814952850342, "logits/rejected": -0.4959707260131836, "logps/chosen": -51.95500946044922, "logps/ref_chosen": -51.991493225097656, "logps/ref_rejected": -86.0406265258789, "logps/rejected": -87.06432342529297, "loss": 0.8938, "margin_dpo/margin_mean": 1.0601813793182373, "margin_dpo/margin_std": 1.1594077348709106, "step": 36 }, { "epoch": 0.05433186490455213, "fcm_dpo/beta": 0.7387670278549194, "fcm_dpo/delta": -0.1298440545797348, "fcm_dpo/margin": 0.9678917527198792, "fcm_dpo/q_t": 0.35818910598754883, "grad_norm": 304.4512939453125, "learning_rate": 2.6086956521739126e-07, "logits/chosen": -0.5167920589447021, "logits/rejected": -0.4735545217990875, "logps/chosen": -62.80792999267578, "logps/ref_chosen": -62.807106018066406, "logps/ref_rejected": -77.89507293701172, "logps/rejected": -78.86378479003906, "loss": 0.982, "margin_dpo/margin_mean": 0.9678912162780762, "margin_dpo/margin_std": 1.311614751815796, "step": 37 }, { "epoch": 0.055800293685756244, "fcm_dpo/beta": 0.7068374156951904, "fcm_dpo/delta": -0.37072476744651794, "fcm_dpo/margin": 1.3253750801086426, "fcm_dpo/q_t": 0.32258689403533936, "grad_norm": 279.52642822265625, "learning_rate": 2.681159420289855e-07, "logits/chosen": -0.5299612879753113, "logits/rejected": -0.49737080931663513, "logps/chosen": -48.24340057373047, "logps/ref_chosen": -48.39051818847656, "logps/ref_rejected": -97.91244506835938, "logps/rejected": -99.0906982421875, "loss": 0.8506, "margin_dpo/margin_mean": 1.3253748416900635, "margin_dpo/margin_std": 1.5660966634750366, "step": 38 }, { "epoch": 0.05726872246696035, "fcm_dpo/beta": 0.6403067111968994, "fcm_dpo/delta": -0.4625703692436218, "fcm_dpo/margin": 1.5791468620300293, "fcm_dpo/q_t": 0.2878713607788086, "grad_norm": 240.9384765625, "learning_rate": 2.753623188405797e-07, "logits/chosen": -0.5889618396759033, "logits/rejected": -0.5511986613273621, "logps/chosen": -50.64854431152344, "logps/ref_chosen": -50.75047302246094, "logps/ref_rejected": -78.56951141357422, "logps/rejected": -80.0467300415039, "loss": 0.7516, "margin_dpo/margin_mean": 1.5791466236114502, "margin_dpo/margin_std": 1.3109757900238037, "step": 39 }, { "epoch": 0.05873715124816446, "fcm_dpo/beta": 0.6059188842773438, "fcm_dpo/delta": -0.3013249933719635, "fcm_dpo/margin": 1.4455113410949707, "fcm_dpo/q_t": 0.3245118260383606, "grad_norm": 211.55166625976562, "learning_rate": 2.8260869565217386e-07, "logits/chosen": -0.5337985754013062, "logits/rejected": -0.5050651431083679, "logps/chosen": -57.77853012084961, "logps/ref_chosen": -57.985069274902344, "logps/ref_rejected": -74.3000717163086, "logps/rejected": -75.53904724121094, "loss": 0.8502, "margin_dpo/margin_mean": 1.445512294769287, "margin_dpo/margin_std": 1.521177053451538, "step": 40 }, { "epoch": 0.06020558002936858, "fcm_dpo/beta": 0.5612444877624512, "fcm_dpo/delta": -0.3904670178890228, "fcm_dpo/margin": 1.6995972394943237, "fcm_dpo/q_t": 0.31129536032676697, "grad_norm": 220.2853546142578, "learning_rate": 2.898550724637681e-07, "logits/chosen": -0.5450239777565002, "logits/rejected": -0.5087831020355225, "logps/chosen": -62.68452835083008, "logps/ref_chosen": -62.69581604003906, "logps/ref_rejected": -97.02352905273438, "logps/rejected": -98.71183776855469, "loss": 0.8303, "margin_dpo/margin_mean": 1.6995975971221924, "margin_dpo/margin_std": 1.7764086723327637, "step": 41 }, { "epoch": 0.06167400881057269, "fcm_dpo/beta": 0.4963580369949341, "fcm_dpo/delta": -0.6783915758132935, "fcm_dpo/margin": 2.4108240604400635, "fcm_dpo/q_t": 0.2684577703475952, "grad_norm": 179.24830627441406, "learning_rate": 2.971014492753623e-07, "logits/chosen": -0.5304057598114014, "logits/rejected": -0.48365843296051025, "logps/chosen": -58.75749588012695, "logps/ref_chosen": -58.966426849365234, "logps/ref_rejected": -109.90837097167969, "logps/rejected": -112.11026000976562, "loss": 0.6984, "margin_dpo/margin_mean": 2.4108238220214844, "margin_dpo/margin_std": 2.168299436569214, "step": 42 }, { "epoch": 0.0631424375917768, "fcm_dpo/beta": 0.44308847188949585, "fcm_dpo/delta": -0.48201635479927063, "fcm_dpo/margin": 2.3211944103240967, "fcm_dpo/q_t": 0.2817467451095581, "grad_norm": 169.86190795898438, "learning_rate": 3.043478260869565e-07, "logits/chosen": -0.5538966059684753, "logits/rejected": -0.5297778844833374, "logps/chosen": -53.645652770996094, "logps/ref_chosen": -54.15599822998047, "logps/ref_rejected": -96.48019409179688, "logps/rejected": -98.29104614257812, "loss": 0.7188, "margin_dpo/margin_mean": 2.3211941719055176, "margin_dpo/margin_std": 1.7973530292510986, "step": 43 }, { "epoch": 0.06461086637298091, "fcm_dpo/beta": 0.4017961323261261, "fcm_dpo/delta": -0.49928557872772217, "fcm_dpo/margin": 2.597970485687256, "fcm_dpo/q_t": 0.2795543372631073, "grad_norm": 164.8052215576172, "learning_rate": 3.115942028985507e-07, "logits/chosen": -0.4567152261734009, "logits/rejected": -0.43692851066589355, "logps/chosen": -49.864402770996094, "logps/ref_chosen": -50.07849884033203, "logps/ref_rejected": -108.78376007080078, "logps/rejected": -111.16763305664062, "loss": 0.713, "margin_dpo/margin_mean": 2.597970962524414, "margin_dpo/margin_std": 1.9955940246582031, "step": 44 }, { "epoch": 0.06607929515418502, "fcm_dpo/beta": 0.3842885196208954, "fcm_dpo/delta": -0.18539269268512726, "fcm_dpo/margin": 2.0074782371520996, "fcm_dpo/q_t": 0.3449398875236511, "grad_norm": 142.70497131347656, "learning_rate": 3.188405797101449e-07, "logits/chosen": -0.4787759780883789, "logits/rejected": -0.46609732508659363, "logps/chosen": -48.27836227416992, "logps/ref_chosen": -48.4149284362793, "logps/ref_rejected": -77.93643188476562, "logps/rejected": -79.80734252929688, "loss": 0.9085, "margin_dpo/margin_mean": 2.0074782371520996, "margin_dpo/margin_std": 2.3263068199157715, "step": 45 }, { "epoch": 0.06754772393538913, "fcm_dpo/beta": 0.3586476445198059, "fcm_dpo/delta": -0.3937702775001526, "fcm_dpo/margin": 2.6675519943237305, "fcm_dpo/q_t": 0.31235700845718384, "grad_norm": 143.95310974121094, "learning_rate": 3.260869565217391e-07, "logits/chosen": -0.5191431045532227, "logits/rejected": -0.46773868799209595, "logps/chosen": -55.786354064941406, "logps/ref_chosen": -55.999427795410156, "logps/ref_rejected": -95.652587890625, "logps/rejected": -98.10707092285156, "loss": 0.8258, "margin_dpo/margin_mean": 2.6675515174865723, "margin_dpo/margin_std": 2.8206050395965576, "step": 46 }, { "epoch": 0.06901615271659324, "fcm_dpo/beta": 0.33721715211868286, "fcm_dpo/delta": -0.31360626220703125, "fcm_dpo/margin": 2.6299524307250977, "fcm_dpo/q_t": 0.31478771567344666, "grad_norm": 143.6584930419922, "learning_rate": 3.333333333333333e-07, "logits/chosen": -0.5552619695663452, "logits/rejected": -0.5014165639877319, "logps/chosen": -57.52300262451172, "logps/ref_chosen": -57.92607879638672, "logps/ref_rejected": -94.67920684814453, "logps/rejected": -96.90608215332031, "loss": 0.8165, "margin_dpo/margin_mean": 2.6299524307250977, "margin_dpo/margin_std": 2.3889541625976562, "step": 47 }, { "epoch": 0.07048458149779736, "fcm_dpo/beta": 0.31545713543891907, "fcm_dpo/delta": -0.23579037189483643, "fcm_dpo/margin": 2.579418659210205, "fcm_dpo/q_t": 0.3256019353866577, "grad_norm": 148.15615844726562, "learning_rate": 3.4057971014492755e-07, "logits/chosen": -0.584140956401825, "logits/rejected": -0.5260648131370544, "logps/chosen": -57.080162048339844, "logps/ref_chosen": -57.188072204589844, "logps/ref_rejected": -88.0166015625, "logps/rejected": -90.48811340332031, "loss": 0.8611, "margin_dpo/margin_mean": 2.5794191360473633, "margin_dpo/margin_std": 2.3878588676452637, "step": 48 }, { "epoch": 0.07195301027900147, "fcm_dpo/beta": 0.30197674036026, "fcm_dpo/delta": -0.30794399976730347, "fcm_dpo/margin": 2.9200515747070312, "fcm_dpo/q_t": 0.31857985258102417, "grad_norm": 391.1684265136719, "learning_rate": 3.478260869565217e-07, "logits/chosen": -0.5332294702529907, "logits/rejected": -0.47411999106407166, "logps/chosen": -61.26190948486328, "logps/ref_chosen": -61.685272216796875, "logps/ref_rejected": -83.76747131347656, "logps/rejected": -86.26417541503906, "loss": 0.8336, "margin_dpo/margin_mean": 2.9200520515441895, "margin_dpo/margin_std": 2.897745132446289, "step": 49 }, { "epoch": 0.07342143906020558, "fcm_dpo/beta": 0.2793174982070923, "fcm_dpo/delta": -0.3477798402309418, "fcm_dpo/margin": 3.2729971408843994, "fcm_dpo/q_t": 0.3099723756313324, "grad_norm": 119.61611938476562, "learning_rate": 3.5507246376811595e-07, "logits/chosen": -0.5257272124290466, "logits/rejected": -0.4897651672363281, "logps/chosen": -58.62849426269531, "logps/ref_chosen": -58.72413635253906, "logps/ref_rejected": -96.35814666748047, "logps/rejected": -99.5354995727539, "loss": 0.8143, "margin_dpo/margin_mean": 3.2729973793029785, "margin_dpo/margin_std": 3.099050521850586, "step": 50 }, { "epoch": 0.07488986784140969, "fcm_dpo/beta": 0.2618306875228882, "fcm_dpo/delta": -0.2724575996398926, "fcm_dpo/margin": 3.226135730743408, "fcm_dpo/q_t": 0.337465763092041, "grad_norm": 97.82896423339844, "learning_rate": 3.6231884057971015e-07, "logits/chosen": -0.5134993195533752, "logits/rejected": -0.48083585500717163, "logps/chosen": -61.43326950073242, "logps/ref_chosen": -61.3736686706543, "logps/ref_rejected": -76.00199890136719, "logps/rejected": -79.28773498535156, "loss": 0.9065, "margin_dpo/margin_mean": 3.2261362075805664, "margin_dpo/margin_std": 3.9684386253356934, "step": 51 }, { "epoch": 0.0763582966226138, "fcm_dpo/beta": 0.2358231097459793, "fcm_dpo/delta": -0.6680800914764404, "fcm_dpo/margin": 5.0320281982421875, "fcm_dpo/q_t": 0.2603490352630615, "grad_norm": 94.4676513671875, "learning_rate": 3.695652173913043e-07, "logits/chosen": -0.557870090007782, "logits/rejected": -0.5032240748405457, "logps/chosen": -51.786834716796875, "logps/ref_chosen": -52.33735656738281, "logps/ref_rejected": -79.97391510009766, "logps/rejected": -84.45541381835938, "loss": 0.6746, "margin_dpo/margin_mean": 5.0320281982421875, "margin_dpo/margin_std": 4.007593154907227, "step": 52 }, { "epoch": 0.07782672540381791, "fcm_dpo/beta": 0.21381568908691406, "fcm_dpo/delta": -0.5255783796310425, "fcm_dpo/margin": 5.014308452606201, "fcm_dpo/q_t": 0.298156201839447, "grad_norm": 91.9140853881836, "learning_rate": 3.7681159420289855e-07, "logits/chosen": -0.6126323342323303, "logits/rejected": -0.5912652015686035, "logps/chosen": -53.282958984375, "logps/ref_chosen": -53.31465148925781, "logps/ref_rejected": -91.78359985351562, "logps/rejected": -96.7662124633789, "loss": 0.7792, "margin_dpo/margin_mean": 5.014307975769043, "margin_dpo/margin_std": 4.957146644592285, "step": 53 }, { "epoch": 0.07929515418502203, "fcm_dpo/beta": 0.1964169442653656, "fcm_dpo/delta": -0.3215777575969696, "fcm_dpo/margin": 4.548583984375, "fcm_dpo/q_t": 0.3135607838630676, "grad_norm": 79.06949615478516, "learning_rate": 3.8405797101449274e-07, "logits/chosen": -0.5691754817962646, "logits/rejected": -0.5145428776741028, "logps/chosen": -50.683597564697266, "logps/ref_chosen": -50.68865966796875, "logps/ref_rejected": -91.71539306640625, "logps/rejected": -96.2589111328125, "loss": 0.8046, "margin_dpo/margin_mean": 4.548584461212158, "margin_dpo/margin_std": 4.153512954711914, "step": 54 }, { "epoch": 0.08076358296622614, "fcm_dpo/beta": 0.1820051074028015, "fcm_dpo/delta": -0.4129885137081146, "fcm_dpo/margin": 5.350862979888916, "fcm_dpo/q_t": 0.31068581342697144, "grad_norm": 79.86705780029297, "learning_rate": 3.9130434782608694e-07, "logits/chosen": -0.6421248912811279, "logits/rejected": -0.5801297426223755, "logps/chosen": -63.01799774169922, "logps/ref_chosen": -62.615234375, "logps/ref_rejected": -88.99349975585938, "logps/rejected": -94.74713134765625, "loss": 0.8358, "margin_dpo/margin_mean": 5.350862503051758, "margin_dpo/margin_std": 5.820072650909424, "step": 55 }, { "epoch": 0.08223201174743025, "fcm_dpo/beta": 0.17025861144065857, "fcm_dpo/delta": -0.2861342430114746, "fcm_dpo/margin": 5.0614013671875, "fcm_dpo/q_t": 0.32851046323776245, "grad_norm": 65.21794891357422, "learning_rate": 3.9855072463768114e-07, "logits/chosen": -0.6073439717292786, "logits/rejected": -0.5643370151519775, "logps/chosen": -58.113746643066406, "logps/ref_chosen": -57.9327278137207, "logps/ref_rejected": -94.1744384765625, "logps/rejected": -99.41685485839844, "loss": 0.8886, "margin_dpo/margin_mean": 5.061400890350342, "margin_dpo/margin_std": 5.794804573059082, "step": 56 }, { "epoch": 0.08370044052863436, "fcm_dpo/beta": 0.16040679812431335, "fcm_dpo/delta": -0.334885835647583, "fcm_dpo/margin": 5.6461501121521, "fcm_dpo/q_t": 0.3076401352882385, "grad_norm": 72.25611877441406, "learning_rate": 4.057971014492754e-07, "logits/chosen": -0.5476347208023071, "logits/rejected": -0.5185646414756775, "logps/chosen": -70.79481506347656, "logps/ref_chosen": -70.49528503417969, "logps/ref_rejected": -95.56546020507812, "logps/rejected": -101.51113891601562, "loss": 0.806, "margin_dpo/margin_mean": 5.646149635314941, "margin_dpo/margin_std": 4.98412561416626, "step": 57 }, { "epoch": 0.08516886930983847, "fcm_dpo/beta": 0.14835116267204285, "fcm_dpo/delta": -0.41062241792678833, "fcm_dpo/margin": 6.553308963775635, "fcm_dpo/q_t": 0.30739498138427734, "grad_norm": 73.45458984375, "learning_rate": 4.1304347826086954e-07, "logits/chosen": -0.5891647338867188, "logits/rejected": -0.513205885887146, "logps/chosen": -62.415225982666016, "logps/ref_chosen": -62.13294219970703, "logps/ref_rejected": -84.61729431152344, "logps/rejected": -91.45289611816406, "loss": 0.8144, "margin_dpo/margin_mean": 6.553309440612793, "margin_dpo/margin_std": 6.610756874084473, "step": 58 }, { "epoch": 0.08663729809104258, "fcm_dpo/beta": 0.13525643944740295, "fcm_dpo/delta": -0.35778504610061646, "fcm_dpo/margin": 6.788931846618652, "fcm_dpo/q_t": 0.3129827380180359, "grad_norm": 67.0660171508789, "learning_rate": 4.2028985507246374e-07, "logits/chosen": -0.6016139984130859, "logits/rejected": -0.558463454246521, "logps/chosen": -52.71605682373047, "logps/ref_chosen": -51.932525634765625, "logps/ref_rejected": -88.88520050048828, "logps/rejected": -96.45765686035156, "loss": 0.8265, "margin_dpo/margin_mean": 6.7889323234558105, "margin_dpo/margin_std": 6.713343620300293, "step": 59 }, { "epoch": 0.0881057268722467, "fcm_dpo/beta": 0.1309395432472229, "fcm_dpo/delta": -0.13870461285114288, "fcm_dpo/margin": 5.555472373962402, "fcm_dpo/q_t": 0.3427530527114868, "grad_norm": 79.94298553466797, "learning_rate": 4.2753623188405794e-07, "logits/chosen": -0.5748249888420105, "logits/rejected": -0.512750506401062, "logps/chosen": -62.536293029785156, "logps/ref_chosen": -60.94218826293945, "logps/ref_rejected": -85.39340209960938, "logps/rejected": -92.54298400878906, "loss": 0.9199, "margin_dpo/margin_mean": 5.555472373962402, "margin_dpo/margin_std": 6.041206359863281, "step": 60 }, { "epoch": 0.08957415565345081, "fcm_dpo/beta": 0.12486925721168518, "fcm_dpo/delta": -0.2642156481742859, "fcm_dpo/margin": 6.724519729614258, "fcm_dpo/q_t": 0.3386513590812683, "grad_norm": 61.240787506103516, "learning_rate": 4.3478260869565214e-07, "logits/chosen": -0.5803092122077942, "logits/rejected": -0.5446271896362305, "logps/chosen": -61.23955535888672, "logps/ref_chosen": -60.633522033691406, "logps/ref_rejected": -89.85249328613281, "logps/rejected": -97.18305206298828, "loss": 0.9087, "margin_dpo/margin_mean": 6.724519729614258, "margin_dpo/margin_std": 8.698123931884766, "step": 61 }, { "epoch": 0.09104258443465492, "fcm_dpo/beta": 0.12395919859409332, "fcm_dpo/delta": -0.035701360553503036, "fcm_dpo/margin": 5.107054233551025, "fcm_dpo/q_t": 0.3664395213127136, "grad_norm": 62.39984893798828, "learning_rate": 4.420289855072464e-07, "logits/chosen": -0.5827194452285767, "logits/rejected": -0.547878086566925, "logps/chosen": -57.02729034423828, "logps/ref_chosen": -56.15077209472656, "logps/ref_rejected": -75.56619262695312, "logps/rejected": -81.54976654052734, "loss": 0.9761, "margin_dpo/margin_mean": 5.107054710388184, "margin_dpo/margin_std": 6.175351619720459, "step": 62 }, { "epoch": 0.09251101321585903, "fcm_dpo/beta": 0.11800666153430939, "fcm_dpo/delta": -0.23345595598220825, "fcm_dpo/margin": 6.8794989585876465, "fcm_dpo/q_t": 0.3267975449562073, "grad_norm": 60.61161422729492, "learning_rate": 4.4927536231884053e-07, "logits/chosen": -0.5791250467300415, "logits/rejected": -0.5316081643104553, "logps/chosen": -74.61734008789062, "logps/ref_chosen": -73.14739227294922, "logps/ref_rejected": -97.61006164550781, "logps/rejected": -105.95950317382812, "loss": 0.8674, "margin_dpo/margin_mean": 6.879498481750488, "margin_dpo/margin_std": 6.8499274253845215, "step": 63 }, { "epoch": 0.09397944199706314, "fcm_dpo/beta": 0.11056903749704361, "fcm_dpo/delta": -0.3205702304840088, "fcm_dpo/margin": 8.005148887634277, "fcm_dpo/q_t": 0.32072484493255615, "grad_norm": 56.07626724243164, "learning_rate": 4.5652173913043473e-07, "logits/chosen": -0.5584267377853394, "logits/rejected": -0.5256708264350891, "logps/chosen": -54.30480194091797, "logps/ref_chosen": -53.998600006103516, "logps/ref_rejected": -93.53019714355469, "logps/rejected": -101.841552734375, "loss": 0.8584, "margin_dpo/margin_mean": 8.005149841308594, "margin_dpo/margin_std": 8.640867233276367, "step": 64 }, { "epoch": 0.09544787077826726, "fcm_dpo/beta": 0.10528378188610077, "fcm_dpo/delta": -0.26119303703308105, "fcm_dpo/margin": 7.948797225952148, "fcm_dpo/q_t": 0.32479870319366455, "grad_norm": 56.537654876708984, "learning_rate": 4.63768115942029e-07, "logits/chosen": -0.6330504417419434, "logits/rejected": -0.6177892684936523, "logps/chosen": -66.75464630126953, "logps/ref_chosen": -64.83599853515625, "logps/ref_rejected": -109.94645690917969, "logps/rejected": -119.81390380859375, "loss": 0.8793, "margin_dpo/margin_mean": 7.948797225952148, "margin_dpo/margin_std": 8.451780319213867, "step": 65 }, { "epoch": 0.09691629955947137, "fcm_dpo/beta": 0.10108963400125504, "fcm_dpo/delta": -0.16579298675060272, "fcm_dpo/margin": 7.422573089599609, "fcm_dpo/q_t": 0.34628963470458984, "grad_norm": 50.0311164855957, "learning_rate": 4.7101449275362313e-07, "logits/chosen": -0.6193344593048096, "logits/rejected": -0.5846483707427979, "logps/chosen": -53.2528076171875, "logps/ref_chosen": -51.44352722167969, "logps/ref_rejected": -75.63629913330078, "logps/rejected": -84.86815643310547, "loss": 0.9152, "margin_dpo/margin_mean": 7.422572135925293, "margin_dpo/margin_std": 8.596309661865234, "step": 66 }, { "epoch": 0.09838472834067548, "fcm_dpo/beta": 0.09882417321205139, "fcm_dpo/delta": -0.17645135521888733, "fcm_dpo/margin": 7.723736763000488, "fcm_dpo/q_t": 0.3419727683067322, "grad_norm": 51.04230499267578, "learning_rate": 4.782608695652174e-07, "logits/chosen": -0.5773788690567017, "logits/rejected": -0.5347466468811035, "logps/chosen": -60.845420837402344, "logps/ref_chosen": -59.34080505371094, "logps/ref_rejected": -72.78728485107422, "logps/rejected": -82.01564025878906, "loss": 0.9005, "margin_dpo/margin_mean": 7.72373628616333, "margin_dpo/margin_std": 8.540781021118164, "step": 67 }, { "epoch": 0.09985315712187959, "fcm_dpo/beta": 0.09611781686544418, "fcm_dpo/delta": -0.1308629810810089, "fcm_dpo/margin": 7.5075178146362305, "fcm_dpo/q_t": 0.34419840574264526, "grad_norm": 48.844181060791016, "learning_rate": 4.855072463768116e-07, "logits/chosen": -0.6071466207504272, "logits/rejected": -0.5482945442199707, "logps/chosen": -66.77816772460938, "logps/ref_chosen": -65.2058334350586, "logps/ref_rejected": -77.20724487304688, "logps/rejected": -86.28710174560547, "loss": 0.8922, "margin_dpo/margin_mean": 7.507518291473389, "margin_dpo/margin_std": 7.301938056945801, "step": 68 }, { "epoch": 0.1013215859030837, "fcm_dpo/beta": 0.09169034659862518, "fcm_dpo/delta": -0.20954345166683197, "fcm_dpo/margin": 8.629631042480469, "fcm_dpo/q_t": 0.3316349387168884, "grad_norm": 50.881256103515625, "learning_rate": 4.927536231884058e-07, "logits/chosen": -0.5818198323249817, "logits/rejected": -0.557062029838562, "logps/chosen": -61.803890228271484, "logps/ref_chosen": -59.81924057006836, "logps/ref_rejected": -103.38886260986328, "logps/rejected": -114.00314331054688, "loss": 0.8551, "margin_dpo/margin_mean": 8.629631042480469, "margin_dpo/margin_std": 8.243066787719727, "step": 69 }, { "epoch": 0.1027900146842878, "fcm_dpo/beta": 0.08815528452396393, "fcm_dpo/delta": -0.25355854630470276, "fcm_dpo/margin": 9.44918155670166, "fcm_dpo/q_t": 0.3277716040611267, "grad_norm": 51.42112350463867, "learning_rate": 5e-07, "logits/chosen": -0.6246175765991211, "logits/rejected": -0.5891268253326416, "logps/chosen": -64.80339050292969, "logps/ref_chosen": -61.930641174316406, "logps/ref_rejected": -91.06078338623047, "logps/rejected": -103.38272094726562, "loss": 0.8569, "margin_dpo/margin_mean": 9.449182510375977, "margin_dpo/margin_std": 9.51607894897461, "step": 70 }, { "epoch": 0.10425844346549193, "fcm_dpo/beta": 0.08285696059465408, "fcm_dpo/delta": -0.30268311500549316, "fcm_dpo/margin": 10.571916580200195, "fcm_dpo/q_t": 0.3174167275428772, "grad_norm": 45.99234390258789, "learning_rate": 4.999967061337492e-07, "logits/chosen": -0.6398484110832214, "logits/rejected": -0.5968210101127625, "logps/chosen": -64.16336822509766, "logps/ref_chosen": -61.750335693359375, "logps/ref_rejected": -97.33662414550781, "logps/rejected": -110.32157135009766, "loss": 0.8213, "margin_dpo/margin_mean": 10.571917533874512, "margin_dpo/margin_std": 9.830376625061035, "step": 71 }, { "epoch": 0.10572687224669604, "fcm_dpo/beta": 0.07784741371870041, "fcm_dpo/delta": -0.2931217849254608, "fcm_dpo/margin": 11.129253387451172, "fcm_dpo/q_t": 0.31700342893600464, "grad_norm": 48.5871696472168, "learning_rate": 4.999868246217933e-07, "logits/chosen": -0.6470623016357422, "logits/rejected": -0.6127582788467407, "logps/chosen": -68.55953216552734, "logps/ref_chosen": -66.05341339111328, "logps/ref_rejected": -95.2869873046875, "logps/rejected": -108.92236328125, "loss": 0.8263, "margin_dpo/margin_mean": 11.129253387451172, "margin_dpo/margin_std": 10.417036056518555, "step": 72 }, { "epoch": 0.10719530102790015, "fcm_dpo/beta": 0.07408735156059265, "fcm_dpo/delta": -0.2234322428703308, "fcm_dpo/margin": 10.851455688476562, "fcm_dpo/q_t": 0.3458544611930847, "grad_norm": 45.48418045043945, "learning_rate": 4.999703557245192e-07, "logits/chosen": -0.6490052938461304, "logits/rejected": -0.6046302914619446, "logps/chosen": -70.05899047851562, "logps/ref_chosen": -66.25627136230469, "logps/ref_rejected": -90.45613098144531, "logps/rejected": -105.11030578613281, "loss": 0.953, "margin_dpo/margin_mean": 10.851455688476562, "margin_dpo/margin_std": 14.541223526000977, "step": 73 }, { "epoch": 0.10866372980910426, "fcm_dpo/beta": 0.07084085047245026, "fcm_dpo/delta": -0.22987580299377441, "fcm_dpo/margin": 11.434507369995117, "fcm_dpo/q_t": 0.3413064479827881, "grad_norm": 47.909542083740234, "learning_rate": 4.999472998758977e-07, "logits/chosen": -0.6333368420600891, "logits/rejected": -0.6184351444244385, "logps/chosen": -57.71470260620117, "logps/ref_chosen": -53.42488098144531, "logps/ref_rejected": -95.94693756103516, "logps/rejected": -111.6712646484375, "loss": 0.9336, "margin_dpo/margin_mean": 11.4345064163208, "margin_dpo/margin_std": 15.739431381225586, "step": 74 }, { "epoch": 0.11013215859030837, "fcm_dpo/beta": 0.06567974388599396, "fcm_dpo/delta": -0.5119467377662659, "fcm_dpo/margin": 16.138355255126953, "fcm_dpo/q_t": 0.291500985622406, "grad_norm": 41.545692443847656, "learning_rate": 4.999176576834721e-07, "logits/chosen": -0.6396368741989136, "logits/rejected": -0.6279274225234985, "logps/chosen": -55.962501525878906, "logps/ref_chosen": -51.861663818359375, "logps/ref_rejected": -111.25398254394531, "logps/rejected": -131.4931640625, "loss": 0.7651, "margin_dpo/margin_mean": 16.138355255126953, "margin_dpo/margin_std": 15.203086853027344, "step": 75 }, { "epoch": 0.11160058737151249, "fcm_dpo/beta": 0.062492579221725464, "fcm_dpo/delta": -0.09926581382751465, "fcm_dpo/margin": 11.07880973815918, "fcm_dpo/q_t": 0.34875011444091797, "grad_norm": 41.94511795043945, "learning_rate": 4.998814299283415e-07, "logits/chosen": -0.7100133895874023, "logits/rejected": -0.6680034399032593, "logps/chosen": -58.421531677246094, "logps/ref_chosen": -53.26603698730469, "logps/ref_rejected": -78.21662902832031, "logps/rejected": -94.45092010498047, "loss": 0.9298, "margin_dpo/margin_mean": 11.078810691833496, "margin_dpo/margin_std": 12.21889877319336, "step": 76 }, { "epoch": 0.1130690161527166, "fcm_dpo/beta": 0.05767790973186493, "fcm_dpo/delta": -0.45353737473487854, "fcm_dpo/margin": 17.38974380493164, "fcm_dpo/q_t": 0.2953811585903168, "grad_norm": 43.50868606567383, "learning_rate": 4.998386175651409e-07, "logits/chosen": -0.6637353897094727, "logits/rejected": -0.6201448440551758, "logps/chosen": -62.73433303833008, "logps/ref_chosen": -58.0966796875, "logps/ref_rejected": -93.77361297607422, "logps/rejected": -115.80101013183594, "loss": 0.7985, "margin_dpo/margin_mean": 17.38974380493164, "margin_dpo/margin_std": 16.823715209960938, "step": 77 }, { "epoch": 0.1145374449339207, "fcm_dpo/beta": 0.0555085651576519, "fcm_dpo/delta": -0.18481989204883575, "fcm_dpo/margin": 13.888815879821777, "fcm_dpo/q_t": 0.3378889262676239, "grad_norm": 39.630531311035156, "learning_rate": 4.997892217220159e-07, "logits/chosen": -0.6263148784637451, "logits/rejected": -0.5966655015945435, "logps/chosen": -60.46125030517578, "logps/ref_chosen": -55.61378479003906, "logps/ref_rejected": -84.93436431884766, "logps/rejected": -103.67064666748047, "loss": 0.8902, "margin_dpo/margin_mean": 13.888814926147461, "margin_dpo/margin_std": 14.465457916259766, "step": 78 }, { "epoch": 0.11600587371512482, "fcm_dpo/beta": 0.05275849997997284, "fcm_dpo/delta": -0.24378372728824615, "fcm_dpo/margin": 15.608173370361328, "fcm_dpo/q_t": 0.3392306864261627, "grad_norm": 33.420074462890625, "learning_rate": 4.997332437005931e-07, "logits/chosen": -0.6634904146194458, "logits/rejected": -0.6324459314346313, "logps/chosen": -60.526432037353516, "logps/ref_chosen": -55.45048522949219, "logps/ref_rejected": -87.64756774902344, "logps/rejected": -108.33168029785156, "loss": 0.9079, "margin_dpo/margin_mean": 15.608173370361328, "margin_dpo/margin_std": 18.54351043701172, "step": 79 }, { "epoch": 0.11747430249632893, "fcm_dpo/beta": 0.05068628117442131, "fcm_dpo/delta": -0.12069609016180038, "fcm_dpo/margin": 13.985533714294434, "fcm_dpo/q_t": 0.3596472144126892, "grad_norm": 37.298439025878906, "learning_rate": 4.996706849759452e-07, "logits/chosen": -0.7097140550613403, "logits/rejected": -0.6619396805763245, "logps/chosen": -66.05055236816406, "logps/ref_chosen": -58.519290924072266, "logps/ref_rejected": -87.54750061035156, "logps/rejected": -109.06429290771484, "loss": 0.9679, "margin_dpo/margin_mean": 13.98553466796875, "margin_dpo/margin_std": 17.724082946777344, "step": 80 }, { "epoch": 0.11894273127753303, "fcm_dpo/beta": 0.047711554914712906, "fcm_dpo/delta": -0.3803252577781677, "fcm_dpo/margin": 19.70598602294922, "fcm_dpo/q_t": 0.31643375754356384, "grad_norm": 38.47947692871094, "learning_rate": 4.996015471965529e-07, "logits/chosen": -0.6981919407844543, "logits/rejected": -0.6639083623886108, "logps/chosen": -73.06269836425781, "logps/ref_chosen": -66.44886779785156, "logps/ref_rejected": -129.66270446777344, "logps/rejected": -155.98251342773438, "loss": 0.8483, "margin_dpo/margin_mean": 19.70598602294922, "margin_dpo/margin_std": 22.23188018798828, "step": 81 }, { "epoch": 0.12041116005873716, "fcm_dpo/beta": 0.04673580825328827, "fcm_dpo/delta": -0.15118417143821716, "fcm_dpo/margin": 15.803937911987305, "fcm_dpo/q_t": 0.3553317189216614, "grad_norm": 42.93745422363281, "learning_rate": 4.995258321842611e-07, "logits/chosen": -0.6455098390579224, "logits/rejected": -0.6292978525161743, "logps/chosen": -61.03137969970703, "logps/ref_chosen": -52.232383728027344, "logps/ref_rejected": -90.74325561523438, "logps/rejected": -115.34617614746094, "loss": 1.018, "margin_dpo/margin_mean": 15.803937911987305, "margin_dpo/margin_std": 23.58461570739746, "step": 82 }, { "epoch": 0.12187958883994127, "fcm_dpo/beta": 0.04404474422335625, "fcm_dpo/delta": -0.23587247729301453, "fcm_dpo/margin": 18.518720626831055, "fcm_dpo/q_t": 0.33574339747428894, "grad_norm": 40.53355407714844, "learning_rate": 4.994435419342304e-07, "logits/chosen": -0.6932875514030457, "logits/rejected": -0.6512936353683472, "logps/chosen": -64.86161804199219, "logps/ref_chosen": -55.82738494873047, "logps/ref_rejected": -103.71589660644531, "logps/rejected": -131.26885986328125, "loss": 0.9031, "margin_dpo/margin_mean": 18.518720626831055, "margin_dpo/margin_std": 21.3782901763916, "step": 83 }, { "epoch": 0.12334801762114538, "fcm_dpo/beta": 0.042480941861867905, "fcm_dpo/delta": -0.12278434634208679, "fcm_dpo/margin": 16.74671173095703, "fcm_dpo/q_t": 0.34350255131721497, "grad_norm": 34.17844772338867, "learning_rate": 4.993546786148857e-07, "logits/chosen": -0.6624174118041992, "logits/rejected": -0.624229907989502, "logps/chosen": -75.19337463378906, "logps/ref_chosen": -67.1761703491211, "logps/ref_rejected": -87.29859924316406, "logps/rejected": -112.06253051757812, "loss": 0.9041, "margin_dpo/margin_mean": 16.74671173095703, "margin_dpo/margin_std": 16.270580291748047, "step": 84 }, { "epoch": 0.12481644640234948, "fcm_dpo/beta": 0.042263105511665344, "fcm_dpo/delta": -0.1225767433643341, "fcm_dpo/margin": 16.87063217163086, "fcm_dpo/q_t": 0.3529721796512604, "grad_norm": 36.07400131225586, "learning_rate": 4.992592445678582e-07, "logits/chosen": -0.645004391670227, "logits/rejected": -0.6133354902267456, "logps/chosen": -67.26490783691406, "logps/ref_chosen": -58.4066162109375, "logps/ref_rejected": -78.63880157470703, "logps/rejected": -104.36772155761719, "loss": 0.9467, "margin_dpo/margin_mean": 16.87063217163086, "margin_dpo/margin_std": 19.575843811035156, "step": 85 }, { "epoch": 0.1262848751835536, "fcm_dpo/beta": 0.04085496813058853, "fcm_dpo/delta": -0.15022027492523193, "fcm_dpo/margin": 18.096515655517578, "fcm_dpo/q_t": 0.3664831817150116, "grad_norm": 40.77984619140625, "learning_rate": 4.991572423079235e-07, "logits/chosen": -0.6773054599761963, "logits/rejected": -0.6616036891937256, "logps/chosen": -67.63337707519531, "logps/ref_chosen": -56.13746643066406, "logps/ref_rejected": -88.12165069580078, "logps/rejected": -117.71407318115234, "loss": 1.0372, "margin_dpo/margin_mean": 18.096515655517578, "margin_dpo/margin_std": 29.443721771240234, "step": 86 }, { "epoch": 0.1277533039647577, "fcm_dpo/beta": 0.03880458325147629, "fcm_dpo/delta": -0.21359695494174957, "fcm_dpo/margin": 20.468475341796875, "fcm_dpo/q_t": 0.3394929766654968, "grad_norm": 33.87302780151367, "learning_rate": 4.990486745229364e-07, "logits/chosen": -0.7122533917427063, "logits/rejected": -0.6807093620300293, "logps/chosen": -67.1455078125, "logps/ref_chosen": -55.63609313964844, "logps/ref_rejected": -95.46757507324219, "logps/rejected": -127.44546508789062, "loss": 0.943, "margin_dpo/margin_mean": 20.468475341796875, "margin_dpo/margin_std": 25.73678207397461, "step": 87 }, { "epoch": 0.12922173274596183, "fcm_dpo/beta": 0.038395434617996216, "fcm_dpo/delta": 0.0005285218358039856, "fcm_dpo/margin": 15.591540336608887, "fcm_dpo/q_t": 0.3785722851753235, "grad_norm": 36.46826171875, "learning_rate": 4.989335440737586e-07, "logits/chosen": -0.6723873615264893, "logits/rejected": -0.6606534719467163, "logps/chosen": -88.06733703613281, "logps/ref_chosen": -73.67115020751953, "logps/ref_rejected": -106.70849609375, "logps/rejected": -136.69622802734375, "loss": 1.0542, "margin_dpo/margin_mean": 15.59153938293457, "margin_dpo/margin_std": 22.799100875854492, "step": 88 }, { "epoch": 0.13069016152716592, "fcm_dpo/beta": 0.03851180896162987, "fcm_dpo/delta": -0.06854115426540375, "fcm_dpo/margin": 17.231027603149414, "fcm_dpo/q_t": 0.36006081104278564, "grad_norm": 31.71763038635254, "learning_rate": 4.988118539941847e-07, "logits/chosen": -0.7132691144943237, "logits/rejected": -0.6763544082641602, "logps/chosen": -69.2666015625, "logps/ref_chosen": -60.624916076660156, "logps/ref_rejected": -82.08354949951172, "logps/rejected": -107.95626831054688, "loss": 0.9577, "margin_dpo/margin_mean": 17.231027603149414, "margin_dpo/margin_std": 20.84075927734375, "step": 89 }, { "epoch": 0.13215859030837004, "fcm_dpo/beta": 0.037114374339580536, "fcm_dpo/delta": -0.26081135869026184, "fcm_dpo/margin": 22.61049461364746, "fcm_dpo/q_t": 0.34644296765327454, "grad_norm": 35.28056716918945, "learning_rate": 4.986836074908615e-07, "logits/chosen": -0.6307741403579712, "logits/rejected": -0.6357797384262085, "logps/chosen": -64.71424102783203, "logps/ref_chosen": -53.285308837890625, "logps/ref_rejected": -111.54470825195312, "logps/rejected": -145.58413696289062, "loss": 0.9619, "margin_dpo/margin_mean": 22.610496520996094, "margin_dpo/margin_std": 31.246440887451172, "step": 90 }, { "epoch": 0.13362701908957417, "fcm_dpo/beta": 0.035702183842659, "fcm_dpo/delta": -0.12644991278648376, "fcm_dpo/margin": 20.09602928161621, "fcm_dpo/q_t": 0.3567592203617096, "grad_norm": 33.433372497558594, "learning_rate": 4.985488079432037e-07, "logits/chosen": -0.6847516894340515, "logits/rejected": -0.6470792889595032, "logps/chosen": -73.15221405029297, "logps/ref_chosen": -61.802955627441406, "logps/ref_rejected": -87.87395477294922, "logps/rejected": -119.31924438476562, "loss": 0.9724, "margin_dpo/margin_mean": 20.096027374267578, "margin_dpo/margin_std": 26.328693389892578, "step": 91 }, { "epoch": 0.13509544787077826, "fcm_dpo/beta": 0.03499949723482132, "fcm_dpo/delta": -0.10029930621385574, "fcm_dpo/margin": 19.803434371948242, "fcm_dpo/q_t": 0.36030054092407227, "grad_norm": 30.455114364624023, "learning_rate": 4.984074589033043e-07, "logits/chosen": -0.7354650497436523, "logits/rejected": -0.7112290859222412, "logps/chosen": -62.292938232421875, "logps/ref_chosen": -51.640769958496094, "logps/ref_rejected": -77.88117980957031, "logps/rejected": -108.3367919921875, "loss": 0.9797, "margin_dpo/margin_mean": 19.803434371948242, "margin_dpo/margin_std": 25.99898910522461, "step": 92 }, { "epoch": 0.13656387665198239, "fcm_dpo/beta": 0.03412316367030144, "fcm_dpo/delta": -0.08976341784000397, "fcm_dpo/margin": 20.03380584716797, "fcm_dpo/q_t": 0.3562747538089752, "grad_norm": 30.054187774658203, "learning_rate": 4.982595640958425e-07, "logits/chosen": -0.7533690333366394, "logits/rejected": -0.6998313069343567, "logps/chosen": -64.41513061523438, "logps/ref_chosen": -52.529239654541016, "logps/ref_rejected": -77.16075134277344, "logps/rejected": -109.0804443359375, "loss": 0.9377, "margin_dpo/margin_mean": 20.03380584716797, "margin_dpo/margin_std": 23.324481964111328, "step": 93 }, { "epoch": 0.13803230543318648, "fcm_dpo/beta": 0.0329783670604229, "fcm_dpo/delta": -0.1528889387845993, "fcm_dpo/margin": 22.42888641357422, "fcm_dpo/q_t": 0.3440149128437042, "grad_norm": 30.21152114868164, "learning_rate": 4.98105127417984e-07, "logits/chosen": -0.6753390431404114, "logits/rejected": -0.6542376279830933, "logps/chosen": -74.20423889160156, "logps/ref_chosen": -61.22261047363281, "logps/ref_rejected": -99.59902954101562, "logps/rejected": -135.00955200195312, "loss": 0.9066, "margin_dpo/margin_mean": 22.42888641357422, "margin_dpo/margin_std": 23.697879791259766, "step": 94 }, { "epoch": 0.1395007342143906, "fcm_dpo/beta": 0.03280120715498924, "fcm_dpo/delta": -0.006751883774995804, "fcm_dpo/margin": 18.477895736694336, "fcm_dpo/q_t": 0.36756014823913574, "grad_norm": 28.341552734375, "learning_rate": 4.979441529392784e-07, "logits/chosen": -0.732563853263855, "logits/rejected": -0.701493501663208, "logps/chosen": -64.28013610839844, "logps/ref_chosen": -52.523643493652344, "logps/ref_rejected": -75.8803482055664, "logps/rejected": -106.11474609375, "loss": 0.981, "margin_dpo/margin_mean": 18.477895736694336, "margin_dpo/margin_std": 21.757883071899414, "step": 95 }, { "epoch": 0.14096916299559473, "fcm_dpo/beta": 0.03166916221380234, "fcm_dpo/delta": -0.20488545298576355, "fcm_dpo/margin": 24.822622299194336, "fcm_dpo/q_t": 0.3381555676460266, "grad_norm": 29.113483428955078, "learning_rate": 4.977766449015534e-07, "logits/chosen": -0.7252618670463562, "logits/rejected": -0.6913504600524902, "logps/chosen": -73.15214538574219, "logps/ref_chosen": -62.15697479248047, "logps/ref_rejected": -96.59601593017578, "logps/rejected": -132.413818359375, "loss": 0.888, "margin_dpo/margin_mean": 24.82262420654297, "margin_dpo/margin_std": 27.286436080932617, "step": 96 }, { "epoch": 0.14243759177679882, "fcm_dpo/beta": 0.03180006146430969, "fcm_dpo/delta": -0.026856690645217896, "fcm_dpo/margin": 19.596725463867188, "fcm_dpo/q_t": 0.36258357763290405, "grad_norm": 30.518535614013672, "learning_rate": 4.976026077188012e-07, "logits/chosen": -0.6421018838882446, "logits/rejected": -0.591498851776123, "logps/chosen": -66.89000701904297, "logps/ref_chosen": -54.646366119384766, "logps/ref_rejected": -76.96475219726562, "logps/rejected": -108.80511474609375, "loss": 0.9602, "margin_dpo/margin_mean": 19.596725463867188, "margin_dpo/margin_std": 20.48422622680664, "step": 97 }, { "epoch": 0.14390602055800295, "fcm_dpo/beta": 0.03093864396214485, "fcm_dpo/delta": -0.09095153957605362, "fcm_dpo/margin": 22.12537384033203, "fcm_dpo/q_t": 0.35298076272010803, "grad_norm": 31.037092208862305, "learning_rate": 4.974220459770639e-07, "logits/chosen": -0.7037702798843384, "logits/rejected": -0.684203028678894, "logps/chosen": -80.09527587890625, "logps/ref_chosen": -65.25862884521484, "logps/ref_rejected": -96.5274887084961, "logps/rejected": -133.48951721191406, "loss": 0.9689, "margin_dpo/margin_mean": 22.12537384033203, "margin_dpo/margin_std": 27.224925994873047, "step": 98 }, { "epoch": 0.14537444933920704, "fcm_dpo/beta": 0.029884759336709976, "fcm_dpo/delta": -0.19739775359630585, "fcm_dpo/margin": 26.147565841674805, "fcm_dpo/q_t": 0.3427448868751526, "grad_norm": 27.304397583007812, "learning_rate": 4.972349644343108e-07, "logits/chosen": -0.694936990737915, "logits/rejected": -0.6911288499832153, "logps/chosen": -57.74766159057617, "logps/ref_chosen": -45.638484954833984, "logps/ref_rejected": -86.43793487548828, "logps/rejected": -124.69467163085938, "loss": 0.9019, "margin_dpo/margin_mean": 26.147567749023438, "margin_dpo/margin_std": 30.577857971191406, "step": 99 }, { "epoch": 0.14684287812041116, "fcm_dpo/beta": 0.0302484929561615, "fcm_dpo/delta": 0.1044018343091011, "fcm_dpo/margin": 16.52963638305664, "fcm_dpo/q_t": 0.3935774266719818, "grad_norm": 31.379667282104492, "learning_rate": 4.970413680203148e-07, "logits/chosen": -0.6751364469528198, "logits/rejected": -0.6300145983695984, "logps/chosen": -70.98727416992188, "logps/ref_chosen": -57.59397888183594, "logps/ref_rejected": -74.06021118164062, "logps/rejected": -103.98313903808594, "loss": 1.0889, "margin_dpo/margin_mean": 16.529634475708008, "margin_dpo/margin_std": 25.98699188232422, "step": 100 }, { "epoch": 0.14831130690161526, "fcm_dpo/beta": 0.030119918286800385, "fcm_dpo/delta": 0.012738943099975586, "fcm_dpo/margin": 19.492538452148438, "fcm_dpo/q_t": 0.382551908493042, "grad_norm": 31.488115310668945, "learning_rate": 4.968412618365215e-07, "logits/chosen": -0.7016730308532715, "logits/rejected": -0.6665774583816528, "logps/chosen": -77.9628677368164, "logps/ref_chosen": -61.64885330200195, "logps/ref_rejected": -83.18968200683594, "logps/rejected": -118.99624633789062, "loss": 1.0464, "margin_dpo/margin_mean": 19.492538452148438, "margin_dpo/margin_std": 29.35474395751953, "step": 101 }, { "epoch": 0.14977973568281938, "fcm_dpo/beta": 0.03130955249071121, "fcm_dpo/delta": 0.14849652349948883, "fcm_dpo/margin": 14.546876907348633, "fcm_dpo/q_t": 0.40587127208709717, "grad_norm": 37.16018295288086, "learning_rate": 4.966346511559149e-07, "logits/chosen": -0.6912135481834412, "logits/rejected": -0.6412276029586792, "logps/chosen": -82.47193145751953, "logps/ref_chosen": -64.0788803100586, "logps/ref_rejected": -68.18707275390625, "logps/rejected": -101.12701416015625, "loss": 1.142, "margin_dpo/margin_mean": 14.546875, "margin_dpo/margin_std": 26.82202911376953, "step": 102 }, { "epoch": 0.1512481644640235, "fcm_dpo/beta": 0.030011659488081932, "fcm_dpo/delta": -0.2606973648071289, "fcm_dpo/margin": 27.926231384277344, "fcm_dpo/q_t": 0.3282929062843323, "grad_norm": 28.890342712402344, "learning_rate": 4.964215414228785e-07, "logits/chosen": -0.6866222620010376, "logits/rejected": -0.6497406959533691, "logps/chosen": -74.54656982421875, "logps/ref_chosen": -61.299278259277344, "logps/ref_rejected": -93.57270812988281, "logps/rejected": -134.74624633789062, "loss": 0.8619, "margin_dpo/margin_mean": 27.926231384277344, "margin_dpo/margin_std": 29.437318801879883, "step": 103 }, { "epoch": 0.1527165932452276, "fcm_dpo/beta": 0.028996918350458145, "fcm_dpo/delta": -0.1786722093820572, "fcm_dpo/margin": 26.393293380737305, "fcm_dpo/q_t": 0.3505978584289551, "grad_norm": 29.279157638549805, "learning_rate": 4.96201938253052e-07, "logits/chosen": -0.7208267450332642, "logits/rejected": -0.683314859867096, "logps/chosen": -69.56233215332031, "logps/ref_chosen": -54.372772216796875, "logps/ref_rejected": -89.5647201538086, "logps/rejected": -131.14756774902344, "loss": 0.9468, "margin_dpo/margin_mean": 26.393293380737305, "margin_dpo/margin_std": 33.748008728027344, "step": 104 }, { "epoch": 0.15418502202643172, "fcm_dpo/beta": 0.026863520964980125, "fcm_dpo/delta": -0.43884339928627014, "fcm_dpo/margin": 37.05156707763672, "fcm_dpo/q_t": 0.28996556997299194, "grad_norm": 27.160659790039062, "learning_rate": 4.959758474331832e-07, "logits/chosen": -0.7008599638938904, "logits/rejected": -0.6714988946914673, "logps/chosen": -68.3555679321289, "logps/ref_chosen": -54.638946533203125, "logps/ref_rejected": -97.97351837158203, "logps/rejected": -148.74169921875, "loss": 0.7501, "margin_dpo/margin_mean": 37.05156707763672, "margin_dpo/margin_std": 30.501293182373047, "step": 105 }, { "epoch": 0.15565345080763582, "fcm_dpo/beta": 0.026156390085816383, "fcm_dpo/delta": -0.01053478755056858, "fcm_dpo/margin": 23.31535530090332, "fcm_dpo/q_t": 0.3665970265865326, "grad_norm": 27.37348747253418, "learning_rate": 4.957432749209755e-07, "logits/chosen": -0.650317370891571, "logits/rejected": -0.6077761650085449, "logps/chosen": -70.25245666503906, "logps/ref_chosen": -54.83289337158203, "logps/ref_rejected": -85.22461700439453, "logps/rejected": -123.95954895019531, "loss": 0.9688, "margin_dpo/margin_mean": 23.315357208251953, "margin_dpo/margin_std": 26.133378982543945, "step": 106 }, { "epoch": 0.15712187958883994, "fcm_dpo/beta": 0.025723904371261597, "fcm_dpo/delta": -0.0867917537689209, "fcm_dpo/margin": 26.438936233520508, "fcm_dpo/q_t": 0.3539083003997803, "grad_norm": 26.59319496154785, "learning_rate": 4.955042268449307e-07, "logits/chosen": -0.7173153162002563, "logits/rejected": -0.6675245761871338, "logps/chosen": -88.4300537109375, "logps/ref_chosen": -69.70780944824219, "logps/ref_rejected": -94.73950958251953, "logps/rejected": -139.90069580078125, "loss": 0.9479, "margin_dpo/margin_mean": 26.438936233520508, "margin_dpo/margin_std": 29.977489471435547, "step": 107 }, { "epoch": 0.15859030837004406, "fcm_dpo/beta": 0.025118088349699974, "fcm_dpo/delta": -0.1593261957168579, "fcm_dpo/margin": 29.753965377807617, "fcm_dpo/q_t": 0.3548673987388611, "grad_norm": 27.372953414916992, "learning_rate": 4.952587095041881e-07, "logits/chosen": -0.6825627088546753, "logits/rejected": -0.6495173573493958, "logps/chosen": -73.43077087402344, "logps/ref_chosen": -56.0098876953125, "logps/ref_rejected": -95.79601287841797, "logps/rejected": -142.97085571289062, "loss": 0.9717, "margin_dpo/margin_mean": 29.75396728515625, "margin_dpo/margin_std": 39.8942756652832, "step": 108 }, { "epoch": 0.16005873715124816, "fcm_dpo/beta": 0.02407240867614746, "fcm_dpo/delta": -0.20850417017936707, "fcm_dpo/margin": 32.8599853515625, "fcm_dpo/q_t": 0.33347180485725403, "grad_norm": 26.96596908569336, "learning_rate": 4.95006729368358e-07, "logits/chosen": -0.6184978485107422, "logits/rejected": -0.5932357311248779, "logps/chosen": -78.72866821289062, "logps/ref_chosen": -62.88549041748047, "logps/ref_rejected": -98.68573760986328, "logps/rejected": -147.38890075683594, "loss": 0.8832, "margin_dpo/margin_mean": 32.8599853515625, "margin_dpo/margin_std": 33.96752166748047, "step": 109 }, { "epoch": 0.16152716593245228, "fcm_dpo/beta": 0.023202266544103622, "fcm_dpo/delta": -0.10694625228643417, "fcm_dpo/margin": 29.92901611328125, "fcm_dpo/q_t": 0.3550639748573303, "grad_norm": 24.584754943847656, "learning_rate": 4.947482930773511e-07, "logits/chosen": -0.6574271321296692, "logits/rejected": -0.6152883768081665, "logps/chosen": -74.33702087402344, "logps/ref_chosen": -58.753684997558594, "logps/ref_rejected": -79.75001525878906, "logps/rejected": -125.26237487792969, "loss": 0.9612, "margin_dpo/margin_mean": 29.92901611328125, "margin_dpo/margin_std": 35.73463439941406, "step": 110 }, { "epoch": 0.16299559471365638, "fcm_dpo/beta": 0.022714221850037575, "fcm_dpo/delta": -0.15530934929847717, "fcm_dpo/margin": 32.667232513427734, "fcm_dpo/q_t": 0.347843736410141, "grad_norm": 27.264070510864258, "learning_rate": 4.944834074412042e-07, "logits/chosen": -0.6808880567550659, "logits/rejected": -0.6555056571960449, "logps/chosen": -87.53121948242188, "logps/ref_chosen": -68.62410736083984, "logps/ref_rejected": -98.42886352539062, "logps/rejected": -150.00320434570312, "loss": 0.9492, "margin_dpo/margin_mean": 32.667232513427734, "margin_dpo/margin_std": 39.969459533691406, "step": 111 }, { "epoch": 0.1644640234948605, "fcm_dpo/beta": 0.022934526205062866, "fcm_dpo/delta": 0.11087077856063843, "fcm_dpo/margin": 21.566303253173828, "fcm_dpo/q_t": 0.39232131838798523, "grad_norm": 25.337993621826172, "learning_rate": 4.942120794399002e-07, "logits/chosen": -0.6598977446556091, "logits/rejected": -0.613295316696167, "logps/chosen": -67.96663665771484, "logps/ref_chosen": -50.24964141845703, "logps/ref_rejected": -64.77442932128906, "logps/rejected": -104.05772399902344, "loss": 1.0543, "margin_dpo/margin_mean": 21.566301345825195, "margin_dpo/margin_std": 29.049222946166992, "step": 112 }, { "epoch": 0.16593245227606462, "fcm_dpo/beta": 0.023325610905885696, "fcm_dpo/delta": 0.040511354804039, "fcm_dpo/margin": 24.083799362182617, "fcm_dpo/q_t": 0.37566256523132324, "grad_norm": 25.780298233032227, "learning_rate": 4.939343162231841e-07, "logits/chosen": -0.647671639919281, "logits/rejected": -0.5998808741569519, "logps/chosen": -87.71575927734375, "logps/ref_chosen": -66.71295166015625, "logps/ref_rejected": -77.96870422363281, "logps/rejected": -123.0553207397461, "loss": 0.9905, "margin_dpo/margin_mean": 24.083797454833984, "margin_dpo/margin_std": 27.016586303710938, "step": 113 }, { "epoch": 0.16740088105726872, "fcm_dpo/beta": 0.02235432155430317, "fcm_dpo/delta": -0.24365481734275818, "fcm_dpo/margin": 36.693424224853516, "fcm_dpo/q_t": 0.33775386214256287, "grad_norm": 25.40207862854004, "learning_rate": 4.936501251103751e-07, "logits/chosen": -0.6867313385009766, "logits/rejected": -0.6511885523796082, "logps/chosen": -77.00772094726562, "logps/ref_chosen": -57.78507995605469, "logps/ref_rejected": -87.10966491699219, "logps/rejected": -143.02572631835938, "loss": 0.8963, "margin_dpo/margin_mean": 36.693424224853516, "margin_dpo/margin_std": 44.51530456542969, "step": 114 }, { "epoch": 0.16886930983847284, "fcm_dpo/beta": 0.022399384528398514, "fcm_dpo/delta": 0.014765216037631035, "fcm_dpo/margin": 26.151029586791992, "fcm_dpo/q_t": 0.38476622104644775, "grad_norm": 38.612613677978516, "learning_rate": 4.933595135901732e-07, "logits/chosen": -0.6795483827590942, "logits/rejected": -0.6489361524581909, "logps/chosen": -91.45716094970703, "logps/ref_chosen": -65.5826416015625, "logps/ref_rejected": -98.56552124023438, "logps/rejected": -150.591064453125, "loss": 1.0837, "margin_dpo/margin_mean": 26.15102767944336, "margin_dpo/margin_std": 44.87915802001953, "step": 115 }, { "epoch": 0.17033773861967694, "fcm_dpo/beta": 0.02236928418278694, "fcm_dpo/delta": -0.02009018138051033, "fcm_dpo/margin": 27.650157928466797, "fcm_dpo/q_t": 0.3669181764125824, "grad_norm": 28.57147979736328, "learning_rate": 4.930624893204624e-07, "logits/chosen": -0.6698657274246216, "logits/rejected": -0.6501831412315369, "logps/chosen": -70.91404724121094, "logps/ref_chosen": -51.40031433105469, "logps/ref_rejected": -80.5218505859375, "logps/rejected": -127.68574523925781, "loss": 0.9627, "margin_dpo/margin_mean": 27.650157928466797, "margin_dpo/margin_std": 31.316654205322266, "step": 116 }, { "epoch": 0.17180616740088106, "fcm_dpo/beta": 0.022425107657909393, "fcm_dpo/delta": 0.05522657185792923, "fcm_dpo/margin": 24.42792510986328, "fcm_dpo/q_t": 0.3833027184009552, "grad_norm": 33.80198287963867, "learning_rate": 4.927590601281083e-07, "logits/chosen": -0.6356416940689087, "logits/rejected": -0.594796895980835, "logps/chosen": -93.27058410644531, "logps/ref_chosen": -69.29840850830078, "logps/ref_rejected": -66.583984375, "logps/rejected": -114.98409271240234, "loss": 