{ "best_global_step": null, "best_metric": null, "best_model_checkpoint": null, "epoch": 1.0, "eval_steps": 100, "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": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.06123492121696472, "fcm_dpo/q_t": 0.4984707534313202, "grad_norm": 38.81842041015625, "learning_rate": 0.0, "logits/chosen": 0.2677389979362488, "logits/rejected": 0.6191319227218628, "logps/chosen": -85.37664031982422, "logps/ref_chosen": -85.43083190917969, "logps/ref_rejected": -79.90458679199219, "logps/rejected": -79.91163635253906, "loss": 1.3807, "margin_dpo/margin_mean": 0.061235010623931885, "margin_dpo/margin_std": 0.4424324035644531, "step": 1 }, { "epoch": 0.002936857562408223, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.12442219257354736, "fcm_dpo/q_t": 0.5031089186668396, "grad_norm": 29.758039474487305, "learning_rate": 7.246376811594203e-09, "logits/chosen": 0.47771453857421875, "logits/rejected": 0.5908545851707458, "logps/chosen": -82.11383056640625, "logps/ref_chosen": -82.06893157958984, "logps/ref_rejected": -81.65457916259766, "logps/rejected": -81.57505798339844, "loss": 1.3992, "margin_dpo/margin_mean": -0.12442246079444885, "margin_dpo/margin_std": 0.41358208656311035, "step": 2 }, { "epoch": 0.004405286343612335, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.033603206276893616, "fcm_dpo/q_t": 0.49916017055511475, "grad_norm": 26.0107479095459, "learning_rate": 1.4492753623188406e-08, "logits/chosen": 0.4268984794616699, "logits/rejected": 0.6501251459121704, "logps/chosen": -93.78929901123047, "logps/ref_chosen": -93.81098937988281, "logps/ref_rejected": -74.22950744628906, "logps/rejected": -74.24141693115234, "loss": 1.3835, "margin_dpo/margin_mean": 0.033603474497795105, "margin_dpo/margin_std": 0.4623871445655823, "step": 3 }, { "epoch": 0.005873715124816446, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.027436256408691406, "fcm_dpo/q_t": 0.5006846189498901, "grad_norm": 32.81400680541992, "learning_rate": 2.1739130434782606e-08, "logits/chosen": 0.646797776222229, "logits/rejected": 0.778215765953064, "logps/chosen": -87.31211853027344, "logps/ref_chosen": -87.29246520996094, "logps/ref_rejected": -93.82425689697266, "logps/rejected": -93.81646728515625, "loss": 1.3895, "margin_dpo/margin_mean": -0.027436137199401855, "margin_dpo/margin_std": 0.4442657232284546, "step": 4 }, { "epoch": 0.007342143906020558, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.08646883070468903, "fcm_dpo/q_t": 0.49783921241760254, "grad_norm": 35.074466705322266, "learning_rate": 2.898550724637681e-08, "logits/chosen": 0.44550371170043945, "logits/rejected": 0.6571714878082275, "logps/chosen": -89.2946548461914, "logps/ref_chosen": -89.33676147460938, "logps/ref_rejected": -88.74783325195312, "logps/rejected": -88.79219818115234, "loss": 1.3781, "margin_dpo/margin_mean": 0.08646932244300842, "margin_dpo/margin_std": 0.42956870794296265, "step": 5 }, { "epoch": 0.00881057268722467, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.04864634573459625, "fcm_dpo/q_t": 0.49878376722335815, "grad_norm": 35.885719299316406, "learning_rate": 3.6231884057971014e-08, "logits/chosen": 0.4864919185638428, "logits/rejected": 0.8014158010482788, "logps/chosen": -97.29067993164062, "logps/ref_chosen": -97.32147216796875, "logps/ref_rejected": -97.88345336914062, "logps/rejected": -97.90130615234375, "loss": 1.3819, "margin_dpo/margin_mean": 0.048646748065948486, "margin_dpo/margin_std": 0.40285322070121765, "step": 6 }, { "epoch": 0.010279001468428781, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.021133705973625183, "fcm_dpo/q_t": 0.49947190284729004, "grad_norm": 38.00736999511719, "learning_rate": 4.347826086956521e-08, "logits/chosen": 0.49776560068130493, "logits/rejected": 0.5453675985336304, "logps/chosen": -86.66970825195312, "logps/ref_chosen": -86.64852905273438, "logps/ref_rejected": -109.61618041992188, "logps/rejected": -109.65848541259766, "loss": 1.3846, "margin_dpo/margin_mean": 0.021133065223693848, "margin_dpo/margin_std": 0.42155858874320984, "step": 7 }, { "epoch": 0.011747430249632892, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.11333945393562317, "fcm_dpo/q_t": 0.5028325915336609, "grad_norm": 32.925907135009766, "learning_rate": 5.0724637681159424e-08, "logits/chosen": 0.41005465388298035, "logits/rejected": 0.7211995124816895, "logps/chosen": -90.00587463378906, "logps/ref_chosen": -89.9236831665039, "logps/ref_rejected": -86.22804260253906, "logps/rejected": -86.19688415527344, "loss": 1.3982, "margin_dpo/margin_mean": -0.11333887279033661, "margin_dpo/margin_std": 0.47709834575653076, "step": 8 }, { "epoch": 0.013215859030837005, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.05302123725414276, "fcm_dpo/q_t": 0.4986753463745117, "grad_norm": 35.99601364135742, "learning_rate": 5.797101449275362e-08, "logits/chosen": 0.6641107797622681, "logits/rejected": 0.8435112237930298, "logps/chosen": -103.84231567382812, "logps/ref_chosen": -103.85713195800781, "logps/ref_rejected": -104.31932067871094, "logps/rejected": -104.35752868652344, "loss": 1.3815, "margin_dpo/margin_mean": 0.05302073061466217, "margin_dpo/margin_std": 0.44739070534706116, "step": 9 }, { "epoch": 0.014684287812041116, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.07061505317687988, "fcm_dpo/q_t": 0.5017644762992859, "grad_norm": 33.54641342163086, "learning_rate": 6.521739130434782e-08, "logits/chosen": 0.49548810720443726, "logits/rejected": 0.7073336839675903, "logps/chosen": -76.18025207519531, "logps/ref_chosen": -76.20436096191406, "logps/ref_rejected": -87.152099609375, "logps/rejected": -87.057373046875, "loss": 1.3938, "margin_dpo/margin_mean": -0.07061484456062317, "margin_dpo/margin_std": 0.4113759398460388, "step": 10 }, { "epoch": 0.016152716593245228, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.044263288378715515, "fcm_dpo/q_t": 0.4988948404788971, "grad_norm": 36.77696990966797, "learning_rate": 7.246376811594203e-08, "logits/chosen": 0.4357748329639435, "logits/rejected": 0.659180223941803, "logps/chosen": -82.36294555664062, "logps/ref_chosen": -82.36650085449219, "logps/ref_rejected": -94.26461791992188, "logps/rejected": -94.30534362792969, "loss": 1.3826, "margin_dpo/margin_mean": 0.044263795018196106, "margin_dpo/margin_std": 0.5451699495315552, "step": 11 }, { "epoch": 0.01762114537444934, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.11056771874427795, "fcm_dpo/q_t": 0.4972396492958069, "grad_norm": 39.76823043823242, "learning_rate": 7.971014492753623e-08, "logits/chosen": 0.24172449111938477, "logits/rejected": 0.3647180199623108, "logps/chosen": -99.05966186523438, "logps/ref_chosen": -99.10549926757812, "logps/ref_rejected": -110.27140808105469, "logps/rejected": -110.33615112304688, "loss": 1.376, "margin_dpo/margin_mean": 0.11056801676750183, "margin_dpo/margin_std": 0.5118380784988403, "step": 12 }, { "epoch": 0.01908957415565345, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.0572160929441452, "fcm_dpo/q_t": 0.4985709488391876, "grad_norm": 41.4147834777832, "learning_rate": 8.695652173913042e-08, "logits/chosen": 0.5275347232818604, "logits/rejected": 0.7880017757415771, "logps/chosen": -90.50440216064453, "logps/ref_chosen": -90.55972290039062, "logps/ref_rejected": -93.69110107421875, "logps/rejected": -93.69297790527344, "loss": 1.3813, "margin_dpo/margin_mean": 0.05721601843833923, "margin_dpo/margin_std": 0.5135507583618164, "step": 13 }, { "epoch": 0.020558002936857563, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.08244989812374115, "fcm_dpo/q_t": 0.502057671546936, "grad_norm": 35.919612884521484, "learning_rate": 9.420289855072464e-08, "logits/chosen": 1.0716733932495117, "logits/rejected": 1.167722225189209, "logps/chosen": -99.8815689086914, "logps/ref_chosen": -99.82717895507812, "logps/ref_rejected": -108.94200134277344, "logps/rejected": -108.91395568847656, "loss": 1.3952, "margin_dpo/margin_mean": -0.0824495255947113, "margin_dpo/margin_std": 0.5182109475135803, "step": 14 }, { "epoch": 0.022026431718061675, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.04384300112724304, "fcm_dpo/q_t": 0.49890750646591187, "grad_norm": 31.643028259277344, "learning_rate": 1.0144927536231885e-07, "logits/chosen": 0.3740409314632416, "logits/rejected": 0.5398809313774109, "logps/chosen": -78.9114990234375, "logps/ref_chosen": -78.90997314453125, "logps/ref_rejected": -90.06234741210938, "logps/rejected": -90.10771179199219, "loss": 1.3825, "margin_dpo/margin_mean": 0.043842852115631104, "margin_dpo/margin_std": 0.49486204981803894, "step": 15 }, { "epoch": 0.023494860499265784, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.020282208919525146, "fcm_dpo/q_t": 0.500505805015564, "grad_norm": 34.51702117919922, "learning_rate": 1.0869565217391303e-07, "logits/chosen": 0.6703887581825256, "logits/rejected": 0.9389725923538208, "logps/chosen": -97.39533996582031, "logps/ref_chosen": -97.42327880859375, "logps/ref_rejected": -90.59945678710938, "logps/rejected": -90.5512466430664, "loss": 1.3889, "margin_dpo/margin_mean": -0.02028217911720276, "margin_dpo/margin_std": 0.45394831895828247, "step": 16 }, { "epoch": 0.024963289280469897, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.03842547535896301, "fcm_dpo/q_t": 0.4990399181842804, "grad_norm": 35.85000228881836, "learning_rate": 1.1594202898550725e-07, "logits/chosen": 0.5046379566192627, "logits/rejected": 0.7362222671508789, "logps/chosen": -104.31991577148438, "logps/ref_chosen": -104.36430358886719, "logps/ref_rejected": -90.46772766113281, "logps/rejected": -90.46174621582031, "loss": 1.3829, "margin_dpo/margin_mean": 0.03842410445213318, "margin_dpo/margin_std": 0.4408981204032898, "step": 17 }, { "epoch": 0.02643171806167401, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.004340857267379761, "fcm_dpo/q_t": 0.4998915195465088, "grad_norm": 41.64680099487305, "learning_rate": 1.2318840579710146e-07, "logits/chosen": 0.2849810719490051, "logits/rejected": 0.6180058121681213, "logps/chosen": -87.09352111816406, "logps/ref_chosen": -87.09195709228516, "logps/ref_rejected": -81.85072326660156, "logps/rejected": -81.85662841796875, "loss": 1.3863, "margin_dpo/margin_mean": 0.004341289401054382, "margin_dpo/margin_std": 0.3960126042366028, "step": 18 }, { "epoch": 0.027900146842878122, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.00902073085308075, "fcm_dpo/q_t": 0.49977564811706543, "grad_norm": 31.44943618774414, "learning_rate": 1.3043478260869563e-07, "logits/chosen": 0.725235104560852, "logits/rejected": 1.009577989578247, "logps/chosen": -105.87781524658203, "logps/ref_chosen": -105.87354278564453, "logps/ref_rejected": -96.93023681640625, "logps/rejected": -96.94352722167969, "loss": 1.3859, "margin_dpo/margin_mean": 0.009020596742630005, "margin_dpo/margin_std": 0.4446555972099304, "step": 19 }, { "epoch": 0.02936857562408223, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.004244685173034668, "fcm_dpo/q_t": 0.49989384412765503, "grad_norm": 32.87926483154297, "learning_rate": 1.3768115942028986e-07, "logits/chosen": 0.3713058829307556, "logits/rejected": 0.6612204909324646, "logps/chosen": -90.76942443847656, "logps/ref_chosen": -90.75811767578125, "logps/ref_rejected": -85.91232299804688, "logps/rejected": -85.9278564453125, "loss": 1.3862, "margin_dpo/margin_mean": 0.0042449235916137695, "margin_dpo/margin_std": 0.38462841510772705, "step": 20 }, { "epoch": 0.030837004405286344, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.12687450647354126, "fcm_dpo/q_t": 0.4968298673629761, "grad_norm": 31.369789123535156, "learning_rate": 1.4492753623188405e-07, "logits/chosen": 0.7661827802658081, "logits/rejected": 0.9624724984169006, "logps/chosen": -80.27702331542969, "logps/ref_chosen": -80.33345794677734, "logps/ref_rejected": -83.93373107910156, "logps/rejected": -84.00416564941406, "loss": 1.374, "margin_dpo/margin_mean": 0.12687447667121887, "margin_dpo/margin_std": 0.39177459478378296, "step": 21 }, { "epoch": 0.032305433186490456, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.11361062526702881, "fcm_dpo/q_t": 0.4971611797809601, "grad_norm": 42.44084548950195, "learning_rate": 1.5217391304347825e-07, "logits/chosen": 0.5324074029922485, "logits/rejected": 0.6477072238922119, "logps/chosen": -95.31715393066406, "logps/ref_chosen": -95.39530181884766, "logps/ref_rejected": -103.47351837158203, "logps/rejected": -103.50899505615234, "loss": 1.3754, "margin_dpo/margin_mean": 0.11361068487167358, "margin_dpo/margin_std": 0.3894934058189392, "step": 22 }, { "epoch": 0.033773861967694566, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.024963736534118652, "fcm_dpo/q_t": 0.499376118183136, "grad_norm": 31.390506744384766, "learning_rate": 1.5942028985507245e-07, "logits/chosen": 0.7837526202201843, "logits/rejected": 0.9659498333930969, "logps/chosen": -90.59944152832031, "logps/ref_chosen": -90.63751220703125, "logps/ref_rejected": -86.59425354003906, "logps/rejected": -86.5811538696289, "loss": 1.3842, "margin_dpo/margin_mean": 0.02496352791786194, "margin_dpo/margin_std": 0.3982018232345581, "step": 23 }, { "epoch": 0.03524229074889868, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.012234330177307129, "fcm_dpo/q_t": 0.4996940493583679, "grad_norm": 44.296112060546875, "learning_rate": 1.6666666666666665e-07, "logits/chosen": 0.667069137096405, "logits/rejected": 0.6968539953231812, "logps/chosen": -69.93962097167969, "logps/ref_chosen": -69.91728973388672, "logps/ref_rejected": -106.60990142822266, "logps/rejected": -106.64446258544922, "loss": 1.3855, "margin_dpo/margin_mean": 0.012234941124916077, "margin_dpo/margin_std": 0.43183019757270813, "step": 24 }, { "epoch": 0.03671071953010279, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.06985443830490112, "fcm_dpo/q_t": 0.4982590675354004, "grad_norm": 36.90232849121094, "learning_rate": 1.7391304347826085e-07, "logits/chosen": 0.5097548365592957, "logits/rejected": 0.6774953007698059, "logps/chosen": -80.84941101074219, "logps/ref_chosen": -80.82548522949219, "logps/ref_rejected": -95.95710754394531, "logps/rejected": -96.05089569091797, "loss": 1.38, "margin_dpo/margin_mean": 0.06985485553741455, "margin_dpo/margin_std": 0.5077825784683228, "step": 25 }, { "epoch": 0.0381791483113069, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.05025067925453186, "fcm_dpo/q_t": 0.49874305725097656, "grad_norm": 41.10307693481445, "learning_rate": 1.8115942028985507e-07, "logits/chosen": 0.1814301759004593, "logits/rejected": 0.38568517565727234, "logps/chosen": -88.87095642089844, "logps/ref_chosen": -88.90116882324219, "logps/ref_rejected": -109.82819366455078, "logps/rejected": -109.84822845458984, "loss": 1.3817, "margin_dpo/margin_mean": 0.05025133490562439, "margin_dpo/margin_std": 0.412225604057312, "step": 26 }, { "epoch": 0.039647577092511016, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.07869970798492432, "fcm_dpo/q_t": 0.4980340600013733, "grad_norm": 44.32426071166992, "learning_rate": 1.8840579710144927e-07, "logits/chosen": 0.33394181728363037, "logits/rejected": 0.5241584777832031, "logps/chosen": -77.59822082519531, "logps/ref_chosen": -77.59600830078125, "logps/ref_rejected": -103.93850708007812, "logps/rejected": -104.01941680908203, "loss": 1.379, "margin_dpo/margin_mean": 0.07869991660118103, "margin_dpo/margin_std": 0.4667752981185913, "step": 27 }, { "epoch": 0.041116005873715125, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.06924325227737427, "fcm_dpo/q_t": 0.4982706308364868, "grad_norm": 36.54985809326172, "learning_rate": 1.9565217391304347e-07, "logits/chosen": 0.5311095714569092, "logits/rejected": 0.7446248531341553, "logps/chosen": -102.18016815185547, "logps/ref_chosen": -102.22856903076172, "logps/ref_rejected": -96.9594955444336, "logps/rejected": -96.98033142089844, "loss": 1.38, "margin_dpo/margin_mean": 0.0692436695098877, "margin_dpo/margin_std": 0.47658565640449524, "step": 28 }, { "epoch": 0.042584434654919234, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.16378992795944214, "fcm_dpo/q_t": 0.49590858817100525, "grad_norm": 42.102882385253906, "learning_rate": 2.028985507246377e-07, "logits/chosen": 0.20089560747146606, "logits/rejected": 0.450194776058197, "logps/chosen": -88.62367248535156, "logps/ref_chosen": -88.64704895019531, "logps/ref_rejected": -102.96011352539062, "logps/rejected": -103.10052490234375, "loss": 1.3705, "margin_dpo/margin_mean": 0.1637905091047287, "margin_dpo/margin_std": 0.46207696199417114, "step": 29 }, { "epoch": 0.04405286343612335, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.13101083040237427, "fcm_dpo/q_t": 0.49672505259513855, "grad_norm": 39.2225341796875, "learning_rate": 2.1014492753623187e-07, "logits/chosen": 0.5814670324325562, "logits/rejected": 0.6965478658676147, "logps/chosen": -88.37109375, "logps/ref_chosen": -88.38838195800781, "logps/ref_rejected": -102.31889343261719, "logps/rejected": -102.4326171875, "loss": 1.3736, "margin_dpo/margin_mean": 0.13101181387901306, "margin_dpo/margin_std": 0.39710086584091187, "step": 30 }, { "epoch": 0.04552129221732746, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.11164575815200806, "fcm_dpo/q_t": 0.4972113370895386, "grad_norm": 31.29022216796875, "learning_rate": 2.1739130434782607e-07, "logits/chosen": 0.6445361971855164, "logits/rejected": 0.8826954960823059, "logps/chosen": -101.128662109375, "logps/ref_chosen": -101.12565612792969, "logps/ref_rejected": -79.8584213256836, "logps/rejected": -79.97307586669922, "loss": 1.3755, "margin_dpo/margin_mean": 0.11164584755897522, "margin_dpo/margin_std": 0.38801681995391846, "step": 31 }, { "epoch": 0.04698972099853157, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.08825841546058655, "fcm_dpo/q_t": 0.4977962076663971, "grad_norm": 35.21088790893555, "learning_rate": 2.2463768115942027e-07, "logits/chosen": 0.5731779932975769, "logits/rejected": 0.7231494188308716, "logps/chosen": -96.67442321777344, "logps/ref_chosen": -96.61702728271484, "logps/ref_rejected": -96.224365234375, "logps/rejected": -96.3700180053711, "loss": 1.3781, "margin_dpo/margin_mean": 0.0882592499256134, "margin_dpo/margin_std": 0.4722122251987457, "step": 32 }, { "epoch": 0.048458149779735685, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.1351313441991806, "fcm_dpo/q_t": 0.49662476778030396, "grad_norm": 33.66917037963867, "learning_rate": 2.318840579710145e-07, "logits/chosen": 0.6991294026374817, "logits/rejected": 0.8825780749320984, "logps/chosen": -81.51029968261719, "logps/ref_chosen": -81.52101135253906, "logps/ref_rejected": -93.80595397949219, "logps/rejected": -93.93037414550781, "loss": 1.3735, "margin_dpo/margin_mean": 0.13513219356536865, "margin_dpo/margin_std": 0.4900803565979004, "step": 33 }, { "epoch": 0.049926578560939794, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.1415727734565735, "fcm_dpo/q_t": 0.4964648485183716, "grad_norm": 41.019248962402344, "learning_rate": 2.391304347826087e-07, "logits/chosen": 0.5307381749153137, "logits/rejected": 0.7028604745864868, "logps/chosen": -77.18989562988281, "logps/ref_chosen": -77.20204162597656, "logps/ref_rejected": -106.71876525878906, "logps/rejected": -106.84819030761719, "loss": 1.3728, "margin_dpo/margin_mean": 0.14157328009605408, "margin_dpo/margin_std": 0.4783652722835541, "step": 34 }, { "epoch": 0.0513950073421439, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.16865116357803345, "fcm_dpo/q_t": 0.4957859516143799, "grad_norm": 42.815826416015625, "learning_rate": 2.463768115942029e-07, "logits/chosen": 0.34624242782592773, "logits/rejected": 0.5021581053733826, "logps/chosen": -77.58589935302734, "logps/ref_chosen": -77.57035064697266, "logps/ref_rejected": -112.18855285644531, "logps/rejected": -112.37275695800781, "loss": 1.37, "margin_dpo/margin_mean": 0.1686515212059021, "margin_dpo/margin_std": 0.45406052470207214, "step": 35 }, { "epoch": 0.05286343612334802, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.12370005249977112, "fcm_dpo/q_t": 0.4969105124473572, "grad_norm": 32.699363708496094, "learning_rate": 2.536231884057971e-07, "logits/chosen": 0.5115445852279663, "logits/rejected": 0.7059195041656494, "logps/chosen": -83.30363464355469, "logps/ref_chosen": -83.28823852539062, "logps/ref_rejected": -91.75741577148438, "logps/rejected": -91.89651489257812, "loss": 1.3746, "margin_dpo/margin_mean": 0.12370041012763977, "margin_dpo/margin_std": 0.4851023554801941, "step": 36 }, { "epoch": 0.05433186490455213, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.2637890577316284, "fcm_dpo/q_t": 0.4934196472167969, "grad_norm": 35.98224639892578, "learning_rate": 2.6086956521739126e-07, "logits/chosen": 0.6918505430221558, "logits/rejected": 0.8543441295623779, "logps/chosen": -94.75133514404297, "logps/ref_chosen": -94.77108764648438, "logps/ref_rejected": -85.71726989746094, "logps/rejected": -85.9613037109375, "loss": 1.3611, "margin_dpo/margin_mean": 0.2637891173362732, "margin_dpo/margin_std": 0.6317564845085144, "step": 37 }, { "epoch": 0.055800293685756244, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.30049073696136475, "fcm_dpo/q_t": 0.49250102043151855, "grad_norm": 36.850074768066406, "learning_rate": 2.681159420289855e-07, "logits/chosen": 0.4852568805217743, "logits/rejected": 0.624308168888092, "logps/chosen": -75.86715698242188, "logps/ref_chosen": -75.97850036621094, "logps/ref_rejected": -104.10401916503906, "logps/rejected": -104.29315948486328, "loss": 1.3573, "margin_dpo/margin_mean": 0.3004909157752991, "margin_dpo/margin_std": 0.5791944861412048, "step": 38 }, { "epoch": 0.05726872246696035, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.2556076645851135, "fcm_dpo/q_t": 0.4936175048351288, "grad_norm": 36.4305534362793, "learning_rate": 2.753623188405797e-07, "logits/chosen": 0.5048327445983887, "logits/rejected": 0.6937540769577026, "logps/chosen": -81.26030731201172, "logps/ref_chosen": -81.18577575683594, "logps/ref_rejected": -84.1959228515625, "logps/rejected": -84.52605438232422, "loss": 1.3617, "margin_dpo/margin_mean": 0.25560712814331055, "margin_dpo/margin_std": 0.5651835203170776, "step": 39 }, { "epoch": 0.05873715124816446, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.2306583821773529, "fcm_dpo/q_t": 0.49423885345458984, "grad_norm": 29.03916358947754, "learning_rate": 2.8260869565217386e-07, "logits/chosen": 0.6887877583503723, "logits/rejected": 0.8908544182777405, "logps/chosen": -83.27583312988281, "logps/ref_chosen": -83.33257293701172, "logps/ref_rejected": -80.25591278076172, "logps/rejected": -80.42982482910156, "loss": 1.3639, "margin_dpo/margin_mean": 0.23065805435180664, "margin_dpo/margin_std": 0.4786723256111145, "step": 40 }, { "epoch": 0.06020558002936858, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.25650691986083984, "fcm_dpo/q_t": 0.4935963749885559, "grad_norm": 33.375675201416016, "learning_rate": 2.898550724637681e-07, "logits/chosen": 0.6165636777877808, "logits/rejected": 0.8028247356414795, "logps/chosen": -93.20872497558594, "logps/ref_chosen": -93.14866638183594, "logps/ref_rejected": -102.07920837402344, "logps/rejected": -102.39578247070312, "loss": 1.3619, "margin_dpo/margin_mean": 0.25650691986083984, "margin_dpo/margin_std": 0.6334136128425598, "step": 41 }, { "epoch": 0.06167400881057269, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.41214534640312195, "fcm_dpo/q_t": 0.48970967531204224, "grad_norm": 40.96079635620117, "learning_rate": 2.971014492753623e-07, "logits/chosen": 0.2511819899082184, "logits/rejected": 0.39002329111099243, "logps/chosen": -90.65541076660156, "logps/ref_chosen": -90.70162200927734, "logps/ref_rejected": -113.8646469116211, "logps/rejected": -114.23058319091797, "loss": 1.3463, "margin_dpo/margin_mean": 0.4121454060077667, "margin_dpo/margin_std": 0.5678550601005554, "step": 42 }, { "epoch": 0.0631424375917768, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.35536667704582214, "fcm_dpo/q_t": 0.4911324381828308, "grad_norm": 38.216575622558594, "learning_rate": 3.043478260869565e-07, "logits/chosen": 0.7120174765586853, "logits/rejected": 0.9082381725311279, "logps/chosen": -89.6702880859375, "logps/ref_chosen": -89.64402770996094, "logps/ref_rejected": -103.95185852050781, "logps/rejected": -104.33348083496094, "loss": 1.3522, "margin_dpo/margin_mean": 0.3553672134876251, "margin_dpo/margin_std": 0.6629294157028198, "step": 43 }, { "epoch": 0.06461086637298091, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.344689279794693, "fcm_dpo/q_t": 0.49139606952667236, "grad_norm": 34.39558792114258, "learning_rate": 3.115942028985507e-07, "logits/chosen": 0.9785443544387817, "logits/rejected": 1.0956333875656128, "logps/chosen": -81.92916870117188, "logps/ref_chosen": -81.8783187866211, "logps/ref_rejected": -113.421630859375, "logps/rejected": -113.81716918945312, "loss": 1.3531, "margin_dpo/margin_mean": 0.34468963742256165, "margin_dpo/margin_std": 0.6063359975814819, "step": 44 }, { "epoch": 0.06607929515418502, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.20663021504878998, "fcm_dpo/q_t": 0.49483978748321533, "grad_norm": 24.50741195678711, "learning_rate": 3.188405797101449e-07, "logits/chosen": 0.5508902072906494, "logits/rejected": 0.6601569056510925, "logps/chosen": -77.46229553222656, "logps/ref_chosen": -77.34459686279297, "logps/ref_rejected": -84.24774169921875, "logps/rejected": -84.57206726074219, "loss": 1.3666, "margin_dpo/margin_mean": 0.20663078129291534, "margin_dpo/margin_std": 0.5758439302444458, "step": 45 }, { "epoch": 0.06754772393538913, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.5048208236694336, "fcm_dpo/q_t": 0.48739558458328247, "grad_norm": 33.225318908691406, "learning_rate": 3.260869565217391e-07, "logits/chosen": 0.35200345516204834, "logits/rejected": 0.5865339040756226, "logps/chosen": -90.34577941894531, "logps/ref_chosen": -90.3408203125, "logps/ref_rejected": -100.676513671875, "logps/rejected": -101.18629455566406, "loss": 1.3374, "margin_dpo/margin_mean": 0.5048205852508545, "margin_dpo/margin_std": 0.6096693873405457, "step": 46 }, { "epoch": 0.06901615271659324, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.35532331466674805, "fcm_dpo/q_t": 0.4911317825317383, "grad_norm": 34.18780517578125, "learning_rate": 3.333333333333333e-07, "logits/chosen": 0.5468966960906982, "logits/rejected": 0.7348242402076721, "logps/chosen": -104.42884826660156, "logps/ref_chosen": -104.41130065917969, "logps/ref_rejected": -101.25489807128906, "logps/rejected": -101.62776947021484, "loss": 1.3522, "margin_dpo/margin_mean": 0.3553228974342346, "margin_dpo/margin_std": 0.6635472178459167, "step": 47 }, { "epoch": 0.07048458149779736, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.5949963927268982, "fcm_dpo/q_t": 0.485162615776062, "grad_norm": 38.76776123046875, "learning_rate": 3.4057971014492755e-07, "logits/chosen": 0.31125134229660034, "logits/rejected": 0.5184098482131958, "logps/chosen": -91.11347961425781, "logps/ref_chosen": -91.10027313232422, "logps/ref_rejected": -95.08058166503906, "logps/rejected": -95.68878173828125, "loss": 1.3294, "margin_dpo/margin_mean": 0.5949971079826355, "margin_dpo/margin_std": 0.8412452936172485, "step": 48 }, { "epoch": 0.07195301027900147, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.8068197965621948, "fcm_dpo/q_t": 0.4799154996871948, "grad_norm": 40.11665344238281, "learning_rate": 3.478260869565217e-07, "logits/chosen": 0.042669445276260376, "logits/rejected": 0.5065343379974365, "logps/chosen": -92.9081802368164, "logps/ref_chosen": -93.00367736816406, "logps/ref_rejected": -91.74899291992188, "logps/rejected": -92.46031188964844, "loss": 1.3099, "margin_dpo/margin_mean": 0.8068201541900635, "margin_dpo/margin_std": 1.0258755683898926, "step": 49 }, { "epoch": 0.07342143906020558, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.646321177482605, "fcm_dpo/q_t": 0.4838915765285492, "grad_norm": 33.23332595825195, "learning_rate": 3.5507246376811595e-07, "logits/chosen": 0.6108515858650208, "logits/rejected": 0.7840231657028198, "logps/chosen": -94.68785095214844, "logps/ref_chosen": -94.62681579589844, "logps/ref_rejected": -103.57435607910156, "logps/rejected": -104.28172302246094, "loss": 1.3248, "margin_dpo/margin_mean": 0.6463215947151184, "margin_dpo/margin_std": 0.9034823179244995, "step": 50 }, { "epoch": 0.07488986784140969, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.6294625997543335, "fcm_dpo/q_t": 0.48433637619018555, "grad_norm": 30.03205680847168, "learning_rate": 3.6231884057971015e-07, "logits/chosen": 0.351556658744812, "logits/rejected": 0.47928231954574585, "logps/chosen": -87.57754516601562, "logps/ref_chosen": -87.50727844238281, "logps/ref_rejected": -83.47235107421875, "logps/rejected": -84.17208862304688, "loss": 1.3273, "margin_dpo/margin_mean": 0.6294622421264648, "margin_dpo/margin_std": 1.0819957256317139, "step": 51 }, { "epoch": 0.0763582966226138, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.7756294012069702, "fcm_dpo/q_t": 0.4806681275367737, "grad_norm": 32.24178695678711, "learning_rate": 3.695652173913043e-07, "logits/chosen": 0.47563832998275757, "logits/rejected": 0.6913837194442749, "logps/chosen": -90.67118835449219, "logps/ref_chosen": -90.63026428222656, "logps/ref_rejected": -87.0390625, "logps/rejected": -87.85562133789062, "loss": 1.3121, "margin_dpo/margin_mean": 0.7756293416023254, "margin_dpo/margin_std": 0.8545148968696594, "step": 52 }, { "epoch": 0.07782672540381791, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.0988894701004028, "fcm_dpo/q_t": 0.47282397747039795, "grad_norm": 39.16925048828125, "learning_rate": 3.7681159420289855e-07, "logits/chosen": 0.2351905107498169, "logits/rejected": 0.35199305415153503, "logps/chosen": -81.59117126464844, "logps/ref_chosen": -81.58307647705078, "logps/ref_rejected": -95.66152954101562, "logps/rejected": -96.76851654052734, "loss": 1.2849, "margin_dpo/margin_mean": 1.0988894701004028, "margin_dpo/margin_std": 1.4093308448791504, "step": 53 }, { "epoch": 0.07929515418502203, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.159604787826538, "fcm_dpo/q_t": 0.4712783694267273, "grad_norm": 39.55070877075195, "learning_rate": 3.8405797101449274e-07, "logits/chosen": 0.44965824484825134, "logits/rejected": 0.6323707699775696, "logps/chosen": -88.9405288696289, "logps/ref_chosen": -88.91017150878906, "logps/ref_rejected": -99.1175537109375, "logps/rejected": -100.30752563476562, "loss": 1.2788, "margin_dpo/margin_mean": 1.159604787826538, "margin_dpo/margin_std": 1.439812421798706, "step": 54 }, { "epoch": 0.08076358296622614, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.234811782836914, "fcm_dpo/q_t": 0.46946150064468384, "grad_norm": 35.18501663208008, "learning_rate": 3.9130434782608694e-07, "logits/chosen": 0.10041927546262741, "logits/rejected": 0.40900927782058716, "logps/chosen": -92.3023910522461, "logps/ref_chosen": -92.45591735839844, "logps/ref_rejected": -92.97093200683594, "logps/rejected": -94.05221557617188, "loss": 1.273, "margin_dpo/margin_mean": 1.2348113059997559, "margin_dpo/margin_std": 1.5631911754608154, "step": 55 }, { "epoch": 0.08223201174743025, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.0258830785751343, "fcm_dpo/q_t": 0.47462257742881775, "grad_norm": 33.30818176269531, "learning_rate": 3.9855072463768114e-07, "logits/chosen": 0.2586688697338104, "logits/rejected": 0.6011684536933899, "logps/chosen": -87.36018371582031, "logps/ref_chosen": -87.23665618896484, "logps/ref_rejected": -100.64553833007812, "logps/rejected": -101.79496002197266, "loss": 1.2928, "margin_dpo/margin_mean": 1.0258824825286865, "margin_dpo/margin_std": 1.6051685810089111, "step": 56 }, { "epoch": 0.08370044052863436, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.1884700059890747, "fcm_dpo/q_t": 0.47055989503860474, "grad_norm": 33.485557556152344, "learning_rate": 4.057971014492754e-07, "logits/chosen": 0.29828542470932007, "logits/rejected": 0.5136665105819702, "logps/chosen": -98.16508483886719, "logps/ref_chosen": -98.15074157714844, "logps/ref_rejected": -101.05284118652344, "logps/rejected": -102.25564575195312, "loss": 1.2775, "margin_dpo/margin_mean": 1.1884701251983643, "margin_dpo/margin_std": 1.6172009706497192, "step": 57 }, { "epoch": 0.08516886930983847, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.5477666854858398, "fcm_dpo/q_t": 0.46193379163742065, "grad_norm": 37.840396881103516, "learning_rate": 4.1304347826086954e-07, "logits/chosen": 0.1909644901752472, "logits/rejected": 0.517054557800293, "logps/chosen": -99.33880615234375, "logps/ref_chosen": -99.58097076416016, "logps/ref_rejected": -91.22227478027344, "logps/rejected": -92.52787780761719, "loss": 1.2468, "margin_dpo/margin_mean": 1.547766923904419, "margin_dpo/margin_std": 1.863322138786316, "step": 58 }, { "epoch": 0.08663729809104258, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.1774803400039673, "fcm_dpo/q_t": 0.47080308198928833, "grad_norm": 34.35539245605469, "learning_rate": 4.2028985507246374e-07, "logits/chosen": 0.21893709897994995, "logits/rejected": 0.41581693291664124, "logps/chosen": -89.77354431152344, "logps/ref_chosen": -89.80232238769531, "logps/ref_rejected": -94.27667236328125, "logps/rejected": -95.42537689208984, "loss": 1.2771, "margin_dpo/margin_mean": 1.1774799823760986, "margin_dpo/margin_std": 1.4213714599609375, "step": 59 }, { "epoch": 0.0881057268722467, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.2050294876098633, "fcm_dpo/q_t": 0.47011345624923706, "grad_norm": 28.579572677612305, "learning_rate": 4.2753623188405794e-07, "logits/chosen": 0.21510683000087738, "logits/rejected": 0.47372835874557495, "logps/chosen": -95.11346435546875, "logps/ref_chosen": -95.15571594238281, "logps/ref_rejected": -91.3724365234375, "logps/rejected": -92.53521728515625, "loss": 1.2744, "margin_dpo/margin_mean": 1.2050297260284424, "margin_dpo/margin_std": 1.4122042655944824, "step": 60 }, { "epoch": 0.08957415565345081, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.0777596235275269, "fcm_dpo/q_t": 0.47358256578445435, "grad_norm": 28.4328670501709, "learning_rate": 4.3478260869565214e-07, "logits/chosen": 0.15031884610652924, "logits/rejected": 0.28975626826286316, "logps/chosen": -85.95419311523438, "logps/ref_chosen": -85.73230743408203, "logps/ref_rejected": -97.96575927734375, "logps/rejected": -99.26539611816406, "loss": 1.2898, "margin_dpo/margin_mean": 1.0777602195739746, "margin_dpo/margin_std": 1.7849442958831787, "step": 61 }, { "epoch": 0.09104258443465492, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.185563325881958, "fcm_dpo/q_t": 0.4706765115261078, "grad_norm": 28.67509651184082, "learning_rate": 4.420289855072464e-07, "logits/chosen": 0.3115650415420532, "logits/rejected": 0.5142878890037537, "logps/chosen": -81.77017211914062, "logps/ref_chosen": -81.63538360595703, "logps/ref_rejected": -84.03831481933594, "logps/rejected": -85.3586654663086, "loss": 1.2803, "margin_dpo/margin_mean": 1.1855626106262207, "margin_dpo/margin_std": 1.801519751548767, "step": 62 }, { "epoch": 0.09251101321585903, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.5173454284667969, "fcm_dpo/q_t": 0.4624067544937134, "grad_norm": 35.08026123046875, "learning_rate": 4.4927536231884053e-07, "logits/chosen": 0.1734933853149414, "logits/rejected": 0.3734012246131897, "logps/chosen": -103.64640808105469, "logps/ref_chosen": -103.62405395507812, "logps/ref_rejected": -103.40303039550781, "logps/rejected": -104.94271850585938, "loss": 1.2463, "margin_dpo/margin_mean": 1.5173449516296387, "margin_dpo/margin_std": 1.5217807292938232, "step": 63 }, { "epoch": 0.09397944199706314, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.6387507915496826, "fcm_dpo/q_t": 0.4597811698913574, "grad_norm": 35.231178283691406, "learning_rate": 4.5652173913043473e-07, "logits/chosen": 0.6302533149719238, "logits/rejected": 0.8535206913948059, "logps/chosen": -87.16665649414062, "logps/ref_chosen": -87.0015869140625, "logps/ref_rejected": -100.5854721069336, "logps/rejected": -102.38929748535156, "loss": 1.2402, "margin_dpo/margin_mean": 1.638750672340393, "margin_dpo/margin_std": 2.034325122833252, "step": 64 }, { "epoch": 0.09544787077826726, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.7660255432128906, "fcm_dpo/q_t": 0.45665818452835083, "grad_norm": 38.46485137939453, "learning_rate": 4.63768115942029e-07, "logits/chosen": 0.02439071610569954, "logits/rejected": 0.16916462779045105, "logps/chosen": -91.31500244140625, "logps/ref_chosen": -91.22191619873047, "logps/ref_rejected": -115.33553314208984, "logps/rejected": -117.19464874267578, "loss": 