1.0417, "margin_dpo/margin_mean": 24.427927017211914, "margin_dpo/margin_std": 34.87114715576172, "step": 117 }, { "epoch": 0.17327459618208516, "fcm_dpo/beta": 0.022451553493738174, "fcm_dpo/delta": -0.012901581823825836, "fcm_dpo/margin": 27.263755798339844, "fcm_dpo/q_t": 0.36932849884033203, "grad_norm": 27.19830894470215, "learning_rate": 4.924492340087524e-07, "logits/chosen": -0.6901913285255432, "logits/rejected": -0.6656088829040527, "logps/chosen": -76.18692779541016, "logps/ref_chosen": -55.6409797668457, "logps/ref_rejected": -75.66905975341797, "logps/rejected": -123.478759765625, "loss": 0.9741, "margin_dpo/margin_mean": 27.263755798339844, "margin_dpo/margin_std": 32.05895233154297, "step": 118 }, { "epoch": 0.17474302496328928, "fcm_dpo/beta": 0.02224084362387657, "fcm_dpo/delta": -0.010049469769001007, "fcm_dpo/margin": 27.355573654174805, "fcm_dpo/q_t": 0.37495267391204834, "grad_norm": 29.576772689819336, "learning_rate": 4.92133019126601e-07, "logits/chosen": -0.6743392944335938, "logits/rejected": -0.6560527682304382, "logps/chosen": -100.41120910644531, "logps/ref_chosen": -73.51019287109375, "logps/ref_rejected": -102.977294921875, "logps/rejected": -157.23387145996094, "loss": 1.0134, "margin_dpo/margin_mean": 27.355573654174805, "margin_dpo/margin_std": 37.5633430480957, "step": 119 }, { "epoch": 0.1762114537444934, "fcm_dpo/beta": 0.02168981544673443, "fcm_dpo/delta": -0.2090209424495697, "fcm_dpo/margin": 36.529258728027344, "fcm_dpo/q_t": 0.3369954228401184, "grad_norm": 27.668405532836914, "learning_rate": 4.918104238142103e-07, "logits/chosen": -0.6739026308059692, "logits/rejected": -0.6356397867202759, "logps/chosen": -105.38459777832031, "logps/ref_chosen": -76.78083801269531, "logps/ref_rejected": -108.02374267578125, "logps/rejected": -173.15676879882812, "loss": 0.9002, "margin_dpo/margin_mean": 36.529258728027344, "margin_dpo/margin_std": 40.862762451171875, "step": 120 }, { "epoch": 0.1776798825256975, "fcm_dpo/beta": 0.020538385957479477, "fcm_dpo/delta": -0.2580873370170593, "fcm_dpo/margin": 40.62395477294922, "fcm_dpo/q_t": 0.33450600504875183, "grad_norm": 27.22615623474121, "learning_rate": 4.91481456572267e-07, "logits/chosen": -0.6262691020965576, "logits/rejected": -0.6172099113464355, "logps/chosen": -89.005126953125, "logps/ref_chosen": -61.789894104003906, "logps/ref_rejected": -109.99456787109375, "logps/rejected": -177.83375549316406, "loss": 0.9049, "margin_dpo/margin_mean": 40.623958587646484, "margin_dpo/margin_std": 47.878326416015625, "step": 121 }, { "epoch": 0.17914831130690162, "fcm_dpo/beta": 0.01918221451342106, "fcm_dpo/delta": -0.4117882251739502, "fcm_dpo/margin": 50.65019989013672, "fcm_dpo/q_t": 0.29864490032196045, "grad_norm": 28.19658660888672, "learning_rate": 4.911461260693638e-07, "logits/chosen": -0.6220955848693848, "logits/rejected": -0.6330730319023132, "logps/chosen": -71.84508514404297, "logps/ref_chosen": -46.9022102355957, "logps/ref_rejected": -106.71418762207031, "logps/rejected": -182.30726623535156, "loss": 0.7772, "margin_dpo/margin_mean": 50.65019989013672, "margin_dpo/margin_std": 44.97718048095703, "step": 122 }, { "epoch": 0.18061674008810572, "fcm_dpo/beta": 0.018608931452035904, "fcm_dpo/delta": -0.027324281632900238, "fcm_dpo/margin": 33.586708068847656, "fcm_dpo/q_t": 0.374822199344635, "grad_norm": 26.818113327026367, "learning_rate": 4.908044411417711e-07, "logits/chosen": -0.612564206123352, "logits/rejected": -0.5866974592208862, "logps/chosen": -88.52189636230469, "logps/ref_chosen": -61.33863830566406, "logps/ref_rejected": -87.775390625, "logps/rejected": -148.5453643798828, "loss": 1.0431, "margin_dpo/margin_mean": 33.586708068847656, "margin_dpo/margin_std": 51.27476501464844, "step": 123 }, { "epoch": 0.18208516886930984, "fcm_dpo/beta": 0.017872150987386703, "fcm_dpo/delta": -0.27416178584098816, "fcm_dpo/margin": 47.546592712402344, "fcm_dpo/q_t": 0.34260714054107666, "grad_norm": 28.516874313354492, "learning_rate": 4.904564107932048e-07, "logits/chosen": -0.6066104173660278, "logits/rejected": -0.6043754816055298, "logps/chosen": -101.30255126953125, "logps/ref_chosen": -71.44833374023438, "logps/ref_rejected": -117.58056640625, "logps/rejected": -194.98138427734375, "loss": 0.9423, "margin_dpo/margin_mean": 47.546592712402344, "margin_dpo/margin_std": 64.18199157714844, "step": 124 }, { "epoch": 0.18355359765051396, "fcm_dpo/beta": 0.017282985150814056, "fcm_dpo/delta": -0.17058980464935303, "fcm_dpo/margin": 43.85457229614258, "fcm_dpo/q_t": 0.3455929458141327, "grad_norm": 24.23063087463379, "learning_rate": 4.90102044194588e-07, "logits/chosen": -0.5887873768806458, "logits/rejected": -0.5830005407333374, "logps/chosen": -74.79535675048828, "logps/ref_chosen": -50.136940002441406, "logps/ref_rejected": -83.98861694335938, "logps/rejected": -152.50160217285156, "loss": 0.9197, "margin_dpo/margin_mean": 43.85457229614258, "margin_dpo/margin_std": 50.49039077758789, "step": 125 }, { "epoch": 0.18502202643171806, "fcm_dpo/beta": 0.016843030229210854, "fcm_dpo/delta": -0.08583886921405792, "fcm_dpo/margin": 40.36263656616211, "fcm_dpo/q_t": 0.35755184292793274, "grad_norm": 26.442380905151367, "learning_rate": 4.897413506838102e-07, "logits/chosen": -0.5875859260559082, "logits/rejected": -0.5705457925796509, "logps/chosen": -82.85122680664062, "logps/ref_chosen": -55.66706848144531, "logps/ref_rejected": -98.1297607421875, "logps/rejected": -165.67654418945312, "loss": 0.9522, "margin_dpo/margin_mean": 40.36263656616211, "margin_dpo/margin_std": 47.85084533691406, "step": 126 }, { "epoch": 0.18649045521292218, "fcm_dpo/beta": 0.0169706791639328, "fcm_dpo/delta": 0.061831459403038025, "fcm_dpo/margin": 31.902957916259766, "fcm_dpo/q_t": 0.38108593225479126, "grad_norm": 26.799118041992188, "learning_rate": 4.89374339765481e-07, "logits/chosen": -0.576623797416687, "logits/rejected": -0.5499337911605835, "logps/chosen": -82.35685729980469, "logps/ref_chosen": -56.55467987060547, "logps/ref_rejected": -76.7957763671875, "logps/rejected": -134.50091552734375, "loss": 1.0288, "margin_dpo/margin_mean": 31.902957916259766, "margin_dpo/margin_std": 41.576568603515625, "step": 127 }, { "epoch": 0.18795888399412627, "fcm_dpo/beta": 0.017191898077726364, "fcm_dpo/delta": 0.03887506574392319, "fcm_dpo/margin": 32.724117279052734, "fcm_dpo/q_t": 0.3826625347137451, "grad_norm": 38.53647994995117, "learning_rate": 4.890010211106795e-07, "logits/chosen": -0.5986074209213257, "logits/rejected": -0.5656411051750183, "logps/chosen": -86.5602035522461, "logps/ref_chosen": -58.12095642089844, "logps/ref_rejected": -76.43896484375, "logps/rejected": -137.60232543945312, "loss": 1.0607, "margin_dpo/margin_mean": 32.72411346435547, "margin_dpo/margin_std": 49.83234405517578, "step": 128 }, { "epoch": 0.1894273127753304, "fcm_dpo/beta": 0.017184410244226456, "fcm_dpo/delta": 0.005901549011468887, "fcm_dpo/margin": 34.58758544921875, "fcm_dpo/q_t": 0.3824169635772705, "grad_norm": 28.598745346069336, "learning_rate": 4.88621404556699e-07, "logits/chosen": -0.6040947437286377, "logits/rejected": -0.5859910845756531, "logps/chosen": -102.27650451660156, "logps/ref_chosen": -66.91637420654297, "logps/ref_rejected": -96.6422119140625, "logps/rejected": -166.58993530273438, "loss": 1.0624, "margin_dpo/margin_mean": 34.58758544921875, "margin_dpo/margin_std": 55.73461151123047, "step": 129 }, { "epoch": 0.19089574155653452, "fcm_dpo/beta": 0.016697879880666733, "fcm_dpo/delta": -0.2792219817638397, "fcm_dpo/margin": 51.21843338012695, "fcm_dpo/q_t": 0.3288246989250183, "grad_norm": 22.612878799438477, "learning_rate": 4.882355001067891e-07, "logits/chosen": -0.5792099237442017, "logits/rejected": -0.5733057260513306, "logps/chosen": -71.08717346191406, "logps/ref_chosen": -44.66685104370117, "logps/ref_rejected": -82.78165435791016, "logps/rejected": -160.42041015625, "loss": 0.8943, "margin_dpo/margin_mean": 51.21843338012695, "margin_dpo/margin_std": 56.39968490600586, "step": 130 }, { "epoch": 0.19236417033773862, "fcm_dpo/beta": 0.015755565837025642, "fcm_dpo/delta": -0.18772238492965698, "fcm_dpo/margin": 49.05485916137695, "fcm_dpo/q_t": 0.3336232304573059, "grad_norm": 26.170124053955078, "learning_rate": 4.878433179298909e-07, "logits/chosen": -0.5605712532997131, "logits/rejected": -0.564369261264801, "logps/chosen": -66.87252044677734, "logps/ref_chosen": -44.924591064453125, "logps/ref_rejected": -88.44401550292969, "logps/rejected": -159.44680786132812, "loss": 0.8818, "margin_dpo/margin_mean": 49.05485916137695, "margin_dpo/margin_std": 49.06263732910156, "step": 131 }, { "epoch": 0.19383259911894274, "fcm_dpo/beta": 0.015453611500561237, "fcm_dpo/delta": -0.05859919637441635, "fcm_dpo/margin": 42.35327911376953, "fcm_dpo/q_t": 0.36758387088775635, "grad_norm": 25.90082359313965, "learning_rate": 4.874448683603694e-07, "logits/chosen": -0.6056761741638184, "logits/rejected": -0.5935735702514648, "logps/chosen": -89.5750732421875, "logps/ref_chosen": -59.00108337402344, "logps/ref_rejected": -87.89215087890625, "logps/rejected": -160.81942749023438, "loss": 0.9903, "margin_dpo/margin_mean": 42.35327911376953, "margin_dpo/margin_std": 57.95384979248047, "step": 132 }, { "epoch": 0.19530102790014683, "fcm_dpo/beta": 0.01545734517276287, "fcm_dpo/delta": 0.0167352557182312, "fcm_dpo/margin": 37.79522705078125, "fcm_dpo/q_t": 0.38009342551231384, "grad_norm": 27.229202270507812, "learning_rate": 4.870401618977415e-07, "logits/chosen": -0.5968517065048218, "logits/rejected": -0.579937219619751, "logps/chosen": -102.51626586914062, "logps/ref_chosen": -66.60449981689453, "logps/ref_rejected": -96.33355712890625, "logps/rejected": -170.04055786132812, "loss": 1.0233, "margin_dpo/margin_mean": 37.795230865478516, "margin_dpo/margin_std": 53.25979995727539, "step": 133 }, { "epoch": 0.19676945668135096, "fcm_dpo/beta": 0.01544961053878069, "fcm_dpo/delta": -0.046767231076955795, "fcm_dpo/margin": 41.66034698486328, "fcm_dpo/q_t": 0.3632383346557617, "grad_norm": 24.29633903503418, "learning_rate": 4.866292092063986e-07, "logits/chosen": -0.562139630317688, "logits/rejected": -0.5372625589370728, "logps/chosen": -80.47048950195312, "logps/ref_chosen": -52.06925582885742, "logps/ref_rejected": -87.6545181274414, "logps/rejected": -157.71609497070312, "loss": 0.9512, "margin_dpo/margin_mean": 41.66035079956055, "margin_dpo/margin_std": 47.288726806640625, "step": 134 }, { "epoch": 0.19823788546255505, "fcm_dpo/beta": 0.014866790734231472, "fcm_dpo/delta": -0.21857857704162598, "fcm_dpo/margin": 53.88473892211914, "fcm_dpo/q_t": 0.3398352265357971, "grad_norm": 25.403532028198242, "learning_rate": 4.862120211153265e-07, "logits/chosen": -0.6001534461975098, "logits/rejected": -0.627263069152832, "logps/chosen": -82.99242401123047, "logps/ref_chosen": -50.353858947753906, "logps/ref_rejected": -115.97975158691406, "logps/rejected": -202.5030517578125, "loss": 0.9102, "margin_dpo/margin_mean": 53.88473892211914, "margin_dpo/margin_std": 62.925498962402344, "step": 135 }, { "epoch": 0.19970631424375918, "fcm_dpo/beta": 0.014702379703521729, "fcm_dpo/delta": 0.05534332990646362, "fcm_dpo/margin": 37.198089599609375, "fcm_dpo/q_t": 0.3941587805747986, "grad_norm": 27.017377853393555, "learning_rate": 4.857886086178193e-07, "logits/chosen": -0.5803062915802002, "logits/rejected": -0.5640683174133301, "logps/chosen": -103.38967895507812, "logps/ref_chosen": -65.072509765625, "logps/ref_rejected": -96.32122802734375, "logps/rejected": -171.83648681640625, "loss": 1.0679, "margin_dpo/margin_mean": 37.198089599609375, "margin_dpo/margin_std": 59.631011962890625, "step": 136 }, { "epoch": 0.2011747430249633, "fcm_dpo/beta": 0.014293761923909187, "fcm_dpo/delta": -0.24502620100975037, "fcm_dpo/margin": 57.65531921386719, "fcm_dpo/q_t": 0.3446000814437866, "grad_norm": 23.374046325683594, "learning_rate": 4.853589828711902e-07, "logits/chosen": -0.5490569472312927, "logits/rejected": -0.5711492300033569, "logps/chosen": -87.5726089477539, "logps/ref_chosen": -48.759117126464844, "logps/ref_rejected": -113.86376953125, "logps/rejected": -210.33258056640625, "loss": 0.9452, "margin_dpo/margin_mean": 57.65531921386719, "margin_dpo/margin_std": 76.925048828125, "step": 137 }, { "epoch": 0.2026431718061674, "fcm_dpo/beta": 0.014147358015179634, "fcm_dpo/delta": -0.0190581101924181, "fcm_dpo/margin": 43.66788864135742, "fcm_dpo/q_t": 0.3662746548652649, "grad_norm": 26.790586471557617, "learning_rate": 4.849231551964771e-07, "logits/chosen": -0.5593977570533752, "logits/rejected": -0.5456332564353943, "logps/chosen": -101.01687622070312, "logps/ref_chosen": -60.519649505615234, "logps/ref_rejected": -93.19694519042969, "logps/rejected": -177.362060546875, "loss": 0.9758, "margin_dpo/margin_mean": 43.667884826660156, "margin_dpo/margin_std": 51.35163879394531, "step": 138 }, { "epoch": 0.20411160058737152, "fcm_dpo/beta": 0.01379378791898489, "fcm_dpo/delta": -0.16102235019207, "fcm_dpo/margin": 54.313804626464844, "fcm_dpo/q_t": 0.34360840916633606, "grad_norm": 22.88560676574707, "learning_rate": 4.844811370781446e-07, "logits/chosen": -0.5527390241622925, "logits/rejected": -0.5373610258102417, "logps/chosen": -79.2032470703125, "logps/ref_chosen": -46.89138412475586, "logps/ref_rejected": -79.72798156738281, "logps/rejected": -166.3536376953125, "loss": 0.906, "margin_dpo/margin_mean": 54.313804626464844, "margin_dpo/margin_std": 59.772552490234375, "step": 139 }, { "epoch": 0.2055800293685756, "fcm_dpo/beta": 0.013468625023961067, "fcm_dpo/delta": -0.0790109634399414, "fcm_dpo/margin": 50.007423400878906, "fcm_dpo/q_t": 0.36189499497413635, "grad_norm": 24.64659309387207, "learning_rate": 4.840329401637809e-07, "logits/chosen": -0.5458736419677734, "logits/rejected": -0.5282671451568604, "logps/chosen": -98.45331573486328, "logps/ref_chosen": -58.97471618652344, "logps/ref_rejected": -83.28410339355469, "logps/rejected": -172.77012634277344, "loss": 0.9788, "margin_dpo/margin_mean": 50.007423400878906, "margin_dpo/margin_std": 64.80023193359375, "step": 140 }, { "epoch": 0.20704845814977973, "fcm_dpo/beta": 0.013357013463973999, "fcm_dpo/delta": -0.04534909874200821, "fcm_dpo/margin": 48.098968505859375, "fcm_dpo/q_t": 0.3626646399497986, "grad_norm": 28.8856258392334, "learning_rate": 4.83578576263792e-07, "logits/chosen": -0.5459233522415161, "logits/rejected": -0.5275824069976807, "logps/chosen": -116.89370727539062, "logps/ref_chosen": -75.07566833496094, "logps/ref_rejected": -98.1922607421875, "logps/rejected": -188.10928344726562, "loss": 0.9921, "margin_dpo/margin_mean": 48.098968505859375, "margin_dpo/margin_std": 62.48729705810547, "step": 141 }, { "epoch": 0.20851688693098386, "fcm_dpo/beta": 0.01312628760933876, "fcm_dpo/delta": -0.13432356715202332, "fcm_dpo/margin": 55.20783233642578, "fcm_dpo/q_t": 0.35684800148010254, "grad_norm": 28.54540252685547, "learning_rate": 4.83118057351089e-07, "logits/chosen": -0.5353626608848572, "logits/rejected": -0.5301727056503296, "logps/chosen": -103.46934509277344, "logps/ref_chosen": -58.027931213378906, "logps/ref_rejected": -94.58222961425781, "logps/rejected": -195.23147583007812, "loss": 0.9917, "margin_dpo/margin_mean": 55.20783233642578, "margin_dpo/margin_std": 75.06365203857422, "step": 142 }, { "epoch": 0.20998531571218795, "fcm_dpo/beta": 0.013182537630200386, "fcm_dpo/delta": 0.14479340612888336, "fcm_dpo/margin": 35.0554313659668, "fcm_dpo/q_t": 0.4067147374153137, "grad_norm": 31.684499740600586, "learning_rate": 4.826513955607734e-07, "logits/chosen": -0.5001394152641296, "logits/rejected": -0.48260974884033203, "logps/chosen": -107.36073303222656, "logps/ref_chosen": -57.59645080566406, "logps/ref_rejected": -78.99957275390625, "logps/rejected": -163.81927490234375, "loss": 1.1328, "margin_dpo/margin_mean": 35.05542755126953, "margin_dpo/margin_std": 64.47492980957031, "step": 143 }, { "epoch": 0.21145374449339208, "fcm_dpo/beta": 0.0133123230189085, "fcm_dpo/delta": -0.005581766366958618, "fcm_dpo/margin": 45.4619255065918, "fcm_dpo/q_t": 0.3696960508823395, "grad_norm": 24.83319854736328, "learning_rate": 4.821786031898176e-07, "logits/chosen": -0.5548040866851807, "logits/rejected": -0.5310955047607422, "logps/chosen": -100.95870971679688, "logps/ref_chosen": -59.90636444091797, "logps/ref_rejected": -82.00025939941406, "logps/rejected": -168.51454162597656, "loss": 0.9842, "margin_dpo/margin_mean": 45.46192932128906, "margin_dpo/margin_std": 54.452301025390625, "step": 144 }, { "epoch": 0.21292217327459617, "fcm_dpo/beta": 0.013245116919279099, "fcm_dpo/delta": -0.03191668167710304, "fcm_dpo/margin": 47.55194854736328, "fcm_dpo/q_t": 0.3665177524089813, "grad_norm": 29.93503189086914, "learning_rate": 4.816996926967401e-07, "logits/chosen": -0.5175589323043823, "logits/rejected": -0.48851436376571655, "logps/chosen": -98.98822021484375, "logps/ref_chosen": -56.60066604614258, "logps/ref_rejected": -77.86631774902344, "logps/rejected": -167.80581665039062, "loss": 0.9815, "margin_dpo/margin_mean": 47.55194854736328, "margin_dpo/margin_std": 58.7222900390625, "step": 145 }, { "epoch": 0.2143906020558003, "fcm_dpo/beta": 0.013389000669121742, "fcm_dpo/delta": 0.0990617573261261, "fcm_dpo/margin": 37.79475021362305, "fcm_dpo/q_t": 0.3926282227039337, "grad_norm": 36.985809326171875, "learning_rate": 4.812146767012779e-07, "logits/chosen": -0.5096883177757263, "logits/rejected": -0.4698811173439026, "logps/chosen": -124.57284545898438, "logps/ref_chosen": -66.00045013427734, "logps/ref_rejected": -81.70278930664062, "logps/rejected": -178.0699462890625, "loss": 1.1025, "margin_dpo/margin_mean": 37.79475021362305, "margin_dpo/margin_std": 62.888580322265625, "step": 146 }, { "epoch": 0.21585903083700442, "fcm_dpo/beta": 0.013290293514728546, "fcm_dpo/delta": -0.0870274156332016, "fcm_dpo/margin": 51.21372985839844, "fcm_dpo/q_t": 0.36140725016593933, "grad_norm": 26.684223175048828, "learning_rate": 4.807235679840536e-07, "logits/chosen": -0.5257313251495361, "logits/rejected": -0.49477505683898926, "logps/chosen": -96.1661605834961, "logps/ref_chosen": -53.405487060546875, "logps/ref_rejected": -71.39060974121094, "logps/rejected": -165.36502075195312, "loss": 0.9801, "margin_dpo/margin_mean": 51.21372985839844, "margin_dpo/margin_std": 67.14807891845703, "step": 147 }, { "epoch": 0.2173274596182085, "fcm_dpo/beta": 0.01351265050470829, "fcm_dpo/delta": 0.03601977974176407, "fcm_dpo/margin": 41.63152313232422, "fcm_dpo/q_t": 0.3840796947479248, "grad_norm": 26.11017608642578, "learning_rate": 4.802263794862384e-07, "logits/chosen": -0.5603736639022827, "logits/rejected": -0.5478102564811707, "logps/chosen": -106.5193862915039, "logps/ref_chosen": -64.93708038330078, "logps/ref_rejected": -103.09384155273438, "logps/rejected": -186.30767822265625, "loss": 1.0332, "margin_dpo/margin_mean": 41.63152313232422, "margin_dpo/margin_std": 55.343299865722656, "step": 148 }, { "epoch": 0.21879588839941264, "fcm_dpo/beta": 0.013042960315942764, "fcm_dpo/delta": -0.0962439477443695, "fcm_dpo/margin": 52.64115905761719, "fcm_dpo/q_t": 0.35281527042388916, "grad_norm": 24.8011417388916, "learning_rate": 4.797231243092118e-07, "logits/chosen": -0.590922474861145, "logits/rejected": -0.5712438225746155, "logps/chosen": -99.46907043457031, "logps/ref_chosen": -58.47376251220703, "logps/ref_rejected": -99.31474304199219, "logps/rejected": -192.95120239257812, "loss": 0.9376, "margin_dpo/margin_mean": 52.64115905761719, "margin_dpo/margin_std": 57.46501922607422, "step": 149 }, { "epoch": 0.22026431718061673, "fcm_dpo/beta": 0.012883552350103855, "fcm_dpo/delta": -0.049887098371982574, "fcm_dpo/margin": 49.994728088378906, "fcm_dpo/q_t": 0.37187010049819946, "grad_norm": 24.784076690673828, "learning_rate": 4.792138157142157e-07, "logits/chosen": -0.557221531867981, "logits/rejected": -0.5565592050552368, "logps/chosen": -82.40767669677734, "logps/ref_chosen": -45.705810546875, "logps/ref_rejected": -83.34759521484375, "logps/rejected": -170.044189453125, "loss": 0.9916, "margin_dpo/margin_mean": 49.994728088378906, "margin_dpo/margin_std": 68.09803771972656, "step": 150 }, { "epoch": 0.22173274596182085, "fcm_dpo/beta": 0.012838941067457199, "fcm_dpo/delta": -0.08010067045688629, "fcm_dpo/margin": 52.55384063720703, "fcm_dpo/q_t": 0.35465726256370544, "grad_norm": 29.252914428710938, "learning_rate": 4.786984671220053e-07, "logits/chosen": -0.6052649021148682, "logits/rejected": -0.569302499294281, "logps/chosen": -114.0267333984375, "logps/ref_chosen": -70.57083129882812, "logps/ref_rejected": -100.46382141113281, "logps/rejected": -196.47357177734375, "loss": 0.9399, "margin_dpo/margin_mean": 52.55384063720703, "margin_dpo/margin_std": 58.15677261352539, "step": 151 }, { "epoch": 0.22320117474302498, "fcm_dpo/beta": 0.012523850426077843, "fcm_dpo/delta": -0.16435271501541138, "fcm_dpo/margin": 60.06843185424805, "fcm_dpo/q_t": 0.34519827365875244, "grad_norm": 23.900650024414062, "learning_rate": 4.78177092112495e-07, "logits/chosen": -0.588593602180481, "logits/rejected": -0.5805606842041016, "logps/chosen": -98.1216049194336, "logps/ref_chosen": -60.16438674926758, "logps/ref_rejected": -106.14045715332031, "logps/rejected": -204.16610717773438, "loss": 0.918, "margin_dpo/margin_mean": 60.06843185424805, "margin_dpo/margin_std": 66.94328308105469, "step": 152 }, { "epoch": 0.22466960352422907, "fcm_dpo/beta": 0.012331589125096798, "fcm_dpo/delta": -0.03408785164356232, "fcm_dpo/margin": 51.24233627319336, "fcm_dpo/q_t": 0.37102195620536804, "grad_norm": 21.188068389892578, "learning_rate": 4.776497044244016e-07, "logits/chosen": -0.576537013053894, "logits/rejected": -0.5661931037902832, "logps/chosen": -94.53947448730469, "logps/ref_chosen": -56.315277099609375, "logps/ref_rejected": -85.65583801269531, "logps/rejected": -175.12237548828125, "loss": 1.0004, "margin_dpo/margin_mean": 51.24233627319336, "margin_dpo/margin_std": 70.23153686523438, "step": 153 }, { "epoch": 0.2261380323054332, "fcm_dpo/beta": 0.012307348661124706, "fcm_dpo/delta": -0.028648383915424347, "fcm_dpo/margin": 50.88260269165039, "fcm_dpo/q_t": 0.37250685691833496, "grad_norm": 26.30771255493164, "learning_rate": 4.771163179548808e-07, "logits/chosen": -0.5932099223136902, "logits/rejected": -0.5894089937210083, "logps/chosen": -107.8296127319336, "logps/ref_chosen": -62.74256896972656, "logps/ref_rejected": -104.24420166015625, "logps/rejected": -200.21383666992188, "loss": 1.0289, "margin_dpo/margin_mean": 50.88260269165039, "margin_dpo/margin_std": 72.57513427734375, "step": 154 }, { "epoch": 0.2276064610866373, "fcm_dpo/beta": 0.012188572436571121, "fcm_dpo/delta": -0.041290249675512314, "fcm_dpo/margin": 52.38173294067383, "fcm_dpo/q_t": 0.3654828369617462, "grad_norm": 24.021343231201172, "learning_rate": 4.7657694675916247e-07, "logits/chosen": -0.6054894924163818, "logits/rejected": -0.581174373626709, "logps/chosen": -101.20684814453125, "logps/ref_chosen": -60.65318298339844, "logps/ref_rejected": -77.49220275878906, "logps/rejected": -170.42758178710938, "loss": 0.9842, "margin_dpo/margin_mean": 52.38173294067383, "margin_dpo/margin_std": 66.93553161621094, "step": 155 }, { "epoch": 0.2290748898678414, "fcm_dpo/beta": 0.012476801872253418, "fcm_dpo/delta": 0.20647725462913513, "fcm_dpo/margin": 32.21952819824219, "fcm_dpo/q_t": 0.41537606716156006, "grad_norm": 38.838287353515625, "learning_rate": 4.7603160505017893e-07, "logits/chosen": -0.531362771987915, "logits/rejected": -0.5084357857704163, "logps/chosen": -126.44786071777344, "logps/ref_chosen": -69.49188232421875, "logps/ref_rejected": -77.16929626464844, "logps/rejected": -166.3448028564453, "loss": 1.2212, "margin_dpo/margin_mean": 32.21952819824219, "margin_dpo/margin_std": 76.02751159667969, "step": 156 }, { "epoch": 0.2305433186490455, "fcm_dpo/beta": 0.012192957103252411, "fcm_dpo/delta": -0.18633562326431274, "fcm_dpo/margin": 63.14281463623047, "fcm_dpo/q_t": 0.33702823519706726, "grad_norm": 27.627710342407227, "learning_rate": 4.7548030719819154e-07, "logits/chosen": -0.5252609252929688, "logits/rejected": -0.5241682529449463, "logps/chosen": -113.81539916992188, "logps/ref_chosen": -61.368438720703125, "logps/ref_rejected": -107.64636993408203, "logps/rejected": -223.23614501953125, "loss": 0.9149, "margin_dpo/margin_mean": 63.14281463623047, "margin_dpo/margin_std": 69.51899719238281, "step": 157 }, { "epoch": 0.23201174743024963, "fcm_dpo/beta": 0.011821776628494263, "fcm_dpo/delta": -0.1891898512840271, "fcm_dpo/margin": 65.51558685302734, "fcm_dpo/q_t": 0.35599738359451294, "grad_norm": 23.29702377319336, "learning_rate": 4.7492306773041136e-07, "logits/chosen": -0.5330479145050049, "logits/rejected": -0.5438076257705688, "logps/chosen": -108.17401123046875, "logps/ref_chosen": -57.612918853759766, "logps/ref_rejected": -113.6946792602539, "logps/rejected": -229.7713623046875, "loss": 0.9692, "margin_dpo/margin_mean": 65.51559448242188, "margin_dpo/margin_std": 92.29185485839844, "step": 158 }, { "epoch": 0.23348017621145375, "fcm_dpo/beta": 0.011839054524898529, "fcm_dpo/delta": 0.03908521309494972, "fcm_dpo/margin": 47.499595642089844, "fcm_dpo/q_t": 0.38455915451049805, "grad_norm": 30.979955673217773, "learning_rate": 4.743599013306165e-07, "logits/chosen": -0.5508168935775757, "logits/rejected": -0.5111355781555176, "logps/chosen": -139.3695526123047, "logps/ref_chosen": -81.56034851074219, "logps/ref_rejected": -88.89871215820312, "logps/rejected": -194.20751953125, "loss": 1.0618, "margin_dpo/margin_mean": 47.499595642089844, "margin_dpo/margin_std": 73.20787811279297, "step": 159 }, { "epoch": 0.23494860499265785, "fcm_dpo/beta": 0.011559359729290009, "fcm_dpo/delta": -0.1065782979130745, "fcm_dpo/margin": 60.404293060302734, "fcm_dpo/q_t": 0.365913450717926, "grad_norm": 25.069011688232422, "learning_rate": 4.737908228387656e-07, "logits/chosen": -0.5436577796936035, "logits/rejected": -0.5286530256271362, "logps/chosen": -123.36406707763672, "logps/ref_chosen": -65.73088073730469, "logps/ref_rejected": -97.21781921386719, "logps/rejected": -215.25531005859375, "loss": 1.0081, "margin_dpo/margin_mean": 60.404296875, "margin_dpo/margin_std": 88.11878967285156, "step": 160 }, { "epoch": 0.23641703377386197, "fcm_dpo/beta": 0.01151471957564354, "fcm_dpo/delta": -0.025925662368535995, "fcm_dpo/margin": 54.21880340576172, "fcm_dpo/q_t": 0.3725304901599884, "grad_norm": 25.731287002563477, "learning_rate": 4.7321584725060594e-07, "logits/chosen": -0.5345560312271118, "logits/rejected": -0.5234242677688599, "logps/chosen": -102.83772277832031, "logps/ref_chosen": -52.43647003173828, "logps/ref_rejected": -83.43095397949219, "logps/rejected": -188.05099487304688, "loss": 1.0083, "margin_dpo/margin_mean": 54.21880340576172, "margin_dpo/margin_std": 74.05145263671875, "step": 161 }, { "epoch": 0.23788546255506607, "fcm_dpo/beta": 0.011299725621938705, "fcm_dpo/delta": -0.03639139235019684, "fcm_dpo/margin": 55.859214782714844, "fcm_dpo/q_t": 0.37161916494369507, "grad_norm": 23.54488182067871, "learning_rate": 4.7263498971727905e-07, "logits/chosen": -0.5820114612579346, "logits/rejected": -0.5577338933944702, "logps/chosen": -109.61683654785156, "logps/ref_chosen": -62.6105842590332, "logps/ref_rejected": -89.39057922363281, "logps/rejected": -192.25604248046875, "loss": 1.0073, "margin_dpo/margin_mean": 55.859214782714844, "margin_dpo/margin_std": 76.0927734375, "step": 162 }, { "epoch": 0.2393538913362702, "fcm_dpo/beta": 0.01139851938933134, "fcm_dpo/delta": -0.009991994127631187, "fcm_dpo/margin": 53.46131134033203, "fcm_dpo/q_t": 0.374041885137558, "grad_norm": 23.255294799804688, "learning_rate": 4.720482655449212e-07, "logits/chosen": -0.5350059270858765, "logits/rejected": -0.510658323764801, "logps/chosen": -106.6119384765625, "logps/ref_chosen": -55.021629333496094, "logps/ref_rejected": -75.418212890625, "logps/rejected": -180.4698486328125, "loss": 1.0059, "margin_dpo/margin_mean": 53.46131134033203, "margin_dpo/margin_std": 71.67129516601562, "step": 163 }, { "epoch": 0.24082232011747431, "fcm_dpo/beta": 0.011088410392403603, "fcm_dpo/delta": -0.1139398068189621, "fcm_dpo/margin": 63.41717529296875, "fcm_dpo/q_t": 0.3512231409549713, "grad_norm": 25.359941482543945, "learning_rate": 4.714556901942599e-07, "logits/chosen": -0.5142285823822021, "logits/rejected": -0.4906259775161743, "logps/chosen": -104.08917236328125, "logps/ref_chosen": -55.64066696166992, "logps/ref_rejected": -79.66463470458984, "logps/rejected": -191.5303192138672, "loss": 0.9321, "margin_dpo/margin_mean": 63.41717529296875, "margin_dpo/margin_std": 71.19700622558594, "step": 164 }, { "epoch": 0.2422907488986784, "fcm_dpo/beta": 0.011331463232636452, "fcm_dpo/delta": 0.1354885995388031, "fcm_dpo/margin": 41.572532653808594, "fcm_dpo/q_t": 0.39925825595855713, "grad_norm": 34.06428527832031, "learning_rate": 4.708572792802069e-07, "logits/chosen": -0.5556213855743408, "logits/rejected": -0.5192099213600159, "logps/chosen": -114.64573669433594, "logps/ref_chosen": -61.310691833496094, "logps/ref_rejected": -73.67060852050781, "logps/rejected": -168.57818603515625, "loss": 1.094, "margin_dpo/margin_mean": 41.572532653808594, "margin_dpo/margin_std": 65.32736206054688, "step": 165 }, { "epoch": 0.24375917767988253, "fcm_dpo/beta": 0.010963313281536102, "fcm_dpo/delta": -0.2387712448835373, "fcm_dpo/margin": 74.51507568359375, "fcm_dpo/q_t": 0.3456820547580719, "grad_norm": 21.995134353637695, "learning_rate": 4.702530485714461e-07, "logits/chosen": -0.49838924407958984, "logits/rejected": -0.5080593824386597, "logps/chosen": -99.90086364746094, "logps/ref_chosen": -50.98360061645508, "logps/ref_rejected": -98.09512329101562, "logps/rejected": -221.5274658203125, "loss": 0.9183, "margin_dpo/margin_mean": 74.51507568359375, "margin_dpo/margin_std": 95.35562896728516, "step": 166 }, { "epoch": 0.24522760646108663, "fcm_dpo/beta": 0.010487915948033333, "fcm_dpo/delta": -0.2514822483062744, "fcm_dpo/margin": 79.16014862060547, "fcm_dpo/q_t": 0.3255418539047241, "grad_norm": 22.585697174072266, "learning_rate": 4.6964301399001877e-07, "logits/chosen": -0.4470037817955017, "logits/rejected": -0.4476546347141266, "logps/chosen": -99.91099548339844, "logps/ref_chosen": -50.424095153808594, "logps/ref_rejected": -96.03042602539062, "logps/rejected": -224.677490234375, "loss": 0.8589, "margin_dpo/margin_mean": 79.16014862060547, "margin_dpo/margin_std": 78.89952087402344, "step": 167 }, { "epoch": 0.24669603524229075, "fcm_dpo/beta": 0.010302037000656128, "fcm_dpo/delta": -0.02494889497756958, "fcm_dpo/margin": 60.4947395324707, "fcm_dpo/q_t": 0.36779898405075073, "grad_norm": 23.62080192565918, "learning_rate": 4.690271916109034e-07, "logits/chosen": -0.4938944876194, "logits/rejected": -0.4785892963409424, "logps/chosen": -104.25721740722656, "logps/ref_chosen": -49.462825775146484, "logps/ref_rejected": -75.30855560302734, "logps/rejected": -190.59768676757812, "loss": 0.9727, "margin_dpo/margin_mean": 60.49474334716797, "margin_dpo/margin_std": 72.37049865722656, "step": 168 }, { "epoch": 0.24816446402349487, "fcm_dpo/beta": 0.010289417579770088, "fcm_dpo/delta": 0.049492061138153076, "fcm_dpo/margin": 53.629798889160156, "fcm_dpo/q_t": 0.3923034071922302, "grad_norm": 26.301158905029297, "learning_rate": 4.6840559766159235e-07, "logits/chosen": -0.4954657554626465, "logits/rejected": -0.4744875431060791, "logps/chosen": -114.426513671875, "logps/ref_chosen": -59.803443908691406, "logps/ref_rejected": -83.34574890136719, "logps/rejected": -191.59860229492188, "loss": 1.0873, "margin_dpo/margin_mean": 53.629798889160156, "margin_dpo/margin_std": 91.4170913696289, "step": 169 }, { "epoch": 0.24963289280469897, "fcm_dpo/beta": 0.010223541408777237, "fcm_dpo/delta": -0.07429465651512146, "fcm_dpo/margin": 65.30518341064453, "fcm_dpo/q_t": 0.35696014761924744, "grad_norm": 24.374975204467773, "learning_rate": 4.6777824852166437e-07, "logits/chosen": -0.4555085003376007, "logits/rejected": -0.4390765428543091, "logps/chosen": -99.52091979980469, "logps/ref_chosen": -49.471771240234375, "logps/ref_rejected": -75.91734313964844, "logps/rejected": -191.27166748046875, "loss": 0.958, "margin_dpo/margin_mean": 65.30518341064453, "margin_dpo/margin_std": 75.40479278564453, "step": 170 }, { "epoch": 0.2511013215859031, "fcm_dpo/beta": 0.010339688509702682, "fcm_dpo/delta": 0.05246017128229141, "fcm_dpo/margin": 53.23592758178711, "fcm_dpo/q_t": 0.39151298999786377, "grad_norm": 31.557279586791992, "learning_rate": 4.6714516072235273e-07, "logits/chosen": -0.5044229030609131, "logits/rejected": -0.47634661197662354, "logps/chosen": -157.2343292236328, "logps/ref_chosen": -84.49931335449219, "logps/ref_rejected": -109.38209533691406, "logps/rejected": -235.35305786132812, "loss": 1.0862, "margin_dpo/margin_mean": 53.23592758178711, "margin_dpo/margin_std": 91.4041748046875, "step": 171 }, { "epoch": 0.2525697503671072, "fcm_dpo/beta": 0.010385138913989067, "fcm_dpo/delta": 0.024514978751540184, "fcm_dpo/margin": 55.545555114746094, "fcm_dpo/q_t": 0.3802782893180847, "grad_norm": 29.966943740844727, "learning_rate": 4.6650635094610966e-07, "logits/chosen": -0.5027618408203125, "logits/rejected": -0.47274085879325867, "logps/chosen": -136.3382568359375, "logps/ref_chosen": -68.65391540527344, "logps/ref_rejected": -85.43667602539062, "logps/rejected": -208.66656494140625, "loss": 1.0408, "margin_dpo/margin_mean": 55.54555892944336, "margin_dpo/margin_std": 81.37893676757812, "step": 172 }, { "epoch": 0.2540381791483113, "fcm_dpo/beta": 0.010488756000995636, "fcm_dpo/delta": 0.0158163383603096, "fcm_dpo/margin": 55.749046325683594, "fcm_dpo/q_t": 0.3747754693031311, "grad_norm": 29.66188621520996, "learning_rate": 4.6586183602616687e-07, "logits/chosen": -0.5061242580413818, "logits/rejected": -0.4692162275314331, "logps/chosen": -123.67233276367188, "logps/ref_chosen": -63.050880432128906, "logps/ref_rejected": -78.68392181396484, "logps/rejected": -195.054443359375, "loss": 0.9973, "margin_dpo/margin_mean": 55.74905014038086, "margin_dpo/margin_std": 69.4229736328125, "step": 173 }, { "epoch": 0.2555066079295154, "fcm_dpo/beta": 0.01044470351189375, "fcm_dpo/delta": -0.0565309077501297, "fcm_dpo/margin": 62.44271469116211, "fcm_dpo/q_t": 0.3679395318031311, "grad_norm": 31.072416305541992, "learning_rate": 4.652116329460919e-07, "logits/chosen": -0.42711180448532104, "logits/rejected": -0.44361674785614014, "logps/chosen": -113.0875015258789, "logps/ref_chosen": -53.36296844482422, "logps/ref_rejected": -101.91120910644531, "logps/rejected": -224.07846069335938, "loss": 0.9957, "margin_dpo/margin_mean": 62.44270706176758, "margin_dpo/margin_std": 83.03677368164062, "step": 174 }, { "epoch": 0.25697503671071953, "fcm_dpo/beta": 0.009896190837025642, "fcm_dpo/delta": -0.282113254070282, "fcm_dpo/margin": 86.63121032714844, "fcm_dpo/q_t": 0.32039257884025574, "grad_norm": 29.841856002807617, "learning_rate": 4.645557588393406e-07, "logits/chosen": -0.39522379636764526, "logits/rejected": -0.38075965642929077, "logps/chosen": -103.10018920898438, "logps/ref_chosen": -45.417762756347656, "logps/ref_rejected": -89.50579833984375, "logps/rejected": -233.8194580078125, "loss": 0.835, "margin_dpo/margin_mean": 86.63121032714844, "margin_dpo/margin_std": 82.89295196533203, "step": 175 }, { "epoch": 0.25844346549192365, "fcm_dpo/beta": 0.009663419798016548, "fcm_dpo/delta": -0.09544707089662552, "fcm_dpo/margin": 71.29586791992188, "fcm_dpo/q_t": 0.3592815101146698, "grad_norm": 28.798011779785156, "learning_rate": 4.638942309888058e-07, "logits/chosen": -0.39833515882492065, "logits/rejected": -0.41467368602752686, "logps/chosen": -110.1692123413086, "logps/ref_chosen": -50.452842712402344, "logps/ref_rejected": -95.5589599609375, "logps/rejected": -226.57119750976562, "loss": 0.9511, "margin_dpo/margin_mean": 71.29586791992188, "margin_dpo/margin_std": 86.2072525024414, "step": 176 }, { "epoch": 0.2599118942731278, "fcm_dpo/beta": 0.009497301653027534, "fcm_dpo/delta": -0.09194030612707138, "fcm_dpo/margin": 72.19395446777344, "fcm_dpo/q_t": 0.35893410444259644, "grad_norm": 32.332763671875, "learning_rate": 4.6322706682636137e-07, "logits/chosen": -0.4460568428039551, "logits/rejected": -0.4327634572982788, "logps/chosen": -128.06985473632812, "logps/ref_chosen": -61.216468811035156, "logps/ref_rejected": -95.89378356933594, "logps/rejected": -234.94113159179688, "loss": 0.9502, "margin_dpo/margin_mean": 72.19395446777344, "margin_dpo/margin_std": 87.10002136230469, "step": 177 }, { "epoch": 0.26138032305433184, "fcm_dpo/beta": 0.009019111283123493, "fcm_dpo/delta": -0.24799524247646332, "fcm_dpo/margin": 91.52853393554688, "fcm_dpo/q_t": 0.3365201950073242, "grad_norm": 31.862321853637695, "learning_rate": 4.6255428393240354e-07, "logits/chosen": -0.3157181143760681, "logits/rejected": -0.31319284439086914, "logps/chosen": -133.81484985351562, "logps/ref_chosen": -58.26478958129883, "logps/ref_rejected": -105.3653335571289, "logps/rejected": -272.44390869140625, "loss": 0.8982, "margin_dpo/margin_mean": 91.52853393554688, "margin_dpo/margin_std": 108.31314086914062, "step": 178 }, { "epoch": 0.26284875183553597, "fcm_dpo/beta": 0.008883455768227577, "fcm_dpo/delta": -0.01396089792251587, "fcm_dpo/margin": 68.92755889892578, "fcm_dpo/q_t": 0.37395423650741577, "grad_norm": 51.664894104003906, "learning_rate": 4.6187590003538724e-07, "logits/chosen": -0.3555856943130493, "logits/rejected": -0.35472238063812256, "logps/chosen": -136.07827758789062, "logps/ref_chosen": -61.05832290649414, "logps/ref_rejected": -90.52782440185547, "logps/rejected": -234.475341796875, "loss": 1.0389, "margin_dpo/margin_mean": 68.92755889892578, "margin_dpo/margin_std": 100.45736694335938, "step": 179 }, { "epoch": 0.2643171806167401, "fcm_dpo/beta": 0.008733205497264862, "fcm_dpo/delta": -0.13286282122135162, "fcm_dpo/margin": 82.76274108886719, "fcm_dpo/q_t": 0.34625113010406494, "grad_norm": 34.209205627441406, "learning_rate": 4.611919330113591e-07, "logits/chosen": -0.39520972967147827, "logits/rejected": -0.3850780725479126, "logps/chosen": -116.90325927734375, "logps/ref_chosen": -54.34272003173828, "logps/ref_rejected": -98.21183776855469, "logps/rejected": -243.5351104736328, "loss": 0.9237, "margin_dpo/margin_mean": 82.76274108886719, "margin_dpo/margin_std": 91.62629699707031, "step": 180 }, { "epoch": 0.2657856093979442, "fcm_dpo/beta": 0.008904640562832355, "fcm_dpo/delta": 0.14196932315826416, "fcm_dpo/margin": 52.12580871582031, "fcm_dpo/q_t": 0.4002281725406647, "grad_norm": 29.242713928222656, "learning_rate": 4.605024008834863e-07, "logits/chosen": -0.4259217381477356, "logits/rejected": -0.3925449252128601, "logps/chosen": -108.0588607788086, "logps/ref_chosen": -55.000457763671875, "logps/ref_rejected": -61.656166076660156, "logps/rejected": -166.8403778076172, "loss": 1.0943, "margin_dpo/margin_mean": 52.12580871582031, "margin_dpo/margin_std": 82.18958282470703, "step": 181 }, { "epoch": 0.26725403817914833, "fcm_dpo/beta": 0.008686801418662071, "fcm_dpo/delta": -0.16562983393669128, "fcm_dpo/margin": 86.55928802490234, "fcm_dpo/q_t": 0.3413204550743103, "grad_norm": 27.281949996948242, "learning_rate": 4.598073218215817e-07, "logits/chosen": -0.3973362445831299, "logits/rejected": -0.4009121060371399, "logps/chosen": -91.49468231201172, "logps/ref_chosen": -41.107852935791016, "logps/ref_rejected": -89.5215835571289, "logps/rejected": -226.4676971435547, "loss": 0.9226, "margin_dpo/margin_mean": 86.55928802490234, "margin_dpo/margin_std": 97.51657104492188, "step": 182 }, { "epoch": 0.2687224669603524, "fcm_dpo/beta": 0.008756536990404129, "fcm_dpo/delta": 0.15965726971626282, "fcm_dpo/margin": 50.974124908447266, "fcm_dpo/q_t": 0.40270137786865234, "grad_norm": 26.794544219970703, "learning_rate": 4.5910671414162484e-07, "logits/chosen": -0.4323645234107971, "logits/rejected": -0.41012632846832275, "logps/chosen": -132.4599151611328, "logps/ref_chosen": -57.52456283569336, "logps/ref_rejected": -75.97572326660156, "logps/rejected": -201.8852081298828, "loss": 1.0928, "margin_dpo/margin_mean": 50.97412872314453, "margin_dpo/margin_std": 70.7024917602539, "step": 183 }, { "epoch": 0.2701908957415565, "fcm_dpo/beta": 0.009184681810438633, "fcm_dpo/delta": 0.13521449267864227, "fcm_dpo/margin": 51.08043670654297, "fcm_dpo/q_t": 0.40079188346862793, "grad_norm": 24.206710815429688, "learning_rate": 4.5840059630527985e-07, "logits/chosen": -0.4838387966156006, "logits/rejected": -0.4702536165714264, "logps/chosen": -119.1988296508789, "logps/ref_chosen": -58.544952392578125, "logps/ref_rejected": -76.63406372070312, "logps/rejected": -188.36837768554688, "loss": 1.083, "margin_dpo/margin_mean": 51.080440521240234, "margin_dpo/margin_std": 76.5762939453125, "step": 184 }, { "epoch": 0.27165932452276065, "fcm_dpo/beta": 0.00946805626153946, "fcm_dpo/delta": 0.21153177320957184, "fcm_dpo/margin": 41.93297576904297, "fcm_dpo/q_t": 0.41936683654785156, "grad_norm": 26.689727783203125, "learning_rate": 4.5768898691940836e-07, "logits/chosen": -0.4485119581222534, "logits/rejected": -0.4117010831832886, "logps/chosen": -128.1175079345703, "logps/ref_chosen": -62.025848388671875, "logps/ref_rejected": -73.7625961303711, "logps/rejected": -181.7872314453125, "loss": 1.1579, "margin_dpo/margin_mean": 41.93297576904297, "margin_dpo/margin_std": 82.34846496582031, "step": 185 }, { "epoch": 0.27312775330396477, "fcm_dpo/beta": 0.009380245581269264, "fcm_dpo/delta": -0.12999294698238373, "fcm_dpo/margin": 76.79740905761719, "fcm_dpo/q_t": 0.3472128212451935, "grad_norm": 49.98681640625, "learning_rate": 4.5697190473557947e-07, "logits/chosen": -0.4748345613479614, "logits/rejected": -0.43923431634902954, "logps/chosen": -130.9991455078125, "logps/ref_chosen": -69.35346984863281, "logps/ref_rejected": -88.07244873046875, "logps/rejected": -226.51553344726562, "loss": 0.9126, "margin_dpo/margin_mean": 76.79740905761719, "margin_dpo/margin_std": 82.54960632324219, "step": 186 }, { "epoch": 0.2745961820851689, "fcm_dpo/beta": 0.00942191667854786, "fcm_dpo/delta": 0.014476820826530457, "fcm_dpo/margin": 62.11200714111328, "fcm_dpo/q_t": 0.3749355673789978, "grad_norm": 33.28017807006836, "learning_rate": 4.5624936864957555e-07, "logits/chosen": -0.4199049770832062, "logits/rejected": -0.41446995735168457, "logps/chosen": -114.2715072631836, "logps/ref_chosen": -52.7564582824707, "logps/ref_rejected": -81.96910095214844, "logps/rejected": -205.59616088867188, "loss": 1.0036, "margin_dpo/margin_mean": 62.11200714111328, "margin_dpo/margin_std": 77.76611328125, "step": 187 }, { "epoch": 0.27606461086637296, "fcm_dpo/beta": 0.009187942370772362, "fcm_dpo/delta": -0.12695354223251343, "fcm_dpo/margin": 78.13199615478516, "fcm_dpo/q_t": 0.35078778862953186, "grad_norm": 41.620758056640625, "learning_rate": 4.5552139770089454e-07, "logits/chosen": -0.4144795536994934, "logits/rejected": -0.4158819019794464, "logps/chosen": -107.33932495117188, "logps/ref_chosen": -49.415489196777344, "logps/ref_rejected": -89.54043579101562, "logps/rejected": -225.59625244140625, "loss": 0.9298, "margin_dpo/margin_mean": 78.13198852539062, "margin_dpo/margin_std": 89.57806396484375, "step": 188 }, { "epoch": 0.2775330396475771, "fcm_dpo/beta": 0.00912393070757389, "fcm_dpo/delta": -0.008689474314451218, "fcm_dpo/margin": 66.65364074707031, "fcm_dpo/q_t": 0.37750983238220215, "grad_norm": 41.65003967285156, "learning_rate": 4.5478801107224794e-07, "logits/chosen": -0.4112040400505066, "logits/rejected": -0.38720983266830444, "logps/chosen": -123.0237045288086, "logps/ref_chosen": -52.39896011352539, "logps/ref_rejected": -72.16735076904297, "logps/rejected": -209.44573974609375, "loss": 1.028, "margin_dpo/margin_mean": 66.65364074707031, "margin_dpo/margin_std": 99.14986419677734, "step": 189 }, { "epoch": 0.2790014684287812, "fcm_dpo/beta": 0.009091001003980637, "fcm_dpo/delta": -0.09938493371009827, "fcm_dpo/margin": 75.96780395507812, "fcm_dpo/q_t": 0.361076295375824, "grad_norm": 30.051088333129883, "learning_rate": 4.5404922808905543e-07, "logits/chosen": -0.43357422947883606, "logits/rejected": -0.4192691147327423, "logps/chosen": -141.0237579345703, "logps/ref_chosen": -64.68305969238281, "logps/ref_rejected": -102.55052185058594, "logps/rejected": -254.85903930664062, "loss": 0.9991, "margin_dpo/margin_mean": 75.96780395507812, "margin_dpo/margin_std": 101.82373809814453, "step": 190 }, { "epoch": 0.28046989720998533, "fcm_dpo/beta": 0.008413515985012054, "fcm_dpo/delta": -0.3635314106941223, "fcm_dpo/margin": 110.20065307617188, "fcm_dpo/q_t": 0.3166946768760681, "grad_norm": 25.67074203491211, "learning_rate": 4.5330506821893565e-07, "logits/chosen": -0.35954105854034424, "logits/rejected": -0.33676308393478394, "logps/chosen": -138.35675048828125, "logps/ref_chosen": -68.65887451171875, "logps/ref_rejected": -110.1396713256836, "logps/rejected": -290.0382080078125, "loss": 0.8336, "margin_dpo/margin_mean": 110.20065307617188, "margin_dpo/margin_std": 117.54473876953125, "step": 191 }, { "epoch": 0.28193832599118945, "fcm_dpo/beta": 0.008292671293020248, "fcm_dpo/delta": -0.028325140476226807, "fcm_dpo/margin": 75.54448699951172, "fcm_dpo/q_t": 0.37382009625434875, "grad_norm": 34.30679702758789, "learning_rate": 4.5255555107119336e-07, "logits/chosen": -0.37166693806648254, "logits/rejected": -0.37043970823287964, "logps/chosen": -160.6333770751953, "logps/ref_chosen": -69.72691345214844, "logps/ref_rejected": -103.32135009765625, "logps/rejected": -269.77227783203125, "loss": 1.0202, "margin_dpo/margin_mean": 75.54447937011719, "margin_dpo/margin_std": 108.79905700683594, "step": 192 }, { "epoch": 0.2834067547723935, "fcm_dpo/beta": 0.008594648912549019, "fcm_dpo/delta": 0.2289629727602005, "fcm_dpo/margin": 43.941375732421875, "fcm_dpo/q_t": 0.4161950647830963, "grad_norm": 38.5386848449707, "learning_rate": 4.5180069639630236e-07, "logits/chosen": -0.37603700160980225, "logits/rejected": -0.35951176285743713, "logps/chosen": -151.70782470703125, "logps/ref_chosen": -60.19049835205078, "logps/ref_rejected": -76.40755462646484, "logps/rejected": -211.86627197265625, "loss": 1.2035, "margin_dpo/margin_mean": 43.941375732421875, "margin_dpo/margin_std": 97.33943176269531, "step": 193 }, { "epoch": 0.28487518355359764, "fcm_dpo/beta": 0.008597111329436302, "fcm_dpo/delta": -0.016570590436458588, "fcm_dpo/margin": 71.57522583007812, "fcm_dpo/q_t": 0.36463361978530884, "grad_norm": 24.157445907592773, "learning_rate": 4.510405240853854e-07, "logits/chosen": -0.28482919931411743, "logits/rejected": -0.2655167579650879, "logps/chosen": -93.96827697753906, "logps/ref_chosen": -37.84037399291992, "logps/ref_rejected": -60.684783935546875, "logps/rejected": -188.38790893554688, "loss": 0.9556, "margin_dpo/margin_mean": 71.57522583007812, "margin_dpo/margin_std": 75.2847671508789, "step": 194 }, { "epoch": 0.28634361233480177, "fcm_dpo/beta": 0.008543387055397034, "fcm_dpo/delta": -0.04339263588190079, "fcm_dpo/margin": 74.96640014648438, "fcm_dpo/q_t": 0.361470490694046, "grad_norm": 31.8635311126709, "learning_rate": 4.5027505416968985e-07, "logits/chosen": -0.3245890140533447, "logits/rejected": -0.3413041830062866, "logps/chosen": -144.32359313964844, "logps/ref_chosen": -54.891571044921875, "logps/ref_rejected": -96.77095794677734, "logps/rejected": -261.1694030761719, "loss": 0.9508, "margin_dpo/margin_mean": 74.96640014648438, "margin_dpo/margin_std": 82.75733947753906, "step": 195 }, { "epoch": 0.2878120411160059, "fcm_dpo/beta": 0.008355004712939262, "fcm_dpo/delta": -0.11748833954334259, "fcm_dpo/margin": 84.77229309082031, "fcm_dpo/q_t": 0.352541983127594, "grad_norm": 28.523649215698242, "learning_rate": 4.495043068200599e-07, "logits/chosen": -0.3666142225265503, "logits/rejected": -0.34821170568466187, "logps/chosen": -121.3379898071289, "logps/ref_chosen": -53.245243072509766, "logps/ref_rejected": -76.05294799804688, "logps/rejected": -228.91798400878906, "loss": 0.9446, "margin_dpo/margin_mean": 84.77229309082031, "margin_dpo/margin_std": 98.6243896484375, "step": 196 }, { "epoch": 0.28928046989721, "fcm_dpo/beta": 0.00841631181538105, "fcm_dpo/delta": 0.04314948618412018, "fcm_dpo/margin": 66.4241714477539, "fcm_dpo/q_t": 0.3789867162704468, "grad_norm": 36.85988235473633, "learning_rate": 4.4872830234640493e-07, "logits/chosen": -0.34630054235458374, "logits/rejected": -0.33658844232559204, "logps/chosen": -127.1322250366211, "logps/ref_chosen": -60.42033386230469, "logps/ref_rejected": -77.20890808105469, "logps/rejected": -210.344970703125, "loss": 1.0009, "margin_dpo/margin_mean": 66.4241714477539, "margin_dpo/margin_std": 79.97479248046875, "step": 197 }, { "epoch": 0.2907488986784141, "fcm_dpo/beta": 0.008350390940904617, "fcm_dpo/delta": -0.068976491689682, "fcm_dpo/margin": 79.5688705444336, "fcm_dpo/q_t": 0.3639752268791199, "grad_norm": 31.69214630126953, "learning_rate": 4.479470611971645e-07, "logits/chosen": -0.36272114515304565, "logits/rejected": -0.36416107416152954, "logps/chosen": -129.2322998046875, "logps/ref_chosen": -55.03618621826172, "logps/ref_rejected": -97.24325561523438, "logps/rejected": -251.00823974609375, "loss": 0.9641, "margin_dpo/margin_mean": 79.5688705444336, "margin_dpo/margin_std": 100.84001922607422, "step": 198 }, { "epoch": 0.2922173274596182, "fcm_dpo/beta": 0.008206663653254509, "fcm_dpo/delta": -0.027660571038722992, "fcm_dpo/margin": 76.12931060791016, "fcm_dpo/q_t": 0.3673062324523926, "grad_norm": 35.12301254272461, "learning_rate": 4.471606039587695e-07, "logits/chosen": -0.3977452516555786, "logits/rejected": -0.38206660747528076, "logps/chosen": -128.30673217773438, "logps/ref_chosen": -56.828826904296875, "logps/ref_rejected": -84.64820861816406, "logps/rejected": -232.25543212890625, "loss": 0.9921, "margin_dpo/margin_mean": 76.12931060791016, "margin_dpo/margin_std": 95.28636169433594, "step": 199 }, { "epoch": 0.2936857562408223, "fcm_dpo/beta": 0.008154574781656265, "fcm_dpo/delta": -0.05656225234270096, "fcm_dpo/margin": 80.00682830810547, "fcm_dpo/q_t": 0.3651021122932434, "grad_norm": 30.504737854003906, "learning_rate": 4.4636895135509966e-07, "logits/chosen": -0.3527507185935974, "logits/rejected": -0.3269976079463959, "logps/chosen": -120.8394775390625, "logps/ref_chosen": -53.06706237792969, "logps/ref_rejected": -80.60843658447266, "logps/rejected": -228.38766479492188, "loss": 0.9914, "margin_dpo/margin_mean": 80.00682830810547, "margin_dpo/margin_std": 106.2452621459961, "step": 200 }, { "epoch": 0.2936857562408223, "eval_fcm_dpo/beta": 0.00816048588603735, "eval_logits/chosen": -0.4719357490539551, "eval_logits/rejected": -0.45578357577323914, "eval_logps/chosen": -172.25338745117188, "eval_logps/ref_chosen": -79.05104064941406, "eval_logps/ref_rejected": -86.79793548583984, "eval_logps/rejected": -225.0161590576172, "eval_loss": 0.6008790135383606, "eval_margin_dpo/margin_mean": 45.01588439941406, "eval_margin_dpo/margin_std": 98.482421875, "eval_runtime": 39.3122, "eval_samples_per_second": 59.498, "eval_steps_per_second": 1.882, "step": 200 }, { "epoch": 0.29515418502202645, "fcm_dpo/beta": 0.008100075647234917, "fcm_dpo/delta": -0.0565091148018837, "fcm_dpo/margin": 80.59832000732422, "fcm_dpo/q_t": 0.3638731837272644, "grad_norm": 30.036048889160156, "learning_rate": 4.455721242469372e-07, "logits/chosen": -0.4626866281032562, "logits/rejected": -0.4563639163970947, "logps/chosen": -146.43814086914062, "logps/ref_chosen": -75.4022216796875, "logps/ref_rejected": -114.80821990966797, "logps/rejected": -266.44244384765625, "loss": 0.9815, "margin_dpo/margin_mean": 80.59831237792969, "margin_dpo/margin_std": 103.2762222290039, "step": 201 }, { "epoch": 0.2966226138032305, "fcm_dpo/beta": 0.00825195387005806, "fcm_dpo/delta": 0.13680000603199005, "fcm_dpo/margin": 56.866981506347656, "fcm_dpo/q_t": 0.40256237983703613, "grad_norm": 26.29439353942871, "learning_rate": 4.4477014363141755e-07, "logits/chosen": -0.3789522051811218, "logits/rejected": -0.3867080807685852, "logps/chosen": -124.50201416015625, "logps/ref_chosen": -50.101318359375, "logps/ref_rejected": -86.98503112792969, "logps/rejected": -218.25270080566406, "loss": 1.1171, "margin_dpo/margin_mean": 56.866981506347656, "margin_dpo/margin_std": 99.0474853515625, "step": 202 }, { "epoch": 0.29809104258443464, "fcm_dpo/beta": 0.008316997438669205, "fcm_dpo/delta": 0.0254436694085598, "fcm_dpo/margin": 69.2591781616211, "fcm_dpo/q_t": 0.3762577772140503, "grad_norm": 31.99873924255371, "learning_rate": 4.439630306414758e-07, "logits/chosen": -0.4376029968261719, "logits/rejected": -0.4240657389163971, "logps/chosen": -134.08056640625, "logps/ref_chosen": -60.60969543457031, "logps/ref_rejected": -85.89596557617188, "logps/rejected": -228.62603759765625, "loss": 0.9972, "margin_dpo/margin_mean": 69.2591781616211, "margin_dpo/margin_std": 84.98655700683594, "step": 203 }, { "epoch": 0.29955947136563876, "fcm_dpo/beta": 0.00838047917932272, "fcm_dpo/delta": 0.04098369926214218, "fcm_dpo/margin": 66.98109436035156, "fcm_dpo/q_t": 0.3864881992340088, "grad_norm": 29.694320678710938, "learning_rate": 4.431508065452897e-07, "logits/chosen": -0.47066575288772583, "logits/rejected": -0.4281098246574402, "logps/chosen": -162.51052856445312, "logps/ref_chosen": -80.16496276855469, "logps/ref_rejected": -87.69590759277344, "logps/rejected": -237.0225830078125, "loss": 1.0491, "margin_dpo/margin_mean": 66.98109436035156, "margin_dpo/margin_std": 100.66291046142578, "step": 204 }, { "epoch": 0.3010279001468429, "fcm_dpo/beta": 0.008169597014784813, "fcm_dpo/delta": -0.11707717180252075, "fcm_dpo/margin": 86.33795166015625, "fcm_dpo/q_t": 0.350330114364624, "grad_norm": 24.040218353271484, "learning_rate": 4.4233349274571974e-07, "logits/chosen": -0.38784509897232056, "logits/rejected": -0.35509926080703735, "logps/chosen": -142.8284454345703, "logps/ref_chosen": -59.384735107421875, "logps/ref_rejected": -85.12505340576172, "logps/rejected": -254.90672302246094, "loss": 0.9583, "margin_dpo/margin_mean": 86.33795166015625, "margin_dpo/margin_std": 101.34950256347656, "step": 205 }, { "epoch": 0.302496328928047, "fcm_dpo/beta": 0.007971297949552536, "fcm_dpo/delta": -0.16260316967964172, "fcm_dpo/margin": 93.97215270996094, "fcm_dpo/q_t": 0.33909979462623596, "grad_norm": 34.61493682861328, "learning_rate": 4.415111107797445e-07, "logits/chosen": -0.3233247995376587, "logits/rejected": -0.32719945907592773, "logps/chosen": -125.01007843017578, "logps/ref_chosen": -46.964500427246094, "logps/ref_rejected": -98.9534912109375, "logps/rejected": -270.9712219238281, "loss": 0.9038, "margin_dpo/margin_mean": 93.97215270996094, "margin_dpo/margin_std": 98.51936340332031, "step": 206 }, { "epoch": 0.3039647577092511, "fcm_dpo/beta": 0.00767237413674593, "fcm_dpo/delta": -0.24557989835739136, "fcm_dpo/margin": 107.59861755371094, "fcm_dpo/q_t": 0.33342668414115906, "grad_norm": 28.77750587463379, "learning_rate": 4.4068368231789365e-07, "logits/chosen": -0.40268674492836, "logits/rejected": -0.37191277742385864, "logps/chosen": -121.83184814453125, "logps/ref_chosen": -56.05625915527344, "logps/ref_rejected": -84.44779968261719, "logps/rejected": -257.822021484375, "loss": 0.8758, "margin_dpo/margin_mean": 107.59861755371094, "margin_dpo/margin_std": 117.77052307128906, "step": 207 }, { "epoch": 0.3054331864904552, "fcm_dpo/beta": 0.007520717103034258, "fcm_dpo/delta": -0.02344403602182865, "fcm_dpo/margin": 82.68033599853516, "fcm_dpo/q_t": 0.36858728528022766, "grad_norm": 29.981359481811523, "learning_rate": 4.398512291636768e-07, "logits/chosen": -0.42680084705352783, "logits/rejected": -0.4093056321144104, "logps/chosen": -167.66326904296875, "logps/ref_chosen": -67.06761169433594, "logps/ref_rejected": -94.28689575195312, "logps/rejected": -277.5628662109375, "loss": 1.007, "margin_dpo/margin_mean": 82.68034362792969, "margin_dpo/margin_std": 110.93801879882812, "step": 208 }, { "epoch": 0.3069016152716593, "fcm_dpo/beta": 0.0075776162557303905, "fcm_dpo/delta": 0.04492693394422531, "fcm_dpo/margin": 73.58404541015625, "fcm_dpo/q_t": 0.3805572986602783, "grad_norm": 32.00458908081055, "learning_rate": 4.3901377325300857e-07, "logits/chosen": -0.33511584997177124, "logits/rejected": -0.3224318027496338, "logps/chosen": -138.41427612304688, "logps/ref_chosen": -56.18169403076172, "logps/ref_rejected": -80.94152069091797, "logps/rejected": -236.75814819335938, "loss": 1.0361, "margin_dpo/margin_mean": 73.58403778076172, "margin_dpo/margin_std": 99.81734466552734, "step": 209 }, { "epoch": 0.30837004405286345, "fcm_dpo/beta": 0.007613741792738438, "fcm_dpo/delta": -0.010615115985274315, "fcm_dpo/margin": 80.08528137207031, "fcm_dpo/q_t": 0.3702564239501953, "grad_norm": 22.40015411376953, "learning_rate": 4.381713366536311e-07, "logits/chosen": -0.3724694550037384, "logits/rejected": -0.36007583141326904, "logps/chosen": -120.55526733398438, "logps/ref_chosen": -46.371822357177734, "logps/ref_rejected": -76.68162536621094, "logps/rejected": -230.95034790039062, "loss": 0.984, "margin_dpo/margin_mean": 80.08528137207031, "margin_dpo/margin_std": 96.79129028320312, "step": 210 }, { "epoch": 0.30983847283406757, "fcm_dpo/beta": 0.0076542324386537075, "fcm_dpo/delta": 0.0667615756392479, "fcm_dpo/margin": 70.13720703125, "fcm_dpo/q_t": 0.38842254877090454, "grad_norm": 34.49944305419922, "learning_rate": 4.373239415645323e-07, "logits/chosen": -0.4161061644554138, "logits/rejected": -0.37069594860076904, "logps/chosen": -185.52003479003906, "logps/ref_chosen": -78.93235778808594, "logps/ref_rejected": -86.82098388671875, "logps/rejected": -263.5458984375, "loss": 1.0509, "margin_dpo/margin_mean": 70.13721466064453, "margin_dpo/margin_std": 100.4876937866211, "step": 211 }, { "epoch": 0.31130690161527164, "fcm_dpo/beta": 0.007427205331623554, "fcm_dpo/delta": -0.15895992517471313, "fcm_dpo/margin": 100.10624694824219, "fcm_dpo/q_t": 0.3454575836658478, "grad_norm": 28.596296310424805, "learning_rate": 4.3647161031536086e-07, "logits/chosen": -0.3710683584213257, "logits/rejected": -0.35650646686553955, "logps/chosen": -147.09378051757812, "logps/ref_chosen": -58.19701385498047, "logps/ref_rejected": -103.05785369873047, "logps/rejected": -292.06085205078125, "loss": 0.9345, "margin_dpo/margin_mean": 100.10624694824219, "margin_dpo/margin_std": 112.30523681640625, "step": 212 }, { "epoch": 0.31277533039647576, "fcm_dpo/beta": 0.007264791056513786, "fcm_dpo/delta": -0.1239488273859024, "fcm_dpo/margin": 98.2468490600586, "fcm_dpo/q_t": 0.34874916076660156, "grad_norm": 25.72871208190918, "learning_rate": 4.3561436536583774e-07, "logits/chosen": -0.41955453157424927, "logits/rejected": -0.3891031742095947, "logps/chosen": -149.52923583984375, "logps/ref_chosen": -67.51271057128906, "logps/ref_rejected": -93.91471862792969, "logps/rejected": -274.1781005859375, "loss": 0.9295, "margin_dpo/margin_mean": 98.24685668945312, "margin_dpo/margin_std": 108.88845825195312, "step": 213 }, { "epoch": 0.3142437591776799, "fcm_dpo/beta": 0.007254892028868198, "fcm_dpo/delta": -0.009785597212612629, "fcm_dpo/margin": 83.96929168701172, "fcm_dpo/q_t": 0.37200796604156494, "grad_norm": 26.340700149536133, "learning_rate": 4.3475222930516473e-07, "logits/chosen": -0.3503088653087616, "logits/rejected": -0.35779234766960144, "logps/chosen": -118.85914611816406, "logps/ref_chosen": -41.604888916015625, "logps/ref_rejected": -77.51741027832031, "logps/rejected": -238.740966796875, "loss": 0.9962, "margin_dpo/margin_mean": 83.96929931640625, "margin_dpo/margin_std": 107.99652099609375, "step": 214 }, { "epoch": 0.315712187958884, "fcm_dpo/beta": 0.007164125796407461, "fcm_dpo/delta": -0.04974304512143135, "fcm_dpo/margin": 90.17284393310547, "fcm_dpo/q_t": 0.3607136309146881, "grad_norm": 27.573640823364258, "learning_rate": 4.3388522485142885e-07, "logits/chosen": -0.3955865502357483, "logits/rejected": -0.3844960331916809, "logps/chosen": -139.45066833496094, "logps/ref_chosen": -53.279266357421875, "logps/ref_rejected": -89.96464538574219, "logps/rejected": -266.30889892578125, "loss": 0.9501, "margin_dpo/margin_mean": 90.1728515625, "margin_dpo/margin_std": 99.1728515625, "step": 215 }, { "epoch": 0.31718061674008813, "fcm_dpo/beta": 0.007151792757213116, "fcm_dpo/delta": -0.04989144578576088, "fcm_dpo/margin": 90.40126037597656, "fcm_dpo/q_t": 0.36354175209999084, "grad_norm": 33.67525863647461, "learning_rate": 4.330133748510036e-07, "logits/chosen": -0.3702380061149597, "logits/rejected": -0.3506859540939331, "logps/chosen": -137.24215698242188, "logps/ref_chosen": -48.887794494628906, "logps/ref_rejected": -77.19892883300781, "logps/rejected": -255.95455932617188, "loss": 0.9844, "margin_dpo/margin_mean": 90.40126037597656, "margin_dpo/margin_std": 113.781982421875, "step": 216 }, { "epoch": 0.3186490455212922, "fcm_dpo/beta": 0.006948284804821014, "fcm_dpo/delta": -0.1394684761762619, "fcm_dpo/margin": 104.92652893066406, "fcm_dpo/q_t": 0.34340259432792664, "grad_norm": 25.96721649169922, "learning_rate": 4.3213670227794757e-07, "logits/chosen": -0.3827437162399292, "logits/rejected": -0.3740556240081787, "logps/chosen": -134.2442626953125, "logps/ref_chosen": -49.845306396484375, "logps/ref_rejected": -100.07832336425781, "logps/rejected": -289.40380859375, "loss": 0.897, "margin_dpo/margin_mean": 104.92652893066406, "margin_dpo/margin_std": 105.56159210205078, "step": 217 }, { "epoch": 0.3201174743024963, "fcm_dpo/beta": 0.0069186268374323845, "fcm_dpo/delta": 0.01962290331721306, "fcm_dpo/margin": 84.04899597167969, "fcm_dpo/q_t": 0.37689316272735596, "grad_norm": 29.522064208984375, "learning_rate": 4.3125523023339815e-07, "logits/chosen": -0.4043159484863281, "logits/rejected": -0.3937210440635681, "logps/chosen": -144.95455932617188, "logps/ref_chosen": -58.576683044433594, "logps/ref_rejected": -87.84639739990234, "logps/rejected": -258.2732849121094, "loss": 1.008, "margin_dpo/margin_mean": 84.04900360107422, "margin_dpo/margin_std": 107.98806762695312, "step": 218 }, { "epoch": 0.32158590308370044, "fcm_dpo/beta": 0.007075803354382515, "fcm_dpo/delta": 0.09832102805376053, "fcm_dpo/margin": 71.45269775390625, "fcm_dpo/q_t": 0.3926897644996643, "grad_norm": 25.61137580871582, "learning_rate": 4.303689819449636e-07, "logits/chosen": -0.4170767664909363, "logits/rejected": -0.4031888246536255, "logps/chosen": -152.66976928710938, "logps/ref_chosen": -61.083858489990234, "logps/ref_rejected": -85.83042907714844, "logps/rejected": -248.86904907226562, "loss": 1.0769, "margin_dpo/margin_mean": 71.45269775390625, "margin_dpo/margin_std": 110.11042785644531, "step": 219 }, { "epoch": 0.32305433186490456, "fcm_dpo/beta": 0.007226555608212948, "fcm_dpo/delta": 0.1684371680021286, "fcm_dpo/margin": 60.78836441040039, "fcm_dpo/q_t": 0.40072399377822876, "grad_norm": 25.96778106689453, "learning_rate": 4.2947798076611047e-07, "logits/chosen": -0.40701720118522644, "logits/rejected": -0.3823140263557434, "logps/chosen": -173.3245849609375, "logps/ref_chosen": -70.03128051757812, "logps/ref_rejected": -87.68551635742188, "logps/rejected": -251.76718139648438, "loss": 1.0723, "margin_dpo/margin_mean": 60.788368225097656, "margin_dpo/margin_std": 78.86666870117188, "step": 220 }, { "epoch": 0.3245227606461087, "fcm_dpo/beta": 0.007043336518108845, "fcm_dpo/delta": -0.2806098461151123, "fcm_dpo/margin": 121.70394897460938, "fcm_dpo/q_t": 0.3166281580924988, "grad_norm": 29.857473373413086, "learning_rate": 4.285822501755485e-07, "logits/chosen": -0.36656078696250916, "logits/rejected": -0.3734094500541687, "logps/chosen": -148.55584716796875, "logps/ref_chosen": -52.15470886230469, "logps/ref_rejected": -106.46768188476562, "logps/rejected": -324.57275390625, "loss": 0.8286, "margin_dpo/margin_mean": 121.70394897460938, "margin_dpo/margin_std": 108.79747009277344, "step": 221 }, { "epoch": 0.32599118942731276, "fcm_dpo/beta": 0.00688206497579813, "fcm_dpo/delta": -0.058866240084171295, "fcm_dpo/margin": 95.16747283935547, "fcm_dpo/q_t": 0.3594551086425781, "grad_norm": 27.359058380126953, "learning_rate": 4.276818137766118e-07, "logits/chosen": -0.4119839668273926, "logits/rejected": -0.40599650144577026, "logps/chosen": -164.09307861328125, "logps/ref_chosen": -60.971099853515625, "logps/ref_rejected": -100.00115203857422, "logps/rejected": -298.29058837890625, "loss": 0.9574, "margin_dpo/margin_mean": 95.16747283935547, "margin_dpo/margin_std": 109.49444580078125, "step": 222 }, { "epoch": 0.3274596182085169, "fcm_dpo/beta": 0.006874423939734697, "fcm_dpo/delta": 0.045850180089473724, "fcm_dpo/margin": 80.98518371582031, "fcm_dpo/q_t": 0.38237255811691284, "grad_norm": 31.334300994873047, "learning_rate": 4.2677669529663686e-07, "logits/chosen": -0.3320853114128113, "logits/rejected": -0.31920111179351807, "logps/chosen": -163.82130432128906, "logps/ref_chosen": -52.64057540893555, "logps/ref_rejected": -82.82502746582031, "logps/rejected": -274.9909362792969, "loss": 1.0635, "margin_dpo/margin_mean": 80.98518371582031, "margin_dpo/margin_std": 123.70716857910156, "step": 223 }, { "epoch": 0.328928046989721, "fcm_dpo/beta": 0.006771251559257507, "fcm_dpo/delta": -0.05909021943807602, "fcm_dpo/margin": 96.40702056884766, "fcm_dpo/q_t": 0.3656306862831116, "grad_norm": 32.546329498291016, "learning_rate": 4.2586691858633747e-07, "logits/chosen": -0.36080002784729004, "logits/rejected": -0.3402705192565918, "logps/chosen": -143.6527862548828, "logps/ref_chosen": -48.59541320800781, "logps/ref_rejected": -77.11648559570312, "logps/rejected": -268.58087158203125, "loss": 0.9886, "margin_dpo/margin_mean": 96.40702056884766, "margin_dpo/margin_std": 124.457275390625, "step": 224 }, { "epoch": 0.3303964757709251, "fcm_dpo/beta": 0.006611201912164688, "fcm_dpo/delta": -0.19814401865005493, "fcm_dpo/margin": 118.33515930175781, "fcm_dpo/q_t": 0.3383178114891052, "grad_norm": 26.60202407836914, "learning_rate": 4.249525076191759e-07, "logits/chosen": -0.35223451256752014, "logits/rejected": -0.33783799409866333, "logps/chosen": -169.36512756347656, "logps/ref_chosen": -58.000465393066406, "logps/ref_rejected": -99.90291595458984, "logps/rejected": -329.60272216796875, "loss": 0.9074, "margin_dpo/margin_mean": 118.33515930175781, "margin_dpo/margin_std": 133.42843627929688, "step": 225 }, { "epoch": 0.33186490455212925, "fcm_dpo/beta": 0.006490599364042282, "fcm_dpo/delta": 0.004586532711982727, "fcm_dpo/margin": 91.46955871582031, "fcm_dpo/q_t": 0.37888455390930176, "grad_norm": 27.73929214477539, "learning_rate": 4.2403348649073167e-07, "logits/chosen": -0.389433890581131, "logits/rejected": -0.3457120954990387, "logps/chosen": -155.97816467285156, "logps/ref_chosen": -58.898799896240234, "logps/ref_rejected": -78.68775939941406, "logps/rejected": -267.2366943359375, "loss": 1.0272, "margin_dpo/margin_mean": 91.46955871582031, "margin_dpo/margin_std": 127.21261596679688, "step": 226 }, { "epoch": 0.3333333333333333, "fcm_dpo/beta": 0.00636161956936121, "fcm_dpo/delta": -0.16426458954811096, "fcm_dpo/margin": 117.93624877929688, "fcm_dpo/q_t": 0.34007027745246887, "grad_norm": 27.874391555786133, "learning_rate": 4.2310987941806615e-07, "logits/chosen": -0.37095752358436584, "logits/rejected": -0.35500913858413696, "logps/chosen": -170.41358947753906, "logps/ref_chosen": -59.072181701660156, "logps/ref_rejected": -99.41236877441406, "logps/rejected": -328.6900329589844, "loss": 0.9131, "margin_dpo/margin_mean": 117.93626403808594, "margin_dpo/margin_std": 129.8984832763672, "step": 227 }, { "epoch": 0.33480176211453744, "fcm_dpo/beta": 0.006433088798075914, "fcm_dpo/delta": 0.07818908244371414, "fcm_dpo/margin": 81.71975708007812, "fcm_dpo/q_t": 0.38696879148483276, "grad_norm": 28.652111053466797, "learning_rate": 4.2218171073908463e-07, "logits/chosen": -0.3662306070327759, "logits/rejected": -0.3457154631614685, "logps/chosen": -179.63031005859375, "logps/ref_chosen": -65.89128875732422, "logps/ref_rejected": -91.04875183105469, "logps/rejected": -286.50750732421875, "loss": 1.0584, "margin_dpo/margin_mean": 81.71975708007812, "margin_dpo/margin_std": 119.07015228271484, "step": 228 }, { "epoch": 0.33627019089574156, "fcm_dpo/beta": 0.006459444761276245, "fcm_dpo/delta": 0.005232607014477253, "fcm_dpo/margin": 92.12015533447266, "fcm_dpo/q_t": 0.3724709153175354, "grad_norm": 33.15553665161133, "learning_rate": 4.212490049118951e-07, "logits/chosen": -0.44420498609542847, "logits/rejected": -0.412958562374115, "logps/chosen": -181.48049926757812, "logps/ref_chosen": -70.70637512207031, "logps/ref_rejected": -84.52741241455078, "logps/rejected": -287.42169189453125, "loss": 1.0091, "margin_dpo/margin_mean": 92.12016296386719, "margin_dpo/margin_std": 121.14804077148438, "step": 229 }, { "epoch": 0.3377386196769457, "fcm_dpo/beta": 0.006237386725842953, "fcm_dpo/delta": -0.20489397644996643, "fcm_dpo/margin": 126.29830169677734, "fcm_dpo/q_t": 0.33031266927719116, "grad_norm": 31.86959457397461, "learning_rate": 4.203117865141635e-07, "logits/chosen": -0.32878124713897705, "logits/rejected": -0.3298182487487793, "logps/chosen": -130.9337158203125, "logps/ref_chosen": -39.282005310058594, "logps/ref_rejected": -85.62191009521484, "logps/rejected": -303.5719299316406, "loss": 0.8704, "margin_dpo/margin_mean": 126.29830932617188, "margin_dpo/margin_std": 123.00277709960938, "step": 230 }, { "epoch": 0.3392070484581498, "fcm_dpo/beta": 0.0062300898134708405, "fcm_dpo/delta": 0.04776204377412796, "fcm_dpo/margin": 89.06995391845703, "fcm_dpo/q_t": 0.3806319832801819, "grad_norm": 26.72955894470215, "learning_rate": 4.1937008024246625e-07, "logits/chosen": -0.36490845680236816, "logits/rejected": -0.33270594477653503, "logps/chosen": -161.59432983398438, "logps/ref_chosen": -63.27644348144531, "logps/ref_rejected": -74.1239013671875, "logps/rejected": -261.5117492675781, "loss": 1.0029, "margin_dpo/margin_mean": 89.06995391845703, "margin_dpo/margin_std": 107.24581909179688, "step": 231 }, { "epoch": 0.3406754772393539, "fcm_dpo/beta": 0.006363061256706715, "fcm_dpo/delta": 0.136710524559021, "fcm_dpo/margin": 73.8527603149414, "fcm_dpo/q_t": 0.40357333421707153, "grad_norm": 31.16952896118164, "learning_rate": 4.1842391091163933e-07, "logits/chosen": -0.3764330744743347, "logits/rejected": -0.3467271029949188, "logps/chosen": -187.44082641601562, "logps/ref_chosen": -70.74876403808594, "logps/ref_rejected": -83.97706604003906, "logps/rejected": -274.52191162109375, "loss": 1.092, "margin_dpo/margin_mean": 73.8527603149414, "margin_dpo/margin_std": 118.7332763671875, "step": 232 }, { "epoch": 0.342143906020558, "fcm_dpo/beta": 0.006316785700619221, "fcm_dpo/delta": -0.11919762194156647, "fcm_dpo/margin": 112.51909637451172, "fcm_dpo/q_t": 0.3590048551559448, "grad_norm": 28.269271850585938, "learning_rate": 4.174733034541245e-07, "logits/chosen": -0.3798902630805969, "logits/rejected": -0.3772153854370117, "logps/chosen": -157.00439453125, "logps/ref_chosen": -54.8829345703125, "logps/ref_rejected": -107.4800796508789, "logps/rejected": -322.12060546875, "loss": 0.9786, "margin_dpo/margin_mean": 112.51909637451172, "margin_dpo/margin_std": 150.15170288085938, "step": 233 }, { "epoch": 0.3436123348017621, "fcm_dpo/beta": 0.006136967800557613, "fcm_dpo/delta": -0.0946032702922821, "fcm_dpo/margin": 111.787109375, "fcm_dpo/q_t": 0.3509105145931244, "grad_norm": 38.759918212890625, "learning_rate": 4.165182829193126e-07, "logits/chosen": -0.3641296625137329, "logits/rejected": -0.39079636335372925, "logps/chosen": -142.52503967285156, "logps/ref_chosen": -44.094520568847656, "logps/ref_rejected": -100.00663757324219, "logps/rejected": -310.2242431640625, "loss": 0.9207, "margin_dpo/margin_mean": 111.787109375, "margin_dpo/margin_std": 113.17666625976562, "step": 234 }, { "epoch": 0.34508076358296624, "fcm_dpo/beta": 0.0063125560991466045, "fcm_dpo/delta": 0.14529383182525635, "fcm_dpo/margin": 72.99315643310547, "fcm_dpo/q_t": 0.40060219168663025, "grad_norm": 33.466941833496094, "learning_rate": 4.1555887447288255e-07, "logits/chosen": -0.4195903241634369, "logits/rejected": -0.39497965574264526, "logps/chosen": -181.33181762695312, "logps/ref_chosen": -62.237911224365234, "logps/ref_rejected": -90.39506530761719, "logps/rejected": -282.48211669921875, "loss": 1.0955, "margin_dpo/margin_mean": 72.9931640625, "margin_dpo/margin_std": 114.59500122070312, "step": 235 }, { "epoch": 0.3465491923641703, "fcm_dpo/beta": 0.006248786114156246, "fcm_dpo/delta": -0.09669020771980286, "fcm_dpo/margin": 110.42259216308594, "fcm_dpo/q_t": 0.3470776677131653, "grad_norm": 50.5585823059082, "learning_rate": 4.1459510339613946e-07, "logits/chosen": -0.36349114775657654, "logits/rejected": -0.3634676933288574, "logps/chosen": -136.66470336914062, "logps/ref_chosen": -49.34136199951172, "logps/ref_rejected": -103.51162719726562, "logps/rejected": -301.257568359375, "loss": 0.8949, "margin_dpo/margin_mean": 110.42259216308594, "margin_dpo/margin_std": 99.5032958984375, "step": 236 }, { "epoch": 0.34801762114537443, "fcm_dpo/beta": 0.006244904361665249, "fcm_dpo/delta": 0.023870917037129402, "fcm_dpo/margin": 92.47233581542969, "fcm_dpo/q_t": 0.37653428316116333, "grad_norm": 30.452253341674805, "learning_rate": 4.136269950853473e-07, "logits/chosen": -0.35491400957107544, "logits/rejected": -0.34882035851478577, "logps/chosen": -170.31382751464844, "logps/ref_chosen": -54.168121337890625, "logps/ref_rejected": -94.78036499023438, "logps/rejected": -303.3984069824219, "loss": 1.0111, "margin_dpo/margin_mean": 92.47232818603516, "margin_dpo/margin_std": 120.4630126953125, "step": 237 }, { "epoch": 0.34948604992657856, "fcm_dpo/beta": 0.006248504854738712, "fcm_dpo/delta": 0.04543028026819229, "fcm_dpo/margin": 89.06282043457031, "fcm_dpo/q_t": 0.3825730085372925, "grad_norm": 31.341205596923828, "learning_rate": 4.126545750510605e-07, "logits/chosen": -0.3699679374694824, "logits/rejected": -0.3880308270454407, "logps/chosen": -163.38356018066406, "logps/ref_chosen": -53.973121643066406, "logps/ref_rejected": -89.41795349121094, "logps/rejected": -287.8912048339844, "loss": 1.0143, "margin_dpo/margin_mean": 89.06282043457031, "margin_dpo/margin_std": 114.02665710449219, "step": 238 }, { "epoch": 0.3509544787077827, "fcm_dpo/beta": 0.0062086135149002075, "fcm_dpo/delta": -0.038354501128196716, "fcm_dpo/margin": 102.07293701171875, "fcm_dpo/q_t": 0.36239856481552124, "grad_norm": 28.831066131591797, "learning_rate": 4.116778689174514e-07, "logits/chosen": -0.34534239768981934, "logits/rejected": -0.32214316725730896, "logps/chosen": -178.08087158203125, "logps/ref_chosen": -58.09782409667969, "logps/ref_rejected": -93.59294128417969, "logps/rejected": -315.64892578125, "loss": 0.9607, "margin_dpo/margin_mean": 102.07293701171875, "margin_dpo/margin_std": 110.16291809082031, "step": 239 }, { "epoch": 0.3524229074889868, "fcm_dpo/beta": 0.006256358698010445, "fcm_dpo/delta": 0.03319564089179039, "fcm_dpo/margin": 90.84452056884766, "fcm_dpo/q_t": 0.37809598445892334, "grad_norm": 34.58857727050781, "learning_rate": 4.106969024216348e-07, "logits/chosen": -0.3797963261604309, "logits/rejected": -0.3545226454734802, "logps/chosen": -188.4025421142578, "logps/ref_chosen": -60.6144905090332, "logps/ref_rejected": -74.1185302734375, "logps/rejected": -292.7510986328125, "loss": 1.0351, "margin_dpo/margin_mean": 90.84451293945312, "margin_dpo/margin_std": 124.30812072753906, "step": 240 }, { "epoch": 0.35389133627019087, "fcm_dpo/beta": 0.006106827408075333, "fcm_dpo/delta": -0.2318933755159378, "fcm_dpo/margin": 133.21487426757812, "fcm_dpo/q_t": 0.33365797996520996, "grad_norm": 27.41342544555664, "learning_rate": 4.097117014129903e-07, "logits/chosen": -0.42394572496414185, "logits/rejected": -0.39460113644599915, "logps/chosen": -171.47601318359375, "logps/ref_chosen": -66.091064453125, "logps/ref_rejected": -88.06088256835938, "logps/rejected": -326.66070556640625, "loss": 0.8839, "margin_dpo/margin_mean": 133.21487426757812, "margin_dpo/margin_std": 142.80030822753906, "step": 241 }, { "epoch": 0.355359765051395, "fcm_dpo/beta": 0.006066991947591305, "fcm_dpo/delta": 0.01960517093539238, "fcm_dpo/margin": 95.71859741210938, "fcm_dpo/q_t": 0.3771246075630188, "grad_norm": 29.23126983642578, "learning_rate": 4.087222918524807e-07, "logits/chosen": -0.3212859630584717, "logits/rejected": -0.2920798659324646, "logps/chosen": -186.41012573242188, "logps/ref_chosen": -67.86392974853516, "logps/ref_rejected": -83.36033630371094, "logps/rejected": -297.6251220703125, "loss": 1.0017, "margin_dpo/margin_mean": 95.71859741210938, "margin_dpo/margin_std": 121.07659912109375, "step": 242 }, { "epoch": 0.3568281938325991, "fcm_dpo/beta": 0.005917678587138653, "fcm_dpo/delta": -0.11988495290279388, "fcm_dpo/margin": 120.15345764160156, "fcm_dpo/q_t": 0.34703272581100464, "grad_norm": 27.846588134765625, "learning_rate": 4.07728699811968e-07, "logits/chosen": -0.36647382378578186, "logits/rejected": -0.3263128399848938, "logps/chosen": -183.94918823242188, "logps/ref_chosen": -63.0842399597168, "logps/ref_rejected": -76.33563232421875, "logps/rejected": -317.35406494140625, "loss": 0.9108, "margin_dpo/margin_mean": 120.1534652709961, "margin_dpo/margin_std": 125.44760131835938, "step": 243 }, { "epoch": 0.35829662261380324, "fcm_dpo/beta": 0.005800644867122173, "fcm_dpo/delta": -0.08758918195962906, "fcm_dpo/margin": 117.43812561035156, "fcm_dpo/q_t": 0.3503507375717163, "grad_norm": 31.091215133666992, "learning_rate": 4.067309514735267e-07, "logits/chosen": -0.42268094420433044, "logits/rejected": -0.4104187488555908, "logps/chosen": -163.81822204589844, "logps/ref_chosen": -61.140689849853516, "logps/ref_rejected": -94.89193725585938, "logps/rejected": -315.00762939453125, "loss": 0.9077, "margin_dpo/margin_mean": 117.43812561035156, "margin_dpo/margin_std": 110.86767578125, "step": 244 }, { "epoch": 0.35976505139500736, "fcm_dpo/beta": 0.00573451304808259, "fcm_dpo/delta": 0.01733044534921646, "fcm_dpo/margin": 101.28717803955078, "fcm_dpo/q_t": 0.3786167502403259, "grad_norm": 33.05402755737305, "learning_rate": 4.057290731287531e-07, "logits/chosen": -0.40215417742729187, "logits/rejected": -0.3637341856956482, "logps/chosen": -191.681396484375, "logps/ref_chosen": -67.26228332519531, "logps/ref_rejected": -87.64010620117188, "logps/rejected": -313.3464050292969, "loss": 1.0265, "margin_dpo/margin_mean": 101.28717041015625, "margin_dpo/margin_std": 134.58218383789062, "step": 245 }, { "epoch": 0.36123348017621143, "fcm_dpo/beta": 0.005821553990244865, "fcm_dpo/delta": 0.027677321806550026, "fcm_dpo/margin": 98.57552337646484, "fcm_dpo/q_t": 0.3820421099662781, "grad_norm": 26.930212020874023, "learning_rate": 4.047230911780736e-07, "logits/chosen": -0.4357486963272095, "logits/rejected": -0.394938200712204, "logps/chosen": -191.74822998046875, "logps/ref_chosen": -66.69696807861328, "logps/ref_rejected": -84.34634399414062, "logps/rejected": -307.97314453125, "loss": 1.0254, "margin_dpo/margin_mean": 98.57553100585938, "margin_dpo/margin_std": 138.3719482421875, "step": 246 }, { "epoch": 0.36270190895741555, "fcm_dpo/beta": 0.005621997639536858, "fcm_dpo/delta": -0.23428848385810852, "fcm_dpo/margin": 144.79278564453125, "fcm_dpo/q_t": 0.33090120553970337, "grad_norm": 26.57573890686035, "learning_rate": 4.0371303213004814e-07, "logits/chosen": -0.3413492441177368, "logits/rejected": -0.33773866295814514, "logps/chosen": -205.19589233398438, "logps/ref_chosen": -56.6053466796875, "logps/ref_rejected": -106.29326629638672, "logps/rejected": -399.6766357421875, "loss": 0.8815, "margin_dpo/margin_mean": 144.79278564453125, "margin_dpo/margin_std": 153.0501708984375, "step": 247 }, { "epoch": 0.3641703377386197, "fcm_dpo/beta": 0.005500612780451775, "fcm_dpo/delta": -0.0632907822728157, "fcm_dpo/margin": 119.71734619140625, "fcm_dpo/q_t": 0.35248565673828125, "grad_norm": 40.54594039916992, "learning_rate": 4.0269892260067197e-07, "logits/chosen": -0.3447737991809845, "logits/rejected": -0.358917772769928, "logps/chosen": -171.75363159179688, "logps/ref_chosen": -44.043216705322266, "logps/ref_rejected": -91.85687255859375, "logps/rejected": -339.2846374511719, "loss": 0.908, "margin_dpo/margin_mean": 119.71734619140625, "margin_dpo/margin_std": 105.92500305175781, "step": 248 }, { "epoch": 0.3656387665198238, "fcm_dpo/beta": 0.005672304891049862, "fcm_dpo/delta": 0.20714300870895386, "fcm_dpo/margin": 70.76122283935547, "fcm_dpo/q_t": 0.4150072932243347, "grad_norm": 52.06844711303711, "learning_rate": 4.0168078931267426e-07, "logits/chosen": -0.3665584921836853, "logits/rejected": -0.3416873514652252, "logps/chosen": -225.1993408203125, "logps/ref_chosen": -62.442352294921875, "logps/ref_rejected": -80.46806335449219, "logps/rejected": -313.98626708984375, "loss": 1.1559, "margin_dpo/margin_mean": 70.76122283935547, "margin_dpo/margin_std": 133.9126739501953, "step": 249 }, { "epoch": 0.3671071953010279, "fcm_dpo/beta": 0.005674234591424465, "fcm_dpo/delta": -0.10982278734445572, "fcm_dpo/margin": 123.74378967285156, "fcm_dpo/q_t": 0.34339314699172974, "grad_norm": 34.09236526489258, "learning_rate": 4.006586590948141e-07, "logits/chosen": -0.38642430305480957, "logits/rejected": -0.33151566982269287, "logps/chosen": -202.39520263671875, "logps/ref_chosen": -65.63668823242188, "logps/ref_rejected": -73.87184143066406, "logps/rejected": -334.3741455078125, "loss": 0.8893, "margin_dpo/margin_mean": 123.74378967285156, "margin_dpo/margin_std": 108.41996765136719, "step": 250 }, { "epoch": 0.368575624082232, "fcm_dpo/beta": 0.005667783785611391, "fcm_dpo/delta": 0.09314079582691193, "fcm_dpo/margin": 90.27076721191406, "fcm_dpo/q_t": 0.3912964463233948, "grad_norm": 54.34296798706055, "learning_rate": 3.9963255888117325e-07, "logits/chosen": -0.33801060914993286, "logits/rejected": -0.3055931329727173, "logps/chosen": -211.67941284179688, "logps/ref_chosen": -57.182716369628906, "logps/ref_rejected": -77.66343688964844, "logps/rejected": -322.430908203125, "loss": 1.0705, "margin_dpo/margin_mean": 90.27076721191406, "margin_dpo/margin_std": 133.482421875, "step": 251 }, { "epoch": 0.3700440528634361, "fcm_dpo/beta": 0.005632590502500534, "fcm_dpo/delta": -0.06894432753324509, "fcm_dpo/margin": 117.80094909667969, "fcm_dpo/q_t": 0.34721171855926514, "grad_norm": 34.09019088745117, "learning_rate": 3.9860251571044666e-07, "logits/chosen": -0.4143436849117279, "logits/rejected": -0.3699311912059784, "logps/chosen": -211.8092041015625, "logps/ref_chosen": -71.68563842773438, "logps/ref_rejected": -84.75799560546875, "logps/rejected": -342.6824951171875, "loss": 0.9072, "margin_dpo/margin_mean": 117.80094909667969, "margin_dpo/margin_std": 102.10981750488281, "step": 252 }, { "epoch": 0.37151248164464024, "fcm_dpo/beta": 0.0056702555157244205, "fcm_dpo/delta": -0.024740278720855713, "fcm_dpo/margin": 109.68414306640625, "fcm_dpo/q_t": 0.3643852472305298, "grad_norm": 29.00065040588379, "learning_rate": 3.9756855672522986e-07, "logits/chosen": -0.40758126974105835, "logits/rejected": -0.3990533649921417, "logps/chosen": -192.28323364257812, "logps/ref_chosen": -69.1339340209961, "logps/ref_rejected": -98.70252990722656, "logps/rejected": -331.5359802246094, "loss": 0.9822, "margin_dpo/margin_mean": 109.68414306640625, "margin_dpo/margin_std": 129.01275634765625, "step": 253 }, { "epoch": 0.37298091042584436, "fcm_dpo/beta": 0.005641551688313484, "fcm_dpo/delta": 0.04004208371043205, "fcm_dpo/margin": 99.6583251953125, "fcm_dpo/q_t": 0.3856009840965271, "grad_norm": 27.635013580322266, "learning_rate": 3.965307091713037e-07, "logits/chosen": -0.41606834530830383, "logits/rejected": -0.397452175617218, "logps/chosen": -183.28872680664062, "logps/ref_chosen": -54.154998779296875, "logps/ref_rejected": -90.30764770507812, "logps/rejected": -319.0997009277344, "loss": 1.0509, "margin_dpo/margin_mean": 99.6583251953125, "margin_dpo/margin_std": 151.13661193847656, "step": 254 }, { "epoch": 0.3744493392070485, "fcm_dpo/beta": 0.005625682417303324, "fcm_dpo/delta": 0.0007314831018447876, "fcm_dpo/margin": 106.40800476074219, "fcm_dpo/q_t": 0.37160855531692505, "grad_norm": 29.136993408203125, "learning_rate": 3.954890003969163e-07, "logits/chosen": -0.3470228314399719, "logits/rejected": -0.3381271958351135, "logps/chosen": -193.8470458984375, "logps/ref_chosen": -57.14167022705078, "logps/ref_rejected": -90.2085952758789, "logps/rejected": -333.32196044921875, "loss": 1.0155, "margin_dpo/margin_mean": 106.40800476074219, "margin_dpo/margin_std": 140.4176483154297, "step": 255 }, { "epoch": 0.37591776798825255, "fcm_dpo/beta": 0.005624064244329929, "fcm_dpo/delta": -0.03853630647063255, "fcm_dpo/margin": 113.09561157226562, "fcm_dpo/q_t": 0.3635770082473755, "grad_norm": 43.678157806396484, "learning_rate": 3.944434578520628e-07, "logits/chosen": -0.3604806661605835, "logits/rejected": -0.3640385866165161, "logps/chosen": -175.4744110107422, "logps/ref_chosen": -55.163490295410156, "logps/ref_rejected": -92.56291961669922, "logps/rejected": -325.969482421875, "loss": 0.9694, "margin_dpo/margin_mean": 113.09561157226562, "margin_dpo/margin_std": 133.815185546875, "step": 256 }, { "epoch": 0.37738619676945667, "fcm_dpo/beta": 0.005505690351128578, "fcm_dpo/delta": -0.06898978352546692, "fcm_dpo/margin": 120.23841857910156, "fcm_dpo/q_t": 0.35991615056991577, "grad_norm": 27.866411209106445, "learning_rate": 3.933941090877615e-07, "logits/chosen": -0.3533473610877991, "logits/rejected": -0.33796215057373047, "logps/chosen": -166.825439453125, "logps/ref_chosen": -49.42369842529297, "logps/ref_rejected": -79.53791809082031, "logps/rejected": -317.1780700683594, "loss": 0.9576, "margin_dpo/margin_mean": 120.23841857910156, "margin_dpo/margin_std": 135.40008544921875, "step": 257 }, { "epoch": 0.3788546255506608, "fcm_dpo/beta": 0.005546920001506805, "fcm_dpo/delta": -0.001775261014699936, "fcm_dpo/margin": 108.45994567871094, "fcm_dpo/q_t": 0.36892908811569214, "grad_norm": 33.50674819946289, "learning_rate": 3.923409817553284e-07, "logits/chosen": -0.35097846388816833, "logits/rejected": -0.34521350264549255, "logps/chosen": -207.83615112304688, "logps/ref_chosen": -59.384124755859375, "logps/ref_rejected": -95.99010467529297, "logps/rejected": -352.902099609375, "loss": 1.0141, "margin_dpo/margin_mean": 108.45994567871094, "margin_dpo/margin_std": 145.33441162109375, "step": 258 }, { "epoch": 0.3803230543318649, "fcm_dpo/beta": 0.005560112651437521, "fcm_dpo/delta": 0.057319387793540955, "fcm_dpo/margin": 98.13949584960938, "fcm_dpo/q_t": 0.38358962535858154, "grad_norm": 33.96443557739258, "learning_rate": 3.9128410360564793e-07, "logits/chosen": -0.41265004873275757, "logits/rejected": -0.40723663568496704, "logps/chosen": -191.66282653808594, "logps/ref_chosen": -52.828346252441406, "logps/ref_rejected": -89.191650390625, "logps/rejected": -326.1656494140625, "loss": 1.0362, "margin_dpo/margin_mean": 98.13948822021484, "margin_dpo/margin_std": 132.66073608398438, "step": 259 }, { "epoch": 0.38179148311306904, "fcm_dpo/beta": 0.005574359092861414, "fcm_dpo/delta": -0.05944538116455078, "fcm_dpo/margin": 117.58509826660156, "fcm_dpo/q_t": 0.35981056094169617, "grad_norm": 32.101783752441406, "learning_rate": 3.9022350248844246e-07, "logits/chosen": -0.37826740741729736, "logits/rejected": -0.3950265347957611, "logps/chosen": -189.1026611328125, "logps/ref_chosen": -47.41767501831055, "logps/ref_rejected": -95.08978271484375, "logps/rejected": -354.35986328125, "loss": 0.9407, "margin_dpo/margin_mean": 117.58509826660156, "margin_dpo/margin_std": 128.20523071289062, "step": 260 }, { "epoch": 0.3832599118942731, "fcm_dpo/beta": 0.005468597635626793, "fcm_dpo/delta": -0.06998618692159653, "fcm_dpo/margin": 121.63386535644531, "fcm_dpo/q_t": 0.35736051201820374, "grad_norm": 23.966012954711914, "learning_rate": 3.891592063515376e-07, "logits/chosen": -0.319491982460022, "logits/rejected": -0.3146633505821228, "logps/chosen": -190.02244567871094, "logps/ref_chosen": -53.03137969970703, "logps/ref_rejected": -88.51494598388672, "logps/rejected": -347.139892578125, "loss": 0.9529, "margin_dpo/margin_mean": 121.63386535644531, "margin_dpo/margin_std": 140.09190368652344, "step": 261 }, { "epoch": 0.38472834067547723, "fcm_dpo/beta": 0.005526003893464804, "fcm_dpo/delta": 0.0573844239115715, "fcm_dpo/margin": 98.6690902709961, "fcm_dpo/q_t": 0.37783992290496826, "grad_norm": 33.454994201660156, "learning_rate": 3.880912432401264e-07, "logits/chosen": -0.2890551686286926, "logits/rejected": -0.26014506816864014, "logps/chosen": -221.21661376953125, "logps/ref_chosen": -59.620140075683594, "logps/ref_rejected": -86.41853332519531, "logps/rejected": -346.68408203125, "loss": 1.0033, "margin_dpo/margin_mean": 98.6690902709961, "margin_dpo/margin_std": 113.80564880371094, "step": 262 }, { "epoch": 0.38619676945668135, "fcm_dpo/beta": 0.005366258788853884, "fcm_dpo/delta": -0.16116134822368622, "fcm_dpo/margin": 139.3936309814453, "fcm_dpo/q_t": 0.3375219702720642, "grad_norm": 30.789506912231445, "learning_rate": 3.870196412960302e-07, "logits/chosen": -0.3338819146156311, "logits/rejected": -0.30775582790374756, "logps/chosen": -204.53103637695312, "logps/ref_chosen": -59.42094421386719, "logps/ref_rejected": -96.85720825195312, "logps/rejected": -381.3609619140625, "loss": 0.8897, "margin_dpo/margin_mean": 139.3936309814453, "margin_dpo/margin_std": 136.70474243164062, "step": 263 }, { "epoch": 0.3876651982378855, "fcm_dpo/beta": 0.0052995504811406136, "fcm_dpo/delta": -0.014157909899950027, "fcm_dpo/margin": 115.54821014404297, "fcm_dpo/q_t": 0.3663901686668396, "grad_norm": 33.63625717163086, "learning_rate": 3.8594442875695665e-07, "logits/chosen": -0.3557894825935364, "logits/rejected": -0.3422415256500244, "logps/chosen": -212.91517639160156, "logps/ref_chosen": -62.722084045410156, "logps/ref_rejected": -93.85620880126953, "logps/rejected": -359.5975341796875, "loss": 0.9682, "margin_dpo/margin_mean": 115.5482177734375, "margin_dpo/margin_std": 127.98780822753906, "step": 264 }, { "epoch": 0.3891336270190896, "fcm_dpo/beta": 0.005331607535481453, "fcm_dpo/delta": -0.0023941383697092533, "fcm_dpo/margin": 112.95767211914062, "fcm_dpo/q_t": 0.3753606081008911, "grad_norm": 48.9280891418457, "learning_rate": 3.848656339557562e-07, "logits/chosen": -0.31224972009658813, "logits/rejected": -0.2951231598854065, "logps/chosen": -235.91409301757812, "logps/ref_chosen": -61.971466064453125, "logps/ref_rejected": -88.02059936523438, "logps/rejected": -374.9208984375, "loss": 1.0295, "margin_dpo/margin_mean": 112.95767211914062, "margin_dpo/margin_std": 163.15493774414062, "step": 265 }, { "epoch": 0.39060205580029367, "fcm_dpo/beta": 0.0053364490158855915, "fcm_dpo/delta": 0.014164826832711697, "fcm_dpo/margin": 109.93594360351562, "fcm_dpo/q_t": 0.37322813272476196, "grad_norm": 41.273826599121094, "learning_rate": 3.8378328531967507e-07, "logits/chosen": -0.343116819858551, "logits/rejected": -0.2906397581100464, "logps/chosen": -235.95025634765625, "logps/ref_chosen": -67.09967041015625, "logps/ref_rejected": -67.97122192382812, "logps/rejected": -346.75775146484375, "loss": 0.9942, "margin_dpo/margin_mean": 109.93594360351562, "margin_dpo/margin_std": 133.6815185546875, "step": 266 }, { "epoch": 0.3920704845814978, "fcm_dpo/beta": 0.005292531102895737, "fcm_dpo/delta": -0.037379078567028046, "fcm_dpo/margin": 119.896484375, "fcm_dpo/q_t": 0.36082565784454346, "grad_norm": 37.10322189331055, "learning_rate": 3.8269741136960646e-07, "logits/chosen": -0.3480778932571411, "logits/rejected": -0.30662471055984497, "logps/chosen": -230.77191162109375, "logps/ref_chosen": -68.97075653076172, "logps/ref_rejected": -90.16844940185547, "logps/rejected": -371.8660888671875, "loss": 0.9604, "margin_dpo/margin_mean": 119.896484375, "margin_dpo/margin_std": 134.846435546875, "step": 267 }, { "epoch": 0.3935389133627019, "fcm_dpo/beta": 0.0052706534042954445, "fcm_dpo/delta": -0.005634669214487076, "fcm_dpo/margin": 114.72981262207031, "fcm_dpo/q_t": 0.37070363759994507, "grad_norm": 36.76900100708008, "learning_rate": 3.8160804071933894e-07, "logits/chosen": -0.3364071846008301, "logits/rejected": -0.34323662519454956, "logps/chosen": -227.5570526123047, "logps/ref_chosen": -55.90031051635742, "logps/ref_rejected": -101.64763641357422, "logps/rejected": -388.0341796875, "loss": 0.9952, "margin_dpo/margin_mean": 114.72982025146484, "margin_dpo/margin_std": 145.24905395507812, "step": 268 }, { "epoch": 0.39500734214390604, "fcm_dpo/beta": 0.005157118663191795, "fcm_dpo/delta": -0.17022442817687988, "fcm_dpo/margin": 146.7906494140625, "fcm_dpo/q_t": 0.3414715826511383, "grad_norm": 42.1302375793457, "learning_rate": 3.8051520207480204e-07, "logits/chosen": -0.3691613972187042, "logits/rejected": -0.3452231287956238, "logps/chosen": -251.450439453125, "logps/ref_chosen": -70.03955841064453, "logps/ref_rejected": -107.34937286376953, "logps/rejected": -435.5509033203125, "loss": 0.925, "margin_dpo/margin_mean": 146.7906494140625, "margin_dpo/margin_std": 168.05130004882812, "step": 269 }, { "epoch": 0.3964757709251101, "fcm_dpo/beta": 0.005182851571589708, "fcm_dpo/delta": 0.05792323499917984, "fcm_dpo/margin": 105.11456298828125, "fcm_dpo/q_t": 0.3828323185443878, "grad_norm": 37.189781188964844, "learning_rate": 3.794189242333106e-07, "logits/chosen": -0.41337841749191284, "logits/rejected": -0.40365487337112427, "logps/chosen": -232.35296630859375, "logps/ref_chosen": -69.53347778320312, "logps/ref_rejected": -109.92864990234375, "logps/rejected": -377.8627014160156, "loss": 1.0389, "margin_dpo/margin_mean": 105.11456298828125, "margin_dpo/margin_std": 145.3082275390625, "step": 270 }, { "epoch": 0.39794419970631423, "fcm_dpo/beta": 0.0051218001171946526, "fcm_dpo/delta": -0.057347435504198074, "fcm_dpo/margin": 127.55348205566406, "fcm_dpo/q_t": 0.35683873295783997, "grad_norm": 39.590396881103516, "learning_rate": 3.7831923608280514e-07, "logits/chosen": -0.34910643100738525, "logits/rejected": -0.3228174149990082, "logps/chosen": -206.6493377685547, "logps/ref_chosen": -56.76456832885742, "logps/ref_rejected": -92.51383972167969, "logps/rejected": -369.95208740234375, "loss": 0.9256, "margin_dpo/margin_mean": 127.55348205566406, "margin_dpo/margin_std": 125.99525451660156, "step": 271 }, { "epoch": 0.39941262848751835, "fcm_dpo/beta": 0.005027889274060726, "fcm_dpo/delta": -0.1585131734609604, "fcm_dpo/margin": 148.55596923828125, "fcm_dpo/q_t": 0.33690333366394043, "grad_norm": 36.002960205078125, "learning_rate": 3.772161666010912e-07, "logits/chosen": -0.331684410572052, "logits/rejected": -0.3400617837905884, "logps/chosen": -184.74874877929688, "logps/ref_chosen": -49.497154235839844, "logps/ref_rejected": -105.54279327392578, "logps/rejected": -389.350341796875, "loss": 0.8841, "margin_dpo/margin_mean": 148.55596923828125, "margin_dpo/margin_std": 140.95240783691406, "step": 272 }, { "epoch": 0.4008810572687225, "fcm_dpo/beta": 0.004825687035918236, "fcm_dpo/delta": -0.1409759819507599, "fcm_dpo/margin": 151.05177307128906, "fcm_dpo/q_t": 0.3416686952114105, "grad_norm": 34.01618576049805, "learning_rate": 3.761097448550755e-07, "logits/chosen": -0.3630342483520508, "logits/rejected": -0.3401733636856079, "logps/chosen": -218.07992553710938, "logps/ref_chosen": -62.97539520263672, "logps/ref_rejected": -92.49858093261719, "logps/rejected": -398.6549072265625, "loss": 0.892, "margin_dpo/margin_mean": 151.05178833007812, "margin_dpo/margin_std": 145.12493896484375, "step": 273 }, { "epoch": 0.4023494860499266, "fcm_dpo/beta": 0.004842686001211405, "fcm_dpo/delta": 0.012656617909669876, "fcm_dpo/margin": 121.4140625, "fcm_dpo/q_t": 0.3711016774177551, "grad_norm": 29.5135555267334, "learning_rate": 3.75e-07, "logits/chosen": -0.2929585874080658, "logits/rejected": -0.26695334911346436, "logps/chosen": -246.96438598632812, "logps/ref_chosen": -55.66770935058594, "logps/ref_rejected": -77.33308410644531, "logps/rejected": -390.0438232421875, "loss": 0.9834, "margin_dpo/margin_mean": 121.4140625, "margin_dpo/margin_std": 138.95596313476562, "step": 274 }, { "epoch": 0.40381791483113066, "fcm_dpo/beta": 0.0048539480194449425, "fcm_dpo/delta": 0.006926249712705612, "fcm_dpo/margin": 122.2330551147461, "fcm_dpo/q_t": 0.3715580701828003, "grad_norm": 31.085887908935547, "learning_rate": 3.738869612786737e-07, "logits/chosen": -0.3700248599052429, "logits/rejected": -0.3734614849090576, "logps/chosen": -204.39918518066406, "logps/ref_chosen": -48.594703674316406, "logps/ref_rejected": -93.30369567871094, "logps/rejected": -371.34124755859375, "loss": 0.995, "margin_dpo/margin_mean": 122.2330551147461, "margin_dpo/margin_std": 148.9960479736328, "step": 275 }, { "epoch": 0.4052863436123348, "fcm_dpo/beta": 0.004785975441336632, "fcm_dpo/delta": -0.05713774263858795, "fcm_dpo/margin": 136.44924926757812, "fcm_dpo/q_t": 0.3614548146724701, "grad_norm": 38.087745666503906, "learning_rate": 3.7277065802070204e-07, "logits/chosen": -0.3316674828529358, "logits/rejected": -0.3020857274532318, "logps/chosen": -219.54293823242188, "logps/ref_chosen": -56.57740783691406, "logps/ref_rejected": -70.36566925048828, "logps/rejected": -369.78045654296875, "loss": 0.9619, "margin_dpo/margin_mean": 136.44924926757812, "margin_dpo/margin_std": 161.114990234375, "step": 276 }, { "epoch": 0.4067547723935389, "fcm_dpo/beta": 0.0047689881175756454, "fcm_dpo/delta": -0.04522915184497833, "fcm_dpo/margin": 134.67898559570312, "fcm_dpo/q_t": 0.362157940864563, "grad_norm": 36.099918365478516, "learning_rate": 3.71651119641714e-07, "logits/chosen": -0.3248156011104584, "logits/rejected": -0.3068217635154724, "logps/chosen": -242.65780639648438, "logps/ref_chosen": -56.27156066894531, "logps/ref_rejected": -92.88127136230469, "logps/rejected": -413.946533203125, "loss": 0.9632, "margin_dpo/margin_mean": 134.67898559570312, "margin_dpo/margin_std": 155.36151123046875, "step": 277 }, { "epoch": 0.40822320117474303, "fcm_dpo/beta": 0.004655085504055023, "fcm_dpo/delta": -0.10972905904054642, "fcm_dpo/margin": 150.7223358154297, "fcm_dpo/q_t": 0.3516245484352112, "grad_norm": 29.135597229003906, "learning_rate": 3.705283756425872e-07, "logits/chosen": -0.3341606557369232, "logits/rejected": -0.336972713470459, "logps/chosen": -221.07388305664062, "logps/ref_chosen": -52.94194030761719, "logps/ref_rejected": -91.25357818603516, "logps/rejected": -410.10784912109375, "loss": 0.9254, "margin_dpo/margin_mean": 