1.2282, "margin_dpo/margin_mean": 1.7660255432128906, "margin_dpo/margin_std": 2.0794177055358887, "step": 65 }, { "epoch": 0.09691629955947137, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.1536918878555298, "fcm_dpo/q_t": 0.4715958833694458, "grad_norm": 25.735258102416992, "learning_rate": 4.7101449275362313e-07, "logits/chosen": 0.2745053470134735, "logits/rejected": 0.48479628562927246, "logps/chosen": -84.15156555175781, "logps/ref_chosen": -83.78422546386719, "logps/ref_rejected": -82.7520980834961, "logps/rejected": -84.27313232421875, "loss": 1.2838, "margin_dpo/margin_mean": 1.1536920070648193, "margin_dpo/margin_std": 1.8838276863098145, "step": 66 }, { "epoch": 0.09838472834067548, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.0333654880523682, "fcm_dpo/q_t": 0.47503939270973206, "grad_norm": 24.520505905151367, "learning_rate": 4.782608695652174e-07, "logits/chosen": 0.3828052282333374, "logits/rejected": 0.6662817001342773, "logps/chosen": -88.6138916015625, "logps/ref_chosen": -87.67295837402344, "logps/ref_rejected": -79.08674621582031, "logps/rejected": -81.06103515625, "loss": 1.3044, "margin_dpo/margin_mean": 1.0333654880523682, "margin_dpo/margin_std": 2.741905689239502, "step": 67 }, { "epoch": 0.09985315712187959, "fcm_dpo/beta": 0.1100722998380661, "fcm_dpo/delta": 0.6364251971244812, "fcm_dpo/margin": 2.0699269771575928, "fcm_dpo/q_t": 0.4486227035522461, "grad_norm": 30.40932273864746, "learning_rate": 4.855072463768116e-07, "logits/chosen": -0.0022230129688978195, "logits/rejected": 0.29651695489883423, "logps/chosen": -97.31098937988281, "logps/ref_chosen": -97.07884216308594, "logps/ref_rejected": -84.11872863769531, "logps/rejected": -86.4207992553711, "loss": 1.1941, "margin_dpo/margin_mean": 2.0699267387390137, "margin_dpo/margin_std": 2.751864433288574, "step": 68 }, { "epoch": 0.1013215859030837, "fcm_dpo/beta": 0.12420888990163803, "fcm_dpo/delta": 0.5869454145431519, "fcm_dpo/margin": 2.242255210876465, "fcm_dpo/q_t": 0.43765681982040405, "grad_norm": 35.120948791503906, "learning_rate": 4.927536231884058e-07, "logits/chosen": 0.10133904963731766, "logits/rejected": 0.21260981261730194, "logps/chosen": -85.975341796875, "logps/ref_chosen": -85.71971130371094, "logps/ref_rejected": -109.48020935058594, "logps/rejected": -111.97808837890625, "loss": 1.1604, "margin_dpo/margin_mean": 2.242255210876465, "margin_dpo/margin_std": 3.020266056060791, "step": 69 }, { "epoch": 0.1027900146842878, "fcm_dpo/beta": 0.13789792358875275, "fcm_dpo/delta": 0.5037232041358948, "fcm_dpo/margin": 2.6392881870269775, "fcm_dpo/q_t": 0.42064177989959717, "grad_norm": 35.125850677490234, "learning_rate": 5e-07, "logits/chosen": 0.376822829246521, "logits/rejected": 0.5594728589057922, "logps/chosen": -95.60337829589844, "logps/ref_chosen": -95.00994873046875, "logps/ref_rejected": -96.21272277832031, "logps/rejected": -99.44544219970703, "loss": 1.1087, "margin_dpo/margin_mean": 2.6392879486083984, "margin_dpo/margin_std": 3.4235267639160156, "step": 70 }, { "epoch": 0.10425844346549193, "fcm_dpo/beta": 0.14798226952552795, "fcm_dpo/delta": 0.3334314823150635, "fcm_dpo/margin": 3.616886854171753, "fcm_dpo/q_t": 0.38527965545654297, "grad_norm": 42.508155822753906, "learning_rate": 4.999967061337492e-07, "logits/chosen": -0.3210897445678711, "logits/rejected": 0.03221406042575836, "logps/chosen": -91.05834197998047, "logps/ref_chosen": -90.97735595703125, "logps/ref_rejected": -102.59103393554688, "logps/rejected": -106.28890991210938, "loss": 0.9942, "margin_dpo/margin_mean": 3.616886615753174, "margin_dpo/margin_std": 3.979628801345825, "step": 71 }, { "epoch": 0.10572687224669604, "fcm_dpo/beta": 0.15995782613754272, "fcm_dpo/delta": 0.3947334587574005, "fcm_dpo/margin": 2.966139316558838, "fcm_dpo/q_t": 0.39579302072525024, "grad_norm": 38.812652587890625, "learning_rate": 4.999868246217933e-07, "logits/chosen": -0.07235760986804962, "logits/rejected": 0.2604253590106964, "logps/chosen": -98.39273071289062, "logps/ref_chosen": -97.89152526855469, "logps/ref_rejected": -100.19171142578125, "logps/rejected": -103.6590576171875, "loss": 1.0302, "margin_dpo/margin_mean": 2.966139316558838, "margin_dpo/margin_std": 3.2803831100463867, "step": 72 }, { "epoch": 0.10719530102790015, "fcm_dpo/beta": 0.17192596197128296, "fcm_dpo/delta": 0.3318787217140198, "fcm_dpo/margin": 3.1142077445983887, "fcm_dpo/q_t": 0.3958582282066345, "grad_norm": 38.59789276123047, "learning_rate": 4.999703557245192e-07, "logits/chosen": -0.3595370054244995, "logits/rejected": -0.05686055123806, "logps/chosen": -96.527099609375, "logps/ref_chosen": -95.7690200805664, "logps/ref_rejected": -95.93243408203125, "logps/rejected": -99.80471801757812, "loss": 1.0633, "margin_dpo/margin_mean": 3.1142072677612305, "margin_dpo/margin_std": 4.913205146789551, "step": 73 }, { "epoch": 0.10866372980910426, "fcm_dpo/beta": 0.18055209517478943, "fcm_dpo/delta": 0.23689761757850647, "fcm_dpo/margin": 3.480034589767456, "fcm_dpo/q_t": 0.37341517210006714, "grad_norm": 39.81556701660156, "learning_rate": 4.999472998758977e-07, "logits/chosen": -0.11234185099601746, "logits/rejected": 0.03054591454565525, "logps/chosen": -79.81861877441406, "logps/ref_chosen": -78.80839538574219, "logps/ref_rejected": -101.64677429199219, "logps/rejected": -106.13703155517578, "loss": 0.9882, "margin_dpo/margin_mean": 3.480034589767456, "margin_dpo/margin_std": 4.480711936950684, "step": 74 }, { "epoch": 0.11013215859030837, "fcm_dpo/beta": 0.1853078305721283, "fcm_dpo/delta": 0.030487246811389923, "fcm_dpo/margin": 4.428960800170898, "fcm_dpo/q_t": 0.3425433039665222, "grad_norm": 40.42939758300781, "learning_rate": 4.999176576834721e-07, "logits/chosen": -0.18196149170398712, "logits/rejected": 0.018665947020053864, "logps/chosen": -79.34150695800781, "logps/ref_chosen": -78.28185272216797, "logps/ref_rejected": -115.40311431884766, "logps/rejected": -120.8917236328125, "loss": 0.8974, "margin_dpo/margin_mean": 4.428961277008057, "margin_dpo/margin_std": 5.278062343597412, "step": 75 }, { "epoch": 0.11160058737151249, "fcm_dpo/beta": 0.19199031591415405, "fcm_dpo/delta": 0.2775818109512329, "fcm_dpo/margin": 3.0665392875671387, "fcm_dpo/q_t": 0.38420790433883667, "grad_norm": 39.92314529418945, "learning_rate": 4.998814299283415e-07, "logits/chosen": -0.6445671319961548, "logits/rejected": -0.2710894048213959, "logps/chosen": -89.29500579833984, "logps/ref_chosen": -87.87713623046875, "logps/ref_rejected": -85.71968078613281, "logps/rejected": -90.20408630371094, "loss": 1.0398, "margin_dpo/margin_mean": 3.066539764404297, "margin_dpo/margin_std": 4.716013431549072, "step": 76 }, { "epoch": 0.1130690161527166, "fcm_dpo/beta": 0.19982174038887024, "fcm_dpo/delta": 0.19942709803581238, "fcm_dpo/margin": 3.324207305908203, "fcm_dpo/q_t": 0.37066739797592163, "grad_norm": 50.32765197753906, "learning_rate": 4.998386175651409e-07, "logits/chosen": -0.15631820261478424, "logits/rejected": 0.21387773752212524, "logps/chosen": -101.25022888183594, "logps/ref_chosen": -99.70034790039062, "logps/ref_rejected": -98.20576477050781, "logps/rejected": -103.07984924316406, "loss": 1.0254, "margin_dpo/margin_mean": 3.324207305908203, "margin_dpo/margin_std": 4.918254852294922, "step": 77 }, { "epoch": 0.1145374449339207, "fcm_dpo/beta": 0.20663225650787354, "fcm_dpo/delta": 0.1321578472852707, "fcm_dpo/margin": 3.520613193511963, "fcm_dpo/q_t": 0.35340678691864014, "grad_norm": 36.7442512512207, "learning_rate": 4.997892217220159e-07, "logits/chosen": 0.2885063588619232, "logits/rejected": 0.5307227373123169, "logps/chosen": -90.95689392089844, "logps/ref_chosen": -90.29670715332031, "logps/ref_rejected": -91.13772583007812, "logps/rejected": -95.31852722167969, "loss": 0.9308, "margin_dpo/margin_mean": 3.5206122398376465, "margin_dpo/margin_std": 4.204378604888916, "step": 78 }, { "epoch": 0.11600587371512482, "fcm_dpo/beta": 0.20485563576221466, "fcm_dpo/delta": -0.12860579788684845, "fcm_dpo/margin": 4.717245101928711, "fcm_dpo/q_t": 0.3210354447364807, "grad_norm": 39.420284271240234, "learning_rate": 4.997332437005931e-07, "logits/chosen": -0.5903613567352295, "logits/rejected": -0.21010416746139526, "logps/chosen": -87.16404724121094, "logps/ref_chosen": -86.37832641601562, "logps/ref_rejected": -94.10777282714844, "logps/rejected": -99.61073303222656, "loss": 0.8702, "margin_dpo/margin_mean": 4.717245578765869, "margin_dpo/margin_std": 5.7245635986328125, "step": 79 }, { "epoch": 0.11747430249632893, "fcm_dpo/beta": 0.20580410957336426, "fcm_dpo/delta": 0.09618805348873138, "fcm_dpo/margin": 3.69810152053833, "fcm_dpo/q_t": 0.3560129404067993, "grad_norm": 43.681617736816406, "learning_rate": 4.996706849759452e-07, "logits/chosen": -0.5645250678062439, "logits/rejected": -0.20316016674041748, "logps/chosen": -95.42378997802734, "logps/ref_chosen": -93.97032165527344, "logps/ref_rejected": -92.57441711425781, "logps/rejected": -97.72599792480469, "loss": 0.9497, "margin_dpo/margin_mean": 3.698101043701172, "margin_dpo/margin_std": 4.941034317016602, "step": 80 }, { "epoch": 0.11894273127753303, "fcm_dpo/beta": 0.19814518094062805, "fcm_dpo/delta": -0.21343490481376648, "fcm_dpo/margin": 5.233097076416016, "fcm_dpo/q_t": 0.3187583088874817, "grad_norm": 39.48776626586914, "learning_rate": 4.996015471965529e-07, "logits/chosen": -0.4514112174510956, "logits/rejected": -0.026859581470489502, "logps/chosen": -100.7833480834961, "logps/ref_chosen": -99.83012390136719, "logps/ref_rejected": -133.67247009277344, "logps/rejected": -139.85879516601562, "loss": 0.8648, "margin_dpo/margin_mean": 5.233097076416016, "margin_dpo/margin_std": 6.736208438873291, "step": 81 }, { "epoch": 0.12041116005873716, "fcm_dpo/beta": 0.2045668065547943, "fcm_dpo/delta": 0.13720954954624176, "fcm_dpo/margin": 3.512045383453369, "fcm_dpo/q_t": 0.35955876111984253, "grad_norm": 42.71525192260742, "learning_rate": 4.995258321842611e-07, "logits/chosen": -0.2967575490474701, "logits/rejected": -0.08636781573295593, "logps/chosen": -84.83177185058594, "logps/ref_chosen": -83.04598236083984, "logps/ref_rejected": -94.52595520019531, "logps/rejected": -99.82378387451172, "loss": 0.9743, "margin_dpo/margin_mean": 3.512044906616211, "margin_dpo/margin_std": 4.646213054656982, "step": 82 }, { "epoch": 0.12187958883994127, "fcm_dpo/beta": 0.20421761274337769, "fcm_dpo/delta": 0.02744770422577858, "fcm_dpo/margin": 4.038933753967285, "fcm_dpo/q_t": 0.3518849313259125, "grad_norm": 58.90052032470703, "learning_rate": 4.994435419342304e-07, "logits/chosen": -0.4892539381980896, "logits/rejected": -0.07342211157083511, "logps/chosen": -94.18898010253906, "logps/ref_chosen": -92.17621612548828, "logps/ref_rejected": -107.74464416503906, "logps/rejected": -113.79634094238281, "loss": 0.9931, "margin_dpo/margin_mean": 4.038934230804443, "margin_dpo/margin_std": 5.922978401184082, "step": 83 }, { "epoch": 0.12334801762114538, "fcm_dpo/beta": 0.20572717487812042, "fcm_dpo/delta": 0.02440953627228737, "fcm_dpo/margin": 2.4480397701263428, "fcm_dpo/q_t": 0.40574872493743896, "grad_norm": 62.624961853027344, "learning_rate": 4.993546786148857e-07, "logits/chosen": -0.3661155104637146, "logits/rejected": 0.011862829327583313, "logps/chosen": -104.15892028808594, "logps/ref_chosen": -101.5264892578125, "logps/ref_rejected": -92.42608642578125, "logps/rejected": -97.50656127929688, "loss": 1.148, "margin_dpo/margin_mean": 2.4480395317077637, "margin_dpo/margin_std": 4.715095043182373, "step": 84 }, { "epoch": 0.12481644640234948, "fcm_dpo/beta": 0.21252544224262238, "fcm_dpo/delta": 0.15723800659179688, "fcm_dpo/margin": 3.297471523284912, "fcm_dpo/q_t": 0.37289494276046753, "grad_norm": 64.13458251953125, "learning_rate": 4.992592445678582e-07, "logits/chosen": -0.2507883310317993, "logits/rejected": 0.07221034169197083, "logps/chosen": -98.31668853759766, "logps/ref_chosen": -96.12738037109375, "logps/ref_rejected": -85.05519104003906, "logps/rejected": -90.54196166992188, "loss": 1.0876, "margin_dpo/margin_mean": 3.2974720001220703, "margin_dpo/margin_std": 5.628973484039307, "step": 85 }, { "epoch": 0.1262848751835536, "fcm_dpo/beta": 0.2112560123205185, "fcm_dpo/delta": -0.015112241730093956, "fcm_dpo/margin": 4.088076591491699, "fcm_dpo/q_t": 0.35769122838974, "grad_norm": 42.73858642578125, "learning_rate": 4.991572423079235e-07, "logits/chosen": -0.5674518346786499, "logits/rejected": -0.3341968059539795, "logps/chosen": -83.38761138916016, "logps/ref_chosen": -81.70426177978516, "logps/ref_rejected": -93.65550231933594, "logps/rejected": -99.42691802978516, "loss": 0.9898, "margin_dpo/margin_mean": 4.088076114654541, "margin_dpo/margin_std": 6.537631988525391, "step": 86 }, { "epoch": 0.1277533039647577, "fcm_dpo/beta": 0.2079107165336609, "fcm_dpo/delta": -0.08900696039199829, "fcm_dpo/margin": 4.47269344329834, "fcm_dpo/q_t": 0.33168458938598633, "grad_norm": 54.897640228271484, "learning_rate": 4.990486745229364e-07, "logits/chosen": -0.6444218158721924, "logits/rejected": -0.2799781858921051, "logps/chosen": -94.55389404296875, "logps/ref_chosen": -92.68596649169922, "logps/ref_rejected": -102.91818237304688, "logps/rejected": -109.2587890625, "loss": 0.9951, "margin_dpo/margin_mean": 4.472694396972656, "margin_dpo/margin_std": 6.770974159240723, "step": 87 }, { "epoch": 0.12922173274596183, "fcm_dpo/beta": 0.20963647961616516, "fcm_dpo/delta": 0.13043585419654846, "fcm_dpo/margin": 3.4725332260131836, "fcm_dpo/q_t": 0.3684794306755066, "grad_norm": 53.998870849609375, "learning_rate": 4.989335440737586e-07, "logits/chosen": -0.18257489800453186, "logits/rejected": -0.004581686109304428, "logps/chosen": -103.09647369384766, "logps/ref_chosen": -100.76298522949219, "logps/ref_rejected": -113.15037536621094, "logps/rejected": -118.95639038085938, "loss": 1.0549, "margin_dpo/margin_mean": 3.4725332260131836, "margin_dpo/margin_std": 5.660479545593262, "step": 88 }, { "epoch": 0.13069016152716592, "fcm_dpo/beta": 0.21444444358348846, "fcm_dpo/delta": 0.058221083134412766, "fcm_dpo/margin": 3.713489055633545, "fcm_dpo/q_t": 0.3624182641506195, "grad_norm": 41.0621223449707, "learning_rate": 4.988118539941847e-07, "logits/chosen": -0.7678704857826233, "logits/rejected": -0.4581834077835083, "logps/chosen": -91.50296020507812, "logps/ref_chosen": -89.69108581542969, "logps/ref_rejected": -88.56832885742188, "logps/rejected": -94.09370422363281, "loss": 1.004, "margin_dpo/margin_mean": 3.7134885787963867, "margin_dpo/margin_std": 6.147014617919922, "step": 89 }, { "epoch": 0.13215859030837004, "fcm_dpo/beta": 0.21235866844654083, "fcm_dpo/delta": -0.1558917909860611, "fcm_dpo/margin": 4.664684295654297, "fcm_dpo/q_t": 0.3254511058330536, "grad_norm": 44.808536529541016, "learning_rate": 4.986836074908615e-07, "logits/chosen": -0.29161787033081055, "logits/rejected": -0.17332673072814941, "logps/chosen": -82.94302368164062, "logps/ref_chosen": -81.38255310058594, "logps/ref_rejected": -117.77714538574219, "logps/rejected": -124.00228881835938, "loss": 0.9002, "margin_dpo/margin_mean": 4.664685249328613, "margin_dpo/margin_std": 6.0813398361206055, "step": 90 }, { "epoch": 0.13362701908957417, "fcm_dpo/beta": 0.2086181640625, "fcm_dpo/delta": -0.03319477662444115, "fcm_dpo/margin": 4.219526290893555, "fcm_dpo/q_t": 0.34347307682037354, "grad_norm": 49.89957046508789, "learning_rate": 4.985488079432037e-07, "logits/chosen": -0.19626545906066895, "logits/rejected": 0.20309990644454956, "logps/chosen": -98.95803833007812, "logps/ref_chosen": -97.22187805175781, "logps/ref_rejected": -92.97674560546875, "logps/rejected": -98.93242645263672, "loss": 0.9327, "margin_dpo/margin_mean": 4.2195258140563965, "margin_dpo/margin_std": 5.781785011291504, "step": 91 }, { "epoch": 0.13509544787077826, "fcm_dpo/beta": 0.21198439598083496, "fcm_dpo/delta": 0.11860194057226181, "fcm_dpo/margin": 3.4903693199157715, "fcm_dpo/q_t": 0.3639296889305115, "grad_norm": 42.03072738647461, "learning_rate": 4.984074589033043e-07, "logits/chosen": -0.795829176902771, "logits/rejected": -0.45754796266555786, "logps/chosen": -86.59548950195312, "logps/ref_chosen": -84.5302734375, "logps/ref_rejected": -84.5013198852539, "logps/rejected": -90.05690002441406, "loss": 1.0081, "margin_dpo/margin_mean": 3.4903693199157715, "margin_dpo/margin_std": 5.338055610656738, "step": 92 }, { "epoch": 0.13656387665198239, "fcm_dpo/beta": 0.21652595698833466, "fcm_dpo/delta": 0.10127636045217514, "fcm_dpo/margin": 3.492115020751953, "fcm_dpo/q_t": 0.3656906485557556, "grad_norm": 58.36591720581055, "learning_rate": 4.982595640958425e-07, "logits/chosen": -0.7161461114883423, "logits/rejected": -0.16940158605575562, "logps/chosen": -92.29219055175781, "logps/ref_chosen": -90.25043487548828, "logps/ref_rejected": -84.0942153930664, "logps/rejected": -89.62808990478516, "loss": 1.068, "margin_dpo/margin_mean": 3.492115020751953, "margin_dpo/margin_std": 5.786944389343262, "step": 93 }, { "epoch": 0.13803230543318648, "fcm_dpo/beta": 0.20668484270572662, "fcm_dpo/delta": -0.22309033572673798, "fcm_dpo/margin": 5.036521911621094, "fcm_dpo/q_t": 0.3098403215408325, "grad_norm": 48.04051971435547, "learning_rate": 4.98105127417984e-07, "logits/chosen": -0.43175843358039856, "logits/rejected": -0.14784827828407288, "logps/chosen": -94.38967895507812, "logps/ref_chosen": -92.4542236328125, "logps/ref_rejected": -105.2472915649414, "logps/rejected": -112.21926879882812, "loss": 0.884, "margin_dpo/margin_mean": 5.036520957946777, "margin_dpo/margin_std": 6.06168794631958, "step": 94 }, { "epoch": 0.1395007342143906, "fcm_dpo/beta": 0.209566131234169, "fcm_dpo/delta": 0.002998150885105133, "fcm_dpo/margin": 4.036772727966309, "fcm_dpo/q_t": 0.34862929582595825, "grad_norm": 42.049312591552734, "learning_rate": 4.979441529392784e-07, "logits/chosen": -0.8037420511245728, "logits/rejected": -0.5004515647888184, "logps/chosen": -80.51316833496094, "logps/ref_chosen": -78.87370300292969, "logps/ref_rejected": -83.59121704101562, "logps/rejected": -89.26744079589844, "loss": 0.9502, "margin_dpo/margin_mean": 4.036772727966309, "margin_dpo/margin_std": 5.532905101776123, "step": 95 }, { "epoch": 0.14096916299559473, "fcm_dpo/beta": 0.20307976007461548, "fcm_dpo/delta": -0.13602276146411896, "fcm_dpo/margin": 4.787096977233887, "fcm_dpo/q_t": 0.3208298683166504, "grad_norm": 39.64480209350586, "learning_rate": 4.977766449015534e-07, "logits/chosen": -0.48742806911468506, "logits/rejected": -0.07584237307310104, "logps/chosen": -107.8035888671875, "logps/ref_chosen": -106.5921630859375, "logps/ref_rejected": -101.76802825927734, "logps/rejected": -107.76654052734375, "loss": 0.8767, "margin_dpo/margin_mean": 4.78709602355957, "margin_dpo/margin_std": 5.776007652282715, "step": 96 }, { "epoch": 0.14243759177679882, "fcm_dpo/beta": 0.20178145170211792, "fcm_dpo/delta": -0.09740416705608368, "fcm_dpo/margin": 4.642391681671143, "fcm_dpo/q_t": 0.3197094202041626, "grad_norm": 41.0906867980957, "learning_rate": 4.976026077188012e-07, "logits/chosen": -0.3646326959133148, "logits/rejected": 0.015910834074020386, "logps/chosen": -90.6146469116211, "logps/ref_chosen": -88.67988586425781, "logps/ref_rejected": -84.81229400634766, "logps/rejected": -91.38944244384766, "loss": 0.8343, "margin_dpo/margin_mean": 4.642391204833984, "margin_dpo/margin_std": 4.824398040771484, "step": 97 }, { "epoch": 0.14390602055800295, "fcm_dpo/beta": 0.19345466792583466, "fcm_dpo/delta": -0.10774275660514832, "fcm_dpo/margin": 4.8845977783203125, "fcm_dpo/q_t": 0.3147512674331665, "grad_norm": 38.62285614013672, "learning_rate": 4.974220459770639e-07, "logits/chosen": -0.33925938606262207, "logits/rejected": -0.09292425215244293, "logps/chosen": -93.5835189819336, "logps/ref_chosen": -92.24249267578125, "logps/ref_rejected": -101.51948547363281, "logps/rejected": -107.74510192871094, "loss": 0.8221, "margin_dpo/margin_mean": 4.884597301483154, "margin_dpo/margin_std": 5.033493518829346, "step": 98 }, { "epoch": 0.14537444933920704, "fcm_dpo/beta": 0.18968458473682404, "fcm_dpo/delta": -0.15581762790679932, "fcm_dpo/margin": 5.220826148986816, "fcm_dpo/q_t": 0.3240944743156433, "grad_norm": 34.411415100097656, "learning_rate": 4.972349644343108e-07, "logits/chosen": -0.3686472177505493, "logits/rejected": -0.16389545798301697, "logps/chosen": -73.22004699707031, "logps/ref_chosen": -72.18464660644531, "logps/ref_rejected": -91.88131713867188, "logps/rejected": -98.13754272460938, "loss": 0.8523, "margin_dpo/margin_mean": 5.220826148986816, "margin_dpo/margin_std": 6.51201057434082, "step": 99 }, { "epoch": 0.14684287812041116, "fcm_dpo/beta": 0.18906737864017487, "fcm_dpo/delta": 0.07484784722328186, "fcm_dpo/margin": 4.128679275512695, "fcm_dpo/q_t": 0.34153318405151367, "grad_norm": 33.857669830322266, "learning_rate": 4.970413680203148e-07, "logits/chosen": -0.4443850517272949, "logits/rejected": 0.00018494389951229095, "logps/chosen": -90.57693481445312, "logps/ref_chosen": -89.51382446289062, "logps/ref_rejected": -81.21713256835938, "logps/rejected": -86.40892028808594, "loss": 0.8851, "margin_dpo/margin_mean": 4.128678798675537, "margin_dpo/margin_std": 4.408847808837891, "step": 100 }, { "epoch": 0.14684287812041116, "eval_fcm_dpo/beta": 0.19106584787368774, "eval_logits/chosen": -0.04890349507331848, "eval_logits/rejected": 0.16974791884422302, "eval_logps/chosen": -102.81055450439453, "eval_logps/ref_chosen": -100.49356842041016, "eval_logps/ref_rejected": -94.06775665283203, "eval_logps/rejected": -99.17350006103516, "eval_loss": 0.5619437098503113, "eval_margin_dpo/margin_mean": 2.788757562637329, "eval_margin_dpo/margin_std": 5.18633508682251, "eval_runtime": 43.7613, "eval_samples_per_second": 53.449, "eval_steps_per_second": 1.691, "step": 100 }, { "epoch": 0.14831130690161526, "fcm_dpo/beta": 0.19471797347068787, "fcm_dpo/delta": 0.16918794810771942, "fcm_dpo/margin": 3.557769536972046, "fcm_dpo/q_t": 0.3666532635688782, "grad_norm": 48.270713806152344, "learning_rate": 4.968412618365215e-07, "logits/chosen": -0.5501360297203064, "logits/rejected": -0.311795175075531, "logps/chosen": -96.4827880859375, "logps/ref_chosen": -94.4031982421875, "logps/ref_rejected": -89.91001892089844, "logps/rejected": -95.54737854003906, "loss": 1.0253, "margin_dpo/margin_mean": 3.557770013809204, "margin_dpo/margin_std": 5.32769775390625, "step": 101 }, { "epoch": 0.14977973568281938, "fcm_dpo/beta": 0.20780380070209503, "fcm_dpo/delta": 0.24079912900924683, "fcm_dpo/margin": 2.968078136444092, "fcm_dpo/q_t": 0.3742523789405823, "grad_norm": 63.03786087036133, "learning_rate": 4.966346511559149e-07, "logits/chosen": -0.6122028827667236, "logits/rejected": -0.29000306129455566, "logps/chosen": -94.04932403564453, "logps/ref_chosen": -91.60942077636719, "logps/ref_rejected": -75.7689208984375, "logps/rejected": -81.17689514160156, "loss": 1.0796, "margin_dpo/margin_mean": 2.9680778980255127, "margin_dpo/margin_std": 4.8967742919921875, "step": 102 }, { "epoch": 0.1512481644640235, "fcm_dpo/beta": 0.20262879133224487, "fcm_dpo/delta": -0.1501111090183258, "fcm_dpo/margin": 4.861822605133057, "fcm_dpo/q_t": 0.3110318183898926, "grad_norm": 44.94874954223633, "learning_rate": 4.964215414228785e-07, "logits/chosen": -0.5446847677230835, "logits/rejected": -0.20542305707931519, "logps/chosen": -98.96780395507812, "logps/ref_chosen": -97.40238952636719, "logps/ref_rejected": -98.981689453125, "logps/rejected": -105.40892028808594, "loss": 0.8525, "margin_dpo/margin_mean": 4.861822605133057, "margin_dpo/margin_std": 5.37109899520874, "step": 103 }, { "epoch": 0.1527165932452276, "fcm_dpo/beta": 0.19483795762062073, "fcm_dpo/delta": -0.2128961682319641, "fcm_dpo/margin": 5.340367317199707, "fcm_dpo/q_t": 0.30966752767562866, "grad_norm": 33.6173210144043, "learning_rate": 4.96201938253052e-07, "logits/chosen": -0.584915816783905, "logits/rejected": -0.27762916684150696, "logps/chosen": -95.84004211425781, "logps/ref_chosen": -94.85636138916016, "logps/ref_rejected": -97.02669525146484, "logps/rejected": -103.35074615478516, "loss": 0.8139, "margin_dpo/margin_mean": 5.340367317199707, "margin_dpo/margin_std": 6.170401573181152, "step": 104 }, { "epoch": 0.15418502202643172, "fcm_dpo/beta": 0.19031333923339844, "fcm_dpo/delta": -0.13976818323135376, "fcm_dpo/margin": 5.128673553466797, "fcm_dpo/q_t": 0.31237882375717163, "grad_norm": 40.28583908081055, "learning_rate": 4.959758474331832e-07, "logits/chosen": -0.6987419724464417, "logits/rejected": -0.28522351384162903, "logps/chosen": -98.16111755371094, "logps/ref_chosen": -96.75790405273438, "logps/ref_rejected": -101.5774917602539, "logps/rejected": -108.10939025878906, "loss": 0.8491, "margin_dpo/margin_mean": 5.128673076629639, "margin_dpo/margin_std": 5.720231056213379, "step": 105 }, { "epoch": 0.15565345080763582, "fcm_dpo/beta": 0.18653175234794617, "fcm_dpo/delta": -0.09469683468341827, "fcm_dpo/margin": 5.013686656951904, "fcm_dpo/q_t": 0.32377827167510986, "grad_norm": 35.92363739013672, "learning_rate": 4.957432749209755e-07, "logits/chosen": -0.38015517592430115, "logits/rejected": 0.013428784906864166, "logps/chosen": -93.89437866210938, "logps/ref_chosen": -92.43791198730469, "logps/ref_rejected": -91.00970458984375, "logps/rejected": -97.47986602783203, "loss": 0.8544, "margin_dpo/margin_mean": 5.0136871337890625, "margin_dpo/margin_std": 5.765174865722656, "step": 106 }, { "epoch": 0.15712187958883994, "fcm_dpo/beta": 0.18005526065826416, "fcm_dpo/delta": -0.16589003801345825, "fcm_dpo/margin": 5.5514326095581055, "fcm_dpo/q_t": 0.32457587122917175, "grad_norm": 35.89270782470703, "learning_rate": 4.955042268449307e-07, "logits/chosen": -0.4841831922531128, "logits/rejected": -0.03617634251713753, "logps/chosen": -105.86752319335938, "logps/ref_chosen": -105.12500762939453, "logps/ref_rejected": -101.1824951171875, "logps/rejected": -107.47644805908203, "loss": 0.8619, "margin_dpo/margin_mean": 5.5514326095581055, "margin_dpo/margin_std": 7.6866655349731445, "step": 107 }, { "epoch": 0.15859030837004406, "fcm_dpo/beta": 0.1751759946346283, "fcm_dpo/delta": -0.052587710320949554, "fcm_dpo/margin": 5.115565776824951, "fcm_dpo/q_t": 0.330201119184494, "grad_norm": 32.2303581237793, "learning_rate": 4.952587095041881e-07, "logits/chosen": -0.5506917238235474, "logits/rejected": -0.06726711988449097, "logps/chosen": -90.73790740966797, "logps/ref_chosen": -89.20936584472656, "logps/ref_rejected": -101.83784484863281, "logps/rejected": -108.48196411132812, "loss": 0.8834, "margin_dpo/margin_mean": 5.115566253662109, "margin_dpo/margin_std": 6.003669738769531, "step": 108 }, { "epoch": 0.16005873715124816, "fcm_dpo/beta": 0.16900481283664703, "fcm_dpo/delta": -0.2684016525745392, "fcm_dpo/margin": 6.4412641525268555, "fcm_dpo/q_t": 0.2995629906654358, "grad_norm": 33.71637725830078, "learning_rate": 4.95006729368358e-07, "logits/chosen": 0.004945341497659683, "logits/rejected": 0.20756444334983826, "logps/chosen": -101.10147094726562, "logps/ref_chosen": -100.18708801269531, "logps/ref_rejected": -103.53739929199219, "logps/rejected": -110.8930435180664, "loss": 0.8032, "margin_dpo/margin_mean": 6.441263675689697, "margin_dpo/margin_std": 6.986750602722168, "step": 109 }, { "epoch": 0.16152716593245228, "fcm_dpo/beta": 0.17018066346645355, "fcm_dpo/delta": 0.16250738501548767, "fcm_dpo/margin": 4.108128547668457, "fcm_dpo/q_t": 0.35990265011787415, "grad_norm": 35.48023986816406, "learning_rate": 4.947482930773511e-07, "logits/chosen": 0.06814084947109222, "logits/rejected": 0.411674439907074, "logps/chosen": -108.29478454589844, "logps/ref_chosen": -106.89572143554688, "logps/ref_rejected": -85.81120300292969, "logps/rejected": -91.31838989257812, "loss": 0.9661, "margin_dpo/margin_mean": 4.108128547668457, "margin_dpo/margin_std": 5.228711128234863, "step": 110 }, { "epoch": 0.16299559471365638, "fcm_dpo/beta": 0.16213089227676392, "fcm_dpo/delta": -0.34137529134750366, "fcm_dpo/margin": 7.0692596435546875, "fcm_dpo/q_t": 0.28307241201400757, "grad_norm": 29.604639053344727, "learning_rate": 4.944834074412042e-07, "logits/chosen": -0.5990385413169861, "logits/rejected": -0.3386729955673218, "logps/chosen": -94.55763244628906, "logps/ref_chosen": -93.3293685913086, "logps/ref_rejected": -104.177490234375, "logps/rejected": -112.47500610351562, "loss": 0.7428, "margin_dpo/margin_mean": 7.069259166717529, "margin_dpo/margin_std": 6.622108459472656, "step": 111 }, { "epoch": 0.1644640234948605, "fcm_dpo/beta": 0.1657009720802307, "fcm_dpo/delta": 0.1916673481464386, "fcm_dpo/margin": 4.050925254821777, "fcm_dpo/q_t": 0.3612492084503174, "grad_norm": 32.211814880371094, "learning_rate": 4.942120794399002e-07, "logits/chosen": -0.8714514374732971, "logits/rejected": -0.47875988483428955, "logps/chosen": -84.03836059570312, "logps/ref_chosen": -82.74703979492188, "logps/ref_rejected": -71.5584487915039, "logps/rejected": -76.90069580078125, "loss": 0.9483, "margin_dpo/margin_mean": 4.050925254821777, "margin_dpo/margin_std": 4.675178527832031, "step": 112 }, { "epoch": 0.16593245227606462, "fcm_dpo/beta": 0.17069506645202637, "fcm_dpo/delta": 0.12468759715557098, "fcm_dpo/margin": 4.301934242248535, "fcm_dpo/q_t": 0.34889668226242065, "grad_norm": 37.83422088623047, "learning_rate": 4.939343162231841e-07, "logits/chosen": -0.5917797088623047, "logits/rejected": -0.1455421894788742, "logps/chosen": -99.65104675292969, "logps/ref_chosen": -97.89949035644531, "logps/ref_rejected": -84.41960144042969, "logps/rejected": -90.47306823730469, "loss": 0.9257, "margin_dpo/margin_mean": 4.301935195922852, "margin_dpo/margin_std": 4.792448997497559, "step": 113 }, { "epoch": 0.16740088105726872, "fcm_dpo/beta": 0.16812041401863098, "fcm_dpo/delta": -0.09158610552549362, "fcm_dpo/margin": 5.540801048278809, "fcm_dpo/q_t": 0.3299529254436493, "grad_norm": 38.36504364013672, "learning_rate": 4.936501251103751e-07, "logits/chosen": -0.43144339323043823, "logits/rejected": 0.10210932791233063, "logps/chosen": -98.132080078125, "logps/ref_chosen": -96.4951171875, "logps/ref_rejected": -92.01536560058594, "logps/rejected": -99.19312286376953, "loss": 0.8925, "margin_dpo/margin_mean": 5.540801048278809, "margin_dpo/margin_std": 6.796547889709473, "step": 114 }, { "epoch": 0.16886930983847284, "fcm_dpo/beta": 0.1658993363380432, "fcm_dpo/delta": -0.11033232510089874, "fcm_dpo/margin": 5.7258405685424805, "fcm_dpo/q_t": 0.31527501344680786, "grad_norm": 31.082290649414062, "learning_rate": 4.933595135901732e-07, "logits/chosen": -0.16044090688228607, "logits/rejected": 0.21673715114593506, "logps/chosen": -104.51092529296875, "logps/ref_chosen": -103.18013000488281, "logps/ref_rejected": -105.71595764160156, "logps/rejected": -112.77259826660156, "loss": 0.8321, "margin_dpo/margin_mean": 5.725839614868164, "margin_dpo/margin_std": 6.0801544189453125, "step": 115 }, { "epoch": 0.17033773861967694, "fcm_dpo/beta": 0.16711236536502838, "fcm_dpo/delta": 0.06599107384681702, "fcm_dpo/margin": 4.719298839569092, "fcm_dpo/q_t": 0.34467944502830505, "grad_norm": 30.001972198486328, "learning_rate": 4.930624893204624e-07, "logits/chosen": -0.4106103777885437, "logits/rejected": -0.010030269622802734, "logps/chosen": -82.48646545410156, "logps/ref_chosen": -81.15837097167969, "logps/ref_rejected": -86.30586242675781, "logps/rejected": -92.3532485961914, "loss": 0.9057, "margin_dpo/margin_mean": 4.71929931640625, "margin_dpo/margin_std": 5.378247261047363, "step": 116 }, { "epoch": 0.17180616740088106, "fcm_dpo/beta": 0.16843508183956146, "fcm_dpo/delta": 0.12693826854228973, "fcm_dpo/margin": 4.342744827270508, "fcm_dpo/q_t": 0.3495209217071533, "grad_norm": 40.84837341308594, "learning_rate": 4.927590601281083e-07, "logits/chosen": -0.2890988290309906, "logits/rejected": 0.05249544978141785, "logps/chosen": -98.79878234863281, "logps/ref_chosen": -97.0809555053711, "logps/ref_rejected": -75.1399154663086, "logps/rejected": -81.20048522949219, "loss": 0.9336, "margin_dpo/margin_mean": 4.342745304107666, "margin_dpo/margin_std": 4.965022563934326, "step": 117 }, { "epoch": 0.17327459618208516, "fcm_dpo/beta": 0.16939415037631989, "fcm_dpo/delta": -0.06589382886886597, "fcm_dpo/margin": 5.3707122802734375, "fcm_dpo/q_t": 0.31360048055648804, "grad_norm": 30.420207977294922, "learning_rate": 4.924492340087524e-07, "logits/chosen": -0.663594663143158, "logits/rejected": -0.3713434636592865, "logps/chosen": -87.47075653076172, "logps/ref_chosen": -86.53623962402344, "logps/ref_rejected": -82.93563842773438, "logps/rejected": -89.24087524414062, "loss": 0.8266, "margin_dpo/margin_mean": 5.370712757110596, "margin_dpo/margin_std": 5.203623294830322, "step": 118 }, { "epoch": 0.17474302496328928, "fcm_dpo/beta": 0.16728976368904114, "fcm_dpo/delta": -0.050966691225767136, "fcm_dpo/margin": 5.35649299621582, "fcm_dpo/q_t": 0.32624873518943787, "grad_norm": 36.59634780883789, "learning_rate": 4.92133019126601e-07, "logits/chosen": -0.38907530903816223, "logits/rejected": -0.10628137737512589, "logps/chosen": -106.18470764160156, "logps/ref_chosen": -104.60908508300781, "logps/ref_rejected": -109.15780639648438, "logps/rejected": -116.08993530273438, "loss": 0.907, "margin_dpo/margin_mean": 5.35649299621582, "margin_dpo/margin_std": 6.589056015014648, "step": 119 }, { "epoch": 0.1762114537444934, "fcm_dpo/beta": 0.15864476561546326, "fcm_dpo/delta": -0.3532995283603668, "fcm_dpo/margin": 7.322626113891602, "fcm_dpo/q_t": 0.29675668478012085, "grad_norm": 42.30807876586914, "learning_rate": 4.918104238142103e-07, "logits/chosen": -0.6697523593902588, "logits/rejected": -0.18009905517101288, "logps/chosen": -115.46632385253906, "logps/ref_chosen": -114.68742370605469, "logps/ref_rejected": -111.18586730957031, "logps/rejected": -119.28739929199219, "loss": 0.823, "margin_dpo/margin_mean": 7.322626113891602, "margin_dpo/margin_std": 8.437666893005371, "step": 120 }, { "epoch": 0.1776798825256975, "fcm_dpo/beta": 0.15266025066375732, "fcm_dpo/delta": -0.10870355367660522, "fcm_dpo/margin": 6.205869674682617, "fcm_dpo/q_t": 0.31219685077667236, "grad_norm": 35.275596618652344, "learning_rate": 4.91481456572267e-07, "logits/chosen": -0.2544522285461426, "logits/rejected": 0.05134699493646622, "logps/chosen": -92.9080810546875, "logps/ref_chosen": -92.11666107177734, "logps/ref_rejected": -114.972900390625, "logps/rejected": -121.97018432617188, "loss": 0.8079, "margin_dpo/margin_mean": 6.205869674682617, "margin_dpo/margin_std": 6.08989143371582, "step": 121 }, { "epoch": 0.17914831130690162, "fcm_dpo/beta": 0.14998239278793335, "fcm_dpo/delta": -0.14806339144706726, "fcm_dpo/margin": 6.558841705322266, "fcm_dpo/q_t": 0.3043653964996338, "grad_norm": 35.02898025512695, "learning_rate": 4.911461260693638e-07, "logits/chosen": -0.5000349283218384, "logits/rejected": -0.36555373668670654, "logps/chosen": -86.06637573242188, "logps/ref_chosen": -84.87757110595703, "logps/ref_rejected": -111.0829849243164, "logps/rejected": -118.83062744140625, "loss": 0.8072, "margin_dpo/margin_mean": 6.558840751647949, "margin_dpo/margin_std": 6.412026405334473, "step": 122 }, { "epoch": 0.18061674008810572, "fcm_dpo/beta": 0.1485665887594223, "fcm_dpo/delta": 0.08498059958219528, "fcm_dpo/margin": 5.186343669891357, "fcm_dpo/q_t": 0.344007283449173, "grad_norm": 31.594209671020508, "learning_rate": 4.908044411417711e-07, "logits/chosen": -0.68192058801651, "logits/rejected": -0.4137950539588928, "logps/chosen": -99.76564025878906, "logps/ref_chosen": -98.47874450683594, "logps/ref_rejected": -93.85417175292969, "logps/rejected": -100.32740783691406, "loss": 0.907, "margin_dpo/margin_mean": 5.186343193054199, "margin_dpo/margin_std": 5.748686790466309, "step": 123 }, { "epoch": 0.18208516886930984, "fcm_dpo/beta": 0.14938881993293762, "fcm_dpo/delta": -0.0725385993719101, "fcm_dpo/margin": 6.131298065185547, "fcm_dpo/q_t": 0.33808284997940063, "grad_norm": 33.024810791015625, "learning_rate": 4.904564107932048e-07, "logits/chosen": -0.5029033422470093, "logits/rejected": -0.2667350769042969, "logps/chosen": -103.74954223632812, "logps/ref_chosen": -102.76290130615234, "logps/ref_rejected": -123.005615234375, "logps/rejected": -130.12356567382812, "loss": 0.9159, "margin_dpo/margin_mean": 6.1312971115112305, "margin_dpo/margin_std": 8.150118827819824, "step": 124 }, { "epoch": 0.18355359765051396, "fcm_dpo/beta": 0.15123271942138672, "fcm_dpo/delta": 0.11664462089538574, "fcm_dpo/margin": 4.905575752258301, "fcm_dpo/q_t": 0.353229820728302, "grad_norm": 33.92650604248047, "learning_rate": 