150.72235107421875, "margin_dpo/margin_std": 164.86074829101562, "step": 278 }, { "epoch": 0.40969162995594716, "fcm_dpo/beta": 0.004513373598456383, "fcm_dpo/delta": -0.13143345713615417, "fcm_dpo/margin": 159.43524169921875, "fcm_dpo/q_t": 0.35075873136520386, "grad_norm": 37.86472702026367, "learning_rate": 3.6940245560867e-07, "logits/chosen": -0.2502022981643677, "logits/rejected": -0.24658802151679993, "logps/chosen": -250.07838439941406, "logps/ref_chosen": -48.641319274902344, "logps/ref_rejected": -87.8514404296875, "logps/rejected": -448.7237548828125, "loss": 0.9479, "margin_dpo/margin_mean": 159.4352569580078, "margin_dpo/margin_std": 191.3142852783203, "step": 279 }, { "epoch": 0.4111600587371512, "fcm_dpo/beta": 0.00445135310292244, "fcm_dpo/delta": -0.09706270694732666, "fcm_dpo/margin": 155.06472778320312, "fcm_dpo/q_t": 0.3463428020477295, "grad_norm": 40.227508544921875, "learning_rate": 3.6827338920900253e-07, "logits/chosen": -0.28654709458351135, "logits/rejected": -0.28727859258651733, "logps/chosen": -259.53009033203125, "logps/ref_chosen": -58.797122955322266, "logps/ref_rejected": -98.61885070800781, "logps/rejected": -454.41656494140625, "loss": 0.9052, "margin_dpo/margin_mean": 155.0647430419922, "margin_dpo/margin_std": 146.40179443359375, "step": 280 }, { "epoch": 0.41262848751835535, "fcm_dpo/beta": 0.004433557391166687, "fcm_dpo/delta": -0.03812451288104057, "fcm_dpo/margin": 143.29185485839844, "fcm_dpo/q_t": 0.3562266230583191, "grad_norm": 30.542402267456055, "learning_rate": 3.6714120619553435e-07, "logits/chosen": -0.34395909309387207, "logits/rejected": -0.3141595721244812, "logps/chosen": -239.7939453125, "logps/ref_chosen": -55.488521575927734, "logps/ref_rejected": -80.88258361816406, "logps/rejected": -408.4798583984375, "loss": 0.9469, "margin_dpo/margin_mean": 143.29185485839844, "margin_dpo/margin_std": 147.6947021484375, "step": 281 }, { "epoch": 0.41409691629955947, "fcm_dpo/beta": 0.004496179986745119, "fcm_dpo/delta": 0.11047537624835968, "fcm_dpo/margin": 109.85005187988281, "fcm_dpo/q_t": 0.39475515484809875, "grad_norm": 32.4976806640625, "learning_rate": 3.660059364023408e-07, "logits/chosen": -0.40445345640182495, "logits/rejected": -0.37312695384025574, "logps/chosen": -266.0508728027344, "logps/ref_chosen": -73.07014465332031, "logps/ref_rejected": -95.35098266601562, "logps/rejected": -398.1817626953125, "loss": 1.0544, "margin_dpo/margin_mean": 109.85005187988281, "margin_dpo/margin_std": 156.09814453125, "step": 282 }, { "epoch": 0.4155653450807636, "fcm_dpo/beta": 0.004433175548911095, "fcm_dpo/delta": -0.12632399797439575, "fcm_dpo/margin": 161.82952880859375, "fcm_dpo/q_t": 0.341400146484375, "grad_norm": 44.17348098754883, "learning_rate": 3.6486760974483685e-07, "logits/chosen": -0.3220086991786957, "logits/rejected": -0.3167068660259247, "logps/chosen": -252.05734252929688, "logps/ref_chosen": -61.89844512939453, "logps/ref_rejected": -96.98655700683594, "logps/rejected": -448.9749755859375, "loss": 0.8904, "margin_dpo/margin_mean": 161.8295135498047, "margin_dpo/margin_std": 149.40249633789062, "step": 283 }, { "epoch": 0.4170337738619677, "fcm_dpo/beta": 0.004331875126808882, "fcm_dpo/delta": -0.0721510797739029, "fcm_dpo/margin": 154.01272583007812, "fcm_dpo/q_t": 0.35365206003189087, "grad_norm": 28.796356201171875, "learning_rate": 3.6372625621898863e-07, "logits/chosen": -0.39249300956726074, "logits/rejected": -0.37775149941444397, "logps/chosen": -239.20748901367188, "logps/ref_chosen": -58.4355354309082, "logps/ref_rejected": -93.46926879882812, "logps/rejected": -428.2539367675781, "loss": 0.9158, "margin_dpo/margin_mean": 154.01272583007812, "margin_dpo/margin_std": 150.7726593017578, "step": 284 }, { "epoch": 0.4185022026431718, "fcm_dpo/beta": 0.004361903294920921, "fcm_dpo/delta": 0.014079738408327103, "fcm_dpo/margin": 134.28880310058594, "fcm_dpo/q_t": 0.3673364818096161, "grad_norm": 34.644283294677734, "learning_rate": 3.625819059005228e-07, "logits/chosen": -0.3506457209587097, "logits/rejected": -0.32794976234436035, "logps/chosen": -276.463134765625, "logps/ref_chosen": -66.23219299316406, "logps/ref_rejected": -99.1268310546875, "logps/rejected": -443.6466064453125, "loss": 0.9694, "margin_dpo/margin_mean": 134.28878784179688, "margin_dpo/margin_std": 137.147216796875, "step": 285 }, { "epoch": 0.4199706314243759, "fcm_dpo/beta": 0.004304747097194195, "fcm_dpo/delta": -0.060156404972076416, "fcm_dpo/margin": 152.447265625, "fcm_dpo/q_t": 0.3558097183704376, "grad_norm": 35.37562561035156, "learning_rate": 3.614345889441346e-07, "logits/chosen": -0.3511343002319336, "logits/rejected": -0.32436245679855347, "logps/chosen": -274.97412109375, "logps/ref_chosen": -72.95100402832031, "logps/ref_rejected": -88.58845520019531, "logps/rejected": -443.058837890625, "loss": 0.9432, "margin_dpo/margin_mean": 152.447265625, "margin_dpo/margin_std": 162.77908325195312, "step": 286 }, { "epoch": 0.42143906020558003, "fcm_dpo/beta": 0.0043206894770264626, "fcm_dpo/delta": 0.0441654808819294, "fcm_dpo/margin": 129.17588806152344, "fcm_dpo/q_t": 0.3783441185951233, "grad_norm": 31.608234405517578, "learning_rate": 3.6028433558269275e-07, "logits/chosen": -0.31515616178512573, "logits/rejected": -0.28630921244621277, "logps/chosen": -264.4310607910156, "logps/ref_chosen": -61.54115295410156, "logps/ref_rejected": -77.69607543945312, "logps/rejected": -409.7618408203125, "loss": 1.0024, "margin_dpo/margin_mean": 129.17587280273438, "margin_dpo/margin_std": 156.08370971679688, "step": 287 }, { "epoch": 0.42290748898678415, "fcm_dpo/beta": 0.00423096539452672, "fcm_dpo/delta": -0.06944943964481354, "fcm_dpo/margin": 156.4290771484375, "fcm_dpo/q_t": 0.3512691259384155, "grad_norm": 27.641883850097656, "learning_rate": 3.5913117612644327e-07, "logits/chosen": -0.32602715492248535, "logits/rejected": -0.3050415515899658, "logps/chosen": -260.18255615234375, "logps/ref_chosen": -56.661224365234375, "logps/ref_rejected": -87.33570098876953, "logps/rejected": -447.2861022949219, "loss": 0.914, "margin_dpo/margin_mean": 156.4290771484375, "margin_dpo/margin_std": 137.47991943359375, "step": 288 }, { "epoch": 0.4243759177679883, "fcm_dpo/beta": 0.0041911243461072445, "fcm_dpo/delta": -0.14678727090358734, "fcm_dpo/margin": 175.65328979492188, "fcm_dpo/q_t": 0.34263110160827637, "grad_norm": 36.062095642089844, "learning_rate": 3.5797514096221024e-07, "logits/chosen": -0.26659348607063293, "logits/rejected": -0.25999683141708374, "logps/chosen": -253.7037353515625, "logps/ref_chosen": -45.23039245605469, "logps/ref_rejected": -87.64266967773438, "logps/rejected": -471.769287109375, "loss": 0.8987, "margin_dpo/margin_mean": 175.65328979492188, "margin_dpo/margin_std": 176.63510131835938, "step": 289 }, { "epoch": 0.42584434654919234, "fcm_dpo/beta": 0.004051923286169767, "fcm_dpo/delta": -0.12010594457387924, "fcm_dpo/margin": 175.46475219726562, "fcm_dpo/q_t": 0.35006120800971985, "grad_norm": 32.5576286315918, "learning_rate": 3.568162605525952e-07, "logits/chosen": -0.24551498889923096, "logits/rejected": -0.262775182723999, "logps/chosen": -283.1232604980469, "logps/ref_chosen": -55.47149658203125, "logps/ref_rejected": -116.70857238769531, "logps/rejected": -519.8250732421875, "loss": 0.9297, "margin_dpo/margin_mean": 175.46475219726562, "margin_dpo/margin_std": 197.48675537109375, "step": 290 }, { "epoch": 0.42731277533039647, "fcm_dpo/beta": 0.004028604365885258, "fcm_dpo/delta": -0.06118813902139664, "fcm_dpo/margin": 163.03326416015625, "fcm_dpo/q_t": 0.35449501872062683, "grad_norm": 26.545326232910156, "learning_rate": 3.5565456543517485e-07, "logits/chosen": -0.30797719955444336, "logits/rejected": -0.2923903167247772, "logps/chosen": -260.06689453125, "logps/ref_chosen": -63.26036834716797, "logps/ref_rejected": -89.29708862304688, "logps/rejected": -449.1368713378906, "loss": 0.9354, "margin_dpo/margin_mean": 163.03326416015625, "margin_dpo/margin_std": 164.8643798828125, "step": 291 }, { "epoch": 0.4287812041116006, "fcm_dpo/beta": 0.003977569285780191, "fcm_dpo/delta": -0.05237273871898651, "fcm_dpo/margin": 163.1212615966797, "fcm_dpo/q_t": 0.3588453233242035, "grad_norm": 34.14498519897461, "learning_rate": 3.5449008622169583e-07, "logits/chosen": -0.262368768453598, "logits/rejected": -0.23994597792625427, "logps/chosen": -277.8984375, "logps/ref_chosen": -53.91852951049805, "logps/ref_rejected": -89.96138000488281, "logps/rejected": -477.06256103515625, "loss": 0.9758, "margin_dpo/margin_mean": 163.12124633789062, "margin_dpo/margin_std": 198.7785186767578, "step": 292 }, { "epoch": 0.4302496328928047, "fcm_dpo/beta": 0.004008471965789795, "fcm_dpo/delta": 0.07511234283447266, "fcm_dpo/margin": 131.8068389892578, "fcm_dpo/q_t": 0.38762548565864563, "grad_norm": 37.924625396728516, "learning_rate": 3.5332285359726846e-07, "logits/chosen": -0.3276376724243164, "logits/rejected": -0.3103697896003723, "logps/chosen": -269.2470703125, "logps/ref_chosen": -60.376033782958984, "logps/ref_rejected": -77.85244750976562, "logps/rejected": -418.53033447265625, "loss": 1.0488, "margin_dpo/margin_mean": 131.8068389892578, "margin_dpo/margin_std": 185.1345672607422, "step": 293 }, { "epoch": 0.43171806167400884, "fcm_dpo/beta": 0.0040005166083574295, "fcm_dpo/delta": 0.008330628275871277, "fcm_dpo/margin": 147.9698486328125, "fcm_dpo/q_t": 0.37461739778518677, "grad_norm": 28.923660278320312, "learning_rate": 3.5215289831955786e-07, "logits/chosen": -0.293660044670105, "logits/rejected": -0.2946998178958893, "logps/chosen": -237.821533203125, "logps/ref_chosen": -48.0875358581543, "logps/ref_rejected": -81.89698791503906, "logps/rejected": -419.600830078125, "loss": 0.9972, "margin_dpo/margin_mean": 147.9698486328125, "margin_dpo/margin_std": 184.6291961669922, "step": 294 }, { "epoch": 0.4331864904552129, "fcm_dpo/beta": 0.0040347035974264145, "fcm_dpo/delta": 0.02709888480603695, "fcm_dpo/margin": 142.37374877929688, "fcm_dpo/q_t": 0.3751312494277954, "grad_norm": 33.199710845947266, "learning_rate": 3.509802512179737e-07, "logits/chosen": -0.3318241834640503, "logits/rejected": -0.3298850655555725, "logps/chosen": -268.1130065917969, "logps/ref_chosen": -49.92467498779297, "logps/ref_rejected": -87.45632934570312, "logps/rejected": -448.01837158203125, "loss": 1.0026, "margin_dpo/margin_mean": 142.3737335205078, "margin_dpo/margin_std": 173.57199096679688, "step": 295 }, { "epoch": 0.434654919236417, "fcm_dpo/beta": 0.0041085234843194485, "fcm_dpo/delta": 0.1538042277097702, "fcm_dpo/margin": 110.34952545166016, "fcm_dpo/q_t": 0.40237492322921753, "grad_norm": 50.468746185302734, "learning_rate": 3.498049431928577e-07, "logits/chosen": -0.38077211380004883, "logits/rejected": -0.3572128415107727, "logps/chosen": -313.87554931640625, "logps/ref_chosen": -65.49124145507812, "logps/ref_rejected": -93.08908081054688, "logps/rejected": -451.82293701171875, "loss": 1.1194, "margin_dpo/margin_mean": 110.34951782226562, "margin_dpo/margin_std": 187.090576171875, "step": 296 }, { "epoch": 0.43612334801762115, "fcm_dpo/beta": 0.004196628928184509, "fcm_dpo/delta": 0.063119076192379, "fcm_dpo/margin": 128.75506591796875, "fcm_dpo/q_t": 0.3794586956501007, "grad_norm": 28.69700813293457, "learning_rate": 3.486270052146694e-07, "logits/chosen": -0.3626411557197571, "logits/rejected": -0.3594970703125, "logps/chosen": -282.9952697753906, "logps/ref_chosen": -56.476951599121094, "logps/ref_rejected": -95.1385498046875, "logps/rejected": -450.41192626953125, "loss": 1.0008, "margin_dpo/margin_mean": 128.7550506591797, "margin_dpo/margin_std": 144.84683227539062, "step": 297 }, { "epoch": 0.43759177679882527, "fcm_dpo/beta": 0.004173451103270054, "fcm_dpo/delta": -0.09957937896251678, "fcm_dpo/margin": 165.98265075683594, "fcm_dpo/q_t": 0.3572956323623657, "grad_norm": 29.900131225585938, "learning_rate": 3.474464683231698e-07, "logits/chosen": -0.4068795442581177, "logits/rejected": -0.42597341537475586, "logps/chosen": -299.30560302734375, "logps/ref_chosen": -67.32516479492188, "logps/ref_rejected": -116.66217041015625, "logps/rejected": -514.625244140625, "loss": 0.954, "margin_dpo/margin_mean": 165.98265075683594, "margin_dpo/margin_std": 202.80752563476562, "step": 298 }, { "epoch": 0.4390602055800294, "fcm_dpo/beta": 0.00413914117962122, "fcm_dpo/delta": -0.04581880569458008, "fcm_dpo/margin": 155.09854125976562, "fcm_dpo/q_t": 0.3642904758453369, "grad_norm": 38.76698684692383, "learning_rate": 3.462633636266041e-07, "logits/chosen": -0.3764153718948364, "logits/rejected": -0.3700462579727173, "logps/chosen": -246.14369201660156, "logps/ref_chosen": -48.96209716796875, "logps/ref_rejected": -84.32823944091797, "logps/rejected": -436.6083984375, "loss": 0.9702, "margin_dpo/margin_mean": 155.0985565185547, "margin_dpo/margin_std": 182.65501403808594, "step": 299 }, { "epoch": 0.44052863436123346, "fcm_dpo/beta": 0.004044054076075554, "fcm_dpo/delta": -0.1144830510020256, "fcm_dpo/margin": 174.70237731933594, "fcm_dpo/q_t": 0.3509829640388489, "grad_norm": 42.70845413208008, "learning_rate": 3.4507772230088147e-07, "logits/chosen": -0.3672061562538147, "logits/rejected": -0.36297306418418884, "logps/chosen": -328.4141845703125, "logps/ref_chosen": -59.07371139526367, "logps/ref_rejected": -95.9664535522461, "logps/rejected": -540.00927734375, "loss": 0.9647, "margin_dpo/margin_mean": 174.70237731933594, "margin_dpo/margin_std": 214.1061248779297, "step": 300 }, { "epoch": 0.4419970631424376, "fcm_dpo/beta": 0.003927802667021751, "fcm_dpo/delta": -0.099482960999012, "fcm_dpo/margin": 176.13919067382812, "fcm_dpo/q_t": 0.3530727028846741, "grad_norm": 30.08980369567871, "learning_rate": 3.4388957558875316e-07, "logits/chosen": -0.40564534068107605, "logits/rejected": -0.3970520496368408, "logps/chosen": -284.28485107421875, "logps/ref_chosen": -57.249366760253906, "logps/ref_rejected": -92.35354614257812, "logps/rejected": -495.5281982421875, "loss": 0.9267, "margin_dpo/margin_mean": 176.1392059326172, "margin_dpo/margin_std": 191.83755493164062, "step": 301 }, { "epoch": 0.4434654919236417, "fcm_dpo/beta": 0.0039380136877298355, "fcm_dpo/delta": 0.05164004862308502, "fcm_dpo/margin": 139.933349609375, "fcm_dpo/q_t": 0.3805837631225586, "grad_norm": 39.73493957519531, "learning_rate": 3.426989547989902e-07, "logits/chosen": -0.4200012683868408, "logits/rejected": -0.4234361946582794, "logps/chosen": -265.1363830566406, "logps/ref_chosen": -51.197994232177734, "logps/ref_rejected": -97.22636413574219, "logps/rejected": -451.0981140136719, "loss": 1.0152, "margin_dpo/margin_mean": 139.933349609375, "margin_dpo/margin_std": 175.3145751953125, "step": 302 }, { "epoch": 0.44493392070484583, "fcm_dpo/beta": 0.003999094944447279, "fcm_dpo/delta": 0.06958974152803421, "fcm_dpo/margin": 133.57589721679688, "fcm_dpo/q_t": 0.386226624250412, "grad_norm": 40.41800308227539, "learning_rate": 3.4150589130555773e-07, "logits/chosen": -0.41884028911590576, "logits/rejected": -0.39589619636535645, "logps/chosen": -288.8157958984375, "logps/ref_chosen": -66.71394348144531, "logps/ref_rejected": -86.94542694091797, "logps/rejected": -442.6231994628906, "loss": 1.0405, "margin_dpo/margin_mean": 133.57591247558594, "margin_dpo/margin_std": 183.11703491210938, "step": 303 }, { "epoch": 0.44640234948604995, "fcm_dpo/beta": 0.0040207491256296635, "fcm_dpo/delta": -0.03515169024467468, "fcm_dpo/margin": 157.3292694091797, "fcm_dpo/q_t": 0.35517168045043945, "grad_norm": 49.5794563293457, "learning_rate": 3.403104165467883e-07, "logits/chosen": -0.42296552658081055, "logits/rejected": -0.4050954580307007, "logps/chosen": -266.1893310546875, "logps/ref_chosen": -71.95069885253906, "logps/ref_rejected": -90.47203063964844, "logps/rejected": -442.0399169921875, "loss": 0.9141, "margin_dpo/margin_mean": 157.32925415039062, "margin_dpo/margin_std": 126.93194580078125, "step": 304 }, { "epoch": 0.447870778267254, "fcm_dpo/beta": 0.003973768092691898, "fcm_dpo/delta": 0.022460877895355225, "fcm_dpo/margin": 145.38482666015625, "fcm_dpo/q_t": 0.3763822913169861, "grad_norm": 28.328157424926758, "learning_rate": 3.391125620245535e-07, "logits/chosen": -0.42863526940345764, "logits/rejected": -0.40163588523864746, "logps/chosen": -273.1666259765625, "logps/ref_chosen": -66.79523468017578, "logps/ref_rejected": -92.75459289550781, "logps/rejected": -444.51080322265625, "loss": 1.0001, "margin_dpo/margin_mean": 145.38482666015625, "margin_dpo/margin_std": 177.77105712890625, "step": 305 }, { "epoch": 0.44933920704845814, "fcm_dpo/beta": 0.004037954844534397, "fcm_dpo/delta": 0.03192281723022461, "fcm_dpo/margin": 141.09954833984375, "fcm_dpo/q_t": 0.37558555603027344, "grad_norm": 32.04235076904297, "learning_rate": 3.3791235930343417e-07, "logits/chosen": -0.3904664218425751, "logits/rejected": -0.3591429889202118, "logps/chosen": -269.64520263671875, "logps/ref_chosen": -69.68389892578125, "logps/ref_rejected": -85.15919494628906, "logps/rejected": -426.2200622558594, "loss": 0.9881, "margin_dpo/margin_mean": 141.09954833984375, "margin_dpo/margin_std": 163.01104736328125, "step": 306 }, { "epoch": 0.45080763582966227, "fcm_dpo/beta": 0.004054022952914238, "fcm_dpo/delta": 0.060959137976169586, "fcm_dpo/margin": 133.7398681640625, "fcm_dpo/q_t": 0.3805789649486542, "grad_norm": 42.768367767333984, "learning_rate": 3.367098400098881e-07, "logits/chosen": -0.36754947900772095, "logits/rejected": -0.3443043828010559, "logps/chosen": -266.11358642578125, "logps/ref_chosen": -70.16542053222656, "logps/ref_rejected": -86.97230529785156, "logps/rejected": -416.6603698730469, "loss": 1.0175, "margin_dpo/margin_mean": 133.7398681640625, "margin_dpo/margin_std": 165.68634033203125, "step": 307 }, { "epoch": 0.4522760646108664, "fcm_dpo/beta": 0.004061142913997173, "fcm_dpo/delta": -0.046599872410297394, "fcm_dpo/margin": 158.4833526611328, "fcm_dpo/q_t": 0.3559141755104065, "grad_norm": 34.26539993286133, "learning_rate": 3.355050358314172e-07, "logits/chosen": -0.3417690396308899, "logits/rejected": -0.32422229647636414, "logps/chosen": -250.93490600585938, "logps/ref_chosen": -55.2449951171875, "logps/ref_rejected": -79.37226104736328, "logps/rejected": -433.5455017089844, "loss": 0.9279, "margin_dpo/margin_mean": 158.48333740234375, "margin_dpo/margin_std": 150.41107177734375, "step": 308 }, { "epoch": 0.45374449339207046, "fcm_dpo/beta": 0.004053506068885326, "fcm_dpo/delta": -0.04356030374765396, "fcm_dpo/margin": 157.86776733398438, "fcm_dpo/q_t": 0.35849449038505554, "grad_norm": 36.413551330566406, "learning_rate": 3.3429797851573183e-07, "logits/chosen": -0.3285176157951355, "logits/rejected": -0.31304115056991577, "logps/chosen": -240.2936248779297, "logps/ref_chosen": -48.959083557128906, "logps/ref_rejected": -82.34072875976562, "logps/rejected": -431.54302978515625, "loss": 0.9503, "margin_dpo/margin_mean": 157.86776733398438, "margin_dpo/margin_std": 165.03907775878906, "step": 309 }, { "epoch": 0.4552129221732746, "fcm_dpo/beta": 0.004054386168718338, "fcm_dpo/delta": 0.03481416404247284, "fcm_dpo/margin": 139.72027587890625, "fcm_dpo/q_t": 0.3711395859718323, "grad_norm": 30.295249938964844, "learning_rate": 3.3308869986991487e-07, "logits/chosen": -0.35680243372917175, "logits/rejected": -0.33559316396713257, "logps/chosen": -279.739501953125, "logps/ref_chosen": -62.74177932739258, "logps/ref_rejected": -79.9302978515625, "logps/rejected": -436.6482849121094, "loss": 0.9662, "margin_dpo/margin_mean": 139.72027587890625, "margin_dpo/margin_std": 132.34963989257812, "step": 310 }, { "epoch": 0.4566813509544787, "fcm_dpo/beta": 0.003984658978879452, "fcm_dpo/delta": -0.08475086092948914, "fcm_dpo/margin": 170.3639373779297, "fcm_dpo/q_t": 0.3561347723007202, "grad_norm": 27.640430450439453, "learning_rate": 3.3187723175958346e-07, "logits/chosen": -0.33530759811401367, "logits/rejected": -0.30850982666015625, "logps/chosen": -278.0128479003906, "logps/ref_chosen": -53.02798080444336, "logps/ref_rejected": -77.43820190429688, "logps/rejected": -472.7869873046875, "loss": 0.9432, "margin_dpo/margin_mean": 170.3639373779297, "margin_dpo/margin_std": 193.53854370117188, "step": 311 }, { "epoch": 0.4581497797356828, "fcm_dpo/beta": 0.003978079650551081, "fcm_dpo/delta": -0.023921042680740356, "fcm_dpo/margin": 156.37979125976562, "fcm_dpo/q_t": 0.36489611864089966, "grad_norm": 29.57942008972168, "learning_rate": 3.306636061080487e-07, "logits/chosen": -0.27948933839797974, "logits/rejected": -0.261229932308197, "logps/chosen": -254.8425750732422, "logps/ref_chosen": -49.39221954345703, "logps/ref_rejected": -75.79280853271484, "logps/rejected": -437.62298583984375, "loss": 0.9616, "margin_dpo/margin_mean": 156.37979125976562, "margin_dpo/margin_std": 174.12229919433594, "step": 312 }, { "epoch": 0.45961820851688695, "fcm_dpo/beta": 0.0040033310651779175, "fcm_dpo/delta": 0.01650937646627426, "fcm_dpo/margin": 145.47708129882812, "fcm_dpo/q_t": 0.3724866807460785, "grad_norm": 33.476741790771484, "learning_rate": 3.2944785489547537e-07, "logits/chosen": -0.3220197558403015, "logits/rejected": -0.3162776529788971, "logps/chosen": -249.13385009765625, "logps/ref_chosen": -50.152740478515625, "logps/ref_rejected": -86.40620422363281, "logps/rejected": -430.8643798828125, "loss": 1.0053, "margin_dpo/margin_mean": 145.47706604003906, "margin_dpo/margin_std": 175.97396850585938, "step": 313 }, { "epoch": 0.461086637298091, "fcm_dpo/beta": 0.0039779325015842915, "fcm_dpo/delta": 0.053729020059108734, "fcm_dpo/margin": 137.9976806640625, "fcm_dpo/q_t": 0.38239920139312744, "grad_norm": 28.024948120117188, "learning_rate": 3.2823001015803857e-07, "logits/chosen": -0.37483978271484375, "logits/rejected": -0.37171846628189087, "logps/chosen": -271.1210632324219, "logps/ref_chosen": -57.237579345703125, "logps/ref_rejected": -97.5965347290039, "logps/rejected": -449.4776611328125, "loss": 1.0322, "margin_dpo/margin_mean": 137.99769592285156, "margin_dpo/margin_std": 185.14695739746094, "step": 314 }, { "epoch": 0.46255506607929514, "fcm_dpo/beta": 0.004043739754706621, "fcm_dpo/delta": 0.06402629613876343, "fcm_dpo/margin": 133.40902709960938, "fcm_dpo/q_t": 0.3819577395915985, "grad_norm": 34.34666442871094, "learning_rate": 3.270101039870797e-07, "logits/chosen": -0.25970029830932617, "logits/rejected": -0.2599654793739319, "logps/chosen": -258.28570556640625, "logps/ref_chosen": -49.06958770751953, "logps/ref_rejected": -85.68087768554688, "logps/rejected": -428.30596923828125, "loss": 1.0081, "margin_dpo/margin_mean": 133.40902709960938, "margin_dpo/margin_std": 158.57159423828125, "step": 315 }, { "epoch": 0.46402349486049926, "fcm_dpo/beta": 0.0039628720842301846, "fcm_dpo/delta": -0.11757272481918335, "fcm_dpo/margin": 178.52488708496094, "fcm_dpo/q_t": 0.34840020537376404, "grad_norm": 29.264596939086914, "learning_rate": 3.2578816852826086e-07, "logits/chosen": -0.30190056562423706, "logits/rejected": -0.3046305775642395, "logps/chosen": -262.7852783203125, "logps/ref_chosen": -54.26074981689453, "logps/ref_rejected": -101.2814712524414, "logps/rejected": -488.33087158203125, "loss": 0.9125, "margin_dpo/margin_mean": 178.52488708496094, "margin_dpo/margin_std": 183.82940673828125, "step": 316 }, { "epoch": 0.4654919236417034, "fcm_dpo/beta": 0.0038559194654226303, "fcm_dpo/delta": -0.21507573127746582, "fcm_dpo/margin": 207.04443359375, "fcm_dpo/q_t": 0.32492557168006897, "grad_norm": 30.406768798828125, "learning_rate": 3.2456423598071783e-07, "logits/chosen": -0.3299999237060547, "logits/rejected": -0.3157616853713989, "logps/chosen": -272.45648193359375, "logps/ref_chosen": -56.094207763671875, "logps/ref_rejected": -100.69905090332031, "logps/rejected": -524.105712890625, "loss": 0.8408, "margin_dpo/margin_mean": 207.04443359375, "margin_dpo/margin_std": 174.73843383789062, "step": 317 }, { "epoch": 0.4669603524229075, "fcm_dpo/beta": 0.0038027470000088215, "fcm_dpo/delta": -0.02990717813372612, "fcm_dpo/margin": 165.0573272705078, "fcm_dpo/q_t": 0.3614851236343384, "grad_norm": 39.84989929199219, "learning_rate": 3.233383385962115e-07, "logits/chosen": -0.39633870124816895, "logits/rejected": -0.3627668023109436, "logps/chosen": -288.56982421875, "logps/ref_chosen": -64.64569854736328, "logps/ref_rejected": -82.76425170898438, "logps/rejected": -471.7456970214844, "loss": 0.9456, "margin_dpo/margin_mean": 165.0573272705078, "margin_dpo/margin_std": 168.24526977539062, "step": 318 }, { "epoch": 0.4684287812041116, "fcm_dpo/beta": 0.0037455919664353132, "fcm_dpo/delta": -0.06087436154484749, "fcm_dpo/margin": 175.37954711914062, "fcm_dpo/q_t": 0.35514286160469055, "grad_norm": 32.62771987915039, "learning_rate": 3.2211050867827805e-07, "logits/chosen": -0.32384127378463745, "logits/rejected": -0.33828234672546387, "logps/chosen": -260.467529296875, "logps/ref_chosen": -49.383758544921875, "logps/ref_rejected": -113.90650939941406, "logps/rejected": -500.3698425292969, "loss": 0.9384, "margin_dpo/margin_mean": 175.37954711914062, "margin_dpo/margin_std": 184.76806640625, "step": 319 }, { "epoch": 0.4698972099853157, "fcm_dpo/beta": 0.0036399452947080135, "fcm_dpo/delta": -0.14344662427902222, "fcm_dpo/margin": 201.1544189453125, "fcm_dpo/q_t": 0.3414286971092224, "grad_norm": 35.252044677734375, "learning_rate": 3.208807785813777e-07, "logits/chosen": -0.34068095684051514, "logits/rejected": -0.34001022577285767, "logps/chosen": -272.8780212402344, "logps/ref_chosen": -59.50489044189453, "logps/ref_rejected": -97.66717529296875, "logps/rejected": -512.1947021484375, "loss": 0.8862, "margin_dpo/margin_mean": 201.1544189453125, "margin_dpo/margin_std": 194.9528045654297, "step": 320 }, { "epoch": 0.4713656387665198, "fcm_dpo/beta": 0.0035794866271317005, "fcm_dpo/delta": -0.025970421731472015, "fcm_dpo/margin": 173.90859985351562, "fcm_dpo/q_t": 0.3651740550994873, "grad_norm": 39.09996795654297, "learning_rate": 3.1964918071004217e-07, "logits/chosen": -0.3051578104496002, "logits/rejected": -0.29157111048698425, "logps/chosen": -323.6648254394531, "logps/ref_chosen": -61.548683166503906, "logps/ref_rejected": -91.64103698730469, "logps/rejected": -527.665771484375, "loss": 0.9826, "margin_dpo/margin_mean": 173.90859985351562, "margin_dpo/margin_std": 204.27359008789062, "step": 321 }, { "epoch": 0.47283406754772395, "fcm_dpo/beta": 0.003550457302480936, "fcm_dpo/delta": -0.07277655601501465, "fcm_dpo/margin": 187.88018798828125, "fcm_dpo/q_t": 0.3536107838153839, "grad_norm": 34.10807800292969, "learning_rate": 3.184157475180207e-07, "logits/chosen": -0.3065683841705322, "logits/rejected": -0.300385057926178, "logps/chosen": -284.23797607421875, "logps/ref_chosen": -57.29003143310547, "logps/ref_rejected": -95.74992370605469, "logps/rejected": -510.5780334472656, "loss": 0.9286, "margin_dpo/margin_mean": 187.88018798828125, "margin_dpo/margin_std": 189.9258270263672, "step": 322 }, { "epoch": 0.47430249632892807, "fcm_dpo/beta": 0.003559872740879655, "fcm_dpo/delta": 0.03238154202699661, "fcm_dpo/margin": 159.92843627929688, "fcm_dpo/q_t": 0.374622106552124, "grad_norm": 46.58732604980469, "learning_rate": 3.171805115074251e-07, "logits/chosen": -0.3481108546257019, "logits/rejected": -0.343014121055603, "logps/chosen": -301.1619873046875, "logps/ref_chosen": -51.23395919799805, "logps/ref_rejected": -75.06192016601562, "logps/rejected": -484.91839599609375, "loss": 1.0045, "margin_dpo/margin_mean": 159.92843627929688, "margin_dpo/margin_std": 189.8189697265625, "step": 323 }, { "epoch": 0.47577092511013214, "fcm_dpo/beta": 0.003606216050684452, "fcm_dpo/delta": -0.016858771443367004, "fcm_dpo/margin": 170.2562713623047, "fcm_dpo/q_t": 0.3706227242946625, "grad_norm": 61.91301345825195, "learning_rate": 3.1594350522787295e-07, "logits/chosen": -0.3303983509540558, "logits/rejected": -0.3054184913635254, "logps/chosen": -341.5509033203125, "logps/ref_chosen": -65.13516998291016, "logps/ref_rejected": -86.47750854492188, "logps/rejected": -533.1495361328125, "loss": 1.0022, "margin_dpo/margin_mean": 170.2562713623047, "margin_dpo/margin_std": 211.9681396484375, "step": 324 }, { "epoch": 0.47723935389133626, "fcm_dpo/beta": 0.003626979887485504, "fcm_dpo/delta": 0.10734298825263977, "fcm_dpo/margin": 137.27645874023438, "fcm_dpo/q_t": 0.38791322708129883, "grad_norm": 33.29164123535156, "learning_rate": 3.147047612756302e-07, "logits/chosen": -0.4132434129714966, "logits/rejected": -0.3767145872116089, "logps/chosen": -273.2174377441406, "logps/ref_chosen": -56.215599060058594, "logps/ref_rejected": -70.08592987060547, "logps/rejected": -424.3642272949219, "loss": 1.0305, "margin_dpo/margin_mean": 137.27645874023438, "margin_dpo/margin_std": 158.93292236328125, "step": 325 }, { "epoch": 0.4787077826725404, "fcm_dpo/beta": 0.0036813640035688877, "fcm_dpo/delta": 0.08160503208637238, "fcm_dpo/margin": 141.9955596923828, "fcm_dpo/q_t": 0.3813520669937134, "grad_norm": 49.46158981323242, "learning_rate": 3.134643122927519e-07, "logits/chosen": -0.42192065715789795, "logits/rejected": -0.38546887040138245, "logps/chosen": -309.13665771484375, "logps/ref_chosen": -72.72496032714844, "logps/ref_rejected": -79.8467788696289, "logps/rejected": -458.25408935546875, "loss": 0.9932, "margin_dpo/margin_mean": 141.9955596923828, "margin_dpo/margin_std": 141.18475341796875, "step": 326 }, { "epoch": 0.4801762114537445, "fcm_dpo/beta": 0.0036152596585452557, "fcm_dpo/delta": -0.12888285517692566, "fcm_dpo/margin": 198.74432373046875, "fcm_dpo/q_t": 0.34069618582725525, "grad_norm": 56.69047546386719, "learning_rate": 3.1222219096622264e-07, "logits/chosen": -0.3953353762626648, "logits/rejected": -0.3728201389312744, "logps/chosen": -278.8756103515625, "logps/ref_chosen": -69.13441467285156, "logps/ref_rejected": -111.93377685546875, "logps/rejected": -520.4193115234375, "loss": 0.8922, "margin_dpo/margin_mean": 198.74432373046875, "margin_dpo/margin_std": 184.82196044921875, "step": 327 }, { "epoch": 0.48164464023494863, "fcm_dpo/beta": 0.0036046169698238373, "fcm_dpo/delta": -0.01882966235280037, "fcm_dpo/margin": 171.35028076171875, "fcm_dpo/q_t": 0.3656262159347534, "grad_norm": 41.368961334228516, "learning_rate": 3.1097843002709427e-07, "logits/chosen": -0.3795957565307617, "logits/rejected": -0.3735014796257019, "logps/chosen": -278.96380615234375, "logps/ref_chosen": -59.68719482421875, "logps/ref_rejected": -90.85499572753906, "logps/rejected": -481.48187255859375, "loss": 0.963, "margin_dpo/margin_mean": 171.35028076171875, "margin_dpo/margin_std": 190.85531616210938, "step": 328 }, { "epoch": 0.4831130690161527, "fcm_dpo/beta": 0.0035419906489551067, "fcm_dpo/delta": -0.06768238544464111, "fcm_dpo/margin": 187.00729370117188, "fcm_dpo/q_t": 0.3545988202095032, "grad_norm": 28.033584594726562, "learning_rate": 3.0973306224962437e-07, "logits/chosen": -0.35659074783325195, "logits/rejected": -0.33611446619033813, "logps/chosen": -300.49114990234375, "logps/ref_chosen": -65.2461929321289, "logps/ref_rejected": -100.69770812988281, "logps/rejected": -522.949951171875, "loss": 0.9397, "margin_dpo/margin_mean": 187.00729370117188, "margin_dpo/margin_std": 195.09939575195312, "step": 329 }, { "epoch": 0.4845814977973568, "fcm_dpo/beta": 0.0035197772085666656, "fcm_dpo/delta": -0.04879715293645859, "fcm_dpo/margin": 183.4234619140625, "fcm_dpo/q_t": 0.3568430542945862, "grad_norm": 40.50788116455078, "learning_rate": 3.084861204504122e-07, "logits/chosen": -0.3393748104572296, "logits/rejected": -0.34215378761291504, "logps/chosen": -263.3137512207031, "logps/ref_chosen": -46.998348236083984, "logps/ref_rejected": -86.87684631347656, "logps/rejected": -486.61572265625, "loss": 0.9342, "margin_dpo/margin_mean": 183.4234619140625, "margin_dpo/margin_std": 184.44473266601562, "step": 330 }, { "epoch": 0.48604992657856094, "fcm_dpo/beta": 0.003468625945970416, "fcm_dpo/delta": -0.08099709451198578, "fcm_dpo/margin": 194.73370361328125, "fcm_dpo/q_t": 0.34509968757629395, "grad_norm": 34.197383880615234, "learning_rate": 3.072376374875335e-07, "logits/chosen": -0.33651411533355713, "logits/rejected": -0.32489413022994995, "logps/chosen": -278.44921875, "logps/ref_chosen": -50.52424621582031, "logps/ref_rejected": -89.01544189453125, "logps/rejected": -511.67413330078125, "loss": 0.8733, "margin_dpo/margin_mean": 194.73370361328125, "margin_dpo/margin_std": 140.58489990234375, "step": 331 }, { "epoch": 0.48751835535976507, "fcm_dpo/beta": 0.003518771380186081, "fcm_dpo/delta": 0.11069121956825256, "fcm_dpo/margin": 140.563232421875, "fcm_dpo/q_t": 0.38987618684768677, "grad_norm": 32.90526580810547, "learning_rate": 3.059876462596758e-07, "logits/chosen": -0.3346155285835266, "logits/rejected": -0.3171945810317993, "logps/chosen": -291.15472412109375, "logps/ref_chosen": -49.18028259277344, "logps/ref_rejected": -76.48515319824219, "logps/rejected": -459.0228271484375, "loss": 1.029, "margin_dpo/margin_mean": 140.563232421875, "margin_dpo/margin_std": 166.2570037841797, "step": 332 }, { "epoch": 0.4889867841409692, "fcm_dpo/beta": 0.0034761871211230755, "fcm_dpo/delta": -0.055231112986803055, "fcm_dpo/margin": 186.94793701171875, "fcm_dpo/q_t": 0.3612428307533264, "grad_norm": 31.33414649963379, "learning_rate": 3.0473617970527015e-07, "logits/chosen": -0.37555959820747375, "logits/rejected": -0.36939913034439087, "logps/chosen": -315.4451904296875, "logps/ref_chosen": -63.75574493408203, "logps/ref_rejected": -95.04411315917969, "logps/rejected": -533.6814575195312, "loss": 0.9718, "margin_dpo/margin_mean": 186.94793701171875, "margin_dpo/margin_std": 219.6304931640625, "step": 333 }, { "epoch": 0.49045521292217326, "fcm_dpo/beta": 0.003508550813421607, "fcm_dpo/delta": 0.022331170737743378, "fcm_dpo/margin": 165.01031494140625, "fcm_dpo/q_t": 0.3745434284210205, "grad_norm": 45.59312438964844, "learning_rate": 3.034832708016243e-07, "logits/chosen": -0.37708473205566406, "logits/rejected": -0.37125474214553833, "logps/chosen": -313.27618408203125, "logps/ref_chosen": -66.97975158691406, "logps/ref_rejected": -95.31692504882812, "logps/rejected": -506.6236877441406, "loss": 1.0114, "margin_dpo/margin_mean": 165.01031494140625, "margin_dpo/margin_std": 209.47122192382812, "step": 334 }, { "epoch": 0.4919236417033774, "fcm_dpo/beta": 0.003607717342674732, "fcm_dpo/delta": 0.14469006657600403, "fcm_dpo/margin": 127.66055297851562, "fcm_dpo/q_t": 0.4004567563533783, "grad_norm": 63.05012512207031, "learning_rate": 3.022289525640531e-07, "logits/chosen": -0.38800978660583496, "logits/rejected": -0.36463358998298645, "logps/chosen": -334.33935546875, "logps/ref_chosen": -62.54248046875, "logps/ref_rejected": -87.61770629882812, "logps/rejected": -487.07513427734375, "loss": 1.0921, "margin_dpo/margin_mean": 127.66055297851562, "margin_dpo/margin_std": 194.76068115234375, "step": 335 }, { "epoch": 0.4933920704845815, "fcm_dpo/beta": 0.0035587395541369915, "fcm_dpo/delta": -0.1215284913778305, "fcm_dpo/margin": 200.35911560058594, "fcm_dpo/q_t": 0.34965717792510986, "grad_norm": 34.31383514404297, "learning_rate": 3.009732580450086e-07, "logits/chosen": -0.3676879405975342, "logits/rejected": -0.36403754353523254, "logps/chosen": -313.09222412109375, "logps/ref_chosen": -54.53115463256836, "logps/ref_rejected": -104.40424346923828, "logps/rejected": -563.3244018554688, "loss": 0.9395, "margin_dpo/margin_mean": 200.35910034179688, "margin_dpo/margin_std": 232.01760864257812, "step": 336 }, { "epoch": 0.4948604992657856, "fcm_dpo/beta": 0.003459386760368943, "fcm_dpo/delta": -0.11083994805812836, "fcm_dpo/margin": 203.0913848876953, "fcm_dpo/q_t": 0.34715068340301514, "grad_norm": 32.61246109008789, "learning_rate": 2.9971622033320914e-07, "logits/chosen": -0.3618990480899811, "logits/rejected": -0.3507819175720215, "logps/chosen": -311.53759765625, "logps/ref_chosen": -65.12869262695312, "logps/ref_rejected": -101.72701263427734, "logps/rejected": -551.2273559570312, "loss": 0.9156, "margin_dpo/margin_mean": 203.09136962890625, "margin_dpo/margin_std": 209.97573852539062, "step": 337 }, { "epoch": 0.49632892804698975, "fcm_dpo/beta": 0.0034060748293995857, "fcm_dpo/delta": -0.08523920178413391, "fcm_dpo/margin": 199.4775848388672, "fcm_dpo/q_t": 0.34894174337387085, "grad_norm": 29.552736282348633, "learning_rate": 2.984578725527675e-07, "logits/chosen": -0.4156501591205597, "logits/rejected": -0.40091991424560547, "logps/chosen": -276.4804382324219, "logps/ref_chosen": -58.422706604003906, "logps/ref_rejected": -89.06854248046875, "logps/rejected": -506.6038513183594, "loss": 0.9049, "margin_dpo/margin_mean": 199.4775848388672, "margin_dpo/margin_std": 184.65594482421875, "step": 338 }, { "epoch": 0.4977973568281938, "fcm_dpo/beta": 0.0033988947980105877, "fcm_dpo/delta": -0.04927036911249161, "fcm_dpo/margin": 189.7136993408203, "fcm_dpo/q_t": 0.35729217529296875, "grad_norm": 32.228424072265625, "learning_rate": 2.9719824786231796e-07, "logits/chosen": -0.40810322761535645, "logits/rejected": -0.3912397623062134, "logps/chosen": -287.6553649902344, "logps/ref_chosen": -59.99531555175781, "logps/ref_rejected": -103.9109115600586, "logps/rejected": -521.28466796875, "loss": 0.9323, "margin_dpo/margin_mean": 189.7136993408203, "margin_dpo/margin_std": 180.0070343017578, "step": 339 }, { "epoch": 0.49926578560939794, "fcm_dpo/beta": 0.0033571631647646427, "fcm_dpo/delta": 0.030759550631046295, "fcm_dpo/margin": 169.98486328125, "fcm_dpo/q_t": 0.3748738765716553, "grad_norm": 34.929901123046875, "learning_rate": 2.959373794541426e-07, "logits/chosen": -0.3299640417098999, "logits/rejected": -0.3012125492095947, "logps/chosen": -285.945556640625, "logps/ref_chosen": -52.83022689819336, "logps/ref_rejected": -73.10723114013672, "logps/rejected": -476.2073974609375, "loss": 0.9978, "margin_dpo/margin_mean": 169.98486328125, "margin_dpo/margin_std": 201.46371459960938, "step": 340 }, { "epoch": 0.5007342143906021, "fcm_dpo/beta": 0.0033384612761437893, "fcm_dpo/delta": -0.0509408563375473, "fcm_dpo/margin": 193.81198120117188, "fcm_dpo/q_t": 0.3578472137451172, "grad_norm": 29.954818725585938, "learning_rate": 2.946753005532965e-07, "logits/chosen": -0.358737051486969, "logits/rejected": -0.3582429587841034, "logps/chosen": -274.4048767089844, "logps/ref_chosen": -47.899803161621094, "logps/ref_rejected": -101.80987548828125, "logps/rejected": -522.126953125, "loss": 0.9392, "margin_dpo/margin_mean": 193.81198120117188, "margin_dpo/margin_std": 198.76821899414062, "step": 341 }, { "epoch": 0.5022026431718062, "fcm_dpo/beta": 0.003372443839907646, "fcm_dpo/delta": 0.02195318415760994, "fcm_dpo/margin": 171.60679626464844, "fcm_dpo/q_t": 0.37457433342933655, "grad_norm": 34.212318420410156, "learning_rate": 2.934120444167326e-07, "logits/chosen": -0.38501685857772827, "logits/rejected": -0.35319679975509644, "logps/chosen": -302.15545654296875, "logps/ref_chosen": -71.99664306640625, "logps/ref_rejected": -92.58959197998047, "logps/rejected": -494.355224609375, "loss": 0.9967, "margin_dpo/margin_mean": 171.60679626464844, "margin_dpo/margin_std": 208.8232421875, "step": 342 }, { "epoch": 0.5036710719530103, "fcm_dpo/beta": 0.0033103451132774353, "fcm_dpo/delta": -0.0885283350944519, "fcm_dpo/margin": 206.1126708984375, "fcm_dpo/q_t": 0.3450053334236145, "grad_norm": 34.08480453491211, "learning_rate": 2.9214764433242476e-07, "logits/chosen": -0.3982047438621521, "logits/rejected": -0.3942341208457947, "logps/chosen": -267.8955383300781, "logps/ref_chosen": -54.405616760253906, "logps/ref_rejected": -111.04142761230469, "logps/rejected": -530.64404296875, "loss": 0.8828, "margin_dpo/margin_mean": 206.1126708984375, "margin_dpo/margin_std": 162.10682678222656, "step": 343 }, { "epoch": 0.5051395007342144, "fcm_dpo/beta": 0.0033293329179286957, "fcm_dpo/delta": -0.014567896723747253, "fcm_dpo/margin": 183.69131469726562, "fcm_dpo/q_t": 0.3665677309036255, "grad_norm": 27.15936851501465, "learning_rate": 2.9088213361849126e-07, "logits/chosen": -0.38978269696235657, "logits/rejected": -0.3856002688407898, "logps/chosen": -273.5009460449219, "logps/ref_chosen": -53.96466827392578, "logps/ref_rejected": -90.62336730957031, "logps/rejected": -493.8509521484375, "loss": 0.9563, "margin_dpo/margin_mean": 183.69131469726562, "margin_dpo/margin_std": 185.71316528320312, "step": 344 }, { "epoch": 0.5066079295154186, "fcm_dpo/beta": 0.0032876995392143726, "fcm_dpo/delta": -0.037146538496017456, "fcm_dpo/margin": 193.00808715820312, "fcm_dpo/q_t": 0.36083394289016724, "grad_norm": 27.978347778320312, "learning_rate": 2.896155456223163e-07, "logits/chosen": -0.39028286933898926, "logits/rejected": -0.3786432147026062, "logps/chosen": -334.9024963378906, "logps/ref_chosen": -61.685699462890625, "logps/ref_rejected": -99.49041748046875, "logps/rejected": -565.7152709960938, "loss": 0.952, "margin_dpo/margin_mean": 193.00807189941406, "margin_dpo/margin_std": 207.6051025390625, "step": 345 }, { "epoch": 0.5080763582966226, "fcm_dpo/beta": 0.003262849524617195, "fcm_dpo/delta": -0.004547563381493092, "fcm_dpo/margin": 185.1959686279297, "fcm_dpo/q_t": 0.3652818500995636, "grad_norm": 34.94870376586914, "learning_rate": 2.883479137196714e-07, "logits/chosen": -0.3476218581199646, "logits/rejected": -0.3306209444999695, "logps/chosen": -337.10003662109375, "logps/ref_chosen": -55.256263732910156, "logps/ref_rejected": -77.41532135009766, "logps/rejected": -544.455078125, "loss": 0.9697, "margin_dpo/margin_mean": 185.1959686279297, "margin_dpo/margin_std": 203.53085327148438, "step": 346 }, { "epoch": 0.5095447870778267, "fcm_dpo/beta": 0.0032533821649849415, "fcm_dpo/delta": -0.015425489284098148, "fcm_dpo/margin": 188.865478515625, "fcm_dpo/q_t": 0.3658721446990967, "grad_norm": 38.90462112426758, "learning_rate": 2.8707927131383614e-07, "logits/chosen": -0.3579314947128296, "logits/rejected": -0.3465426564216614, "logps/chosen": -343.6216125488281, "logps/ref_chosen": -57.56623840332031, "logps/ref_rejected": -92.35509490966797, "logps/rejected": -567.2759399414062, "loss": 0.9762, "margin_dpo/margin_mean": 188.865478515625, "margin_dpo/margin_std": 218.76934814453125, "step": 347 }, { "epoch": 0.5110132158590308, "fcm_dpo/beta": 0.0032581479754298925, "fcm_dpo/delta": 0.0629473477602005, "fcm_dpo/margin": 165.63992309570312, "fcm_dpo/q_t": 0.381051242351532, "grad_norm": 30.090953826904297, "learning_rate": 2.858096518347179e-07, "logits/chosen": -0.4058574438095093, "logits/rejected": -0.4066629707813263, "logps/chosen": -306.64599609375, "logps/ref_chosen": -56.31770324707031, "logps/ref_rejected": -89.13836669921875, "logps/rejected": -505.1065673828125, "loss": 1.0098, "margin_dpo/margin_mean": 165.63992309570312, "margin_dpo/margin_std": 191.34750366210938, "step": 348 }, { "epoch": 0.5124816446402349, "fcm_dpo/beta": 0.003329505678266287, "fcm_dpo/delta": 0.02792828157544136, "fcm_dpo/margin": 172.01544189453125, "fcm_dpo/q_t": 0.3782421946525574, "grad_norm": 26.204023361206055, "learning_rate": 2.845390887379706e-07, "logits/chosen": -0.3858356475830078, "logits/rejected": -0.3851817548274994, "logps/chosen": -302.1793212890625, "logps/ref_chosen": -58.025516510009766, "logps/ref_rejected": -97.50515747070312, "logps/rejected": -513.6744384765625, "loss": 1.0217, "margin_dpo/margin_mean": 172.01544189453125, "margin_dpo/margin_std": 228.81277465820312, "step": 349 }, { "epoch": 0.5139500734214391, "fcm_dpo/beta": 0.003294072113931179, "fcm_dpo/delta": -0.006476100534200668, "fcm_dpo/margin": 183.84515380859375, "fcm_dpo/q_t": 0.36599087715148926, "grad_norm": 39.012115478515625, "learning_rate": 2.8326761550411346e-07, "logits/chosen": -0.4358096122741699, "logits/rejected": -0.43361157178878784, "logps/chosen": -314.4185485839844, "logps/ref_chosen": -64.33049011230469, "logps/ref_rejected": -89.87164306640625, "logps/rejected": -523.8048706054688, "loss": 0.9903, "margin_dpo/margin_mean": 183.84518432617188, "margin_dpo/margin_std": 221.12252807617188, "step": 350 }, { "epoch": 0.5154185022026432, "fcm_dpo/beta": 0.00327136623673141, "fcm_dpo/delta": -0.07544407248497009, "fcm_dpo/margin": 204.92010498046875, "fcm_dpo/q_t": 0.355930358171463, "grad_norm": 33.651607513427734, "learning_rate": 2.819952656376487e-07, "logits/chosen": -0.40549007058143616, "logits/rejected": -0.3955671489238739, "logps/chosen": -296.75994873046875, "logps/ref_chosen": -60.6721305847168, "logps/ref_rejected": -101.5654296875, "logps/rejected": -542.5733642578125, "loss": 0.9437, "margin_dpo/margin_mean": 204.92010498046875, "margin_dpo/margin_std": 227.87014770507812, "step": 351 }, { "epoch": 0.5168869309838473, "fcm_dpo/beta": 0.003319011302664876, "fcm_dpo/delta": 0.15692806243896484, "fcm_dpo/margin": 135.69808959960938, "fcm_dpo/q_t": 0.40290194749832153, "grad_norm": 41.35541915893555, "learning_rate": 2.8072207266617854e-07, "logits/chosen": -0.39126765727996826, "logits/rejected": -0.35691601037979126, "logps/chosen": -348.217041015625, "logps/ref_chosen": -70.9434585571289, "logps/ref_rejected": -76.6419677734375, "logps/rejected": -489.6136169433594, "loss": 1.0962, "margin_dpo/margin_mean": 135.69808959960938, "margin_dpo/margin_std": 207.16409301757812, "step": 352 }, { "epoch": 0.5183553597650514, "fcm_dpo/beta": 0.003364278469234705, "fcm_dpo/delta": 0.043343670666217804, "fcm_dpo/margin": 166.03475952148438, "fcm_dpo/q_t": 0.3799568712711334, "grad_norm": 41.904701232910156, "learning_rate": 2.794480701395219e-07, "logits/chosen": -0.43321138620376587, "logits/rejected": -0.41663485765457153, "logps/chosen": -319.3279113769531, "logps/ref_chosen": -58.39533996582031, "logps/ref_rejected": -80.33553314208984, "logps/rejected": -507.3028259277344, "loss": 1.0207, "margin_dpo/margin_mean": 166.03475952148438, "margin_dpo/margin_std": 212.48361206054688, "step": 353 }, { "epoch": 0.5198237885462555, "fcm_dpo/beta": 0.0033673213329166174, "fcm_dpo/delta": -0.036777447909116745, "fcm_dpo/margin": 188.36825561523438, "fcm_dpo/q_t": 0.3579559922218323, "grad_norm": 39.02664566040039, "learning_rate": 2.781732916288303e-07, "logits/chosen": -0.4032512307167053, "logits/rejected": -0.383810818195343, "logps/chosen": -275.3306579589844, "logps/ref_chosen": -59.80299377441406, "logps/ref_rejected": -88.75750732421875, "logps/rejected": -492.6534423828125, "loss": 0.9294, "margin_dpo/margin_mean": 188.36827087402344, "margin_dpo/margin_std": 177.9414520263672, "step": 354 }, { "epoch": 0.5212922173274597, "fcm_dpo/beta": 0.003320076037198305, "fcm_dpo/delta": -0.06938491016626358, "fcm_dpo/margin": 200.06072998046875, "fcm_dpo/q_t": 0.3513967990875244, "grad_norm": 39.97414779663086, "learning_rate": 2.7689777072570284e-07, "logits/chosen": -0.4452924132347107, "logits/rejected": -0.42499247193336487, "logps/chosen": -272.80706787109375, "logps/ref_chosen": -54.12849807739258, "logps/ref_rejected": -82.40606689453125, "logps/rejected": -501.1453552246094, "loss": 0.9087, "margin_dpo/margin_mean": 200.06072998046875, "margin_dpo/margin_std": 179.74571228027344, "step": 355 }, { "epoch": 0.5227606461086637, "fcm_dpo/beta": 0.0034437617287039757, "fcm_dpo/delta": 0.1814279407262802, "fcm_dpo/margin": 122.82986450195312, "fcm_dpo/q_t": 0.41210052371025085, "grad_norm": 50.92268753051758, "learning_rate": 2.7562154104130176e-07, "logits/chosen": -0.4147285223007202, "logits/rejected": -0.39140522480010986, "logps/chosen": -330.02154541015625, "logps/ref_chosen": -64.6738052368164, "logps/ref_rejected": -75.89926147460938, "logps/rejected": -464.07684326171875, "loss": 1.1595, "margin_dpo/margin_mean": 122.82986450195312, "margin_dpo/margin_std": 233.60147094726562, "step": 356 }, { "epoch": 0.5242290748898678, "fcm_dpo/beta": 0.0034146038815379143, "fcm_dpo/delta": -0.019891854375600815, "fcm_dpo/margin": 180.96995544433594, "fcm_dpo/q_t": 0.3629704713821411, "grad_norm": 36.10367202758789, "learning_rate": 2.7434463620546594e-07, "logits/chosen": -0.4508202075958252, "logits/rejected": -0.43361175060272217, "logps/chosen": -294.203857421875, "logps/ref_chosen": -52.725799560546875, "logps/ref_rejected": -86.84115600585938, "logps/rejected": -509.2891845703125, "loss": 0.9528, "margin_dpo/margin_mean": 180.96995544433594, "margin_dpo/margin_std": 186.50103759765625, "step": 357 }, { "epoch": 0.5256975036710719, "fcm_dpo/beta": 0.003465921152383089, "fcm_dpo/delta": 0.06651946902275085, "fcm_dpo/margin": 154.9437255859375, "fcm_dpo/q_t": 0.3825710117816925, "grad_norm": 31.915424346923828, "learning_rate": 2.730670898658255e-07, "logits/chosen": -0.4602763056755066, "logits/rejected": -0.44098126888275146, "logps/chosen": -284.6385192871094, "logps/ref_chosen": -63.20543670654297, "logps/ref_rejected": -88.373291015625, "logps/rejected": -464.7501220703125, "loss": 1.0108, "margin_dpo/margin_mean": 154.9437255859375, "margin_dpo/margin_std": 186.69895935058594, "step": 358 }, { "epoch": 0.527165932452276, "fcm_dpo/beta": 0.003439284861087799, "fcm_dpo/delta": -0.05646362528204918, "fcm_dpo/margin": 189.73446655273438, "fcm_dpo/q_t": 0.36093398928642273, "grad_norm": 38.223670959472656, "learning_rate": 2.717889356869146e-07, "logits/chosen": -0.39621424674987793, "logits/rejected": -0.3747130036354065, "logps/chosen": -310.32684326171875, "logps/ref_chosen": -56.370216369628906, "logps/ref_rejected": -82.17375183105469, "logps/rejected": -525.8648681640625, "loss": 0.9534, "margin_dpo/margin_mean": 189.73448181152344, "margin_dpo/margin_std": 211.576416015625, "step": 359 }, { "epoch": 0.5286343612334802, "fcm_dpo/beta": 0.0034922873601317406, "fcm_dpo/delta": 0.09579764306545258, "fcm_dpo/margin": 145.714599609375, "fcm_dpo/q_t": 0.3853462338447571, "grad_norm": 46.432003021240234, "learning_rate": 2.7051020734928443e-07, "logits/chosen": -0.37489083409309387, "logits/rejected": -0.3580286502838135, "logps/chosen": -298.4937744140625, "logps/ref_chosen": -51.460384368896484, "logps/ref_rejected": -69.83892059326172, "logps/rejected": -462.5869445800781, "loss": 1.0106, "margin_dpo/margin_mean": 145.714599609375, "margin_dpo/margin_std": 157.42027282714844, "step": 360 }, { "epoch": 0.5301027900146843, "fcm_dpo/beta": 0.0035505923442542553, "fcm_dpo/delta": 0.06202498450875282, "fcm_dpo/margin": 152.30416870117188, "fcm_dpo/q_t": 0.3825526237487793, "grad_norm": 37.47382736206055, "learning_rate": 2.6923093854861593e-07, "logits/chosen": -0.3717784285545349, "logits/rejected": -0.36584901809692383, "logps/chosen": -310.5164794921875, "logps/ref_chosen": -53.86951446533203, "logps/ref_rejected": -90.7692642211914, "logps/rejected": -499.7203674316406, "loss": 1.0306, "margin_dpo/margin_mean": 152.30416870117188, "margin_dpo/margin_std": 200.70108032226562, "step": 361 }, { "epoch": 0.5315712187958884, "fcm_dpo/beta": 0.0034302864223718643, "fcm_dpo/delta": -0.19361282885074615, "fcm_dpo/margin": 226.57215881347656, "fcm_dpo/q_t": 0.3321119248867035, "grad_norm": 34.013301849365234, "learning_rate": 2.679511629948319e-07, "logits/chosen": -0.4100894331932068, "logits/rejected": -0.4160076677799225, "logps/chosen": -279.9183654785156, "logps/ref_chosen": -58.639060974121094, "logps/ref_rejected": -105.58195495605469, "logps/rejected": -553.4334106445312, "loss": 0.8572, "margin_dpo/margin_mean": 226.57215881347656, "margin_dpo/margin_std": 208.89010620117188, "step": 362 }, { "epoch": 0.5330396475770925, "fcm_dpo/beta": 0.003321536350995302, "fcm_dpo/delta": -0.20300859212875366, "fcm_dpo/margin": 237.05946350097656, "fcm_dpo/q_t": 0.32788795232772827, "grad_norm": 25.153287887573242, "learning_rate": 2.6667091441120816e-07, "logits/chosen": -0.3687742054462433, "logits/rejected": -0.3484545350074768, "logps/chosen": -246.83908081054688, "logps/ref_chosen": -44.558380126953125, "logps/ref_rejected": -74.69496154785156, "logps/rejected": -514.03515625, "loss": 0.8467, "margin_dpo/margin_mean": 237.0594482421875, "margin_dpo/margin_std": 207.8511962890625, "step": 363 }, { "epoch": 0.5345080763582967, "fcm_dpo/beta": 0.003306190250441432, "fcm_dpo/delta": 0.02047722041606903, "fcm_dpo/margin": 175.40988159179688, "fcm_dpo/q_t": 0.37260884046554565, "grad_norm": 35.79054641723633, "learning_rate": 2.6539022653348575e-07, "logits/chosen": -0.40862590074539185, "logits/rejected": -0.41578084230422974, "logps/chosen": -270.4530944824219, "logps/ref_chosen": -48.894622802734375, "logps/ref_rejected": -91.395751953125, "logps/rejected": -488.36407470703125, "loss": 0.9919, "margin_dpo/margin_mean": 175.40988159179688, "margin_dpo/margin_std": 205.0936737060547, "step": 364 }, { "epoch": 0.5359765051395007, "fcm_dpo/beta": 0.0033004698343575, "fcm_dpo/delta": 0.016225282102823257, "fcm_dpo/margin": 177.1582794189453, "fcm_dpo/q_t": 0.3734648823738098, "grad_norm": 36.35362243652344, "learning_rate": 2.641091331089811e-07, "logits/chosen": -0.38673701882362366, "logits/rejected": -0.3943500518798828, "logps/chosen": -276.78192138671875, "logps/ref_chosen": -51.49274444580078, "logps/ref_rejected": -92.70166778564453, "logps/rejected": -495.1491394042969, "loss": 0.9778, "margin_dpo/margin_mean": 177.1582794189453, "margin_dpo/margin_std": 200.8518524169922, "step": 365 }, { "epoch": 0.5374449339207048, "fcm_dpo/beta": 0.0032841768115758896, "fcm_dpo/delta": 0.0016945749521255493, "fcm_dpo/margin": 182.0388946533203, "fcm_dpo/q_t": 0.3690670430660248, "grad_norm": 28.885610580444336, "learning_rate": 2.6282766789569736e-07, "logits/chosen": -0.36415767669677734, "logits/rejected": -0.37583303451538086, "logps/chosen": -248.95938110351562, "logps/ref_chosen": -44.7205696105957, "logps/ref_rejected": -83.31040954589844, "logps/rejected": -469.588134765625, "loss": 0.9792, "margin_dpo/margin_mean": 182.0388946533203, "margin_dpo/margin_std": 206.19607543945312, "step": 366 }, { "epoch": 0.5389133627019089, "fcm_dpo/beta": 0.0033485644962638617, "fcm_dpo/delta": 0.08677735924720764, "fcm_dpo/margin": 154.576416015625, "fcm_dpo/q_t": 0.3839992582798004, "grad_norm": 24.783180236816406, "learning_rate": 2.615458646614349e-07, "logits/chosen": -0.39714837074279785, "logits/rejected": -0.37421467900276184, "logps/chosen": -266.13519287109375, "logps/ref_chosen": -58.405418395996094, "logps/ref_rejected": -76.75132751464844, "logps/rejected": -439.0574951171875, "loss": 1.016, "margin_dpo/margin_mean": 154.576416015625, "margin_dpo/margin_std": 178.5806884765625, "step": 367 }, { "epoch": 0.540381791483113, "fcm_dpo/beta": 0.003304690821096301, "fcm_dpo/delta": -0.12389971315860748, "fcm_dpo/margin": 216.4133758544922, "fcm_dpo/q_t": 0.33748477697372437, "grad_norm": 53.478485107421875, "learning_rate": 2.6026375718290083e-07, "logits/chosen": -0.4093177318572998, "logits/rejected": -0.4137042760848999, "logps/chosen": -243.46835327148438, "logps/ref_chosen": -44.452518463134766, "logps/ref_rejected": -98.55526733398438, "logps/rejected": -513.9844970703125, "loss": 0.8591, "margin_dpo/margin_mean": 216.4133758544922, "margin_dpo/margin_std": 163.95318603515625, "step": 368 }, { "epoch": 0.5418502202643172, "fcm_dpo/beta": 0.0033651897683739662, "fcm_dpo/delta": 0.14462685585021973, "fcm_dpo/margin": 136.97042846679688, "fcm_dpo/q_t": 0.39847463369369507, "grad_norm": 42.55653762817383, "learning_rate": 2.589813792448196e-07, "logits/chosen": -0.3915749788284302, "logits/rejected": -0.3701251447200775, "logps/chosen": -312.446533203125, "logps/ref_chosen": -71.38150024414062, "logps/ref_rejected": -91.29582214355469, "logps/rejected": -469.33123779296875, "loss": 1.083, "margin_dpo/margin_mean": 136.9704132080078, "margin_dpo/margin_std": 200.96856689453125, "step": 369 }, { "epoch": 0.5433186490455213, "fcm_dpo/beta": 0.003478851169347763, "fcm_dpo/delta": 0.18450786173343658, "fcm_dpo/margin": 121.46247863769531, "fcm_dpo/q_t": 0.40680354833602905, "grad_norm": 36.18484878540039, "learning_rate": 2.5769876463904263e-07, "logits/chosen": -0.4049254357814789, "logits/rejected": -0.39879292249679565, "logps/chosen": -330.7537841796875, "logps/ref_chosen": -71.60749816894531, "logps/ref_rejected": -97.25978088378906, "logps/rejected": -477.86859130859375, "loss": 1.1047, "margin_dpo/margin_mean": 121.46247863769531, "margin_dpo/margin_std": 185.816162109375, "step": 370 }, { "epoch": 0.5447870778267254, "fcm_dpo/beta": 0.003517968812957406, "fcm_dpo/delta": 0.012149915099143982, "fcm_dpo/margin": 167.24020385742188, "fcm_dpo/q_t": 0.37212252616882324, "grad_norm": 32.21125030517578, "learning_rate": 2.5641594716365744e-07, "logits/chosen": -0.41469210386276245, "logits/rejected": -0.39996105432510376, "logps/chosen": -320.701171875, "logps/ref_chosen": -69.41448974609375, "logps/ref_rejected": -99.17217254638672, "logps/rejected": -517.6990966796875, "loss": 1.0022, "margin_dpo/margin_mean": 167.24020385742188, "margin_dpo/margin_std": 208.5806121826172, "step": 371 }, { "epoch": 0.5462555066079295, "fcm_dpo/beta": 0.00341993011534214, "fcm_dpo/delta": -0.14982974529266357, "fcm_dpo/margin": 215.7982635498047, "fcm_dpo/q_t": 0.3453604578971863, "grad_norm": 25.910696029663086, "learning_rate": 2.551329606220976e-07, "logits/chosen": -0.4080229103565216, "logits/rejected": -0.38077276945114136, "logps/chosen": -302.6246643066406, "logps/ref_chosen": -61.8179931640625, "logps/ref_rejected": -78.53948974609375, "logps/rejected": -535.1444091796875, "loss": 0.9163, "margin_dpo/margin_mean": 215.79824829101562, "margin_dpo/margin_std": 238.89315795898438, "step": 372 }, { "epoch": 0.5477239353891337, "fcm_dpo/beta": 0.0033825889695435762, "fcm_dpo/delta": -0.09976015239953995, "fcm_dpo/margin": 204.6912841796875, "fcm_dpo/q_t": 0.3472912907600403, "grad_norm": 46.24806213378906, "learning_rate": 2.538498388222517e-07, "logits/chosen": -0.3898068368434906, "logits/rejected": -0.36087289452552795, "logps/chosen": -328.5521240234375, "logps/ref_chosen": -64.21713256835938, "logps/ref_rejected": -85.95960998535156, "logps/rejected": -554.9859619140625, "loss": 0.9146, "margin_dpo/margin_mean": 204.6912841796875, "margin_dpo/margin_std": 197.73797607421875, "step": 373 }, { "epoch": 0.5491923641703378, "fcm_dpo/beta": 0.0033088005147874355, "fcm_dpo/delta": -0.009856566786766052, "fcm_dpo/margin": 183.782958984375, "fcm_dpo/q_t": 0.3736165761947632, "grad_norm": 32.12522888183594, "learning_rate": 2.525666155755725e-07, "logits/chosen": -0.46630632877349854, "logits/rejected": -0.44315189123153687, "logps/chosen": -309.981201171875, "logps/ref_chosen": -70.65018463134766, "logps/ref_rejected": -93.64016723632812, "logps/rejected": -516.754150390625, "loss": 1.0042, "margin_dpo/margin_mean": 183.782958984375, "margin_dpo/margin_std": 240.5712890625, "step": 374 }, { "epoch": 0.5506607929515418, "fcm_dpo/beta": 0.003304019570350647, "fcm_dpo/delta": 0.007909119129180908, "fcm_dpo/margin": 178.82217407226562, "fcm_dpo/q_t": 0.3718445301055908, "grad_norm": 32.83172607421875, "learning_rate": 2.512833246961859e-07, "logits/chosen": -0.4039704203605652, "logits/rejected": -0.40034106373786926, "logps/chosen": -325.0075378417969, "logps/ref_chosen": -60.080223083496094, "logps/ref_rejected": -88.93830871582031, "logps/rejected": -532.6878051757812, "loss": 1.0158, "margin_dpo/margin_mean": 178.8221893310547, "margin_dpo/margin_std": 229.82867431640625, "step": 375 }, { "epoch": 0.5521292217327459, "fcm_dpo/beta": 0.003269658889621496, "fcm_dpo/delta": -0.13697855174541473, "fcm_dpo/margin": 222.34677124023438, "fcm_dpo/q_t": 0.3445650041103363, "grad_norm": 41.24235153198242, "learning_rate": 2.5e-07, "logits/chosen": -0.42362266778945923, "logits/rejected": -0.41283369064331055, "logps/chosen": -312.84063720703125, "logps/ref_chosen": -62.660308837890625, "logps/ref_rejected": -105.52660369873047, "logps/rejected": -578.0537109375, "loss": 0.9182, "margin_dpo/margin_mean": 222.34677124023438, "margin_dpo/margin_std": 240.65061950683594, "step": 376 }, { "epoch": 0.55359765051395, "fcm_dpo/beta": 0.0032267069909721613, "fcm_dpo/delta": -0.05978459492325783, "fcm_dpo/margin": 203.26560974121094, "fcm_dpo/q_t": 0.3565753102302551, "grad_norm": 30.29025650024414, "learning_rate": 2.487166753038141e-07, "logits/chosen": -0.37672150135040283, "logits/rejected": -0.3738042712211609, "logps/chosen": -322.83746337890625, "logps/ref_chosen": -54.478736877441406, "logps/ref_rejected": -98.70335388183594, "logps/rejected": -570.3276977539062, "loss": 0.9434, "margin_dpo/margin_mean": 203.265625, "margin_dpo/margin_std": 216.9762420654297, "step": 377 }, { "epoch": 0.5550660792951542, "fcm_dpo/beta": 0.0031417347490787506, "fcm_dpo/delta": -0.11842157691717148, "fcm_dpo/margin": 225.80239868164062, "fcm_dpo/q_t": 0.3420916795730591, "grad_norm": 36.97312545776367, "learning_rate": 2.4743338442442754e-07, "logits/chosen": -0.37459754943847656, "logits/rejected": -0.3812897205352783, "logps/chosen": -291.92535400390625, "logps/ref_chosen": -45.02053451538086, "logps/ref_rejected": -88.0469741821289, "logps/rejected": -560.754150390625, "loss": 0.8968, "margin_dpo/margin_mean": 225.80239868164062, "margin_dpo/margin_std": 211.446533203125, "step": 378 }, { "epoch": 0.5565345080763583, "fcm_dpo/beta": 0.0030931192450225353, "fcm_dpo/delta": -0.05514880642294884, "fcm_dpo/margin": 210.42945861816406, "fcm_dpo/q_t": 0.35938382148742676, "grad_norm": 45.417083740234375, "learning_rate": 2.461501611777483e-07, "logits/chosen": -0.34937772154808044, "logits/rejected": -0.36817657947540283, "logps/chosen": -330.56414794921875, "logps/ref_chosen": -53.182098388671875, "logps/ref_rejected": -114.3001708984375, "logps/rejected": -602.1116333007812, "loss": 0.9546, "margin_dpo/margin_mean": 210.429443359375, "margin_dpo/margin_std": 237.06201171875, "step": 379 }, { "epoch": 0.5580029368575624, "fcm_dpo/beta": 0.0030490856152027845, "fcm_dpo/delta": -0.1288592517375946, "fcm_dpo/margin": 236.06048583984375, "fcm_dpo/q_t": 0.34213754534721375, "grad_norm": 34.14671325683594, "learning_rate": 2.4486703937790243e-07, "logits/chosen": -0.38159555196762085, "logits/rejected": -0.4034666419029236, "logps/chosen": -325.86065673828125, "logps/ref_chosen": -51.3530387878418, "logps/ref_rejected": -104.19169616699219, "logps/rejected": -614.759765625, "loss": 0.9184, "margin_dpo/margin_mean": 236.06048583984375, "margin_dpo/margin_std": 250.18896484375, "step": 380 }, { "epoch": 0.5594713656387665, "fcm_dpo/beta": 0.0030402536503970623, "fcm_dpo/delta": 0.06344518810510635, "fcm_dpo/margin": 177.61572265625, "fcm_dpo/q_t": 0.38519516587257385, "grad_norm": 35.91572189331055, "learning_rate": 2.435840528363426e-07, "logits/chosen": -0.4085123538970947, "logits/rejected": -0.386846125125885, "logps/chosen": -332.9157409667969, "logps/ref_chosen": -57.80306625366211, "logps/ref_rejected": -79.21940612792969, "logps/rejected": -531.9478149414062, "loss": 1.0584, "margin_dpo/margin_mean": 177.61572265625, "margin_dpo/margin_std": 262.201904296875, "step": 381 }, { "epoch": 0.5609397944199707, "fcm_dpo/beta": 0.0030455852393060923, "fcm_dpo/delta": -0.024747373536229134, "fcm_dpo/margin": 204.61878967285156, "fcm_dpo/q_t": 0.3605649173259735, "grad_norm": 28.64333724975586, "learning_rate": 2.4230123536095745e-07, "logits/chosen": -0.43623343110084534, "logits/rejected": -0.43505585193634033, "logps/chosen": -338.3182067871094, "logps/ref_chosen": -66.02030181884766, "logps/ref_rejected": -110.71016693115234, "logps/rejected": -587.6268310546875, "loss": 0.9376, "margin_dpo/margin_mean": 204.6187744140625, "margin_dpo/margin_std": 196.5201416015625, "step": 382 }, { "epoch": 0.5624082232011748, "fcm_dpo/beta": 0.003025210928171873, "fcm_dpo/delta": -0.03023645095527172, "fcm_dpo/margin": 207.67013549804688, "fcm_dpo/q_t": 0.36507272720336914, "grad_norm": 36.634037017822266, "learning_rate": 2.4101862075518037e-07, "logits/chosen": -0.38792383670806885, "logits/rejected": -0.39373254776000977, "logps/chosen": -328.1868896484375, "logps/ref_chosen": -50.39148712158203, "logps/ref_rejected": -93.71589660644531, "logps/rejected": -579.181396484375, "loss": 0.9913, "margin_dpo/margin_mean": 207.67013549804688, "margin_dpo/margin_std": 263.74371337890625, "step": 383 }, { "epoch": 0.5638766519823789, "fcm_dpo/beta": 0.003078867681324482, "fcm_dpo/delta": 0.08402273058891296, "fcm_dpo/margin": 168.70809936523438, "fcm_dpo/q_t": 0.38155868649482727, "grad_norm": 27.592374801635742, "learning_rate": 2.397362428170992e-07, "logits/chosen": -0.425597608089447, "logits/rejected": -0.4135643243789673, "logps/chosen": -330.5281677246094, "logps/ref_chosen": -52.046104431152344, "logps/ref_rejected": -85.76089477539062, "logps/rejected": -532.9510498046875, "loss": 0.9975, "margin_dpo/margin_mean": 168.70809936523438, "margin_dpo/margin_std": 170.51783752441406, "step": 384 }, { "epoch": 0.5653450807635829, "fcm_dpo/beta": 0.0030736164189875126, "fcm_dpo/delta": -0.00494825653731823, "fcm_dpo/margin": 196.7218475341797, "fcm_dpo/q_t": 0.3634437322616577, "grad_norm": 29.260663986206055, "learning_rate": 2.3845413533856514e-07, "logits/chosen": -0.4318080544471741, "logits/rejected": -0.39295560121536255, "logps/chosen": -317.27215576171875, "logps/ref_chosen": -65.55215454101562, "logps/ref_rejected": -77.82792663574219, "logps/rejected": -526.269775390625, "loss": 0.9452, "margin_dpo/margin_mean": 196.7218475341797, "margin_dpo/margin_std": 186.86444091796875, "step": 385 }, { "epoch": 0.566813509544787, "fcm_dpo/beta": 0.003056345507502556, "fcm_dpo/delta": -0.04953234642744064, "fcm_dpo/margin": 211.47879028320312, "fcm_dpo/q_t": 0.3594558835029602, "grad_norm": 32.3004264831543, "learning_rate": 2.3717233210430254e-07, "logits/chosen": -0.4355229139328003, "logits/rejected": -0.4245448112487793, "logps/chosen": -326.9559326171875, "logps/ref_chosen": -58.22185516357422, "logps/ref_rejected": -92.32742309570312, "logps/rejected": -572.5403442382812, "loss": 0.9513, "margin_dpo/margin_mean": 211.4788055419922, "margin_dpo/margin_std": 234.755859375, "step": 386 }, { "epoch": 0.5682819383259912, "fcm_dpo/beta": 0.003059773938730359, "fcm_dpo/delta": 0.06919734925031662, "fcm_dpo/margin": 174.63983154296875, "fcm_dpo/q_t": 0.37939631938934326, "grad_norm": 37.251277923583984, "learning_rate": 2.3589086689101889e-07, "logits/chosen": -0.5007586479187012, "logits/rejected": -0.4742280840873718, "logps/chosen": -350.56463623046875, "logps/ref_chosen": -66.41944885253906, "logps/ref_rejected": -92.16915893554688, "logps/rejected": -550.9541625976562, "loss": 0.9959, "margin_dpo/margin_mean": 174.63983154296875, "margin_dpo/margin_std": 184.31121826171875, "step": 387 }, { "epoch": 0.5697503671071953, "fcm_dpo/beta": 0.0030251401476562023, "fcm_dpo/delta": -0.11198361963033676, "fcm_dpo/margin": 232.57754516601562, "fcm_dpo/q_t": 0.34893184900283813, "grad_norm": 34.62775421142578, "learning_rate": 2.3460977346651428e-07, "logits/chosen": -0.37753403186798096, "logits/rejected": -0.384876012802124, "logps/chosen": -325.88348388671875, "logps/ref_chosen": -50.129459381103516, "logps/ref_rejected": -104.43305969238281, "logps/rejected": -612.7646484375, "loss": 0.9198, "margin_dpo/margin_mean": 232.57754516601562, "margin_dpo/margin_std": 249.87030029296875, "step": 388 }, { "epoch": 0.5712187958883994, "fcm_dpo/beta": 0.0030121058225631714, "fcm_dpo/delta": -0.0019872207194566727, "fcm_dpo/margin": 199.80288696289062, "fcm_dpo/q_t": 0.36984607577323914, "grad_norm": 36.19541549682617, "learning_rate": 2.3332908558879177e-07, "logits/chosen": -0.45936498045921326, "logits/rejected": -0.44065529108047485, "logps/chosen": -367.11614990234375, "logps/ref_chosen": -57.906593322753906, "logps/ref_rejected": -77.91454315185547, "logps/rejected": -586.927001953125, "loss": 0.9889, "margin_dpo/margin_mean": 199.80288696289062, "margin_dpo/margin_std": 242.24429321289062, "step": 389 }, { "epoch": 0.5726872246696035, "fcm_dpo/beta": 0.0030045639723539352, "fcm_dpo/delta": 0.008383665233850479, "fcm_dpo/margin": 196.9191131591797, "fcm_dpo/q_t": 0.37548089027404785, "grad_norm": 35.78683853149414, "learning_rate": 2.320488370051681e-07, "logits/chosen": -0.4140510559082031, "logits/rejected": -0.40045636892318726, "logps/chosen": -360.5129699707031, "logps/ref_chosen": -49.22591781616211, "logps/ref_rejected": -85.5281982421875, "logps/rejected": -593.734375, "loss": 1.0231, "margin_dpo/margin_mean": 196.9191131591797, "margin_dpo/margin_std": 272.89666748046875, "step": 390 }, { "epoch": 0.5741556534508077, "fcm_dpo/beta": 0.0031170903239399195, "fcm_dpo/delta": 0.19101738929748535, "fcm_dpo/margin": 133.54165649414062, "fcm_dpo/q_t": 0.41037672758102417, "grad_norm": 45.97058868408203, "learning_rate": 2.3076906145138405e-07, "logits/chosen": -0.41732552647590637, "logits/rejected": -0.40580475330352783, "logps/chosen": -364.70263671875, "logps/ref_chosen": -64.32965087890625, "logps/ref_rejected": -86.73820495605469, "logps/rejected": -520.65283203125, "loss": 1.1084, "margin_dpo/margin_mean": 133.54165649414062, "margin_dpo/margin_std": 206.95831298828125, "step": 391 }, { "epoch": 0.5756240822320118, "fcm_dpo/beta": 0.003100549802184105, "fcm_dpo/delta": -0.10810496658086777, "fcm_dpo/margin": 225.9717559814453, "fcm_dpo/q_t": 0.3456001877784729, "grad_norm": 33.09029769897461, "learning_rate": 2.294897926507156e-07, "logits/chosen": -0.4238108992576599, "logits/rejected": -0.41437631845474243, "logps/chosen": -324.51971435546875, "logps/ref_chosen": -53.50397872924805, "logps/ref_rejected": -102.34584045410156, "logps/rejected": -599.3333129882812, "loss": 0.8946, "margin_dpo/margin_mean": 225.9717559814453, "margin_dpo/margin_std": 207.6659393310547, "step": 392 }, { "epoch": 0.5770925110132159, "fcm_dpo/beta": 0.003088606521487236, "fcm_dpo/delta": 0.04707473888993263, "fcm_dpo/margin": 179.875244140625, "fcm_dpo/q_t": 0.3865922689437866, "grad_norm": 37.300777435302734, "learning_rate": 2.2821106431308543e-07, "logits/chosen": -0.4085625410079956, "logits/rejected": -0.40135449171066284, "logps/chosen": -313.4761962890625, "logps/ref_chosen": -46.473915100097656, "logps/ref_rejected": -71.96885681152344, "logps/rejected": -518.846435546875, "loss": 1.0513, "margin_dpo/margin_mean": 179.87525939941406, "margin_dpo/margin_std": 273.9698791503906, "step": 393 }, { "epoch": 0.57856093979442, "fcm_dpo/beta": 0.003088343655690551, "fcm_dpo/delta": -0.005594061687588692, "fcm_dpo/margin": 195.913330078125, "fcm_dpo/q_t": 0.371346116065979, "grad_norm": 37.6540641784668, "learning_rate": 2.2693291013417452e-07, "logits/chosen": -0.4252457022666931, "logits/rejected": -0.4192967116832733, "logps/chosen": -353.8863525390625, "logps/ref_chosen": -52.91154861450195, "logps/ref_rejected": -90.8226318359375, "logps/rejected": -587.7107543945312, "loss": 0.9933, "margin_dpo/margin_mean": 195.913330078125, "margin_dpo/margin_std": 244.0496826171875, "step": 394 }, { "epoch": 0.580029368575624, "fcm_dpo/beta": 0.0030713113956153393, "fcm_dpo/delta": -0.03419704735279083, "fcm_dpo/margin": 205.683837890625, "fcm_dpo/q_t": 0.36777931451797485, "grad_norm": 43.52488708496094, "learning_rate": 2.2565536379453404e-07, "logits/chosen": -0.43619823455810547, "logits/rejected": -0.4260198175907135, "logps/chosen": -357.0216369628906, "logps/ref_chosen": -62.546112060546875, "logps/ref_rejected": -83.78262329101562, "logps/rejected": -583.9420166015625, "loss": 0.9959, "margin_dpo/margin_mean": 205.683837890625, "margin_dpo/margin_std": 268.2707214355469, "step": 395 }, { "epoch": 0.5814977973568282, "fcm_dpo/beta": 0.0030712145380675793, "fcm_dpo/delta": 0.00153273344039917, "fcm_dpo/margin": 194.87359619140625, "fcm_dpo/q_t": 0.36835265159606934, "grad_norm": 34.06664276123047, "learning_rate": 2.2437845895869825e-07, "logits/chosen": -0.4365978240966797, "logits/rejected": -0.40913087129592896, "logps/chosen": -365.53350830078125, "logps/ref_chosen": -68.99594116210938, "logps/ref_rejected": -88.64665985107422, "logps/rejected": -580.0578002929688, "loss": 0.9636, "margin_dpo/margin_mean": 194.87359619140625, "margin_dpo/margin_std": 210.96783447265625, "step": 396 }, { "epoch": 0.5829662261380323, "fcm_dpo/beta": 0.0029885326512157917, "fcm_dpo/delta": -0.15819747745990753, "fcm_dpo/margin": 249.2646942138672, "fcm_dpo/q_t": 0.3350715637207031, "grad_norm": 35.845821380615234, "learning_rate": 2.2310222927429716e-07, "logits/chosen": -0.40235501527786255, "logits/rejected": -0.3899458944797516, "logps/chosen": -331.36700439453125, "logps/ref_chosen": -61.27716827392578, "logps/ref_rejected": -103.11612701416016, "logps/rejected": -622.4706420898438, "loss": 0.8687, "margin_dpo/margin_mean": 249.26467895507812, "margin_dpo/margin_std": 219.42138671875, "step": 397 }, { "epoch": 0.5844346549192364, "fcm_dpo/beta": 0.0029659317806363106, "fcm_dpo/delta": -0.029825370758771896, "fcm_dpo/margin": 211.71212768554688, "fcm_dpo/q_t": 0.3643757104873657, "grad_norm": 33.558929443359375, "learning_rate": 2.2182670837116972e-07, "logits/chosen": -0.46954572200775146, "logits/rejected": -0.458831787109375, "logps/chosen": -372.77703857421875, "logps/ref_chosen": -68.15155029296875, "logps/ref_rejected": -108.52360534667969, "logps/rejected": -624.8612060546875, "loss": 0.9628, "margin_dpo/margin_mean": 211.71212768554688, "margin_dpo/margin_std": 241.5102081298828, "step": 398 }, { "epoch": 0.5859030837004405, "fcm_dpo/beta": 0.0029565119184553623, "fcm_dpo/delta": 0.0046921707689762115, "fcm_dpo/margin": 201.39321899414062, "fcm_dpo/q_t": 0.3706039786338806, "grad_norm": 38.02647399902344, "learning_rate": 2.2055192986047804e-07, "logits/chosen": -0.4315683841705322, "logits/rejected": -0.38041120767593384, "logps/chosen": -330.4461975097656, "logps/ref_chosen": -60.889801025390625, "logps/ref_rejected": -77.965576171875, "logps/rejected": -548.9151611328125, "loss": 0.9914, "margin_dpo/margin_mean": 201.39321899414062, "margin_dpo/margin_std": 242.23544311523438, "step": 399 }, { "epoch": 0.5873715124816447, "fcm_dpo/beta": 0.002852116245776415, "fcm_dpo/delta": -0.23298129439353943, "fcm_dpo/margin": 284.9051818847656, "fcm_dpo/q_t": 0.31991565227508545, "grad_norm": 30.26348876953125, "learning_rate": 2.192779273338215e-07, "logits/chosen": -0.40956538915634155, "logits/rejected": -0.3953632712364197, "logps/chosen": -309.8514404296875, "logps/ref_chosen": -63.64359664916992, "logps/ref_rejected": -105.252685546875, "logps/rejected": -636.36572265625, "loss": 0.8283, "margin_dpo/margin_mean": 284.9051818847656, "margin_dpo/margin_std": 236.2659912109375, "step": 400 }, { "epoch": 0.5873715124816447, "eval_fcm_dpo/beta": 0.0028311661444604397, "eval_logits/chosen": -0.44635966420173645, "eval_logits/rejected": -0.42632707953453064, "eval_logps/chosen": -410.6361999511719, "eval_logps/ref_chosen": -79.05104064941406, "eval_logps/ref_rejected": -86.79793548583984, "eval_logps/rejected": -548.2955932617188, "eval_loss": 0.579971969127655, "eval_margin_dpo/margin_mean": 129.91244506835938, "eval_margin_dpo/margin_std": 237.20855712890625, "eval_runtime": 39.3402, "eval_samples_per_second": 59.456, "eval_steps_per_second": 1.881, "step": 400 }, { "epoch": 0.5888399412628488, "fcm_dpo/beta": 0.0028878701850771904, "fcm_dpo/delta": 0.15109677612781525, "fcm_dpo/margin": 157.91604614257812, "fcm_dpo/q_t": 0.40476858615875244, "grad_norm": 35.52607727050781, "learning_rate": 2.1800473436235136e-07, "logits/chosen": -0.39845049381256104, "logits/rejected": -0.3905218243598938, "logps/chosen": -345.0669250488281, "logps/ref_chosen": -57.16303253173828, "logps/ref_rejected": -83.79249572753906, "logps/rejected": -529.6124267578125, "loss": 1.1349, "margin_dpo/margin_mean": 157.91604614257812, "margin_dpo/margin_std": 286.06787109375, "step": 401 }, { "epoch": 0.5903083700440529, "fcm_dpo/beta": 0.002814686391502619, "fcm_dpo/delta": -0.2406725436449051, "fcm_dpo/margin": 291.61102294921875, "fcm_dpo/q_t": 0.3189072012901306, "grad_norm": 26.38422203063965, "learning_rate": 2.1673238449588665e-07, "logits/chosen": -0.423443078994751, "logits/rejected": -0.3997713625431061, "logps/chosen": -261.9085998535156, "logps/ref_chosen": -50.74037170410156, "logps/ref_rejected": -81.0460433959961, "logps/rejected": -583.8253173828125, "loss": 0.814, "margin_dpo/margin_mean": 291.61102294921875, "margin_dpo/margin_std": 232.8087158203125, "step": 402 }, { "epoch": 0.591776798825257, "fcm_dpo/beta": 0.002770167775452137, "fcm_dpo/delta": -0.014987404458224773, "fcm_dpo/margin": 221.64427185058594, "fcm_dpo/q_t": 0.36400288343429565, "grad_norm": 28.025609970092773, "learning_rate": 2.154609112620295e-07, "logits/chosen": -0.3887700140476227, "logits/rejected": -0.3797057271003723, "logps/chosen": -287.8169860839844, "logps/ref_chosen": -47.14731216430664, "logps/ref_rejected": -77.2666015625, "logps/rejected": -539.58056640625, "loss": 0.9562, "margin_dpo/margin_mean": 221.64425659179688, "margin_dpo/margin_std": 230.0044708251953, "step": 403 }, { "epoch": 0.593245227606461, "fcm_dpo/beta": 0.0027550768572837114, "fcm_dpo/delta": -0.033548858016729355, "fcm_dpo/margin": 229.15975952148438, "fcm_dpo/q_t": 0.36247456073760986, "grad_norm": 29.232345581054688, "learning_rate": 2.1419034816528218e-07, "logits/chosen": -0.35594630241394043, "logits/rejected": -0.3386983871459961, "logps/chosen": -307.4197998046875, "logps/ref_chosen": -47.875274658203125, "logps/ref_rejected": -77.15499877929688, "logps/rejected": -565.8592529296875, "loss": 0.9693, "margin_dpo/margin_mean": 229.15975952148438, "margin_dpo/margin_std": 263.5882568359375, "step": 404 }, { "epoch": 0.5947136563876652, "fcm_dpo/beta": 0.0028160212095826864, "fcm_dpo/delta": 0.0803837925195694, "fcm_dpo/margin": 184.99761962890625, "fcm_dpo/q_t": 0.38957273960113525, "grad_norm": 46.26492691040039, "learning_rate": 2.129207286861638e-07, "logits/chosen": -0.3900597095489502, "logits/rejected": -0.37792617082595825, "logps/chosen": -363.26409912109375, "logps/ref_chosen": -65.16290283203125, "logps/ref_rejected": -87.18678283691406, "logps/rejected": -570.2855834960938, "loss": 1.0657, "margin_dpo/margin_mean": 184.99761962890625, "margin_dpo/margin_std": 266.91748046875, "step": 405 }, { "epoch": 0.5961820851688693, "fcm_dpo/beta": 0.002798373345285654, "fcm_dpo/delta": -0.047089073807001114, "fcm_dpo/margin": 229.93914794921875, "fcm_dpo/q_t": 0.3619244396686554, "grad_norm": 38.67859649658203, "learning_rate": 2.1165208628032861e-07, "logits/chosen": -0.4114735722541809, "logits/rejected": -0.4083394408226013, "logps/chosen": -326.46710205078125, "logps/ref_chosen": -49.740814208984375, "logps/ref_rejected": -92.07862854003906, "logps/rejected": -598.744140625, "loss": 0.9584, "margin_dpo/margin_mean": 229.93914794921875, "margin_dpo/margin_std": 259.556640625, "step": 406 }, { "epoch": 0.5976505139500734, "fcm_dpo/beta": 0.002859012223780155, "fcm_dpo/delta": 0.14943772554397583, "fcm_dpo/margin": 159.05093383789062, "fcm_dpo/q_t": 0.3990488648414612, "grad_norm": 51.8092041015625, "learning_rate": 2.1038445437768375e-07, "logits/chosen": -0.41653358936309814, "logits/rejected": -0.38666877150535583, "logps/chosen": -360.2216796875, "logps/ref_chosen": -56.33069610595703, "logps/ref_rejected": -77.51209259033203, "logps/rejected": -540.4539794921875, "loss": 1.1054, "margin_dpo/margin_mean": 159.05093383789062, "margin_dpo/margin_std": 249.28366088867188, "step": 407 }, { "epoch": 0.5991189427312775, "fcm_dpo/beta": 0.0029165763407945633, "fcm_dpo/delta": 0.1471884399652481, "fcm_dpo/margin": 157.62078857421875, "fcm_dpo/q_t": 0.39709725975990295, "grad_norm": 37.17081832885742, "learning_rate": 2.0911786638150872e-07, "logits/chosen": -0.43049564957618713, "logits/rejected": -0.4026661217212677, "logps/chosen": -369.4244384765625, "logps/ref_chosen": -69.789306640625, "logps/ref_rejected": -90.09693908691406, "logps/rejected": -547.3529052734375, "loss": 1.0566, "margin_dpo/margin_mean": 157.62078857421875, "margin_dpo/margin_std": 198.9236602783203, "step": 408 }, { "epoch": 0.6005873715124816, "fcm_dpo/beta": 0.0029886472038924694, "fcm_dpo/delta": 0.08729657530784607, "fcm_dpo/margin": 172.93191528320312, "fcm_dpo/q_t": 0.38554543256759644, "grad_norm": 46.94636154174805, "learning_rate": 2.0785235566757517e-07, "logits/chosen": -0.3939806818962097, "logits/rejected": -0.3780689239501953, "logps/chosen": -353.55657958984375, "logps/ref_chosen": -67.31744384765625, "logps/ref_rejected": -84.904296875, "logps/rejected": -544.0753173828125, "loss": 1.0233, "margin_dpo/margin_mean": 172.93191528320312, "margin_dpo/margin_std": 209.38848876953125, "step": 409 }, { "epoch": 0.6020558002936858, "fcm_dpo/beta": 0.0030058538541197777, "fcm_dpo/delta": 0.03576916456222534, "fcm_dpo/margin": 188.38836669921875, "fcm_dpo/q_t": 0.3729361891746521, "grad_norm": 36.37736129760742, "learning_rate": 2.065879555832674e-07, "logits/chosen": -0.3966106176376343, "logits/rejected": -0.39567676186561584, "logps/chosen": -317.6330261230469, "logps/ref_chosen": -51.465354919433594, "logps/ref_rejected": -83.198974609375, "logps/rejected": -537.7550048828125, "loss": 0.9902, "margin_dpo/margin_mean": 188.38836669921875, "margin_dpo/margin_std": 211.78468322753906, "step": 410 }, { "epoch": 0.6035242290748899, "fcm_dpo/beta": 0.0030007443856447935, "fcm_dpo/delta": 0.03217388689517975, "fcm_dpo/margin": 189.33251953125, "fcm_dpo/q_t": 0.37786054611206055, "grad_norm": 50.11013412475586, "learning_rate": 2.0532469944670343e-07, "logits/chosen": -0.3935418725013733, "logits/rejected": -0.3961487412452698, "logps/chosen": -325.54656982421875, "logps/ref_chosen": -52.30727005004883, "logps/ref_rejected": -80.69495391845703, "logps/rejected": -543.2667846679688, "loss": 1.0067, "margin_dpo/margin_mean": 189.33250427246094, "margin_dpo/margin_std": 229.78692626953125, "step": 411 }, { "epoch": 0.604992657856094, "fcm_dpo/beta": 0.0030541394371539354, "fcm_dpo/delta": 0.038746751844882965, "fcm_dpo/margin": 184.47250366210938, "fcm_dpo/q_t": 0.37744957208633423, "grad_norm": 60.06779861450195, "learning_rate": 2.0406262054585738e-07, "logits/chosen": -0.4395294785499573, "logits/rejected": -0.45951640605926514, "logps/chosen": -333.88372802734375, "logps/ref_chosen": -53.144126892089844, "logps/ref_rejected": -100.0608139038086, "logps/rejected": -565.27294921875, "loss": 1.0095, "margin_dpo/margin_mean": 184.47250366210938, "margin_dpo/margin_std": 230.4259490966797, "step": 412 }, { "epoch": 0.6064610866372981, "fcm_dpo/beta": 0.003098301589488983, "fcm_dpo/delta": 0.049318261444568634, "fcm_dpo/margin": 178.3287353515625, "fcm_dpo/q_t": 0.37706446647644043, "grad_norm": 51.303245544433594, "learning_rate": 2.0280175213768205e-07, "logits/chosen": -0.4190981388092041, "logits/rejected": -0.414195716381073, "logps/chosen": -356.578369140625, "logps/ref_chosen": -61.58196258544922, "logps/ref_rejected": -99.47340393066406, "logps/rejected": -572.7985229492188, "loss": 1.0196, "margin_dpo/margin_mean": 178.3287353515625, "margin_dpo/margin_std": 226.3348388671875, "step": 413 }, { "epoch": 0.6079295154185022, "fcm_dpo/beta": 0.0030625397339463234, "fcm_dpo/delta": -0.10935955494642258, "fcm_dpo/margin": 229.00564575195312, "fcm_dpo/q_t": 0.3444908857345581, "grad_norm": 45.593589782714844, "learning_rate": 2.0154212744723247e-07, "logits/chosen": -0.377156138420105, "logits/rejected": -0.3661056160926819, "logps/chosen": -312.2870178222656, "logps/ref_chosen": -46.63148498535156, "logps/ref_rejected": -87.64653015136719, "logps/rejected": -582.3077392578125, "loss": 0.9087, "margin_dpo/margin_mean": 229.00564575195312, "margin_dpo/margin_std": 222.5440216064453, "step": 414 }, { "epoch": 0.6093979441997063, "fcm_dpo/beta": 0.0030559361912310123, "fcm_dpo/delta": 0.09887358546257019, "fcm_dpo/margin": 165.65187072753906, "fcm_dpo/q_t": 0.3890509009361267, "grad_norm": 40.08522033691406, "learning_rate": 2.002837796667909e-07, "logits/chosen": -0.43923091888427734, "logits/rejected": -0.4275917410850525, "logps/chosen": -386.06353759765625, "logps/ref_chosen": -78.6182861328125, "logps/ref_rejected": -100.47752380371094, "logps/rejected": -573.5745849609375, "loss": 1.0445, "margin_dpo/margin_mean": 165.65187072753906, "margin_dpo/margin_std": 219.1352081298828, "step": 415 }, { "epoch": 0.6108663729809104, "fcm_dpo/beta": 0.003010387998074293, "fcm_dpo/delta": -0.19164395332336426, "fcm_dpo/margin": 258.158447265625, "fcm_dpo/q_t": 0.3281788229942322, "grad_norm": 34.98270797729492, "learning_rate": 1.990267419549914e-07, "logits/chosen": -0.40908828377723694, "logits/rejected": -0.39938801527023315, "logps/chosen": -331.32080078125, "logps/ref_chosen": -58.27912521362305, "logps/ref_rejected": -90.56871795654297, "logps/rejected": -621.7688598632812, "loss": 0.8446, "margin_dpo/margin_mean": 258.158447265625, "margin_dpo/margin_std": 212.94345092773438, "step": 416 }, { "epoch": 0.6123348017621145, "fcm_dpo/beta": 0.002966091735288501, "fcm_dpo/delta": -0.016102882102131844, "fcm_dpo/margin": 207.3574981689453, "fcm_dpo/q_t": 0.360975980758667, "grad_norm": 30.207260131835938, "learning_rate": 1.9777104743594686e-07, "logits/chosen": -0.3575609624385834, "logits/rejected": -0.3195306360721588, "logps/chosen": -326.48291015625, "logps/ref_chosen": -50.1987190246582, "logps/ref_rejected": -68.15184020996094, "logps/rejected": -551.7935791015625, "loss": 0.935, "margin_dpo/margin_mean": 207.35751342773438, "margin_dpo/margin_std": 190.68011474609375, "step": 417 }, { "epoch": 0.6138032305433186, "fcm_dpo/beta": 0.002962034661322832, "fcm_dpo/delta": -0.04637201130390167, "fcm_dpo/margin": 216.85317993164062, "fcm_dpo/q_t": 0.3637494742870331, "grad_norm": 25.070138931274414, "learning_rate": 1.965167291983757e-07, "logits/chosen": -0.46340426802635193, "logits/rejected": -0.4286314845085144, "logps/chosen": -367.762451171875, "logps/ref_chosen": -81.97846984863281, "logps/ref_rejected": -104.69148254394531, "logps/rejected": -607.32861328125, "loss": 0.9707, "margin_dpo/margin_mean": 216.85317993164062, "margin_dpo/margin_std": 255.80337524414062, "step": 418 }, { "epoch": 0.6152716593245228, "fcm_dpo/beta": 0.0028951792046427727, "fcm_dpo/delta": -0.08015348017215729, "fcm_dpo/margin": 233.0614013671875, "fcm_dpo/q_t": 0.3508976697921753, "grad_norm": 24.830020904541016, "learning_rate": 1.9526382029472988e-07, "logits/chosen": -0.39471104741096497, "logits/rejected": -0.3786032199859619, "logps/chosen": -315.84515380859375, "logps/ref_chosen": -52.948646545410156, "logps/ref_rejected": -91.58309936523438, "logps/rejected": -587.541015625, "loss": 0.9221, "margin_dpo/margin_mean": 233.0614013671875, "margin_dpo/margin_std": 232.522705078125, "step": 419 }, { "epoch": 0.6167400881057269, "fcm_dpo/beta": 0.0029447104316204786, "fcm_dpo/delta": 0.1310030221939087, "fcm_dpo/margin": 161.28619384765625, "fcm_dpo/q_t": 0.40055570006370544, "grad_norm": 52.39372253417969, "learning_rate": 1.9401235374032425e-07, "logits/chosen": -0.44154638051986694, "logits/rejected": -0.3922814130783081, "logps/chosen": -384.56658935546875, "logps/ref_chosen": -77.7699203491211, "logps/ref_rejected": -69.31985473632812, "logps/rejected": -537.4027099609375, "loss": 1.0989, "margin_dpo/margin_mean": 161.28619384765625, "margin_dpo/margin_std": 265.3115234375, "step": 420 }, { "epoch": 0.618208516886931, "fcm_dpo/beta": 0.0030234779696911573, "fcm_dpo/delta": 0.11057442426681519, "fcm_dpo/margin": 163.34596252441406, "fcm_dpo/q_t": 0.3888086974620819, "grad_norm": 33.57830810546875, "learning_rate": 1.9276236251246653e-07, "logits/chosen": -0.45666855573654175, "logits/rejected": -0.43738096952438354, "logps/chosen": -311.1793212890625, "logps/ref_chosen": -53.765865325927734, "logps/ref_rejected": -89.28144836425781, "logps/rejected": -510.0408630371094, "loss": 1.0469, "margin_dpo/margin_mean": 163.34596252441406, "margin_dpo/margin_std": 208.17727661132812, "step": 421 }, { "epoch": 0.6196769456681351, "fcm_dpo/beta": 0.0030494979582726955, "fcm_dpo/delta": 0.04655279591679573, "fcm_dpo/margin": 182.31634521484375, "fcm_dpo/q_t": 0.3765103816986084, "grad_norm": 35.29560852050781, "learning_rate": 1.9151387954958792e-07, "logits/chosen": -0.4439750909805298, "logits/rejected": -0.42819076776504517, "logps/chosen": -358.15643310546875, "logps/ref_chosen": -68.6337661743164, "logps/ref_rejected": -87.86351013183594, "logps/rejected": -559.7025146484375, "loss": 1.0112, "margin_dpo/margin_mean": 182.31634521484375, "margin_dpo/margin_std": 225.18934631347656, "step": 422 }, { "epoch": 0.6211453744493393, "fcm_dpo/beta": 0.003037120681256056, "fcm_dpo/delta": -0.029378529638051987, "fcm_dpo/margin": 206.6036376953125, "fcm_dpo/q_t": 0.3606680631637573, "grad_norm": 30.540699005126953, "learning_rate": 1.902669377503756e-07, "logits/chosen": -0.42354506254196167, "logits/rejected": -0.42130446434020996, "logps/chosen": -310.9537353515625, "logps/ref_chosen": -54.99030303955078, "logps/ref_rejected": -86.30654907226562, "logps/rejected": -548.8735961914062, "loss": 0.9463, "margin_dpo/margin_mean": 206.6036376953125, "margin_dpo/margin_std": 212.5541229248047, "step": 423 }, { "epoch": 0.6226138032305433, "fcm_dpo/beta": 0.0030368948355317116, "fcm_dpo/delta": 0.032666608691215515, "fcm_dpo/margin": 187.3394775390625, "fcm_dpo/q_t": 0.37737834453582764, "grad_norm": 32.117645263671875, "learning_rate": 1.890215699729057e-07, "logits/chosen": -0.4487920105457306, "logits/rejected": -0.41544824838638306, "logps/chosen": -282.232177734375, "logps/ref_chosen": -56.01192092895508, "logps/ref_rejected": -66.47896575927734, "logps/rejected": -480.0386962890625, "loss": 1.0066, "margin_dpo/margin_mean": 187.3394775390625, "margin_dpo/margin_std": 233.7342529296875, "step": 424 }, { "epoch": 0.6240822320117474, "fcm_dpo/beta": 0.0031079701147973537, "fcm_dpo/delta": 0.09193390607833862, "fcm_dpo/margin": 164.8294677734375, "fcm_dpo/q_t": 0.3847385048866272, "grad_norm": 31.431589126586914, "learning_rate": 1.8777780903377732e-07, "logits/chosen": -0.38863712549209595, "logits/rejected": -0.3869234323501587, "logps/chosen": -284.51678466796875, "logps/ref_chosen": -46.86899948120117, "logps/ref_rejected": -95.92545318603516, "logps/rejected": -498.4027099609375, "loss": 1.0322, "margin_dpo/margin_mean": 164.8294677734375, "margin_dpo/margin_std": 204.4449920654297, "step": 425 }, { "epoch": 0.6255506607929515, "fcm_dpo/beta": 0.0031212703324854374, "fcm_dpo/delta": -0.033512186259031296, "fcm_dpo/margin": 201.99557495117188, "fcm_dpo/q_t": 0.3606271743774414, "grad_norm": 38.98627853393555, "learning_rate": 1.8653568770724803e-07, "logits/chosen": -0.4529041051864624, "logits/rejected": -0.4022233486175537, "logps/chosen": -299.26812744140625, "logps/ref_chosen": -76.58354187011719, "logps/ref_rejected": -81.26658630371094, "logps/rejected": -505.94671630859375, "loss": 0.9505, "margin_dpo/margin_mean": 201.99557495117188, "margin_dpo/margin_std": 206.97454833984375, "step": 426 }, { "epoch": 0.6270190895741556, "fcm_dpo/beta": 0.003162715584039688, "fcm_dpo/delta": 0.15372638404369354, "fcm_dpo/margin": 143.31716918945312, "fcm_dpo/q_t": 0.3991009294986725, "grad_norm": 34.75946044921875, "learning_rate": 1.8529523872436977e-07, "logits/chosen": -0.45945197343826294, "logits/rejected": -0.4228222370147705, "logps/chosen": -276.22698974609375, "logps/ref_chosen": -64.8538818359375, "logps/ref_rejected": -78.5660171508789, "logps/rejected": -433.2562561035156, "loss": 1.0598, "margin_dpo/margin_mean": 143.31715393066406, "margin_dpo/margin_std": 183.68850708007812, "step": 427 }, { "epoch": 0.6284875183553598, "fcm_dpo/beta": 0.003189402399584651, "fcm_dpo/delta": 0.009054990485310555, "fcm_dpo/margin": 185.44857788085938, "fcm_dpo/q_t": 0.37188664078712463, "grad_norm": 29.212984085083008, "learning_rate": 1.8405649477212697e-07, "logits/chosen": -0.4812716245651245, "logits/rejected": -0.47259581089019775, "logps/chosen": -344.0758972167969, "logps/ref_chosen": -62.63666534423828, "logps/ref_rejected": -103.28181457519531, "logps/rejected": -570.169677734375, "loss": 1.0084, "margin_dpo/margin_mean": 185.44857788085938, "margin_dpo/margin_std": 239.06033325195312, "step": 428 }, { "epoch": 0.6299559471365639, "fcm_dpo/beta": 0.0032644562888890505, "fcm_dpo/delta": 0.08702676743268967, "fcm_dpo/margin": 157.81967163085938, "fcm_dpo/q_t": 0.3883768320083618, "grad_norm": 37.52164077758789, "learning_rate": 1.828194884925749e-07, "logits/chosen": -0.46438342332839966, "logits/rejected": -0.4186452627182007, "logps/chosen": -361.667236328125, "logps/ref_chosen": -81.23401641845703, "logps/ref_rejected": -91.79493713378906, "logps/rejected": -530.0478515625, "loss": 1.0732, "margin_dpo/margin_mean": 157.81967163085938, "margin_dpo/margin_std": 228.81576538085938, "step": 429 }, { "epoch": 0.631424375917768, "fcm_dpo/beta": 0.0032723741605877876, "fcm_dpo/delta": 0.02481374703347683, "fcm_dpo/margin": 176.12640380859375, "fcm_dpo/q_t": 0.37325915694236755, "grad_norm": 30.306285858154297, "learning_rate": 1.8158425248197928e-07, "logits/chosen": -0.4715721011161804, "logits/rejected": -0.46700191497802734, "logps/chosen": -303.973388671875, "logps/ref_chosen": -60.920326232910156, "logps/ref_rejected": -104.42280578613281, "logps/rejected": -523.602294921875, "loss": 0.9759, "margin_dpo/margin_mean": 176.12640380859375, "margin_dpo/margin_std": 191.27938842773438, "step": 430 }, { "epoch": 0.6328928046989721, "fcm_dpo/beta": 0.0031659076921641827, "fcm_dpo/delta": -0.19049368798732758, "fcm_dpo/margin": 244.8009033203125, "fcm_dpo/q_t": 0.3305261731147766, "grad_norm": 26.351268768310547, "learning_rate": 1.8035081928995788e-07, "logits/chosen": -0.3872508406639099, "logits/rejected": -0.3752667307853699, "logps/chosen": -293.5878601074219, "logps/ref_chosen": -57.34874725341797, "logps/ref_rejected": -92.84022521972656, "logps/rejected": -573.8802490234375, "loss": 0.8614, "margin_dpo/margin_mean": 244.8009033203125, "margin_dpo/margin_std": 219.5264892578125, "step": 431 }, { "epoch": 0.6343612334801763, "fcm_dpo/beta": 0.0031368029303848743, "fcm_dpo/delta": -0.083712637424469, "fcm_dpo/margin": 215.79534912109375, "fcm_dpo/q_t": 0.35200947523117065, "grad_norm": 28.590843200683594, "learning_rate": 1.791192214186223e-07, "logits/chosen": -0.4346217215061188, "logits/rejected": -0.4086345434188843, "logps/chosen": -320.74261474609375, "logps/ref_chosen": -71.07479095458984, "logps/ref_rejected": -98.57952880859375, "logps/rejected": -564.042724609375, "loss": 0.9165, "margin_dpo/margin_mean": 215.79534912109375, "margin_dpo/margin_std": 203.4432373046875, "step": 432 }, { "epoch": 0.6358296622613803, "fcm_dpo/beta": 0.0031476817093789577, "fcm_dpo/delta": 0.1059950441122055, "fcm_dpo/margin": 158.52908325195312, "fcm_dpo/q_t": 0.3886904716491699, "grad_norm": 33.54300308227539, "learning_rate": 1.7788949132172193e-07, "logits/chosen": -0.3967306911945343, "logits/rejected": -0.37517887353897095, "logps/chosen": -373.89337158203125, "logps/ref_chosen": -58.273193359375, "logps/ref_rejected": -95.95089721679688, "logps/rejected": -570.1001586914062, "loss": 1.0645, "margin_dpo/margin_mean": 158.52908325195312, "margin_dpo/margin_std": 224.95269775390625, "step": 433 }, { "epoch": 0.6372980910425844, "fcm_dpo/beta": 0.0031643174588680267, "fcm_dpo/delta": 0.033846087753772736, "fcm_dpo/margin": 179.49761962890625, "fcm_dpo/q_t": 0.38102903962135315, "grad_norm": 28.94676399230957, "learning_rate": 1.7666166140378853e-07, "logits/chosen": -0.42005491256713867, "logits/rejected": -0.40054017305374146, "logps/chosen": -327.9237060546875, "logps/ref_chosen": -61.97370147705078, "logps/ref_rejected": -78.49861145019531, "logps/rejected": -523.9462280273438, "loss": 1.0064, "margin_dpo/margin_mean": 179.49761962890625, "margin_dpo/margin_std": 235.10498046875, "step": 434 }, { "epoch": 0.6387665198237885, "fcm_dpo/beta": 0.0031498861499130726, "fcm_dpo/delta": -0.05088559538125992, "fcm_dpo/margin": 205.5423583984375, "fcm_dpo/q_t": 0.3588791489601135, "grad_norm": 28.799325942993164, "learning_rate": 1.7543576401928218e-07, "logits/chosen": -0.4303112328052521, "logits/rejected": -0.40882986783981323, "logps/chosen": -297.64056396484375, "logps/ref_chosen": -51.502052307128906, "logps/ref_rejected": -87.56689453125, "logps/rejected": -539.247802734375, "loss": 0.9575, "margin_dpo/margin_mean": 205.54234313964844, "margin_dpo/margin_std": 227.2718048095703, "step": 435 }, { "epoch": 0.6402349486049926, "fcm_dpo/beta": 0.003141575027257204, "fcm_dpo/delta": -0.0027038920670747757, "fcm_dpo/margin": 191.78115844726562, "fcm_dpo/q_t": 0.36727362871170044, "grad_norm": 53.2458610534668, "learning_rate": 1.742118314717391e-07, "logits/chosen": -0.4502115249633789, "logits/rejected": -0.4046837091445923, "logps/chosen": -322.1340026855469, "logps/ref_chosen": -71.40371704101562, "logps/ref_rejected": -82.72775268554688, "logps/rejected": -525.2391967773438, "loss": 0.9633, "margin_dpo/margin_mean": 191.78115844726562, "margin_dpo/margin_std": 206.23292541503906, "step": 436 }, { "epoch": 0.6417033773861968, "fcm_dpo/beta": 0.0031595812179148197, "fcm_dpo/delta": 0.026493970304727554, "fcm_dpo/margin": 181.99075317382812, "fcm_dpo/q_t": 0.372455894947052, "grad_norm": 30.233713150024414, "learning_rate": 1.7298989601292036e-07, "logits/chosen": -0.43642017245292664, "logits/rejected": -0.4024639129638672, "logps/chosen": -321.0672912597656, "logps/ref_chosen": -64.7442626953125, "logps/ref_rejected": -82.04356384277344, "logps/rejected": -520.3573608398438, "loss": 0.9863, "margin_dpo/margin_mean": 181.99075317382812, "margin_dpo/margin_std": 205.34609985351562, "step": 437 }, { "epoch": 0.6431718061674009, "fcm_dpo/beta": 0.00310856057330966, "fcm_dpo/delta": -0.08261559903621674, "fcm_dpo/margin": 217.52000427246094, "fcm_dpo/q_t": 0.34879451990127563, "grad_norm": 41.74407196044922, "learning_rate": 1.7176998984196144e-07, "logits/chosen": -0.3906601667404175, "logits/rejected": -0.35248345136642456, "logps/chosen": -311.93084716796875, "logps/ref_chosen": -59.0186653137207, "logps/ref_rejected": -83.07682800292969, "logps/rejected": -553.5089721679688, "loss": 0.9029, "margin_dpo/margin_mean": 217.51998901367188, "margin_dpo/margin_std": 194.8534393310547, "step": 438 }, { "epoch": 0.644640234948605, "fcm_dpo/beta": 0.0031026997603476048, "fcm_dpo/delta": 0.05992882698774338, "fcm_dpo/margin": 174.4813232421875, "fcm_dpo/q_t": 0.38289695978164673, "grad_norm": 30.436159133911133, "learning_rate": 1.7055214510452458e-07, "logits/chosen": -0.38404345512390137, "logits/rejected": -0.37690842151641846, "logps/chosen": -328.8792724609375, "logps/ref_chosen": -53.78407669067383, "logps/ref_rejected": -83.98545837402344, "logps/rejected": -533.56201171875, "loss": 