4.90102044194588e-07, "logits/chosen": -0.37007012963294983, "logits/rejected": -0.20071515440940857, "logps/chosen": -92.7006607055664, "logps/ref_chosen": -90.42627716064453, "logps/ref_rejected": -89.24901580810547, "logps/rejected": -96.42897033691406, "loss": 0.9455, "margin_dpo/margin_mean": 4.905576705932617, "margin_dpo/margin_std": 6.141561031341553, "step": 125 }, { "epoch": 0.18502202643171806, "fcm_dpo/beta": 0.14798884093761444, "fcm_dpo/delta": -0.19207918643951416, "fcm_dpo/margin": 6.910327434539795, "fcm_dpo/q_t": 0.2891598343849182, "grad_norm": 30.2923583984375, "learning_rate": 4.897413506838102e-07, "logits/chosen": -0.3179720640182495, "logits/rejected": -0.07812147587537766, "logps/chosen": -95.60994720458984, "logps/ref_chosen": -94.42288208007812, "logps/ref_rejected": -103.88156127929688, "logps/rejected": -111.97894287109375, "loss": 0.7351, "margin_dpo/margin_mean": 6.910326957702637, "margin_dpo/margin_std": 5.4942731857299805, "step": 126 }, { "epoch": 0.18649045521292218, "fcm_dpo/beta": 0.147809237241745, "fcm_dpo/delta": 0.008888594806194305, "fcm_dpo/margin": 5.684311866760254, "fcm_dpo/q_t": 0.3361806869506836, "grad_norm": 35.29644775390625, "learning_rate": 4.89374339765481e-07, "logits/chosen": -0.5459603071212769, "logits/rejected": -0.2612873911857605, "logps/chosen": -86.01332092285156, "logps/ref_chosen": -84.75105285644531, "logps/ref_rejected": -83.54523468017578, "logps/rejected": -90.49181365966797, "loss": 0.9171, "margin_dpo/margin_mean": 5.684311866760254, "margin_dpo/margin_std": 6.909788608551025, "step": 127 }, { "epoch": 0.18795888399412627, "fcm_dpo/beta": 0.14771845936775208, "fcm_dpo/delta": 0.032183483242988586, "fcm_dpo/margin": 5.5521039962768555, "fcm_dpo/q_t": 0.33484911918640137, "grad_norm": 35.253360748291016, "learning_rate": 4.890010211106795e-07, "logits/chosen": -0.11950550228357315, "logits/rejected": 0.2775838077068329, "logps/chosen": -91.87654113769531, "logps/ref_chosen": -90.61964416503906, "logps/ref_rejected": -83.10347747802734, "logps/rejected": -89.9124755859375, "loss": 0.9291, "margin_dpo/margin_mean": 5.552103042602539, "margin_dpo/margin_std": 6.96898078918457, "step": 128 }, { "epoch": 0.1894273127753304, "fcm_dpo/beta": 0.14885255694389343, "fcm_dpo/delta": 0.012820325791835785, "fcm_dpo/margin": 5.623868942260742, "fcm_dpo/q_t": 0.33776140213012695, "grad_norm": 29.690258026123047, "learning_rate": 4.88621404556699e-07, "logits/chosen": -0.12152446806430817, "logits/rejected": 0.14454413950443268, "logps/chosen": -100.38304138183594, "logps/ref_chosen": -98.91870880126953, "logps/ref_rejected": -103.28419494628906, "logps/rejected": -110.37239837646484, "loss": 0.8899, "margin_dpo/margin_mean": 5.623868942260742, "margin_dpo/margin_std": 6.273656845092773, "step": 129 }, { "epoch": 0.19089574155653452, "fcm_dpo/beta": 0.14546489715576172, "fcm_dpo/delta": -0.19717153906822205, "fcm_dpo/margin": 7.054594993591309, "fcm_dpo/q_t": 0.3090992569923401, "grad_norm": 29.339309692382812, "learning_rate": 4.882355001067891e-07, "logits/chosen": -0.4405710697174072, "logits/rejected": -0.1895657330751419, "logps/chosen": -74.39476013183594, "logps/ref_chosen": -73.18368530273438, "logps/ref_rejected": -88.02154541015625, "logps/rejected": -96.28720092773438, "loss": 0.8141, "margin_dpo/margin_mean": 7.054594993591309, "margin_dpo/margin_std": 7.372723579406738, "step": 130 }, { "epoch": 0.19236417033773862, "fcm_dpo/beta": 0.14382612705230713, "fcm_dpo/delta": 0.0152159184217453, "fcm_dpo/margin": 5.799963474273682, "fcm_dpo/q_t": 0.33881255984306335, "grad_norm": 29.034713745117188, "learning_rate": 4.878433179298909e-07, "logits/chosen": -0.49546778202056885, "logits/rejected": -0.2850472331047058, "logps/chosen": -74.73524475097656, "logps/ref_chosen": -73.9564437866211, "logps/ref_rejected": -94.45582580566406, "logps/rejected": -101.03459167480469, "loss": 0.8991, "margin_dpo/margin_mean": 5.799962997436523, "margin_dpo/margin_std": 6.754213333129883, "step": 131 }, { "epoch": 0.19383259911894274, "fcm_dpo/beta": 0.14022360742092133, "fcm_dpo/delta": -0.07685554027557373, "fcm_dpo/margin": 6.553079605102539, "fcm_dpo/q_t": 0.31034043431282043, "grad_norm": 31.93946075439453, "learning_rate": 4.874448683603694e-07, "logits/chosen": -0.5433443188667297, "logits/rejected": -0.17930719256401062, "logps/chosen": -93.70853424072266, "logps/ref_chosen": -92.43948364257812, "logps/ref_rejected": -94.13252258300781, "logps/rejected": -101.95465087890625, "loss": 0.8211, "margin_dpo/margin_mean": 6.553079605102539, "margin_dpo/margin_std": 6.320675849914551, "step": 132 }, { "epoch": 0.19530102790014683, "fcm_dpo/beta": 0.13906362652778625, "fcm_dpo/delta": -0.09275015443563461, "fcm_dpo/margin": 6.717145919799805, "fcm_dpo/q_t": 0.31701478362083435, "grad_norm": 29.01922607421875, "learning_rate": 4.870401618977415e-07, "logits/chosen": -0.8589807748794556, "logits/rejected": -0.5169089436531067, "logps/chosen": -93.42756652832031, "logps/ref_chosen": -93.11666107177734, "logps/ref_rejected": -102.09490966796875, "logps/rejected": -109.12296295166016, "loss": 0.8198, "margin_dpo/margin_mean": 6.717145919799805, "margin_dpo/margin_std": 6.780609130859375, "step": 133 }, { "epoch": 0.19676945668135096, "fcm_dpo/beta": 0.13517756760120392, "fcm_dpo/delta": -0.17839491367340088, "fcm_dpo/margin": 7.475937366485596, "fcm_dpo/q_t": 0.2990626096725464, "grad_norm": 26.34817886352539, "learning_rate": 4.866292092063986e-07, "logits/chosen": -1.06900954246521, "logits/rejected": -0.7887668609619141, "logps/chosen": -84.31695556640625, "logps/ref_chosen": -83.55012512207031, "logps/ref_rejected": -93.8815689086914, "logps/rejected": -102.12433624267578, "loss": 0.777, "margin_dpo/margin_mean": 7.475937366485596, "margin_dpo/margin_std": 6.868247985839844, "step": 134 }, { "epoch": 0.19823788546255505, "fcm_dpo/beta": 0.13384297490119934, "fcm_dpo/delta": 0.0007069706916809082, "fcm_dpo/margin": 6.340190887451172, "fcm_dpo/q_t": 0.3378380239009857, "grad_norm": 36.35404586791992, "learning_rate": 4.862120211153265e-07, "logits/chosen": -0.6139770746231079, "logits/rejected": -0.4515921473503113, "logps/chosen": -78.59407043457031, "logps/ref_chosen": -77.58262634277344, "logps/ref_rejected": -121.05648040771484, "logps/rejected": -128.4081268310547, "loss": 0.8986, "margin_dpo/margin_mean": 6.34019136428833, "margin_dpo/margin_std": 7.819980144500732, "step": 135 }, { "epoch": 0.19970631424375918, "fcm_dpo/beta": 0.1300199031829834, "fcm_dpo/delta": -0.14350877702236176, "fcm_dpo/margin": 7.531214714050293, "fcm_dpo/q_t": 0.29828688502311707, "grad_norm": 25.63352394104004, "learning_rate": 4.857886086178193e-07, "logits/chosen": -0.6523430347442627, "logits/rejected": -0.24616163969039917, "logps/chosen": -94.6618423461914, "logps/ref_chosen": -94.1727294921875, "logps/ref_rejected": -103.26428985595703, "logps/rejected": -111.28461456298828, "loss": 0.7668, "margin_dpo/margin_mean": 7.531215190887451, "margin_dpo/margin_std": 6.398444175720215, "step": 136 }, { "epoch": 0.2011747430249633, "fcm_dpo/beta": 0.11987289786338806, "fcm_dpo/delta": -0.43336063623428345, "fcm_dpo/margin": 10.205141067504883, "fcm_dpo/q_t": 0.2635684013366699, "grad_norm": 27.441190719604492, "learning_rate": 4.853589828711902e-07, "logits/chosen": -0.8289228677749634, "logits/rejected": -0.5975626707077026, "logps/chosen": -78.52019500732422, "logps/ref_chosen": -78.11874389648438, "logps/ref_rejected": -117.92237091064453, "logps/rejected": -128.52896118164062, "loss": 0.6749, "margin_dpo/margin_mean": 10.205141067504883, "margin_dpo/margin_std": 8.54926872253418, "step": 137 }, { "epoch": 0.2026431718061674, "fcm_dpo/beta": 0.11853648722171783, "fcm_dpo/delta": 0.01834501326084137, "fcm_dpo/margin": 7.02825403213501, "fcm_dpo/q_t": 0.3304741680622101, "grad_norm": 24.61989974975586, "learning_rate": 4.849231551964771e-07, "logits/chosen": -0.22558237612247467, "logits/rejected": -0.01865313947200775, "logps/chosen": -94.71937561035156, "logps/ref_chosen": -92.82994079589844, "logps/ref_rejected": -98.64840698242188, "logps/rejected": -107.56608581542969, "loss": 0.8528, "margin_dpo/margin_mean": 7.02825403213501, "margin_dpo/margin_std": 6.998868942260742, "step": 138 }, { "epoch": 0.20411160058737152, "fcm_dpo/beta": 0.11699052155017853, "fcm_dpo/delta": -0.09517881274223328, "fcm_dpo/margin": 8.000020980834961, "fcm_dpo/q_t": 0.30926352739334106, "grad_norm": 25.54629898071289, "learning_rate": 4.844811370781446e-07, "logits/chosen": -0.6505274772644043, "logits/rejected": -0.3552773594856262, "logps/chosen": -79.84962463378906, "logps/ref_chosen": -79.00301361083984, "logps/ref_rejected": -85.82733154296875, "logps/rejected": -94.67395782470703, "loss": 0.8347, "margin_dpo/margin_mean": 8.000020980834961, "margin_dpo/margin_std": 8.218055725097656, "step": 139 }, { "epoch": 0.2055800293685756, "fcm_dpo/beta": 0.11667490005493164, "fcm_dpo/delta": -0.006098256446421146, "fcm_dpo/margin": 7.332958221435547, "fcm_dpo/q_t": 0.3275277018547058, "grad_norm": 24.43271827697754, "learning_rate": 4.840329401637809e-07, "logits/chosen": -0.656264066696167, "logits/rejected": -0.28980007767677307, "logps/chosen": -89.24510192871094, "logps/ref_chosen": -88.35244750976562, "logps/ref_rejected": -90.57476806640625, "logps/rejected": -98.80037689208984, "loss": 0.8603, "margin_dpo/margin_mean": 7.332958221435547, "margin_dpo/margin_std": 7.586727619171143, "step": 140 }, { "epoch": 0.20704845814977973, "fcm_dpo/beta": 0.1137249618768692, "fcm_dpo/delta": -0.17630261182785034, "fcm_dpo/margin": 8.869564056396484, "fcm_dpo/q_t": 0.29895448684692383, "grad_norm": 23.59160804748535, "learning_rate": 4.83578576263792e-07, "logits/chosen": -0.40679359436035156, "logits/rejected": -0.06718328595161438, "logps/chosen": -108.21107482910156, "logps/ref_chosen": -107.22016143798828, "logps/ref_rejected": -104.71223449707031, "logps/rejected": -114.57270812988281, "loss": 0.7693, "margin_dpo/margin_mean": 8.8695650100708, "margin_dpo/margin_std": 8.169231414794922, "step": 141 }, { "epoch": 0.20851688693098386, "fcm_dpo/beta": 0.11140327155590057, "fcm_dpo/delta": -0.03288538381457329, "fcm_dpo/margin": 7.893212795257568, "fcm_dpo/q_t": 0.3255746364593506, "grad_norm": 25.964059829711914, "learning_rate": 4.83118057351089e-07, "logits/chosen": -0.4473114609718323, "logits/rejected": -0.21175801753997803, "logps/chosen": -90.84941101074219, "logps/ref_chosen": -88.72894287109375, "logps/ref_rejected": -99.73038482666016, "logps/rejected": -109.74406433105469, "loss": 0.8421, "margin_dpo/margin_mean": 7.893212795257568, "margin_dpo/margin_std": 8.307409286499023, "step": 142 }, { "epoch": 0.20998531571218795, "fcm_dpo/beta": 0.11170312762260437, "fcm_dpo/delta": 0.037874262779951096, "fcm_dpo/margin": 7.293005466461182, "fcm_dpo/q_t": 0.33844050765037537, "grad_norm": 23.84907341003418, "learning_rate": 4.826513955607734e-07, "logits/chosen": -0.87675940990448, "logits/rejected": -0.46402356028556824, "logps/chosen": -85.97489929199219, "logps/ref_chosen": -85.50457763671875, "logps/ref_rejected": -85.8056640625, "logps/rejected": -93.56900024414062, "loss": 0.9032, "margin_dpo/margin_mean": 7.293004989624023, "margin_dpo/margin_std": 8.4471435546875, "step": 143 }, { "epoch": 0.21145374449339208, "fcm_dpo/beta": 0.11300264298915863, "fcm_dpo/delta": -0.06249871850013733, "fcm_dpo/margin": 8.005437850952148, "fcm_dpo/q_t": 0.3288303315639496, "grad_norm": 29.966264724731445, "learning_rate": 4.821786031898176e-07, "logits/chosen": -0.6753568053245544, "logits/rejected": -0.2179480940103531, "logps/chosen": -105.02165222167969, "logps/ref_chosen": -104.68708801269531, "logps/ref_rejected": -87.71908569335938, "logps/rejected": -96.05908203125, "loss": 0.8923, "margin_dpo/margin_mean": 8.005437850952148, "margin_dpo/margin_std": 9.631026268005371, "step": 144 }, { "epoch": 0.21292217327459617, "fcm_dpo/beta": 0.10966965556144714, "fcm_dpo/delta": -0.08353471755981445, "fcm_dpo/margin": 8.439840316772461, "fcm_dpo/q_t": 0.3153953552246094, "grad_norm": 25.04039764404297, "learning_rate": 4.816996926967401e-07, "logits/chosen": -0.5359320044517517, "logits/rejected": 0.0038733258843421936, "logps/chosen": -89.20046997070312, "logps/ref_chosen": -88.51988983154297, "logps/ref_rejected": -83.76513671875, "logps/rejected": -92.88556671142578, "loss": 0.8383, "margin_dpo/margin_mean": 8.439842224121094, "margin_dpo/margin_std": 8.859256744384766, "step": 145 }, { "epoch": 0.2143906020558003, "fcm_dpo/beta": 0.10515066981315613, "fcm_dpo/delta": -0.19317114353179932, "fcm_dpo/margin": 9.692916870117188, "fcm_dpo/q_t": 0.2934582829475403, "grad_norm": 22.523944854736328, "learning_rate": 4.812146767012779e-07, "logits/chosen": -0.8558279275894165, "logits/rejected": -0.34566935896873474, "logps/chosen": -103.20064544677734, "logps/ref_chosen": -102.90586853027344, "logps/ref_rejected": -88.9493637084961, "logps/rejected": -98.93706512451172, "loss": 0.7536, "margin_dpo/margin_mean": 9.692916870117188, "margin_dpo/margin_std": 8.176064491271973, "step": 146 }, { "epoch": 0.21585903083700442, "fcm_dpo/beta": 0.10192033648490906, "fcm_dpo/delta": -0.21661145985126495, "fcm_dpo/margin": 10.242698669433594, "fcm_dpo/q_t": 0.2886643707752228, "grad_norm": 21.00252342224121, "learning_rate": 4.807235679840536e-07, "logits/chosen": -0.9141993522644043, "logits/rejected": -0.40335512161254883, "logps/chosen": -90.03712463378906, "logps/ref_chosen": -89.26219177246094, "logps/ref_rejected": -76.50011444091797, "logps/rejected": -87.51775360107422, "loss": 0.754, "margin_dpo/margin_mean": 10.242698669433594, "margin_dpo/margin_std": 8.800421714782715, "step": 147 }, { "epoch": 0.2173274596182085, "fcm_dpo/beta": 0.10235866904258728, "fcm_dpo/delta": 0.02111022174358368, "fcm_dpo/margin": 8.084315299987793, "fcm_dpo/q_t": 0.338812917470932, "grad_norm": 25.80347442626953, "learning_rate": 4.802263794862384e-07, "logits/chosen": -0.9320132732391357, "logits/rejected": -0.5644128322601318, "logps/chosen": -91.48237609863281, "logps/ref_chosen": -91.76287841796875, "logps/ref_rejected": -110.38552856445312, "logps/rejected": -118.18933868408203, "loss": 0.8942, "margin_dpo/margin_mean": 8.08431625366211, "margin_dpo/margin_std": 9.086051940917969, "step": 148 }, { "epoch": 0.21879588839941264, "fcm_dpo/beta": 0.10116075724363327, "fcm_dpo/delta": 0.01745833456516266, "fcm_dpo/margin": 8.242944717407227, "fcm_dpo/q_t": 0.3417103588581085, "grad_norm": 25.629846572875977, "learning_rate": 4.797231243092118e-07, "logits/chosen": -0.5707268714904785, "logits/rejected": -0.29623883962631226, "logps/chosen": -94.82536315917969, "logps/ref_chosen": -93.48401641845703, "logps/ref_rejected": -105.6436996459961, "logps/rejected": -115.22799682617188, "loss": 0.9025, "margin_dpo/margin_mean": 8.24294662475586, "margin_dpo/margin_std": 9.996318817138672, "step": 149 }, { "epoch": 0.22026431718061673, "fcm_dpo/beta": 0.10018637031316757, "fcm_dpo/delta": -0.07022020220756531, "fcm_dpo/margin": 9.116851806640625, "fcm_dpo/q_t": 0.3224765360355377, "grad_norm": 21.188783645629883, "learning_rate": 4.792138157142157e-07, "logits/chosen": -0.36832523345947266, "logits/rejected": -0.0916675478219986, "logps/chosen": -76.63916015625, "logps/ref_chosen": -75.7786865234375, "logps/ref_rejected": -90.02676391601562, "logps/rejected": -100.00409698486328, "loss": 0.8448, "margin_dpo/margin_mean": 9.116851806640625, "margin_dpo/margin_std": 10.08909797668457, "step": 150 }, { "epoch": 0.22173274596182085, "fcm_dpo/beta": 0.10023315250873566, "fcm_dpo/delta": -0.0017028320580720901, "fcm_dpo/margin": 8.495565414428711, "fcm_dpo/q_t": 0.3360986113548279, "grad_norm": 28.195302963256836, "learning_rate": 4.786984671220053e-07, "logits/chosen": -1.2456610202789307, "logits/rejected": -0.7084729671478271, "logps/chosen": -104.43856811523438, "logps/ref_chosen": -102.73422241210938, "logps/ref_rejected": -106.03781127929688, "logps/rejected": -116.23772430419922, "loss": 0.8976, "margin_dpo/margin_mean": 8.495565414428711, "margin_dpo/margin_std": 10.185565948486328, "step": 151 }, { "epoch": 0.22320117474302498, "fcm_dpo/beta": 0.0979384109377861, "fcm_dpo/delta": -0.1397112011909485, "fcm_dpo/margin": 9.964031219482422, "fcm_dpo/q_t": 0.30377432703971863, "grad_norm": 23.11962127685547, "learning_rate": 4.78177092112495e-07, "logits/chosen": -0.8004452586174011, "logits/rejected": -0.4601547420024872, "logps/chosen": -97.12671661376953, "logps/ref_chosen": -97.59074401855469, "logps/ref_rejected": -110.49913024902344, "logps/rejected": -119.9991455078125, "loss": 0.7848, "margin_dpo/margin_mean": 9.964032173156738, "margin_dpo/margin_std": 9.50239372253418, "step": 152 }, { "epoch": 0.22466960352422907, "fcm_dpo/beta": 0.0991874635219574, "fcm_dpo/delta": 0.12609340250492096, "fcm_dpo/margin": 7.391656875610352, "fcm_dpo/q_t": 0.3547060787677765, "grad_norm": 22.816524505615234, "learning_rate": 4.776497044244016e-07, "logits/chosen": -1.251391887664795, "logits/rejected": -0.9469621181488037, "logps/chosen": -88.66514587402344, "logps/ref_chosen": -87.02560424804688, "logps/ref_rejected": -91.06854248046875, "logps/rejected": -100.09973907470703, "loss": 0.9558, "margin_dpo/margin_mean": 7.391656875610352, "margin_dpo/margin_std": 9.42247200012207, "step": 153 }, { "epoch": 0.2261380323054332, "fcm_dpo/beta": 0.096708282828331, "fcm_dpo/delta": -0.32071927189826965, "fcm_dpo/margin": 11.730560302734375, "fcm_dpo/q_t": 0.2757687270641327, "grad_norm": 22.773046493530273, "learning_rate": 4.771163179548808e-07, "logits/chosen": -0.9404456615447998, "logits/rejected": -0.6278473734855652, "logps/chosen": -91.3301010131836, "logps/ref_chosen": -91.10566711425781, "logps/ref_rejected": -111.37238311767578, "logps/rejected": -123.32737731933594, "loss": 0.7088, "margin_dpo/margin_mean": 11.730561256408691, "margin_dpo/margin_std": 9.345605850219727, "step": 154 }, { "epoch": 0.2276064610866373, "fcm_dpo/beta": 0.09173603355884552, "fcm_dpo/delta": -0.07401001453399658, "fcm_dpo/margin": 9.972280502319336, "fcm_dpo/q_t": 0.3208170533180237, "grad_norm": 28.64071273803711, "learning_rate": 4.7657694675916247e-07, "logits/chosen": -0.5264374017715454, "logits/rejected": -0.19909337162971497, "logps/chosen": -95.87799072265625, "logps/ref_chosen": -94.56097412109375, "logps/ref_rejected": -84.77359008789062, "logps/rejected": -96.06289672851562, "loss": 0.8954, "margin_dpo/margin_mean": 9.972282409667969, "margin_dpo/margin_std": 11.702651977539062, "step": 155 }, { "epoch": 0.2290748898678414, "fcm_dpo/beta": 0.0929301306605339, "fcm_dpo/delta": 0.06338099390268326, "fcm_dpo/margin": 8.519001007080078, "fcm_dpo/q_t": 0.3422147035598755, "grad_norm": 27.464357376098633, "learning_rate": 4.7603160505017893e-07, "logits/chosen": -0.9737367630004883, "logits/rejected": -0.5267705321311951, "logps/chosen": -97.08377075195312, "logps/ref_chosen": -94.65538787841797, "logps/ref_rejected": -84.41934204101562, "logps/rejected": -95.36672973632812, "loss": 0.918, "margin_dpo/margin_mean": 8.519001007080078, "margin_dpo/margin_std": 10.034339904785156, "step": 156 }, { "epoch": 0.2305433186490455, "fcm_dpo/beta": 0.09153124690055847, "fcm_dpo/delta": -0.16219165921211243, "fcm_dpo/margin": 10.88486385345459, "fcm_dpo/q_t": 0.30015528202056885, "grad_norm": 23.287403106689453, "learning_rate": 4.7548030719819154e-07, "logits/chosen": -1.3203184604644775, "logits/rejected": -0.8898429870605469, "logps/chosen": -99.07041931152344, "logps/ref_chosen": -97.01239013671875, "logps/ref_rejected": -111.70133972167969, "logps/rejected": -124.64424133300781, "loss": 0.7785, "margin_dpo/margin_mean": 10.884864807128906, "margin_dpo/margin_std": 9.999959945678711, "step": 157 }, { "epoch": 0.23201174743024963, "fcm_dpo/beta": 0.0887395441532135, "fcm_dpo/delta": -0.15405318140983582, "fcm_dpo/margin": 11.145757675170898, "fcm_dpo/q_t": 0.297393798828125, "grad_norm": 23.948062896728516, "learning_rate": 4.7492306773041136e-07, "logits/chosen": -1.0246334075927734, "logits/rejected": -0.7327722311019897, "logps/chosen": -96.69784545898438, "logps/ref_chosen": -94.71296691894531, "logps/ref_rejected": -120.93733978271484, "logps/rejected": -134.06797790527344, "loss": 0.7641, "margin_dpo/margin_mean": 11.145757675170898, "margin_dpo/margin_std": 9.565597534179688, "step": 158 }, { "epoch": 0.23348017621145375, "fcm_dpo/beta": 0.08581481873989105, "fcm_dpo/delta": -0.12313660234212875, "fcm_dpo/margin": 11.194160461425781, "fcm_dpo/q_t": 0.31181782484054565, "grad_norm": 20.498376846313477, "learning_rate": 4.743599013306165e-07, "logits/chosen": -1.4571746587753296, "logits/rejected": -1.0124716758728027, "logps/chosen": -109.18254852294922, "logps/ref_chosen": -108.2182388305664, "logps/ref_rejected": -94.83901977539062, "logps/rejected": -106.99748229980469, "loss": 0.8284, "margin_dpo/margin_mean": 11.194160461425781, "margin_dpo/margin_std": 11.655582427978516, "step": 159 }, { "epoch": 0.23494860499265785, "fcm_dpo/beta": 0.08508424460887909, "fcm_dpo/delta": -0.025002997368574142, "fcm_dpo/margin": 10.25672721862793, "fcm_dpo/q_t": 0.3192032277584076, "grad_norm": 20.28183364868164, "learning_rate": 4.737908228387656e-07, "logits/chosen": -1.2797681093215942, "logits/rejected": -0.7797454595565796, "logps/chosen": -98.57347106933594, "logps/ref_chosen": -98.16340637207031, "logps/ref_rejected": -102.28097534179688, "logps/rejected": -112.9477767944336, "loss": 0.8396, "margin_dpo/margin_mean": 10.25672721862793, "margin_dpo/margin_std": 9.986027717590332, "step": 160 }, { "epoch": 0.23641703377386197, "fcm_dpo/beta": 0.0846460685133934, "fcm_dpo/delta": 0.005213946104049683, "fcm_dpo/margin": 9.975399017333984, "fcm_dpo/q_t": 0.3322407305240631, "grad_norm": 22.53128433227539, "learning_rate": 4.7321584725060594e-07, "logits/chosen": -1.2773491144180298, "logits/rejected": -0.8251501321792603, "logps/chosen": -88.7325439453125, "logps/ref_chosen": -86.59187316894531, "logps/ref_rejected": -89.01643371582031, "logps/rejected": -101.13250732421875, "loss": 0.8939, "margin_dpo/margin_mean": 9.975399017333984, "margin_dpo/margin_std": 11.190435409545898, "step": 161 }, { "epoch": 0.23788546255506607, "fcm_dpo/beta": 0.08452028781175613, "fcm_dpo/delta": -0.03935426473617554, "fcm_dpo/margin": 10.478342056274414, "fcm_dpo/q_t": 0.3177115321159363, "grad_norm": 20.945478439331055, "learning_rate": 4.7263498971727905e-07, "logits/chosen": -0.6811781525611877, "logits/rejected": -0.29194605350494385, "logps/chosen": -98.99219512939453, "logps/ref_chosen": -97.40288543701172, "logps/ref_rejected": -95.86923217773438, "logps/rejected": -107.93688201904297, "loss": 0.8561, "margin_dpo/margin_mean": 10.478342056274414, "margin_dpo/margin_std": 10.870805740356445, "step": 162 }, { "epoch": 0.2393538913362702, "fcm_dpo/beta": 0.08143004029989243, "fcm_dpo/delta": -0.242702916264534, "fcm_dpo/margin": 13.094063758850098, "fcm_dpo/q_t": 0.2967764139175415, "grad_norm": 17.559616088867188, "learning_rate": 4.720482655449212e-07, "logits/chosen": -1.1392264366149902, "logits/rejected": -0.7611545324325562, "logps/chosen": -83.36274719238281, "logps/ref_chosen": -82.1753921508789, "logps/ref_rejected": -82.86473846435547, "logps/rejected": -97.14615631103516, "loss": 0.7736, "margin_dpo/margin_mean": 13.094064712524414, "margin_dpo/margin_std": 13.233580589294434, "step": 163 }, { "epoch": 0.24082232011747431, "fcm_dpo/beta": 0.08133898675441742, "fcm_dpo/delta": 0.14469550549983978, "fcm_dpo/margin": 8.78353214263916, "fcm_dpo/q_t": 0.3501974046230316, "grad_norm": 20.816787719726562, "learning_rate": 4.714556901942599e-07, "logits/chosen": -0.9682873487472534, "logits/rejected": -0.5573686957359314, "logps/chosen": -94.12632751464844, "logps/ref_chosen": -90.47427368164062, "logps/ref_rejected": -87.02809143066406, "logps/rejected": -99.46368408203125, "loss": 0.9361, "margin_dpo/margin_mean": 8.783533096313477, "margin_dpo/margin_std": 9.51855182647705, "step": 164 }, { "epoch": 0.2422907488986784, "fcm_dpo/beta": 0.08096069097518921, "fcm_dpo/delta": -0.20310097932815552, "fcm_dpo/margin": 12.753755569458008, "fcm_dpo/q_t": 0.3054801821708679, "grad_norm": 18.420368194580078, "learning_rate": 4.708572792802069e-07, "logits/chosen": -1.2523810863494873, "logits/rejected": -0.8975613117218018, "logps/chosen": -88.87324523925781, "logps/ref_chosen": -87.313232421875, "logps/ref_rejected": -81.902587890625, "logps/rejected": -96.21635437011719, "loss": 0.7986, "margin_dpo/margin_mean": 12.753755569458008, "margin_dpo/margin_std": 14.236822128295898, "step": 165 }, { "epoch": 0.24375917767988253, "fcm_dpo/beta": 0.0783786028623581, "fcm_dpo/delta": -0.11244317889213562, "fcm_dpo/margin": 12.143659591674805, "fcm_dpo/q_t": 0.322204053401947, "grad_norm": 20.77262306213379, "learning_rate": 4.702530485714461e-07, "logits/chosen": -0.9464759230613708, "logits/rejected": -0.7219276428222656, "logps/chosen": -86.5793228149414, "logps/ref_chosen": -81.58606719970703, "logps/ref_rejected": -103.16517639160156, "logps/rejected": -120.30209350585938, "loss": 0.8471, "margin_dpo/margin_mean": 12.143659591674805, "margin_dpo/margin_std": 14.19489860534668, "step": 166 }, { "epoch": 0.24522760646108663, "fcm_dpo/beta": 0.0751272439956665, "fcm_dpo/delta": -0.183059424161911, "fcm_dpo/margin": 13.46567440032959, "fcm_dpo/q_t": 0.2956295907497406, "grad_norm": 20.29990005493164, "learning_rate": 4.6964301399001877e-07, "logits/chosen": -0.7697356939315796, "logits/rejected": -0.5615593194961548, "logps/chosen": -87.39271545410156, "logps/ref_chosen": -83.56167602539062, "logps/ref_rejected": -99.05305480957031, "logps/rejected": -116.34976959228516, "loss": 0.7512, "margin_dpo/margin_mean": 13.465675354003906, "margin_dpo/margin_std": 11.612232208251953, "step": 167 }, { "epoch": 0.24669603524229075, "fcm_dpo/beta": 0.07545551657676697, "fcm_dpo/delta": 0.006653144955635071, "fcm_dpo/margin": 11.17927360534668, "fcm_dpo/q_t": 0.33007046580314636, "grad_norm": 22.19552230834961, "learning_rate": 4.690271916109034e-07, "logits/chosen": -0.9670852422714233, "logits/rejected": -0.6387625336647034, "logps/chosen": -83.56878662109375, "logps/ref_chosen": -79.93123626708984, "logps/ref_rejected": -80.10216522216797, "logps/rejected": -94.91899108886719, "loss": 0.8696, "margin_dpo/margin_mean": 11.17927360534668, "margin_dpo/margin_std": 11.96432113647461, "step": 168 }, { "epoch": 0.24816446402349487, "fcm_dpo/beta": 0.07446470856666565, "fcm_dpo/delta": -0.08862879127264023, "fcm_dpo/margin": 12.494746208190918, "fcm_dpo/q_t": 0.3135685920715332, "grad_norm": 20.065807342529297, "learning_rate": 4.6840559766159235e-07, "logits/chosen": -0.9137426018714905, "logits/rejected": -0.6862479448318481, "logps/chosen": -85.2474365234375, "logps/ref_chosen": -82.90370178222656, "logps/ref_rejected": -90.12191009521484, "logps/rejected": -104.96038818359375, "loss": 0.8135, "margin_dpo/margin_mean": 12.494745254516602, "margin_dpo/margin_std": 12.275469779968262, "step": 169 }, { "epoch": 0.24963289280469897, "fcm_dpo/beta": 0.07337189465761185, "fcm_dpo/delta": 0.003176748752593994, "fcm_dpo/margin": 11.533134460449219, "fcm_dpo/q_t": 0.3243024945259094, "grad_norm": 20.527284622192383, "learning_rate": 4.6777824852166437e-07, "logits/chosen": -1.0314369201660156, "logits/rejected": -0.7973779439926147, "logps/chosen": -79.22702026367188, "logps/ref_chosen": -74.98197174072266, "logps/ref_rejected": -82.16560363769531, "logps/rejected": -97.94378662109375, "loss": 0.8447, "margin_dpo/margin_mean": 11.533134460449219, "margin_dpo/margin_std": 10.969135284423828, "step": 170 }, { "epoch": 0.2511013215859031, "fcm_dpo/beta": 0.07268837094306946, "fcm_dpo/delta": -0.11282499134540558, "fcm_dpo/margin": 13.098483085632324, "fcm_dpo/q_t": 0.32213613390922546, "grad_norm": 22.617277145385742, "learning_rate": 4.6714516072235273e-07, "logits/chosen": -1.1407670974731445, "logits/rejected": -0.6412278413772583, "logps/chosen": -123.88941955566406, "logps/ref_chosen": -118.97853088378906, "logps/ref_rejected": -115.76930236816406, "logps/rejected": -133.77867126464844, "loss": 0.8676, "margin_dpo/margin_mean": 13.098483085632324, "margin_dpo/margin_std": 15.950417518615723, "step": 171 }, { "epoch": 0.2525697503671072, "fcm_dpo/beta": 0.0713869035243988, "fcm_dpo/delta": -0.04091132432222366, "fcm_dpo/margin": 12.417065620422363, "fcm_dpo/q_t": 0.30859920382499695, "grad_norm": 16.54987907409668, "learning_rate": 4.6650635094610966e-07, "logits/chosen": -1.8607536554336548, "logits/rejected": -1.2343878746032715, "logps/chosen": -106.86568450927734, "logps/ref_chosen": -101.15379333496094, "logps/ref_rejected": -91.00568389892578, "logps/rejected": -109.1346435546875, "loss": 0.7824, "margin_dpo/margin_mean": 12.417064666748047, "margin_dpo/margin_std": 9.314888000488281, "step": 172 }, { "epoch": 0.2540381791483113, "fcm_dpo/beta": 0.07276591658592224, "fcm_dpo/delta": 0.10682529211044312, "fcm_dpo/margin": 10.32400131225586, "fcm_dpo/q_t": 0.34479019045829773, "grad_norm": 18.29812240600586, "learning_rate": 4.6586183602616687e-07, "logits/chosen": -0.8446004390716553, "logits/rejected": -0.4139299988746643, "logps/chosen": -104.5085678100586, "logps/ref_chosen": -99.05668640136719, "logps/ref_rejected": -83.8507308959961, "logps/rejected": -99.62661743164062, "loss": 0.9065, "margin_dpo/margin_mean": 10.32400131225586, "margin_dpo/margin_std": 11.158248901367188, "step": 173 }, { "epoch": 0.2555066079295154, "fcm_dpo/beta": 0.07212045043706894, "fcm_dpo/delta": -0.11132963001728058, "fcm_dpo/margin": 13.184260368347168, "fcm_dpo/q_t": 0.3065446615219116, "grad_norm": 21.124048233032227, "learning_rate": 4.652116329460919e-07, "logits/chosen": -1.4102823734283447, "logits/rejected": -1.3270244598388672, "logps/chosen": -85.65483856201172, "logps/ref_chosen": -80.40890502929688, "logps/ref_rejected": -108.20171356201172, "logps/rejected": -126.63190460205078, "loss": 0.8031, "margin_dpo/margin_mean": 13.184259414672852, "margin_dpo/margin_std": 12.659872055053711, "step": 174 }, { "epoch": 0.25697503671071953, "fcm_dpo/beta": 0.07079780846834183, "fcm_dpo/delta": -0.12347761541604996, "fcm_dpo/margin": 13.580202102661133, "fcm_dpo/q_t": 0.30228978395462036, "grad_norm": 21.37623405456543, "learning_rate": 4.645557588393406e-07, "logits/chosen": -0.736810564994812, "logits/rejected": -0.47296372056007385, "logps/chosen": -90.78504943847656, "logps/ref_chosen": -82.71649169921875, "logps/ref_rejected": -96.66712951660156, "logps/rejected": -118.31588745117188, "loss": 0.7768, "margin_dpo/margin_mean": 13.580202102661133, "margin_dpo/margin_std": 11.554082870483398, "step": 175 }, { "epoch": 0.25844346549192365, "fcm_dpo/beta": 0.06876523047685623, "fcm_dpo/delta": -0.11029569059610367, "fcm_dpo/margin": 13.813409805297852, "fcm_dpo/q_t": 0.3185492753982544, "grad_norm": 17.933225631713867, "learning_rate": 4.638942309888058e-07, "logits/chosen": -1.2823872566223145, "logits/rejected": -1.0857605934143066, "logps/chosen": -85.75364685058594, "logps/ref_chosen": -79.62930297851562, "logps/ref_rejected": -102.6885986328125, "logps/rejected": -122.62635040283203, "loss": 0.8524, "margin_dpo/margin_mean": 13.813409805297852, "margin_dpo/margin_std": 15.199995040893555, "step": 176 }, { "epoch": 0.2599118942731278, "fcm_dpo/beta": 0.06667226552963257, "fcm_dpo/delta": -0.22711145877838135, "fcm_dpo/margin": 15.801198959350586, "fcm_dpo/q_t": 0.3023749887943268, "grad_norm": 16.617761611938477, "learning_rate": 4.6322706682636137e-07, "logits/chosen": -1.1350969076156616, "logits/rejected": -0.7791039943695068, "logps/chosen": -99.27539825439453, "logps/ref_chosen": -92.56109619140625, "logps/ref_rejected": -103.33839416503906, "logps/rejected": -125.85388946533203, "loss": 0.7779, "margin_dpo/margin_mean": 15.801198959350586, "margin_dpo/margin_std": 16.495779037475586, "step": 177 }, { "epoch": 0.26138032305433184, "fcm_dpo/beta": 0.0618768036365509, "fcm_dpo/delta": -0.3499090075492859, "fcm_dpo/margin": 18.723678588867188, "fcm_dpo/q_t": 0.2736192047595978, "grad_norm": 19.030345916748047, "learning_rate": 4.6255428393240354e-07, "logits/chosen": -1.4820408821105957, "logits/rejected": -1.190525770187378, "logps/chosen": -90.46121215820312, "logps/ref_chosen": -84.26579284667969, "logps/ref_rejected": -110.79730987548828, "logps/rejected": -135.71640014648438, "loss": 0.6981, "margin_dpo/margin_mean": 18.723678588867188, "margin_dpo/margin_std": 16.16928482055664, "step": 178 }, { "epoch": 0.26284875183553597, "fcm_dpo/beta": 0.06120564043521881, "fcm_dpo/delta": 0.0761781707406044, "fcm_dpo/margin": 12.736023902893066, "fcm_dpo/q_t": 0.34814584255218506, "grad_norm": 19.742097854614258, "learning_rate": 4.6187590003538724e-07, "logits/chosen": -1.2575550079345703, "logits/rejected": -0.8413559198379517, "logps/chosen": -102.05105590820312, "logps/ref_chosen": -93.4479751586914, "logps/ref_rejected": -94.25541687011719, "logps/rejected": -115.59451293945312, "loss": 0.9468, "margin_dpo/margin_mean": 12.73602294921875, "margin_dpo/margin_std": 15.995349884033203, "step": 179 }, { "epoch": 0.2643171806167401, "fcm_dpo/beta": 0.06053383648395538, "fcm_dpo/delta": -0.10252348333597183, "fcm_dpo/margin": 15.557160377502441, "fcm_dpo/q_t": 0.3080785870552063, "grad_norm": 20.96506118774414, "learning_rate": 4.611919330113591e-07, "logits/chosen": -1.0803053379058838, "logits/rejected": -0.8237367272377014, "logps/chosen": -95.61346435546875, "logps/ref_chosen": -88.83554077148438, "logps/ref_rejected": -103.90696716308594, "logps/rejected": -126.24205017089844, "loss": 0.7905, "margin_dpo/margin_mean": 15.557161331176758, "margin_dpo/margin_std": 14.514302253723145, "step": 180 }, { "epoch": 0.2657856093979442, "fcm_dpo/beta": 0.06162776052951813, "fcm_dpo/delta": 0.10773488879203796, "fcm_dpo/margin": 12.171871185302734, "fcm_dpo/q_t": 0.34699884057044983, "grad_norm": 19.01604652404785, "learning_rate": 4.605024008834863e-07, "logits/chosen": -0.9782782793045044, "logits/rejected": -0.5275274515151978, "logps/chosen": -86.51460266113281, "logps/ref_chosen": -79.23396301269531, "logps/ref_rejected": -69.17442321777344, "logps/rejected": -88.6269302368164, "loss": 0.9037, "margin_dpo/margin_mean": 12.171870231628418, "margin_dpo/margin_std": 13.407722473144531, "step": 181 }, { "epoch": 0.26725403817914833, "fcm_dpo/beta": 0.06035807728767395, "fcm_dpo/delta": -0.1120065227150917, "fcm_dpo/margin": 15.727120399475098, "fcm_dpo/q_t": 0.3143818974494934, "grad_norm": 16.87702178955078, "learning_rate": 4.598073218215817e-07, "logits/chosen": -0.9282777905464172, "logits/rejected": -0.6541332006454468, "logps/chosen": -86.01435089111328, "logps/ref_chosen": -76.08218383789062, "logps/ref_rejected": -94.9686050415039, "logps/rejected": -120.62789916992188, "loss": 0.8428, "margin_dpo/margin_mean": 15.727121353149414, "margin_dpo/margin_std": 16.90331268310547, "step": 182 }, { "epoch": 0.2687224669603524, "fcm_dpo/beta": 0.05917463079094887, "fcm_dpo/delta": -0.11101389676332474, "fcm_dpo/margin": 16.041793823242188, "fcm_dpo/q_t": 0.3082314133644104, "grad_norm": 18.056352615356445, "learning_rate": 4.5910671414162484e-07, "logits/chosen": -1.2095495462417603, "logits/rejected": -0.6235491037368774, "logps/chosen": -103.69279479980469, "logps/ref_chosen": -93.90790557861328, "logps/ref_rejected": -83.56061553955078, "logps/rejected": -109.38729858398438, "loss": 0.7971, "margin_dpo/margin_mean": 16.041793823242188, "margin_dpo/margin_std": 14.786785125732422, "step": 183 }, { "epoch": 0.2701908957415565, "fcm_dpo/beta": 0.061080534011125565, "fcm_dpo/delta": 0.21639560163021088, "fcm_dpo/margin": 10.603964805603027, "fcm_dpo/q_t": 0.36788150668144226, "grad_norm": 20.342153549194336, "learning_rate": 4.5840059630527985e-07, "logits/chosen": -1.476677656173706, "logits/rejected": -1.261820912361145, "logps/chosen": -95.60768127441406, "logps/ref_chosen": -86.86790466308594, "logps/ref_rejected": -84.28456115722656, "logps/rejected": -103.6282958984375, "loss": 1.0314, "margin_dpo/margin_mean": 10.603963851928711, "margin_dpo/margin_std": 15.710251808166504, "step": 184 }, { "epoch": 0.27165932452276065, "fcm_dpo/beta": 0.06259152293205261, "fcm_dpo/delta": 0.04567991569638252, "fcm_dpo/margin": 12.893394470214844, "fcm_dpo/q_t": 0.34303832054138184, "grad_norm": 23.124479293823242, "learning_rate": 4.5768898691940836e-07, "logits/chosen": -1.7452738285064697, "logits/rejected": -1.1792840957641602, "logps/chosen": -103.65872955322266, "logps/ref_chosen": -94.02142333984375, "logps/ref_rejected": -81.45668029785156, "logps/rejected": -103.98738098144531, "loss": 0.943, "margin_dpo/margin_mean": 12.89339542388916, "margin_dpo/margin_std": 