1.0235, "margin_dpo/margin_mean": 174.4813232421875, "margin_dpo/margin_std": 211.0643768310547, "step": 439 }, { "epoch": 0.6461086637298091, "fcm_dpo/beta": 0.0031752916984260082, "fcm_dpo/delta": 0.037204399704933167, "fcm_dpo/margin": 177.81304931640625, "fcm_dpo/q_t": 0.3790428638458252, "grad_norm": 36.150543212890625, "learning_rate": 1.6933639389195134e-07, "logits/chosen": -0.45105117559432983, "logits/rejected": -0.4272584915161133, "logps/chosen": -372.9900207519531, "logps/ref_chosen": -78.56671905517578, "logps/ref_rejected": -96.49775695800781, "logps/rejected": -568.734130859375, "loss": 1.0121, "margin_dpo/margin_mean": 177.81304931640625, "margin_dpo/margin_std": 225.9781494140625, "step": 440 }, { "epoch": 0.6475770925110133, "fcm_dpo/beta": 0.0031558992341160774, "fcm_dpo/delta": -0.05471445247530937, "fcm_dpo/margin": 206.31964111328125, "fcm_dpo/q_t": 0.36530476808547974, "grad_norm": 41.47607421875, "learning_rate": 1.681227682404166e-07, "logits/chosen": -0.42267531156539917, "logits/rejected": -0.4021593928337097, "logps/chosen": -371.01458740234375, "logps/ref_chosen": -60.824440002441406, "logps/ref_rejected": -96.47080993652344, "logps/rejected": -612.9805908203125, "loss": 0.9977, "margin_dpo/margin_mean": 206.31964111328125, "margin_dpo/margin_std": 277.037841796875, "step": 441 }, { "epoch": 0.6490455212922174, "fcm_dpo/beta": 0.0031145289540290833, "fcm_dpo/delta": -0.10305148363113403, "fcm_dpo/margin": 223.28561401367188, "fcm_dpo/q_t": 0.3511419892311096, "grad_norm": 28.58584213256836, "learning_rate": 1.669113001300851e-07, "logits/chosen": -0.4091102182865143, "logits/rejected": -0.3904363512992859, "logps/chosen": -326.81689453125, "logps/ref_chosen": -47.01121520996094, "logps/ref_rejected": -76.53926086425781, "logps/rejected": -579.6304931640625, "loss": 0.9206, "margin_dpo/margin_mean": 223.28561401367188, "margin_dpo/margin_std": 229.4536590576172, "step": 442 }, { "epoch": 0.6505139500734214, "fcm_dpo/beta": 0.003117314074188471, "fcm_dpo/delta": 0.14508318901062012, "fcm_dpo/margin": 148.06362915039062, "fcm_dpo/q_t": 0.40172627568244934, "grad_norm": 37.50111389160156, "learning_rate": 1.6570202148426815e-07, "logits/chosen": -0.408527135848999, "logits/rejected": -0.38471612334251404, "logps/chosen": -397.83453369140625, "logps/ref_chosen": -71.27301788330078, "logps/ref_rejected": -86.679931640625, "logps/rejected": -561.3050537109375, "loss": 1.1228, "margin_dpo/margin_mean": 148.06362915039062, "margin_dpo/margin_std": 257.1429138183594, "step": 443 }, { "epoch": 0.6519823788546255, "fcm_dpo/beta": 0.00309735513292253, "fcm_dpo/delta": -0.12234368920326233, "fcm_dpo/margin": 230.2260284423828, "fcm_dpo/q_t": 0.34877198934555054, "grad_norm": 35.36920928955078, "learning_rate": 1.6449496416858282e-07, "logits/chosen": -0.36250776052474976, "logits/rejected": -0.3549155592918396, "logps/chosen": -372.5677185058594, "logps/ref_chosen": -57.213706970214844, "logps/ref_rejected": -97.25489807128906, "logps/rejected": -642.8349609375, "loss": 0.93, "margin_dpo/margin_mean": 230.22604370117188, "margin_dpo/margin_std": 260.1424865722656, "step": 444 }, { "epoch": 0.6534508076358296, "fcm_dpo/beta": 0.003053246531635523, "fcm_dpo/delta": -0.07174232602119446, "fcm_dpo/margin": 218.43179321289062, "fcm_dpo/q_t": 0.3583186864852905, "grad_norm": 34.992069244384766, "learning_rate": 1.6329015999011182e-07, "logits/chosen": -0.39381998777389526, "logits/rejected": -0.37214112281799316, "logps/chosen": -351.34881591796875, "logps/ref_chosen": -67.29979705810547, "logps/ref_rejected": -92.68267059326172, "logps/rejected": -595.1634521484375, "loss": 0.9516, "margin_dpo/margin_mean": 218.43177795410156, "margin_dpo/margin_std": 246.44183349609375, "step": 445 }, { "epoch": 0.6549192364170338, "fcm_dpo/beta": 0.0030171829275786877, "fcm_dpo/delta": -0.10393056273460388, "fcm_dpo/margin": 230.83352661132812, "fcm_dpo/q_t": 0.34617435932159424, "grad_norm": 26.35415267944336, "learning_rate": 1.6208764069656578e-07, "logits/chosen": -0.43713754415512085, "logits/rejected": -0.43820661306381226, "logps/chosen": -321.90740966796875, "logps/ref_chosen": -59.098487854003906, "logps/ref_rejected": -101.26419067382812, "logps/rejected": -594.9066162109375, "loss": 0.8993, "margin_dpo/margin_mean": 230.83352661132812, "margin_dpo/margin_std": 212.1956024169922, "step": 446 }, { "epoch": 0.6563876651982379, "fcm_dpo/beta": 0.0029287966899573803, "fcm_dpo/delta": -0.08693499863147736, "fcm_dpo/margin": 232.25527954101562, "fcm_dpo/q_t": 0.35467979311943054, "grad_norm": 32.44998550415039, "learning_rate": 1.608874379754465e-07, "logits/chosen": -0.421570360660553, "logits/rejected": -0.42356860637664795, "logps/chosen": -343.432861328125, "logps/ref_chosen": -56.07533264160156, "logps/ref_rejected": -98.69475555419922, "logps/rejected": -618.3075561523438, "loss": 0.9272, "margin_dpo/margin_mean": 232.2552947998047, "margin_dpo/margin_std": 253.9686737060547, "step": 447 }, { "epoch": 0.657856093979442, "fcm_dpo/beta": 0.002891901880502701, "fcm_dpo/delta": -0.11446341872215271, "fcm_dpo/margin": 244.29922485351562, "fcm_dpo/q_t": 0.34647947549819946, "grad_norm": 37.00398254394531, "learning_rate": 1.5968958345321177e-07, "logits/chosen": -0.3976057767868042, "logits/rejected": -0.39330196380615234, "logps/chosen": -364.0402526855469, "logps/ref_chosen": -60.00384521484375, "logps/ref_rejected": -102.26465606689453, "logps/rejected": -650.6002807617188, "loss": 0.9, "margin_dpo/margin_mean": 244.29922485351562, "margin_dpo/margin_std": 234.91090393066406, "step": 448 }, { "epoch": 0.6593245227606461, "fcm_dpo/beta": 0.002847596537321806, "fcm_dpo/delta": -0.026260854676365852, "fcm_dpo/margin": 219.3227081298828, "fcm_dpo/q_t": 0.3709370791912079, "grad_norm": 35.01579284667969, "learning_rate": 1.584941086944423e-07, "logits/chosen": -0.40996721386909485, "logits/rejected": -0.3869406580924988, "logps/chosen": -385.76275634765625, "logps/ref_chosen": -67.52661895751953, "logps/ref_rejected": -88.59690856933594, "logps/rejected": -626.15576171875, "loss": 1.0099, "margin_dpo/margin_mean": 219.3227081298828, "margin_dpo/margin_std": 304.9261474609375, "step": 449 }, { "epoch": 0.6607929515418502, "fcm_dpo/beta": 0.0028021461330354214, "fcm_dpo/delta": -0.12278854101896286, "fcm_dpo/margin": 254.86483764648438, "fcm_dpo/q_t": 0.33995717763900757, "grad_norm": 35.38984680175781, "learning_rate": 1.573010452010098e-07, "logits/chosen": -0.4261053204536438, "logits/rejected": -0.42196720838546753, "logps/chosen": -306.01397705078125, "logps/ref_chosen": -57.10811996459961, "logps/ref_rejected": -102.75494384765625, "logps/rejected": -606.525634765625, "loss": 0.8839, "margin_dpo/margin_mean": 254.86483764648438, "margin_dpo/margin_std": 228.66526794433594, "step": 450 }, { "epoch": 0.6622613803230544, "fcm_dpo/beta": 0.002811973448842764, "fcm_dpo/delta": 0.04851696640253067, "fcm_dpo/margin": 196.84996032714844, "fcm_dpo/q_t": 0.3752724528312683, "grad_norm": 42.07421112060547, "learning_rate": 1.5611042441124687e-07, "logits/chosen": -0.415117084980011, "logits/rejected": -0.38188982009887695, "logps/chosen": -367.1541748046875, "logps/ref_chosen": -58.46883010864258, "logps/ref_rejected": -72.92941284179688, "logps/rejected": -578.4647216796875, "loss": 1.0271, "margin_dpo/margin_mean": 196.8499755859375, "margin_dpo/margin_std": 257.9520263671875, "step": 451 }, { "epoch": 0.6637298091042585, "fcm_dpo/beta": 0.002792142331600189, "fcm_dpo/delta": -0.020108355209231377, "fcm_dpo/margin": 221.60552978515625, "fcm_dpo/q_t": 0.36079761385917664, "grad_norm": 25.38531494140625, "learning_rate": 1.549222776991186e-07, "logits/chosen": -0.3913271129131317, "logits/rejected": -0.3999473750591278, "logps/chosen": -272.1903076171875, "logps/ref_chosen": -50.39055252075195, "logps/ref_rejected": -97.77142333984375, "logps/rejected": -541.1766967773438, "loss": 0.9344, "margin_dpo/margin_mean": 221.60552978515625, "margin_dpo/margin_std": 206.4978790283203, "step": 452 }, { "epoch": 0.6651982378854625, "fcm_dpo/beta": 0.0028024991042912006, "fcm_dpo/delta": 0.04751697555184364, "fcm_dpo/margin": 198.021484375, "fcm_dpo/q_t": 0.37886279821395874, "grad_norm": 26.131412506103516, "learning_rate": 1.5373663637339584e-07, "logits/chosen": -0.4181209206581116, "logits/rejected": -0.38588768243789673, "logps/chosen": -332.87738037109375, "logps/ref_chosen": -57.71485137939453, "logps/ref_rejected": -82.20741271972656, "logps/rejected": -555.3914184570312, "loss": 0.9948, "margin_dpo/margin_mean": 198.021484375, "margin_dpo/margin_std": 228.34547424316406, "step": 453 }, { "epoch": 0.6666666666666666, "fcm_dpo/beta": 0.0027867662720382214, "fcm_dpo/delta": -0.059162285178899765, "fcm_dpo/margin": 234.98402404785156, "fcm_dpo/q_t": 0.35987627506256104, "grad_norm": 29.379961013793945, "learning_rate": 1.5255353167683017e-07, "logits/chosen": -0.40170955657958984, "logits/rejected": -0.3783451318740845, "logps/chosen": -366.53466796875, "logps/ref_chosen": -60.945648193359375, "logps/ref_rejected": -84.95079040527344, "logps/rejected": -625.5238037109375, "loss": 0.9475, "margin_dpo/margin_mean": 234.98403930664062, "margin_dpo/margin_std": 266.738037109375, "step": 454 }, { "epoch": 0.6681350954478708, "fcm_dpo/beta": 0.002752738306298852, "fcm_dpo/delta": -0.13085488975048065, "fcm_dpo/margin": 262.0626220703125, "fcm_dpo/q_t": 0.34647172689437866, "grad_norm": 29.516231536865234, "learning_rate": 1.5137299478533064e-07, "logits/chosen": -0.4226793646812439, "logits/rejected": -0.4276910424232483, "logps/chosen": -327.73028564453125, "logps/ref_chosen": -44.88671112060547, "logps/ref_rejected": -115.30147552490234, "logps/rejected": -660.2076416015625, "loss": 0.9117, "margin_dpo/margin_mean": 262.0626220703125, "margin_dpo/margin_std": 271.49981689453125, "step": 455 }, { "epoch": 0.6696035242290749, "fcm_dpo/beta": 0.0026842241641134024, "fcm_dpo/delta": -0.08822280168533325, "fcm_dpo/margin": 254.17987060546875, "fcm_dpo/q_t": 0.3486155569553375, "grad_norm": 30.855884552001953, "learning_rate": 1.5019505680714232e-07, "logits/chosen": -0.42566657066345215, "logits/rejected": -0.4341641068458557, "logps/chosen": -339.21624755859375, "logps/ref_chosen": -57.036781311035156, "logps/ref_rejected": -105.21784210205078, "logps/rejected": -641.5771484375, "loss": 0.8963, "margin_dpo/margin_mean": 254.17987060546875, "margin_dpo/margin_std": 227.32037353515625, "step": 456 }, { "epoch": 0.671071953010279, "fcm_dpo/beta": 0.002638116478919983, "fcm_dpo/delta": -0.051346272230148315, "fcm_dpo/margin": 245.4560546875, "fcm_dpo/q_t": 0.3536643385887146, "grad_norm": 43.32398223876953, "learning_rate": 1.4901974878202627e-07, "logits/chosen": -0.4410599172115326, "logits/rejected": -0.4189218282699585, "logps/chosen": -335.3031311035156, "logps/ref_chosen": -54.24253845214844, "logps/ref_rejected": -85.10956573486328, "logps/rejected": -611.626220703125, "loss": 0.9209, "margin_dpo/margin_mean": 245.4560546875, "margin_dpo/margin_std": 225.08139038085938, "step": 457 }, { "epoch": 0.6725403817914831, "fcm_dpo/beta": 0.0026078918017446995, "fcm_dpo/delta": -0.07165485620498657, "fcm_dpo/margin": 255.62677001953125, "fcm_dpo/q_t": 0.35423409938812256, "grad_norm": 27.470394134521484, "learning_rate": 1.4784710168044212e-07, "logits/chosen": -0.42076390981674194, "logits/rejected": -0.40564727783203125, "logps/chosen": -329.96343994140625, "logps/ref_chosen": -55.40888214111328, "logps/ref_rejected": -97.68325805664062, "logps/rejected": -627.8646240234375, "loss": 0.9208, "margin_dpo/margin_mean": 255.6267852783203, "margin_dpo/margin_std": 259.33697509765625, "step": 458 }, { "epoch": 0.6740088105726872, "fcm_dpo/beta": 0.0025895205326378345, "fcm_dpo/delta": -0.016186371445655823, "fcm_dpo/margin": 237.4794158935547, "fcm_dpo/q_t": 0.36387407779693604, "grad_norm": 28.839862823486328, "learning_rate": 1.466771464027316e-07, "logits/chosen": -0.4365195035934448, "logits/rejected": -0.43345052003860474, "logps/chosen": -352.8772277832031, "logps/ref_chosen": -46.55748748779297, "logps/ref_rejected": -86.16854095458984, "logps/rejected": -629.9677124023438, "loss": 0.9595, "margin_dpo/margin_mean": 237.4794158935547, "margin_dpo/margin_std": 249.88954162597656, "step": 459 }, { "epoch": 0.6754772393538914, "fcm_dpo/beta": 0.0025813106913119555, "fcm_dpo/delta": -0.04352150857448578, "fcm_dpo/margin": 248.228271484375, "fcm_dpo/q_t": 0.35783928632736206, "grad_norm": 29.873188018798828, "learning_rate": 1.4550991377830423e-07, "logits/chosen": -0.45997023582458496, "logits/rejected": -0.47314104437828064, "logps/chosen": -400.25897216796875, "logps/ref_chosen": -51.63489532470703, "logps/ref_rejected": -104.11935424804688, "logps/rejected": -700.9716796875, "loss": 0.9376, "margin_dpo/margin_mean": 248.22825622558594, "margin_dpo/margin_std": 249.94369506835938, "step": 460 }, { "epoch": 0.6769456681350955, "fcm_dpo/beta": 0.002606459427624941, "fcm_dpo/delta": 0.06772121787071228, "fcm_dpo/margin": 205.47274780273438, "fcm_dpo/q_t": 0.3833613097667694, "grad_norm": 33.35320281982422, "learning_rate": 1.4434543456482518e-07, "logits/chosen": -0.49127471446990967, "logits/rejected": -0.4890371561050415, "logps/chosen": -408.4715576171875, "logps/ref_chosen": -55.18195724487305, "logps/ref_rejected": -86.47689819335938, "logps/rejected": -645.2392578125, "loss": 1.0183, "margin_dpo/margin_mean": 205.47274780273438, "margin_dpo/margin_std": 255.91586303710938, "step": 461 }, { "epoch": 0.6784140969162996, "fcm_dpo/beta": 0.002664759289473295, "fcm_dpo/delta": 0.15602310001850128, "fcm_dpo/margin": 169.3271026611328, "fcm_dpo/q_t": 0.40088993310928345, "grad_norm": 56.87949752807617, "learning_rate": 1.4318373944740484e-07, "logits/chosen": -0.5266599655151367, "logits/rejected": -0.5125139951705933, "logps/chosen": -429.41357421875, "logps/ref_chosen": -69.92803192138672, "logps/ref_rejected": -78.84111022949219, "logps/rejected": -607.6537475585938, "loss": 1.0795, "margin_dpo/margin_mean": 169.3271026611328, "margin_dpo/margin_std": 241.06423950195312, "step": 462 }, { "epoch": 0.6798825256975036, "fcm_dpo/beta": 0.0027028373442590237, "fcm_dpo/delta": -0.00691063329577446, "fcm_dpo/margin": 224.18450927734375, "fcm_dpo/q_t": 0.37208259105682373, "grad_norm": 59.45559310913086, "learning_rate": 1.4202485903778976e-07, "logits/chosen": -0.48024556040763855, "logits/rejected": -0.4725441634654999, "logps/chosen": -393.7677001953125, "logps/ref_chosen": -55.27437210083008, "logps/ref_rejected": -89.02497863769531, "logps/rejected": -651.7028198242188, "loss": 1.0023, "margin_dpo/margin_mean": 224.18450927734375, "margin_dpo/margin_std": 288.2208557128906, "step": 463 }, { "epoch": 0.6813509544787077, "fcm_dpo/beta": 0.002575058490037918, "fcm_dpo/delta": -0.24060693383216858, "fcm_dpo/margin": 317.8370361328125, "fcm_dpo/q_t": 0.32363659143447876, "grad_norm": 35.653228759765625, "learning_rate": 1.4086882387355658e-07, "logits/chosen": -0.4769009053707123, "logits/rejected": -0.5046235918998718, "logps/chosen": -376.1098327636719, "logps/ref_chosen": -50.91230010986328, "logps/ref_rejected": -102.4893798828125, "logps/rejected": -745.52392578125, "loss": 0.8444, "margin_dpo/margin_mean": 317.8370361328125, "margin_dpo/margin_std": 292.25225830078125, "step": 464 }, { "epoch": 0.6828193832599119, "fcm_dpo/beta": 0.0025294728111475706, "fcm_dpo/delta": -0.11184176802635193, "fcm_dpo/margin": 278.3428039550781, "fcm_dpo/q_t": 0.34381166100502014, "grad_norm": 32.800621032714844, "learning_rate": 1.3971566441730714e-07, "logits/chosen": -0.4514947533607483, "logits/rejected": -0.4547080993652344, "logps/chosen": -367.2335205078125, "logps/ref_chosen": -60.116851806640625, "logps/ref_rejected": -113.94602966308594, "logps/rejected": -699.405517578125, "loss": 0.8945, "margin_dpo/margin_mean": 278.3428039550781, "margin_dpo/margin_std": 257.3295593261719, "step": 465 }, { "epoch": 0.684287812041116, "fcm_dpo/beta": 0.0025112878065556288, "fcm_dpo/delta": 0.04471685737371445, "fcm_dpo/margin": 222.06112670898438, "fcm_dpo/q_t": 0.37593716382980347, "grad_norm": 33.81809616088867, "learning_rate": 1.3856541105586545e-07, "logits/chosen": -0.46073633432388306, "logits/rejected": -0.45000267028808594, "logps/chosen": -391.29534912109375, "logps/ref_chosen": -52.920921325683594, "logps/ref_rejected": -90.3154296875, "logps/rejected": -650.7509765625, "loss": 1.0039, "margin_dpo/margin_mean": 222.06109619140625, "margin_dpo/margin_std": 262.95599365234375, "step": 466 }, { "epoch": 0.6857562408223201, "fcm_dpo/beta": 0.00250377063639462, "fcm_dpo/delta": 0.003041490912437439, "fcm_dpo/margin": 237.8001251220703, "fcm_dpo/q_t": 0.37523841857910156, "grad_norm": 46.20204544067383, "learning_rate": 1.3741809409947729e-07, "logits/chosen": -0.5411770343780518, "logits/rejected": -0.5251420736312866, "logps/chosen": -482.17987060546875, "logps/ref_chosen": -78.7158203125, "logps/ref_rejected": -102.86019897460938, "logps/rejected": -744.1243896484375, "loss": 1.0294, "margin_dpo/margin_mean": 237.80014038085938, "margin_dpo/margin_std": 325.20404052734375, "step": 467 }, { "epoch": 0.6872246696035242, "fcm_dpo/beta": 0.002495876979082823, "fcm_dpo/delta": -0.055481918156147, "fcm_dpo/margin": 260.9195251464844, "fcm_dpo/q_t": 0.35720133781433105, "grad_norm": 33.269073486328125, "learning_rate": 1.362737437810114e-07, "logits/chosen": -0.5177669525146484, "logits/rejected": -0.49751150608062744, "logps/chosen": -376.0140075683594, "logps/ref_chosen": -69.93536376953125, "logps/ref_rejected": -101.02880859375, "logps/rejected": -668.0269775390625, "loss": 0.9373, "margin_dpo/margin_mean": 260.9195556640625, "margin_dpo/margin_std": 270.485595703125, "step": 468 }, { "epoch": 0.6886930983847284, "fcm_dpo/beta": 0.002456499030813575, "fcm_dpo/delta": -0.07530368864536285, "fcm_dpo/margin": 272.3406982421875, "fcm_dpo/q_t": 0.35292261838912964, "grad_norm": 29.80097770690918, "learning_rate": 1.351323902551631e-07, "logits/chosen": -0.5581041574478149, "logits/rejected": -0.5368698835372925, "logps/chosen": -412.0981140136719, "logps/ref_chosen": -68.12469482421875, "logps/ref_rejected": -104.78640747070312, "logps/rejected": -721.1005249023438, "loss": 0.9203, "margin_dpo/margin_mean": 272.3406982421875, "margin_dpo/margin_std": 266.16925048828125, "step": 469 }, { "epoch": 0.6901615271659325, "fcm_dpo/beta": 0.0024513239040970802, "fcm_dpo/delta": -0.03624919056892395, "fcm_dpo/margin": 258.62164306640625, "fcm_dpo/q_t": 0.35686296224594116, "grad_norm": 25.734210968017578, "learning_rate": 1.339940635976592e-07, "logits/chosen": -0.5032815933227539, "logits/rejected": -0.49432438611984253, "logps/chosen": -312.69384765625, "logps/ref_chosen": -43.791927337646484, "logps/ref_rejected": -82.70285034179688, "logps/rejected": -610.2263793945312, "loss": 0.9388, "margin_dpo/margin_mean": 258.62164306640625, "margin_dpo/margin_std": 256.03778076171875, "step": 470 }, { "epoch": 0.6916299559471366, "fcm_dpo/beta": 0.0024561528116464615, "fcm_dpo/delta": 0.06438558548688889, "fcm_dpo/margin": 219.4058074951172, "fcm_dpo/q_t": 0.38015252351760864, "grad_norm": 27.724517822265625, "learning_rate": 1.3285879380446563e-07, "logits/chosen": -0.5248172879219055, "logits/rejected": -0.5118055939674377, "logps/chosen": -382.17425537109375, "logps/ref_chosen": -63.33952331542969, "logps/ref_rejected": -83.61048126220703, "logps/rejected": -621.85107421875, "loss": 1.001, "margin_dpo/margin_mean": 219.40579223632812, "margin_dpo/margin_std": 245.6639404296875, "step": 471 }, { "epoch": 0.6930983847283406, "fcm_dpo/beta": 0.0024520205333828926, "fcm_dpo/delta": -0.06850513815879822, "fcm_dpo/margin": 270.7828063964844, "fcm_dpo/q_t": 0.3614341616630554, "grad_norm": 26.621925354003906, "learning_rate": 1.317266107909975e-07, "logits/chosen": -0.5987716317176819, "logits/rejected": -0.5613675117492676, "logps/chosen": -406.96942138671875, "logps/ref_chosen": -83.66610717773438, "logps/ref_rejected": -117.20919799804688, "logps/rejected": -711.2952880859375, "loss": 0.9546, "margin_dpo/margin_mean": 270.78277587890625, "margin_dpo/margin_std": 321.0069885253906, "step": 472 }, { "epoch": 0.6945668135095447, "fcm_dpo/beta": 0.00249496940523386, "fcm_dpo/delta": 0.14937041699886322, "fcm_dpo/margin": 183.41294860839844, "fcm_dpo/q_t": 0.40312713384628296, "grad_norm": 36.99402618408203, "learning_rate": 1.3059754439133002e-07, "logits/chosen": -0.5006664395332336, "logits/rejected": -0.4722933769226074, "logps/chosen": -418.47216796875, "logps/ref_chosen": -63.49696731567383, "logps/ref_rejected": -81.14657592773438, "logps/rejected": -619.5347900390625, "loss": 1.1136, "margin_dpo/margin_mean": 183.4129638671875, "margin_dpo/margin_std": 317.38916015625, "step": 473 }, { "epoch": 0.6960352422907489, "fcm_dpo/beta": 0.002567752031609416, "fcm_dpo/delta": 0.09946488589048386, "fcm_dpo/margin": 196.4961700439453, "fcm_dpo/q_t": 0.38934361934661865, "grad_norm": 33.70315933227539, "learning_rate": 1.2947162435741277e-07, "logits/chosen": -0.5230387449264526, "logits/rejected": -0.5138710737228394, "logps/chosen": -361.4368591308594, "logps/ref_chosen": -52.6119384765625, "logps/ref_rejected": -90.08041381835938, "logps/rejected": -595.4014892578125, "loss": 1.0559, "margin_dpo/margin_mean": 196.49615478515625, "margin_dpo/margin_std": 272.44683837890625, "step": 474 }, { "epoch": 0.697503671071953, "fcm_dpo/beta": 0.002530063269659877, "fcm_dpo/delta": -0.09672711789608002, "fcm_dpo/margin": 272.7248840332031, "fcm_dpo/q_t": 0.3485596776008606, "grad_norm": 23.911319732666016, "learning_rate": 1.2834888035828596e-07, "logits/chosen": -0.48518672585487366, "logits/rejected": -0.49074888229370117, "logps/chosen": -290.80499267578125, "logps/ref_chosen": -42.49519348144531, "logps/ref_rejected": -90.06294250488281, "logps/rejected": -611.09765625, "loss": 0.9061, "margin_dpo/margin_mean": 272.7248840332031, "margin_dpo/margin_std": 265.2176208496094, "step": 475 }, { "epoch": 0.6989720998531571, "fcm_dpo/beta": 0.002525830641388893, "fcm_dpo/delta": 0.016785459592938423, "fcm_dpo/margin": 231.28424072265625, "fcm_dpo/q_t": 0.3694440722465515, "grad_norm": 32.7138786315918, "learning_rate": 1.2722934197929802e-07, "logits/chosen": -0.5376179814338684, "logits/rejected": -0.5307664275169373, "logps/chosen": -305.538818359375, "logps/ref_chosen": -42.94938278198242, "logps/ref_rejected": -73.71023559570312, "logps/rejected": -567.5839233398438, "loss": 0.9606, "margin_dpo/margin_mean": 231.28424072265625, "margin_dpo/margin_std": 231.99273681640625, "step": 476 }, { "epoch": 0.7004405286343612, "fcm_dpo/beta": 0.0025543358642607927, "fcm_dpo/delta": 0.06146734952926636, "fcm_dpo/margin": 212.0760498046875, "fcm_dpo/q_t": 0.3796027898788452, "grad_norm": 30.43485450744629, "learning_rate": 1.2611303872132631e-07, "logits/chosen": -0.5261760354042053, "logits/rejected": -0.4716786742210388, "logps/chosen": -365.0345458984375, "logps/ref_chosen": -70.77261352539062, "logps/ref_rejected": -76.13737487792969, "logps/rejected": -582.475341796875, "loss": 1.02, "margin_dpo/margin_mean": 212.0760498046875, "margin_dpo/margin_std": 266.8735656738281, "step": 477 }, { "epoch": 0.7019089574155654, "fcm_dpo/beta": 0.0025457185693085194, "fcm_dpo/delta": -0.05643880367279053, "fcm_dpo/margin": 256.39349365234375, "fcm_dpo/q_t": 0.3551124632358551, "grad_norm": 30.269575119018555, "learning_rate": 1.2500000000000005e-07, "logits/chosen": -0.4699116349220276, "logits/rejected": -0.4670166075229645, "logps/chosen": -282.0042724609375, "logps/ref_chosen": -41.440513610839844, "logps/ref_rejected": -85.36196899414062, "logps/rejected": -582.3192138671875, "loss": 0.9133, "margin_dpo/margin_mean": 256.39349365234375, "margin_dpo/margin_std": 231.91416931152344, "step": 478 }, { "epoch": 0.7033773861967695, "fcm_dpo/beta": 0.002581070177257061, "fcm_dpo/delta": 0.0580781027674675, "fcm_dpo/margin": 209.93270874023438, "fcm_dpo/q_t": 0.3800834119319916, "grad_norm": 34.0640754699707, "learning_rate": 1.2389025514492456e-07, "logits/chosen": -0.49057966470718384, "logits/rejected": -0.49215784668922424, "logps/chosen": -384.4983215332031, "logps/ref_chosen": -53.907920837402344, "logps/ref_rejected": -95.1163330078125, "logps/rejected": -635.639404296875, "loss": 1.0201, "margin_dpo/margin_mean": 209.93270874023438, "margin_dpo/margin_std": 248.3424530029297, "step": 479 }, { "epoch": 0.7048458149779736, "fcm_dpo/beta": 0.002632645657286048, "fcm_dpo/delta": 0.1320512592792511, "fcm_dpo/margin": 179.1344757080078, "fcm_dpo/q_t": 0.3967488408088684, "grad_norm": 46.27069091796875, "learning_rate": 1.227838333989088e-07, "logits/chosen": -0.4810282588005066, "logits/rejected": -0.4517630636692047, "logps/chosen": -410.599853515625, "logps/ref_chosen": -58.682701110839844, "logps/ref_rejected": -82.93248748779297, "logps/rejected": -613.984130859375, "loss": 1.083, "margin_dpo/margin_mean": 179.1344757080078, "margin_dpo/margin_std": 258.1438293457031, "step": 480 }, { "epoch": 0.7063142437591777, "fcm_dpo/beta": 0.002601025393232703, "fcm_dpo/delta": -0.051007017493247986, "fcm_dpo/margin": 248.92288208007812, "fcm_dpo/q_t": 0.3576533794403076, "grad_norm": 35.464439392089844, "learning_rate": 1.2168076391719489e-07, "logits/chosen": -0.5199482440948486, "logits/rejected": -0.5128802061080933, "logps/chosen": -346.0648498535156, "logps/ref_chosen": -54.964271545410156, "logps/ref_rejected": -92.42044067382812, "logps/rejected": -632.44384765625, "loss": 0.94, "margin_dpo/margin_mean": 248.92288208007812, "margin_dpo/margin_std": 260.282958984375, "step": 481 }, { "epoch": 0.7077826725403817, "fcm_dpo/beta": 0.002659988822415471, "fcm_dpo/delta": 0.1587131917476654, "fcm_dpo/margin": 168.63143920898438, "fcm_dpo/q_t": 0.4008188843727112, "grad_norm": 38.83755111694336, "learning_rate": 1.2058107576668938e-07, "logits/chosen": -0.47180038690567017, "logits/rejected": -0.45411762595176697, "logps/chosen": -398.88873291015625, "logps/ref_chosen": -67.553466796875, "logps/ref_rejected": -87.58953857421875, "logps/rejected": -587.5562133789062, "loss": 1.0902, "margin_dpo/margin_mean": 168.63143920898438, "margin_dpo/margin_std": 254.7269287109375, "step": 482 }, { "epoch": 0.7092511013215859, "fcm_dpo/beta": 0.002619517268612981, "fcm_dpo/delta": -0.1474650502204895, "fcm_dpo/margin": 281.0007019042969, "fcm_dpo/q_t": 0.34216371178627014, "grad_norm": 40.65428161621094, "learning_rate": 1.194847979251979e-07, "logits/chosen": -0.5000494718551636, "logits/rejected": -0.48498857021331787, "logps/chosen": -357.7904052734375, "logps/ref_chosen": -63.32981872558594, "logps/ref_rejected": -95.78697204589844, "logps/rejected": -671.248291015625, "loss": 0.8912, "margin_dpo/margin_mean": 281.0007019042969, "margin_dpo/margin_std": 279.4383544921875, "step": 483 }, { "epoch": 0.71071953010279, "fcm_dpo/beta": 0.0026109693571925163, "fcm_dpo/delta": -0.03564995154738426, "fcm_dpo/margin": 242.32601928710938, "fcm_dpo/q_t": 0.3604073226451874, "grad_norm": 27.125112533569336, "learning_rate": 1.1839195928066101e-07, "logits/chosen": -0.48026251792907715, "logits/rejected": -0.46726343035697937, "logps/chosen": -332.4265441894531, "logps/ref_chosen": -59.13812255859375, "logps/ref_rejected": -84.37144470214844, "logps/rejected": -599.9859008789062, "loss": 0.9377, "margin_dpo/margin_mean": 242.32603454589844, "margin_dpo/margin_std": 237.0671844482422, "step": 484 }, { "epoch": 0.7121879588839941, "fcm_dpo/beta": 0.0025863349437713623, "fcm_dpo/delta": -0.01730707660317421, "fcm_dpo/margin": 238.25384521484375, "fcm_dpo/q_t": 0.36559098958969116, "grad_norm": 36.989036560058594, "learning_rate": 1.1730258863039347e-07, "logits/chosen": -0.468616247177124, "logits/rejected": -0.466255784034729, "logps/chosen": -326.2464904785156, "logps/ref_chosen": -58.849571228027344, "logps/ref_rejected": -103.36408233642578, "logps/rejected": -609.0148315429688, "loss": 0.967, "margin_dpo/margin_mean": 238.25384521484375, "margin_dpo/margin_std": 268.4663391113281, "step": 485 }, { "epoch": 0.7136563876651982, "fcm_dpo/beta": 0.0025734733790159225, "fcm_dpo/delta": -0.033280279487371445, "fcm_dpo/margin": 245.24191284179688, "fcm_dpo/q_t": 0.36465585231781006, "grad_norm": 40.85445022583008, "learning_rate": 1.1621671468032493e-07, "logits/chosen": -0.50041663646698, "logits/rejected": -0.48742035031318665, "logps/chosen": -367.93310546875, "logps/ref_chosen": -55.25966262817383, "logps/ref_rejected": -92.13936614990234, "logps/rejected": -650.0546875, "loss": 0.9882, "margin_dpo/margin_mean": 245.24192810058594, "margin_dpo/margin_std": 306.00677490234375, "step": 486 }, { "epoch": 0.7151248164464024, "fcm_dpo/beta": 0.002564340131357312, "fcm_dpo/delta": 0.024611998349428177, "fcm_dpo/margin": 224.8999786376953, "fcm_dpo/q_t": 0.3688669800758362, "grad_norm": 35.55867004394531, "learning_rate": 1.1513436604424378e-07, "logits/chosen": -0.48910582065582275, "logits/rejected": -0.4775072932243347, "logps/chosen": -351.23388671875, "logps/ref_chosen": -53.06330871582031, "logps/ref_rejected": -92.41883087158203, "logps/rejected": -615.4893798828125, "loss": 0.9727, "margin_dpo/margin_mean": 224.8999786376953, "margin_dpo/margin_std": 231.9244842529297, "step": 487 }, { "epoch": 0.7165932452276065, "fcm_dpo/beta": 0.0025914846919476986, "fcm_dpo/delta": 0.017736222594976425, "fcm_dpo/margin": 224.94769287109375, "fcm_dpo/q_t": 0.3664684593677521, "grad_norm": 39.95408248901367, "learning_rate": 1.1405557124304335e-07, "logits/chosen": -0.4504004716873169, "logits/rejected": -0.43368983268737793, "logps/chosen": -295.21221923828125, "logps/ref_chosen": -52.22815704345703, "logps/ref_rejected": -84.00656127929688, "logps/rejected": -551.9383544921875, "loss": 0.9446, "margin_dpo/margin_mean": 224.94769287109375, "margin_dpo/margin_std": 193.76791381835938, "step": 488 }, { "epoch": 0.7180616740088106, "fcm_dpo/beta": 0.00260374927893281, "fcm_dpo/delta": 0.053067468106746674, "fcm_dpo/margin": 211.18069458007812, "fcm_dpo/q_t": 0.378859281539917, "grad_norm": 33.997283935546875, "learning_rate": 1.1298035870396985e-07, "logits/chosen": -0.49653559923171997, "logits/rejected": -0.47625601291656494, "logps/chosen": -304.00372314453125, "logps/ref_chosen": -55.989627838134766, "logps/ref_rejected": -79.39812469482422, "logps/rejected": -538.5928955078125, "loss": 0.9855, "margin_dpo/margin_mean": 211.18069458007812, "margin_dpo/margin_std": 227.5222930908203, "step": 489 }, { "epoch": 0.7195301027900147, "fcm_dpo/beta": 0.002616376616060734, "fcm_dpo/delta": -0.01173383742570877, "fcm_dpo/margin": 233.45465087890625, "fcm_dpo/q_t": 0.36978480219841003, "grad_norm": 51.60263442993164, "learning_rate": 1.1190875675987355e-07, "logits/chosen": -0.5432174801826477, "logits/rejected": -0.5576708316802979, "logps/chosen": -326.5618896484375, "logps/ref_chosen": -52.36639404296875, "logps/ref_rejected": -110.4090576171875, "logps/rejected": -618.0592041015625, "loss": 0.995, "margin_dpo/margin_mean": 233.45465087890625, "margin_dpo/margin_std": 293.9273376464844, "step": 490 }, { "epoch": 0.7209985315712188, "fcm_dpo/beta": 0.002662881277501583, "fcm_dpo/delta": 0.1412055492401123, "fcm_dpo/margin": 174.71432495117188, "fcm_dpo/q_t": 0.39850908517837524, "grad_norm": 46.16297912597656, "learning_rate": 1.1084079364846241e-07, "logits/chosen": -0.4562872052192688, "logits/rejected": -0.4420876204967499, "logps/chosen": -348.5367431640625, "logps/ref_chosen": -60.11626434326172, "logps/ref_rejected": -73.27278900146484, "logps/rejected": -536.4075927734375, "loss": 1.072, "margin_dpo/margin_mean": 174.71432495117188, "margin_dpo/margin_std": 248.17523193359375, "step": 491 }, { "epoch": 0.7224669603524229, "fcm_dpo/beta": 0.002773004351183772, "fcm_dpo/delta": 0.22383955121040344, "fcm_dpo/margin": 138.69735717773438, "fcm_dpo/q_t": 0.4172734022140503, "grad_norm": 42.61263656616211, "learning_rate": 1.097764975115576e-07, "logits/chosen": -0.44189101457595825, "logits/rejected": -0.4200980067253113, "logps/chosen": -361.65570068359375, "logps/ref_chosen": -53.994178771972656, "logps/ref_rejected": -72.65962219238281, "logps/rejected": -519.0184936523438, "loss": 1.1625, "margin_dpo/margin_mean": 138.69735717773438, "margin_dpo/margin_std": 266.28662109375, "step": 492 }, { "epoch": 0.723935389133627, "fcm_dpo/beta": 0.0028385447803884745, "fcm_dpo/delta": 0.1169082373380661, "fcm_dpo/margin": 172.2346954345703, "fcm_dpo/q_t": 0.39189258217811584, "grad_norm": 43.42572021484375, "learning_rate": 1.0871589639435203e-07, "logits/chosen": -0.5289523601531982, "logits/rejected": -0.5010632276535034, "logps/chosen": -380.6925048828125, "logps/ref_chosen": -75.49723815917969, "logps/ref_rejected": -87.32301330566406, "logps/rejected": -564.7529296875, "loss": 1.0419, "margin_dpo/margin_mean": 172.2346954345703, "margin_dpo/margin_std": 214.199951171875, "step": 493 }, { "epoch": 0.7254038179148311, "fcm_dpo/beta": 0.00280939694494009, "fcm_dpo/delta": -0.12965027987957, "fcm_dpo/margin": 256.29547119140625, "fcm_dpo/q_t": 0.3384782075881958, "grad_norm": 34.547096252441406, "learning_rate": 1.0765901824467166e-07, "logits/chosen": -0.4465922713279724, "logits/rejected": -0.45816731452941895, "logps/chosen": -296.73284912109375, "logps/ref_chosen": -41.35926818847656, "logps/ref_rejected": -86.09136962890625, "logps/rejected": -597.7604370117188, "loss": 0.8787, "margin_dpo/margin_mean": 256.2955017089844, "margin_dpo/margin_std": 221.27516174316406, "step": 494 }, { "epoch": 0.7268722466960352, "fcm_dpo/beta": 0.002777573186904192, "fcm_dpo/delta": -0.08846628665924072, "fcm_dpo/margin": 245.58688354492188, "fcm_dpo/q_t": 0.35333937406539917, "grad_norm": 34.23904037475586, "learning_rate": 1.0660589091223854e-07, "logits/chosen": -0.493415504693985, "logits/rejected": -0.4882367253303528, "logps/chosen": -356.96539306640625, "logps/ref_chosen": -63.53507995605469, "logps/ref_rejected": -91.42443084716797, "logps/rejected": -630.441650390625, "loss": 0.929, "margin_dpo/margin_mean": 245.58688354492188, "margin_dpo/margin_std": 258.7152099609375, "step": 495 }, { "epoch": 0.7283406754772394, "fcm_dpo/beta": 0.002794483909383416, "fcm_dpo/delta": 0.13067297637462616, "fcm_dpo/margin": 170.2359619140625, "fcm_dpo/q_t": 0.39404138922691345, "grad_norm": 46.63887405395508, "learning_rate": 1.0555654214793722e-07, "logits/chosen": -0.4721938669681549, "logits/rejected": -0.4511212706565857, "logps/chosen": -414.6033630371094, "logps/ref_chosen": -72.5919189453125, "logps/ref_rejected": -84.32933807373047, "logps/rejected": -596.5767822265625, "loss": 1.0508, "margin_dpo/margin_mean": 170.2359619140625, "margin_dpo/margin_std": 214.48468017578125, "step": 496 }, { "epoch": 0.7298091042584435, "fcm_dpo/beta": 0.0028514941222965717, "fcm_dpo/delta": 0.13596858084201813, "fcm_dpo/margin": 164.8284912109375, "fcm_dpo/q_t": 0.3949023485183716, "grad_norm": 35.32782745361328, "learning_rate": 1.0451099960308374e-07, "logits/chosen": -0.4963914752006531, "logits/rejected": -0.479754239320755, "logps/chosen": -384.9673156738281, "logps/ref_chosen": -58.59397506713867, "logps/ref_rejected": -76.28836822509766, "logps/rejected": -567.490234375, "loss": 1.0561, "margin_dpo/margin_mean": 164.8284912109375, "margin_dpo/margin_std": 201.14633178710938, "step": 497 }, { "epoch": 0.7312775330396476, "fcm_dpo/beta": 0.0028984027449041605, "fcm_dpo/delta": 0.004962385166436434, "fcm_dpo/margin": 205.40037536621094, "fcm_dpo/q_t": 0.37139469385147095, "grad_norm": 38.93318557739258, "learning_rate": 1.0346929082869641e-07, "logits/chosen": -0.4890047311782837, "logits/rejected": -0.4707816243171692, "logps/chosen": -396.84356689453125, "logps/ref_chosen": -71.20565795898438, "logps/ref_rejected": -83.95803833007812, "logps/rejected": -614.996337890625, "loss": 1.0052, "margin_dpo/margin_mean": 205.40037536621094, "margin_dpo/margin_std": 263.6328125, "step": 498 }, { "epoch": 0.7327459618208517, "fcm_dpo/beta": 0.0028373058885335922, "fcm_dpo/delta": -0.13358387351036072, "fcm_dpo/margin": 255.02593994140625, "fcm_dpo/q_t": 0.3414595425128937, "grad_norm": 47.647823333740234, "learning_rate": 1.0243144327477013e-07, "logits/chosen": -0.4490722417831421, "logits/rejected": -0.45660898089408875, "logps/chosen": -363.71142578125, "logps/ref_chosen": -51.25519561767578, "logps/ref_rejected": -101.07870483398438, "logps/rejected": -668.5609130859375, "loss": 0.9152, "margin_dpo/margin_mean": 255.02593994140625, "margin_dpo/margin_std": 267.2959899902344, "step": 499 }, { "epoch": 0.7342143906020558, "fcm_dpo/beta": 0.0028089380357414484, "fcm_dpo/delta": -0.024006161838769913, "fcm_dpo/margin": 221.58847045898438, "fcm_dpo/q_t": 0.3665502965450287, "grad_norm": 52.78643798828125, "learning_rate": 1.0139748428955333e-07, "logits/chosen": -0.4591681957244873, "logits/rejected": -0.4673272371292114, "logps/chosen": -429.0184326171875, "logps/ref_chosen": -57.027442932128906, "logps/ref_rejected": -93.93421173095703, "logps/rejected": -687.513671875, "loss": 0.9865, "margin_dpo/margin_mean": 221.58847045898438, "margin_dpo/margin_std": 274.5337829589844, "step": 500 }, { "epoch": 0.73568281938326, "fcm_dpo/beta": 0.002789230551570654, "fcm_dpo/delta": -0.06634011119604111, "fcm_dpo/margin": 237.33590698242188, "fcm_dpo/q_t": 0.3573167026042938, "grad_norm": 34.69628143310547, "learning_rate": 1.0036744111882672e-07, "logits/chosen": -0.45637276768684387, "logits/rejected": -0.4298150837421417, "logps/chosen": -364.97088623046875, "logps/ref_chosen": -54.359527587890625, "logps/ref_rejected": -80.15670013427734, "logps/rejected": -628.10400390625, "loss": 0.9556, "margin_dpo/margin_mean": 237.33590698242188, "margin_dpo/margin_std": 273.72589111328125, "step": 501 }, { "epoch": 0.737151248164464, "fcm_dpo/beta": 0.002761396113783121, "fcm_dpo/delta": -0.06494729220867157, "fcm_dpo/margin": 239.1594696044922, "fcm_dpo/q_t": 0.3486095666885376, "grad_norm": 28.551965713500977, "learning_rate": 9.934134090518592e-08, "logits/chosen": -0.4536617398262024, "logits/rejected": -0.4335412085056305, "logps/chosen": -356.140380859375, "logps/ref_chosen": -67.60050964355469, "logps/ref_rejected": -82.94876098632812, "logps/rejected": -610.6480712890625, "loss": 0.8913, "margin_dpo/margin_mean": 239.1594696044922, "margin_dpo/margin_std": 181.16015625, "step": 502 }, { "epoch": 0.7386196769456681, "fcm_dpo/beta": 0.002728442195802927, "fcm_dpo/delta": 0.006640031933784485, "fcm_dpo/margin": 217.5033721923828, "fcm_dpo/q_t": 0.36779502034187317, "grad_norm": 33.08012008666992, "learning_rate": 9.831921068732571e-08, "logits/chosen": -0.43443652987480164, "logits/rejected": -0.4157460927963257, "logps/chosen": -348.41259765625, "logps/ref_chosen": -55.078407287597656, "logps/ref_rejected": -82.50544738769531, "logps/rejected": -593.343017578125, "loss": 0.9555, "margin_dpo/margin_mean": 217.50338745117188, "margin_dpo/margin_std": 215.96286010742188, "step": 503 }, { "epoch": 0.7400881057268722, "fcm_dpo/beta": 0.0026825086679309607, "fcm_dpo/delta": -0.13238787651062012, "fcm_dpo/margin": 269.34735107421875, "fcm_dpo/q_t": 0.34110093116760254, "grad_norm": 35.52935791015625, "learning_rate": 9.730107739932805e-08, "logits/chosen": -0.45639944076538086, "logits/rejected": -0.4605451822280884, "logps/chosen": -375.00860595703125, "logps/ref_chosen": -59.96575164794922, "logps/ref_rejected": -103.76212310791016, "logps/rejected": -688.15234375, "loss": 0.8908, "margin_dpo/margin_mean": 269.34735107421875, "margin_dpo/margin_std": 249.264892578125, "step": 504 }, { "epoch": 0.7415565345080763, "fcm_dpo/beta": 0.0027532954700291157, "fcm_dpo/delta": 0.1878781020641327, "fcm_dpo/margin": 152.27743530273438, "fcm_dpo/q_t": 0.4108385741710663, "grad_norm": 54.627689361572266, "learning_rate": 9.628696786995188e-08, "logits/chosen": -0.5202739834785461, "logits/rejected": -0.49124228954315186, "logps/chosen": -429.90838623046875, "logps/ref_chosen": -76.1549072265625, "logps/ref_rejected": -88.58537292480469, "logps/rejected": -594.6163330078125, "loss": 1.1332, "margin_dpo/margin_mean": 152.27743530273438, "margin_dpo/margin_std": 267.0574951171875, "step": 505 }, { "epoch": 0.7430249632892805, "fcm_dpo/beta": 0.002741964068263769, "fcm_dpo/delta": -0.04416107013821602, "fcm_dpo/margin": 233.71510314941406, "fcm_dpo/q_t": 0.358504056930542, "grad_norm": 38.65158462524414, "learning_rate": 9.527690882192635e-08, "logits/chosen": -0.47684890031814575, "logits/rejected": -0.4706017076969147, "logps/chosen": -328.3465576171875, "logps/ref_chosen": -48.96050262451172, "logps/ref_rejected": -78.41505432128906, "logps/rejected": -591.5161743164062, "loss": 0.9438, "margin_dpo/margin_mean": 233.71510314941406, "margin_dpo/margin_std": 243.26914978027344, "step": 506 }, { "epoch": 0.7444933920704846, "fcm_dpo/beta": 0.002767894882708788, "fcm_dpo/delta": 0.04769434034824371, "fcm_dpo/margin": 200.48585510253906, "fcm_dpo/q_t": 0.38139885663986206, "grad_norm": 31.428619384765625, "learning_rate": 9.427092687124691e-08, "logits/chosen": -0.4897286891937256, "logits/rejected": -0.4771530032157898, "logps/chosen": -389.1927185058594, "logps/ref_chosen": -66.80149841308594, "logps/ref_rejected": -95.37289428710938, "logps/rejected": -618.25, "loss": 1.0113, "margin_dpo/margin_mean": 200.48583984375, "margin_dpo/margin_std": 256.5914001464844, "step": 507 }, { "epoch": 0.7459618208516887, "fcm_dpo/beta": 0.002809921745210886, "fcm_dpo/delta": 0.07570519298315048, "fcm_dpo/margin": 187.93719482421875, "fcm_dpo/q_t": 0.38831907510757446, "grad_norm": 37.678199768066406, "learning_rate": 9.326904852647344e-08, "logits/chosen": -0.4942491054534912, "logits/rejected": -0.48562514781951904, "logps/chosen": -381.1417236328125, "logps/ref_chosen": -71.303466796875, "logps/ref_rejected": -95.6275405883789, "logps/rejected": -593.4030151367188, "loss": 1.0468, "margin_dpo/margin_mean": 187.93719482421875, "margin_dpo/margin_std": 266.2950744628906, "step": 508 }, { "epoch": 0.7474302496328928, "fcm_dpo/beta": 0.0028523635119199753, "fcm_dpo/delta": 0.05893818289041519, "fcm_dpo/margin": 190.60777282714844, "fcm_dpo/q_t": 0.38169676065444946, "grad_norm": 39.39438247680664, "learning_rate": 9.227130018803195e-08, "logits/chosen": -0.5167714357376099, "logits/rejected": -0.5027569532394409, "logps/chosen": -326.60260009765625, "logps/ref_chosen": -63.81895065307617, "logps/ref_rejected": -83.25643920898438, "logps/rejected": -536.6478881835938, "loss": 1.0086, "margin_dpo/margin_mean": 190.60775756835938, "margin_dpo/margin_std": 227.0200653076172, "step": 509 }, { "epoch": 0.748898678414097, "fcm_dpo/beta": 0.002829122357070446, "fcm_dpo/delta": -0.050209492444992065, "fcm_dpo/margin": 228.68527221679688, "fcm_dpo/q_t": 0.35418280959129333, "grad_norm": 33.37362289428711, "learning_rate": 9.127770814751932e-08, "logits/chosen": -0.47170090675354004, "logits/rejected": -0.4742484390735626, "logps/chosen": -364.83221435546875, "logps/ref_chosen": -51.878448486328125, "logps/ref_rejected": -102.7651596069336, "logps/rejected": -644.4041748046875, "loss": 0.9209, "margin_dpo/margin_mean": 228.68527221679688, "margin_dpo/margin_std": 210.78187561035156, "step": 510 }, { "epoch": 0.750367107195301, "fcm_dpo/beta": 0.0028122500516474247, "fcm_dpo/delta": -0.04460298642516136, "fcm_dpo/margin": 228.15643310546875, "fcm_dpo/q_t": 0.3623583912849426, "grad_norm": 28.944910049438477, "learning_rate": 9.028829858700973e-08, "logits/chosen": -0.4922507405281067, "logits/rejected": -0.4849390387535095, "logps/chosen": -339.91571044921875, "logps/ref_chosen": -60.23811721801758, "logps/ref_rejected": -92.85676574707031, "logps/rejected": -600.6907958984375, "loss": 0.9745, "margin_dpo/margin_mean": 228.15643310546875, "margin_dpo/margin_std": 276.27374267578125, "step": 511 }, { "epoch": 0.7518355359765051, "fcm_dpo/beta": 0.002760728355497122, "fcm_dpo/delta": -0.08452171087265015, "fcm_dpo/margin": 245.87234497070312, "fcm_dpo/q_t": 0.3484812378883362, "grad_norm": 27.64395523071289, "learning_rate": 8.930309757836516e-08, "logits/chosen": -0.4628169536590576, "logits/rejected": -0.44934043288230896, "logps/chosen": -307.468994140625, "logps/ref_chosen": -54.905494689941406, "logps/ref_rejected": -81.87586975097656, "logps/rejected": -580.3117065429688, "loss": 0.8986, "margin_dpo/margin_mean": 245.87234497070312, "margin_dpo/margin_std": 219.15728759765625, "step": 512 }, { "epoch": 0.7533039647577092, "fcm_dpo/beta": 0.002783479169011116, "fcm_dpo/delta": 0.1268652230501175, "fcm_dpo/margin": 172.15696716308594, "fcm_dpo/q_t": 0.39400118589401245, "grad_norm": 46.16716766357422, "learning_rate": 8.832213108254863e-08, "logits/chosen": -0.4615207314491272, "logits/rejected": -0.42703330516815186, "logps/chosen": -352.5402526855469, "logps/ref_chosen": -64.91644287109375, "logps/ref_rejected": -76.06245422363281, "logps/rejected": -535.84326171875, "loss": 1.0662, "margin_dpo/margin_mean": 172.15696716308594, "margin_dpo/margin_std": 235.25601196289062, "step": 513 }, { "epoch": 0.7547723935389133, "fcm_dpo/beta": 0.0028532061260193586, "fcm_dpo/delta": 0.07234474271535873, "fcm_dpo/margin": 186.24154663085938, "fcm_dpo/q_t": 0.38589781522750854, "grad_norm": 40.779319763183594, "learning_rate": 8.734542494893954e-08, "logits/chosen": -0.47675463557243347, "logits/rejected": -0.44973114132881165, "logps/chosen": -357.24749755859375, "logps/ref_chosen": -74.22957611083984, "logps/ref_rejected": -78.945556640625, "logps/rejected": -548.2050170898438, "loss": 1.0352, "margin_dpo/margin_mean": 186.24154663085938, "margin_dpo/margin_std": 250.20767211914062, "step": 514 }, { "epoch": 0.7562408223201175, "fcm_dpo/beta": 0.0028960055205971003, "fcm_dpo/delta": 0.09402909129858017, "fcm_dpo/margin": 176.40408325195312, "fcm_dpo/q_t": 0.3851974606513977, "grad_norm": 31.811119079589844, "learning_rate": 8.637300491465272e-08, "logits/chosen": -0.47848212718963623, "logits/rejected": -0.47844403982162476, "logps/chosen": -294.98284912109375, "logps/ref_chosen": -50.40156555175781, "logps/ref_rejected": -87.09774780273438, "logps/rejected": -508.0831604003906, "loss": 1.0331, "margin_dpo/margin_mean": 176.40408325195312, "margin_dpo/margin_std": 221.04446411132812, "step": 515 }, { "epoch": 0.7577092511013216, "fcm_dpo/beta": 0.00288777775131166, "fcm_dpo/delta": -0.08516449481248856, "fcm_dpo/margin": 235.27685546875, "fcm_dpo/q_t": 0.3478805422782898, "grad_norm": 32.073699951171875, "learning_rate": 8.540489660386064e-08, "logits/chosen": -0.50160813331604, "logits/rejected": -0.5122185945510864, "logps/chosen": -312.9991455078125, "logps/ref_chosen": -64.64956665039062, "logps/ref_rejected": -111.72237396240234, "logps/rejected": -595.3488159179688, "loss": 0.8995, "margin_dpo/margin_mean": 235.27685546875, "margin_dpo/margin_std": 209.3588104248047, "step": 516 }, { "epoch": 0.7591776798825257, "fcm_dpo/beta": 0.002824044320732355, "fcm_dpo/delta": -0.06978069990873337, "fcm_dpo/margin": 235.31272888183594, "fcm_dpo/q_t": 0.35607290267944336, "grad_norm": 34.7307014465332, "learning_rate": 8.444112552711752e-08, "logits/chosen": -0.498099148273468, "logits/rejected": -0.47780922055244446, "logps/chosen": -338.8596496582031, "logps/ref_chosen": -60.913551330566406, "logps/ref_rejected": -89.08308410644531, "logps/rejected": -602.3419189453125, "loss": 0.9334, "margin_dpo/margin_mean": 235.31271362304688, "margin_dpo/margin_std": 250.22042846679688, "step": 517 }, { "epoch": 0.7606461086637298, "fcm_dpo/beta": 0.002825916511937976, "fcm_dpo/delta": 0.01667827181518078, "fcm_dpo/margin": 206.73744201660156, "fcm_dpo/q_t": 0.37291136384010315, "grad_norm": 30.245616912841797, "learning_rate": 8.348171708068747e-08, "logits/chosen": -0.48336124420166016, "logits/rejected": -0.4803961515426636, "logps/chosen": -331.5633850097656, "logps/ref_chosen": -57.45589065551758, "logps/ref_rejected": -85.31269836425781, "logps/rejected": -566.1575927734375, "loss": 1.0028, "margin_dpo/margin_mean": 206.73744201660156, "margin_dpo/margin_std": 258.0599365234375, "step": 518 }, { "epoch": 0.762114537444934, "fcm_dpo/beta": 0.0028772333171218634, "fcm_dpo/delta": 0.08668302744626999, "fcm_dpo/margin": 179.88906860351562, "fcm_dpo/q_t": 0.3874738812446594, "grad_norm": 30.495445251464844, "learning_rate": 8.25266965458755e-08, "logits/chosen": -0.49129921197891235, "logits/rejected": -0.470879465341568, "logps/chosen": -341.64581298828125, "logps/ref_chosen": -74.06331634521484, "logps/ref_rejected": -104.44416809082031, "logps/rejected": -551.9157104492188, "loss": 1.035, "margin_dpo/margin_mean": 179.88906860351562, "margin_dpo/margin_std": 234.95318603515625, "step": 519 }, { "epoch": 0.7635829662261381, "fcm_dpo/beta": 0.002877956721931696, "fcm_dpo/delta": 0.02885432541370392, "fcm_dpo/margin": 198.79312133789062, "fcm_dpo/q_t": 0.3750660717487335, "grad_norm": 30.533952713012695, "learning_rate": 8.15760890883607e-08, "logits/chosen": -0.4489268660545349, "logits/rejected": -0.4413859248161316, "logps/chosen": -339.8787841796875, "logps/ref_chosen": -70.2998275756836, "logps/ref_rejected": -99.98133850097656, "logps/rejected": -568.3533935546875, "loss": 0.9855, "margin_dpo/margin_mean": 198.79312133789062, "margin_dpo/margin_std": 