16.560266494750977, "step": 185 }, { "epoch": 0.27312775330396477, "fcm_dpo/beta": 0.06136384233832359, "fcm_dpo/delta": -0.06328608095645905, "fcm_dpo/margin": 14.769530296325684, "fcm_dpo/q_t": 0.3183685541152954, "grad_norm": 21.316608428955078, "learning_rate": 4.5697190473557947e-07, "logits/chosen": -2.0271401405334473, "logits/rejected": -1.2335484027862549, "logps/chosen": -125.817626953125, "logps/ref_chosen": -114.19168090820312, "logps/ref_rejected": -93.33912658691406, "logps/rejected": -119.73458862304688, "loss": 0.8408, "margin_dpo/margin_mean": 14.769529342651367, "margin_dpo/margin_std": 15.083984375, "step": 186 }, { "epoch": 0.2745961820851689, "fcm_dpo/beta": 0.06224139779806137, "fcm_dpo/delta": 0.0447879321873188, "fcm_dpo/margin": 12.98681354522705, "fcm_dpo/q_t": 0.34211674332618713, "grad_norm": 18.177583694458008, "learning_rate": 4.5624936864957555e-07, "logits/chosen": -1.5803933143615723, "logits/rejected": -1.344294786453247, "logps/chosen": -86.82991790771484, "logps/ref_chosen": -76.86955261230469, "logps/ref_rejected": -88.16494750976562, "logps/rejected": -111.11212158203125, "loss": 0.9328, "margin_dpo/margin_mean": 12.986814498901367, "margin_dpo/margin_std": 16.151668548583984, "step": 187 }, { "epoch": 0.27606461086637296, "fcm_dpo/beta": 0.06076783686876297, "fcm_dpo/delta": -0.08916963636875153, "fcm_dpo/margin": 15.28190803527832, "fcm_dpo/q_t": 0.3073975741863251, "grad_norm": 17.986310958862305, "learning_rate": 4.5552139770089454e-07, "logits/chosen": -1.4565420150756836, "logits/rejected": -1.1993229389190674, "logps/chosen": -86.23423767089844, "logps/ref_chosen": -79.1948013305664, "logps/ref_rejected": -96.82889556884766, "logps/rejected": -119.1502456665039, "loss": 0.8328, "margin_dpo/margin_mean": 15.281907081604004, "margin_dpo/margin_std": 14.838127136230469, "step": 188 }, { "epoch": 0.2775330396475771, "fcm_dpo/beta": 0.06060121953487396, "fcm_dpo/delta": -0.12682175636291504, "fcm_dpo/margin": 15.897202491760254, "fcm_dpo/q_t": 0.3206084072589874, "grad_norm": 16.118820190429688, "learning_rate": 4.5478801107224794e-07, "logits/chosen": -1.2777214050292969, "logits/rejected": -0.6506960988044739, "logps/chosen": -93.4096450805664, "logps/ref_chosen": -83.8927993774414, "logps/ref_rejected": -79.46589660644531, "logps/rejected": -104.87994384765625, "loss": 0.8542, "margin_dpo/margin_mean": 15.89720344543457, "margin_dpo/margin_std": 18.334991455078125, "step": 189 }, { "epoch": 0.2790014684287812, "fcm_dpo/beta": 0.05765543505549431, "fcm_dpo/delta": -0.19643330574035645, "fcm_dpo/margin": 17.790395736694336, "fcm_dpo/q_t": 0.29139959812164307, "grad_norm": 17.58658790588379, "learning_rate": 4.5404922808905543e-07, "logits/chosen": -1.5000858306884766, "logits/rejected": -0.9203025102615356, "logps/chosen": -110.40105438232422, "logps/ref_chosen": -103.44146728515625, "logps/ref_rejected": -109.21502685546875, "logps/rejected": -133.9650115966797, "loss": 0.7545, "margin_dpo/margin_mean": 17.790395736694336, "margin_dpo/margin_std": 15.214214324951172, "step": 190 }, { "epoch": 0.28046989720998533, "fcm_dpo/beta": 0.05509544536471367, "fcm_dpo/delta": -0.27554386854171753, "fcm_dpo/margin": 19.890722274780273, "fcm_dpo/q_t": 0.28646570444107056, "grad_norm": 18.667753219604492, "learning_rate": 4.5330506821893565e-07, "logits/chosen": -1.3162940740585327, "logits/rejected": -0.9771643280982971, "logps/chosen": -105.68276977539062, "logps/ref_chosen": -99.78551483154297, "logps/ref_rejected": -116.00028228759766, "logps/rejected": -141.78826904296875, "loss": 0.7436, "margin_dpo/margin_mean": 19.89072036743164, "margin_dpo/margin_std": 18.173255920410156, "step": 191 }, { "epoch": 0.28193832599118945, "fcm_dpo/beta": 0.05472166836261749, "fcm_dpo/delta": 0.08773260563611984, "fcm_dpo/margin": 14.054750442504883, "fcm_dpo/q_t": 0.3466401994228363, "grad_norm": 20.07705307006836, "learning_rate": 4.5255555107119336e-07, "logits/chosen": -1.1917040348052979, "logits/rejected": -0.9342058897018433, "logps/chosen": -113.84033203125, "logps/ref_chosen": -100.86032104492188, "logps/ref_rejected": -109.44500732421875, "logps/rejected": -136.47976684570312, "loss": 0.9351, "margin_dpo/margin_mean": 14.054752349853516, "margin_dpo/margin_std": 17.19097137451172, "step": 192 }, { "epoch": 0.2834067547723935, "fcm_dpo/beta": 0.05689571797847748, "fcm_dpo/delta": 0.17024913430213928, "fcm_dpo/margin": 12.118256568908691, "fcm_dpo/q_t": 0.3621511459350586, "grad_norm": 21.269683837890625, "learning_rate": 4.5180069639630236e-07, "logits/chosen": -1.5149171352386475, "logits/rejected": -0.9463554620742798, "logps/chosen": -102.46221160888672, "logps/ref_chosen": -92.93831634521484, "logps/ref_rejected": -83.10917663574219, "logps/rejected": -104.75133514404297, "loss": 1.0219, "margin_dpo/margin_mean": 12.118255615234375, "margin_dpo/margin_std": 17.301712036132812, "step": 193 }, { "epoch": 0.28487518355359764, "fcm_dpo/beta": 0.05779755860567093, "fcm_dpo/delta": 0.03512994945049286, "fcm_dpo/margin": 14.118998527526855, "fcm_dpo/q_t": 0.33766230940818787, "grad_norm": 17.252609252929688, "learning_rate": 4.510405240853854e-07, "logits/chosen": -2.011484146118164, "logits/rejected": -1.64884614944458, "logps/chosen": -73.9320297241211, "logps/ref_chosen": -63.79157257080078, "logps/ref_rejected": -68.72665405273438, "logps/rejected": -92.98609924316406, "loss": 0.9078, "margin_dpo/margin_mean": 14.118997573852539, "margin_dpo/margin_std": 16.229263305664062, "step": 194 }, { "epoch": 0.28634361233480177, "fcm_dpo/beta": 0.05644117295742035, "fcm_dpo/delta": -0.14289246499538422, "fcm_dpo/margin": 17.347488403320312, "fcm_dpo/q_t": 0.3069530725479126, "grad_norm": 17.122520446777344, "learning_rate": 4.5027505416968985e-07, "logits/chosen": -1.070321798324585, "logits/rejected": -0.9118151664733887, "logps/chosen": -93.67105102539062, "logps/ref_chosen": -83.88545989990234, "logps/ref_rejected": -104.00125122070312, "logps/rejected": -131.13433837890625, "loss": 0.8372, "margin_dpo/margin_mean": 17.347488403320312, "margin_dpo/margin_std": 18.290515899658203, "step": 195 }, { "epoch": 0.2878120411160059, "fcm_dpo/beta": 0.05479153245687485, "fcm_dpo/delta": -0.11973267793655396, "fcm_dpo/margin": 17.488330841064453, "fcm_dpo/q_t": 0.3044339418411255, "grad_norm": 14.206181526184082, "learning_rate": 4.495043068200599e-07, "logits/chosen": -0.9788313508033752, "logits/rejected": -0.5322003960609436, "logps/chosen": -100.54130554199219, "logps/ref_chosen": -89.4013671875, "logps/ref_rejected": -83.34735107421875, "logps/rejected": -111.97561645507812, "loss": 0.7886, "margin_dpo/margin_mean": 17.488330841064453, "margin_dpo/margin_std": 15.582159042358398, "step": 196 }, { "epoch": 0.28928046989721, "fcm_dpo/beta": 0.0564768947660923, "fcm_dpo/delta": 0.24834321439266205, "fcm_dpo/margin": 10.927431106567383, "fcm_dpo/q_t": 0.36820292472839355, "grad_norm": 18.218454360961914, "learning_rate": 4.4872830234640493e-07, "logits/chosen": -1.0139613151550293, "logits/rejected": -0.7055380344390869, "logps/chosen": -98.77421569824219, "logps/ref_chosen": -86.84070587158203, "logps/ref_rejected": -82.79289245605469, "logps/rejected": -105.65383911132812, "loss": 1.0011, "margin_dpo/margin_mean": 10.927431106567383, "margin_dpo/margin_std": 14.292451858520508, "step": 197 }, { "epoch": 0.2907488986784141, "fcm_dpo/beta": 0.0565267875790596, "fcm_dpo/delta": -0.09485112130641937, "fcm_dpo/margin": 16.559494018554688, "fcm_dpo/q_t": 0.3122091293334961, "grad_norm": 20.28647804260254, "learning_rate": 4.479470611971645e-07, "logits/chosen": -1.3765474557876587, "logits/rejected": -1.164352297782898, "logps/chosen": -99.16595458984375, "logps/ref_chosen": -85.04824829101562, "logps/ref_rejected": -104.10365295410156, "logps/rejected": -134.78085327148438, "loss": 0.8164, "margin_dpo/margin_mean": 16.55949592590332, "margin_dpo/margin_std": 16.41266441345215, "step": 198 }, { "epoch": 0.2922173274596182, "fcm_dpo/beta": 0.05642756074666977, "fcm_dpo/delta": 0.020511126145720482, "fcm_dpo/margin": 14.728890419006348, "fcm_dpo/q_t": 0.33377328515052795, "grad_norm": 18.18387222290039, "learning_rate": 4.471606039587695e-07, "logits/chosen": -1.5072638988494873, "logits/rejected": -1.2555978298187256, "logps/chosen": -95.25479125976562, "logps/ref_chosen": -81.78669738769531, "logps/ref_rejected": -90.790771484375, "logps/rejected": -118.98776245117188, "loss": 0.8888, "margin_dpo/margin_mean": 14.728889465332031, "margin_dpo/margin_std": 16.250869750976562, "step": 199 }, { "epoch": 0.2936857562408223, "fcm_dpo/beta": 0.055008381605148315, "fcm_dpo/delta": -0.11476686596870422, "fcm_dpo/margin": 17.302513122558594, "fcm_dpo/q_t": 0.3144181966781616, "grad_norm": 15.63186264038086, "learning_rate": 4.4636895135509966e-07, "logits/chosen": -1.6174651384353638, "logits/rejected": -1.1445786952972412, "logps/chosen": -96.79918670654297, "logps/ref_chosen": -84.26809692382812, "logps/ref_rejected": -86.96786499023438, "logps/rejected": -116.80148315429688, "loss": 0.8588, "margin_dpo/margin_mean": 17.302513122558594, "margin_dpo/margin_std": 18.646202087402344, "step": 200 }, { "epoch": 0.2936857562408223, "eval_fcm_dpo/beta": 0.05511855334043503, "eval_logits/chosen": -1.145448923110962, "eval_logits/rejected": -0.8443020582199097, "eval_logps/chosen": -115.77802276611328, "eval_logps/ref_chosen": -100.49356842041016, "eval_logps/ref_rejected": -94.06775665283203, "eval_logps/rejected": -121.74003601074219, "eval_loss": 0.5018326044082642, "eval_margin_dpo/margin_mean": 12.387815475463867, "eval_margin_dpo/margin_std": 17.043996810913086, "eval_runtime": 43.7855, "eval_samples_per_second": 53.42, "eval_steps_per_second": 1.69, "step": 200 }, { "epoch": 0.29515418502202645, "fcm_dpo/beta": 0.05490724742412567, "fcm_dpo/delta": -0.03479336202144623, "fcm_dpo/margin": 16.05862045288086, "fcm_dpo/q_t": 0.3281078338623047, "grad_norm": 19.102737426757812, "learning_rate": 4.455721242469372e-07, "logits/chosen": -1.08148193359375, "logits/rejected": -0.6773242950439453, "logps/chosen": -122.24678039550781, "logps/ref_chosen": -107.77249145507812, "logps/ref_rejected": -119.79248046875, "logps/rejected": -150.32537841796875, "loss": 0.9179, "margin_dpo/margin_mean": 16.05862045288086, "margin_dpo/margin_std": 19.691104888916016, "step": 201 }, { "epoch": 0.2966226138032305, "fcm_dpo/beta": 0.05517326667904854, "fcm_dpo/delta": 0.041220732033252716, "fcm_dpo/margin": 14.717092514038086, "fcm_dpo/q_t": 0.340792715549469, "grad_norm": 17.2861328125, "learning_rate": 4.4477014363141755e-07, "logits/chosen": -1.263124942779541, "logits/rejected": -1.1197762489318848, "logps/chosen": -89.5303955078125, "logps/ref_chosen": -75.97245788574219, "logps/ref_rejected": -94.4599838256836, "logps/rejected": -122.73501586914062, "loss": 0.9107, "margin_dpo/margin_mean": 14.717092514038086, "margin_dpo/margin_std": 17.112964630126953, "step": 202 }, { "epoch": 0.29809104258443464, "fcm_dpo/beta": 0.054431818425655365, "fcm_dpo/delta": -0.08932656049728394, "fcm_dpo/margin": 17.103050231933594, "fcm_dpo/q_t": 0.31066063046455383, "grad_norm": 15.397363662719727, "learning_rate": 4.439630306414758e-07, "logits/chosen": -1.5567055940628052, "logits/rejected": -1.1206634044647217, "logps/chosen": -106.02995300292969, "logps/ref_chosen": -94.96715545654297, "logps/ref_rejected": -92.8876724243164, "logps/rejected": -121.05351257324219, "loss": 0.8194, "margin_dpo/margin_mean": 17.103050231933594, "margin_dpo/margin_std": 16.513595581054688, "step": 203 }, { "epoch": 0.29955947136563876, "fcm_dpo/beta": 0.054940156638622284, "fcm_dpo/delta": 0.06543414294719696, "fcm_dpo/margin": 14.365680694580078, "fcm_dpo/q_t": 0.34287261962890625, "grad_norm": 21.70539665222168, "learning_rate": 4.431508065452897e-07, "logits/chosen": -1.2361195087432861, "logits/rejected": -0.5873404741287231, "logps/chosen": -131.6267852783203, "logps/ref_chosen": -116.35719299316406, "logps/ref_rejected": -93.39759063720703, "logps/rejected": -123.03286743164062, "loss": 0.94, "margin_dpo/margin_mean": 14.365680694580078, "margin_dpo/margin_std": 17.704381942749023, "step": 204 }, { "epoch": 0.3010279001468429, "fcm_dpo/beta": 0.05323153734207153, "fcm_dpo/delta": -0.18734723329544067, "fcm_dpo/margin": 19.112470626831055, "fcm_dpo/q_t": 0.28901487588882446, "grad_norm": 14.131471633911133, "learning_rate": 4.4233349274571974e-07, "logits/chosen": -1.951904535293579, "logits/rejected": -1.5425834655761719, "logps/chosen": -102.00930786132812, "logps/ref_chosen": -88.85934448242188, "logps/ref_rejected": -91.8544921875, "logps/rejected": -124.1169204711914, "loss": 0.7412, "margin_dpo/margin_mean": 19.112470626831055, "margin_dpo/margin_std": 15.50132942199707, "step": 205 }, { "epoch": 0.302496328928047, "fcm_dpo/beta": 0.052048660814762115, "fcm_dpo/delta": -0.06987091153860092, "fcm_dpo/margin": 17.530094146728516, "fcm_dpo/q_t": 0.30802714824676514, "grad_norm": 19.163528442382812, "learning_rate": 4.415111107797445e-07, "logits/chosen": -1.3110017776489258, "logits/rejected": -1.1106235980987549, "logps/chosen": -90.1287841796875, "logps/ref_chosen": -76.54634857177734, "logps/ref_rejected": -102.95314025878906, "logps/rejected": -134.065673828125, "loss": 0.7868, "margin_dpo/margin_mean": 17.530094146728516, "margin_dpo/margin_std": 14.55693531036377, "step": 206 }, { "epoch": 0.3039647577092511, "fcm_dpo/beta": 0.051451846957206726, "fcm_dpo/delta": -0.09747027605772018, "fcm_dpo/margin": 18.237651824951172, "fcm_dpo/q_t": 0.32565534114837646, "grad_norm": 22.39365005493164, "learning_rate": 4.4068368231789365e-07, "logits/chosen": -1.6113066673278809, "logits/rejected": -1.1651461124420166, "logps/chosen": -101.52843475341797, "logps/ref_chosen": -86.23164367675781, "logps/ref_rejected": -90.65512084960938, "logps/rejected": -124.18955993652344, "loss": 0.893, "margin_dpo/margin_mean": 18.237651824951172, "margin_dpo/margin_std": 22.46469497680664, "step": 207 }, { "epoch": 0.3054331864904552, "fcm_dpo/beta": 0.050518698990345, "fcm_dpo/delta": -0.14123016595840454, "fcm_dpo/margin": 19.343996047973633, "fcm_dpo/q_t": 0.2982601523399353, "grad_norm": 17.023468017578125, "learning_rate": 4.398512291636768e-07, "logits/chosen": -1.5571520328521729, "logits/rejected": -1.182025671005249, "logps/chosen": -109.33761596679688, "logps/ref_chosen": -94.1595458984375, "logps/ref_rejected": -100.96233367919922, "logps/rejected": -135.48440551757812, "loss": 0.7912, "margin_dpo/margin_mean": 19.343996047973633, "margin_dpo/margin_std": 17.425840377807617, "step": 208 }, { "epoch": 0.3069016152716593, "fcm_dpo/beta": 0.04946083575487137, "fcm_dpo/delta": 0.011037945747375488, "fcm_dpo/margin": 16.961767196655273, "fcm_dpo/q_t": 0.33165478706359863, "grad_norm": 15.817395210266113, "learning_rate": 4.3901377325300857e-07, "logits/chosen": -1.7526047229766846, "logits/rejected": -1.2873879671096802, "logps/chosen": -97.22732543945312, "logps/ref_chosen": -84.17056274414062, "logps/ref_rejected": -87.61955261230469, "logps/rejected": -117.63807678222656, "loss": 0.8862, "margin_dpo/margin_mean": 16.961767196655273, "margin_dpo/margin_std": 18.429622650146484, "step": 209 }, { "epoch": 0.30837004405286345, "fcm_dpo/beta": 0.049842916429042816, "fcm_dpo/delta": 0.037534188479185104, "fcm_dpo/margin": 16.353267669677734, "fcm_dpo/q_t": 0.33743494749069214, "grad_norm": 15.248353958129883, "learning_rate": 4.381713366536311e-07, "logits/chosen": -1.5737419128417969, "logits/rejected": -1.1732791662216187, "logps/chosen": -95.85877990722656, "logps/ref_chosen": -81.17117309570312, "logps/ref_rejected": -84.17478942871094, "logps/rejected": -115.21566009521484, "loss": 0.8877, "margin_dpo/margin_mean": 16.353267669677734, "margin_dpo/margin_std": 17.86672592163086, "step": 210 }, { "epoch": 0.30983847283406757, "fcm_dpo/beta": 0.05027975142002106, "fcm_dpo/delta": 0.0004709279164671898, "fcm_dpo/margin": 16.896448135375977, "fcm_dpo/q_t": 0.33016592264175415, "grad_norm": 16.873620986938477, "learning_rate": 4.373239415645323e-07, "logits/chosen": -1.735175371170044, "logits/rejected": -1.1475024223327637, "logps/chosen": -125.00082397460938, "logps/ref_chosen": -108.71270751953125, "logps/ref_rejected": -93.55564880371094, "logps/rejected": -126.74020385742188, "loss": 0.8806, "margin_dpo/margin_mean": 16.896448135375977, "margin_dpo/margin_std": 18.872371673583984, "step": 211 }, { "epoch": 0.31130690161527164, "fcm_dpo/beta": 0.04791909083724022, "fcm_dpo/delta": -0.2870158553123474, "fcm_dpo/margin": 23.030895233154297, "fcm_dpo/q_t": 0.28026482462882996, "grad_norm": 13.476890563964844, "learning_rate": 4.3647161031536086e-07, "logits/chosen": -1.4513945579528809, "logits/rejected": -1.0123846530914307, "logps/chosen": -108.43475341796875, "logps/ref_chosen": -98.36195373535156, "logps/ref_rejected": -109.88999938964844, "logps/rejected": -142.99371337890625, "loss": 0.7281, "margin_dpo/margin_mean": 23.030893325805664, "margin_dpo/margin_std": 19.890701293945312, "step": 212 }, { "epoch": 0.31277533039647576, "fcm_dpo/beta": 0.04703017696738243, "fcm_dpo/delta": -0.029142513871192932, "fcm_dpo/margin": 18.629619598388672, "fcm_dpo/q_t": 0.3192410171031952, "grad_norm": 20.05600357055664, "learning_rate": 4.3561436536583774e-07, "logits/chosen": -1.5639327764511108, "logits/rejected": -0.8274385929107666, "logps/chosen": -121.58412170410156, "logps/ref_chosen": -108.05531311035156, "logps/ref_rejected": -100.14414978027344, "logps/rejected": -132.30258178710938, "loss": 0.8464, "margin_dpo/margin_mean": 18.629619598388672, "margin_dpo/margin_std": 18.510005950927734, "step": 213 }, { "epoch": 0.3142437591776799, "fcm_dpo/beta": 0.047874145209789276, "fcm_dpo/delta": 0.08061401546001434, "fcm_dpo/margin": 16.192214965820312, "fcm_dpo/q_t": 0.3512595295906067, "grad_norm": 16.127655029296875, "learning_rate": 4.3475222930516473e-07, "logits/chosen": -1.3913880586624146, "logits/rejected": -1.266923189163208, "logps/chosen": -92.62548828125, "logps/ref_chosen": -77.80473327636719, "logps/ref_rejected": -84.27578735351562, "logps/rejected": -115.28874206542969, "loss": 0.9673, "margin_dpo/margin_mean": 16.192214965820312, "margin_dpo/margin_std": 21.93199348449707, "step": 214 }, { "epoch": 0.315712187958884, "fcm_dpo/beta": 0.046690285205841064, "fcm_dpo/delta": -0.16893821954727173, "fcm_dpo/margin": 21.456758499145508, "fcm_dpo/q_t": 0.2978704571723938, "grad_norm": 14.699638366699219, "learning_rate": 4.3388522485142885e-07, "logits/chosen": -1.5712075233459473, "logits/rejected": -1.0760598182678223, "logps/chosen": -97.10356140136719, "logps/ref_chosen": -85.1138916015625, "logps/ref_rejected": -96.86151123046875, "logps/rejected": -130.30795288085938, "loss": 0.7625, "margin_dpo/margin_mean": 21.456756591796875, "margin_dpo/margin_std": 19.08936309814453, "step": 215 }, { "epoch": 0.31718061674008813, "fcm_dpo/beta": 0.0466710701584816, "fcm_dpo/delta": 0.01144382357597351, "fcm_dpo/margin": 17.97332191467285, "fcm_dpo/q_t": 0.3272448778152466, "grad_norm": 13.613797187805176, "learning_rate": 4.330133748510036e-07, "logits/chosen": -1.5654511451721191, "logits/rejected": -1.1545265913009644, "logps/chosen": -94.93984985351562, "logps/ref_chosen": -80.59230041503906, "logps/ref_rejected": -81.41983795166016, "logps/rejected": -113.74070739746094, "loss": 0.8434, "margin_dpo/margin_mean": 17.97332000732422, "margin_dpo/margin_std": 17.3824462890625, "step": 216 }, { "epoch": 0.3186490455212922, "fcm_dpo/beta": 0.04719318449497223, "fcm_dpo/delta": 0.07027594745159149, "fcm_dpo/margin": 16.616840362548828, "fcm_dpo/q_t": 0.34417039155960083, "grad_norm": 17.98988151550293, "learning_rate": 4.3213670227794757e-07, "logits/chosen": -1.81255042552948, "logits/rejected": -1.5056719779968262, "logps/chosen": -110.1501693725586, "logps/ref_chosen": -93.47257995605469, "logps/ref_rejected": -103.488525390625, "logps/rejected": -136.782958984375, "loss": 0.9225, "margin_dpo/margin_mean": 16.616840362548828, "margin_dpo/margin_std": 19.80947494506836, "step": 217 }, { "epoch": 0.3201174743024963, "fcm_dpo/beta": 0.04671292006969452, "fcm_dpo/delta": -0.0367327444255352, "fcm_dpo/margin": 18.908458709716797, "fcm_dpo/q_t": 0.3229714035987854, "grad_norm": 16.900039672851562, "learning_rate": 4.3125523023339815e-07, "logits/chosen": -1.0308185815811157, "logits/rejected": -0.6163084506988525, "logps/chosen": -104.5977783203125, "logps/ref_chosen": -89.05883026123047, "logps/ref_rejected": -94.30680847167969, "logps/rejected": -128.7542266845703, "loss": 0.8795, "margin_dpo/margin_mean": 18.908458709716797, "margin_dpo/margin_std": 20.99794578552246, "step": 218 }, { "epoch": 0.32158590308370044, "fcm_dpo/beta": 0.04732338711619377, "fcm_dpo/delta": 0.014991968870162964, "fcm_dpo/margin": 17.627521514892578, "fcm_dpo/q_t": 0.33559125661849976, "grad_norm": 22.87677764892578, "learning_rate": 4.303689819449636e-07, "logits/chosen": -1.599898099899292, "logits/rejected": -1.204871654510498, "logps/chosen": -115.39610290527344, "logps/ref_chosen": -101.00733947753906, "logps/ref_rejected": -92.46794128417969, "logps/rejected": -124.48422241210938, "loss": 0.9291, "margin_dpo/margin_mean": 17.627521514892578, "margin_dpo/margin_std": 21.77505874633789, "step": 219 }, { "epoch": 0.32305433186490456, "fcm_dpo/beta": 0.04759221524000168, "fcm_dpo/delta": 0.10844795405864716, "fcm_dpo/margin": 15.754082679748535, "fcm_dpo/q_t": 0.3434942364692688, "grad_norm": 16.797428131103516, "learning_rate": 4.2947798076611047e-07, "logits/chosen": -1.211072325706482, "logits/rejected": -0.8570730090141296, "logps/chosen": -110.34193420410156, "logps/ref_chosen": -95.53721618652344, "logps/ref_rejected": -94.30703735351562, "logps/rejected": -124.86582946777344, "loss": 0.901, "margin_dpo/margin_mean": 15.754082679748535, "margin_dpo/margin_std": 16.457195281982422, "step": 220 }, { "epoch": 0.3245227606461087, "fcm_dpo/beta": 0.046583935618400574, "fcm_dpo/delta": -0.26324182748794556, "fcm_dpo/margin": 23.2933349609375, "fcm_dpo/q_t": 0.2837619185447693, "grad_norm": 15.483804702758789, "learning_rate": 4.285822501755485e-07, "logits/chosen": -1.017369270324707, "logits/rejected": -0.8445158004760742, "logps/chosen": -97.19378662109375, "logps/ref_chosen": -82.84486389160156, "logps/ref_rejected": -110.81179809570312, "logps/rejected": -148.45407104492188, "loss": 0.7347, "margin_dpo/margin_mean": 23.2933349609375, "margin_dpo/margin_std": 19.25533676147461, "step": 221 }, { "epoch": 0.32599118942731276, "fcm_dpo/beta": 0.045510318130254745, "fcm_dpo/delta": -0.0532805472612381, "fcm_dpo/margin": 19.718502044677734, "fcm_dpo/q_t": 0.32435593008995056, "grad_norm": 23.144872665405273, "learning_rate": 4.276818137766118e-07, "logits/chosen": -1.720402479171753, "logits/rejected": -1.287516474723816, "logps/chosen": -113.3494873046875, "logps/ref_chosen": -95.14198303222656, "logps/ref_rejected": -106.80441284179688, "logps/rejected": -144.73040771484375, "loss": 0.9196, "margin_dpo/margin_mean": 19.7185001373291, "margin_dpo/margin_std": 24.308671951293945, "step": 222 }, { "epoch": 0.3274596182085169, "fcm_dpo/beta": 0.045756347477436066, "fcm_dpo/delta": 0.13066641986370087, "fcm_dpo/margin": 15.932899475097656, "fcm_dpo/q_t": 0.3532084822654724, "grad_norm": 15.819351196289062, "learning_rate": 4.2677669529663686e-07, "logits/chosen": -2.0023083686828613, "logits/rejected": -1.5972869396209717, "logps/chosen": -102.81034851074219, "logps/ref_chosen": -85.57511138916016, "logps/ref_rejected": -86.45237731933594, "logps/rejected": -119.62051391601562, "loss": 0.9563, "margin_dpo/margin_mean": 15.932899475097656, "margin_dpo/margin_std": 19.981346130371094, "step": 223 }, { "epoch": 0.328928046989721, "fcm_dpo/beta": 0.04605341702699661, "fcm_dpo/delta": -0.03015466220676899, "fcm_dpo/margin": 19.053667068481445, "fcm_dpo/q_t": 0.3233034014701843, "grad_norm": 16.37089729309082, "learning_rate": 4.2586691858633747e-07, "logits/chosen": -1.5879364013671875, "logits/rejected": -1.1492383480072021, "logps/chosen": -99.3148193359375, "logps/ref_chosen": -82.72380065917969, "logps/ref_rejected": -82.59538269042969, "logps/rejected": -118.24006652832031, "loss": 0.8544, "margin_dpo/margin_mean": 19.053668975830078, "margin_dpo/margin_std": 20.131755828857422, "step": 224 }, { "epoch": 0.3303964757709251, "fcm_dpo/beta": 0.045157790184020996, "fcm_dpo/delta": -0.1496858149766922, "fcm_dpo/margin": 21.81402587890625, "fcm_dpo/q_t": 0.3073573112487793, "grad_norm": 16.657323837280273, "learning_rate": 4.249525076191759e-07, "logits/chosen": -1.70444917678833, "logits/rejected": -1.226616382598877, "logps/chosen": -114.02447509765625, "logps/ref_chosen": -95.67768859863281, "logps/ref_rejected": -105.09687805175781, "logps/rejected": -145.2576904296875, "loss": 0.8305, "margin_dpo/margin_mean": 21.81402587890625, "margin_dpo/margin_std": 23.370323181152344, "step": 225 }, { "epoch": 0.33186490455212925, "fcm_dpo/beta": 0.04409945756196976, "fcm_dpo/delta": -0.017373599112033844, "fcm_dpo/margin": 19.608936309814453, "fcm_dpo/q_t": 0.32043924927711487, "grad_norm": 14.112857818603516, "learning_rate": 4.2403348649073167e-07, "logits/chosen": -1.6110211610794067, "logits/rejected": -1.0937122106552124, "logps/chosen": -108.33169555664062, "logps/ref_chosen": -93.46092224121094, "logps/ref_rejected": -86.7017593383789, "logps/rejected": -121.18147277832031, "loss": 0.8342, "margin_dpo/margin_mean": 19.608936309814453, "margin_dpo/margin_std": 18.334829330444336, "step": 226 }, { "epoch": 0.3333333333333333, "fcm_dpo/beta": 0.044242776930332184, "fcm_dpo/delta": -0.023924846202135086, "fcm_dpo/margin": 19.70592498779297, "fcm_dpo/q_t": 0.3298208713531494, "grad_norm": 16.250980377197266, "learning_rate": 4.2310987941806615e-07, "logits/chosen": -1.4222127199172974, "logits/rejected": -1.094136357307434, "logps/chosen": -114.05935668945312, "logps/ref_chosen": -92.81427001953125, "logps/ref_rejected": -104.73693084716797, "logps/rejected": -145.68792724609375, "loss": 0.8972, "margin_dpo/margin_mean": 19.7059268951416, "margin_dpo/margin_std": 22.74959945678711, "step": 227 }, { "epoch": 0.33480176211453744, "fcm_dpo/beta": 0.04522550851106644, "fcm_dpo/delta": 0.13866844773292542, "fcm_dpo/margin": 15.936595916748047, "fcm_dpo/q_t": 0.34774237871170044, "grad_norm": 16.323331832885742, "learning_rate": 4.2218171073908463e-07, "logits/chosen": -1.6550034284591675, "logits/rejected": -1.2243372201919556, "logps/chosen": -113.41688537597656, "logps/ref_chosen": -94.03712463378906, "logps/ref_rejected": -96.02151489257812, "logps/rejected": -131.33787536621094, "loss": 0.9329, "margin_dpo/margin_mean": 15.936595916748047, "margin_dpo/margin_std": 17.845264434814453, "step": 228 }, { "epoch": 0.33627019089574156, "fcm_dpo/beta": 0.043626703321933746, "fcm_dpo/delta": -0.2818097770214081, "fcm_dpo/margin": 25.238357543945312, "fcm_dpo/q_t": 0.28765520453453064, "grad_norm": 13.921157836914062, "learning_rate": 4.212490049118951e-07, "logits/chosen": -1.539412021636963, "logits/rejected": -0.9021925926208496, "logps/chosen": -111.91642761230469, "logps/ref_chosen": -95.57766723632812, "logps/ref_rejected": -89.17379760742188, "logps/rejected": -130.75091552734375, "loss": 0.7562, "margin_dpo/margin_mean": 25.238357543945312, "margin_dpo/margin_std": 23.75576400756836, "step": 229 }, { "epoch": 0.3377386196769457, "fcm_dpo/beta": 0.04161980003118515, "fcm_dpo/delta": -0.20371034741401672, "fcm_dpo/margin": 24.80811309814453, "fcm_dpo/q_t": 0.28747209906578064, "grad_norm": 15.197046279907227, "learning_rate": 4.203117865141635e-07, "logits/chosen": -1.5952270030975342, "logits/rejected": -1.4780795574188232, "logps/chosen": -81.2739028930664, "logps/ref_chosen": -63.713626861572266, "logps/ref_rejected": -91.90870666503906, "logps/rejected": -134.277099609375, "loss": 0.7406, "margin_dpo/margin_mean": 24.80811309814453, "margin_dpo/margin_std": 20.30365753173828, "step": 230 }, { "epoch": 0.3392070484581498, "fcm_dpo/beta": 0.04121699184179306, "fcm_dpo/delta": 0.03683965280652046, "fcm_dpo/margin": 19.79311752319336, "fcm_dpo/q_t": 0.3291405439376831, "grad_norm": 13.925002098083496, "learning_rate": 4.1937008024246625e-07, "logits/chosen": -1.4601212739944458, "logits/rejected": -0.9463979005813599, "logps/chosen": -116.09357452392578, "logps/ref_chosen": -95.45567321777344, "logps/ref_rejected": -80.95568084716797, "logps/rejected": -121.3866958618164, "loss": 0.846, "margin_dpo/margin_mean": 19.79311752319336, "margin_dpo/margin_std": 18.4172306060791, "step": 231 }, { "epoch": 0.3406754772393539, "fcm_dpo/beta": 0.04198054224252701, "fcm_dpo/delta": 0.0867529809474945, "fcm_dpo/margin": 18.341995239257812, "fcm_dpo/q_t": 0.3432447910308838, "grad_norm": 15.27546215057373, "learning_rate": 4.1842391091163933e-07, "logits/chosen": -1.2151339054107666, "logits/rejected": -0.749545693397522, "logps/chosen": -116.75117492675781, "logps/ref_chosen": -96.89726257324219, "logps/ref_rejected": -89.76461791992188, "logps/rejected": -127.96051788330078, "loss": 0.9195, "margin_dpo/margin_mean": 18.341995239257812, "margin_dpo/margin_std": 20.981639862060547, "step": 232 }, { "epoch": 0.342143906020558, "fcm_dpo/beta": 0.040949709713459015, "fcm_dpo/delta": -0.18371593952178955, "fcm_dpo/margin": 24.765342712402344, "fcm_dpo/q_t": 0.29807668924331665, "grad_norm": 14.443903923034668, "learning_rate": 4.174733034541245e-07, "logits/chosen": -1.4963655471801758, "logits/rejected": -1.233656644821167, "logps/chosen": -110.30583190917969, "logps/ref_chosen": -89.05032348632812, "logps/ref_rejected": -112.75917053222656, "logps/rejected": -158.77999877929688, "loss": 0.7868, "margin_dpo/margin_mean": 24.765342712402344, "margin_dpo/margin_std": 23.211368560791016, "step": 233 }, { "epoch": 0.3436123348017621, "fcm_dpo/beta": 0.0389174222946167, "fcm_dpo/delta": -0.19708919525146484, "fcm_dpo/margin": 26.18871307373047, "fcm_dpo/q_t": 0.30054956674575806, "grad_norm": 13.936091423034668, "learning_rate": 4.165182829193126e-07, "logits/chosen": -1.3125296831130981, "logits/rejected": -1.2004642486572266, "logps/chosen": -94.45536804199219, "logps/ref_chosen": -74.318359375, "logps/ref_rejected": -106.38758850097656, "logps/rejected": -152.7133026123047, "loss": 0.8011, "margin_dpo/margin_mean": 26.18871307373047, "margin_dpo/margin_std": 25.016361236572266, "step": 234 }, { "epoch": 0.34508076358296624, "fcm_dpo/beta": 0.03963829576969147, "fcm_dpo/delta": 0.07630567252635956, "fcm_dpo/margin": 19.665775299072266, "fcm_dpo/q_t": 0.3366130590438843, "grad_norm": 13.919268608093262, "learning_rate": 4.1555887447288255e-07, "logits/chosen": -1.2706738710403442, "logits/rejected": -0.7282851934432983, "logps/chosen": -122.56187438964844, "logps/ref_chosen": -98.217041015625, "logps/ref_rejected": -97.24677276611328, "logps/rejected": -141.25738525390625, "loss": 0.8743, "margin_dpo/margin_mean": 19.665773391723633, "margin_dpo/margin_std": 19.334230422973633, "step": 235 }, { "epoch": 0.3465491923641703, "fcm_dpo/beta": 0.03917301446199417, "fcm_dpo/delta": -0.12850745022296906, "fcm_dpo/margin": 24.66576385498047, "fcm_dpo/q_t": 0.3191804885864258, "grad_norm": 14.044709205627441, "learning_rate": 4.1459510339613946e-07, "logits/chosen": -1.3997790813446045, "logits/rejected": -1.3525149822235107, "logps/chosen": -98.97593688964844, "logps/ref_chosen": -78.83773040771484, "logps/ref_rejected": -109.06343078613281, "logps/rejected": -153.86740112304688, "loss": 0.8368, "margin_dpo/margin_mean": 24.66576385498047, "margin_dpo/margin_std": 28.19268035888672, "step": 236 }, { "epoch": 0.34801762114537443, "fcm_dpo/beta": 0.0390484444797039, "fcm_dpo/delta": 0.07136844098567963, "fcm_dpo/margin": 20.084590911865234, "fcm_dpo/q_t": 0.35290807485580444, "grad_norm": 15.641119003295898, "learning_rate": 4.136269950853473e-07, "logits/chosen": -1.503021240234375, "logits/rejected": -1.09787118434906, "logps/chosen": -108.82412719726562, "logps/ref_chosen": -85.21128845214844, "logps/ref_rejected": -99.90999603271484, "logps/rejected": -143.60743713378906, "loss": 0.9832, "margin_dpo/margin_mean": 20.084590911865234, "margin_dpo/margin_std": 28.457015991210938, "step": 237 }, { "epoch": 0.34948604992657856, "fcm_dpo/beta": 0.03921890631318092, "fcm_dpo/delta": -0.0024761436507105827, "fcm_dpo/margin": 21.72978401184082, "fcm_dpo/q_t": 0.33124884963035583, "grad_norm": 13.521815299987793, "learning_rate": 4.126545750510605e-07, "logits/chosen": -1.3132723569869995, "logits/rejected": -1.1299904584884644, "logps/chosen": -104.59783935546875, "logps/ref_chosen": -78.73123168945312, "logps/ref_rejected": -95.41840362548828, "logps/rejected": -143.01480102539062, "loss": 0.8794, "margin_dpo/margin_mean": 21.729782104492188, "margin_dpo/margin_std": 24.03033447265625, "step": 238 }, { "epoch": 0.3509544787077827, "fcm_dpo/beta": 0.03920455649495125, "fcm_dpo/delta": 0.04809597134590149, "fcm_dpo/margin": 20.524173736572266, "fcm_dpo/q_t": 0.33716443181037903, "grad_norm": 14.763937950134277, "learning_rate": 4.116778689174514e-07, "logits/chosen": -1.7374763488769531, "logits/rejected": -1.263242244720459, "logps/chosen": -116.79473876953125, "logps/ref_chosen": -92.60093688964844, "logps/ref_rejected": -100.51769256591797, "logps/rejected": -145.23565673828125, "loss": 0.9014, "margin_dpo/margin_mean": 20.524173736572266, "margin_dpo/margin_std": 22.263370513916016, "step": 239 }, { "epoch": 0.3524229074889868, "fcm_dpo/beta": 0.040073223412036896, "fcm_dpo/delta": 0.10381583124399185, "fcm_dpo/margin": 18.81613540649414, "fcm_dpo/q_t": 0.34458890557289124, "grad_norm": 18.043434143066406, "learning_rate": 4.106969024216348e-07, "logits/chosen": -1.850398063659668, "logits/rejected": -1.5278594493865967, "logps/chosen": -110.80873107910156, "logps/ref_chosen": -86.15977478027344, "logps/ref_rejected": -80.45567321777344, "logps/rejected": -123.92076110839844, "loss": 0.9218, "margin_dpo/margin_mean": 18.81613540649414, "margin_dpo/margin_std": 21.0098876953125, "step": 240 }, { "epoch": 0.35389133627019087, "fcm_dpo/beta": 0.03847186267375946, "fcm_dpo/delta": -0.420759916305542, "fcm_dpo/margin": 31.7427978515625, "fcm_dpo/q_t": 0.2649722695350647, "grad_norm": 13.3577299118042, "learning_rate": 4.097117014129903e-07, "logits/chosen": -1.5077540874481201, "logits/rejected": -0.8860681056976318, "logps/chosen": -121.54011535644531, "logps/ref_chosen": -101.04594421386719, "logps/ref_rejected": -94.04934692382812, "logps/rejected": -146.28631591796875, "loss": 0.6882, "margin_dpo/margin_mean": 31.7427978515625, "margin_dpo/margin_std": 26.658802032470703, "step": 241 }, { "epoch": 0.355359765051395, "fcm_dpo/beta": 0.03743920102715492, "fcm_dpo/delta": -0.012355677783489227, "fcm_dpo/margin": 22.97677993774414, "fcm_dpo/q_t": 0.32101404666900635, "grad_norm": 14.553584098815918, "learning_rate": 4.087222918524807e-07, "logits/chosen": -1.6811339855194092, "logits/rejected": -1.0449347496032715, "logps/chosen": -121.91845703125, "logps/ref_chosen": -95.67266845703125, "logps/ref_rejected": -90.65454864501953, "logps/rejected": -139.8771209716797, "loss": 0.8409, "margin_dpo/margin_mean": 22.97677993774414, "margin_dpo/margin_std": 21.574127197265625, "step": 242 }, { "epoch": 0.3568281938325991, "fcm_dpo/beta": 0.03756431117653847, "fcm_dpo/delta": -0.008666589856147766, "fcm_dpo/margin": 22.756086349487305, "fcm_dpo/q_t": 0.32629698514938354, "grad_norm": 13.971424102783203, "learning_rate": 4.07728699811968e-07, "logits/chosen": -1.6577210426330566, "logits/rejected": -0.8642560243606567, "logps/chosen": -123.25171661376953, "logps/ref_chosen": -98.03140258789062, "logps/ref_rejected": -83.18806457519531, "logps/rejected": -131.16445922851562, "loss": 0.8891, "margin_dpo/margin_mean": 22.756086349487305, "margin_dpo/margin_std": 24.21973419189453, "step": 243 }, { "epoch": 0.35829662261380324, "fcm_dpo/beta": 0.0368984192609787, "fcm_dpo/delta": -0.012304404750466347, "fcm_dpo/margin": 23.336746215820312, "fcm_dpo/q_t": 0.32805073261260986, "grad_norm": 13.78069019317627, "learning_rate": 4.067309514735267e-07, "logits/chosen": -1.9630992412567139, "logits/rejected": -1.6161465644836426, "logps/chosen": -115.62930297851562, "logps/ref_chosen": -88.8939208984375, "logps/ref_rejected": -102.57278442382812, "logps/rejected": -152.64491271972656, "loss": 0.8663, "margin_dpo/margin_mean": 23.336746215820312, "margin_dpo/margin_std": 24.9739933013916, "step": 244 }, { "epoch": 0.35976505139500736, "fcm_dpo/beta": 0.03636704385280609, "fcm_dpo/delta": -0.11052923649549484, "fcm_dpo/margin": 26.122928619384766, "fcm_dpo/q_t": 0.3047110438346863, "grad_norm": 13.41275691986084, "learning_rate": 4.057290731287531e-07, "logits/chosen": -2.0480587482452393, "logits/rejected": -1.2761878967285156, "logps/chosen": -126.63717651367188, "logps/ref_chosen": -104.19400024414062, "logps/ref_rejected": -92.65645599365234, "logps/rejected": -141.22256469726562, "loss": 0.7874, "margin_dpo/margin_mean": 26.122928619384766, "margin_dpo/margin_std": 22.928936004638672, "step": 245 }, { "epoch": 0.36123348017621143, "fcm_dpo/beta": 0.03667043149471283, "fcm_dpo/delta": 0.12247386574745178, "fcm_dpo/margin": 20.089324951171875, "fcm_dpo/q_t": 0.3481150269508362, "grad_norm": 15.194877624511719, "learning_rate": 4.047230911780736e-07, "logits/chosen": -1.530800461769104, "logits/rejected": -1.014233946800232, "logps/chosen": -132.90155029296875, "logps/ref_chosen": -103.21904754638672, "logps/ref_rejected": -90.99221801757812, "logps/rejected": -140.7640380859375, "loss": 0.9273, "margin_dpo/margin_mean": 20.089324951171875, "margin_dpo/margin_std": 22.886226654052734, "step": 246 }, { "epoch": 0.36270190895741555, "fcm_dpo/beta": 0.03562181442975998, "fcm_dpo/delta": -0.22008347511291504, "fcm_dpo/margin": 29.32817268371582, "fcm_dpo/q_t": 0.2943752408027649, "grad_norm": 12.687663078308105, "learning_rate": 4.0371303213004814e-07, "logits/chosen": -2.017427921295166, "logits/rejected": -1.7663198709487915, "logps/chosen": -113.63011169433594, "logps/ref_chosen": -86.99437713623047, "logps/ref_rejected": -111.33802795410156, "logps/rejected": -167.30191040039062, "loss": 0.7633, "margin_dpo/margin_mean": 29.32817268371582, "margin_dpo/margin_std": 27.21298599243164, "step": 247 }, { "epoch": 0.3641703377386197, "fcm_dpo/beta": 0.03615882247686386, "fcm_dpo/delta": 0.11238771677017212, "fcm_dpo/margin": 20.618698120117188, "fcm_dpo/q_t": 0.3473028540611267, "grad_norm": 15.50047492980957, "learning_rate": 4.0269892260067197e-07, "logits/chosen": -1.745754599571228, "logits/rejected": -1.6654396057128906, "logps/chosen": -105.63675689697266, "logps/ref_chosen": -74.7855224609375, "logps/ref_rejected": -98.27689361572266, "logps/rejected": -149.746826171875, "loss": 0.9163, "margin_dpo/margin_mean": 20.618698120117188, "margin_dpo/margin_std": 23.03470230102539, "step": 248 }, { "epoch": 0.3656387665198238, "fcm_dpo/beta": 0.03689437732100487, "fcm_dpo/delta": 0.13981816172599792, "fcm_dpo/margin": 19.527952194213867, "fcm_dpo/q_t": 0.3553723394870758, "grad_norm": 17.28803825378418, "learning_rate": 4.0168078931267426e-07, "logits/chosen": -2.111118793487549, "logits/rejected": -1.628953456878662, "logps/chosen": -128.77549743652344, "logps/ref_chosen": -95.70379638671875, "logps/ref_rejected": -87.14646911621094, "logps/rejected": -139.74612426757812, "loss": 0.9751, "margin_dpo/margin_mean": 19.5279541015625, "margin_dpo/margin_std": 25.350357055664062, "step": 249 }, { "epoch": 0.3671071953010279, "fcm_dpo/beta": 0.03755343705415726, "fcm_dpo/delta": 0.10731877386569977, "fcm_dpo/margin": 19.97228240966797, "fcm_dpo/q_t": 0.34544017910957336, "grad_norm": 20.92696189880371, "learning_rate": 4.006586590948141e-07, "logits/chosen": -1.7340309619903564, "logits/rejected": -0.8914738893508911, "logps/chosen": -144.86044311523438, "logps/ref_chosen": -114.05221557617188, "logps/ref_rejected": -81.08767700195312, "logps/rejected": -131.86819458007812, "loss": 0.9467, "margin_dpo/margin_mean": 19.97228240966797, "margin_dpo/margin_std": 23.23900032043457, "step": 250 }, { "epoch": 0.368575624082232, "fcm_dpo/beta": 0.0382770374417305, "fcm_dpo/delta": 0.0928206592798233, "fcm_dpo/margin": 19.94355583190918, "fcm_dpo/q_t": 0.3502398431301117, "grad_norm": 21.311126708984375, "learning_rate": 3.9963255888117325e-07, "logits/chosen": -1.6832396984100342, "logits/rejected": -1.006043553352356, "logps/chosen": -127.50480651855469, "logps/ref_chosen": -97.71128845214844, "logps/ref_rejected": -83.52742004394531, "logps/rejected": -133.26449584960938, "loss": 0.9595, "margin_dpo/margin_mean": 19.94355583190918, "margin_dpo/margin_std": 25.059249877929688, "step": 251 }, { "epoch": 0.3700440528634361, "fcm_dpo/beta": 0.03917098790407181, "fcm_dpo/delta": 0.06389188766479492, "fcm_dpo/margin": 20.19826889038086, "fcm_dpo/q_t": 0.33797770738601685, "grad_norm": 16.346147537231445, "learning_rate": 3.9860251571044666e-07, "logits/chosen": -1.793283462524414, "logits/rejected": -1.3007917404174805, "logps/chosen": -138.48342895507812, "logps/ref_chosen": -108.9861068725586, "logps/ref_rejected": -91.56423950195312, "logps/rejected": -141.25982666015625, "loss": 0.9099, "margin_dpo/margin_mean": 20.19826889038086, "margin_dpo/margin_std": 22.625965118408203, "step": 252 }, { "epoch": 0.37151248164464024, "fcm_dpo/beta": 0.04027726128697395, "fcm_dpo/delta": 0.06578607112169266, "fcm_dpo/margin": 19.479793548583984, "fcm_dpo/q_t": 0.3372243344783783, "grad_norm": 16.55797004699707, "learning_rate": 3.9756855672522986e-07, "logits/chosen": -1.936989426612854, "logits/rejected": -1.7282047271728516, "logps/chosen": -132.63905334472656, "logps/ref_chosen": -100.21630096435547, "logps/ref_rejected": -105.67670440673828, "logps/rejected": -157.57925415039062, "loss": 0.9225, "margin_dpo/margin_mean": 19.479793548583984, "margin_dpo/margin_std": 21.367464065551758, "step": 253 }, { "epoch": 0.37298091042584436, "fcm_dpo/beta": 0.04016454517841339, "fcm_dpo/delta": 0.1159537136554718, "fcm_dpo/margin": 18.474327087402344, "fcm_dpo/q_t": 0.34890618920326233, "grad_norm": 16.3812255859375, "learning_rate": 3.965307091713037e-07, "logits/chosen": -1.952610731124878, "logits/rejected": -1.4213777780532837, "logps/chosen": -130.75515747070312, "logps/ref_chosen": -98.73518371582031, "logps/ref_rejected": -93.73825073242188, "logps/rejected": -144.23255920410156, "loss": 0.9529, "margin_dpo/margin_mean": 18.474328994750977, "margin_dpo/margin_std": 22.191251754760742, "step": 254 }, { "epoch": 0.3744493392070485, "fcm_dpo/beta": 0.03990204632282257, "fcm_dpo/delta": -0.15113875269889832, "fcm_dpo/margin": 24.718217849731445, "fcm_dpo/q_t": 0.30290794372558594, "grad_norm": 16.08772087097168, "learning_rate": 3.954890003969163e-07, "logits/chosen": -1.6271870136260986, "logits/rejected": -1.2749335765838623, "logps/chosen": -124.388427734375, "logps/ref_chosen": -90.382568359375, "logps/ref_rejected": -97.07624816894531, "logps/rejected": -155.80032348632812, "loss": 0.8062, "margin_dpo/margin_mean": 24.718219757080078, "margin_dpo/margin_std": 23.882701873779297, "step": 255 }, { "epoch": 0.37591776798825255, "fcm_dpo/beta": 0.03883664309978485, "fcm_dpo/delta": -0.16052256524562836, "fcm_dpo/margin": 25.614931106567383, "fcm_dpo/q_t": 0.3072439134120941, "grad_norm": 16.105266571044922, "learning_rate": 3.944434578520628e-07, "logits/chosen": -1.953477144241333, "logits/rejected": -1.5968413352966309, "logps/chosen": -119.0958023071289, "logps/ref_chosen": -88.7528076171875, "logps/ref_rejected": -98.49383544921875, "logps/rejected": -154.45175170898438, "loss": 0.8515, "margin_dpo/margin_mean": 25.614931106567383, "margin_dpo/margin_std": 27.952342987060547, "step": 256 }, { "epoch": 0.37738619676945667, "fcm_dpo/beta": 0.03744344413280487, "fcm_dpo/delta": -0.061067692935466766, "fcm_dpo/margin": 24.065053939819336, "fcm_dpo/q_t": 0.32052546739578247, "grad_norm": 17.075380325317383, "learning_rate": 3.933941090877615e-07, "logits/chosen": -1.7229259014129639, "logits/rejected": -1.3566169738769531, "logps/chosen": -115.55632019042969, "logps/ref_chosen": -82.80352020263672, "logps/ref_rejected": -85.8677978515625, "logps/rejected": -142.68563842773438, "loss": 0.864, "margin_dpo/margin_mean": 24.065053939819336, "margin_dpo/margin_std": 25.16031265258789, "step": 257 }, { "epoch": 0.3788546255506608, "fcm_dpo/beta": 0.03788026422262192, "fcm_dpo/delta": -0.03581918030977249, "fcm_dpo/margin": 23.28742027282715, "fcm_dpo/q_t": 0.31651657819747925, "grad_norm": 15.347949981689453, "learning_rate": 3.923409817553284e-07, "logits/chosen": -1.954576015472412, "logits/rejected": -1.6011340618133545, "logps/chosen": -122.21308135986328, "logps/ref_chosen": -90.187744140625, "logps/ref_rejected": -103.47068786621094, "logps/rejected": -158.783447265625, "loss": 0.8404, "margin_dpo/margin_mean": 23.28742218017578, "margin_dpo/margin_std": 22.27655601501465, "step": 258 }, { "epoch": 0.3803230543318649, "fcm_dpo/beta": 0.03793717548251152, "fcm_dpo/delta": 0.11459681391716003, "fcm_dpo/margin": 19.601879119873047, "fcm_dpo/q_t": 0.3489711880683899, "grad_norm": 16.28948974609375, "learning_rate": 3.9128410360564793e-07, "logits/chosen": -1.276770830154419, "logits/rejected": -1.0280699729919434, "logps/chosen": -126.50442504882812, "logps/ref_chosen": -90.77254486083984, "logps/ref_rejected": -94.58816528320312, "logps/rejected": -149.9219207763672, "loss": 0.9648, "margin_dpo/margin_mean": 19.601879119873047, "margin_dpo/margin_std": 25.395401000976562, "step": 259 }, { "epoch": 0.38179148311306904, "fcm_dpo/beta": 0.039135679602622986, "fcm_dpo/delta": 0.13330310583114624, "fcm_dpo/margin": 18.564983367919922, "fcm_dpo/q_t": 0.35755765438079834, "grad_norm": 18.12778663635254, "learning_rate": 3.9022350248844246e-07, "logits/chosen": -1.5422040224075317, "logits/rejected": -1.5151578187942505, "logps/chosen": -111.06421661376953, "logps/ref_chosen": -75.59269714355469, "logps/ref_rejected": -100.84554290771484, "logps/rejected": -154.8820343017578, "loss": 1.0363, "margin_dpo/margin_mean": 18.564983367919922, "margin_dpo/margin_std": 27.767108917236328, "step": 260 }, { "epoch": 0.3832599118942731, "fcm_dpo/beta": 0.038379594683647156, "fcm_dpo/delta": -0.1646619439125061, "fcm_dpo/margin": 25.985095977783203, "fcm_dpo/q_t": 0.2999562621116638, "grad_norm": 14.235591888427734, "learning_rate": 3.891592063515376e-07, "logits/chosen": -2.0275070667266846, "logits/rejected": -1.6710395812988281, "logps/chosen": -123.13007354736328, "logps/ref_chosen": -94.25491333007812, "logps/ref_rejected": -93.65699005126953, "logps/rejected": -148.51724243164062, "loss": 0.7772, "margin_dpo/margin_mean": 25.985097885131836, "margin_dpo/margin_std": 24.243711471557617, "step": 261 }, { "epoch": 0.38472834067547723, "fcm_dpo/beta": 0.037830717861652374, "fcm_dpo/delta": -0.05457323044538498, "fcm_dpo/margin": 23.777236938476562, "fcm_dpo/q_t": 0.3187333941459656, "grad_norm": 15.138924598693848, "learning_rate": 3.880912432401264e-07, "logits/chosen": -1.6778860092163086, "logits/rejected": -1.1244826316833496, "logps/chosen": -117.18390655517578, "logps/ref_chosen": -85.26730346679688, "logps/ref_rejected": -90.82609558105469, "logps/rejected": -146.51992797851562, "loss": 0.8426, "margin_dpo/margin_mean": 23.777236938476562, "margin_dpo/margin_std": 24.42955780029297, "step": 262 }, { "epoch": 0.38619676945668135, "fcm_dpo/beta": 0.03582540154457092, "fcm_dpo/delta": -0.3481602668762207, "fcm_dpo/margin": 32.29640197753906, "fcm_dpo/q_t": 0.2730727195739746, "grad_norm": 13.927531242370605, "learning_rate": 3.870196412960302e-07, "logits/chosen": -1.8874688148498535, "logits/rejected": -1.3352882862091064, "logps/chosen": -121.56033325195312, "logps/ref_chosen": -95.75790405273438, "logps/ref_rejected": -101.83377075195312, "logps/rejected": -159.93260192871094, "loss": 0.7319, "margin_dpo/margin_mean": 32.29640197753906, "margin_dpo/margin_std": 28.878860473632812, "step": 263 }, { "epoch": 0.3876651982378855, "fcm_dpo/beta": 0.03467824310064316, "fcm_dpo/delta": -0.09107442945241928, "fcm_dpo/margin": 26.880929946899414, "fcm_dpo/q_t": 0.31235480308532715, "grad_norm": 13.216853141784668, "learning_rate": 3.8594442875695665e-07, "logits/chosen": -1.9384407997131348, "logits/rejected": -1.6622414588928223, "logps/chosen": -118.2774658203125, "logps/ref_chosen": -90.6226577758789, "logps/ref_rejected": -100.32554626464844, "logps/rejected": -154.8612823486328, "loss": 0.8316, "margin_dpo/margin_mean": 26.88092803955078, "margin_dpo/margin_std": 26.818939208984375, "step": 264 }, { "epoch": 0.3891336270190896, "fcm_dpo/beta": 0.034085676074028015, "fcm_dpo/delta": -0.047373127192258835, "fcm_dpo/margin": 26.153051376342773, "fcm_dpo/q_t": 0.3252333402633667, "grad_norm": 12.807720184326172, "learning_rate": 3.848656339557562e-07, "logits/chosen": -1.829958200454712, "logits/rejected": -1.4463666677474976, "logps/chosen": -122.73773193359375, "logps/ref_chosen": -92.37232971191406, "logps/ref_rejected": -94.62757873535156, "logps/rejected": -151.14602661132812, "loss": 0.8721, "margin_dpo/margin_mean": 26.153051376342773, "margin_dpo/margin_std": 28.59259605407715, "step": 265 }, { "epoch": 0.39060205580029367, "fcm_dpo/beta": 0.03454948216676712, "fcm_dpo/delta": 0.06971698999404907, "fcm_dpo/margin": 22.744022369384766, "fcm_dpo/q_t": 0.3368097245693207, "grad_norm": 16.510560989379883, "learning_rate": 3.8378328531967507e-07, "logits/chosen": -2.1291229724884033, "logits/rejected": -1.4051086902618408, "logps/chosen": -136.4056396484375, "logps/ref_chosen": -102.20002746582031, "logps/ref_rejected": -74.36642456054688, "logps/rejected": -131.31605529785156, "loss": 0.8945, "margin_dpo/margin_mean": 22.744022369384766, "margin_dpo/margin_std": 23.8177433013916, "step": 266 }, { "epoch": 0.3920704845814978, "fcm_dpo/beta": 0.034008704125881195, "fcm_dpo/delta": -0.09747851639986038, "fcm_dpo/margin": 27.53805160522461, "fcm_dpo/q_t": 0.30594033002853394, "grad_norm": 13.239112854003906, "learning_rate": 3.8269741136960646e-07, "logits/chosen": -1.465944766998291, "logits/rejected": -0.8757386207580566, "logps/chosen": -133.18771362304688, "logps/ref_chosen": -104.28599548339844, "logps/ref_rejected": -95.98719024658203, "logps/rejected": -152.4269561767578, "loss": 0.7845, "margin_dpo/margin_mean": 27.538053512573242, "margin_dpo/margin_std": 23.816219329833984, "step": 267 }, { "epoch": 0.3935389133627019, "fcm_dpo/beta": 0.03355433791875839, "fcm_dpo/delta": -0.09911436587572098, "fcm_dpo/margin": 28.003337860107422, "fcm_dpo/q_t": 0.30674782395362854, "grad_norm": 13.598925590515137, "learning_rate": 3.8160804071933894e-07, "logits/chosen": -1.3583710193634033, "logits/rejected": -1.1435065269470215, "logps/chosen": -114.95004272460938, "logps/ref_chosen": -86.69622039794922, "logps/ref_rejected": -109.19184112548828, "logps/rejected": -165.44898986816406, "loss": 0.7982, "margin_dpo/margin_mean": 28.003337860107422, "margin_dpo/margin_std": 25.024091720581055, "step": 268 }, { "epoch": 0.39500734214390604, "fcm_dpo/beta": 0.03295326232910156, "fcm_dpo/delta": -0.11522669345140457, "fcm_dpo/margin": 28.961397171020508, "fcm_dpo/q_t": 0.3074827194213867, "grad_norm": 13.537002563476562, "learning_rate": 3.8051520207480204e-07, "logits/chosen": -1.6168208122253418, "logits/rejected": -1.1353833675384521, "logps/chosen": -134.0711669921875, "logps/ref_chosen": -104.97181701660156, "logps/ref_rejected": -112.47647094726562, "logps/rejected": -170.53720092773438, "loss": 0.808, "margin_dpo/margin_mean": 28.961397171020508, "margin_dpo/margin_std": 27.48983383178711, "step": 269 }, { "epoch": 0.3964757709251101, "fcm_dpo/beta": 0.03269010782241821, "fcm_dpo/delta": -0.0122775137424469, "fcm_dpo/margin": 26.342090606689453, "fcm_dpo/q_t": 0.3317517638206482, "grad_norm": 14.160609245300293, "learning_rate": 3.794189242333106e-07, "logits/chosen": -2.4010086059570312, "logits/rejected": -1.9900360107421875, "logps/chosen": -132.8267822265625, "logps/ref_chosen": -101.07383728027344, "logps/ref_rejected": -117.75289154052734, "logps/rejected": -175.84793090820312, "loss": 0.9059, "margin_dpo/margin_mean": 26.342090606689453, "margin_dpo/margin_std": 31.022964477539062, "step": 270 }, { "epoch": 0.39794419970631423, "fcm_dpo/beta": 0.0318414531648159, "fcm_dpo/delta": -0.1332307755947113, "fcm_dpo/margin": 30.452823638916016, "fcm_dpo/q_t": 0.2993617653846741, "grad_norm": 12.96033763885498, "learning_rate": 3.7831923608280514e-07, "logits/chosen": -1.2467565536499023, "logits/rejected": -0.7933920621871948, "logps/chosen": -123.05168151855469, "logps/ref_chosen": -96.72459411621094, "logps/ref_rejected": -98.5244140625, "logps/rejected": -155.3043212890625, "loss": 0.8195, "margin_dpo/margin_mean": 30.45282745361328, "margin_dpo/margin_std": 29.1860294342041, "step": 271 }, { "epoch": 0.39941262848751835, "fcm_dpo/beta": 0.03125782310962677, "fcm_dpo/delta": -0.09906208515167236, "fcm_dpo/margin": 30.06342124938965, "fcm_dpo/q_t": 0.31359294056892395, "grad_norm": 14.11276912689209, "learning_rate": 3.772161666010912e-07, "logits/chosen": -1.4559786319732666, "logits/rejected": -1.2856930494308472, "logps/chosen": -102.26274871826172, "logps/ref_chosen": -80.97721862792969, "logps/ref_rejected": -108.55535888671875, "logps/rejected": -159.90431213378906, "loss": 0.831, "margin_dpo/margin_mean": 30.06342124938965, "margin_dpo/margin_std": 30.461017608642578, "step": 272 }, { "epoch": 0.4008810572687225, "fcm_dpo/beta": 0.030184127390384674, "fcm_dpo/delta": -0.10749184340238571, "fcm_dpo/margin": 31.21113395690918, "fcm_dpo/q_t": 0.311345636844635, "grad_norm": 13.089473724365234, "learning_rate": 3.761097448550755e-07, "logits/chosen": -1.5064098834991455, "logits/rejected": -1.1125901937484741, "logps/chosen": -120.93928527832031, "logps/ref_chosen": -92.22460174560547, "logps/ref_rejected": -97.3630599975586, "logps/rejected": -157.28887939453125, "loss": 0.8571, "margin_dpo/margin_mean": 31.211132049560547, "margin_dpo/margin_std": 32.5693359375, "step": 273 }, { "epoch": 0.4023494860499266, "fcm_dpo/beta": 0.030762720853090286, "fcm_dpo/delta": 0.08655337244272232, "fcm_dpo/margin": 25.03656578063965, "fcm_dpo/q_t": 0.3384164571762085, "grad_norm": 14.684450149536133, "learning_rate": 3.75e-07, "logits/chosen": -1.8288625478744507, "logits/rejected": -1.4254413843154907, "logps/chosen": -112.06315612792969, "logps/ref_chosen": -82.26089477539062, "logps/ref_rejected": -83.87699890136719, "logps/rejected": -138.7158203125, "loss": 0.8814, "margin_dpo/margin_mean": 25.03656768798828, "margin_dpo/margin_std": 25.01040267944336, "step": 274 }, { "epoch": 0.40381791483113066, "fcm_dpo/beta": 0.03131476789712906, "fcm_dpo/delta": 0.010461606085300446, "fcm_dpo/margin": 26.782379150390625, "fcm_dpo/q_t": 0.32816171646118164, "grad_norm": 14.658917427062988, "learning_rate": 3.738869612786737e-07, "logits/chosen": -1.936241626739502, "logits/rejected": -1.6021487712860107, "logps/chosen": -104.04966735839844, "logps/ref_chosen": -79.68696594238281, "logps/ref_rejected": -98.7509765625, "logps/rejected": -149.89605712890625, "loss": 0.8889, "margin_dpo/margin_mean": 26.782379150390625, "margin_dpo/margin_std": 28.50400161743164, "step": 275 }, { "epoch": 0.4052863436123348, "fcm_dpo/beta": 0.0309858750551939, "fcm_dpo/delta": 0.06398846209049225, "fcm_dpo/margin": 25.471071243286133, "fcm_dpo/q_t": 0.33972838521003723, "grad_norm": 15.50623607635498, "learning_rate": 3.7277065802070204e-07, "logits/chosen": -1.787797212600708, "logits/rejected": -1.2390464544296265, "logps/chosen": -118.15924835205078, "logps/ref_chosen": -86.53970336914062, "logps/ref_rejected": -77.85394287109375, "logps/rejected": -134.94456481933594, "loss": 0.9182, "margin_dpo/margin_mean": 25.471071243286133, "margin_dpo/margin_std": 28.662944793701172, "step": 276 }, { "epoch": 0.4067547723935389, "fcm_dpo/beta": 0.0317692831158638, "fcm_dpo/delta": 0.08927221596240997, "fcm_dpo/margin": 24.16355323791504, "fcm_dpo/q_t": 0.3370756506919861, "grad_norm": 16.35001564025879, "learning_rate": 3.71651119641714e-07, "logits/chosen": -2.0091605186462402, "logits/rejected": -1.5972063541412354, "logps/chosen": -115.12983703613281, "logps/ref_chosen": -84.24411010742188, "logps/ref_rejected": -98.83421325683594, "logps/rejected": -153.88348388671875, "loss": 0.8847, "margin_dpo/margin_mean": 24.163555145263672, "margin_dpo/margin_std": 23.77988624572754, "step": 277 }, { "epoch": 0.40822320117474303, "fcm_dpo/beta": 0.0317951962351799, "fcm_dpo/delta": 0.00917709618806839, "fcm_dpo/margin": 26.432132720947266, "fcm_dpo/q_t": 0.32769787311553955, "grad_norm": 14.036833763122559, "learning_rate": 3.705283756425872e-07, "logits/chosen": -1.9561946392059326, "logits/rejected": -1.744471788406372, "logps/chosen": -112.0636978149414, "logps/ref_chosen": -82.431884765625, "logps/ref_rejected": -97.85691833496094, "logps/rejected": -153.92086791992188, "loss": 0.8902, "margin_dpo/margin_mean": 26.432132720947266, "margin_dpo/margin_std": 28.450267791748047, "step": 278 }, { "epoch": 0.40969162995594716, "fcm_dpo/beta": 0.03182877600193024, "fcm_dpo/delta": -0.011635363101959229, "fcm_dpo/margin": 27.01772689819336, "fcm_dpo/q_t": 0.3264973759651184, "grad_norm": 14.092531204223633, "learning_rate": 3.6940245560867e-07, "logits/chosen": -1.6284323930740356, "logits/rejected": -1.3937866687774658, "logps/chosen": -116.97288513183594, "logps/ref_chosen": -85.16799926757812, "logps/ref_rejected": -94.12664794921875, "logps/rejected": -152.94924926757812, "loss": 0.8698, "margin_dpo/margin_mean": 27.01772689819336, "margin_dpo/margin_std": 28.68284034729004, "step": 279 }, { "epoch": 0.4111600587371512, "fcm_dpo/beta": 0.03129285201430321, "fcm_dpo/delta": -0.15949808061122894, "fcm_dpo/margin": 31.758270263671875, "fcm_dpo/q_t": 0.29896456003189087, "grad_norm": 13.670494079589844, "learning_rate": 3.6827338920900253e-07, "logits/chosen": -1.997270107269287, "logits/rejected": -1.7795908451080322, "logps/chosen": -115.86480712890625, "logps/ref_chosen": -85.85641479492188, "logps/ref_rejected": -104.11859130859375, "logps/rejected": -165.88525390625, "loss": 0.7889, "margin_dpo/margin_mean": 31.758270263671875, "margin_dpo/margin_std": 28.985366821289062, "step": 280 }, { "epoch": 0.41262848751835535, "fcm_dpo/beta": 0.03030802309513092, "fcm_dpo/delta": -0.14983683824539185, "fcm_dpo/margin": 32.501609802246094, "fcm_dpo/q_t": 0.30702778697013855, "grad_norm": 11.849176406860352, "learning_rate": 3.6714120619553435e-07, "logits/chosen": -1.8329683542251587, "logits/rejected": -1.104137659072876, "logps/chosen": -117.03921508789062, "logps/ref_chosen": -86.55082702636719, "logps/ref_rejected": -88.62866973876953, "logps/rejected": -151.61868286132812, "loss": 0.8201, "margin_dpo/margin_mean": 32.501609802246094, "margin_dpo/margin_std": 33.797786712646484, "step": 281 }, { "epoch": 0.41409691629955947, "fcm_dpo/beta": 0.028760947287082672, "fcm_dpo/delta": -0.26554107666015625, "fcm_dpo/margin": 37.71876907348633, "fcm_dpo/q_t": 0.28032368421554565, "grad_norm": 11.502086639404297, "learning_rate": 3.660059364023408e-07, "logits/chosen": -1.680617332458496, "logits/rejected": -1.3186070919036865, "logps/chosen": -135.4449462890625, "logps/ref_chosen": -105.10511016845703, "logps/ref_rejected": -102.85336303710938, "logps/rejected": -170.91197204589844, "loss": 0.7132, "margin_dpo/margin_mean": 37.71876525878906, "margin_dpo/margin_std": 30.542362213134766, "step": 282 }, { "epoch": 0.4155653450807636, "fcm_dpo/beta": 0.029016414657235146, "fcm_dpo/delta": 0.11088653653860092, "fcm_dpo/margin": 25.754066467285156, "fcm_dpo/q_t": 0.35040855407714844, "grad_norm": 17.384960174560547, "learning_rate": 3.6486760974483685e-07, "logits/chosen": -1.991034746170044, "logits/rejected": -1.643927812576294, "logps/chosen": -130.67234802246094, "logps/ref_chosen": -95.05259704589844, "logps/ref_rejected": -103.54454803466797, "logps/rejected": -164.91836547851562, "loss": 0.9866, "margin_dpo/margin_mean": 25.754066467285156, "margin_dpo/margin_std": 34.688026428222656, "step": 283 }, { "epoch": 0.4170337738619677, "fcm_dpo/beta": 0.028016947209835052, "fcm_dpo/delta": -0.22263917326927185, "fcm_dpo/margin": 37.3685188293457, "fcm_dpo/q_t": 0.28344595432281494, "grad_norm": 11.067302703857422, "learning_rate": 3.6372625621898863e-07, "logits/chosen": -1.5473668575286865, "logits/rejected": -1.2372541427612305, "logps/chosen": -116.82695007324219, "logps/ref_chosen": -87.66647338867188, "logps/ref_rejected": -98.75103759765625, "logps/rejected": -165.280029296875, "loss": 0.7178, "margin_dpo/margin_mean": 37.36852264404297, "margin_dpo/margin_std": 29.709064483642578, "step": 284 }, { "epoch": 0.4185022026431718, "fcm_dpo/beta": 0.028491143137216568, "fcm_dpo/delta": 0.10801464319229126, "fcm_dpo/margin": 26.28765106201172, "fcm_dpo/q_t": 0.3376986086368561, "grad_norm": 18.838285446166992, "learning_rate": 3.625819059005228e-07, "logits/chosen": -2.3627262115478516, "logits/rejected": -2.08278751373291, "logps/chosen": -128.71075439453125, "logps/ref_chosen": -94.43303680419922, "logps/ref_rejected": -104.07194519042969, "logps/rejected": -164.6373291015625, "loss": 0.8901, "margin_dpo/margin_mean": 26.28765106201172, "margin_dpo/margin_std": 25.488826751708984, "step": 285 }, { "epoch": 0.4199706314243759, "fcm_dpo/beta": 0.02817067690193653, "fcm_dpo/delta": -0.06162068247795105, "fcm_dpo/margin": 32.15382766723633, "fcm_dpo/q_t": 0.312236487865448, "grad_norm": 14.387185096740723, "learning_rate": 3.614345889441346e-07, "logits/chosen": -1.7623761892318726, "logits/rejected": -1.3088512420654297, "logps/chosen": -140.90280151367188, "logps/ref_chosen": -103.72039794921875, "logps/ref_rejected": -96.25775909423828, "logps/rejected": -165.593994140625, "loss": 0.8213, "margin_dpo/margin_mean": 32.153831481933594, "margin_dpo/margin_std": 29.81972885131836, "step": 286 }, { "epoch": 0.42143906020558003, "fcm_dpo/beta": 0.0293489508330822, "fcm_dpo/delta": 0.2245098203420639, "fcm_dpo/margin": 21.69069480895996, "fcm_dpo/q_t": 0.37523770332336426, "grad_norm": 15.388336181640625, "learning_rate": 3.6028433558269275e-07, "logits/chosen": -2.1079835891723633, "logits/rejected": -1.5590211153030396, "logps/chosen": -137.75592041015625, "logps/ref_chosen": -93.88988494873047, "logps/ref_rejected": -83.33365631103516, "logps/rejected": -148.89036560058594, "loss": 1.0341, "margin_dpo/margin_mean": 21.69069480895996, "margin_dpo/margin_std": 31.85369873046875, "step": 287 }, { "epoch": 0.42290748898678415, "fcm_dpo/beta": 0.028778456151485443, "fcm_dpo/delta": -0.13858668506145477, "fcm_dpo/margin": 33.86872100830078, "fcm_dpo/q_t": 0.3073822855949402, "grad_norm": 13.900410652160645, "learning_rate": 3.5913117612644327e-07, "logits/chosen": -1.7789084911346436, "logits/rejected": -1.3899840116500854, "logps/chosen": -129.65771484375, "logps/ref_chosen": -88.15602111816406, "logps/ref_rejected": -93.28195190429688, "logps/rejected": -168.65237426757812, "loss": 0.8138, "margin_dpo/margin_mean": 33.86872100830078, "margin_dpo/margin_std": 33.53106689453125, "step": 288 }, { "epoch": 0.4243759177679883, "fcm_dpo/beta": 0.02798629179596901, "fcm_dpo/delta": -0.18543705344200134, "fcm_dpo/margin": 36.33726501464844, "fcm_dpo/q_t": 0.31428423523902893, "grad_norm": 12.854353904724121, "learning_rate": 3.5797514096221024e-07, "logits/chosen": -2.2386109828948975, "logits/rejected": -2.0394985675811768, "logps/chosen": -116.66321563720703, "logps/ref_chosen": -75.39292907714844, "logps/ref_rejected": -93.15428161621094, "logps/rejected": -170.7618408203125, "loss": 0.8394, "margin_dpo/margin_mean": 36.33726501464844, "margin_dpo/margin_std": 40.785675048828125, "step": 289 }, { "epoch": 0.42584434654919234, "fcm_dpo/beta": 0.02720276080071926, "fcm_dpo/delta": -0.1718001663684845, "fcm_dpo/margin": 36.90747833251953, "fcm_dpo/q_t": 0.30523180961608887, "grad_norm": 13.576156616210938, "learning_rate": 3.568162605525952e-07, "logits/chosen": -1.2316092252731323, "logits/rejected": -1.231179118156433, "logps/chosen": -133.9448699951172, "logps/ref_chosen": -88.04197692871094, "logps/ref_rejected": -123.212158203125, "logps/rejected": -206.02252197265625, "loss": 0.8157, "margin_dpo/margin_mean": 36.90747833251953, "margin_dpo/margin_std": 38.05936050415039, "step": 290 }, { "epoch": 0.42731277533039647, "fcm_dpo/beta": 0.027082961052656174, "fcm_dpo/delta": 0.08572201430797577, "fcm_dpo/margin": 28.44838523864746, "fcm_dpo/q_t": 0.33763566613197327, "grad_norm": 14.184231758117676, "learning_rate": 3.5565456543517485e-07, "logits/chosen": -2.211665153503418, "logits/rejected": -1.781091570854187, "logps/chosen": -138.7436065673828, "logps/ref_chosen": -94.09524536132812, "logps/ref_rejected": -96.05006408691406, "logps/rejected": -169.14682006835938, "loss": 0.9222, "margin_dpo/margin_mean": 28.448383331298828, "margin_dpo/margin_std": 32.05262756347656, "step": 291 }, { "epoch": 0.4287812041116006, "fcm_dpo/beta": 0.02674213796854019, "fcm_dpo/delta": -0.09594199061393738, "fcm_dpo/margin": 35.03672790527344, "fcm_dpo/q_t": 0.3002822995185852, "grad_norm": 12.321151733398438, "learning_rate": 3.5449008622169583e-07, "logits/chosen": -2.063217878341675, "logits/rejected": -1.6173973083496094, "logps/chosen": -132.1011505126953, "logps/ref_chosen": -88.25041198730469, "logps/ref_rejected": -96.41764068603516, "logps/rejected": -175.30511474609375, "loss": 0.7616, "margin_dpo/margin_mean": 35.03672790527344, "margin_dpo/margin_std": 26.89871597290039, "step": 292 }, { "epoch": 0.4302496328928047, "fcm_dpo/beta": 0.027519384399056435, "fcm_dpo/delta": 0.16997066140174866, "fcm_dpo/margin": 25.03589630126953, "fcm_dpo/q_t": 0.364579975605011, "grad_norm": 12.596840858459473, "learning_rate": 3.5332285359726846e-07, "logits/chosen": -2.3722639083862305, "logits/rejected": -1.8270535469055176, "logps/chosen": -134.981689453125, "logps/ref_chosen": -87.37654113769531, "logps/ref_rejected": -85.75579833984375, "logps/rejected": -158.39686584472656, "loss": 0.9934, "margin_dpo/margin_mean": 25.03589630126953, "margin_dpo/margin_std": 33.70214080810547, "step": 293 }, { "epoch": 0.43171806167400884, "fcm_dpo/beta": 0.02852361649274826, "fcm_dpo/delta": 0.2223101556301117, "fcm_dpo/margin": 22.482465744018555, "fcm_dpo/q_t": 0.37413203716278076, "grad_norm": 14.340274810791016, "learning_rate": 3.5215289831955786e-07, "logits/chosen": -1.5175267457962036, "logits/rejected": -1.3194074630737305, "logps/chosen": -123.25398254394531, "logps/ref_chosen": -73.50794982910156, "logps/ref_rejected": -88.08877563476562, "logps/rejected": -160.31727600097656, "loss": 1.0366, "margin_dpo/margin_mean": 22.482467651367188, "margin_dpo/margin_std": 34.42567443847656, "step": 294 }, { "epoch": 0.4331864904552129, "fcm_dpo/beta": 0.028234336525201797, "fcm_dpo/delta": -0.08705528825521469, "fcm_dpo/margin": 32.85896301269531, "fcm_dpo/q_t": 0.3156839907169342, "grad_norm": 13.020578384399414, "learning_rate": 3.509802512179737e-07, "logits/chosen": -1.6198558807373047, "logits/rejected": -1.4796977043151855, "logps/chosen": -125.44113159179688, "logps/ref_chosen": -77.76548767089844, "logps/ref_rejected": -94.24726867675781, "logps/rejected": -174.78189086914062, "loss": 0.8424, "margin_dpo/margin_mean": 32.85896301269531, "margin_dpo/margin_std": 34.66865539550781, "step": 295 }, { "epoch": 0.434654919236417, "fcm_dpo/beta": 0.02758461982011795, "fcm_dpo/delta": -0.12351512908935547, "fcm_dpo/margin": 34.80379104614258, "fcm_dpo/q_t": 0.3101541996002197, "grad_norm": 13.690773963928223, "learning_rate": 3.498049431928577e-07, "logits/chosen": -2.44696044921875, "logits/rejected": -2.000422716140747, "logps/chosen": -149.84738159179688, "logps/ref_chosen": -97.85641479492188, "logps/ref_rejected": -100.81631469726562, "logps/rejected": -187.61105346679688, "loss": 0.8245, "margin_dpo/margin_mean": 34.80379104614258, "margin_dpo/margin_std": 36.020755767822266, "step": 296 }, { "epoch": 0.43612334801762115, "fcm_dpo/beta": 0.027776416391134262, "fcm_dpo/delta": 0.026497282087802887, "fcm_dpo/margin": 29.71851921081543, "fcm_dpo/q_t": 0.3335018455982208, "grad_norm": 14.7246675491333, "learning_rate": 3.486270052146694e-07, "logits/chosen": -0.898391604423523, "logits/rejected": -0.5614018440246582, "logps/chosen": -142.22756958007812, "logps/ref_chosen": -88.56583404541016, "logps/ref_rejected": -101.55656433105469, "logps/rejected": -184.9368133544922, "loss": 0.8784, "margin_dpo/margin_mean": 29.718517303466797, "margin_dpo/margin_std": 31.882976531982422, "step": 297 }, { "epoch": 0.43759177679882527, "fcm_dpo/beta": 0.027368562296032906, "fcm_dpo/delta": -0.12677723169326782, "fcm_dpo/margin": 35.24578857421875, "fcm_dpo/q_t": 0.31449317932128906, "grad_norm": 16.305553436279297, "learning_rate": 3.474464683231698e-07, "logits/chosen": -1.7980400323867798, "logits/rejected": -1.693596363067627, "logps/chosen": -143.30047607421875, "logps/ref_chosen": -94.88043975830078, "logps/ref_rejected": -122.31101989746094, "logps/rejected": -205.97686767578125, "loss": 0.8283, "margin_dpo/margin_mean": 35.24578857421875, "margin_dpo/margin_std": 37.54370880126953, "step": 298 }, { "epoch": 0.4390602055800294, "fcm_dpo/beta": 0.026351025328040123, "fcm_dpo/delta": -0.15279431641101837, "fcm_dpo/margin": 37.456703186035156, "fcm_dpo/q_t": 0.3005375564098358, "grad_norm": 12.541773796081543, "learning_rate": 3.462633636266041e-07, "logits/chosen": -1.1973588466644287, "logits/rejected": -0.9508827924728394, "logps/chosen": -130.56175231933594, "logps/ref_chosen": -80.40835571289062, "logps/ref_rejected": -89.53717041015625, "logps/rejected": -177.1472625732422, "loss": 0.7802, "margin_dpo/margin_mean": 37.456703186035156, "margin_dpo/margin_std": 33.8513298034668, "step": 299 }, { "epoch": 0.44052863436123346, "fcm_dpo/beta": 0.026305727660655975, "fcm_dpo/delta": 0.02844477817416191, "fcm_dpo/margin": 31.3201904296875, "fcm_dpo/q_t": 0.33717983961105347, "grad_norm": 13.94998836517334, "learning_rate": 3.4507772230088147e-07, "logits/chosen": -1.4799878597259521, "logits/rejected": -1.256316065788269, "logps/chosen": -142.23712158203125, "logps/ref_chosen": -88.15890502929688, "logps/ref_rejected": -100.93919372558594, "logps/rejected": -186.3376007080078, "loss": 0.9115, "margin_dpo/margin_mean": 31.3201904296875, "margin_dpo/margin_std": 36.975074768066406, "step": 300 }, { "epoch": 0.44052863436123346, "eval_fcm_dpo/beta": 0.026378802955150604, "eval_logits/chosen": -1.7435120344161987, "eval_logits/rejected": -1.4185765981674194, "eval_logps/chosen": -155.74705505371094, "eval_logps/ref_chosen": -100.49356842041016, "eval_logps/ref_rejected": -94.06775665283203, "eval_logps/rejected": -175.83782958984375, "eval_loss": 0.4887809157371521, "eval_margin_dpo/margin_mean": 26.516590118408203, "eval_margin_dpo/margin_std": 33.869022369384766, "eval_runtime": 43.7724, "eval_samples_per_second": 53.435, "eval_steps_per_second": 1.691, "step": 300 }, { "epoch": 0.4419970631424376, "fcm_dpo/beta": 0.025531355291604996, "fcm_dpo/delta": -0.1449037790298462, "fcm_dpo/margin": 38.26066589355469, "fcm_dpo/q_t": 0.3068852424621582, "grad_norm": 12.821732521057129, "learning_rate": 3.4388957558875316e-07, "logits/chosen": -1.85687255859375, "logits/rejected": -1.4238810539245605, "logps/chosen": -135.0188751220703, "logps/ref_chosen": -85.50491333007812, "logps/ref_rejected": -98.91683959960938, "logps/rejected": -186.6914825439453, "loss": 0.8059, "margin_dpo/margin_mean": 38.26066589355469, "margin_dpo/margin_std": 37.27458953857422, "step": 301 }, { "epoch": 0.4434654919236417, "fcm_dpo/beta": 0.025708243250846863, "fcm_dpo/delta": 0.04868558049201965, "fcm_dpo/margin": 31.31707763671875, "fcm_dpo/q_t": 0.3379690647125244, "grad_norm": 14.223493576049805, "learning_rate": 3.426989547989902e-07, "logits/chosen": -1.2909713983535767, "logits/rejected": -1.102597951889038, "logps/chosen": -131.14703369140625, "logps/ref_chosen": -79.45040893554688, "logps/ref_rejected": -103.0591049194336, "logps/rejected": -186.07278442382812, "loss": 0.9164, "margin_dpo/margin_mean": 31.317073822021484, "margin_dpo/margin_std": 36.476539611816406, "step": 302 }, { "epoch": 0.44493392070484583, "fcm_dpo/beta": 0.025652453303337097, "fcm_dpo/delta": -0.0899156928062439, "fcm_dpo/margin": 36.312965393066406, "fcm_dpo/q_t": 0.3102775514125824, "grad_norm": 12.295307159423828, "learning_rate": 3.4150589130555773e-07, "logits/chosen": -1.6473402976989746, "logits/rejected": -1.2242443561553955, "logps/chosen": -141.21902465820312, "logps/ref_chosen": -95.59829711914062, "logps/ref_rejected": -92.0645751953125, "logps/rejected": -173.998291015625, "loss": 0.8163, "margin_dpo/margin_mean": 36.312965393066406, "margin_dpo/margin_std": 34.15446090698242, "step": 303 }, { "epoch": 0.44640234948604995, "fcm_dpo/beta": 0.025181569159030914, "fcm_dpo/delta": -0.0828293189406395, "fcm_dpo/margin": 36.74197769165039, "fcm_dpo/q_t": 0.30937981605529785, "grad_norm": 13.311530113220215, "learning_rate": 3.403104165467883e-07, "logits/chosen": -2.078594446182251, "logits/rejected": -1.6698689460754395, "logps/chosen": -146.96170043945312, "logps/ref_chosen": -99.4531478881836, "logps/ref_rejected": -98.08226013183594, "logps/rejected": -182.33279418945312, "loss": 