215.6772918701172, "step": 520 }, { "epoch": 0.7650513950073421, "fcm_dpo/beta": 0.0029383916407823563, "fcm_dpo/delta": 0.026043057441711426, "fcm_dpo/margin": 195.25540161132812, "fcm_dpo/q_t": 0.3743501305580139, "grad_norm": 37.878440856933594, "learning_rate": 8.062991975753378e-08, "logits/chosen": -0.46159619092941284, "logits/rejected": -0.44562065601348877, "logps/chosen": -301.5888366699219, "logps/ref_chosen": -58.14292526245117, "logps/ref_rejected": -83.28060913085938, "logps/rejected": -521.98193359375, "loss": 0.9898, "margin_dpo/margin_mean": 195.2554168701172, "margin_dpo/margin_std": 218.83047485351562, "step": 521 }, { "epoch": 0.7665198237885462, "fcm_dpo/beta": 0.002963225357234478, "fcm_dpo/delta": 0.09250913560390472, "fcm_dpo/margin": 172.7684326171875, "fcm_dpo/q_t": 0.3848268687725067, "grad_norm": 41.32555389404297, "learning_rate": 7.968821348583643e-08, "logits/chosen": -0.45410579442977905, "logits/rejected": -0.44396814703941345, "logps/chosen": -321.1343994140625, "logps/ref_chosen": -46.54766845703125, "logps/ref_rejected": -66.01388549804688, "logps/rejected": -513.3690185546875, "loss": 1.0304, "margin_dpo/margin_mean": 172.7684326171875, "margin_dpo/margin_std": 213.2279052734375, "step": 522 }, { "epoch": 0.7679882525697503, "fcm_dpo/beta": 0.002987142652273178, "fcm_dpo/delta": 0.034964967519044876, "fcm_dpo/margin": 189.8165283203125, "fcm_dpo/q_t": 0.3737178146839142, "grad_norm": 38.12350082397461, "learning_rate": 7.875099508810484e-08, "logits/chosen": -0.5127274394035339, "logits/rejected": -0.48612135648727417, "logps/chosen": -336.416015625, "logps/ref_chosen": -61.76960372924805, "logps/ref_rejected": -83.76141357421875, "logps/rejected": -548.2243041992188, "loss": 1.0043, "margin_dpo/margin_mean": 189.8165283203125, "margin_dpo/margin_std": 229.8765869140625, "step": 523 }, { "epoch": 0.7694566813509545, "fcm_dpo/beta": 0.002963773440569639, "fcm_dpo/delta": -0.03581948205828667, "fcm_dpo/margin": 213.60824584960938, "fcm_dpo/q_t": 0.3576499819755554, "grad_norm": 43.33149719238281, "learning_rate": 7.781828926091535e-08, "logits/chosen": -0.5263937711715698, "logits/rejected": -0.49376797676086426, "logps/chosen": -350.075439453125, "logps/ref_chosen": -78.0720443725586, "logps/ref_rejected": -81.30198669433594, "logps/rejected": -566.9136352539062, "loss": 0.9439, "margin_dpo/margin_mean": 213.60824584960938, "margin_dpo/margin_std": 214.0634765625, "step": 524 }, { "epoch": 0.7709251101321586, "fcm_dpo/beta": 0.0028939242474734783, "fcm_dpo/delta": -0.140090212225914, "fcm_dpo/margin": 251.78009033203125, "fcm_dpo/q_t": 0.3388723134994507, "grad_norm": 34.86455154418945, "learning_rate": 7.689012058193384e-08, "logits/chosen": -0.4875791668891907, "logits/rejected": -0.4962403476238251, "logps/chosen": -318.8263244628906, "logps/ref_chosen": -50.827857971191406, "logps/ref_rejected": -100.05294036865234, "logps/rejected": -619.83154296875, "loss": 0.8763, "margin_dpo/margin_mean": 251.7801055908203, "margin_dpo/margin_std": 222.5386505126953, "step": 525 }, { "epoch": 0.7723935389133627, "fcm_dpo/beta": 0.0028644688427448273, "fcm_dpo/delta": -0.102485790848732, "fcm_dpo/margin": 242.66949462890625, "fcm_dpo/q_t": 0.34745296835899353, "grad_norm": 30.96253776550293, "learning_rate": 7.596651350926836e-08, "logits/chosen": -0.5117580890655518, "logits/rejected": -0.48653119802474976, "logps/chosen": -344.20849609375, "logps/ref_chosen": -63.167236328125, "logps/ref_rejected": -86.30934143066406, "logps/rejected": -610.0201416015625, "loss": 0.9091, "margin_dpo/margin_mean": 242.6695098876953, "margin_dpo/margin_std": 234.477294921875, "step": 526 }, { "epoch": 0.7738619676945668, "fcm_dpo/beta": 0.002855823142454028, "fcm_dpo/delta": 0.0754547268152237, "fcm_dpo/margin": 185.0948486328125, "fcm_dpo/q_t": 0.38190531730651855, "grad_norm": 37.24220275878906, "learning_rate": 7.504749238082414e-08, "logits/chosen": -0.551037073135376, "logits/rejected": -0.5188230276107788, "logps/chosen": -381.26397705078125, "logps/ref_chosen": -71.12867736816406, "logps/ref_rejected": -78.3425521850586, "logps/rejected": -573.57275390625, "loss": 1.004, "margin_dpo/margin_mean": 185.0948486328125, "margin_dpo/margin_std": 205.111572265625, "step": 527 }, { "epoch": 0.775330396475771, "fcm_dpo/beta": 0.0028600182849913836, "fcm_dpo/delta": -0.007810359820723534, "fcm_dpo/margin": 212.27090454101562, "fcm_dpo/q_t": 0.37373942136764526, "grad_norm": 35.34698486328125, "learning_rate": 7.413308141366254e-08, "logits/chosen": -0.5303002595901489, "logits/rejected": -0.509202241897583, "logps/chosen": -384.4117736816406, "logps/ref_chosen": -68.0894546508789, "logps/ref_rejected": -93.91006469726562, "logps/rejected": -622.5032958984375, "loss": 1.0079, "margin_dpo/margin_mean": 212.27090454101562, "margin_dpo/margin_std": 285.313720703125, "step": 528 }, { "epoch": 0.7767988252569751, "fcm_dpo/beta": 0.002924061380326748, "fcm_dpo/delta": 0.13557088375091553, "fcm_dpo/margin": 161.04995727539062, "fcm_dpo/q_t": 0.3966936767101288, "grad_norm": 46.321571350097656, "learning_rate": 7.322330470336313e-08, "logits/chosen": -0.49917709827423096, "logits/rejected": -0.49793556332588196, "logps/chosen": -410.2076416015625, "logps/ref_chosen": -55.57495880126953, "logps/ref_rejected": -89.20909118652344, "logps/rejected": -604.8917236328125, "loss": 1.0838, "margin_dpo/margin_mean": 161.04995727539062, "margin_dpo/margin_std": 237.48895263671875, "step": 529 }, { "epoch": 0.7782672540381792, "fcm_dpo/beta": 0.0028766831383109093, "fcm_dpo/delta": -0.14135703444480896, "fcm_dpo/margin": 253.9014892578125, "fcm_dpo/q_t": 0.3434082269668579, "grad_norm": 33.014739990234375, "learning_rate": 7.231818622338822e-08, "logits/chosen": -0.47442853450775146, "logits/rejected": -0.45949307084083557, "logps/chosen": -333.4599609375, "logps/ref_chosen": -47.601417541503906, "logps/ref_rejected": -87.2845230102539, "logps/rejected": -627.0445556640625, "loss": 0.9293, "margin_dpo/margin_mean": 253.90150451660156, "margin_dpo/margin_std": 285.0568542480469, "step": 530 }, { "epoch": 0.7797356828193832, "fcm_dpo/beta": 0.0028520459309220314, "fcm_dpo/delta": -0.02389775589108467, "fcm_dpo/margin": 218.2183837890625, "fcm_dpo/q_t": 0.36491817235946655, "grad_norm": 39.8703498840332, "learning_rate": 7.141774982445147e-08, "logits/chosen": -0.5686095952987671, "logits/rejected": -0.5440300107002258, "logps/chosen": -381.2401123046875, "logps/ref_chosen": -55.246063232421875, "logps/ref_rejected": -70.60598754882812, "logps/rejected": -614.8184204101562, "loss": 0.9667, "margin_dpo/margin_mean": 218.2183837890625, "margin_dpo/margin_std": 248.57521057128906, "step": 531 }, { "epoch": 0.7812041116005873, "fcm_dpo/beta": 0.0027955423574894667, "fcm_dpo/delta": -0.043103598058223724, "fcm_dpo/margin": 227.61962890625, "fcm_dpo/q_t": 0.36520931124687195, "grad_norm": 60.421783447265625, "learning_rate": 7.052201923388953e-08, "logits/chosen": -0.5076245069503784, "logits/rejected": -0.478635311126709, "logps/chosen": -414.247314453125, "logps/ref_chosen": -70.28601837158203, "logps/ref_rejected": -86.5913314819336, "logps/rejected": -658.1722412109375, "loss": 0.998, "margin_dpo/margin_mean": 227.61961364746094, "margin_dpo/margin_std": 282.49871826171875, "step": 532 }, { "epoch": 0.7826725403817915, "fcm_dpo/beta": 0.0028768163174390793, "fcm_dpo/delta": 0.06540602445602417, "fcm_dpo/margin": 186.47171020507812, "fcm_dpo/q_t": 0.38702893257141113, "grad_norm": 41.66936111450195, "learning_rate": 6.963101805503646e-08, "logits/chosen": -0.4856870174407959, "logits/rejected": -0.45800524950027466, "logps/chosen": -395.514892578125, "logps/ref_chosen": -64.8551025390625, "logps/ref_rejected": -76.58805847167969, "logps/rejected": -593.7196044921875, "loss": 1.0558, "margin_dpo/margin_mean": 186.47171020507812, "margin_dpo/margin_std": 268.0407409667969, "step": 533 }, { "epoch": 0.7841409691629956, "fcm_dpo/beta": 0.0028785793110728264, "fcm_dpo/delta": 0.0326891764998436, "fcm_dpo/margin": 197.72434997558594, "fcm_dpo/q_t": 0.37530991435050964, "grad_norm": 38.83979034423828, "learning_rate": 6.874476976660184e-08, "logits/chosen": -0.510763943195343, "logits/rejected": -0.49589037895202637, "logps/chosen": -392.6068115234375, "logps/ref_chosen": -60.119388580322266, "logps/ref_rejected": -78.54347229003906, "logps/rejected": -608.7552490234375, "loss": 1.0056, "margin_dpo/margin_mean": 197.724365234375, "margin_dpo/margin_std": 245.38973999023438, "step": 534 }, { "epoch": 0.7856093979441997, "fcm_dpo/beta": 0.0028537982143461704, "fcm_dpo/delta": -0.11794188618659973, "fcm_dpo/margin": 248.56651306152344, "fcm_dpo/q_t": 0.34491124749183655, "grad_norm": 37.186885833740234, "learning_rate": 6.786329772205246e-08, "logits/chosen": -0.5178133845329285, "logits/rejected": -0.49928510189056396, "logps/chosen": -351.3582763671875, "logps/ref_chosen": -54.330238342285156, "logps/ref_rejected": -96.30763244628906, "logps/rejected": -641.9022216796875, "loss": 0.899, "margin_dpo/margin_mean": 248.56649780273438, "margin_dpo/margin_std": 232.72039794921875, "step": 535 }, { "epoch": 0.7870778267254038, "fcm_dpo/beta": 0.002728927880525589, "fcm_dpo/delta": -0.21693715453147888, "fcm_dpo/margin": 293.05352783203125, "fcm_dpo/q_t": 0.33220839500427246, "grad_norm": 31.354650497436523, "learning_rate": 6.698662514899638e-08, "logits/chosen": -0.5063174962997437, "logits/rejected": -0.5071820616722107, "logps/chosen": -320.23358154296875, "logps/ref_chosen": -47.08053207397461, "logps/ref_rejected": -89.09783935546875, "logps/rejected": -655.304443359375, "loss": 0.8628, "margin_dpo/margin_mean": 293.05352783203125, "margin_dpo/margin_std": 300.212890625, "step": 536 }, { "epoch": 0.788546255506608, "fcm_dpo/beta": 0.0027053924277424812, "fcm_dpo/delta": -0.014062870293855667, "fcm_dpo/margin": 226.46954345703125, "fcm_dpo/q_t": 0.3674009442329407, "grad_norm": 38.02225875854492, "learning_rate": 6.611477514857114e-08, "logits/chosen": -0.47698020935058594, "logits/rejected": -0.4474596381187439, "logps/chosen": -329.1002197265625, "logps/ref_chosen": -57.747467041015625, "logps/ref_rejected": -70.43838500976562, "logps/rejected": -568.2606811523438, "loss": 0.9875, "margin_dpo/margin_mean": 226.46954345703125, "margin_dpo/margin_std": 272.7801513671875, "step": 537 }, { "epoch": 0.7900146842878121, "fcm_dpo/beta": 0.0026679965667426586, "fcm_dpo/delta": -0.04222983494400978, "fcm_dpo/margin": 239.66842651367188, "fcm_dpo/q_t": 0.35988813638687134, "grad_norm": 38.96491241455078, "learning_rate": 6.524777069483525e-08, "logits/chosen": -0.5299701690673828, "logits/rejected": -0.5038875341415405, "logps/chosen": -405.4573669433594, "logps/ref_chosen": -66.41594696044922, "logps/ref_rejected": -84.22808837890625, "logps/rejected": -662.9379272460938, "loss": 0.9442, "margin_dpo/margin_mean": 239.66842651367188, "margin_dpo/margin_std": 251.71209716796875, "step": 538 }, { "epoch": 0.7914831130690162, "fcm_dpo/beta": 0.0026902840472757816, "fcm_dpo/delta": 0.07448217272758484, "fcm_dpo/margin": 196.82864379882812, "fcm_dpo/q_t": 0.3829989731311798, "grad_norm": 29.287826538085938, "learning_rate": 6.438563463416221e-08, "logits/chosen": -0.5244389176368713, "logits/rejected": -0.4966394901275635, "logps/chosen": -368.57196044921875, "logps/ref_chosen": -58.492855072021484, "logps/ref_rejected": -91.85395050048828, "logps/rejected": -598.76171875, "loss": 1.0163, "margin_dpo/margin_mean": 196.82864379882812, "margin_dpo/margin_std": 235.43844604492188, "step": 539 }, { "epoch": 0.7929515418502202, "fcm_dpo/beta": 0.002647720742970705, "fcm_dpo/delta": -0.18606501817703247, "fcm_dpo/margin": 291.5929260253906, "fcm_dpo/q_t": 0.33625712990760803, "grad_norm": 37.2298698425293, "learning_rate": 6.352838968463919e-08, "logits/chosen": -0.5106328129768372, "logits/rejected": -0.5080406665802002, "logps/chosen": -361.15386962890625, "logps/ref_chosen": -63.482513427734375, "logps/ref_rejected": -116.42999267578125, "logps/rejected": -705.6942749023438, "loss": 0.8864, "margin_dpo/margin_mean": 291.5928955078125, "margin_dpo/margin_std": 289.0450744628906, "step": 540 }, { "epoch": 0.7944199706314243, "fcm_dpo/beta": 0.0026333509013056755, "fcm_dpo/delta": 0.12476900219917297, "fcm_dpo/margin": 182.57830810546875, "fcm_dpo/q_t": 0.3981369137763977, "grad_norm": 36.667198181152344, "learning_rate": 6.267605843546767e-08, "logits/chosen": -0.5467498302459717, "logits/rejected": -0.5339305400848389, "logps/chosen": -430.84771728515625, "logps/ref_chosen": -78.28036499023438, "logps/ref_rejected": -103.273681640625, "logps/rejected": -638.4193725585938, "loss": 1.0807, "margin_dpo/margin_mean": 182.57830810546875, "margin_dpo/margin_std": 271.44097900390625, "step": 541 }, { "epoch": 0.7958883994126285, "fcm_dpo/beta": 0.002633909694850445, "fcm_dpo/delta": -0.05343947932124138, "fcm_dpo/margin": 246.30075073242188, "fcm_dpo/q_t": 0.3598957359790802, "grad_norm": 37.33884811401367, "learning_rate": 6.182866334636888e-08, "logits/chosen": -0.5174318552017212, "logits/rejected": -0.5168588161468506, "logps/chosen": -388.119384765625, "logps/ref_chosen": -57.48497009277344, "logps/ref_rejected": -96.47506713867188, "logps/rejected": -673.4102783203125, "loss": 0.9588, "margin_dpo/margin_mean": 246.30075073242188, "margin_dpo/margin_std": 279.5341796875, "step": 542 }, { "epoch": 0.7973568281938326, "fcm_dpo/beta": 0.0026966403238475323, "fcm_dpo/delta": 0.13497576117515564, "fcm_dpo/margin": 174.83229064941406, "fcm_dpo/q_t": 0.40193819999694824, "grad_norm": 33.94786834716797, "learning_rate": 6.098622674699147e-08, "logits/chosen": -0.519938588142395, "logits/rejected": -0.5251273512840271, "logps/chosen": -390.4471435546875, "logps/ref_chosen": -60.61750793457031, "logps/ref_rejected": -105.59896850585938, "logps/rejected": -610.2608642578125, "loss": 1.099, "margin_dpo/margin_mean": 174.83230590820312, "margin_dpo/margin_std": 286.04852294921875, "step": 543 }, { "epoch": 0.7988252569750367, "fcm_dpo/beta": 0.002676740288734436, "fcm_dpo/delta": -0.09121545404195786, "fcm_dpo/margin": 255.8384246826172, "fcm_dpo/q_t": 0.3496483564376831, "grad_norm": 39.29571533203125, "learning_rate": 6.01487708363232e-08, "logits/chosen": -0.4991987347602844, "logits/rejected": -0.4953753650188446, "logps/chosen": -407.407470703125, "logps/ref_chosen": -59.642303466796875, "logps/ref_rejected": -100.95469665527344, "logps/rejected": -704.5582275390625, "loss": 0.9054, "margin_dpo/margin_mean": 255.8384246826172, "margin_dpo/margin_std": 249.53448486328125, "step": 544 }, { "epoch": 0.8002936857562408, "fcm_dpo/beta": 0.0026431018486618996, "fcm_dpo/delta": -0.0723901242017746, "fcm_dpo/margin": 252.58734130859375, "fcm_dpo/q_t": 0.3531061112880707, "grad_norm": 27.791589736938477, "learning_rate": 5.9316317682106294e-08, "logits/chosen": -0.49018025398254395, "logits/rejected": -0.4865438938140869, "logps/chosen": -402.95513916015625, "logps/ref_chosen": -67.64859771728516, "logps/ref_rejected": -95.90800476074219, "logps/rejected": -683.8018798828125, "loss": 0.9299, "margin_dpo/margin_mean": 252.58734130859375, "margin_dpo/margin_std": 258.5457763671875, "step": 545 }, { "epoch": 0.801762114537445, "fcm_dpo/beta": 0.0026444008108228445, "fcm_dpo/delta": 0.05075571686029434, "fcm_dpo/margin": 208.76841735839844, "fcm_dpo/q_t": 0.37396320700645447, "grad_norm": 34.60557174682617, "learning_rate": 5.848888922025552e-08, "logits/chosen": -0.48540985584259033, "logits/rejected": -0.47076770663261414, "logps/chosen": -358.11322021484375, "logps/ref_chosen": -50.744232177734375, "logps/ref_rejected": -81.86622619628906, "logps/rejected": -598.003662109375, "loss": 0.9881, "margin_dpo/margin_mean": 208.7684326171875, "margin_dpo/margin_std": 221.3150634765625, "step": 546 }, { "epoch": 0.8032305433186491, "fcm_dpo/beta": 0.002690389985218644, "fcm_dpo/delta": 0.04553506523370743, "fcm_dpo/margin": 206.67840576171875, "fcm_dpo/q_t": 0.37883201241493225, "grad_norm": 47.69929122924805, "learning_rate": 5.7666507254280265e-08, "logits/chosen": -0.5229824185371399, "logits/rejected": -0.5123274326324463, "logps/chosen": -416.7408447265625, "logps/ref_chosen": -73.6877212524414, "logps/ref_rejected": -90.76136779785156, "logps/rejected": -640.492919921875, "loss": 1.0268, "margin_dpo/margin_mean": 206.67840576171875, "margin_dpo/margin_std": 272.1181945800781, "step": 547 }, { "epoch": 0.8046989720998532, "fcm_dpo/beta": 0.0026915257330983877, "fcm_dpo/delta": 0.021523747593164444, "fcm_dpo/margin": 215.3766632080078, "fcm_dpo/q_t": 0.37684351205825806, "grad_norm": 32.53667449951172, "learning_rate": 5.684919345471029e-08, "logits/chosen": -0.5300377011299133, "logits/rejected": -0.512967586517334, "logps/chosen": -406.23773193359375, "logps/ref_chosen": -65.24634552001953, "logps/ref_rejected": -94.11807250976562, "logps/rejected": -650.486083984375, "loss": 1.0016, "margin_dpo/margin_mean": 215.3766632080078, "margin_dpo/margin_std": 280.05401611328125, "step": 548 }, { "epoch": 0.8061674008810573, "fcm_dpo/beta": 0.00273230392485857, "fcm_dpo/delta": 0.08910378813743591, "fcm_dpo/margin": 188.5994873046875, "fcm_dpo/q_t": 0.38830626010894775, "grad_norm": 39.737220764160156, "learning_rate": 5.603696935852426e-08, "logits/chosen": -0.4929801821708679, "logits/rejected": -0.4741705656051636, "logps/chosen": -383.95849609375, "logps/ref_chosen": -49.21235656738281, "logps/ref_rejected": -73.91031646728516, "logps/rejected": -597.2559814453125, "loss": 1.0374, "margin_dpo/margin_mean": 188.5994873046875, "margin_dpo/margin_std": 250.06410217285156, "step": 549 }, { "epoch": 0.8076358296622613, "fcm_dpo/beta": 0.0027629181277006865, "fcm_dpo/delta": 0.0815601795911789, "fcm_dpo/margin": 189.18296813964844, "fcm_dpo/q_t": 0.38146719336509705, "grad_norm": 38.10478973388672, "learning_rate": 5.5229856368582376e-08, "logits/chosen": -0.4668349623680115, "logits/rejected": -0.4747786521911621, "logps/chosen": -374.70623779296875, "logps/ref_chosen": -56.80695343017578, "logps/ref_rejected": -95.12580871582031, "logps/rejected": -602.2080688476562, "loss": 1.0057, "margin_dpo/margin_mean": 189.18296813964844, "margin_dpo/margin_std": 203.63247680664062, "step": 550 }, { "epoch": 0.8091042584434655, "fcm_dpo/beta": 0.0027016932144761086, "fcm_dpo/delta": -0.18426547944545746, "fcm_dpo/margin": 284.74884033203125, "fcm_dpo/q_t": 0.3298325836658478, "grad_norm": 41.22262191772461, "learning_rate": 5.4427875753062734e-08, "logits/chosen": -0.4827745258808136, "logits/rejected": -0.48983511328697205, "logps/chosen": -374.94500732421875, "logps/ref_chosen": -59.10633087158203, "logps/ref_rejected": -111.67280578613281, "logps/rejected": -712.2603149414062, "loss": 0.8561, "margin_dpo/margin_mean": 284.74884033203125, "margin_dpo/margin_std": 247.297607421875, "step": 551 }, { "epoch": 0.8105726872246696, "fcm_dpo/beta": 0.002569342264905572, "fcm_dpo/delta": -0.22732627391815186, "fcm_dpo/margin": 313.23541259765625, "fcm_dpo/q_t": 0.32994934916496277, "grad_norm": 25.86652946472168, "learning_rate": 5.363104864490034e-08, "logits/chosen": -0.4956032335758209, "logits/rejected": -0.4835580885410309, "logps/chosen": -360.0677490234375, "logps/ref_chosen": -62.35459899902344, "logps/ref_rejected": -104.56210327148438, "logps/rejected": -715.5106201171875, "loss": 0.8604, "margin_dpo/margin_mean": 313.23541259765625, "margin_dpo/margin_std": 302.6081237792969, "step": 552 }, { "epoch": 0.8120411160058737, "fcm_dpo/beta": 0.0025847742799669504, "fcm_dpo/delta": 0.07955043762922287, "fcm_dpo/margin": 202.94935607910156, "fcm_dpo/q_t": 0.38608789443969727, "grad_norm": 36.25725555419922, "learning_rate": 5.2839396041230415e-08, "logits/chosen": -0.4990794062614441, "logits/rejected": -0.48621460795402527, "logps/chosen": -384.55615234375, "logps/ref_chosen": -68.25881958007812, "logps/ref_rejected": -98.0971450805664, "logps/rejected": -617.3438110351562, "loss": 1.0248, "margin_dpo/margin_mean": 202.94935607910156, "margin_dpo/margin_std": 253.7549285888672, "step": 553 }, { "epoch": 0.8135095447870778, "fcm_dpo/beta": 0.0025886285584419966, "fcm_dpo/delta": -0.09040947258472443, "fcm_dpo/margin": 264.222900390625, "fcm_dpo/q_t": 0.35030698776245117, "grad_norm": 31.871034622192383, "learning_rate": 5.205293880283551e-08, "logits/chosen": -0.5079976916313171, "logits/rejected": -0.4783214330673218, "logps/chosen": -389.82989501953125, "logps/ref_chosen": -67.94767761230469, "logps/ref_rejected": -89.78272247314453, "logps/rejected": -675.8878173828125, "loss": 0.9239, "margin_dpo/margin_mean": 264.222900390625, "margin_dpo/margin_std": 268.23663330078125, "step": 554 }, { "epoch": 0.8149779735682819, "fcm_dpo/beta": 0.0025238655507564545, "fcm_dpo/delta": -0.07741110026836395, "fcm_dpo/margin": 266.2066650390625, "fcm_dpo/q_t": 0.35779866576194763, "grad_norm": 37.653892517089844, "learning_rate": 5.127169765359515e-08, "logits/chosen": -0.5288249254226685, "logits/rejected": -0.5379273891448975, "logps/chosen": -376.24407958984375, "logps/ref_chosen": -53.33049011230469, "logps/ref_rejected": -108.47937774658203, "logps/rejected": -697.599609375, "loss": 0.9471, "margin_dpo/margin_mean": 266.2066650390625, "margin_dpo/margin_std": 305.3093566894531, "step": 555 }, { "epoch": 0.8164464023494861, "fcm_dpo/beta": 0.002562987618148327, "fcm_dpo/delta": 0.11464083939790726, "fcm_dpo/margin": 191.5848388671875, "fcm_dpo/q_t": 0.394670307636261, "grad_norm": 48.968658447265625, "learning_rate": 5.049569317994012e-08, "logits/chosen": -0.45059263706207275, "logits/rejected": -0.43395426869392395, "logps/chosen": -389.5137939453125, "logps/ref_chosen": -58.64447021484375, "logps/ref_rejected": -101.34040832519531, "logps/rejected": -623.7945556640625, "loss": 1.0517, "margin_dpo/margin_mean": 191.5848388671875, "margin_dpo/margin_std": 260.8348388671875, "step": 556 }, { "epoch": 0.8179148311306902, "fcm_dpo/beta": 0.002571512944996357, "fcm_dpo/delta": 0.0015640445053577423, "fcm_dpo/margin": 232.67929077148438, "fcm_dpo/q_t": 0.37047192454338074, "grad_norm": 40.12662887573242, "learning_rate": 4.9724945830310144e-08, "logits/chosen": -0.49551674723625183, "logits/rejected": -0.49951496720314026, "logps/chosen": -406.4974365234375, "logps/ref_chosen": -67.84066009521484, "logps/ref_rejected": -109.93965911865234, "logps/rejected": -681.2757568359375, "loss": 0.9819, "margin_dpo/margin_mean": 232.6792755126953, "margin_dpo/margin_std": 272.00067138671875, "step": 557 }, { "epoch": 0.8193832599118943, "fcm_dpo/beta": 0.0024929968640208244, "fcm_dpo/delta": -0.20179688930511475, "fcm_dpo/margin": 314.78643798828125, "fcm_dpo/q_t": 0.3285095691680908, "grad_norm": 42.07298278808594, "learning_rate": 4.8959475914614554e-08, "logits/chosen": -0.5093335509300232, "logits/rejected": -0.4982109069824219, "logps/chosen": -371.7452392578125, "logps/ref_chosen": -62.36824035644531, "logps/ref_rejected": -102.16102600097656, "logps/rejected": -726.324462890625, "loss": 0.866, "margin_dpo/margin_mean": 314.78643798828125, "margin_dpo/margin_std": 296.3161926269531, "step": 558 }, { "epoch": 0.8208516886930984, "fcm_dpo/beta": 0.0024765529669821262, "fcm_dpo/delta": 0.002801853232085705, "fcm_dpo/margin": 241.2025146484375, "fcm_dpo/q_t": 0.3673396706581116, "grad_norm": 28.609155654907227, "learning_rate": 4.8199303603697614e-08, "logits/chosen": -0.5868000984191895, "logits/rejected": -0.5612949728965759, "logps/chosen": -414.205322265625, "logps/ref_chosen": -60.752323150634766, "logps/ref_rejected": -93.44229125976562, "logps/rejected": -688.0977783203125, "loss": 0.9684, "margin_dpo/margin_mean": 241.2025146484375, "margin_dpo/margin_std": 260.7628173828125, "step": 559 }, { "epoch": 0.8223201174743024, "fcm_dpo/beta": 0.0025273123756051064, "fcm_dpo/delta": 0.15692362189292908, "fcm_dpo/margin": 178.22430419921875, "fcm_dpo/q_t": 0.3967198431491852, "grad_norm": 32.44375991821289, "learning_rate": 4.7444448928806615e-08, "logits/chosen": -0.5102473497390747, "logits/rejected": -0.4792303442955017, "logps/chosen": -362.985595703125, "logps/ref_chosen": -58.10382080078125, "logps/ref_rejected": -79.99122619628906, "logps/rejected": -563.0972900390625, "loss": 1.0412, "margin_dpo/margin_mean": 178.22430419921875, "margin_dpo/margin_std": 187.90447998046875, "step": 560 }, { "epoch": 0.8237885462555066, "fcm_dpo/beta": 0.0026212262455374002, "fcm_dpo/delta": 0.1312703937292099, "fcm_dpo/margin": 180.79673767089844, "fcm_dpo/q_t": 0.3930835723876953, "grad_norm": 34.945152282714844, "learning_rate": 4.669493178106432e-08, "logits/chosen": -0.46148979663848877, "logits/rejected": -0.4639427661895752, "logps/chosen": -401.359619140625, "logps/ref_chosen": -50.912879943847656, "logps/ref_rejected": -99.06856536865234, "logps/rejected": -630.3120727539062, "loss": 1.0486, "margin_dpo/margin_mean": 180.79672241210938, "margin_dpo/margin_std": 226.11021423339844, "step": 561 }, { "epoch": 0.8252569750367107, "fcm_dpo/beta": 0.0025770908687263727, "fcm_dpo/delta": -0.08462567627429962, "fcm_dpo/margin": 262.9320068359375, "fcm_dpo/q_t": 0.35398316383361816, "grad_norm": 43.5445442199707, "learning_rate": 4.5950771910944596e-08, "logits/chosen": -0.5065794587135315, "logits/rejected": -0.4925631284713745, "logps/chosen": -369.410400390625, "logps/ref_chosen": -59.46440124511719, "logps/ref_rejected": -96.54266357421875, "logps/rejected": -669.420654296875, "loss": 0.9218, "margin_dpo/margin_mean": 262.9320068359375, "margin_dpo/margin_std": 273.57208251953125, "step": 562 }, { "epoch": 0.8267254038179148, "fcm_dpo/beta": 0.002649601548910141, "fcm_dpo/delta": 0.11059555411338806, "fcm_dpo/margin": 185.759521484375, "fcm_dpo/q_t": 0.38974693417549133, "grad_norm": 38.421958923339844, "learning_rate": 4.521198892775202e-08, "logits/chosen": -0.4199938178062439, "logits/rejected": -0.41303732991218567, "logps/chosen": -402.4117736816406, "logps/ref_chosen": -60.60819625854492, "logps/ref_rejected": -94.56770324707031, "logps/rejected": -622.1307983398438, "loss": 1.0664, "margin_dpo/margin_mean": 185.759521484375, "margin_dpo/margin_std": 253.02633666992188, "step": 563 }, { "epoch": 0.8281938325991189, "fcm_dpo/beta": 0.002637513680383563, "fcm_dpo/delta": -0.013331831432878971, "fcm_dpo/margin": 232.2291717529297, "fcm_dpo/q_t": 0.35960835218429565, "grad_norm": 34.64234161376953, "learning_rate": 4.447860229910544e-08, "logits/chosen": -0.5244062542915344, "logits/rejected": -0.4892103672027588, "logps/chosen": -395.12603759765625, "logps/ref_chosen": -74.26837921142578, "logps/ref_rejected": -93.23818969726562, "logps/rejected": -646.3250732421875, "loss": 0.9258, "margin_dpo/margin_mean": 232.22918701171875, "margin_dpo/margin_std": 195.64129638671875, "step": 564 }, { "epoch": 0.8296622613803231, "fcm_dpo/beta": 0.0026156511157751083, "fcm_dpo/delta": -0.02939961664378643, "fcm_dpo/margin": 239.86215209960938, "fcm_dpo/q_t": 0.37034720182418823, "grad_norm": 41.29433059692383, "learning_rate": 4.375063135042445e-08, "logits/chosen": -0.5047638416290283, "logits/rejected": -0.48947322368621826, "logps/chosen": -412.850830078125, "logps/ref_chosen": -69.0199203491211, "logps/ref_rejected": -85.7789306640625, "logps/rejected": -669.4719848632812, "loss": 0.9974, "margin_dpo/margin_mean": 239.86215209960938, "margin_dpo/margin_std": 316.59368896484375, "step": 565 }, { "epoch": 0.8311306901615272, "fcm_dpo/beta": 0.002625175751745701, "fcm_dpo/delta": -0.04155733436346054, "fcm_dpo/margin": 242.91168212890625, "fcm_dpo/q_t": 0.36505499482154846, "grad_norm": 34.86541748046875, "learning_rate": 4.3028095264420525e-08, "logits/chosen": -0.46465927362442017, "logits/rejected": -0.46675699949264526, "logps/chosen": -391.67828369140625, "logps/ref_chosen": -66.5453109741211, "logps/ref_rejected": -103.86932373046875, "logps/rejected": -671.9140014648438, "loss": 0.9961, "margin_dpo/margin_mean": 242.91168212890625, "margin_dpo/margin_std": 304.49200439453125, "step": 566 }, { "epoch": 0.8325991189427313, "fcm_dpo/beta": 0.0026026610285043716, "fcm_dpo/delta": 0.01975536160171032, "fcm_dpo/margin": 223.38568115234375, "fcm_dpo/q_t": 0.3706703186035156, "grad_norm": 39.21678924560547, "learning_rate": 4.231101308059165e-08, "logits/chosen": -0.4942345917224884, "logits/rejected": -0.4810565710067749, "logps/chosen": -371.0935974121094, "logps/ref_chosen": -52.85829544067383, "logps/ref_rejected": -85.37095642089844, "logps/rejected": -626.991943359375, "loss": 0.9786, "margin_dpo/margin_mean": 223.38568115234375, "margin_dpo/margin_std": 245.55088806152344, "step": 567 }, { "epoch": 0.8340675477239354, "fcm_dpo/beta": 0.00256351288408041, "fcm_dpo/delta": -0.08483800292015076, "fcm_dpo/margin": 264.6658630371094, "fcm_dpo/q_t": 0.34867846965789795, "grad_norm": 28.661598205566406, "learning_rate": 4.1599403694720145e-08, "logits/chosen": -0.44072189927101135, "logits/rejected": -0.44052451848983765, "logps/chosen": -348.7388916015625, "logps/ref_chosen": -45.1923828125, "logps/ref_rejected": -89.09236907958984, "logps/rejected": -657.3046875, "loss": 0.9103, "margin_dpo/margin_mean": 264.66583251953125, "margin_dpo/margin_std": 249.04701232910156, "step": 568 }, { "epoch": 0.8355359765051396, "fcm_dpo/beta": 0.0026110014878213406, "fcm_dpo/delta": 0.09269944578409195, "fcm_dpo/margin": 195.74871826171875, "fcm_dpo/q_t": 0.3880962133407593, "grad_norm": 41.08034896850586, "learning_rate": 4.089328585837512e-08, "logits/chosen": -0.4750403165817261, "logits/rejected": -0.46098893880844116, "logps/chosen": -413.0689697265625, "logps/ref_chosen": -63.72056198120117, "logps/ref_rejected": -79.10325622558594, "logps/rejected": -624.2003173828125, "loss": 1.0448, "margin_dpo/margin_mean": 195.74871826171875, "margin_dpo/margin_std": 259.3326110839844, "step": 569 }, { "epoch": 0.8370044052863436, "fcm_dpo/beta": 0.002593663055449724, "fcm_dpo/delta": -0.01997148059308529, "fcm_dpo/margin": 238.44491577148438, "fcm_dpo/q_t": 0.3628031015396118, "grad_norm": 33.436527252197266, "learning_rate": 4.019267817841834e-08, "logits/chosen": -0.5121582746505737, "logits/rejected": -0.4927957057952881, "logps/chosen": -382.0532531738281, "logps/ref_chosen": -61.61454391479492, "logps/ref_rejected": -82.14186096191406, "logps/rejected": -641.0254516601562, "loss": 0.9467, "margin_dpo/margin_mean": 238.44491577148438, "margin_dpo/margin_std": 236.92184448242188, "step": 570 }, { "epoch": 0.8384728340675477, "fcm_dpo/beta": 0.002562489826232195, "fcm_dpo/delta": -0.06662685424089432, "fcm_dpo/margin": 258.15692138671875, "fcm_dpo/q_t": 0.3554573655128479, "grad_norm": 45.022804260253906, "learning_rate": 3.9497599116513705e-08, "logits/chosen": -0.4584212005138397, "logits/rejected": -0.4587005376815796, "logps/chosen": -373.08074951171875, "logps/ref_chosen": -53.05406188964844, "logps/ref_rejected": -91.33682250976562, "logps/rejected": -669.5204467773438, "loss": 0.9544, "margin_dpo/margin_mean": 258.15692138671875, "margin_dpo/margin_std": 291.7584228515625, "step": 571 }, { "epoch": 0.8399412628487518, "fcm_dpo/beta": 0.0025576986372470856, "fcm_dpo/delta": -0.0027488693594932556, "fcm_dpo/margin": 235.54122924804688, "fcm_dpo/q_t": 0.3760080337524414, "grad_norm": 36.91169357299805, "learning_rate": 3.880806698864086e-08, "logits/chosen": -0.4423763155937195, "logits/rejected": -0.43832147121429443, "logps/chosen": -395.4063720703125, "logps/ref_chosen": -48.45928955078125, "logps/ref_rejected": -83.55703735351562, "logps/rejected": -666.0453491210938, "loss": 1.0158, "margin_dpo/margin_mean": 235.54122924804688, "margin_dpo/margin_std": 325.2923889160156, "step": 572 }, { "epoch": 0.8414096916299559, "fcm_dpo/beta": 0.002582044806331396, "fcm_dpo/delta": 0.0030916333198547363, "fcm_dpo/margin": 230.921142578125, "fcm_dpo/q_t": 0.36895716190338135, "grad_norm": 34.18012237548828, "learning_rate": 3.812409996461275e-08, "logits/chosen": -0.4861369729042053, "logits/rejected": -0.4677738547325134, "logps/chosen": -373.53863525390625, "logps/ref_chosen": -51.62262725830078, "logps/ref_rejected": -85.32499694824219, "logps/rejected": -638.162109375, "loss": 0.9723, "margin_dpo/margin_mean": 230.921142578125, "margin_dpo/margin_std": 252.47198486328125, "step": 573 }, { "epoch": 0.8428781204111601, "fcm_dpo/beta": 0.0025438209995627403, "fcm_dpo/delta": -0.07464863359928131, "fcm_dpo/margin": 263.26629638671875, "fcm_dpo/q_t": 0.3528655171394348, "grad_norm": 30.160154342651367, "learning_rate": 3.74457160675965e-08, "logits/chosen": -0.5087519288063049, "logits/rejected": -0.5088317394256592, "logps/chosen": -369.17791748046875, "logps/ref_chosen": -51.04446029663086, "logps/ref_rejected": -92.80640411376953, "logps/rejected": -674.2061157226562, "loss": 0.9268, "margin_dpo/margin_mean": 263.26629638671875, "margin_dpo/margin_std": 267.4271240234375, "step": 574 }, { "epoch": 0.8443465491923642, "fcm_dpo/beta": 0.0025178524665534496, "fcm_dpo/delta": 0.048557281494140625, "fcm_dpo/margin": 219.31741333007812, "fcm_dpo/q_t": 0.3802528381347656, "grad_norm": 39.644771575927734, "learning_rate": 3.677293317363864e-08, "logits/chosen": -0.4346257448196411, "logits/rejected": -0.4225711226463318, "logps/chosen": -427.237060546875, "logps/ref_chosen": -71.7901382446289, "logps/ref_rejected": -95.38619995117188, "logps/rejected": -670.1505737304688, "loss": 1.0547, "margin_dpo/margin_mean": 219.31741333007812, "margin_dpo/margin_std": 301.30224609375, "step": 575 }, { "epoch": 0.8458149779735683, "fcm_dpo/beta": 0.0025599338114261627, "fcm_dpo/delta": 0.06943172216415405, "fcm_dpo/margin": 208.48135375976562, "fcm_dpo/q_t": 0.38438957929611206, "grad_norm": 35.327213287353516, "learning_rate": 3.6105769011194224e-08, "logits/chosen": -0.4814299941062927, "logits/rejected": -0.4900047481060028, "logps/chosen": -380.19384765625, "logps/ref_chosen": -54.262962341308594, "logps/ref_rejected": -100.75428009033203, "logps/rejected": -635.16650390625, "loss": 1.0487, "margin_dpo/margin_mean": 208.48135375976562, "margin_dpo/margin_std": 286.2848205566406, "step": 576 }, { "epoch": 0.8472834067547724, "fcm_dpo/beta": 0.0025888322852551937, "fcm_dpo/delta": -0.0014976090751588345, "fcm_dpo/margin": 232.30636596679688, "fcm_dpo/q_t": 0.36417728662490845, "grad_norm": 31.62501335144043, "learning_rate": 3.5444241160659304e-08, "logits/chosen": -0.4902263581752777, "logits/rejected": -0.4602411389350891, "logps/chosen": -351.259033203125, "logps/ref_chosen": -61.909706115722656, "logps/ref_rejected": -84.07069396972656, "logps/rejected": -605.7263793945312, "loss": 0.9617, "margin_dpo/margin_mean": 232.30638122558594, "margin_dpo/margin_std": 235.41848754882812, "step": 577 }, { "epoch": 0.8487518355359766, "fcm_dpo/beta": 0.0025621510576456785, "fcm_dpo/delta": -0.015007011592388153, "fcm_dpo/margin": 239.20761108398438, "fcm_dpo/q_t": 0.3611851930618286, "grad_norm": 30.414794921875, "learning_rate": 3.478836705390808e-08, "logits/chosen": -0.3913380205631256, "logits/rejected": -0.3857693076133728, "logps/chosen": -340.91937255859375, "logps/ref_chosen": -49.26368713378906, "logps/ref_rejected": -83.4362564086914, "logps/rejected": -614.299560546875, "loss": 0.9374, "margin_dpo/margin_mean": 239.20761108398438, "margin_dpo/margin_std": 213.71923828125, "step": 578 }, { "epoch": 0.8502202643171806, "fcm_dpo/beta": 0.0026131728664040565, "fcm_dpo/delta": 0.09434732794761658, "fcm_dpo/margin": 195.37240600585938, "fcm_dpo/q_t": 0.3860264718532562, "grad_norm": 37.3227424621582, "learning_rate": 3.41381639738331e-08, "logits/chosen": -0.47116217017173767, "logits/rejected": -0.4621252417564392, "logps/chosen": -367.91162109375, "logps/ref_chosen": -58.88581848144531, "logps/ref_rejected": -94.78762817382812, "logps/rejected": -599.185791015625, "loss": 1.0322, "margin_dpo/margin_mean": 195.37240600585938, "margin_dpo/margin_std": 244.83892822265625, "step": 579 }, { "epoch": 0.8516886930983847, "fcm_dpo/beta": 0.0025751434732228518, "fcm_dpo/delta": -0.1056961864233017, "fcm_dpo/margin": 270.815673828125, "fcm_dpo/q_t": 0.34934502840042114, "grad_norm": 27.840450286865234, "learning_rate": 3.349364905389032e-08, "logits/chosen": -0.4681975841522217, "logits/rejected": -0.4637313485145569, "logps/chosen": -287.3629150390625, "logps/ref_chosen": -48.70683670043945, "logps/ref_rejected": -81.7583999633789, "logps/rejected": -591.2301635742188, "loss": 0.9179, "margin_dpo/margin_mean": 270.815673828125, "margin_dpo/margin_std": 285.8489685058594, "step": 580 }, { "epoch": 0.8531571218795888, "fcm_dpo/beta": 0.002611548872664571, "fcm_dpo/delta": 0.10047388076782227, "fcm_dpo/margin": 193.2278289794922, "fcm_dpo/q_t": 0.3879375159740448, "grad_norm": 31.592443466186523, "learning_rate": 3.285483927764726e-08, "logits/chosen": -0.5031832456588745, "logits/rejected": -0.4919954836368561, "logps/chosen": -390.05816650390625, "logps/ref_chosen": -62.22235107421875, "logps/ref_rejected": -91.73568725585938, "logps/rejected": -612.79931640625, "loss": 1.0351, "margin_dpo/margin_mean": 193.22781372070312, "margin_dpo/margin_std": 241.17733764648438, "step": 581 }, { "epoch": 0.8546255506607929, "fcm_dpo/beta": 0.002609074581414461, "fcm_dpo/delta": -0.10906706750392914, "fcm_dpo/margin": 268.527587890625, "fcm_dpo/q_t": 0.34740155935287476, "grad_norm": 30.773542404174805, "learning_rate": 3.222175147833556e-08, "logits/chosen": -0.4649215638637543, "logits/rejected": -0.4725267291069031, "logps/chosen": -334.1369323730469, "logps/ref_chosen": -58.228660583496094, "logps/ref_rejected": -110.06959533691406, "logps/rejected": -654.50537109375, "loss": 0.9329, "margin_dpo/margin_mean": 268.527587890625, "margin_dpo/margin_std": 281.925048828125, "step": 582 }, { "epoch": 0.856093979441997, "fcm_dpo/beta": 0.0025950425770133734, "fcm_dpo/delta": 0.12001574784517288, "fcm_dpo/margin": 187.0833282470703, "fcm_dpo/q_t": 0.39638835191726685, "grad_norm": 38.84540557861328, "learning_rate": 3.159440233840763e-08, "logits/chosen": -0.4593685269355774, "logits/rejected": -0.44617611169815063, "logps/chosen": -376.9676513671875, "logps/ref_chosen": -56.86286163330078, "logps/ref_rejected": -88.4039306640625, "logps/rejected": -595.5921020507812, "loss": 1.0772, "margin_dpo/margin_mean": 187.0833282470703, "margin_dpo/margin_std": 274.46728515625, "step": 583 }, { "epoch": 0.8575624082232012, "fcm_dpo/beta": 0.0025689757894724607, "fcm_dpo/delta": -0.1522265374660492, "fcm_dpo/margin": 288.31964111328125, "fcm_dpo/q_t": 0.335731565952301, "grad_norm": 27.584877014160156, "learning_rate": 3.0972808389096635e-08, "logits/chosen": -0.48634427785873413, "logits/rejected": -0.47195422649383545, "logps/chosen": -321.5301818847656, "logps/ref_chosen": -56.90068054199219, "logps/ref_rejected": -97.63606262207031, "logps/rejected": -650.585205078125, "loss": 0.8585, "margin_dpo/margin_mean": 288.3196105957031, "margin_dpo/margin_std": 241.53466796875, "step": 584 }, { "epoch": 0.8590308370044053, "fcm_dpo/beta": 0.0025610639713704586, "fcm_dpo/delta": 0.02514045313000679, "fcm_dpo/margin": 225.02957153320312, "fcm_dpo/q_t": 0.37413886189460754, "grad_norm": 35.07202911376953, "learning_rate": 3.035698600998121e-08, "logits/chosen": -0.49202650785446167, "logits/rejected": -0.4775129556655884, "logps/chosen": -386.05023193359375, "logps/ref_chosen": -60.973968505859375, "logps/ref_rejected": -84.16952514648438, "logps/rejected": -634.275390625, "loss": 0.9943, "margin_dpo/margin_mean": 225.02957153320312, "margin_dpo/margin_std": 258.579345703125, "step": 585 }, { "epoch": 0.8604992657856094, "fcm_dpo/beta": 0.0026022857055068016, "fcm_dpo/delta": 0.09514912962913513, "fcm_dpo/margin": 195.86416625976562, "fcm_dpo/q_t": 0.38909363746643066, "grad_norm": 41.226749420166016, "learning_rate": 2.974695142855388e-08, "logits/chosen": -0.44712215662002563, "logits/rejected": -0.4477439820766449, "logps/chosen": -386.0147705078125, "logps/ref_chosen": -56.85559844970703, "logps/ref_rejected": -91.80261993408203, "logps/rejected": -616.825927734375, "loss": 1.0511, "margin_dpo/margin_mean": 195.86416625976562, "margin_dpo/margin_std": 270.68035888671875, "step": 586 }, { "epoch": 0.8619676945668135, "fcm_dpo/beta": 0.0026168739423155785, "fcm_dpo/delta": 0.0030817219521850348, "fcm_dpo/margin": 228.1727294921875, "fcm_dpo/q_t": 0.368298202753067, "grad_norm": 30.468931198120117, "learning_rate": 2.9142720719793122e-08, "logits/chosen": -0.48182883858680725, "logits/rejected": -0.48729079961776733, "logps/chosen": -289.46923828125, "logps/ref_chosen": -44.69159698486328, "logps/ref_rejected": -82.62385559082031, "logps/rejected": -555.57421875, "loss": 0.9793, "margin_dpo/margin_mean": 228.17274475097656, "margin_dpo/margin_std": 261.90203857421875, "step": 587 }, { "epoch": 0.8634361233480177, "fcm_dpo/beta": 0.0026173333171755075, "fcm_dpo/delta": 0.0325515940785408, "fcm_dpo/margin": 217.41366577148438, "fcm_dpo/q_t": 0.3760989308357239, "grad_norm": 35.79490280151367, "learning_rate": 2.8544309805740018e-08, "logits/chosen": -0.5010935664176941, "logits/rejected": -0.5093272924423218, "logps/chosen": -365.68310546875, "logps/ref_chosen": -50.29494857788086, "logps/ref_rejected": -107.36988067626953, "logps/rejected": -640.1717529296875, "loss": 0.9967, "margin_dpo/margin_mean": 217.41366577148438, "margin_dpo/margin_std": 257.3772277832031, "step": 588 }, { "epoch": 0.8649045521292217, "fcm_dpo/beta": 0.002576986327767372, "fcm_dpo/delta": -0.15281669795513153, "fcm_dpo/margin": 287.77593994140625, "fcm_dpo/q_t": 0.34083396196365356, "grad_norm": 29.585044860839844, "learning_rate": 2.7951734455078786e-08, "logits/chosen": -0.48489910364151, "logits/rejected": -0.47208333015441895, "logps/chosen": -363.57073974609375, "logps/ref_chosen": -59.929908752441406, "logps/ref_rejected": -111.65534973144531, "logps/rejected": -703.0721435546875, "loss": 0.8936, "margin_dpo/margin_mean": 287.77593994140625, "margin_dpo/margin_std": 287.7421875, "step": 589 }, { "epoch": 0.8663729809104258, "fcm_dpo/beta": 0.0025251221377402544, "fcm_dpo/delta": -0.0640040785074234, "fcm_dpo/margin": 261.2105712890625, "fcm_dpo/q_t": 0.3543946146965027, "grad_norm": 30.528892517089844, "learning_rate": 2.736501028272095e-08, "logits/chosen": -0.5115870237350464, "logits/rejected": -0.5127223134040833, "logps/chosen": -343.80242919921875, "logps/ref_chosen": -55.80979537963867, "logps/ref_rejected": -106.06282043457031, "logps/rejected": -655.2659912109375, "loss": 0.9296, "margin_dpo/margin_mean": 261.2105712890625, "margin_dpo/margin_std": 264.1900939941406, "step": 590 }, { "epoch": 0.8678414096916299, "fcm_dpo/beta": 0.002512441948056221, "fcm_dpo/delta": -0.01745918020606041, "fcm_dpo/margin": 245.31549072265625, "fcm_dpo/q_t": 0.3611946403980255, "grad_norm": 34.154884338378906, "learning_rate": 2.678415274939408e-08, "logits/chosen": -0.4548659920692444, "logits/rejected": -0.4189767837524414, "logps/chosen": -348.8348693847656, "logps/ref_chosen": -56.24061965942383, "logps/ref_rejected": -83.78629302978516, "logps/rejected": -621.696044921875, "loss": 0.9438, "margin_dpo/margin_mean": 245.3155059814453, "margin_dpo/margin_std": 240.31198120117188, "step": 591 }, { "epoch": 0.869309838472834, "fcm_dpo/beta": 0.0025183933321386576, "fcm_dpo/delta": 0.03698636591434479, "fcm_dpo/margin": 224.35015869140625, "fcm_dpo/q_t": 0.37596046924591064, "grad_norm": 36.619808197021484, "learning_rate": 2.6209177161234442e-08, "logits/chosen": -0.41440728306770325, "logits/rejected": -0.39510399103164673, "logps/chosen": -373.01190185546875, "logps/ref_chosen": -47.94025421142578, "logps/ref_rejected": -75.73287963867188, "logps/rejected": -625.1546630859375, "loss": 1.0503, "margin_dpo/margin_mean": 224.35015869140625, "margin_dpo/margin_std": 322.49993896484375, "step": 592 }, { "epoch": 0.8707782672540382, "fcm_dpo/beta": 0.002595474012196064, "fcm_dpo/delta": 0.17214420437812805, "fcm_dpo/margin": 167.77935791015625, "fcm_dpo/q_t": 0.4035273790359497, "grad_norm": 44.139076232910156, "learning_rate": 2.564009866938349e-08, "logits/chosen": -0.4011484682559967, "logits/rejected": -0.3781154155731201, "logps/chosen": -341.73126220703125, "logps/ref_chosen": -48.690757751464844, "logps/ref_rejected": -60.90800094604492, "logps/rejected": -521.7278442382812, "loss": 1.0966, "margin_dpo/margin_mean": 167.7793426513672, "margin_dpo/margin_std": 248.49221801757812, "step": 593 }, { "epoch": 0.8722466960352423, "fcm_dpo/beta": 0.00264896173030138, "fcm_dpo/delta": -0.018487006425857544, "fcm_dpo/margin": 231.80612182617188, "fcm_dpo/q_t": 0.36759766936302185, "grad_norm": 32.71717071533203, "learning_rate": 2.5076932269588708e-08, "logits/chosen": -0.5203255414962769, "logits/rejected": -0.49811041355133057, "logps/chosen": -332.0072021484375, "logps/ref_chosen": -54.93488693237305, "logps/ref_rejected": -86.09967803955078, "logps/rejected": -594.9780883789062, "loss": 0.9936, "margin_dpo/margin_mean": 231.80612182617188, "margin_dpo/margin_std": 272.37933349609375, "step": 594 }, { "epoch": 0.8737151248164464, "fcm_dpo/beta": 0.002612186362966895, "fcm_dpo/delta": 0.022652659565210342, "fcm_dpo/margin": 221.48776245117188, "fcm_dpo/q_t": 0.37508952617645264, "grad_norm": 33.132080078125, "learning_rate": 2.451969280180849e-08, "logits/chosen": -0.4913725256919861, "logits/rejected": -0.4721420407295227, "logps/chosen": -338.8521728515625, "logps/ref_chosen": -49.4204216003418, "logps/ref_rejected": -80.62731170654297, "logps/rejected": -591.546875, "loss": 0.9833, "margin_dpo/margin_mean": 221.48776245117188, "margin_dpo/margin_std": 257.7921142578125, "step": 595 }, { "epoch": 0.8751835535976505, "fcm_dpo/beta": 0.002664618194103241, "fcm_dpo/delta": 0.08821612596511841, "fcm_dpo/margin": 193.6279296875, "fcm_dpo/q_t": 0.38701266050338745, "grad_norm": 41.59341049194336, "learning_rate": 2.396839494982103e-08, "logits/chosen": -0.5178302526473999, "logits/rejected": -0.48967409133911133, "logps/chosen": -358.09478759765625, "logps/ref_chosen": -59.791683197021484, "logps/ref_rejected": -80.09111785888672, "logps/rejected": -572.0221557617188, "loss": 1.0508, "margin_dpo/margin_mean": 193.6279296875, "margin_dpo/margin_std": 269.0942687988281, "step": 596 }, { "epoch": 0.8766519823788547, "fcm_dpo/beta": 0.0026114017236977816, "fcm_dpo/delta": -0.08313372731208801, "fcm_dpo/margin": 258.4861755371094, "fcm_dpo/q_t": 0.3541925251483917, "grad_norm": 33.791534423828125, "learning_rate": 2.3423053240837514e-08, "logits/chosen": -0.43104076385498047, "logits/rejected": -0.4429478645324707, "logps/chosen": -359.1304931640625, "logps/ref_chosen": -57.26078796386719, "logps/ref_rejected": -100.6937255859375, "logps/rejected": -661.049560546875, "loss": 0.9329, "margin_dpo/margin_mean": 258.4861755371094, "margin_dpo/margin_std": 268.7314758300781, "step": 597 }, { "epoch": 0.8781204111600588, "fcm_dpo/beta": 0.002603329950943589, "fcm_dpo/delta": 0.007749959826469421, "fcm_dpo/margin": 227.01988220214844, "fcm_dpo/q_t": 0.36725500226020813, "grad_norm": 40.18560791015625, "learning_rate": 2.2883682045119062e-08, "logits/chosen": -0.5023446083068848, "logits/rejected": -0.4874638319015503, "logps/chosen": -370.9491882324219, "logps/ref_chosen": -52.51850509643555, "logps/ref_rejected": -89.44385528564453, "logps/rejected": -634.8944091796875, "loss": 0.9904, "margin_dpo/margin_mean": 227.01986694335938, "margin_dpo/margin_std": 255.60830688476562, "step": 598 }, { "epoch": 0.8795888399412628, "fcm_dpo/beta": 0.002633871277794242, "fcm_dpo/delta": 0.049679867923259735, "fcm_dpo/margin": 209.84701538085938, "fcm_dpo/q_t": 0.3737853169441223, "grad_norm": 29.854141235351562, "learning_rate": 2.2350295575598367e-08, "logits/chosen": -0.4743267297744751, "logits/rejected": -0.46898123621940613, "logps/chosen": -339.2664794921875, "logps/ref_chosen": -49.802677154541016, "logps/ref_rejected": -82.978515625, "logps/rejected": -582.2893676757812, "loss": 0.9804, "margin_dpo/margin_mean": 209.8470458984375, "margin_dpo/margin_std": 206.9398193359375, "step": 599 }, { "epoch": 0.8810572687224669, "fcm_dpo/beta": 0.0026514409109950066, "fcm_dpo/delta": 0.004689887166023254, "fcm_dpo/margin": 224.59609985351562, "fcm_dpo/q_t": 0.3712194561958313, "grad_norm": 36.80059051513672, "learning_rate": 2.1822907887504932e-08, "logits/chosen": -0.5273237824440002, "logits/rejected": -0.5121384263038635, "logps/chosen": -400.91802978515625, "logps/ref_chosen": -66.43487548828125, "logps/ref_rejected": -85.45649719238281, "logps/rejected": -644.5357666015625, "loss": 0.998, "margin_dpo/margin_mean": 224.59609985351562, "margin_dpo/margin_std": 278.3258056640625, "step": 600 }, { "epoch": 0.8810572687224669, "eval_fcm_dpo/beta": 0.002657976234331727, "eval_logits/chosen": -0.5060350298881531, "eval_logits/rejected": -0.4877537786960602, "eval_logps/chosen": -436.8470153808594, "eval_logps/ref_chosen": -79.05104064941406, "eval_logps/ref_rejected": -86.79793548583984, "eval_logps/rejected": -597.329345703125, "eval_loss": 0.569061279296875, "eval_margin_dpo/margin_mean": 152.7354736328125, "eval_margin_dpo/margin_std": 265.09954833984375, "eval_runtime": 39.2811, "eval_samples_per_second": 59.545, "eval_steps_per_second": 1.884, "step": 600 }, { "epoch": 0.882525697503671, "fcm_dpo/beta": 