0.8224, "margin_dpo/margin_mean": 36.74197769165039, "margin_dpo/margin_std": 34.821163177490234, "step": 304 }, { "epoch": 0.447870778267254, "fcm_dpo/beta": 0.024396728724241257, "fcm_dpo/delta": -0.0825144499540329, "fcm_dpo/margin": 37.75819778442383, "fcm_dpo/q_t": 0.31057485938072205, "grad_norm": 12.317713737487793, "learning_rate": 3.391125620245535e-07, "logits/chosen": -1.8674237728118896, "logits/rejected": -1.3910233974456787, "logps/chosen": -151.20004272460938, "logps/ref_chosen": -103.26339721679688, "logps/ref_rejected": -97.82897186279297, "logps/rejected": -183.52383422851562, "loss": 0.8155, "margin_dpo/margin_mean": 37.75819778442383, "margin_dpo/margin_std": 33.5549430847168, "step": 305 }, { "epoch": 0.44933920704845814, "fcm_dpo/beta": 0.023559659719467163, "fcm_dpo/delta": -0.24917837977409363, "fcm_dpo/margin": 45.43587875366211, "fcm_dpo/q_t": 0.2730311453342438, "grad_norm": 11.243393898010254, "learning_rate": 3.3791235930343417e-07, "logits/chosen": -2.1616029739379883, "logits/rejected": -1.6345137357711792, "logps/chosen": -141.40054321289062, "logps/ref_chosen": -100.3038330078125, "logps/ref_rejected": -92.23086547851562, "logps/rejected": -178.76345825195312, "loss": 0.6856, "margin_dpo/margin_mean": 45.435882568359375, "margin_dpo/margin_std": 30.269792556762695, "step": 306 }, { "epoch": 0.45080763582966227, "fcm_dpo/beta": 0.023632019758224487, "fcm_dpo/delta": 0.08696854114532471, "fcm_dpo/margin": 32.57453918457031, "fcm_dpo/q_t": 0.34471505880355835, "grad_norm": 13.035873413085938, "learning_rate": 3.367098400098881e-07, "logits/chosen": -1.7763289213180542, "logits/rejected": -1.4540574550628662, "logps/chosen": -144.75022888183594, "logps/ref_chosen": -96.64976501464844, "logps/ref_rejected": -93.34860229492188, "logps/rejected": -174.0236053466797, "loss": 0.9365, "margin_dpo/margin_mean": 32.57453918457031, "margin_dpo/margin_std": 39.58002471923828, "step": 307 }, { "epoch": 0.4522760646108664, "fcm_dpo/beta": 0.02385834977030754, "fcm_dpo/delta": 0.08316093683242798, "fcm_dpo/margin": 32.360687255859375, "fcm_dpo/q_t": 0.34850144386291504, "grad_norm": 14.49838638305664, "learning_rate": 3.355050358314172e-07, "logits/chosen": -1.7061140537261963, "logits/rejected": -1.4719178676605225, "logps/chosen": -134.83767700195312, "logps/ref_chosen": -85.72467041015625, "logps/ref_rejected": -86.42317962646484, "logps/rejected": -167.89688110351562, "loss": 0.9396, "margin_dpo/margin_mean": 32.360687255859375, "margin_dpo/margin_std": 39.152732849121094, "step": 308 }, { "epoch": 0.45374449339207046, "fcm_dpo/beta": 0.024311821907758713, "fcm_dpo/delta": 0.010310536250472069, "fcm_dpo/margin": 34.5657958984375, "fcm_dpo/q_t": 0.32924583554267883, "grad_norm": 13.426247596740723, "learning_rate": 3.3429797851573183e-07, "logits/chosen": -1.7498160600662231, "logits/rejected": -1.3319708108901978, "logps/chosen": -136.19140625, "logps/ref_chosen": -83.15145874023438, "logps/ref_rejected": -86.58602142333984, "logps/rejected": -174.19175720214844, "loss": 0.8826, "margin_dpo/margin_mean": 34.5657958984375, "margin_dpo/margin_std": 37.12953567504883, "step": 309 }, { "epoch": 0.4552129221732746, "fcm_dpo/beta": 0.024511564522981644, "fcm_dpo/delta": 0.053582563996315, "fcm_dpo/margin": 32.65531921386719, "fcm_dpo/q_t": 0.33659428358078003, "grad_norm": 13.409820556640625, "learning_rate": 3.3308869986991487e-07, "logits/chosen": -2.131115198135376, "logits/rejected": -1.7278515100479126, "logps/chosen": -136.57640075683594, "logps/ref_chosen": -89.73799133300781, "logps/ref_rejected": -87.07083129882812, "logps/rejected": -166.5645751953125, "loss": 0.9131, "margin_dpo/margin_mean": 32.65532302856445, "margin_dpo/margin_std": 37.27210235595703, "step": 310 }, { "epoch": 0.4566813509544787, "fcm_dpo/beta": 0.02398892678320408, "fcm_dpo/delta": -0.13488170504570007, "fcm_dpo/margin": 40.49799346923828, "fcm_dpo/q_t": 0.30414602160453796, "grad_norm": 13.267024993896484, "learning_rate": 3.3187723175958346e-07, "logits/chosen": -1.6566599607467651, "logits/rejected": -0.9458609819412231, "logps/chosen": -147.330078125, "logps/ref_chosen": -88.52644348144531, "logps/ref_rejected": -82.26608276367188, "logps/rejected": -181.56771850585938, "loss": 0.7842, "margin_dpo/margin_mean": 40.49799346923828, "margin_dpo/margin_std": 37.099029541015625, "step": 311 }, { "epoch": 0.4581497797356828, "fcm_dpo/beta": 0.023747332394123077, "fcm_dpo/delta": -0.0758887305855751, "fcm_dpo/margin": 38.68158721923828, "fcm_dpo/q_t": 0.3152478337287903, "grad_norm": 13.058525085449219, "learning_rate": 3.306636061080487e-07, "logits/chosen": -1.466660737991333, "logits/rejected": -0.9308065176010132, "logps/chosen": -134.54989624023438, "logps/ref_chosen": -80.40069580078125, "logps/ref_rejected": -82.4471435546875, "logps/rejected": -175.27792358398438, "loss": 0.8461, "margin_dpo/margin_mean": 38.68158721923828, "margin_dpo/margin_std": 39.664390563964844, "step": 312 }, { "epoch": 0.45961820851688695, "fcm_dpo/beta": 0.023635469377040863, "fcm_dpo/delta": 0.02790502831339836, "fcm_dpo/margin": 34.8759765625, "fcm_dpo/q_t": 0.3363736867904663, "grad_norm": 11.786361694335938, "learning_rate": 3.2944785489547537e-07, "logits/chosen": -2.3114123344421387, "logits/rejected": -1.8879140615463257, "logps/chosen": -134.4079132080078, "logps/ref_chosen": -84.14190673828125, "logps/ref_rejected": -93.42684173583984, "logps/rejected": -178.56883239746094, "loss": 0.9262, "margin_dpo/margin_mean": 34.8759765625, "margin_dpo/margin_std": 42.29779815673828, "step": 313 }, { "epoch": 0.461086637298091, "fcm_dpo/beta": 0.02345544472336769, "fcm_dpo/delta": -0.014135785400867462, "fcm_dpo/margin": 36.76209259033203, "fcm_dpo/q_t": 0.32853952050209045, "grad_norm": 13.441848754882812, "learning_rate": 3.2823001015803857e-07, "logits/chosen": -1.984985113143921, "logits/rejected": -1.779584288597107, "logps/chosen": -139.60719299316406, "logps/ref_chosen": -88.80972290039062, "logps/ref_rejected": -104.15892028808594, "logps/rejected": -191.71847534179688, "loss": 0.8832, "margin_dpo/margin_mean": 36.76209259033203, "margin_dpo/margin_std": 40.621002197265625, "step": 314 }, { "epoch": 0.46255506607929514, "fcm_dpo/beta": 0.023481205105781555, "fcm_dpo/delta": -0.0164218470454216, "fcm_dpo/margin": 36.83607482910156, "fcm_dpo/q_t": 0.3287695348262787, "grad_norm": 13.707772254943848, "learning_rate": 3.270101039870797e-07, "logits/chosen": -1.4009931087493896, "logits/rejected": -1.214354395866394, "logps/chosen": -131.05386352539062, "logps/ref_chosen": -80.793701171875, "logps/ref_rejected": -92.13711547851562, "logps/rejected": -179.23336791992188, "loss": 0.9178, "margin_dpo/margin_mean": 36.83607482910156, "margin_dpo/margin_std": 44.07396697998047, "step": 315 }, { "epoch": 0.46402349486049926, "fcm_dpo/beta": 0.023532241582870483, "fcm_dpo/delta": -0.0018831361085176468, "fcm_dpo/margin": 36.19043731689453, "fcm_dpo/q_t": 0.3307906985282898, "grad_norm": 13.95568561553955, "learning_rate": 3.2578816852826086e-07, "logits/chosen": -1.5216803550720215, "logits/rejected": -1.5168416500091553, "logps/chosen": -138.3435516357422, "logps/ref_chosen": -84.18424987792969, "logps/ref_rejected": -105.6908187866211, "logps/rejected": -196.04055786132812, "loss": 0.8913, "margin_dpo/margin_mean": 36.19043731689453, "margin_dpo/margin_std": 41.48989486694336, "step": 316 }, { "epoch": 0.4654919236417034, "fcm_dpo/beta": 0.022490374743938446, "fcm_dpo/delta": -0.31194061040878296, "fcm_dpo/margin": 50.12665557861328, "fcm_dpo/q_t": 0.28289079666137695, "grad_norm": 13.945904731750488, "learning_rate": 3.2456423598071783e-07, "logits/chosen": -1.6591770648956299, "logits/rejected": -1.1557483673095703, "logps/chosen": -140.0654754638672, "logps/ref_chosen": -87.87348937988281, "logps/ref_rejected": -108.26519775390625, "logps/rejected": -210.58383178710938, "loss": 0.7508, "margin_dpo/margin_mean": 50.12665557861328, "margin_dpo/margin_std": 46.59101486206055, "step": 317 }, { "epoch": 0.4669603524229075, "fcm_dpo/beta": 0.02168363705277443, "fcm_dpo/delta": -0.09581036865711212, "fcm_dpo/margin": 43.18408966064453, "fcm_dpo/q_t": 0.30946460366249084, "grad_norm": 11.313399314880371, "learning_rate": 3.233383385962115e-07, "logits/chosen": -1.9839439392089844, "logits/rejected": -1.3179357051849365, "logps/chosen": -149.8084716796875, "logps/ref_chosen": -98.20553588867188, "logps/ref_rejected": -88.13629150390625, "logps/rejected": -182.92332458496094, "loss": 0.8149, "margin_dpo/margin_mean": 43.18408966064453, "margin_dpo/margin_std": 41.31195831298828, "step": 318 }, { "epoch": 0.4684287812041116, "fcm_dpo/beta": 0.021182440221309662, "fcm_dpo/delta": -0.12014549970626831, "fcm_dpo/margin": 45.22450256347656, "fcm_dpo/q_t": 0.30883777141571045, "grad_norm": 12.289721488952637, "learning_rate": 3.2211050867827805e-07, "logits/chosen": -1.5484849214553833, "logits/rejected": -1.4470410346984863, "logps/chosen": -126.79381561279297, "logps/ref_chosen": -82.50337219238281, "logps/ref_rejected": -118.65068054199219, "logps/rejected": -208.16561889648438, "loss": 0.8169, "margin_dpo/margin_mean": 45.22450256347656, "margin_dpo/margin_std": 44.638736724853516, "step": 319 }, { "epoch": 0.4698972099853157, "fcm_dpo/beta": 0.0210844948887825, "fcm_dpo/delta": 0.016227949410676956, "fcm_dpo/margin": 39.59357452392578, "fcm_dpo/q_t": 0.3285983204841614, "grad_norm": 11.511106491088867, "learning_rate": 3.208807785813777e-07, "logits/chosen": -2.1302032470703125, "logits/rejected": -1.9397833347320557, "logps/chosen": -132.902099609375, "logps/ref_chosen": -87.13722229003906, "logps/ref_rejected": -103.22412109375, "logps/rejected": -188.58255004882812, "loss": 0.8644, "margin_dpo/margin_mean": 39.59357452392578, "margin_dpo/margin_std": 41.0483512878418, "step": 320 }, { "epoch": 0.4713656387665198, "fcm_dpo/beta": 0.020797066390514374, "fcm_dpo/delta": -0.08268754184246063, "fcm_dpo/margin": 44.4124641418457, "fcm_dpo/q_t": 0.3112069368362427, "grad_norm": 11.961592674255371, "learning_rate": 3.1964918071004217e-07, "logits/chosen": -2.0953972339630127, "logits/rejected": -1.7250018119812012, "logps/chosen": -144.9522705078125, "logps/ref_chosen": -93.10466766357422, "logps/ref_rejected": -97.4681396484375, "logps/rejected": -193.72821044921875, "loss": 0.8341, "margin_dpo/margin_mean": 44.4124641418457, "margin_dpo/margin_std": 43.544097900390625, "step": 321 }, { "epoch": 0.47283406754772395, "fcm_dpo/beta": 0.020121555775403976, "fcm_dpo/delta": -0.17728671431541443, "fcm_dpo/margin": 50.00999450683594, "fcm_dpo/q_t": 0.2999124526977539, "grad_norm": 13.548306465148926, "learning_rate": 3.184157475180207e-07, "logits/chosen": -1.853193759918213, "logits/rejected": -1.5143083333969116, "logps/chosen": -141.55194091796875, "logps/ref_chosen": -92.19291687011719, "logps/ref_rejected": -100.89321899414062, "logps/rejected": -200.26223754882812, "loss": 0.8025, "margin_dpo/margin_mean": 50.00999450683594, "margin_dpo/margin_std": 47.50038146972656, "step": 322 }, { "epoch": 0.47430249632892807, "fcm_dpo/beta": 0.01991206780076027, "fcm_dpo/delta": -0.04079621657729149, "fcm_dpo/margin": 44.52705383300781, "fcm_dpo/q_t": 0.32042044401168823, "grad_norm": 11.607592582702637, "learning_rate": 3.171805115074251e-07, "logits/chosen": -2.05534291267395, "logits/rejected": -1.6881399154663086, "logps/chosen": -135.66567993164062, "logps/ref_chosen": -80.27210235595703, "logps/ref_rejected": -80.3033447265625, "logps/rejected": -180.22398376464844, "loss": 0.8341, "margin_dpo/margin_mean": 44.52705383300781, "margin_dpo/margin_std": 43.4844856262207, "step": 323 }, { "epoch": 0.47577092511013214, "fcm_dpo/beta": 0.0196855291724205, "fcm_dpo/delta": -0.09823101758956909, "fcm_dpo/margin": 47.70295715332031, "fcm_dpo/q_t": 0.3015575706958771, "grad_norm": 11.840632438659668, "learning_rate": 3.1594350522787295e-07, "logits/chosen": -1.6086432933807373, "logits/rejected": -0.9486087560653687, "logps/chosen": -153.7623748779297, "logps/ref_chosen": -95.72935485839844, "logps/ref_rejected": -94.16924285888672, "logps/rejected": -199.90521240234375, "loss": 0.7789, "margin_dpo/margin_mean": 47.70295715332031, "margin_dpo/margin_std": 39.34478759765625, "step": 324 }, { "epoch": 0.47723935389133626, "fcm_dpo/beta": 0.019420579075813293, "fcm_dpo/delta": -0.04867015779018402, "fcm_dpo/margin": 46.04854202270508, "fcm_dpo/q_t": 0.3144415020942688, "grad_norm": 10.232751846313477, "learning_rate": 3.147047612756302e-07, "logits/chosen": -1.746485710144043, "logits/rejected": -1.050299882888794, "logps/chosen": -152.50927734375, "logps/ref_chosen": -98.70686340332031, "logps/ref_rejected": -76.40809631347656, "logps/rejected": -176.25906372070312, "loss": 0.829, "margin_dpo/margin_mean": 46.04853820800781, "margin_dpo/margin_std": 43.221412658691406, "step": 325 }, { "epoch": 0.4787077826725404, "fcm_dpo/beta": 0.019502099603414536, "fcm_dpo/delta": 0.041523873805999756, "fcm_dpo/margin": 41.62855529785156, "fcm_dpo/q_t": 0.33376142382621765, "grad_norm": 13.21248722076416, "learning_rate": 3.134643122927519e-07, "logits/chosen": -2.4767727851867676, "logits/rejected": -1.8699431419372559, "logps/chosen": -168.19497680664062, "logps/ref_chosen": -104.66130828857422, "logps/ref_rejected": -86.82537841796875, "logps/rejected": -191.98760986328125, "loss": 0.8896, "margin_dpo/margin_mean": 41.62855529785156, "margin_dpo/margin_std": 44.87831115722656, "step": 326 }, { "epoch": 0.4801762114537445, "fcm_dpo/beta": 0.019457612186670303, "fcm_dpo/delta": -0.01857585832476616, "fcm_dpo/margin": 44.55260467529297, "fcm_dpo/q_t": 0.32346054911613464, "grad_norm": 12.107725143432617, "learning_rate": 3.1222219096622264e-07, "logits/chosen": -1.6513781547546387, "logits/rejected": -1.0370820760726929, "logps/chosen": -167.53587341308594, "logps/ref_chosen": -107.44023895263672, "logps/ref_rejected": -117.05323028564453, "logps/rejected": -221.70147705078125, "loss": 0.8525, "margin_dpo/margin_mean": 44.5526008605957, "margin_dpo/margin_std": 44.991912841796875, "step": 327 }, { "epoch": 0.48164464023494863, "fcm_dpo/beta": 0.01904640533030033, "fcm_dpo/delta": -0.14952132105827332, "fcm_dpo/margin": 51.70508575439453, "fcm_dpo/q_t": 0.30180302262306213, "grad_norm": 12.407594680786133, "learning_rate": 3.1097843002709427e-07, "logits/chosen": -1.4869141578674316, "logits/rejected": -1.2007091045379639, "logps/chosen": -150.2059783935547, "logps/ref_chosen": -86.6106948852539, "logps/ref_rejected": -97.40188598632812, "logps/rejected": -212.70225524902344, "loss": 0.7873, "margin_dpo/margin_mean": 51.70508575439453, "margin_dpo/margin_std": 47.76759338378906, "step": 328 }, { "epoch": 0.4831130690161527, "fcm_dpo/beta": 0.01887112483382225, "fcm_dpo/delta": 0.027751173824071884, "fcm_dpo/margin": 43.669822692871094, "fcm_dpo/q_t": 0.32791298627853394, "grad_norm": 13.570395469665527, "learning_rate": 3.0973306224962437e-07, "logits/chosen": -1.5214825868606567, "logits/rejected": -1.0124340057373047, "logps/chosen": -165.36264038085938, "logps/ref_chosen": -100.61308288574219, "logps/ref_rejected": -104.95370483398438, "logps/rejected": -213.37307739257812, "loss": 0.8621, "margin_dpo/margin_mean": 43.669822692871094, "margin_dpo/margin_std": 42.60850524902344, "step": 329 }, { "epoch": 0.4845814977973568, "fcm_dpo/beta": 0.018791306763887405, "fcm_dpo/delta": -0.051450639963150024, "fcm_dpo/margin": 47.70458984375, "fcm_dpo/q_t": 0.30650538206100464, "grad_norm": 11.095928192138672, "learning_rate": 3.084861204504122e-07, "logits/chosen": -1.7994372844696045, "logits/rejected": -1.7166423797607422, "logps/chosen": -135.48019409179688, "logps/ref_chosen": -76.95843505859375, "logps/ref_rejected": -92.81494140625, "logps/rejected": -199.041259765625, "loss": 0.7905, "margin_dpo/margin_mean": 47.704586029052734, "margin_dpo/margin_std": 37.720497131347656, "step": 330 }, { "epoch": 0.48604992657856094, "fcm_dpo/beta": 0.018632695078849792, "fcm_dpo/delta": -0.10063387453556061, "fcm_dpo/margin": 50.50764465332031, "fcm_dpo/q_t": 0.3032960295677185, "grad_norm": 15.277036666870117, "learning_rate": 3.072376374875335e-07, "logits/chosen": -1.9191899299621582, "logits/rejected": -1.6125640869140625, "logps/chosen": -145.8758544921875, "logps/ref_chosen": -83.05941772460938, "logps/ref_rejected": -94.13113403320312, "logps/rejected": -207.45521545410156, "loss": 0.7858, "margin_dpo/margin_mean": 50.50764465332031, "margin_dpo/margin_std": 42.87828063964844, "step": 331 }, { "epoch": 0.48751835535976507, "fcm_dpo/beta": 0.018833328038454056, "fcm_dpo/delta": 0.14665086567401886, "fcm_dpo/margin": 37.903587341308594, "fcm_dpo/q_t": 0.35199999809265137, "grad_norm": 12.529155731201172, "learning_rate": 3.059876462596758e-07, "logits/chosen": -2.0765371322631836, "logits/rejected": -1.6660399436950684, "logps/chosen": -149.06369018554688, "logps/ref_chosen": -81.93089294433594, "logps/ref_rejected": -82.84590148925781, "logps/rejected": -187.8822784423828, "loss": 0.9459, "margin_dpo/margin_mean": 37.903587341308594, "margin_dpo/margin_std": 45.02606201171875, "step": 332 }, { "epoch": 0.4889867841409692, "fcm_dpo/beta": 0.018546540290117264, "fcm_dpo/delta": -0.0819404274225235, "fcm_dpo/margin": 49.67053985595703, "fcm_dpo/q_t": 0.31946393847465515, "grad_norm": 12.655741691589355, "learning_rate": 3.0473617970527015e-07, "logits/chosen": -1.6623640060424805, "logits/rejected": -1.3435744047164917, "logps/chosen": -165.89602661132812, "logps/ref_chosen": -88.53257751464844, "logps/ref_rejected": -102.46601867675781, "logps/rejected": -229.5, "loss": 0.8978, "margin_dpo/margin_mean": 49.67053985595703, "margin_dpo/margin_std": 58.06535339355469, "step": 333 }, { "epoch": 0.49045521292217326, "fcm_dpo/beta": 0.018588945269584656, "fcm_dpo/delta": 0.01159542053937912, "fcm_dpo/margin": 45.11811828613281, "fcm_dpo/q_t": 0.3292047381401062, "grad_norm": 13.874351501464844, "learning_rate": 3.034832708016243e-07, "logits/chosen": -1.2766565084457397, "logits/rejected": -0.9840788841247559, "logps/chosen": -168.4891357421875, "logps/ref_chosen": -97.75778198242188, "logps/ref_rejected": -100.0058364868164, "logps/rejected": -215.85531616210938, "loss": 0.9167, "margin_dpo/margin_mean": 45.11811828613281, "margin_dpo/margin_std": 52.88778305053711, "step": 334 }, { "epoch": 0.4919236417033774, "fcm_dpo/beta": 0.01887522265315056, "fcm_dpo/delta": 0.058015674352645874, "fcm_dpo/margin": 42.20411682128906, "fcm_dpo/q_t": 0.3390098512172699, "grad_norm": 14.231380462646484, "learning_rate": 3.022289525640531e-07, "logits/chosen": -2.1570115089416504, "logits/rejected": -1.7516398429870605, "logps/chosen": -167.72457885742188, "logps/ref_chosen": -96.37603759765625, "logps/ref_rejected": -94.92401123046875, "logps/rejected": -208.47665405273438, "loss": 0.9201, "margin_dpo/margin_mean": 42.20411682128906, "margin_dpo/margin_std": 49.45446014404297, "step": 335 }, { "epoch": 0.4933920704845815, "fcm_dpo/beta": 0.01868092268705368, "fcm_dpo/delta": -0.11260117590427399, "fcm_dpo/margin": 50.96186828613281, "fcm_dpo/q_t": 0.3083822429180145, "grad_norm": 13.978559494018555, "learning_rate": 3.009732580450086e-07, "logits/chosen": -1.5539782047271729, "logits/rejected": -1.3770310878753662, "logps/chosen": -159.64161682128906, "logps/ref_chosen": -88.52952575683594, "logps/ref_rejected": -109.5557861328125, "logps/rejected": -231.62973022460938, "loss": 0.8231, "margin_dpo/margin_mean": 50.96186828613281, "margin_dpo/margin_std": 52.15556335449219, "step": 336 }, { "epoch": 0.4948604992657856, "fcm_dpo/beta": 0.01848072186112404, "fcm_dpo/delta": -0.04322163388133049, "fcm_dpo/margin": 48.11244583129883, "fcm_dpo/q_t": 0.3242299556732178, "grad_norm": 13.638964653015137, "learning_rate": 2.9971622033320914e-07, "logits/chosen": -2.27642822265625, "logits/rejected": -1.9854902029037476, "logps/chosen": -167.85414123535156, "logps/ref_chosen": -97.3123779296875, "logps/ref_rejected": -110.13134765625, "logps/rejected": -228.78553771972656, "loss": 0.8944, "margin_dpo/margin_mean": 48.11244583129883, "margin_dpo/margin_std": 54.82648468017578, "step": 337 }, { "epoch": 0.49632892804698975, "fcm_dpo/beta": 0.01835198514163494, "fcm_dpo/delta": 0.01314034778624773, "fcm_dpo/margin": 45.65751647949219, "fcm_dpo/q_t": 0.325469970703125, "grad_norm": 14.516743659973145, "learning_rate": 2.984578725527675e-07, "logits/chosen": -1.8865281343460083, "logits/rejected": -1.7105586528778076, "logps/chosen": -157.10043334960938, "logps/ref_chosen": -83.63011169433594, "logps/ref_rejected": -93.69235229492188, "logps/rejected": -212.8201904296875, "loss": 0.8727, "margin_dpo/margin_mean": 45.65751647949219, "margin_dpo/margin_std": 46.734107971191406, "step": 338 }, { "epoch": 0.4977973568281938, "fcm_dpo/beta": 0.018329763785004616, "fcm_dpo/delta": -0.07409132272005081, "fcm_dpo/margin": 50.01941680908203, "fcm_dpo/q_t": 0.31882256269454956, "grad_norm": 13.026541709899902, "learning_rate": 2.9719824786231796e-07, "logits/chosen": -2.9330520629882812, "logits/rejected": -2.518876552581787, "logps/chosen": -174.89187622070312, "logps/ref_chosen": -99.85182189941406, "logps/ref_rejected": -109.0743637084961, "logps/rejected": -234.1338348388672, "loss": 0.8771, "margin_dpo/margin_mean": 50.01941680908203, "margin_dpo/margin_std": 55.33876037597656, "step": 339 }, { "epoch": 0.49926578560939794, "fcm_dpo/beta": 0.018268339335918427, "fcm_dpo/delta": 0.02735849656164646, "fcm_dpo/margin": 45.13690185546875, "fcm_dpo/q_t": 0.33072808384895325, "grad_norm": 16.115903854370117, "learning_rate": 2.959373794541426e-07, "logits/chosen": -1.4078471660614014, "logits/rejected": -0.9559658765792847, "logps/chosen": -167.7802734375, "logps/ref_chosen": -84.97396850585938, "logps/ref_rejected": -78.80033874511719, "logps/rejected": -206.7435760498047, "loss": 0.8937, "margin_dpo/margin_mean": 45.13690185546875, "margin_dpo/margin_std": 51.02478790283203, "step": 340 }, { "epoch": 0.5007342143906021, "fcm_dpo/beta": 0.018367266282439232, "fcm_dpo/delta": 0.05237448215484619, "fcm_dpo/margin": 43.65577697753906, "fcm_dpo/q_t": 0.34013432264328003, "grad_norm": 15.512502670288086, "learning_rate": 2.946753005532965e-07, "logits/chosen": -1.4719302654266357, "logits/rejected": -1.1835095882415771, "logps/chosen": -173.26119995117188, "logps/ref_chosen": -85.26399230957031, "logps/ref_rejected": -106.08595275878906, "logps/rejected": -237.7389373779297, "loss": 0.9403, "margin_dpo/margin_mean": 43.65577697753906, "margin_dpo/margin_std": 53.599098205566406, "step": 341 }, { "epoch": 0.5022026431718062, "fcm_dpo/beta": 0.017494451254606247, "fcm_dpo/delta": -0.25881823897361755, "fcm_dpo/margin": 61.412132263183594, "fcm_dpo/q_t": 0.2814210057258606, "grad_norm": 13.322992324829102, "learning_rate": 2.934120444167326e-07, "logits/chosen": -1.374617338180542, "logits/rejected": -0.6580771803855896, "logps/chosen": -170.48294067382812, "logps/ref_chosen": -101.03860473632812, "logps/ref_rejected": -98.89851379394531, "logps/rejected": -229.75497436523438, "loss": 0.7611, "margin_dpo/margin_mean": 61.41213607788086, "margin_dpo/margin_std": 52.07002639770508, "step": 342 }, { "epoch": 0.5036710719530103, "fcm_dpo/beta": 0.017505202442407608, "fcm_dpo/delta": -0.06715908646583557, "fcm_dpo/margin": 51.96100616455078, "fcm_dpo/q_t": 0.3151499032974243, "grad_norm": 15.006378173828125, "learning_rate": 2.9214764433242476e-07, "logits/chosen": -1.9575207233428955, "logits/rejected": -1.8041001558303833, "logps/chosen": -157.88031005859375, "logps/ref_chosen": -83.74736022949219, "logps/ref_rejected": -116.57908630371094, "logps/rejected": -242.67303466796875, "loss": 0.8391, "margin_dpo/margin_mean": 51.96100997924805, "margin_dpo/margin_std": 50.270721435546875, "step": 343 }, { "epoch": 0.5051395007342144, "fcm_dpo/beta": 0.01725601777434349, "fcm_dpo/delta": -0.0286573339253664, "fcm_dpo/margin": 50.76178741455078, "fcm_dpo/q_t": 0.32474368810653687, "grad_norm": 13.056037902832031, "learning_rate": 2.9088213361849126e-07, "logits/chosen": -1.6103906631469727, "logits/rejected": -1.3356447219848633, "logps/chosen": -158.83035278320312, "logps/ref_chosen": -78.25396728515625, "logps/ref_rejected": -95.42870330810547, "logps/rejected": -226.76687622070312, "loss": 0.8756, "margin_dpo/margin_mean": 50.76179122924805, "margin_dpo/margin_std": 55.72649383544922, "step": 344 }, { "epoch": 0.5066079295154186, "fcm_dpo/beta": 0.01681181788444519, "fcm_dpo/delta": -0.10634002834558487, "fcm_dpo/margin": 56.23204803466797, "fcm_dpo/q_t": 0.3105340003967285, "grad_norm": 13.938704490661621, "learning_rate": 2.896155456223163e-07, "logits/chosen": -1.5161519050598145, "logits/rejected": -1.2150441408157349, "logps/chosen": -175.97979736328125, "logps/ref_chosen": -98.0794906616211, "logps/ref_rejected": -104.03477478027344, "logps/rejected": -238.16714477539062, "loss": 0.8146, "margin_dpo/margin_mean": 56.23204803466797, "margin_dpo/margin_std": 54.27198791503906, "step": 345 }, { "epoch": 0.5080763582966226, "fcm_dpo/beta": 0.017007581889629364, "fcm_dpo/delta": 0.06959997862577438, "fcm_dpo/margin": 46.195072174072266, "fcm_dpo/q_t": 0.3423931896686554, "grad_norm": 12.621767044067383, "learning_rate": 2.883479137196714e-07, "logits/chosen": -1.956493854522705, "logits/rejected": -1.6527358293533325, "logps/chosen": -170.6133270263672, "logps/ref_chosen": -80.68348693847656, "logps/ref_rejected": -85.89260864257812, "logps/rejected": -222.01751708984375, "loss": 0.9513, "margin_dpo/margin_mean": 46.19507598876953, "margin_dpo/margin_std": 58.70746994018555, "step": 346 }, { "epoch": 0.5095447870778267, "fcm_dpo/beta": 0.017326023429632187, "fcm_dpo/delta": 0.09608689695596695, "fcm_dpo/margin": 43.91358184814453, "fcm_dpo/q_t": 0.3499487340450287, "grad_norm": 14.319795608520508, "learning_rate": 2.8707927131383614e-07, "logits/chosen": -1.9093542098999023, "logits/rejected": -1.5683306455612183, "logps/chosen": -178.8655242919922, "logps/ref_chosen": -93.67938232421875, "logps/ref_rejected": -97.54251098632812, "logps/rejected": -226.64224243164062, "loss": 0.9717, "margin_dpo/margin_mean": 43.91358184814453, "margin_dpo/margin_std": 58.20018005371094, "step": 347 }, { "epoch": 0.5110132158590308, "fcm_dpo/beta": 0.01721569523215294, "fcm_dpo/delta": 0.02329825609922409, "fcm_dpo/margin": 47.98047637939453, "fcm_dpo/q_t": 0.3315523862838745, "grad_norm": 12.443164825439453, "learning_rate": 2.858096518347179e-07, "logits/chosen": -1.9739258289337158, "logits/rejected": -1.6834808588027954, "logps/chosen": -165.66465759277344, "logps/ref_chosen": -90.25224304199219, "logps/ref_rejected": -95.77027893066406, "logps/rejected": -219.1631622314453, "loss": 0.9045, "margin_dpo/margin_mean": 47.98047637939453, "margin_dpo/margin_std": 52.16603088378906, "step": 348 }, { "epoch": 0.5124816446402349, "fcm_dpo/beta": 0.01769782230257988, "fcm_dpo/delta": 0.037826307117938995, "fcm_dpo/margin": 45.96704864501953, "fcm_dpo/q_t": 0.3365427255630493, "grad_norm": 13.886954307556152, "learning_rate": 2.845390887379706e-07, "logits/chosen": -2.102787733078003, "logits/rejected": -1.881716251373291, "logps/chosen": -173.50393676757812, "logps/ref_chosen": -92.05809020996094, "logps/ref_rejected": -103.96527099609375, "logps/rejected": -231.378173828125, "loss": 0.9215, "margin_dpo/margin_mean": 45.96704864501953, "margin_dpo/margin_std": 53.656681060791016, "step": 349 }, { "epoch": 0.5139500734214391, "fcm_dpo/beta": 0.01799216866493225, "fcm_dpo/delta": 0.16198407113552094, "fcm_dpo/margin": 38.87583923339844, "fcm_dpo/q_t": 0.35764336585998535, "grad_norm": 15.808060646057129, "learning_rate": 2.8326761550411346e-07, "logits/chosen": -1.8025670051574707, "logits/rejected": -1.573468804359436, "logps/chosen": -175.3384552001953, "logps/ref_chosen": -87.05419921875, "logps/ref_rejected": -95.72193145751953, "logps/rejected": -222.88204956054688, "loss": 0.9587, "margin_dpo/margin_mean": 38.87583923339844, "margin_dpo/margin_std": 47.68621826171875, "step": 350 }, { "epoch": 0.5154185022026432, "fcm_dpo/beta": 0.017786063253879547, "fcm_dpo/delta": -0.17867089807987213, "fcm_dpo/margin": 56.837730407714844, "fcm_dpo/q_t": 0.29999619722366333, "grad_norm": 13.050240516662598, "learning_rate": 2.819952656376487e-07, "logits/chosen": -1.5670181512832642, "logits/rejected": -1.2213714122772217, "logps/chosen": -173.60598754882812, "logps/ref_chosen": -89.16903686523438, "logps/ref_rejected": -108.29801940917969, "logps/rejected": -249.57269287109375, "loss": 0.8264, "margin_dpo/margin_mean": 56.837730407714844, "margin_dpo/margin_std": 57.681922912597656, "step": 351 }, { "epoch": 0.5168869309838473, "fcm_dpo/beta": 0.01727548986673355, "fcm_dpo/delta": -0.09042805433273315, "fcm_dpo/margin": 53.94491195678711, "fcm_dpo/q_t": 0.3155178427696228, "grad_norm": 13.539352416992188, "learning_rate": 2.8072207266617854e-07, "logits/chosen": -1.7938660383224487, "logits/rejected": -1.2727184295654297, "logps/chosen": -175.17919921875, "logps/ref_chosen": -96.81147766113281, "logps/ref_rejected": -84.40483093261719, "logps/rejected": -216.71746826171875, "loss": 0.8342, "margin_dpo/margin_mean": 53.94491195678711, "margin_dpo/margin_std": 54.587371826171875, "step": 352 }, { "epoch": 0.5183553597650514, "fcm_dpo/beta": 0.017876289784908295, "fcm_dpo/delta": 0.26869261264801025, "fcm_dpo/margin": 33.43932342529297, "fcm_dpo/q_t": 0.3762671947479248, "grad_norm": 15.362466812133789, "learning_rate": 2.794480701395219e-07, "logits/chosen": -2.049345016479492, "logits/rejected": -1.686617136001587, "logps/chosen": -176.68536376953125, "logps/ref_chosen": -85.36248779296875, "logps/ref_rejected": -87.36732482910156, "logps/rejected": -212.1295166015625, "loss": 1.059, "margin_dpo/margin_mean": 33.43932342529297, "margin_dpo/margin_std": 50.67942810058594, "step": 353 }, { "epoch": 0.5198237885462555, "fcm_dpo/beta": 0.017914768308401108, "fcm_dpo/delta": -0.11532653868198395, "fcm_dpo/margin": 53.27729797363281, "fcm_dpo/q_t": 0.3020152449607849, "grad_norm": 12.614726066589355, "learning_rate": 2.781732916288303e-07, "logits/chosen": -1.9569498300552368, "logits/rejected": -1.506282091140747, "logps/chosen": -176.7467803955078, "logps/ref_chosen": -95.8011703491211, "logps/ref_rejected": -94.86614227294922, "logps/rejected": -229.0890350341797, "loss": 0.7736, "margin_dpo/margin_mean": 53.27729797363281, "margin_dpo/margin_std": 45.16783142089844, "step": 354 }, { "epoch": 0.5212922173274597, "fcm_dpo/beta": 0.017441321164369583, "fcm_dpo/delta": -0.04620569944381714, "fcm_dpo/margin": 51.005733489990234, "fcm_dpo/q_t": 0.3192770779132843, "grad_norm": 12.877270698547363, "learning_rate": 2.7689777072570284e-07, "logits/chosen": -2.404881477355957, "logits/rejected": -1.9403038024902344, "logps/chosen": -166.10281372070312, "logps/ref_chosen": -86.185546875, "logps/ref_rejected": -88.4615707397461, "logps/rejected": -219.3845672607422, "loss": 0.8567, "margin_dpo/margin_mean": 51.005733489990234, "margin_dpo/margin_std": 50.32048416137695, "step": 355 }, { "epoch": 0.5227606461086637, "fcm_dpo/beta": 0.018100222572684288, "fcm_dpo/delta": 0.17513401806354523, "fcm_dpo/margin": 37.909202575683594, "fcm_dpo/q_t": 0.3618093430995941, "grad_norm": 16.413511276245117, "learning_rate": 2.7562154104130176e-07, "logits/chosen": -1.8032923936843872, "logits/rejected": -1.3581966161727905, "logps/chosen": -185.53103637695312, "logps/ref_chosen": -90.818115234375, "logps/ref_rejected": -82.1273193359375, "logps/rejected": -214.7494354248047, "loss": 1.0311, "margin_dpo/margin_mean": 37.909202575683594, "margin_dpo/margin_std": 56.4036979675293, "step": 356 }, { "epoch": 0.5242290748898678, "fcm_dpo/beta": 0.01725354790687561, "fcm_dpo/delta": -0.25837036967277527, "fcm_dpo/margin": 62.15252685546875, "fcm_dpo/q_t": 0.28101831674575806, "grad_norm": 11.764604568481445, "learning_rate": 2.7434463620546594e-07, "logits/chosen": -2.014652729034424, "logits/rejected": -1.663172960281372, "logps/chosen": -164.86306762695312, "logps/ref_chosen": -87.5047607421875, "logps/ref_rejected": -92.24937438964844, "logps/rejected": -231.76019287109375, "loss": 0.7364, "margin_dpo/margin_mean": 62.15252685546875, "margin_dpo/margin_std": 49.077301025390625, "step": 357 }, { "epoch": 0.5256975036710719, "fcm_dpo/beta": 0.01754681020975113, "fcm_dpo/delta": 0.09242334961891174, "fcm_dpo/margin": 43.57078552246094, "fcm_dpo/q_t": 0.351772665977478, "grad_norm": 13.389106750488281, "learning_rate": 2.730670898658255e-07, "logits/chosen": -1.5166475772857666, "logits/rejected": -1.0220069885253906, "logps/chosen": -184.1082763671875, "logps/ref_chosen": -95.05546569824219, "logps/ref_rejected": -96.47611999511719, "logps/rejected": -229.0997314453125, "loss": 0.946, "margin_dpo/margin_mean": 43.57078552246094, "margin_dpo/margin_std": 55.526466369628906, "step": 358 }, { "epoch": 0.527165932452276, "fcm_dpo/beta": 0.017719190567731857, "fcm_dpo/delta": 0.1193242222070694, "fcm_dpo/margin": 41.656089782714844, "fcm_dpo/q_t": 0.36246833205223083, "grad_norm": 16.13714599609375, "learning_rate": 2.717889356869146e-07, "logits/chosen": -1.5216652154922485, "logits/rejected": -0.9881617426872253, "logps/chosen": -185.498046875, "logps/ref_chosen": -87.37332916259766, "logps/ref_rejected": -88.77188110351562, "logps/rejected": -228.55267333984375, "loss": 1.008, "margin_dpo/margin_mean": 41.65609359741211, "margin_dpo/margin_std": 59.36506652832031, "step": 359 }, { "epoch": 0.5286343612334802, "fcm_dpo/beta": 0.017821190878748894, "fcm_dpo/delta": -0.08467458188533783, "fcm_dpo/margin": 52.00859069824219, "fcm_dpo/q_t": 0.3092251420021057, "grad_norm": 14.783651351928711, "learning_rate": 2.7051020734928443e-07, "logits/chosen": -1.7419157028198242, "logits/rejected": -1.4053552150726318, "logps/chosen": -169.15045166015625, "logps/ref_chosen": -75.22344970703125, "logps/ref_rejected": -76.09400177001953, "logps/rejected": -222.02957153320312, "loss": 0.8292, "margin_dpo/margin_mean": 52.00859069824219, "margin_dpo/margin_std": 51.10163879394531, "step": 360 }, { "epoch": 0.5301027900146843, "fcm_dpo/beta": 0.017817631363868713, "fcm_dpo/delta": 0.017418846487998962, "fcm_dpo/margin": 46.802886962890625, "fcm_dpo/q_t": 0.34032896161079407, "grad_norm": 15.830314636230469, "learning_rate": 2.6923093854861593e-07, "logits/chosen": -1.6299736499786377, "logits/rejected": -1.3665854930877686, "logps/chosen": -176.39230346679688, "logps/ref_chosen": -81.95333099365234, "logps/ref_rejected": -96.67919921875, "logps/rejected": -237.92105102539062, "loss": 0.9434, "margin_dpo/margin_mean": 46.802886962890625, "margin_dpo/margin_std": 60.25688552856445, "step": 361 }, { "epoch": 0.5315712187958884, "fcm_dpo/beta": 0.017633792012929916, "fcm_dpo/delta": -0.048090770840644836, "fcm_dpo/margin": 50.67647933959961, "fcm_dpo/q_t": 0.3179241418838501, "grad_norm": 13.564910888671875, "learning_rate": 2.679511629948319e-07, "logits/chosen": -1.8477286100387573, "logits/rejected": -1.6221222877502441, "logps/chosen": -186.0591278076172, "logps/ref_chosen": -89.33971405029297, "logps/ref_rejected": -110.50196838378906, "logps/rejected": -257.89788818359375, "loss": 0.8685, "margin_dpo/margin_mean": 50.67647933959961, "margin_dpo/margin_std": 53.38725280761719, "step": 362 }, { "epoch": 0.5330396475770925, "fcm_dpo/beta": 0.016984201967716217, "fcm_dpo/delta": -0.2905126214027405, "fcm_dpo/margin": 65.30412292480469, "fcm_dpo/q_t": 0.28682398796081543, "grad_norm": 12.831151008605957, "learning_rate": 2.6667091441120816e-07, "logits/chosen": -1.8855948448181152, "logits/rejected": -1.4528388977050781, "logps/chosen": -172.280517578125, "logps/ref_chosen": -80.77499389648438, "logps/ref_rejected": -79.43267822265625, "logps/rejected": -236.2423095703125, "loss": 0.7683, "margin_dpo/margin_mean": 65.30412292480469, "margin_dpo/margin_std": 62.15265655517578, "step": 363 }, { "epoch": 0.5345080763582967, "fcm_dpo/beta": 0.016796786338090897, "fcm_dpo/delta": -0.002676934003829956, "fcm_dpo/margin": 50.632606506347656, "fcm_dpo/q_t": 0.32234978675842285, "grad_norm": 15.04246997833252, "learning_rate": 2.6539022653348575e-07, "logits/chosen": -1.8682522773742676, "logits/rejected": -1.6991798877716064, "logps/chosen": -169.2406005859375, "logps/ref_chosen": -76.52960205078125, "logps/ref_rejected": -97.56173706054688, "logps/rejected": -240.90536499023438, "loss": 0.8435, "margin_dpo/margin_mean": 50.632606506347656, "margin_dpo/margin_std": 46.46898651123047, "step": 364 }, { "epoch": 0.5359765051395007, "fcm_dpo/beta": 0.016505403444170952, "fcm_dpo/delta": -0.07894276827573776, "fcm_dpo/margin": 