0.002635272918269038, "fcm_dpo/delta": -0.03896784782409668, "fcm_dpo/margin": 241.43040466308594, "fcm_dpo/q_t": 0.3611557185649872, "grad_norm": 41.90599822998047, "learning_rate": 2.1301532877994742e-08, "logits/chosen": -0.47748884558677673, "logits/rejected": -0.46581321954727173, "logps/chosen": -405.7626953125, "logps/ref_chosen": -59.13361358642578, "logps/ref_rejected": -94.69093322753906, "logps/rejected": -682.7504272460938, "loss": 0.9629, "margin_dpo/margin_mean": 241.43038940429688, "margin_dpo/margin_std": 275.83514404296875, "step": 601 }, { "epoch": 0.8839941262848752, "fcm_dpo/beta": 0.002600977662950754, "fcm_dpo/delta": -0.12249189615249634, "fcm_dpo/margin": 274.4678955078125, "fcm_dpo/q_t": 0.34391018748283386, "grad_norm": 34.37300491333008, "learning_rate": 2.0786184285784298e-08, "logits/chosen": -0.4986855983734131, "logits/rejected": -0.4927312731742859, "logps/chosen": -303.8367919921875, "logps/ref_chosen": -48.59352111816406, "logps/ref_rejected": -87.6685562133789, "logps/rejected": -617.3796997070312, "loss": 0.8916, "margin_dpo/margin_mean": 274.4678955078125, "margin_dpo/margin_std": 258.61602783203125, "step": 602 }, { "epoch": 0.8854625550660793, "fcm_dpo/beta": 0.002573822159320116, "fcm_dpo/delta": -0.00591304013505578, "fcm_dpo/margin": 235.269287109375, "fcm_dpo/q_t": 0.3697740435600281, "grad_norm": 44.532958984375, "learning_rate": 2.0276875690788204e-08, "logits/chosen": -0.5375059843063354, "logits/rejected": -0.5155242681503296, "logps/chosen": -360.972900390625, "logps/ref_chosen": -70.41461944580078, "logps/ref_rejected": -100.32559967041016, "logps/rejected": -626.1531982421875, "loss": 0.9847, "margin_dpo/margin_mean": 235.269287109375, "margin_dpo/margin_std": 282.26904296875, "step": 603 }, { "epoch": 0.8869309838472834, "fcm_dpo/beta": 0.002521086484193802, "fcm_dpo/delta": -0.11240590363740921, "fcm_dpo/margin": 279.26800537109375, "fcm_dpo/q_t": 0.34469854831695557, "grad_norm": 37.802711486816406, "learning_rate": 1.977362051376158e-08, "logits/chosen": -0.4898557662963867, "logits/rejected": -0.4911998510360718, "logps/chosen": -313.27166748046875, "logps/ref_chosen": -46.45808029174805, "logps/ref_rejected": -91.8544921875, "logps/rejected": -637.9360961914062, "loss": 0.9107, "margin_dpo/margin_mean": 279.26800537109375, "margin_dpo/margin_std": 277.7120056152344, "step": 604 }, { "epoch": 0.8883994126284875, "fcm_dpo/beta": 0.0025093178264796734, "fcm_dpo/delta": 0.017898954451084137, "fcm_dpo/margin": 232.27915954589844, "fcm_dpo/q_t": 0.37571078538894653, "grad_norm": 30.63051414489746, "learning_rate": 1.9276432015946446e-08, "logits/chosen": -0.5367037057876587, "logits/rejected": -0.5347219109535217, "logps/chosen": -390.2449645996094, "logps/ref_chosen": -66.24933624267578, "logps/ref_rejected": -102.30496978759766, "logps/rejected": -658.5797119140625, "loss": 0.9934, "margin_dpo/margin_mean": 232.27914428710938, "margin_dpo/margin_std": 289.82879638671875, "step": 605 }, { "epoch": 0.8898678414096917, "fcm_dpo/beta": 0.002527826000005007, "fcm_dpo/delta": -0.03301798552274704, "fcm_dpo/margin": 249.30398559570312, "fcm_dpo/q_t": 0.36113643646240234, "grad_norm": 36.33147430419922, "learning_rate": 1.8785323298722093e-08, "logits/chosen": -0.5017037391662598, "logits/rejected": -0.4989688992500305, "logps/chosen": -350.18060302734375, "logps/ref_chosen": -54.819122314453125, "logps/ref_rejected": -98.37146759033203, "logps/rejected": -643.0369873046875, "loss": 0.9437, "margin_dpo/margin_mean": 249.30398559570312, "margin_dpo/margin_std": 251.91043090820312, "step": 606 }, { "epoch": 0.8913362701908958, "fcm_dpo/beta": 0.002542435657233, "fcm_dpo/delta": 0.07603956758975983, "fcm_dpo/margin": 207.49200439453125, "fcm_dpo/q_t": 0.382717490196228, "grad_norm": 28.992673873901367, "learning_rate": 1.8300307303259904e-08, "logits/chosen": -0.5028606057167053, "logits/rejected": -0.48178744316101074, "logps/chosen": -371.967041015625, "logps/ref_chosen": -58.08403778076172, "logps/ref_rejected": -79.777099609375, "logps/rejected": -601.1521606445312, "loss": 1.0122, "margin_dpo/margin_mean": 207.4920196533203, "margin_dpo/margin_std": 238.70074462890625, "step": 607 }, { "epoch": 0.8928046989720999, "fcm_dpo/beta": 0.0025535556487739086, "fcm_dpo/delta": 0.010348204523324966, "fcm_dpo/margin": 231.10964965820312, "fcm_dpo/q_t": 0.3670482039451599, "grad_norm": 27.258190155029297, "learning_rate": 1.7821396810182437e-08, "logits/chosen": -0.536124587059021, "logits/rejected": -0.5197678208351135, "logps/chosen": -358.027099609375, "logps/ref_chosen": -57.450836181640625, "logps/ref_rejected": -94.77339172363281, "logps/rejected": -626.4592895507812, "loss": 0.9573, "margin_dpo/margin_mean": 231.10964965820312, "margin_dpo/margin_std": 227.2630615234375, "step": 608 }, { "epoch": 0.8942731277533039, "fcm_dpo/beta": 0.0025187768042087555, "fcm_dpo/delta": -0.08650342375040054, "fcm_dpo/margin": 270.2357177734375, "fcm_dpo/q_t": 0.3506140112876892, "grad_norm": 41.33991241455078, "learning_rate": 1.7348604439226617e-08, "logits/chosen": -0.5273410677909851, "logits/rejected": -0.516228199005127, "logps/chosen": -351.49822998046875, "logps/ref_chosen": -58.805355072021484, "logps/ref_rejected": -88.81600952148438, "logps/rejected": -651.74462890625, "loss": 0.9151, "margin_dpo/margin_mean": 270.2357177734375, "margin_dpo/margin_std": 274.7491760253906, "step": 609 }, { "epoch": 0.895741556534508, "fcm_dpo/beta": 0.002538463566452265, "fcm_dpo/delta": 0.07885053008794785, "fcm_dpo/margin": 206.88258361816406, "fcm_dpo/q_t": 0.3827626705169678, "grad_norm": 40.440711975097656, "learning_rate": 1.6881942648911074e-08, "logits/chosen": -0.5288474559783936, "logits/rejected": -0.4975678324699402, "logps/chosen": -368.1615905761719, "logps/ref_chosen": -65.69503784179688, "logps/ref_rejected": -83.40538787841797, "logps/rejected": -592.7545166015625, "loss": 1.0173, "margin_dpo/margin_mean": 206.88258361816406, "margin_dpo/margin_std": 242.43508911132812, "step": 610 }, { "epoch": 0.8972099853157122, "fcm_dpo/beta": 0.002536176238209009, "fcm_dpo/delta": -0.05313975363969803, "fcm_dpo/margin": 256.1167907714844, "fcm_dpo/q_t": 0.35713207721710205, "grad_norm": 35.82498550415039, "learning_rate": 1.6421423736208e-08, "logits/chosen": -0.5013567209243774, "logits/rejected": -0.5022546052932739, "logps/chosen": -356.3680725097656, "logps/ref_chosen": -52.59946823120117, "logps/ref_rejected": -86.33099365234375, "logps/rejected": -646.2163696289062, "loss": 0.9377, "margin_dpo/margin_mean": 256.1167907714844, "margin_dpo/margin_std": 261.8753662109375, "step": 611 }, { "epoch": 0.8986784140969163, "fcm_dpo/beta": 0.0024938895367085934, "fcm_dpo/delta": -0.07635810971260071, "fcm_dpo/margin": 269.17584228515625, "fcm_dpo/q_t": 0.3497973084449768, "grad_norm": 25.489120483398438, "learning_rate": 1.5967059836219042e-08, "logits/chosen": -0.49843016266822815, "logits/rejected": -0.48337119817733765, "logps/chosen": -366.25396728515625, "logps/ref_chosen": -59.32372283935547, "logps/ref_rejected": -88.31239318847656, "logps/rejected": -664.41845703125, "loss": 0.9038, "margin_dpo/margin_mean": 269.17584228515625, "margin_dpo/margin_std": 241.643310546875, "step": 612 }, { "epoch": 0.9001468428781204, "fcm_dpo/beta": 0.0024755061604082584, "fcm_dpo/delta": -0.025314215570688248, "fcm_dpo/margin": 251.95530700683594, "fcm_dpo/q_t": 0.3586321473121643, "grad_norm": 35.185272216796875, "learning_rate": 1.551886292185553e-08, "logits/chosen": -0.5439824461936951, "logits/rejected": -0.5609041452407837, "logps/chosen": -351.8963317871094, "logps/ref_chosen": -59.72996520996094, "logps/ref_rejected": -105.10752868652344, "logps/rejected": -649.229248046875, "loss": 0.9211, "margin_dpo/margin_mean": 251.95529174804688, "margin_dpo/margin_std": 217.81700134277344, "step": 613 }, { "epoch": 0.9016152716593245, "fcm_dpo/beta": 0.002444417215883732, "fcm_dpo/delta": -0.04796981438994408, "fcm_dpo/margin": 263.7240905761719, "fcm_dpo/q_t": 0.35601553320884705, "grad_norm": 34.938133239746094, "learning_rate": 1.507684480352292e-08, "logits/chosen": -0.4674204885959625, "logits/rejected": -0.49593475461006165, "logps/chosen": -351.73431396484375, "logps/ref_chosen": -52.93898010253906, "logps/ref_rejected": -104.67938232421875, "logps/rejected": -667.1987915039062, "loss": 0.932, "margin_dpo/margin_mean": 263.7240905761719, "margin_dpo/margin_std": 258.8504638671875, "step": 614 }, { "epoch": 0.9030837004405287, "fcm_dpo/beta": 0.0024889023043215275, "fcm_dpo/delta": 0.09434535354375839, "fcm_dpo/margin": 204.8232421875, "fcm_dpo/q_t": 0.38655292987823486, "grad_norm": 31.439695358276367, "learning_rate": 1.4641017128809801e-08, "logits/chosen": -0.5380944609642029, "logits/rejected": -0.5339647531509399, "logps/chosen": -367.2992248535156, "logps/ref_chosen": -65.81727600097656, "logps/ref_rejected": -95.17749786376953, "logps/rejected": -601.482666015625, "loss": 1.0194, "margin_dpo/margin_mean": 204.8232421875, "margin_dpo/margin_std": 236.53875732421875, "step": 615 }, { "epoch": 0.9045521292217328, "fcm_dpo/beta": 0.002577736508101225, "fcm_dpo/delta": 0.2006855607032776, "fcm_dpo/margin": 157.602294921875, "fcm_dpo/q_t": 0.4117230772972107, "grad_norm": 39.64345932006836, "learning_rate": 1.4211391382180637e-08, "logits/chosen": -0.5193255543708801, "logits/rejected": -0.4902590811252594, "logps/chosen": -436.3590393066406, "logps/ref_chosen": -65.13285827636719, "logps/ref_rejected": -74.70050048828125, "logps/rejected": -603.5289916992188, "loss": 1.1199, "margin_dpo/margin_mean": 157.602294921875, "margin_dpo/margin_std": 252.67433166503906, "step": 616 }, { "epoch": 0.9060205580029369, "fcm_dpo/beta": 0.0026704384945333004, "fcm_dpo/delta": 0.1679898351430893, "fcm_dpo/margin": 164.12026977539062, "fcm_dpo/q_t": 0.4053176939487457, "grad_norm": 46.60618209838867, "learning_rate": 1.378797888467345e-08, "logits/chosen": -0.5051152110099792, "logits/rejected": -0.47557491064071655, "logps/chosen": -400.8250732421875, "logps/ref_chosen": -63.005550384521484, "logps/ref_rejected": -64.234130859375, "logps/rejected": -566.1739501953125, "loss": 1.0936, "margin_dpo/margin_mean": 164.12026977539062, "margin_dpo/margin_std": 247.26060485839844, "step": 617 }, { "epoch": 0.9074889867841409, "fcm_dpo/beta": 0.002672924194484949, "fcm_dpo/delta": -0.059617504477500916, "fcm_dpo/margin": 245.20562744140625, "fcm_dpo/q_t": 0.3566618263721466, "grad_norm": 40.86361312866211, "learning_rate": 1.3370790793601371e-08, "logits/chosen": -0.4829975962638855, "logits/rejected": -0.4715951681137085, "logps/chosen": -403.0682373046875, "logps/ref_chosen": -67.10134887695312, "logps/ref_rejected": -92.15340423583984, "logps/rejected": -673.325927734375, "loss": 0.9513, "margin_dpo/margin_mean": 245.2056427001953, "margin_dpo/margin_std": 266.8808898925781, "step": 618 }, { "epoch": 0.908957415565345, "fcm_dpo/beta": 0.0026871771551668644, "fcm_dpo/delta": 0.07144356518983841, "fcm_dpo/margin": 197.93942260742188, "fcm_dpo/q_t": 0.38919079303741455, "grad_norm": 40.54890441894531, "learning_rate": 1.2959838102258535e-08, "logits/chosen": -0.5091539025306702, "logits/rejected": -0.5012193918228149, "logps/chosen": -386.0754699707031, "logps/ref_chosen": -55.978233337402344, "logps/ref_rejected": -93.1854019165039, "logps/rejected": -621.2220458984375, "loss": 1.0637, "margin_dpo/margin_mean": 197.93943786621094, "margin_dpo/margin_std": 297.71466064453125, "step": 619 }, { "epoch": 0.9104258443465492, "fcm_dpo/beta": 0.0026975993532687426, "fcm_dpo/delta": 0.0638522133231163, "fcm_dpo/margin": 199.90643310546875, "fcm_dpo/q_t": 0.3846666216850281, "grad_norm": 46.63349914550781, "learning_rate": 1.2555131639630567e-08, "logits/chosen": -0.5148506164550781, "logits/rejected": -0.4951644539833069, "logps/chosen": -366.35540771484375, "logps/ref_chosen": -59.79750061035156, "logps/ref_rejected": -78.41075134277344, "logps/rejected": -584.8751220703125, "loss": 1.0283, "margin_dpo/margin_mean": 199.90643310546875, "margin_dpo/margin_std": 263.07427978515625, "step": 620 }, { "epoch": 0.9118942731277533, "fcm_dpo/beta": 0.0026699635200202465, "fcm_dpo/delta": -0.17680245637893677, "fcm_dpo/margin": 285.97552490234375, "fcm_dpo/q_t": 0.33413994312286377, "grad_norm": 35.8352165222168, "learning_rate": 1.2156682070109086e-08, "logits/chosen": -0.5005359649658203, "logits/rejected": -0.5020206570625305, "logps/chosen": -341.9903564453125, "logps/ref_chosen": -53.93375778198242, "logps/ref_rejected": -88.36951446533203, "logps/rejected": -662.401611328125, "loss": 0.8752, "margin_dpo/margin_mean": 285.97552490234375, "margin_dpo/margin_std": 268.96636962890625, "step": 621 }, { "epoch": 0.9133627019089574, "fcm_dpo/beta": 0.0026113735511898994, "fcm_dpo/delta": 0.0012344121932983398, "fcm_dpo/margin": 229.01025390625, "fcm_dpo/q_t": 0.3693368434906006, "grad_norm": 33.05309295654297, "learning_rate": 1.1764499893210878e-08, "logits/chosen": -0.48245564103126526, "logits/rejected": -0.4655565917491913, "logps/chosen": -363.42919921875, "logps/ref_chosen": -60.28582000732422, "logps/ref_rejected": -85.51873779296875, "logps/rejected": -617.67236328125, "loss": 0.9679, "margin_dpo/margin_mean": 229.01025390625, "margin_dpo/margin_std": 250.83233642578125, "step": 622 }, { "epoch": 0.9148311306901615, "fcm_dpo/beta": 0.002670406363904476, "fcm_dpo/delta": 0.12260958552360535, "fcm_dpo/margin": 181.0537109375, "fcm_dpo/q_t": 0.39631474018096924, "grad_norm": 40.9214973449707, "learning_rate": 1.1378595443300998e-08, "logits/chosen": -0.48193132877349854, "logits/rejected": -0.4616912603378296, "logps/chosen": -400.8602294921875, "logps/ref_chosen": -64.1569595336914, "logps/ref_rejected": -85.08304595947266, "logps/rejected": -602.840087890625, "loss": 1.0693, "margin_dpo/margin_mean": 181.05372619628906, "margin_dpo/margin_std": 264.8545227050781, "step": 623 }, { "epoch": 0.9162995594713657, "fcm_dpo/beta": 0.0026587401516735554, "fcm_dpo/delta": -0.08800569176673889, "fcm_dpo/margin": 256.51275634765625, "fcm_dpo/q_t": 0.34991389513015747, "grad_norm": 39.98982620239258, "learning_rate": 1.0998978889320582e-08, "logits/chosen": -0.5772846937179565, "logits/rejected": -0.5503325462341309, "logps/chosen": -415.0184020996094, "logps/ref_chosen": -71.91862487792969, "logps/ref_rejected": -97.13203430175781, "logps/rejected": -696.7445678710938, "loss": 0.9264, "margin_dpo/margin_mean": 256.51275634765625, "margin_dpo/margin_std": 264.0021057128906, "step": 624 }, { "epoch": 0.9177679882525698, "fcm_dpo/beta": 0.0026332507841289043, "fcm_dpo/delta": -0.023220915347337723, "fcm_dpo/margin": 236.09141540527344, "fcm_dpo/q_t": 0.36515945196151733, "grad_norm": 42.99685287475586, "learning_rate": 1.0625660234518913e-08, "logits/chosen": -0.4856789708137512, "logits/rejected": -0.468561589717865, "logps/chosen": -366.0799560546875, "logps/ref_chosen": -58.342071533203125, "logps/ref_rejected": -86.09038543701172, "logps/rejected": -629.919677734375, "loss": 0.9517, "margin_dpo/margin_mean": 236.09140014648438, "margin_dpo/margin_std": 255.3983154296875, "step": 625 }, { "epoch": 0.9192364170337739, "fcm_dpo/beta": 0.0026790527626872063, "fcm_dpo/delta": 0.09501801431179047, "fcm_dpo/margin": 190.15899658203125, "fcm_dpo/q_t": 0.3896622657775879, "grad_norm": 35.053749084472656, "learning_rate": 1.0258649316189721e-08, "logits/chosen": -0.516830325126648, "logits/rejected": -0.4992465078830719, "logps/chosen": -417.03143310546875, "logps/ref_chosen": -75.11260986328125, "logps/ref_rejected": -99.188720703125, "logps/rejected": -631.2666015625, "loss": 1.0376, "margin_dpo/margin_mean": 190.15899658203125, "margin_dpo/margin_std": 242.58265686035156, "step": 626 }, { "epoch": 0.920704845814978, "fcm_dpo/beta": 0.0026376736350357533, "fcm_dpo/delta": -0.13061486184597015, "fcm_dpo/margin": 273.473388671875, "fcm_dpo/q_t": 0.35445448756217957, "grad_norm": 37.661746978759766, "learning_rate": 9.897955805412e-09, "logits/chosen": -0.4934673309326172, "logits/rejected": -0.5177198648452759, "logps/chosen": -314.9818115234375, "logps/ref_chosen": -47.74314880371094, "logps/ref_rejected": -106.75448608398438, "logps/rejected": -647.466552734375, "loss": 0.9501, "margin_dpo/margin_mean": 273.4734191894531, "margin_dpo/margin_std": 342.03204345703125, "step": 627 }, { "epoch": 0.922173274596182, "fcm_dpo/beta": 0.0025675692595541477, "fcm_dpo/delta": -0.09277446568012238, "fcm_dpo/margin": 267.0002746582031, "fcm_dpo/q_t": 0.34974387288093567, "grad_norm": 32.36410140991211, "learning_rate": 9.543589206795238e-09, "logits/chosen": -0.5133162140846252, "logits/rejected": -0.5017916560173035, "logps/chosen": -382.24267578125, "logps/ref_chosen": -60.182945251464844, "logps/ref_rejected": -101.55467224121094, "logps/rejected": -690.6146240234375, "loss": 0.9174, "margin_dpo/margin_mean": 267.0002746582031, "margin_dpo/margin_std": 266.8199768066406, "step": 628 }, { "epoch": 0.9236417033773862, "fcm_dpo/beta": 0.002568417927250266, "fcm_dpo/delta": 0.013333270326256752, "fcm_dpo/margin": 228.7070770263672, "fcm_dpo/q_t": 0.3673741817474365, "grad_norm": 39.50351333618164, "learning_rate": 9.19555885822887e-09, "logits/chosen": -0.4918622374534607, "logits/rejected": -0.47989362478256226, "logps/chosen": -393.0993957519531, "logps/ref_chosen": -64.21354675292969, "logps/ref_rejected": -91.65367126464844, "logps/rejected": -649.24658203125, "loss": 0.9515, "margin_dpo/margin_mean": 228.70706176757812, "margin_dpo/margin_std": 213.66384887695312, "step": 629 }, { "epoch": 0.9251101321585903, "fcm_dpo/beta": 0.0026454501785337925, "fcm_dpo/delta": 0.16232366859912872, "fcm_dpo/margin": 167.9291534423828, "fcm_dpo/q_t": 0.4042319357395172, "grad_norm": 37.71936798095703, "learning_rate": 8.85387393063622e-09, "logits/chosen": -0.5329476594924927, "logits/rejected": -0.5061321258544922, "logps/chosen": -354.07684326171875, "logps/ref_chosen": -59.29100036621094, "logps/ref_rejected": -83.59829711914062, "logps/rejected": -546.3132934570312, "loss": 1.0856, "margin_dpo/margin_mean": 167.9291534423828, "margin_dpo/margin_std": 248.06668090820312, "step": 630 }, { "epoch": 0.9265785609397944, "fcm_dpo/beta": 0.002692385343834758, "fcm_dpo/delta": 0.05998341366648674, "fcm_dpo/margin": 201.62759399414062, "fcm_dpo/q_t": 0.3826253414154053, "grad_norm": 39.2678337097168, "learning_rate": 8.518543427732949e-09, "logits/chosen": -0.5022942423820496, "logits/rejected": -0.48305743932724, "logps/chosen": -390.76043701171875, "logps/ref_chosen": -59.45360565185547, "logps/ref_rejected": -80.95156860351562, "logps/rejected": -613.885986328125, "loss": 1.0448, "margin_dpo/margin_mean": 201.62759399414062, "margin_dpo/margin_std": 280.7120361328125, "step": 631 }, { "epoch": 0.9280469897209985, "fcm_dpo/beta": 0.0027071628719568253, "fcm_dpo/delta": 0.05815380439162254, "fcm_dpo/margin": 201.3222198486328, "fcm_dpo/q_t": 0.3803539276123047, "grad_norm": 38.19340133666992, "learning_rate": 8.189576185789637e-09, "logits/chosen": -0.4966999888420105, "logits/rejected": -0.47419142723083496, "logps/chosen": -389.7144775390625, "logps/ref_chosen": -61.35155487060547, "logps/ref_rejected": -86.16017150878906, "logps/rejected": -615.8453369140625, "loss": 1.0424, "margin_dpo/margin_mean": 201.3222198486328, "margin_dpo/margin_std": 276.93914794921875, "step": 632 }, { "epoch": 0.9295154185022027, "fcm_dpo/beta": 0.0027928436174988747, "fcm_dpo/delta": 0.12002822011709213, "fcm_dpo/margin": 173.24588012695312, "fcm_dpo/q_t": 0.39484164118766785, "grad_norm": 57.099609375, "learning_rate": 7.866980873399015e-09, "logits/chosen": -0.5218715667724609, "logits/rejected": -0.5239961743354797, "logps/chosen": -393.3670349121094, "logps/ref_chosen": -57.27816390991211, "logps/ref_rejected": -91.58395385742188, "logps/rejected": -600.918701171875, "loss": 1.0778, "margin_dpo/margin_mean": 173.24588012695312, "margin_dpo/margin_std": 253.44149780273438, "step": 633 }, { "epoch": 0.9309838472834068, "fcm_dpo/beta": 0.002813713625073433, "fcm_dpo/delta": 0.08731327205896378, "fcm_dpo/margin": 183.8138885498047, "fcm_dpo/q_t": 0.3883476257324219, "grad_norm": 39.00724411010742, "learning_rate": 7.550765991247654e-09, "logits/chosen": -0.5172115564346313, "logits/rejected": -0.5073580741882324, "logps/chosen": -427.5379333496094, "logps/ref_chosen": -66.61896514892578, "logps/ref_rejected": -107.12564849853516, "logps/rejected": -651.8585205078125, "loss": 1.0418, "margin_dpo/margin_mean": 183.8138885498047, "margin_dpo/margin_std": 243.1580047607422, "step": 634 }, { "epoch": 0.9324522760646109, "fcm_dpo/beta": 0.0028590408619493246, "fcm_dpo/delta": 0.05354948714375496, "fcm_dpo/margin": 192.16888427734375, "fcm_dpo/q_t": 0.38193368911743164, "grad_norm": 41.086822509765625, "learning_rate": 7.240939871891699e-09, "logits/chosen": -0.5438894629478455, "logits/rejected": -0.5025356411933899, "logps/chosen": -382.0107421875, "logps/ref_chosen": -73.95551300048828, "logps/ref_rejected": -82.50045776367188, "logps/rejected": -582.724609375, "loss": 1.0099, "margin_dpo/margin_mean": 192.16888427734375, "margin_dpo/margin_std": 238.52110290527344, "step": 635 }, { "epoch": 0.933920704845815, "fcm_dpo/beta": 0.0028688067104667425, "fcm_dpo/delta": -0.04967202991247177, "fcm_dpo/margin": 225.1034698486328, "fcm_dpo/q_t": 0.35912269353866577, "grad_norm": 42.04436492919922, "learning_rate": 6.937510679537628e-09, "logits/chosen": -0.4912013113498688, "logits/rejected": -0.4742831587791443, "logps/chosen": -358.6704406738281, "logps/ref_chosen": -59.628910064697266, "logps/ref_rejected": -81.97883605957031, "logps/rejected": -606.1238403320312, "loss": 0.9427, "margin_dpo/margin_mean": 225.10345458984375, "margin_dpo/margin_std": 237.44015502929688, "step": 636 }, { "epoch": 0.9353891336270191, "fcm_dpo/beta": 0.002806810662150383, "fcm_dpo/delta": -0.06823452562093735, "fcm_dpo/margin": 236.39791870117188, "fcm_dpo/q_t": 0.3558032512664795, "grad_norm": 40.245174407958984, "learning_rate": 6.640486409826785e-09, "logits/chosen": -0.4979478716850281, "logits/rejected": -0.496917724609375, "logps/chosen": -365.4517822265625, "logps/ref_chosen": -49.652687072753906, "logps/ref_rejected": -98.40513610839844, "logps/rejected": -650.6021728515625, "loss": 0.9336, "margin_dpo/margin_mean": 236.39791870117188, "margin_dpo/margin_std": 248.04531860351562, "step": 637 }, { "epoch": 0.9368575624082232, "fcm_dpo/beta": 0.0028112411964684725, "fcm_dpo/delta": -0.03967977315187454, "fcm_dpo/margin": 226.1969757080078, "fcm_dpo/q_t": 0.357632577419281, "grad_norm": 33.084903717041016, "learning_rate": 6.349874889624962e-09, "logits/chosen": -0.4676443338394165, "logits/rejected": -0.4407140016555786, "logps/chosen": -347.86865234375, "logps/ref_chosen": -58.156639099121094, "logps/ref_rejected": -79.3014907836914, "logps/rejected": -595.2105102539062, "loss": 0.9258, "margin_dpo/margin_mean": 226.1969757080078, "margin_dpo/margin_std": 198.74960327148438, "step": 638 }, { "epoch": 0.9383259911894273, "fcm_dpo/beta": 0.0028432803228497505, "fcm_dpo/delta": 0.1298457235097885, "fcm_dpo/margin": 167.3184356689453, "fcm_dpo/q_t": 0.3954669237136841, "grad_norm": 51.73390197753906, "learning_rate": 6.065683776815933e-09, "logits/chosen": -0.4740576148033142, "logits/rejected": -0.434129536151886, "logps/chosen": -424.7391357421875, "logps/ref_chosen": -72.32319641113281, "logps/ref_rejected": -74.2749252319336, "logps/rejected": -594.00927734375, "loss": 1.0733, "margin_dpo/margin_mean": 167.3184356689453, "margin_dpo/margin_std": 239.7473602294922, "step": 639 }, { "epoch": 0.9397944199706314, "fcm_dpo/beta": 0.002794375643134117, "fcm_dpo/delta": -0.18512314558029175, "fcm_dpo/margin": 275.9643249511719, "fcm_dpo/q_t": 0.3381561040878296, "grad_norm": 32.32111358642578, "learning_rate": 5.7879205600998296e-09, "logits/chosen": -0.4964972138404846, "logits/rejected": -0.4891301393508911, "logps/chosen": -369.9661560058594, "logps/ref_chosen": -56.13436508178711, "logps/ref_rejected": -108.60014343261719, "logps/rejected": -698.396240234375, "loss": 0.8868, "margin_dpo/margin_mean": 275.9643249511719, "margin_dpo/margin_std": 286.1241149902344, "step": 640 }, { "epoch": 0.9412628487518355, "fcm_dpo/beta": 0.002765617100521922, "fcm_dpo/delta": 0.06337472051382065, "fcm_dpo/margin": 195.2603759765625, "fcm_dpo/q_t": 0.38529226183891296, "grad_norm": 54.78212356567383, "learning_rate": 5.516592558795746e-09, "logits/chosen": -0.49673527479171753, "logits/rejected": -0.4844779968261719, "logps/chosen": -419.86749267578125, "logps/ref_chosen": -64.99689483642578, "logps/ref_rejected": -86.99232482910156, "logps/rejected": -637.123291015625, "loss": 1.0383, "margin_dpo/margin_mean": 195.26036071777344, "margin_dpo/margin_std": 274.37164306640625, "step": 641 }, { "epoch": 0.9427312775330396, "fcm_dpo/beta": 0.002786886878311634, "fcm_dpo/delta": -0.06086990237236023, "fcm_dpo/margin": 235.2342987060547, "fcm_dpo/q_t": 0.3679524064064026, "grad_norm": 43.42346954345703, "learning_rate": 5.251706922648868e-09, "logits/chosen": -0.5038265585899353, "logits/rejected": -0.5031158328056335, "logps/chosen": -400.65997314453125, "logps/ref_chosen": -65.68924713134766, "logps/ref_rejected": -110.24205017089844, "logps/rejected": -680.4470825195312, "loss": 0.9975, "margin_dpo/margin_mean": 235.23428344726562, "margin_dpo/margin_std": 318.88861083984375, "step": 642 }, { "epoch": 0.9441997063142438, "fcm_dpo/beta": 0.002727922983467579, "fcm_dpo/delta": 0.0013057217001914978, "fcm_dpo/margin": 218.98019409179688, "fcm_dpo/q_t": 0.36833906173706055, "grad_norm": 39.67738342285156, "learning_rate": 4.993270631642038e-09, "logits/chosen": -0.5181025266647339, "logits/rejected": -0.5136141180992126, "logps/chosen": -359.638427734375, "logps/ref_chosen": -51.94999694824219, "logps/ref_rejected": -87.46833801269531, "logps/rejected": -614.136962890625, "loss": 0.9763, "margin_dpo/margin_mean": 218.98019409179688, "margin_dpo/margin_std": 238.99154663085938, "step": 643 }, { "epoch": 0.9456681350954479, "fcm_dpo/beta": 0.0027766618877649307, "fcm_dpo/delta": 0.07305951416492462, "fcm_dpo/margin": 191.19378662109375, "fcm_dpo/q_t": 0.38764306902885437, "grad_norm": 49.11812210083008, "learning_rate": 4.741290495811873e-09, "logits/chosen": -0.511053204536438, "logits/rejected": -0.5009315013885498, "logps/chosen": -381.0382080078125, "logps/ref_chosen": -59.017662048339844, "logps/ref_rejected": -87.13668823242188, "logps/rejected": -600.3509521484375, "loss": 1.0642, "margin_dpo/margin_mean": 191.19378662109375, "margin_dpo/margin_std": 282.96331787109375, "step": 644 }, { "epoch": 0.947136563876652, "fcm_dpo/beta": 0.0028436644934117794, "fcm_dpo/delta": 0.16144075989723206, "fcm_dpo/margin": 156.83892822265625, "fcm_dpo/q_t": 0.40858685970306396, "grad_norm": 49.89131546020508, "learning_rate": 4.495773155069299e-09, "logits/chosen": -0.4825410842895508, "logits/rejected": -0.47511184215545654, "logps/chosen": -378.183349609375, "logps/ref_chosen": -55.87602233886719, "logps/ref_rejected": -97.78080749511719, "logps/rejected": -576.9270629882812, "loss": 1.1329, "margin_dpo/margin_mean": 156.83892822265625, "margin_dpo/margin_std": 280.7100524902344, "step": 645 }, { "epoch": 0.9486049926578561, "fcm_dpo/beta": 0.002883891575038433, "fcm_dpo/delta": 0.029555249959230423, "fcm_dpo/margin": 198.30703735351562, "fcm_dpo/q_t": 0.3727361261844635, "grad_norm": 39.009521484375, "learning_rate": 4.256725079024553e-09, "logits/chosen": -0.48562538623809814, "logits/rejected": -0.46007299423217773, "logps/chosen": -364.4145812988281, "logps/ref_chosen": -61.275787353515625, "logps/ref_rejected": -77.50580596923828, "logps/rejected": -578.95166015625, "loss": 0.9785, "margin_dpo/margin_mean": 198.30703735351562, "margin_dpo/margin_std": 211.89578247070312, "step": 646 }, { "epoch": 0.9500734214390602, "fcm_dpo/beta": 0.002909876639023423, "fcm_dpo/delta": 0.007972326129674911, "fcm_dpo/margin": 203.5854949951172, "fcm_dpo/q_t": 0.3704141676425934, "grad_norm": 34.949928283691406, "learning_rate": 4.024152566816791e-09, "logits/chosen": -0.4611695408821106, "logits/rejected": -0.4655107855796814, "logps/chosen": -345.0274353027344, "logps/ref_chosen": -54.8524169921875, "logps/ref_rejected": -93.5194091796875, "logps/rejected": -587.2799072265625, "loss": 0.9958, "margin_dpo/margin_mean": 203.5854949951172, "margin_dpo/margin_std": 246.224853515625, "step": 647 }, { "epoch": 0.9515418502202643, "fcm_dpo/beta": 0.002797810360789299, "fcm_dpo/delta": -0.25719988346099854, "fcm_dpo/margin": 298.71856689453125, "fcm_dpo/q_t": 0.3337755799293518, "grad_norm": 35.38324737548828, "learning_rate": 3.798061746947995e-09, "logits/chosen": -0.4994831681251526, "logits/rejected": -0.5130796432495117, "logps/chosen": -361.3632507324219, "logps/ref_chosen": -54.17146682739258, "logps/ref_rejected": -98.7127914428711, "logps/rejected": -704.6231689453125, "loss": 0.8967, "margin_dpo/margin_mean": 298.71856689453125, "margin_dpo/margin_std": 349.7194519042969, "step": 648 }, { "epoch": 0.9530102790014684, "fcm_dpo/beta": 0.002828485332429409, "fcm_dpo/delta": 0.1316825896501541, "fcm_dpo/margin": 167.37564086914062, "fcm_dpo/q_t": 0.40139663219451904, "grad_norm": 42.975502014160156, "learning_rate": 3.5784585771215235e-09, "logits/chosen": -0.5309435725212097, "logits/rejected": -0.5060242414474487, "logps/chosen": -379.24853515625, "logps/ref_chosen": -62.480350494384766, "logps/ref_rejected": -80.07717895507812, "logps/rejected": -564.2210083007812, "loss": 1.09, "margin_dpo/margin_mean": 167.37564086914062, "margin_dpo/margin_std": 259.5540466308594, "step": 649 }, { "epoch": 0.9544787077826725, "fcm_dpo/beta": 0.002783937146887183, "fcm_dpo/delta": -0.10079372674226761, "fcm_dpo/margin": 249.080322265625, "fcm_dpo/q_t": 0.3533610701560974, "grad_norm": 34.299888610839844, "learning_rate": 3.3653488440851253e-09, "logits/chosen": -0.48052090406417847, "logits/rejected": -0.47680503129959106, "logps/chosen": -385.91802978515625, "logps/ref_chosen": -56.09281921386719, "logps/ref_rejected": -98.26483917236328, "logps/rejected": -677.1703491210938, "loss": 0.9576, "margin_dpo/margin_mean": 249.08029174804688, "margin_dpo/margin_std": 299.54718017578125, "step": 650 }, { "epoch": 0.9559471365638766, "fcm_dpo/beta": 0.0027128136716783047, "fcm_dpo/delta": -0.12499568611383438, "fcm_dpo/margin": 263.5880126953125, "fcm_dpo/q_t": 0.34237995743751526, "grad_norm": 38.067604064941406, "learning_rate": 3.158738163478475e-09, "logits/chosen": -0.5105220079421997, "logits/rejected": -0.5287940502166748, "logps/chosen": -301.7446594238281, "logps/ref_chosen": -43.42544937133789, "logps/ref_rejected": -99.95791625976562, "logps/rejected": -621.8651123046875, "loss": 0.8879, "margin_dpo/margin_mean": 263.5880126953125, "margin_dpo/margin_std": 240.21612548828125, "step": 651 }, { "epoch": 0.9574155653450808, "fcm_dpo/beta": 0.0027055356185883284, "fcm_dpo/delta": -0.019109947606921196, "fcm_dpo/margin": 228.38198852539062, "fcm_dpo/q_t": 0.36907005310058594, "grad_norm": 36.77167510986328, "learning_rate": 2.9586319796851555e-09, "logits/chosen": -0.5113554000854492, "logits/rejected": -0.504870593547821, "logps/chosen": -353.9315185546875, "logps/ref_chosen": -62.57680892944336, "logps/ref_rejected": -111.76779174804688, "logps/rejected": -631.5044555664062, "loss": 0.9818, "margin_dpo/margin_mean": 228.38198852539062, "margin_dpo/margin_std": 280.3888244628906, "step": 652 }, { "epoch": 0.9588839941262849, "fcm_dpo/beta": 0.0026833987794816494, "fcm_dpo/delta": -0.041953183710575104, "fcm_dpo/margin": 238.23910522460938, "fcm_dpo/q_t": 0.3634333610534668, "grad_norm": 42.57564163208008, "learning_rate": 2.7650355656892166e-09, "logits/chosen": -0.5182297825813293, "logits/rejected": -0.517985999584198, "logps/chosen": -397.2801818847656, "logps/ref_chosen": -61.11295700073242, "logps/ref_rejected": -103.24960327148438, "logps/rejected": -677.6558837890625, "loss": 0.9644, "margin_dpo/margin_mean": 238.23910522460938, "margin_dpo/margin_std": 277.93377685546875, "step": 653 }, { "epoch": 0.960352422907489, "fcm_dpo/beta": 0.002681623911485076, "fcm_dpo/delta": 0.03697919845581055, "fcm_dpo/margin": 210.70355224609375, "fcm_dpo/q_t": 0.3804095983505249, "grad_norm": 35.654144287109375, "learning_rate": 2.577954022936174e-09, "logits/chosen": -0.49398699402809143, "logits/rejected": -0.4834080934524536, "logps/chosen": -394.8516845703125, "logps/ref_chosen": -61.7281379699707, "logps/ref_rejected": -98.7738037109375, "logps/rejected": -642.600830078125, "loss": 1.0183, "margin_dpo/margin_mean": 210.70355224609375, "margin_dpo/margin_std": 280.0821533203125, "step": 654 }, { "epoch": 0.9618208516886931, "fcm_dpo/beta": 0.002694209571927786, "fcm_dpo/delta": 0.017087338492274284, "fcm_dpo/margin": 216.68463134765625, "fcm_dpo/q_t": 0.3784465789794922, "grad_norm": 33.220149993896484, "learning_rate": 2.397392281198729e-09, "logits/chosen": -0.45762312412261963, "logits/rejected": -0.4692486524581909, "logps/chosen": -356.33929443359375, "logps/ref_chosen": -49.576812744140625, "logps/ref_rejected": -98.29183197021484, "logps/rejected": -621.7389526367188, "loss": 1.0201, "margin_dpo/margin_mean": 216.6846466064453, "margin_dpo/margin_std": 295.111083984375, "step": 655 }, { "epoch": 0.9632892804698973, "fcm_dpo/beta": 0.002611855510622263, "fcm_dpo/delta": -0.25648045539855957, "fcm_dpo/margin": 319.9178771972656, "fcm_dpo/q_t": 0.3181704878807068, "grad_norm": 54.623451232910156, "learning_rate": 2.223355098446622e-09, "logits/chosen": -0.4365518093109131, "logits/rejected": -0.44737708568573, "logps/chosen": -353.51763916015625, "logps/ref_chosen": -52.54943084716797, "logps/ref_rejected": -113.67464447021484, "logps/rejected": -734.560791015625, "loss": 0.8148, "margin_dpo/margin_mean": 319.91790771484375, "margin_dpo/margin_std": 263.46356201171875, "step": 656 }, { "epoch": 0.9647577092511013, "fcm_dpo/beta": 0.002502152230590582, "fcm_dpo/delta": -0.15946225821971893, "fcm_dpo/margin": 298.534423828125, "fcm_dpo/q_t": 0.33715760707855225, "grad_norm": 33.79457473754883, "learning_rate": 2.055847060721566e-09, "logits/chosen": -0.5489492416381836, "logits/rejected": -0.5547805428504944, "logps/chosen": -332.32037353515625, "logps/ref_chosen": -46.700538635253906, "logps/ref_rejected": -97.91487121582031, "logps/rejected": -682.0692138671875, "loss": 0.8675, "margin_dpo/margin_mean": 298.534423828125, "margin_dpo/margin_std": 273.04095458984375, "step": 657 }, { "epoch": 0.9662261380323054, "fcm_dpo/beta": 0.0024999987799674273, "fcm_dpo/delta": 0.03055243194103241, "fcm_dpo/margin": 228.47642517089844, "fcm_dpo/q_t": 0.37107935547828674, "grad_norm": 32.88031768798828, "learning_rate": 1.8948725820160662e-09, "logits/chosen": -0.4807060956954956, "logits/rejected": -0.4754422903060913, "logps/chosen": -388.13995361328125, "logps/ref_chosen": -60.95820999145508, "logps/ref_rejected": -95.93949127197266, "logps/rejected": -651.59765625, "loss": 0.9682, "margin_dpo/margin_mean": 228.4764404296875, "margin_dpo/margin_std": 229.4976806640625, "step": 658 }, { "epoch": 0.9676945668135095, "fcm_dpo/beta": 0.002503772731870413, "fcm_dpo/delta": 0.029256850481033325, "fcm_dpo/margin": 228.52227783203125, "fcm_dpo/q_t": 0.37436485290527344, "grad_norm": 36.82291030883789, "learning_rate": 1.7404359041573723e-09, "logits/chosen": -0.509990394115448, "logits/rejected": -0.4739938974380493, "logps/chosen": -371.61309814453125, "logps/ref_chosen": -76.74298095703125, "logps/ref_rejected": -87.4709701538086, "logps/rejected": -610.8634033203125, "loss": 0.9845, "margin_dpo/margin_mean": 228.52227783203125, "margin_dpo/margin_std": 254.27716064453125, "step": 659 }, { "epoch": 0.9691629955947136, "fcm_dpo/beta": 0.002496623434126377, "fcm_dpo/delta": -0.0633494034409523, "fcm_dpo/margin": 264.0384521484375, "fcm_dpo/q_t": 0.35319751501083374, "grad_norm": 33.35771942138672, "learning_rate": 1.592541096695571e-09, "logits/chosen": -0.49027347564697266, "logits/rejected": -0.4790500998497009, "logps/chosen": -362.43536376953125, "logps/ref_chosen": -59.04788589477539, "logps/ref_rejected": -75.96005249023438, "logps/rejected": -643.385986328125, "loss": 0.9142, "margin_dpo/margin_mean": 264.0384521484375, "margin_dpo/margin_std": 247.07907104492188, "step": 660 }, { "epoch": 0.9706314243759178, "fcm_dpo/beta": 0.0024961913004517555, "fcm_dpo/delta": -0.03317072242498398, "fcm_dpo/margin": 252.34422302246094, "fcm_dpo/q_t": 0.3640047013759613, "grad_norm": 30.544677734375, "learning_rate": 1.4511920567963908e-09, "logits/chosen": -0.5381925702095032, "logits/rejected": -0.519443154335022, "logps/chosen": -326.90234375, "logps/ref_chosen": -50.673973083496094, "logps/ref_rejected": -86.00569152832031, "logps/rejected": -614.5782470703125, "loss": 0.9623, "margin_dpo/margin_mean": 252.34423828125, "margin_dpo/margin_std": 285.3419189453125, "step": 661 }, { "epoch": 0.9720998531571219, "fcm_dpo/beta": 0.0025062900967895985, "fcm_dpo/delta": 0.08712703734636307, "fcm_dpo/margin": 206.3653106689453, "fcm_dpo/q_t": 0.3861844539642334, "grad_norm": 34.15724563598633, "learning_rate": 1.3163925091384532e-09, "logits/chosen": -0.5257574319839478, "logits/rejected": -0.5023434162139893, "logps/chosen": -401.38140869140625, "logps/ref_chosen": -69.26106262207031, "logps/ref_rejected": -89.05593872070312, "logps/rejected": -627.5416259765625, "loss": 1.0401, "margin_dpo/margin_mean": 206.36529541015625, "margin_dpo/margin_std": 271.6153564453125, "step": 662 }, { "epoch": 0.973568281938326, "fcm_dpo/beta": 0.002501696813851595, "fcm_dpo/delta": -0.016493409872055054, "fcm_dpo/margin": 245.9669189453125, "fcm_dpo/q_t": 0.3665703535079956, "grad_norm": 30.06166648864746, "learning_rate": 1.1881460058152382e-09, "logits/chosen": -0.5630896687507629, "logits/rejected": -0.5610999464988708, "logps/chosen": -358.4510498046875, "logps/ref_chosen": -64.87890625, "logps/ref_rejected": -113.92536926269531, "logps/rejected": -653.4644165039062, "loss": 0.969, "margin_dpo/margin_mean": 245.9669189453125, "margin_dpo/margin_std": 283.71075439453125, "step": 663 }, { "epoch": 0.9750367107195301, "fcm_dpo/beta": 0.0024649640545248985, "fcm_dpo/delta": -0.09488554298877716, "fcm_dpo/margin": 279.1371765136719, "fcm_dpo/q_t": 0.35125353932380676, "grad_norm": 34.83463668823242, "learning_rate": 1.066455926241383e-09, "logits/chosen": -0.4723320007324219, "logits/rejected": -0.4694727063179016, "logps/chosen": -382.6121826171875, "logps/ref_chosen": -60.88847351074219, "logps/ref_rejected": -105.521728515625, "logps/rejected": -706.382568359375, "loss": 0.9418, "margin_dpo/margin_mean": 279.1371765136719, "margin_dpo/margin_std": 316.78741455078125, "step": 664 }, { "epoch": 0.9765051395007343, "fcm_dpo/beta": 0.0024584224447607994, "fcm_dpo/delta": 0.006961781997233629, "fcm_dpo/margin": 241.38986206054688, "fcm_dpo/q_t": 0.3676924705505371, "grad_norm": 32.79265594482422, "learning_rate": 9.513254770636137e-10, "logits/chosen": -0.4960812032222748, "logits/rejected": -0.4816419780254364, "logps/chosen": -358.48394775390625, "logps/ref_chosen": -60.56413269042969, "logps/ref_rejected": -84.80882263183594, "logps/rejected": -624.1184692382812, "loss": 0.9606, "margin_dpo/margin_mean": 241.38986206054688, "margin_dpo/margin_std": 244.98712158203125, "step": 665 }, { "epoch": 0.9779735682819384, "fcm_dpo/beta": 0.002456535119563341, "fcm_dpo/delta": 0.005466273054480553, "fcm_dpo/margin": 242.0909423828125, "fcm_dpo/q_t": 0.36501848697662354, "grad_norm": 38.639163970947266, "learning_rate": 8.427576920763956e-10, "logits/chosen": -0.4671856760978699, "logits/rejected": -0.4504069983959198, "logps/chosen": -372.948486328125, "logps/ref_chosen": -64.41996002197266, "logps/ref_rejected": -95.8916244506836, "logps/rejected": -646.5111083984375, "loss": 0.9541, "margin_dpo/margin_mean": 242.0909423828125, "margin_dpo/margin_std": 233.2517852783203, "step": 666 }, { "epoch": 0.9794419970631424, "fcm_dpo/beta": 0.002473963890224695, "fcm_dpo/delta": 0.02551562525331974, "fcm_dpo/margin": 232.80560302734375, "fcm_dpo/q_t": 0.3697671890258789, "grad_norm": 38.66985321044922, "learning_rate": 7.407554321417764e-10, "logits/chosen": -0.47391679883003235, "logits/rejected": -0.4410842955112457, "logps/chosen": -402.3163146972656, "logps/ref_chosen": -69.27702331542969, "logps/ref_rejected": -87.83549499511719, "logps/rejected": -653.680419921875, "loss": 0.9638, "margin_dpo/margin_mean": 232.80560302734375, "margin_dpo/margin_std": 229.169677734375, "step": 667 }, { "epoch": 0.9809104258443465, "fcm_dpo/beta": 0.00251585291698575, "fcm_dpo/delta": 0.06850365549325943, "fcm_dpo/margin": 212.354736328125, "fcm_dpo/q_t": 0.3838024139404297, "grad_norm": 52.99601745605469, "learning_rate": 6.453213851142225e-10, "logits/chosen": -0.5375820398330688, "logits/rejected": -0.529564380645752, "logps/chosen": -411.6194763183594, "logps/ref_chosen": -72.60400390625, "logps/ref_rejected": -103.73905944824219, "logps/rejected": -655.1092529296875, "loss": 1.0277, "margin_dpo/margin_mean": 212.354736328125, "margin_dpo/margin_std": 273.1512145996094, "step": 668 }, { "epoch": 0.9823788546255506, "fcm_dpo/beta": 0.0024861646816134453, "fcm_dpo/delta": -0.06441953778266907, "fcm_dpo/margin": 265.5224609375, "fcm_dpo/q_t": 0.35320982336997986, "grad_norm": 32.41237258911133, "learning_rate": 5.564580657695939e-10, "logits/chosen": -0.49402615427970886, "logits/rejected": -0.4743247926235199, "logps/chosen": -311.386962890625, "logps/ref_chosen": -46.116416931152344, "logps/ref_rejected": -77.92434692382812, "logps/rejected": -608.7173461914062, "loss": 0.9308, "margin_dpo/margin_mean": 265.5224609375, "margin_dpo/margin_std": 272.043701171875, "step": 669 }, { "epoch": 0.9838472834067548, "fcm_dpo/beta": 0.0024540587328374386, "fcm_dpo/delta": -0.09894061833620071, "fcm_dpo/margin": 282.0305480957031, "fcm_dpo/q_t": 0.3481866717338562, "grad_norm": 27.054269790649414, "learning_rate": 4.741678157389739e-10, "logits/chosen": -0.4560254216194153, "logits/rejected": -0.4461694657802582, "logps/chosen": -328.38433837890625, "logps/ref_chosen": -62.34575271606445, "logps/ref_rejected": -96.9405517578125, "logps/rejected": -645.0096435546875, "loss": 0.9109, "margin_dpo/margin_mean": 282.0305480957031, "margin_dpo/margin_std": 275.9473876953125, "step": 670 }, { "epoch": 0.9853157121879589, "fcm_dpo/beta": 0.0024509113281965256, "fcm_dpo/delta": 0.019564248621463776, "fcm_dpo/margin": 237.01344299316406, "fcm_dpo/q_t": 0.3713776171207428, "grad_norm": 38.91750717163086, "learning_rate": 3.9845280344705245e-10, "logits/chosen": -0.47259026765823364, "logits/rejected": -0.46925845742225647, "logps/chosen": -369.33929443359375, "logps/ref_chosen": -48.00010681152344, "logps/ref_rejected": -83.81932067871094, "logps/rejected": -642.1719970703125, "loss": 0.9848, "margin_dpo/margin_mean": 237.013427734375, "margin_dpo/margin_std": 265.4302673339844, "step": 671 }, { "epoch": 0.986784140969163, "fcm_dpo/beta": 0.0024676902685314417, "fcm_dpo/delta": 0.07117144763469696, "fcm_dpo/margin": 215.81497192382812, "fcm_dpo/q_t": 0.3837113380432129, "grad_norm": 31.851991653442383, "learning_rate": 3.293150240547549e-10, "logits/chosen": -0.5248799920082092, "logits/rejected": -0.5065342783927917, "logps/chosen": -408.06689453125, "logps/ref_chosen": -58.58328628540039, "logps/ref_rejected": -93.14015197753906, "logps/rejected": -658.438720703125, "loss": 1.0327, "margin_dpo/margin_mean": 215.81497192382812, "margin_dpo/margin_std": 279.0846252441406, "step": 672 }, { "epoch": 0.9882525697503671, "fcm_dpo/beta": 0.0024534822441637516, "fcm_dpo/delta": -0.05598102882504463, "fcm_dpo/margin": 265.86773681640625, "fcm_dpo/q_t": 0.3518218994140625, "grad_norm": 31.22127914428711, "learning_rate": 2.6675629940689504e-10, "logits/chosen": -0.4801812767982483, "logits/rejected": -0.4714806079864502, "logps/chosen": -353.5696105957031, "logps/ref_chosen": -46.72320556640625, "logps/ref_rejected": -85.29623413085938, "logps/rejected": -658.0103759765625, "loss": 0.9048, "margin_dpo/margin_mean": 265.86773681640625, "margin_dpo/margin_std": 224.6098175048828, "step": 673 }, { "epoch": 0.9897209985315712, "fcm_dpo/beta": 0.002430280437693, "fcm_dpo/delta": -0.05990447849035263, "fcm_dpo/margin": 269.932861328125, "fcm_dpo/q_t": 0.3552195131778717, "grad_norm": 27.59811019897461, "learning_rate": 2.1077827798404725e-10, "logits/chosen": -0.45021355152130127, "logits/rejected": -0.43945032358169556, "logps/chosen": -328.8429870605469, "logps/ref_chosen": -45.445526123046875, "logps/ref_rejected": -70.04593658447266, "logps/rejected": -623.3762817382812, "loss": 0.9216, "margin_dpo/margin_mean": 269.932861328125, "margin_dpo/margin_std": 263.62127685546875, "step": 674 }, { "epoch": 0.9911894273127754, "fcm_dpo/beta": 0.0023772665299475193, "fcm_dpo/delta": -0.08401606976985931, "fcm_dpo/margin": 285.0059814453125, "fcm_dpo/q_t": 0.3517681956291199, "grad_norm": 29.97161865234375, "learning_rate": 1.6138243485910863e-10, "logits/chosen": -0.4851476550102234, "logits/rejected": -0.47465410828590393, "logps/chosen": -329.2435607910156, "logps/ref_chosen": -44.17628479003906, "logps/ref_rejected": -74.09197998046875, "logps/rejected": -644.165283203125, "loss": 0.9136, "margin_dpo/margin_mean": 285.0060119628906, "margin_dpo/margin_std": 279.90576171875, "step": 675 }, { "epoch": 0.9926578560939795, "fcm_dpo/beta": 0.0023738746531307697, "fcm_dpo/delta": 0.0118600744754076, "fcm_dpo/margin": 247.96432495117188, "fcm_dpo/q_t": 0.3671633303165436, "grad_norm": 31.170412063598633, "learning_rate": 1.1857007165852472e-10, "logits/chosen": -0.5280739068984985, "logits/rejected": -0.5125676393508911, "logps/chosen": -399.6715087890625, "logps/ref_chosen": -71.39852905273438, "logps/ref_rejected": -88.3587646484375, "logps/rejected": -664.5960693359375, "loss": 0.9577, "margin_dpo/margin_mean": 247.96432495117188, "margin_dpo/margin_std": 239.91629028320312, "step": 676 }, { "epoch": 0.9941262848751835, "fcm_dpo/beta": 0.0024010459892451763, "fcm_dpo/delta": 0.033741872757673264, "fcm_dpo/margin": 236.56593322753906, "fcm_dpo/q_t": 0.37002164125442505, "grad_norm": 29.955596923828125, "learning_rate": 8.23423165278725e-11, "logits/chosen": -0.4702920615673065, "logits/rejected": -0.43533623218536377, "logps/chosen": -371.1539306640625, "logps/ref_chosen": -56.527435302734375, "logps/ref_rejected": -78.22654724121094, "logps/rejected": -629.4189453125, "loss": 0.9604, "margin_dpo/margin_mean": 236.56593322753906, "margin_dpo/margin_std": 216.29861450195312, "step": 677 }, { "epoch": 0.9955947136563876, "fcm_dpo/beta": 0.0023945937864482403, "fcm_dpo/delta": -0.025561237707734108, "fcm_dpo/margin": 260.56817626953125, "fcm_dpo/q_t": 0.36279577016830444, "grad_norm": 30.772653579711914, "learning_rate": 5.270012410216185e-11, "logits/chosen": -0.4658196270465851, "logits/rejected": -0.4596235156059265, "logps/chosen": -331.82763671875, "logps/ref_chosen": -46.13447570800781, "logps/ref_rejected": -80.60462951660156, "logps/rejected": -626.865966796875, "loss": 0.96, "margin_dpo/margin_mean": 260.5681457519531, "margin_dpo/margin_std": 285.284423828125, "step": 678 }, { "epoch": 0.9970631424375918, "fcm_dpo/beta": 0.002417163457721472, "fcm_dpo/delta": 0.0805218368768692, "fcm_dpo/margin": 216.64544677734375, "fcm_dpo/q_t": 0.3846209943294525, "grad_norm": 31.97832489013672, "learning_rate": 2.9644275480772416e-11, "logits/chosen": -0.498549222946167, "logits/rejected": -0.4737178087234497, "logps/chosen": -354.89007568359375, "logps/ref_chosen": -50.294921875, "logps/ref_rejected": -76.59813690185547, "logps/rejected": -597.8387451171875, "loss": 1.0171, "margin_dpo/margin_mean": 216.6454315185547, "margin_dpo/margin_std": 261.185546875, "step": 679 }, { "epoch": 0.9985315712187959, "fcm_dpo/beta": 0.002414638176560402, "fcm_dpo/delta": -0.01086747832596302, "fcm_dpo/margin": 252.61300659179688, "fcm_dpo/q_t": 0.36631903052330017, "grad_norm": 35.76852798461914, "learning_rate": 1.31753782067201e-11, "logits/chosen": -0.5048670768737793, "logits/rejected": -0.5000746250152588, "logps/chosen": -395.61883544921875, "logps/ref_chosen": -76.91569519042969, "logps/ref_rejected": -112.384765625, "logps/rejected": -683.700927734375, "loss": 0.9757, "margin_dpo/margin_mean": 252.61300659179688, "margin_dpo/margin_std": 294.8744201660156, "step": 680 }, { "epoch": 1.0, "fcm_dpo/beta": 0.002459340263158083, "fcm_dpo/delta": 0.04384385794401169, "fcm_dpo/margin": 226.318603515625, "fcm_dpo/q_t": 0.3781166672706604, "grad_norm": 32.83413314819336, "learning_rate": 3.2938662507808745e-12, "logits/chosen": -0.53011155128479, "logits/rejected": -0.5230584740638733, "logps/chosen": -388.7578125, "logps/ref_chosen": -60.957279205322266, "logps/ref_rejected": -88.55797576904297, "logps/rejected": -642.6771240234375, "loss": 1.0058, "margin_dpo/margin_mean": 226.318603515625, "margin_dpo/margin_std": 260.7975769042969, "step": 681 }, { "epoch": 1.0, "step": 681, "total_flos": 0.0, "train_loss": 0.9836619676647382, "train_runtime": 1868.3255, "train_samples_per_second": 23.335, "train_steps_per_second": 0.364 } ], "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 }