55.836181640625, "fcm_dpo/q_t": 0.317300021648407, "grad_norm": 13.384997367858887, "learning_rate": 2.641091331089811e-07, "logits/chosen": -1.6708288192749023, "logits/rejected": -1.5057486295700073, "logps/chosen": -172.15521240234375, "logps/ref_chosen": -80.54489135742188, "logps/ref_rejected": -99.32316589355469, "logps/rejected": -246.76966857910156, "loss": 0.8627, "margin_dpo/margin_mean": 55.836181640625, "margin_dpo/margin_std": 60.313575744628906, "step": 365 }, { "epoch": 0.5374449339207048, "fcm_dpo/beta": 0.01634395495057106, "fcm_dpo/delta": 0.0015336787328124046, "fcm_dpo/margin": 51.919151306152344, "fcm_dpo/q_t": 0.3336575925350189, "grad_norm": 11.990065574645996, "learning_rate": 2.6282766789569736e-07, "logits/chosen": -1.8974790573120117, "logits/rejected": -1.8965535163879395, "logps/chosen": -162.96044921875, "logps/ref_chosen": -72.23806762695312, "logps/ref_rejected": -91.67182922363281, "logps/rejected": -234.3133544921875, "loss": 0.8653, "margin_dpo/margin_mean": 51.919151306152344, "margin_dpo/margin_std": 56.507080078125, "step": 366 }, { "epoch": 0.5389133627019089, "fcm_dpo/beta": 0.016237348318099976, "fcm_dpo/delta": -0.053293175995349884, "fcm_dpo/margin": 55.335411071777344, "fcm_dpo/q_t": 0.31658363342285156, "grad_norm": 12.581396102905273, "learning_rate": 2.615458646614349e-07, "logits/chosen": -1.756191611289978, "logits/rejected": -1.4468399286270142, "logps/chosen": -173.73377990722656, "logps/ref_chosen": -85.42201232910156, "logps/ref_rejected": -83.51779174804688, "logps/rejected": -227.16497802734375, "loss": 0.8241, "margin_dpo/margin_mean": 55.335411071777344, "margin_dpo/margin_std": 53.84605407714844, "step": 367 }, { "epoch": 0.540381791483113, "fcm_dpo/beta": 0.016097472980618477, "fcm_dpo/delta": -0.09254693984985352, "fcm_dpo/margin": 57.9866828918457, "fcm_dpo/q_t": 0.30905455350875854, "grad_norm": 14.102049827575684, "learning_rate": 2.6026375718290083e-07, "logits/chosen": -1.4592063426971436, "logits/rejected": -1.4003169536590576, "logps/chosen": -160.42959594726562, "logps/ref_chosen": -73.84112548828125, "logps/ref_rejected": -104.69528198242188, "logps/rejected": -249.2704315185547, "loss": 0.8385, "margin_dpo/margin_mean": 57.98668670654297, "margin_dpo/margin_std": 56.874576568603516, "step": 368 }, { "epoch": 0.5418502202643172, "fcm_dpo/beta": 0.015790659934282303, "fcm_dpo/delta": -0.06228688731789589, "fcm_dpo/margin": 57.41761779785156, "fcm_dpo/q_t": 0.3094673752784729, "grad_norm": 13.252199172973633, "learning_rate": 2.589813792448196e-07, "logits/chosen": -1.6390339136123657, "logits/rejected": -1.2733685970306396, "logps/chosen": -190.71282958984375, "logps/ref_chosen": -98.46195983886719, "logps/ref_rejected": -99.15672302246094, "logps/rejected": -248.82521057128906, "loss": 0.7867, "margin_dpo/margin_mean": 57.41761779785156, "margin_dpo/margin_std": 47.635658264160156, "step": 369 }, { "epoch": 0.5433186490455213, "fcm_dpo/beta": 0.0161726176738739, "fcm_dpo/delta": 0.18013787269592285, "fcm_dpo/margin": 42.137481689453125, "fcm_dpo/q_t": 0.3603252172470093, "grad_norm": 13.428014755249023, "learning_rate": 2.5769876463904263e-07, "logits/chosen": -1.8658051490783691, "logits/rejected": -1.6577569246292114, "logps/chosen": -198.522705078125, "logps/ref_chosen": -96.77574157714844, "logps/ref_rejected": -102.74411010742188, "logps/rejected": -246.62857055664062, "loss": 1.002, "margin_dpo/margin_mean": 42.137481689453125, "margin_dpo/margin_std": 57.25678253173828, "step": 370 }, { "epoch": 0.5447870778267254, "fcm_dpo/beta": 0.01677476242184639, "fcm_dpo/delta": 0.16446369886398315, "fcm_dpo/margin": 41.4566535949707, "fcm_dpo/q_t": 0.3613440692424774, "grad_norm": 15.086153030395508, "learning_rate": 2.5641594716365744e-07, "logits/chosen": -1.7068711519241333, "logits/rejected": -1.3464912176132202, "logps/chosen": -201.3501434326172, "logps/ref_chosen": -102.99020385742188, "logps/ref_rejected": -105.35062408447266, "logps/rejected": -245.167236328125, "loss": 0.9969, "margin_dpo/margin_mean": 41.45665740966797, "margin_dpo/margin_std": 56.3448600769043, "step": 371 }, { "epoch": 0.5462555066079295, "fcm_dpo/beta": 0.01618875563144684, "fcm_dpo/delta": -0.26988279819488525, "fcm_dpo/margin": 67.3760986328125, "fcm_dpo/q_t": 0.28218111395835876, "grad_norm": 11.737579345703125, "learning_rate": 2.551329606220976e-07, "logits/chosen": -1.7287421226501465, "logits/rejected": -1.1179594993591309, "logps/chosen": -174.77841186523438, "logps/ref_chosen": -91.02623748779297, "logps/ref_rejected": -85.6102294921875, "logps/rejected": -236.738525390625, "loss": 0.7396, "margin_dpo/margin_mean": 67.37609100341797, "margin_dpo/margin_std": 57.7528076171875, "step": 372 }, { "epoch": 0.5477239353891337, "fcm_dpo/beta": 0.015897128731012344, "fcm_dpo/delta": -0.0048202937468886375, "fcm_dpo/margin": 53.741973876953125, "fcm_dpo/q_t": 0.32611092925071716, "grad_norm": 13.899093627929688, "learning_rate": 2.538498388222517e-07, "logits/chosen": -2.0247459411621094, "logits/rejected": -1.520031213760376, "logps/chosen": -195.14593505859375, "logps/ref_chosen": -100.07307434082031, "logps/ref_rejected": -91.11634826660156, "logps/rejected": -239.93118286132812, "loss": 0.8742, "margin_dpo/margin_mean": 53.741973876953125, "margin_dpo/margin_std": 56.848350524902344, "step": 373 }, { "epoch": 0.5491923641703378, "fcm_dpo/beta": 0.015747731551527977, "fcm_dpo/delta": -0.030241642147302628, "fcm_dpo/margin": 55.66587829589844, "fcm_dpo/q_t": 0.31644949316978455, "grad_norm": 11.643533706665039, "learning_rate": 2.525666155755725e-07, "logits/chosen": -2.1849594116210938, "logits/rejected": -1.7997386455535889, "logps/chosen": -183.28863525390625, "logps/ref_chosen": -98.01234436035156, "logps/ref_rejected": -100.75224304199219, "logps/rejected": -241.69439697265625, "loss": 0.8144, "margin_dpo/margin_mean": 55.66587829589844, "margin_dpo/margin_std": 48.722808837890625, "step": 374 }, { "epoch": 0.5506607929515418, "fcm_dpo/beta": 0.0154128298163414, "fcm_dpo/delta": -0.0929117500782013, "fcm_dpo/margin": 60.347686767578125, "fcm_dpo/q_t": 0.3120383322238922, "grad_norm": 12.819354057312012, "learning_rate": 2.512833246961859e-07, "logits/chosen": -1.465893268585205, "logits/rejected": -1.2448041439056396, "logps/chosen": -173.93695068359375, "logps/ref_chosen": -84.39627838134766, "logps/ref_rejected": -95.7188491821289, "logps/rejected": -245.60720825195312, "loss": 0.8251, "margin_dpo/margin_mean": 60.347686767578125, "margin_dpo/margin_std": 57.37383270263672, "step": 375 }, { "epoch": 0.5521292217327459, "fcm_dpo/beta": 0.015097381547093391, "fcm_dpo/delta": -0.16518311202526093, "fcm_dpo/margin": 66.08840942382812, "fcm_dpo/q_t": 0.29718315601348877, "grad_norm": 12.352048873901367, "learning_rate": 2.5e-07, "logits/chosen": -1.7917392253875732, "logits/rejected": -1.50510835647583, "logps/chosen": -180.96434020996094, "logps/ref_chosen": -95.73770141601562, "logps/ref_rejected": -111.97331237792969, "logps/rejected": -263.2883605957031, "loss": 0.7832, "margin_dpo/margin_mean": 66.08840942382812, "margin_dpo/margin_std": 59.88749313354492, "step": 376 }, { "epoch": 0.55359765051395, "fcm_dpo/beta": 0.01455208845436573, "fcm_dpo/delta": -0.15709523856639862, "fcm_dpo/margin": 67.94343566894531, "fcm_dpo/q_t": 0.3027934730052948, "grad_norm": 13.825796127319336, "learning_rate": 2.487166753038141e-07, "logits/chosen": -1.534698724746704, "logits/rejected": -1.4000335931777954, "logps/chosen": -172.19622802734375, "logps/ref_chosen": -78.92694854736328, "logps/ref_rejected": -105.89892578125, "logps/rejected": -267.11163330078125, "loss": 0.8043, "margin_dpo/margin_mean": 67.94343566894531, "margin_dpo/margin_std": 64.65965270996094, "step": 377 }, { "epoch": 0.5550660792951542, "fcm_dpo/beta": 0.014748094603419304, "fcm_dpo/delta": 0.0158497616648674, "fcm_dpo/margin": 56.504417419433594, "fcm_dpo/q_t": 0.32408106327056885, "grad_norm": 12.6776704788208, "learning_rate": 2.4743338442442754e-07, "logits/chosen": -1.9457380771636963, "logits/rejected": -1.6649649143218994, "logps/chosen": -166.82522583007812, "logps/ref_chosen": -75.37240600585938, "logps/ref_rejected": -93.70409393310547, "logps/rejected": -241.6613311767578, "loss": 0.8687, "margin_dpo/margin_mean": 56.50442123413086, "margin_dpo/margin_std": 55.60420227050781, "step": 378 }, { "epoch": 0.5565345080763583, "fcm_dpo/beta": 0.01469915546476841, "fcm_dpo/delta": 0.051874738186597824, "fcm_dpo/margin": 54.58234786987305, "fcm_dpo/q_t": 0.3433055281639099, "grad_norm": 12.682414054870605, "learning_rate": 2.461501611777483e-07, "logits/chosen": -1.7052133083343506, "logits/rejected": -1.5816318988800049, "logps/chosen": -190.00857543945312, "logps/ref_chosen": -88.85568237304688, "logps/ref_rejected": -121.38114166259766, "logps/rejected": -277.11639404296875, "loss": 0.9467, "margin_dpo/margin_mean": 54.58234786987305, "margin_dpo/margin_std": 68.8104019165039, "step": 379 }, { "epoch": 0.5580029368575624, "fcm_dpo/beta": 0.014686446636915207, "fcm_dpo/delta": -0.03623585402965546, "fcm_dpo/margin": 60.12630081176758, "fcm_dpo/q_t": 0.3278425931930542, "grad_norm": 14.065608024597168, "learning_rate": 2.4486703937790243e-07, "logits/chosen": -1.41129732131958, "logits/rejected": -1.3641774654388428, "logps/chosen": -190.62673950195312, "logps/ref_chosen": -79.83321380615234, "logps/ref_rejected": -110.7447509765625, "logps/rejected": -281.6645812988281, "loss": 0.8771, "margin_dpo/margin_mean": 60.12629699707031, "margin_dpo/margin_std": 68.94208526611328, "step": 380 }, { "epoch": 0.5594713656387665, "fcm_dpo/beta": 0.014767487533390522, "fcm_dpo/delta": 0.05422665923833847, "fcm_dpo/margin": 54.18187713623047, "fcm_dpo/q_t": 0.3340100944042206, "grad_norm": 14.278923034667969, "learning_rate": 2.435840528363426e-07, "logits/chosen": -1.7580182552337646, "logits/rejected": -1.2122209072113037, "logps/chosen": -183.7044677734375, "logps/ref_chosen": -86.56683349609375, "logps/ref_rejected": -84.416748046875, "logps/rejected": -235.73626708984375, "loss": 0.887, "margin_dpo/margin_mean": 54.18187713623047, "margin_dpo/margin_std": 57.312503814697266, "step": 381 }, { "epoch": 0.5609397944199707, "fcm_dpo/beta": 0.01462157815694809, "fcm_dpo/delta": -0.017210952937602997, "fcm_dpo/margin": 59.04779052734375, "fcm_dpo/q_t": 0.3251911401748657, "grad_norm": 12.774848937988281, "learning_rate": 2.4230123536095745e-07, "logits/chosen": -1.765809178352356, "logits/rejected": -1.4844367504119873, "logps/chosen": -201.82684326171875, "logps/ref_chosen": -104.46180725097656, "logps/ref_rejected": -114.60223388671875, "logps/rejected": -271.0150451660156, "loss": 0.8592, "margin_dpo/margin_mean": 59.047786712646484, "margin_dpo/margin_std": 59.06350326538086, "step": 382 }, { "epoch": 0.5624082232011748, "fcm_dpo/beta": 0.014921590685844421, "fcm_dpo/delta": 0.09733906388282776, "fcm_dpo/margin": 50.940895080566406, "fcm_dpo/q_t": 0.34485840797424316, "grad_norm": 13.45592212677002, "learning_rate": 2.4101862075518037e-07, "logits/chosen": -1.6310830116271973, "logits/rejected": -1.5183734893798828, "logps/chosen": -186.19158935546875, "logps/ref_chosen": -82.11607360839844, "logps/ref_rejected": -98.77716064453125, "logps/rejected": -253.79356384277344, "loss": 0.9232, "margin_dpo/margin_mean": 50.940895080566406, "margin_dpo/margin_std": 57.292198181152344, "step": 383 }, { "epoch": 0.5638766519823789, "fcm_dpo/beta": 0.01533111184835434, "fcm_dpo/delta": 0.04174236208200455, "fcm_dpo/margin": 52.73126983642578, "fcm_dpo/q_t": 0.33828550577163696, "grad_norm": 15.4530029296875, "learning_rate": 2.397362428170992e-07, "logits/chosen": -1.7482203245162964, "logits/rejected": -1.5194519758224487, "logps/chosen": -191.6592254638672, "logps/ref_chosen": -80.58525085449219, "logps/ref_rejected": -91.44789123535156, "logps/rejected": -255.2531280517578, "loss": 0.9616, "margin_dpo/margin_mean": 52.73126983642578, "margin_dpo/margin_std": 66.3970718383789, "step": 384 }, { "epoch": 0.5653450807635829, "fcm_dpo/beta": 0.014804725535213947, "fcm_dpo/delta": -0.0903623104095459, "fcm_dpo/margin": 62.7065544128418, "fcm_dpo/q_t": 0.31812626123428345, "grad_norm": 13.666119575500488, "learning_rate": 2.3845413533856514e-07, "logits/chosen": -2.33363676071167, "logits/rejected": -1.75101637840271, "logps/chosen": -186.3187255859375, "logps/ref_chosen": -92.76016235351562, "logps/ref_rejected": -84.45997619628906, "logps/rejected": -240.72511291503906, "loss": 0.858, "margin_dpo/margin_mean": 62.70655822753906, "margin_dpo/margin_std": 65.78327941894531, "step": 385 }, { "epoch": 0.566813509544787, "fcm_dpo/beta": 0.014840572141110897, "fcm_dpo/delta": -0.0528637170791626, "fcm_dpo/margin": 60.51496505737305, "fcm_dpo/q_t": 0.3133012652397156, "grad_norm": 11.866787910461426, "learning_rate": 2.3717233210430254e-07, "logits/chosen": -1.5037336349487305, "logits/rejected": -1.2165539264678955, "logps/chosen": -182.65261840820312, "logps/ref_chosen": -85.35896301269531, "logps/ref_rejected": -96.99065399169922, "logps/rejected": -254.79925537109375, "loss": 0.8583, "margin_dpo/margin_mean": 60.51496124267578, "margin_dpo/margin_std": 61.9644775390625, "step": 386 }, { "epoch": 0.5682819383259912, "fcm_dpo/beta": 0.014781072735786438, "fcm_dpo/delta": -0.010366106405854225, "fcm_dpo/margin": 58.13372039794922, "fcm_dpo/q_t": 0.3218954801559448, "grad_norm": 12.85347843170166, "learning_rate": 2.3589086689101889e-07, "logits/chosen": -2.2964346408843994, "logits/rejected": -1.622833490371704, "logps/chosen": -212.4945068359375, "logps/ref_chosen": -102.68814086914062, "logps/ref_rejected": -99.42858123779297, "logps/rejected": -267.3686828613281, "loss": 0.8492, "margin_dpo/margin_mean": 58.13372039794922, "margin_dpo/margin_std": 56.552650451660156, "step": 387 }, { "epoch": 0.5697503671071953, "fcm_dpo/beta": 0.01442895270884037, "fcm_dpo/delta": -0.12326682358980179, "fcm_dpo/margin": 66.6114501953125, "fcm_dpo/q_t": 0.3023766279220581, "grad_norm": 12.544069290161133, "learning_rate": 2.3460977346651428e-07, "logits/chosen": -1.5236001014709473, "logits/rejected": -1.368851900100708, "logps/chosen": -191.8466796875, "logps/ref_chosen": -83.58523559570312, "logps/ref_rejected": -110.10037231445312, "logps/rejected": -284.9732666015625, "loss": 0.7869, "margin_dpo/margin_mean": 66.6114501953125, "margin_dpo/margin_std": 58.575645446777344, "step": 388 }, { "epoch": 0.5712187958883994, "fcm_dpo/beta": 0.014386516064405441, "fcm_dpo/delta": 0.0063179126009345055, "fcm_dpo/margin": 58.680145263671875, "fcm_dpo/q_t": 0.33595648407936096, "grad_norm": 13.45885944366455, "learning_rate": 2.3332908558879177e-07, "logits/chosen": -2.3460850715637207, "logits/rejected": -1.9088528156280518, "logps/chosen": -203.8497314453125, "logps/ref_chosen": -89.822265625, "logps/ref_rejected": -84.85376739501953, "logps/rejected": -257.5613708496094, "loss": 0.9254, "margin_dpo/margin_mean": 58.680145263671875, "margin_dpo/margin_std": 72.35084533691406, "step": 389 }, { "epoch": 0.5726872246696035, "fcm_dpo/beta": 0.014441296458244324, "fcm_dpo/delta": 0.09785392880439758, "fcm_dpo/margin": 52.516563415527344, "fcm_dpo/q_t": 0.3560144901275635, "grad_norm": 13.203153610229492, "learning_rate": 2.320488370051681e-07, "logits/chosen": -1.580621361732483, "logits/rejected": -1.3598695993423462, "logps/chosen": -193.17552185058594, "logps/ref_chosen": -77.53244018554688, "logps/ref_rejected": -92.58392333984375, "logps/rejected": -260.7435607910156, "loss": 0.9781, "margin_dpo/margin_mean": 52.51655960083008, "margin_dpo/margin_std": 71.42833709716797, "step": 390 }, { "epoch": 0.5741556534508077, "fcm_dpo/beta": 0.01456206664443016, "fcm_dpo/delta": -0.014334239065647125, "fcm_dpo/margin": 59.23310852050781, "fcm_dpo/q_t": 0.3304310441017151, "grad_norm": 13.462164878845215, "learning_rate": 2.3076906145138405e-07, "logits/chosen": -1.8457610607147217, "logits/rejected": -1.3921658992767334, "logps/chosen": -193.5087890625, "logps/ref_chosen": -88.2329330444336, "logps/ref_rejected": -93.00756072998047, "logps/rejected": -257.51654052734375, "loss": 0.8957, "margin_dpo/margin_mean": 59.23310852050781, "margin_dpo/margin_std": 68.13093566894531, "step": 391 }, { "epoch": 0.5756240822320118, "fcm_dpo/beta": 0.014434915035963058, "fcm_dpo/delta": -0.12481804192066193, "fcm_dpo/margin": 66.70355224609375, "fcm_dpo/q_t": 0.30818355083465576, "grad_norm": 14.184881210327148, "learning_rate": 2.294897926507156e-07, "logits/chosen": -1.6148556470870972, "logits/rejected": -1.4798262119293213, "logps/chosen": -193.44656372070312, "logps/ref_chosen": -87.74442291259766, "logps/ref_rejected": -108.57071685791016, "logps/rejected": -280.9764404296875, "loss": 0.8476, "margin_dpo/margin_mean": 66.70355224609375, "margin_dpo/margin_std": 70.49634552001953, "step": 392 }, { "epoch": 0.5770925110132159, "fcm_dpo/beta": 0.014169027097523212, "fcm_dpo/delta": -0.032773129642009735, "fcm_dpo/margin": 62.09144592285156, "fcm_dpo/q_t": 0.32115113735198975, "grad_norm": 11.957456588745117, "learning_rate": 2.2821106431308543e-07, "logits/chosen": -1.8445684909820557, "logits/rejected": -1.5407391786575317, "logps/chosen": -170.1377410888672, "logps/ref_chosen": -76.56381225585938, "logps/ref_rejected": -79.72618103027344, "logps/rejected": -235.39157104492188, "loss": 0.8532, "margin_dpo/margin_mean": 62.09144592285156, "margin_dpo/margin_std": 63.155155181884766, "step": 393 }, { "epoch": 0.57856093979442, "fcm_dpo/beta": 0.013979214243590832, "fcm_dpo/delta": -0.07188640534877777, "fcm_dpo/margin": 65.45079040527344, "fcm_dpo/q_t": 0.3089185357093811, "grad_norm": 14.39035701751709, "learning_rate": 2.2693291013417452e-07, "logits/chosen": -1.5587331056594849, "logits/rejected": -1.3254432678222656, "logps/chosen": -182.34947204589844, "logps/ref_chosen": -84.69866180419922, "logps/ref_rejected": -97.457763671875, "logps/rejected": -260.55938720703125, "loss": 0.802, "margin_dpo/margin_mean": 65.45079040527344, "margin_dpo/margin_std": 57.37220764160156, "step": 394 }, { "epoch": 0.580029368575624, "fcm_dpo/beta": 0.014071676880121231, "fcm_dpo/delta": 0.0893394872546196, "fcm_dpo/margin": 54.5306282043457, "fcm_dpo/q_t": 0.3583162724971771, "grad_norm": 14.767304420471191, "learning_rate": 2.2565536379453404e-07, "logits/chosen": -2.528858184814453, "logits/rejected": -2.1432571411132812, "logps/chosen": -189.65113830566406, "logps/ref_chosen": -88.88763427734375, "logps/ref_rejected": -91.0455322265625, "logps/rejected": -246.33966064453125, "loss": 1.0618, "margin_dpo/margin_mean": 54.53062438964844, "margin_dpo/margin_std": 86.91376495361328, "step": 395 }, { "epoch": 0.5814977973568282, "fcm_dpo/beta": 0.014645865187048912, "fcm_dpo/delta": 0.19641628861427307, "fcm_dpo/margin": 45.51504135131836, "fcm_dpo/q_t": 0.3645893633365631, "grad_norm": 17.304218292236328, "learning_rate": 2.2437845895869825e-07, "logits/chosen": -1.776867151260376, "logits/rejected": -1.2883884906768799, "logps/chosen": -211.24041748046875, "logps/ref_chosen": -102.47132873535156, "logps/ref_rejected": -93.23478698730469, "logps/rejected": -247.51890563964844, "loss": 1.0063, "margin_dpo/margin_mean": 45.515037536621094, "margin_dpo/margin_std": 62.76597595214844, "step": 396 }, { "epoch": 0.5829662261380323, "fcm_dpo/beta": 0.013984641060233116, "fcm_dpo/delta": -0.33509910106658936, "fcm_dpo/margin": 81.79728698730469, "fcm_dpo/q_t": 0.2831918001174927, "grad_norm": 13.865079879760742, "learning_rate": 2.2310222927429716e-07, "logits/chosen": -2.201831817626953, "logits/rejected": -1.599250078201294, "logps/chosen": -198.48558044433594, "logps/ref_chosen": -101.04441833496094, "logps/ref_rejected": -108.8511734008789, "logps/rejected": -288.0896301269531, "loss": 0.7855, "margin_dpo/margin_mean": 81.79728698730469, "margin_dpo/margin_std": 80.89474487304688, "step": 397 }, { "epoch": 0.5844346549192364, "fcm_dpo/beta": 0.013811508193612099, "fcm_dpo/delta": 0.005452306941151619, "fcm_dpo/margin": 61.17559814453125, "fcm_dpo/q_t": 0.3352923095226288, "grad_norm": 14.45501708984375, "learning_rate": 2.2182670837116972e-07, "logits/chosen": -1.9161015748977661, "logits/rejected": -1.6111736297607422, "logps/chosen": -197.73312377929688, "logps/ref_chosen": -95.04922485351562, "logps/ref_rejected": -114.00563049316406, "logps/rejected": -277.8651123046875, "loss": 0.9266, "margin_dpo/margin_mean": 61.175594329833984, "margin_dpo/margin_std": 75.55554962158203, "step": 398 }, { "epoch": 0.5859030837004405, "fcm_dpo/beta": 0.013482306152582169, "fcm_dpo/delta": -0.218872532248497, "fcm_dpo/margin": 77.61836242675781, "fcm_dpo/q_t": 0.284706175327301, "grad_norm": 12.366701126098633, "learning_rate": 2.2055192986047804e-07, "logits/chosen": -2.152554512023926, "logits/rejected": -1.3891927003860474, "logps/chosen": -189.8502960205078, "logps/ref_chosen": -96.63854217529297, "logps/ref_rejected": -83.53645324707031, "logps/rejected": -254.3665771484375, "loss": 0.7319, "margin_dpo/margin_mean": 77.61836242675781, "margin_dpo/margin_std": 61.290470123291016, "step": 399 }, { "epoch": 0.5873715124816447, "fcm_dpo/beta": 0.012850401923060417, "fcm_dpo/delta": -0.1984182745218277, "fcm_dpo/margin": 79.97297668457031, "fcm_dpo/q_t": 0.29666027426719666, "grad_norm": 14.71058464050293, "learning_rate": 2.192779273338215e-07, "logits/chosen": -2.1020350456237793, "logits/rejected": -1.615342140197754, "logps/chosen": -198.56103515625, "logps/ref_chosen": -105.30314636230469, "logps/ref_rejected": -108.6555404663086, "logps/rejected": -281.88641357421875, "loss": 0.8025, "margin_dpo/margin_mean": 79.97298431396484, "margin_dpo/margin_std": 77.14982604980469, "step": 400 }, { "epoch": 0.5873715124816447, "eval_fcm_dpo/beta": 0.012737083248794079, "eval_logits/chosen": -1.7851052284240723, "eval_logits/rejected": -1.422616958618164, "eval_logps/chosen": -216.16845703125, "eval_logps/ref_chosen": -100.49356842041016, "eval_logps/ref_rejected": -94.06775665283203, "eval_logps/rejected": -261.1965026855469, "eval_loss": 0.49569863080978394, "eval_margin_dpo/margin_mean": 51.453895568847656, "eval_margin_dpo/margin_std": 65.75733184814453, "eval_runtime": 43.747, "eval_samples_per_second": 53.466, "eval_steps_per_second": 1.692, "step": 400 }, { "epoch": 0.5888399412628488, "fcm_dpo/beta": 0.013059141114354134, "fcm_dpo/delta": 0.20135712623596191, "fcm_dpo/margin": 50.72594451904297, "fcm_dpo/q_t": 0.3744737505912781, "grad_norm": 15.875528335571289, "learning_rate": 2.1800473436235136e-07, "logits/chosen": -1.5951911211013794, "logits/rejected": -1.3578665256500244, "logps/chosen": -190.99041748046875, "logps/ref_chosen": -84.28916931152344, "logps/ref_rejected": -90.943115234375, "logps/rejected": -248.3703155517578, "loss": 1.0408, "margin_dpo/margin_mean": 50.72594451904297, "margin_dpo/margin_std": 78.53496551513672, "step": 401 }, { "epoch": 0.5903083700440529, "fcm_dpo/beta": 0.012583300471305847, "fcm_dpo/delta": -0.35893315076828003, "fcm_dpo/margin": 92.74884033203125, "fcm_dpo/q_t": 0.2584311366081238, "grad_norm": 12.396269798278809, "learning_rate": 2.1673238449588665e-07, "logits/chosen": -2.003387451171875, "logits/rejected": -1.5525648593902588, "logps/chosen": -179.36959838867188, "logps/ref_chosen": -83.59312438964844, "logps/ref_rejected": -87.81027221679688, "logps/rejected": -276.3355712890625, "loss": 0.6644, "margin_dpo/margin_mean": 92.74884033203125, "margin_dpo/margin_std": 64.43993377685547, "step": 402 }, { "epoch": 0.591776798825257, "fcm_dpo/beta": 0.012310395017266273, "fcm_dpo/delta": 0.009639285504817963, "fcm_dpo/margin": 68.3082046508789, "fcm_dpo/q_t": 0.3275585174560547, "grad_norm": 11.257796287536621, "learning_rate": 2.154609112620295e-07, "logits/chosen": -2.2196717262268066, "logits/rejected": -1.8389365673065186, "logps/chosen": -173.17076110839844, "logps/ref_chosen": -73.75308227539062, "logps/ref_rejected": -83.92012786865234, "logps/rejected": -251.64601135253906, "loss": 0.8697, "margin_dpo/margin_mean": 68.3082046508789, "margin_dpo/margin_std": 70.12299346923828, "step": 403 }, { "epoch": 0.593245227606461, "fcm_dpo/beta": 0.012473889626562595, "fcm_dpo/delta": 0.08444693684577942, "fcm_dpo/margin": 61.89647674560547, "fcm_dpo/q_t": 0.34659212827682495, "grad_norm": 16.581104278564453, "learning_rate": 2.1419034816528218e-07, "logits/chosen": -2.2287228107452393, "logits/rejected": -1.9110980033874512, "logps/chosen": -196.9864044189453, "logps/ref_chosen": -79.67617797851562, "logps/ref_rejected": -84.280517578125, "logps/rejected": -263.4872131347656, "loss": 0.9672, "margin_dpo/margin_mean": 61.89647674560547, "margin_dpo/margin_std": 81.55596923828125, "step": 404 }, { "epoch": 0.5947136563876652, "fcm_dpo/beta": 0.012832706794142723, "fcm_dpo/delta": 0.08121472597122192, "fcm_dpo/margin": 60.25043487548828, "fcm_dpo/q_t": 0.3399437665939331, "grad_norm": 11.338379859924316, "learning_rate": 2.129207286861638e-07, "logits/chosen": -1.6208863258361816, "logits/rejected": -1.1806436777114868, "logps/chosen": -217.16705322265625, "logps/ref_chosen": -96.46195220947266, "logps/ref_rejected": -92.87071228027344, "logps/rejected": -273.82623291015625, "loss": 0.8836, "margin_dpo/margin_mean": 60.25043487548828, "margin_dpo/margin_std": 60.95580291748047, "step": 405 }, { "epoch": 0.5961820851688693, "fcm_dpo/beta": 0.01305477600544691, "fcm_dpo/delta": 0.13485847413539886, "fcm_dpo/margin": 55.51643371582031, "fcm_dpo/q_t": 0.3540048599243164, "grad_norm": 16.996318817138672, "learning_rate": 2.1165208628032861e-07, "logits/chosen": -1.942261815071106, "logits/rejected": -1.7632594108581543, "logps/chosen": -192.12899780273438, "logps/ref_chosen": -78.13396453857422, "logps/ref_rejected": -98.28359985351562, "logps/rejected": -267.7950744628906, "loss": 0.9845, "margin_dpo/margin_mean": 55.51643371582031, "margin_dpo/margin_std": 73.61194610595703, "step": 406 }, { "epoch": 0.5976505139500734, "fcm_dpo/beta": 0.013465587049722672, "fcm_dpo/delta": 0.17369112372398376, "fcm_dpo/margin": 51.12046432495117, "fcm_dpo/q_t": 0.36146247386932373, "grad_norm": 15.288115501403809, "learning_rate": 2.1038445437768375e-07, "logits/chosen": -2.192194700241089, "logits/rejected": -1.68497633934021, "logps/chosen": -205.4015350341797, "logps/ref_chosen": -84.01283264160156, "logps/ref_rejected": -82.78103637695312, "logps/rejected": -255.29019165039062, "loss": 0.9881, "margin_dpo/margin_mean": 51.120460510253906, "margin_dpo/margin_std": 69.12421417236328, "step": 407 }, { "epoch": 0.5991189427312775, "fcm_dpo/beta": 0.013818442821502686, "fcm_dpo/delta": 0.13276906311511993, "fcm_dpo/margin": 52.61656188964844, "fcm_dpo/q_t": 0.3611217141151428, "grad_norm": 15.027425765991211, "learning_rate": 2.0911786638150872e-07, "logits/chosen": -2.2886483669281006, "logits/rejected": -1.7039839029312134, "logps/chosen": -217.49435424804688, "logps/ref_chosen": -104.46174621582031, "logps/ref_rejected": -96.37217712402344, "logps/rejected": -262.0213317871094, "loss": 1.0044, "margin_dpo/margin_mean": 52.61656188964844, "margin_dpo/margin_std": 76.11251831054688, "step": 408 }, { "epoch": 0.6005873715124816, "fcm_dpo/beta": 0.014177883975207806, "fcm_dpo/delta": 0.11815360188484192, "fcm_dpo/margin": 52.24652862548828, "fcm_dpo/q_t": 0.34808868169784546, "grad_norm": 14.006386756896973, "learning_rate": 2.0785235566757517e-07, "logits/chosen": -1.5552918910980225, "logits/rejected": -1.1278228759765625, "logps/chosen": -222.04092407226562, "logps/ref_chosen": -97.66830444335938, "logps/ref_rejected": -90.04583740234375, "logps/rejected": -266.66497802734375, "loss": 1.0119, "margin_dpo/margin_mean": 52.24652862548828, "margin_dpo/margin_std": 73.68948364257812, "step": 409 }, { "epoch": 0.6020558002936858, "fcm_dpo/beta": 0.014412937685847282, "fcm_dpo/delta": 0.04549999535083771, "fcm_dpo/margin": 56.066070556640625, "fcm_dpo/q_t": 0.33421647548675537, "grad_norm": 14.076881408691406, "learning_rate": 2.065879555832674e-07, "logits/chosen": -1.8263776302337646, "logits/rejected": -1.6934200525283813, "logps/chosen": -198.36102294921875, "logps/ref_chosen": -76.46923828125, "logps/ref_rejected": -88.64064025878906, "logps/rejected": -266.5985107421875, "loss": 0.9166, "margin_dpo/margin_mean": 56.066070556640625, "margin_dpo/margin_std": 65.75325775146484, "step": 410 }, { "epoch": 0.6035242290748899, "fcm_dpo/beta": 0.014379952102899551, "fcm_dpo/delta": 0.015824299305677414, "fcm_dpo/margin": 58.06561279296875, "fcm_dpo/q_t": 0.3332013487815857, "grad_norm": 13.497848510742188, "learning_rate": 2.0532469944670343e-07, "logits/chosen": -2.267174243927002, "logits/rejected": -1.8730146884918213, "logps/chosen": -215.80242919921875, "logps/ref_chosen": -87.16630554199219, "logps/ref_rejected": -87.09603118896484, "logps/rejected": -273.79779052734375, "loss": 0.884, "margin_dpo/margin_mean": 58.06561279296875, "margin_dpo/margin_std": 62.760345458984375, "step": 411 }, { "epoch": 0.604992657856094, "fcm_dpo/beta": 0.014468617737293243, "fcm_dpo/delta": 0.05017668381333351, "fcm_dpo/margin": 55.50398635864258, "fcm_dpo/q_t": 0.34975796937942505, "grad_norm": 15.654382705688477, "learning_rate": 2.0406262054585738e-07, "logits/chosen": -2.643434524536133, "logits/rejected": -2.4237518310546875, "logps/chosen": -199.4673309326172, "logps/ref_chosen": -78.94734191894531, "logps/ref_rejected": -106.10554504394531, "logps/rejected": -282.1295166015625, "loss": 0.9861, "margin_dpo/margin_mean": 55.503990173339844, "margin_dpo/margin_std": 77.02133178710938, "step": 412 }, { "epoch": 0.6064610866372981, "fcm_dpo/beta": 0.014438673853874207, "fcm_dpo/delta": -0.0511392205953598, "fcm_dpo/margin": 62.048255920410156, "fcm_dpo/q_t": 0.3228393793106079, "grad_norm": 14.533951759338379, "learning_rate": 2.0280175213768205e-07, "logits/chosen": -1.3055384159088135, "logits/rejected": -1.0555856227874756, "logps/chosen": -225.90713500976562, "logps/ref_chosen": -95.69470977783203, "logps/ref_rejected": -107.96085357666016, "logps/rejected": -300.2215270996094, "loss": 0.9045, "margin_dpo/margin_mean": 62.048255920410156, "margin_dpo/margin_std": 75.0443115234375, "step": 413 }, { "epoch": 0.6079295154185022, "fcm_dpo/beta": 0.01450310554355383, "fcm_dpo/delta": 0.014196475967764854, "fcm_dpo/margin": 57.7109260559082, "fcm_dpo/q_t": 0.34177166223526, "grad_norm": 19.222719192504883, "learning_rate": 2.0154212744723247e-07, "logits/chosen": -1.5118329524993896, "logits/rejected": -1.1614055633544922, "logps/chosen": -208.80809020996094, "logps/ref_chosen": -88.27667236328125, "logps/ref_rejected": -92.87004089355469, "logps/rejected": -271.11236572265625, "loss": 0.9818, "margin_dpo/margin_mean": 57.71092224121094, "margin_dpo/margin_std": 80.63750457763672, "step": 414 }, { "epoch": 0.6093979441997063, "fcm_dpo/beta": 0.014715914614498615, "fcm_dpo/delta": 0.07649455219507217, "fcm_dpo/margin": 52.944557189941406, "fcm_dpo/q_t": 0.3491487503051758, "grad_norm": 15.72915267944336, "learning_rate": 2.002837796667909e-07, "logits/chosen": -1.5652879476547241, "logits/rejected": -1.1960127353668213, "logps/chosen": -235.4866180419922, "logps/ref_chosen": -108.91590118408203, "logps/ref_rejected": -107.47135925292969, "logps/rejected": -286.98663330078125, "loss": 0.9785, "margin_dpo/margin_mean": 52.944557189941406, "margin_dpo/margin_std": 71.51063537597656, "step": 415 }, { "epoch": 0.6108663729809104, "fcm_dpo/beta": 0.014534441754221916, "fcm_dpo/delta": -0.19304054975509644, "fcm_dpo/margin": 70.30528259277344, "fcm_dpo/q_t": 0.3036141097545624, "grad_norm": 13.811197280883789, "learning_rate": 1.990267419549914e-07, "logits/chosen": -1.9736833572387695, "logits/rejected": -1.4895861148834229, "logps/chosen": -219.60324096679688, "logps/ref_chosen": -93.39888000488281, "logps/ref_rejected": -97.6729736328125, "logps/rejected": -294.1826171875, "loss": 0.8409, "margin_dpo/margin_mean": 70.30529022216797, "margin_dpo/margin_std": 73.31067657470703, "step": 416 }, { "epoch": 0.6123348017621145, "fcm_dpo/beta": 0.01407422311604023, "fcm_dpo/delta": -0.07601834833621979, "fcm_dpo/margin": 65.29705810546875, "fcm_dpo/q_t": 0.3138529360294342, "grad_norm": 13.441536903381348, "learning_rate": 1.9777104743594686e-07, "logits/chosen": -2.2062695026397705, "logits/rejected": -1.3112609386444092, "logps/chosen": -203.29037475585938, "logps/ref_chosen": -83.53533172607422, "logps/ref_rejected": -74.44184875488281, "logps/rejected": -259.49395751953125, "loss": 0.8486, "margin_dpo/margin_mean": 65.29705810546875, "margin_dpo/margin_std": 66.7147216796875, "step": 417 }, { "epoch": 0.6138032305433186, "fcm_dpo/beta": 0.013649359345436096, "fcm_dpo/delta": -0.2670527398586273, "fcm_dpo/margin": 79.63304138183594, "fcm_dpo/q_t": 0.2834809422492981, "grad_norm": 15.17289924621582, "learning_rate": 1.965167291983757e-07, "logits/chosen": -2.136475086212158, "logits/rejected": -1.491228461265564, "logps/chosen": -221.35855102539062, "logps/ref_chosen": -108.22152709960938, "logps/ref_rejected": -111.8646469116211, "logps/rejected": -304.63470458984375, "loss": 0.7466, "margin_dpo/margin_mean": 79.63304138183594, "margin_dpo/margin_std": 66.44700622558594, "step": 418 }, { "epoch": 0.6152716593245228, "fcm_dpo/beta": 0.013165738433599472, "fcm_dpo/delta": -0.06787989288568497, "fcm_dpo/margin": 69.23770141601562, "fcm_dpo/q_t": 0.3166205883026123, "grad_norm": 12.66903305053711, "learning_rate": 1.9526382029472988e-07, "logits/chosen": -1.1017417907714844, "logits/rejected": -0.6296178698539734, "logps/chosen": -218.44863891601562, "logps/ref_chosen": -97.18328094482422, "logps/ref_rejected": -98.18531799316406, "logps/rejected": -288.6883544921875, "loss": 0.8503, "margin_dpo/margin_mean": 69.2376937866211, "margin_dpo/margin_std": 71.49453735351562, "step": 419 }, { "epoch": 0.6167400881057269, "fcm_dpo/beta": 0.013205429539084435, "fcm_dpo/delta": 0.017094507813453674, "fcm_dpo/margin": 63.06053924560547, "fcm_dpo/q_t": 0.3347446918487549, "grad_norm": 15.853597640991211, "learning_rate": 1.9401235374032425e-07, "logits/chosen": -2.1705939769744873, "logits/rejected": -1.1035213470458984, "logps/chosen": -238.99166870117188, "logps/ref_chosen": -114.30848693847656, "logps/ref_rejected": -75.68355560302734, "logps/rejected": -263.42730712890625, "loss": 0.9019, "margin_dpo/margin_mean": 63.060543060302734, "margin_dpo/margin_std": 71.72982788085938, "step": 420 }, { "epoch": 0.618208516886931, "fcm_dpo/beta": 0.012998622842133045, "fcm_dpo/delta": -0.0270537156611681, "fcm_dpo/margin": 67.26036834716797, "fcm_dpo/q_t": 0.32487133145332336, "grad_norm": 12.69776439666748, "learning_rate": 1.9276236251246653e-07, "logits/chosen": -1.688891887664795, "logits/rejected": -1.3223713636398315, "logps/chosen": -209.80950927734375, "logps/ref_chosen": -85.87985229492188, "logps/ref_rejected": -96.33648681640625, "logps/rejected": -287.5264892578125, "loss": 0.8675, "margin_dpo/margin_mean": 67.26036834716797, "margin_dpo/margin_std": 70.79811096191406, "step": 421 }, { "epoch": 0.6196769456681351, "fcm_dpo/beta": 0.013068338856101036, "fcm_dpo/delta": -0.010718483477830887, "fcm_dpo/margin": 65.76750183105469, "fcm_dpo/q_t": 0.3269606828689575, "grad_norm": 12.487133979797363, "learning_rate": 1.9151387954958792e-07, "logits/chosen": -2.3033037185668945, "logits/rejected": -1.9288625717163086, "logps/chosen": -220.82708740234375, "logps/ref_chosen": -100.48060607910156, "logps/ref_rejected": -94.40821838378906, "logps/rejected": -280.5221862792969, "loss": 0.8611, "margin_dpo/margin_mean": 65.76750183105469, "margin_dpo/margin_std": 69.10296630859375, "step": 422 }, { "epoch": 0.6211453744493393, "fcm_dpo/beta": 0.013151784427464008, "fcm_dpo/delta": 0.04228432476520538, "fcm_dpo/margin": 61.62306213378906, "fcm_dpo/q_t": 0.3433518409729004, "grad_norm": 13.384033203125, "learning_rate": 1.902669377503756e-07, "logits/chosen": -1.8323371410369873, "logits/rejected": -1.7185084819793701, "logps/chosen": -195.48995971679688, "logps/ref_chosen": -78.44993591308594, "logps/ref_rejected": -92.04652404785156, "logps/rejected": -270.7095947265625, "loss": 0.943, "margin_dpo/margin_mean": 61.6230583190918, "margin_dpo/margin_std": 77.90271759033203, "step": 423 }, { "epoch": 0.6226138032305433, "fcm_dpo/beta": 0.013109794817864895, "fcm_dpo/delta": -0.008033724501729012, "fcm_dpo/margin": 65.39669799804688, "fcm_dpo/q_t": 0.3260476589202881, "grad_norm": 12.125853538513184, "learning_rate": 1.890215699729057e-07, "logits/chosen": -1.810041904449463, "logits/rejected": -1.2230205535888672, "logps/chosen": -215.9375457763672, "logps/ref_chosen": -87.64236450195312, "logps/ref_rejected": -72.36566162109375, "logps/rejected": -266.05755615234375, "loss": 0.8727, "margin_dpo/margin_mean": 65.39669799804688, "margin_dpo/margin_std": 69.99500274658203, "step": 424 }, { "epoch": 0.6240822320117474, "fcm_dpo/beta": 0.013596277683973312, "fcm_dpo/delta": 0.27160927653312683, "fcm_dpo/margin": 43.7700309753418, "fcm_dpo/q_t": 0.3926636874675751, "grad_norm": 15.634125709533691, "learning_rate": 1.8777780903377732e-07, "logits/chosen": -1.4916131496429443, "logits/rejected": -1.4072489738464355, "logps/chosen": -212.03672790527344, "logps/ref_chosen": -78.51979064941406, "logps/ref_rejected": -102.74864196777344, "logps/rejected": -280.0356140136719, "loss": 1.136, "margin_dpo/margin_mean": 43.77003479003906, "margin_dpo/margin_std": 82.93934631347656, "step": 425 }, { "epoch": 0.6255506607929515, "fcm_dpo/beta": 0.013757694512605667, "fcm_dpo/delta": -0.045442745089530945, "fcm_dpo/margin": 64.79296875, "fcm_dpo/q_t": 0.3165437877178192, "grad_norm": 15.151692390441895, "learning_rate": 1.8653568770724803e-07, "logits/chosen": -1.5372415781021118, "logits/rejected": -0.826823353767395, "logps/chosen": -223.7195587158203, "logps/ref_chosen": -108.50582885742188, "logps/ref_rejected": -88.300048828125, "logps/rejected": -268.3067321777344, "loss": 0.8451, "margin_dpo/margin_mean": 64.79296875, "margin_dpo/margin_std": 64.40265655517578, "step": 426 }, { "epoch": 0.6270190895741556, "fcm_dpo/beta": 0.01376783475279808, "fcm_dpo/delta": 0.018252946436405182, "fcm_dpo/margin": 60.51248550415039, "fcm_dpo/q_t": 0.32682928442955017, "grad_norm": 13.044880867004395, "learning_rate": 1.8529523872436977e-07, "logits/chosen": -2.440342903137207, "logits/rejected": -1.8359763622283936, "logps/chosen": -198.5206298828125, "logps/ref_chosen": -99.12046813964844, "logps/ref_rejected": -85.724609375, "logps/rejected": -245.63723754882812, "loss": 0.8555, "margin_dpo/margin_mean": 60.51248550415039, "margin_dpo/margin_std": 59.35211181640625, "step": 427 }, { "epoch": 0.6284875183553598, "fcm_dpo/beta": 0.01399192400276661, "fcm_dpo/delta": 0.18102213740348816, "fcm_dpo/margin": 48.68280029296875, "fcm_dpo/q_t": 0.36542826890945435, "grad_norm": 14.462920188903809, "learning_rate": 1.8405649477212697e-07, "logits/chosen": -1.4676676988601685, "logits/rejected": -1.1950892210006714, "logps/chosen": -239.3568115234375, "logps/ref_chosen": -105.96925354003906, "logps/ref_rejected": -109.10218048095703, "logps/rejected": -291.17254638671875, "loss": 1.0476, "margin_dpo/margin_mean": 48.68279266357422, "margin_dpo/margin_std": 73.93551635742188, "step": 428 }, { "epoch": 0.6299559471365639, "fcm_dpo/beta": 0.014181070029735565, "fcm_dpo/delta": -0.12901651859283447, "fcm_dpo/margin": 68.02422332763672, "fcm_dpo/q_t": 0.3178831934928894, "grad_norm": 14.960230827331543, "learning_rate": 1.828194884925749e-07, "logits/chosen": -1.7683967351913452, "logits/rejected": -1.0629897117614746, "logps/chosen": -233.390869140625, "logps/ref_chosen": -113.54487609863281, "logps/ref_rejected": -98.24201965332031, "logps/rejected": -286.11224365234375, "loss": 0.8717, "margin_dpo/margin_mean": 68.02421569824219, "margin_dpo/margin_std": 75.7259292602539, "step": 429 }, { "epoch": 0.631424375917768, "fcm_dpo/beta": 0.014241167344152927, "fcm_dpo/delta": 0.14151397347450256, "fcm_dpo/margin": 50.407920837402344, "fcm_dpo/q_t": 0.35696834325790405, "grad_norm": 16.362232208251953, "learning_rate": 1.8158425248197928e-07, "logits/chosen": -1.3190555572509766, "logits/rejected": -1.270359992980957, "logps/chosen": -212.43923950195312, "logps/ref_chosen": -91.31936645507812, "logps/ref_rejected": -110.1096420288086, "logps/rejected": -281.637451171875, "loss": 0.9831, "margin_dpo/margin_mean": 50.407920837402344, "margin_dpo/margin_std": 67.11074829101562, "step": 430 }, { "epoch": 0.6328928046989721, "fcm_dpo/beta": 0.014219909906387329, "fcm_dpo/delta": -0.0074679069221019745, "fcm_dpo/margin": 60.23283386230469, "fcm_dpo/q_t": 0.33199840784072876, "grad_norm": 14.520343780517578, "learning_rate": 1.8035081928995788e-07, "logits/chosen": -1.8971822261810303, "logits/rejected": -1.5131553411483765, "logps/chosen": -212.03729248046875, "logps/ref_chosen": -93.18122100830078, "logps/ref_rejected": -98.13226318359375, "logps/rejected": -277.22119140625, "loss": 0.9109, "margin_dpo/margin_mean": 60.23283386230469, "margin_dpo/margin_std": 71.3617935180664, "step": 431 }, { "epoch": 0.6343612334801763, "fcm_dpo/beta": 0.01398632675409317, "fcm_dpo/delta": -0.17663273215293884, "fcm_dpo/margin": 72.14842987060547, "fcm_dpo/q_t": 0.30276134610176086, "grad_norm": 13.15101432800293, "learning_rate": 1.791192214186223e-07, "logits/chosen": -1.2233517169952393, "logits/rejected": -0.6508141756057739, "logps/chosen": -211.9017333984375, "logps/ref_chosen": -104.43478393554688, "logps/ref_rejected": -105.08955383300781, "logps/rejected": -284.7049560546875, "loss": 0.8083, "margin_dpo/margin_mean": 72.14842987060547, "margin_dpo/margin_std": 70.37438201904297, "step": 432 }, { "epoch": 0.6358296622613803, "fcm_dpo/beta": 0.014024633914232254, "fcm_dpo/delta": 0.04999150335788727, "fcm_dpo/margin": 57.14216613769531, "fcm_dpo/q_t": 0.34285709261894226, "grad_norm": 14.23776912689209, "learning_rate": 1.7788949132172193e-07, "logits/chosen": -1.6155874729156494, "logits/rejected": -1.2833582162857056, "logps/chosen": -216.49232482910156, "logps/ref_chosen": -89.84322357177734, "logps/ref_rejected": -101.73345947265625, "logps/rejected": -285.52471923828125, "loss": 0.9633, "margin_dpo/margin_mean": 57.142173767089844, "margin_dpo/margin_std": 73.58992004394531, "step": 433 }, { "epoch": 0.6372980910425844, "fcm_dpo/beta": 0.013989130035042763, "fcm_dpo/delta": 0.024751663208007812, "fcm_dpo/margin": 59.13129806518555, "fcm_dpo/q_t": 0.342306524515152, "grad_norm": 13.736560821533203, "learning_rate": 1.7666166140378853e-07, "logits/chosen": -2.4307074546813965, "logits/rejected": -1.8070075511932373, "logps/chosen": -215.8836212158203, "logps/ref_chosen": -97.6925277709961, "logps/ref_rejected": -84.09129333496094, "logps/rejected": -261.4136962890625, "loss": 0.9414, "margin_dpo/margin_mean": 59.13129806518555, "margin_dpo/margin_std": 77.16365814208984, "step": 434 }, { "epoch": 0.6387665198237885, "fcm_dpo/beta": 0.01387035846710205, "fcm_dpo/delta": -0.06785215437412262, "fcm_dpo/margin": 65.7310791015625, "fcm_dpo/q_t": 0.3216383159160614, "grad_norm": 14.38983154296875, "learning_rate": 1.7543576401928218e-07, "logits/chosen": -2.151081085205078, "logits/rejected": -1.6833906173706055, "logps/chosen": -198.29730224609375, "logps/ref_chosen": -86.17192077636719, "logps/ref_rejected": -93.751708984375, "logps/rejected": -271.608154296875, "loss": 0.8643, "margin_dpo/margin_mean": 65.7310791015625, "margin_dpo/margin_std": 72.66617584228516, "step": 435 }, { "epoch": 0.6402349486049926, "fcm_dpo/beta": 0.01344333216547966, "fcm_dpo/delta": -0.18730014562606812, "fcm_dpo/margin": 75.75979614257812, "fcm_dpo/q_t": 0.289669394493103, "grad_norm": 12.1524076461792, "learning_rate": 1.742118314717391e-07, "logits/chosen": -1.5553090572357178, "logits/rejected": -0.6732375025749207, "logps/chosen": -197.84014892578125, "logps/ref_chosen": -105.78710174560547, "logps/ref_rejected": -88.62471008300781, "logps/rejected": -256.43756103515625, "loss": 0.7541, "margin_dpo/margin_mean": 75.75979614257812, "margin_dpo/margin_std": 62.77748107910156, "step": 436 }, { "epoch": 0.6417033773861968, "fcm_dpo/beta": 0.013649407774209976, "fcm_dpo/delta": 0.1684824675321579, "fcm_dpo/margin": 50.75785446166992, "fcm_dpo/q_t": 0.3533422648906708, "grad_norm": 14.654732704162598, "learning_rate": 1.7298989601292036e-07, "logits/chosen": -2.0003044605255127, "logits/rejected": -1.368834376335144, "logps/chosen": -206.8590087890625, "logps/ref_chosen": -96.06204223632812, "logps/ref_rejected": -89.01220703125, "logps/rejected": -250.5670166015625, "loss": 0.954, "margin_dpo/margin_mean": 50.757850646972656, "margin_dpo/margin_std": 59.297645568847656, "step": 437 }, { "epoch": 0.6431718061674009, "fcm_dpo/beta": 0.013722064904868603, "fcm_dpo/delta": 0.016998998820781708, "fcm_dpo/margin": 60.7908935546875, "fcm_dpo/q_t": 0.32923266291618347, "grad_norm": 13.463798522949219, "learning_rate": 1.7176998984196144e-07, "logits/chosen": -2.1621651649475098, "logits/rejected": -1.3619840145111084, "logps/chosen": -206.91421508789062, "logps/ref_chosen": -101.85537719726562, "logps/ref_rejected": -89.44766235351562, "logps/rejected": -255.2974090576172, "loss": 0.8886, "margin_dpo/margin_mean": 60.790897369384766, "margin_dpo/margin_std": 65.22203063964844, "step": 438 }, { "epoch": 0.644640234948605, "fcm_dpo/beta": 0.013775939121842384, "fcm_dpo/delta": 0.07153499871492386, "fcm_dpo/margin": 56.78171920776367, "fcm_dpo/q_t": 0.3450607657432556, "grad_norm": 14.867757797241211, "learning_rate": 1.7055214510452458e-07, "logits/chosen": -1.614975929260254, "logits/rejected": -1.3566641807556152, "logps/chosen": -200.1914520263672, "logps/ref_chosen": -81.75563049316406, "logps/ref_rejected": -90.58636474609375, "logps/rejected": -265.80389404296875, "loss": 0.9414, "margin_dpo/margin_mean": 56.78171920776367, "margin_dpo/margin_std": 68.44754028320312, "step": 439 }, { "epoch": 0.6461086637298091, "fcm_dpo/beta": 0.013727106153964996, "fcm_dpo/delta": -0.06820373237133026, "fcm_dpo/margin": 66.29185485839844, "fcm_dpo/q_t": 0.3204452395439148, "grad_norm": 12.376969337463379, "learning_rate": 1.6933639389195134e-07, "logits/chosen": -2.5066826343536377, "logits/rejected": -2.0377886295318604, "logps/chosen": -211.48703002929688, "logps/ref_chosen": -105.64108276367188, "logps/ref_rejected": -103.4010009765625, "logps/rejected": -275.538818359375, "loss": 0.8662, "margin_dpo/margin_mean": 66.29185485839844, "margin_dpo/margin_std": 71.9673080444336, "step": 440 }, { "epoch": 0.6475770925110133, "fcm_dpo/beta": 0.013530462980270386, "fcm_dpo/delta": -0.10768324136734009, "fcm_dpo/margin": 69.98019409179688, "fcm_dpo/q_t": 0.30917853116989136, "grad_norm": 11.863908767700195, "learning_rate": 1.681227682404166e-07, "logits/chosen": -1.7193225622177124, "logits/rejected": -1.3498790264129639, "logps/chosen": -214.11981201171875, "logps/ref_chosen": -91.529541015625, "logps/ref_rejected": -103.619384765625, "logps/rejected": -296.18988037109375, "loss": 0.787, "margin_dpo/margin_mean": 69.98018646240234, "margin_dpo/margin_std": 65.251708984375, "step": 441 }, { "epoch": 0.6490455212922174, "fcm_dpo/beta": 0.013135121203958988, "fcm_dpo/delta": -0.1771937906742096, "fcm_dpo/margin": 76.81739044189453, "fcm_dpo/q_t": 0.29648613929748535, "grad_norm": 12.602315902709961, "learning_rate": 1.669113001300851e-07, "logits/chosen": -1.4338107109069824, "logits/rejected": -0.9611091613769531, "logps/chosen": -208.39720153808594, "logps/ref_chosen": -84.77755737304688, "logps/ref_rejected": -83.82415771484375, "logps/rejected": -284.26116943359375, "loss": 0.7703, "margin_dpo/margin_mean": 76.8173828125, "margin_dpo/margin_std": 68.45848083496094, "step": 442 }, { "epoch": 0.6505139500734214, "fcm_dpo/beta": 0.013111368753015995, "fcm_dpo/delta": 0.11141438782215118, "fcm_dpo/margin": 56.881134033203125, "fcm_dpo/q_t": 0.34306538105010986, "grad_norm": 13.224976539611816, "learning_rate": 1.6570202148426815e-07, "logits/chosen": -1.735137701034546, "logits/rejected": -1.1993635892868042, "logps/chosen": -220.87890625, "logps/ref_chosen": -102.64927673339844, "logps/ref_rejected": -93.03807067871094, "logps/rejected": -268.14886474609375, "loss": 0.9271, "margin_dpo/margin_mean": 56.881134033203125, "margin_dpo/margin_std": 61.88277816772461, "step": 443 }, { "epoch": 0.6519823788546255, "fcm_dpo/beta": 0.0135607048869133, "fcm_dpo/delta": 0.10103952884674072, "fcm_dpo/margin": 55.769325256347656, "fcm_dpo/q_t": 0.35067448019981384, "grad_norm": 14.165532112121582, "learning_rate": 1.6449496416858282e-07, "logits/chosen": -1.592213749885559, "logits/rejected": -1.420748233795166, "logps/chosen": -214.36148071289062, "logps/ref_chosen": -87.91971588134766, "logps/ref_rejected": -103.32345581054688, "logps/rejected": -285.5345458984375, "loss": 0.9687, "margin_dpo/margin_mean": 55.769325256347656, "margin_dpo/margin_std": 72.71159362792969, "step": 444 }, { "epoch": 0.6534508076358296, "fcm_dpo/beta": 0.01349145732820034, "fcm_dpo/delta": -0.027875246480107307, "fcm_dpo/margin": 64.85955810546875, "fcm_dpo/q_t": 0.32111674547195435, "grad_norm": 12.245442390441895, "learning_rate": 1.6329015999011182e-07, "logits/chosen": -1.9300236701965332, "logits/rejected": -1.5292446613311768, "logps/chosen": -211.0179443359375, "logps/ref_chosen": -101.40086364746094, "logps/ref_rejected": -99.03790283203125, "logps/rejected": -273.5145263671875, "loss": 0.8467, "margin_dpo/margin_mean": 64.85956573486328, "margin_dpo/margin_std": 64.7419662475586, "step": 445 }, { "epoch": 0.6549192364170338, "fcm_dpo/beta": 0.013596500270068645, "fcm_dpo/delta": 0.05426015704870224, "fcm_dpo/margin": 58.844242095947266, "fcm_dpo/q_t": 0.33681419491767883, "grad_norm": 14.447946548461914, "learning_rate": 1.6208764069656578e-07, "logits/chosen": -1.9224543571472168, "logits/rejected": -1.767820119857788, "logps/chosen": -194.24234008789062, "logps/ref_chosen": -87.4223403930664, "logps/ref_rejected": -106.70075988769531, "logps/rejected": -272.364990234375, "loss": 0.9015, "margin_dpo/margin_mean": 58.84424591064453, "margin_dpo/margin_std": 65.06867980957031, "step": 446 }, { "epoch": 0.6563876651982379, "fcm_dpo/beta": 0.01339960377663374, "fcm_dpo/delta": -0.11742748320102692, "fcm_dpo/margin": 71.3275375366211, "fcm_dpo/q_t": 0.31018170714378357, "grad_norm": 13.609336853027344, "learning_rate": 1.608874379754465e-07, "logits/chosen": -2.158440113067627, "logits/rejected": -2.107996940612793, "logps/chosen": -191.38192749023438, "logps/ref_chosen": -83.6152572631836, "logps/ref_rejected": -104.91239929199219, "logps/rejected": -284.006591796875, "loss": 0.8392, "margin_dpo/margin_mean": 71.3275375366211, "margin_dpo/margin_std": 73.93170166015625, "step": 447 }, { "epoch": 0.657856093979442, "fcm_dpo/beta": 0.013280214741826057, "fcm_dpo/delta": -0.09408734738826752, "fcm_dpo/margin": 70.38318634033203, "fcm_dpo/q_t": 0.32261860370635986, "grad_norm": 14.318117141723633, "learning_rate": 1.5968958345321177e-07, "logits/chosen": -1.5758308172225952, "logits/rejected": -1.3030452728271484, "logps/chosen": -207.17013549804688, "logps/ref_chosen": -92.5757827758789, "logps/ref_rejected": -107.68977355957031, "logps/rejected": -292.6673278808594, "loss": 0.9036, "margin_dpo/margin_mean": 70.38318634033203, "margin_dpo/margin_std": 83.00077819824219, "step": 448 }, { "epoch": 0.6593245227606461, "fcm_dpo/beta": 0.012970153242349625, "fcm_dpo/delta": -0.09647231549024582, "fcm_dpo/margin": 72.28010559082031, "fcm_dpo/q_t": 0.3125626742839813, "grad_norm": 12.015176773071289, "learning_rate": 1.584941086944423e-07, "logits/chosen": -1.4746041297912598, "logits/rejected": -1.0464733839035034, "logps/chosen": -215.36099243164062, "logps/ref_chosen": -102.39893341064453, "logps/ref_rejected": -95.14886474609375, "logps/rejected": -280.3910217285156, "loss": 0.8473, "margin_dpo/margin_mean": 72.28010559082031, "margin_dpo/margin_std": 75.04696655273438, "step": 449 }, { "epoch": 0.6607929515418502, "fcm_dpo/beta": 0.012733325362205505, "fcm_dpo/delta": -0.0715692788362503, "fcm_dpo/margin": 71.86397552490234, "fcm_dpo/q_t": 0.3133106827735901, "grad_norm": 12.985596656799316, "learning_rate": 1.573010452010098e-07, "logits/chosen": -2.383078098297119, "logits/rejected": -2.1813578605651855, "logps/chosen": -186.05169677734375, "logps/ref_chosen": -86.99285888671875, "logps/ref_rejected": -108.53203582763672, "logps/rejected": -279.4548645019531, "loss": 0.8488, "margin_dpo/margin_mean": 71.86397552490234, "margin_dpo/margin_std": 73.41386413574219, "step": 450 }, { "epoch": 0.6622613803230544, "fcm_dpo/beta": 0.012889644131064415, "fcm_dpo/delta": 0.07636302709579468, "fcm_dpo/margin": 60.40147399902344, "fcm_dpo/q_t": 0.3498266339302063, "grad_norm": 18.933948516845703, "learning_rate": 1.5611042441124687e-07, "logits/chosen": -2.3939194679260254, "logits/rejected": -1.90476655960083, "logps/chosen": -199.50680541992188, "logps/ref_chosen": -86.811279296875, "logps/ref_rejected": -79.8555908203125, "logps/rejected": -252.95257568359375, "loss": 0.9826, "margin_dpo/margin_mean": 60.40147399902344, "margin_dpo/margin_std": 81.97018432617188, "step": 451 }, { "epoch": 0.6637298091042585, "fcm_dpo/beta": 0.012789755128324032, "fcm_dpo/delta": 0.04102400690317154, "fcm_dpo/margin": 63.32386779785156, "fcm_dpo/q_t": 0.3287479281425476, "grad_norm": 12.005148887634277, "learning_rate": 1.549222776991186e-07, "logits/chosen": -1.8443999290466309, "logits/rejected": -1.7731636762619019, "logps/chosen": -179.27359008789062, "logps/ref_chosen": -79.379638671875, "logps/ref_rejected": -103.71539306640625, "logps/rejected": -266.9331970214844, "loss": 0.8507, "margin_dpo/margin_mean": 63.32386779785156, "margin_dpo/margin_std": 55.35037612915039, "step": 452 }, { "epoch": 0.6651982378854625, "fcm_dpo/beta": 0.012718062847852707, "fcm_dpo/delta": -0.1492374986410141, "fcm_dpo/margin": 77.42842102050781, "fcm_dpo/q_t": 0.30019137263298035, "grad_norm": 12.277626037597656, "learning_rate": 1.5373663637339584e-07, "logits/chosen": -1.9240939617156982, "logits/rejected": -1.4621145725250244, "logps/chosen": -204.21820068359375, "logps/ref_chosen": -87.6951904296875, "logps/ref_rejected": -90.0582275390625, "logps/rejected": -284.0096435546875, "loss": 0.7741, "margin_dpo/margin_mean": 77.42842102050781, "margin_dpo/margin_std": 67.35591125488281, "step": 453 }, { "epoch": 0.6666666666666666, "fcm_dpo/beta": 0.012760400772094727, "fcm_dpo/delta": 0.06223556026816368, "fcm_dpo/margin": 62.07777786254883, "fcm_dpo/q_t": 0.3461794853210449, "grad_norm": 14.331741333007812, "learning_rate": 1.5255353167683017e-07, "logits/chosen": -1.8063360452651978, "logits/rejected": -1.3538455963134766, "logps/chosen": -215.82162475585938, "logps/ref_chosen": -89.56623840332031, "logps/ref_rejected": -92.92105102539062, "logps/rejected": -281.25421142578125, "loss": 0.9868, "margin_dpo/margin_mean": 62.07777786254883, "margin_dpo/margin_std": 85.04985809326172, "step": 454 }, { "epoch": 0.6681350954478708, "fcm_dpo/beta": 0.012872504070401192, "fcm_dpo/delta": 0.028366755694150925, "fcm_dpo/margin": 63.95558547973633, "fcm_dpo/q_t": 0.3338191509246826, "grad_norm": 13.07781982421875, "learning_rate": 1.5137299478533064e-07, "logits/chosen": -1.8931918144226074, "logits/rejected": -1.8260715007781982, "logps/chosen": -190.75283813476562, "logps/ref_chosen": -77.62998962402344, "logps/ref_rejected": -118.97795104980469, "logps/rejected": -296.0563659667969, "loss": 0.9057, "margin_dpo/margin_mean": 63.95558166503906, "margin_dpo/margin_std": 71.64030456542969, "step": 455 }, { "epoch": 0.6696035242290749, "fcm_dpo/beta": 0.013063902035355568, "fcm_dpo/delta": 0.09057056158781052, "fcm_dpo/margin": 58.58759307861328, "fcm_dpo/q_t": 0.3430330455303192, "grad_norm": 15.235799789428711, "learning_rate": 1.5019505680714232e-07, "logits/chosen": -2.0362932682037354, "logits/rejected": -1.9036009311676025, "logps/chosen": -206.73255920410156, "logps/ref_chosen": -89.61686706542969, "logps/ref_rejected": -109.5597152709961, "logps/rejected": -285.26300048828125, "loss": 0.9192, "margin_dpo/margin_mean": 58.58759307861328, "margin_dpo/margin_std": 65.98350524902344, "step": 456 }, { "epoch": 0.671071953010279, "fcm_dpo/beta": 0.012593207880854607, "fcm_dpo/delta": -0.1489265263080597, "fcm_dpo/margin": 77.71039581298828, "fcm_dpo/q_t": 0.2934274673461914, "grad_norm": 12.465642929077148, "learning_rate": 1.4901974878202627e-07, "logits/chosen": -1.9858686923980713, "logits/rejected": -1.677903413772583, "logps/chosen": -187.23529052734375, "logps/ref_chosen": -87.32168579101562, "logps/ref_rejected": -90.76660919189453, "logps/rejected": -268.390625, "loss": 0.753, "margin_dpo/margin_mean": 77.71039581298828, "margin_dpo/margin_std": 58.36017608642578, "step": 457 }, { "epoch": 0.6725403817914831, "fcm_dpo/beta": 0.012880630791187286, "fcm_dpo/delta": 0.09830697625875473, "fcm_dpo/margin": 58.92164611816406, "fcm_dpo/q_t": 0.3527994155883789, "grad_norm": 14.763307571411133, "learning_rate": 1.4784710168044212e-07, "logits/chosen": -1.6106715202331543, "logits/rejected": -1.3593188524246216, "logps/chosen": -205.40960693359375, "logps/ref_chosen": -93.52044677734375, "logps/ref_rejected": -103.36898803710938, "logps/rejected": -274.1798095703125, "loss": 0.9732, "margin_dpo/margin_mean": 58.92164611816406, "margin_dpo/margin_std": 79.73458099365234, "step": 458 }, { "epoch": 0.6740088105726872, "fcm_dpo/beta": 0.01270090602338314, "fcm_dpo/delta": -0.07352010160684586, "fcm_dpo/margin": 72.08202362060547, "fcm_dpo/q_t": 0.3198687434196472, "grad_norm": 10.96422290802002, "learning_rate": 1.466771464027316e-07, "logits/chosen": -1.7052292823791504, "logits/rejected": -1.4467270374298096, "logps/chosen": -184.02252197265625, "logps/ref_chosen": -75.68820190429688, "logps/ref_rejected": -92.17048645019531, "logps/rejected": -272.5868225097656, "loss": 0.8545, "margin_dpo/margin_mean": 72.08201599121094, "margin_dpo/margin_std": 76.00017547607422, "step": 459 }, { "epoch": 0.6754772393538914, "fcm_dpo/beta": 0.012723826803267002, "fcm_dpo/delta": 0.032767970114946365, "fcm_dpo/margin": 64.40177917480469, "fcm_dpo/q_t": 0.3371509313583374, "grad_norm": 14.807209014892578, "learning_rate": 1.4550991377830423e-07, "logits/chosen": -1.999427318572998, "logits/rejected": -1.9956591129302979, "logps/chosen": -194.11785888671875, "logps/ref_chosen": -81.11788940429688, "logps/ref_rejected": -110.31238555908203, "logps/rejected": -287.714111328125, "loss": 0.9115, "margin_dpo/margin_mean": 64.40177917480469, "margin_dpo/margin_std": 75.08444213867188, "step": 460 }, { "epoch": 0.6769456681350955, "fcm_dpo/beta": 0.013109425082802773, "fcm_dpo/delta": 0.17610467970371246, "fcm_dpo/margin": 52.35367965698242, "fcm_dpo/q_t": 0.36317408084869385, "grad_norm": 14.377598762512207, "learning_rate": 1.4434543456482518e-07, "logits/chosen": -1.51175057888031, "logits/rejected": -1.312682867050171, "logps/chosen": -216.15902709960938, "logps/ref_chosen": -81.58352661132812, "logps/ref_rejected": -93.87710571289062, "logps/rejected": -280.8062744140625, "loss": 0.9802, "margin_dpo/margin_mean": 52.35367965698242, "margin_dpo/margin_std": 67.33370971679688, "step": 461 }, { "epoch": 0.6784140969162996, "fcm_dpo/beta": 0.01316889002919197, "fcm_dpo/delta": -0.08168931305408478, "fcm_dpo/margin": 70.1772232055664, "fcm_dpo/q_t": 0.3104817271232605, "grad_norm": 12.93285846710205, "learning_rate": 1.4318373944740484e-07, "logits/chosen": -2.101900100708008, "logits/rejected": -1.5979928970336914, "logps/chosen": -198.15518188476562, "logps/ref_chosen": -94.19854736328125, "logps/ref_rejected": -85.63162994384766, "logps/rejected": -259.7654724121094, "loss": 0.8176, "margin_dpo/margin_mean": 70.17721557617188, "margin_dpo/margin_std": 65.97860717773438, "step": 462 }, { "epoch": 0.6798825256975036, "fcm_dpo/beta": 0.012822076678276062, "fcm_dpo/delta": -0.21826551854610443, "fcm_dpo/margin": 81.48209381103516, "fcm_dpo/q_t": 0.2897373139858246, "grad_norm": 13.616190910339355, "learning_rate": 1.4202485903778976e-07, "logits/chosen": -1.8920228481292725, "logits/rejected": -1.4664556980133057, "logps/chosen": -187.51708984375, "logps/ref_chosen": -85.92474365234375, "logps/ref_rejected": -96.90184020996094, "logps/rejected": -279.97625732421875, "loss": 0.8029, "margin_dpo/margin_mean": 81.48209381103516, "margin_dpo/margin_std": 75.05322265625, "step": 463 }, { "epoch": 0.6813509544787077, "fcm_dpo/beta": 0.012513959780335426, "fcm_dpo/delta": 0.027589473873376846, "fcm_dpo/margin": 65.88970947265625, "fcm_dpo/q_t": 0.3308594822883606, "grad_norm": 17.59402847290039, "learning_rate": 1.4086882387355658e-07, "logits/chosen": -1.7415244579315186, "logits/rejected": -1.8106495141983032, "logps/chosen": -202.68325805664062, "logps/ref_chosen": -79.68919372558594, "logps/ref_rejected": -107.29232025146484, "logps/rejected": -296.1761169433594, "loss": 0.8853, "margin_dpo/margin_mean": 65.88970947265625, "margin_dpo/margin_std": 71.70936584472656, "step": 464 }, { "epoch": 0.6828193832599119, "fcm_dpo/beta": 0.012278483249247074, "fcm_dpo/delta": -0.1325981318950653, "fcm_dpo/margin": 78.90789794921875, "fcm_dpo/q_t": 0.3039529621601105, "grad_norm": 13.62932014465332, "learning_rate": 1.3971566441730714e-07, "logits/chosen": -1.6111962795257568, "logits/rejected": -1.3117473125457764, "logps/chosen": -204.054443359375, "logps/ref_chosen": -91.8602294921875, "logps/ref_rejected": -118.71001434326172, "logps/rejected": -309.8121337890625, "loss": 0.7876, "margin_dpo/margin_mean": 78.90789794921875, "margin_dpo/margin_std": 71.70303344726562, "step": 465 }, { "epoch": 0.684287812041116, "fcm_dpo/beta": 0.011934199370443821, "fcm_dpo/delta": -0.1437564343214035, "fcm_dpo/margin": 82.00619506835938, "fcm_dpo/q_t": 0.30160197615623474, "grad_norm": 10.586161613464355, "learning_rate": 1.3856541105586545e-07, "logits/chosen": -1.4708101749420166, "logits/rejected": -1.1374900341033936, "logps/chosen": -186.422607421875, "logps/ref_chosen": -84.70140838623047, "logps/ref_rejected": -96.05084228515625, "logps/rejected": -279.77825927734375, "loss": 0.7936, "margin_dpo/margin_mean": 82.00619506835938, "margin_dpo/margin_std": 75.4998779296875, "step": 466 }, { "epoch": 0.6857562408223201, "fcm_dpo/beta": 0.011811432428658009, "fcm_dpo/delta": -0.001256793737411499, "fcm_dpo/margin": 71.97554016113281, "fcm_dpo/q_t": 0.3301178514957428, "grad_norm": 13.264272689819336, "learning_rate": 1.3741809409947729e-07, "logits/chosen": -2.3040666580200195, "logits/rejected": -1.7428507804870605, "logps/chosen": -234.81190490722656, "logps/ref_chosen": -109.29832458496094, "logps/ref_rejected": -108.8436508178711, "logps/rejected": -306.332763671875, "loss": 0.9072, "margin_dpo/margin_mean": 71.97554016113281, "margin_dpo/margin_std": 82.74515533447266, "step": 467 }, { "epoch": 0.6872246696035242, "fcm_dpo/beta": 0.012133453041315079, "fcm_dpo/delta": 0.14525851607322693, "fcm_dpo/margin": 58.95384979248047, "fcm_dpo/q_t": 0.3619929850101471, "grad_norm": 13.532750129699707, "learning_rate": 1.362737437810114e-07, "logits/chosen": -1.8041236400604248, "logits/rejected": -1.618350625038147, "logps/chosen": -202.35598754882812, "logps/ref_chosen": -98.32164764404297, "logps/ref_rejected": -106.68048095703125, "logps/rejected": -269.6686706542969, "loss": 0.9891, "margin_dpo/margin_mean": 58.953853607177734, "margin_dpo/margin_std": 81.2957992553711, "step": 468 }, { "epoch": 0.6886930983847284, "fcm_dpo/beta": 0.012059122323989868, "fcm_dpo/delta": -0.059398531913757324, "fcm_dpo/margin": 74.865966796875, "fcm_dpo/q_t": 0.3162824809551239, "grad_norm": 15.22911548614502, "learning_rate": 1.351323902551631e-07, "logits/chosen": -2.2128915786743164, "logits/rejected": -1.7914154529571533, "logps/chosen": -218.97763061523438, "logps/ref_chosen": -96.76420593261719, "logps/ref_rejected": -109.59500885009766, "logps/rejected": -306.6744079589844, "loss": 0.8504, "margin_dpo/margin_mean": 74.86597442626953, "margin_dpo/margin_std": 76.03158569335938, "step": 469 }, { "epoch": 0.6901615271659325, "fcm_dpo/beta": 0.012082107365131378, "fcm_dpo/delta": -0.016534948721528053, "fcm_dpo/margin": 71.60215759277344, "fcm_dpo/q_t": 0.3221762180328369, "grad_norm": 12.687642097473145, "learning_rate": 1.339940635976592e-07, "logits/chosen": -1.6372177600860596, "logits/rejected": -1.2168526649475098, "logps/chosen": -199.10205078125, "logps/ref_chosen": -83.49665832519531, "logps/ref_rejected": -88.48578643798828, "logps/rejected": -275.6933288574219, "loss": 0.8488, "margin_dpo/margin_mean": 71.60215759277344, "margin_dpo/margin_std": 71.88697052001953, "step": 470 }, { "epoch": 0.6916299559471366, "fcm_dpo/beta": 0.01205938495695591, "fcm_dpo/delta": 0.02074136957526207, "fcm_dpo/margin": 68.89513397216797, "fcm_dpo/q_t": 0.3252991735935211, "grad_norm": 12.872730255126953, "learning_rate": 1.3285879380446563e-07, "logits/chosen": -1.4987744092941284, "logits/rejected": -1.1069977283477783, "logps/chosen": -221.21546936035156, "logps/ref_chosen": -88.47430419921875, "logps/ref_rejected": -90.48171997070312, "logps/rejected": -292.1180419921875, "loss": 0.852, "margin_dpo/margin_mean": 68.89513397216797, "margin_dpo/margin_std": 65.54840087890625, "step": 471 }, { "epoch": 0.6930983847283406, "fcm_dpo/beta": 0.011771542951464653, "fcm_dpo/delta": -0.16034545004367828, "fcm_dpo/margin": 84.37684631347656, "fcm_dpo/q_t": 0.2922556400299072, "grad_norm": 10.080926895141602, "learning_rate": 1.317266107909975e-07, "logits/chosen": -2.158059597015381, "logits/rejected": -1.4828379154205322, "logps/chosen": -244.9671630859375, "logps/ref_chosen": -125.23370361328125, "logps/ref_rejected": -121.05349731445312, "logps/rejected": -325.163818359375, "loss": 0.7379, "margin_dpo/margin_mean": 84.37684631347656, "margin_dpo/margin_std": 64.8412857055664, "step": 472 }, { "epoch": 0.6945668135095447, "fcm_dpo/beta": 0.012129031121730804, "fcm_dpo/delta": 0.19991233944892883, "fcm_dpo/margin": 54.608848571777344, "fcm_dpo/q_t": 0.3692033588886261, "grad_norm": 14.75636100769043, "learning_rate": 1.3059754439133002e-07, "logits/chosen": -1.7300171852111816, "logits/rejected": -1.2311158180236816, "logps/chosen": -231.66799926757812, "logps/ref_chosen": -95.61137390136719, "logps/ref_rejected": -88.15115356445312, "logps/rejected": -278.816650390625, "loss": 1.0173, "margin_dpo/margin_mean": 54.608848571777344, "margin_dpo/margin_std": 77.77999114990234, "step": 473 }, { "epoch": 0.6960352422907489, "fcm_dpo/beta": 0.012801304459571838, "fcm_dpo/delta": 0.23721067607402802, "fcm_dpo/margin": 48.58485412597656, "fcm_dpo/q_t": 0.3742219805717468, "grad_norm": 16.70572853088379, "learning_rate": 1.2947162435741277e-07, "logits/chosen": -1.5113579034805298, "logits/rejected": -1.3194079399108887, "logps/chosen": -215.53375244140625, "logps/ref_chosen": -81.47975158691406, "logps/ref_rejected": -96.46562957763672, "logps/rejected": -279.1044921875, "loss": 1.0623, "margin_dpo/margin_mean": 48.58485412597656, "margin_dpo/margin_std": 72.38153076171875, "step": 474 }, { "epoch": 0.697503671071953, "fcm_dpo/beta": 0.012611234560608864, "fcm_dpo/delta": -0.10543165355920792, "fcm_dpo/margin": 74.97346496582031, "fcm_dpo/q_t": 0.3156423270702362, "grad_norm": 13.300975799560547, "learning_rate": 1.2834888035828596e-07, "logits/chosen": -2.169642448425293, "logits/rejected": -2.1165261268615723, "logps/chosen": -175.56845092773438, "logps/ref_chosen": -74.19598388671875, "logps/ref_rejected": -94.69242095947266, "logps/rejected": -271.038330078125, "loss": 0.8354, "margin_dpo/margin_mean": 74.97346496582031, "margin_dpo/margin_std": 78.37078857421875, "step": 475 }, { "epoch": 0.6989720998531571, "fcm_dpo/beta": 0.012492308393120766, "fcm_dpo/delta": 0.027760498225688934, "fcm_dpo/margin": 65.93070983886719, "fcm_dpo/q_t": 0.341197669506073, "grad_norm": 14.65475082397461, "learning_rate": 1.2722934197929802e-07, "logits/chosen": -2.071859836578369, "logits/rejected": -1.693770170211792, "logps/chosen": -192.81143188476562, "logps/ref_chosen": -71.97109985351562, "logps/ref_rejected": -80.26224517822266, "logps/rejected": -267.0332946777344, "loss": 0.9165, "margin_dpo/margin_mean": 65.93070983886719, "margin_dpo/margin_std": 79.12750244140625, "step": 476 }, { "epoch": 0.7004405286343612, "fcm_dpo/beta": 0.012532323598861694, "fcm_dpo/delta": -0.045053720474243164, "fcm_dpo/margin": 71.10099029541016, "fcm_dpo/q_t": 0.31960827112197876, "grad_norm": 13.954968452453613, "learning_rate": 1.2611303872132631e-07, "logits/chosen": -2.056334972381592, "logits/rejected": -1.1700620651245117, "logps/chosen": -223.75918579101562, "logps/ref_chosen": -105.00555419921875, "logps/ref_rejected": -81.87843322753906, "logps/rejected": -271.73309326171875, "loss": 0.8799, "margin_dpo/margin_mean": 71.10099029541016, "margin_dpo/margin_std": 78.1356201171875, "step": 477 }, { "epoch": 0.7019089574155654, "fcm_dpo/beta": 0.012363728135824203, "fcm_dpo/delta": -0.06362555921077728, "fcm_dpo/margin": 73.42304992675781, "fcm_dpo/q_t": 0.3152768015861511, "grad_norm": 12.457188606262207, "learning_rate": 1.2500000000000005e-07, "logits/chosen": -1.6319268941879272, "logits/rejected": -1.4650232791900635, "logps/chosen": -189.54151916503906, "logps/ref_chosen": -76.7882080078125, "logps/ref_rejected": -90.43994140625, "logps/rejected": -276.6163024902344, "loss": 0.8413, "margin_dpo/margin_mean": 73.42304992675781, "margin_dpo/margin_std": 74.97042083740234, "step": 478 }, { "epoch": 0.7033773861967695, "fcm_dpo/beta": 0.01251962035894394, "fcm_dpo/delta": 0.05877025052905083, "fcm_dpo/margin": 63.48487091064453, "fcm_dpo/q_t": 0.33374714851379395, "grad_norm": 16.574216842651367, "learning_rate": 1.2389025514492456e-07, "logits/chosen": -1.4575786590576172, "logits/rejected": -1.3220562934875488, "logps/chosen": -202.30274963378906, "logps/ref_chosen": -81.3623046875, "logps/ref_rejected": -101.09114074707031, "logps/rejected": -285.5164489746094, "loss": 0.9021, "margin_dpo/margin_mean": 63.48487091064453, "margin_dpo/margin_std": 68.2608871459961, "step": 479 }, { "epoch": 0.7048458149779736, "fcm_dpo/beta": 0.012569671496748924, "fcm_dpo/delta": 0.05630197376012802, "fcm_dpo/margin": 63.504146575927734, "fcm_dpo/q_t": 0.3408883213996887, "grad_norm": 16.51081085205078, "learning_rate": 1.227838333989088e-07, "logits/chosen": -1.806662917137146, "logits/rejected": -1.1421974897384644, "logps/chosen": -230.1032257080078, "logps/ref_chosen": -96.7739028930664, "logps/ref_rejected": -86.40473937988281, "logps/rejected": -283.23822021484375, "loss": 0.9275, "margin_dpo/margin_mean": 63.504150390625, "margin_dpo/margin_std": 75.69766235351562, "step": 480 }, { "epoch": 0.7063142437591777, "fcm_dpo/beta": 0.012597526423633099, "fcm_dpo/delta": -0.020995954051613808, "fcm_dpo/margin": 68.99488830566406, "fcm_dpo/q_t": 0.32530438899993896, "grad_norm": 14.493057250976562, "learning_rate": 1.2168076391719489e-07, "logits/chosen": -1.962436318397522, "logits/rejected": -1.5211691856384277, "logps/chosen": -224.07144165039062, "logps/ref_chosen": -91.670166015625, "logps/ref_rejected": -98.69490051269531, "logps/rejected": -300.091064453125, "loss": 0.8948, "margin_dpo/margin_mean": 68.99488830566406, "margin_dpo/margin_std": 78.63343048095703, "step": 481 }, { "epoch": 0.7077826725403817, "fcm_dpo/beta": 0.012524547055363655, "fcm_dpo/delta": 0.005113374441862106, "fcm_dpo/margin": 67.46731567382812, "fcm_dpo/q_t": 0.3271167278289795, "grad_norm": 13.560710906982422, "learning_rate": 1.2058107576668938e-07, "logits/chosen": -1.5037896633148193, "logits/rejected": -1.1416594982147217, "logps/chosen": -226.85171508789062, "logps/ref_chosen": -98.52011108398438, "logps/ref_rejected": -94.82945251464844, "logps/rejected": -290.62835693359375, "loss": 0.8862, "margin_dpo/margin_mean": 67.46731567382812, "margin_dpo/margin_std": 73.31199645996094, "step": 482 }, { "epoch": 0.7092511013215859, "fcm_dpo/beta": 0.01257384568452835, "fcm_dpo/delta": -0.05768094211816788, "fcm_dpo/margin": 71.71238708496094, "fcm_dpo/q_t": 0.315372496843338, "grad_norm": 12.017552375793457, "learning_rate": 1.194847979251979e-07, "logits/chosen": -2.2574076652526855, "logits/rejected": -1.6553820371627808, "logps/chosen": -239.170654296875, "logps/ref_chosen": -107.11860656738281, "logps/ref_rejected": -101.11499786376953, "logps/rejected": -304.87945556640625, "loss": 0.8387, "margin_dpo/margin_mean": 71.71238708496094, "margin_dpo/margin_std": 70.38067626953125, "step": 483 }, { "epoch": 0.71071953010279, "fcm_dpo/beta": 0.012338344939053059, "fcm_dpo/delta": -0.10980929434299469, "fcm_dpo/margin": 76.90882873535156, "fcm_dpo/q_t": 0.31271493434906006, "grad_norm": 11.736170768737793, "learning_rate": 1.1839195928066101e-07, "logits/chosen": -2.2439277172088623, "logits/rejected": -1.7512518167495728, "logps/chosen": -212.62677001953125, "logps/ref_chosen": -86.97991943359375, "logps/ref_rejected": -90.72367095947266, "logps/rejected": -293.27935791015625, "loss": 0.8462, "margin_dpo/margin_mean": 76.90882873535156, "margin_dpo/margin_std": 79.86015319824219, "step": 484 }, { "epoch": 0.7121879588839941, "fcm_dpo/beta": 0.011784037575125694, "fcm_dpo/delta": -0.212799072265625, "fcm_dpo/margin": 88.29972839355469, "fcm_dpo/q_t": 0.28984811902046204, "grad_norm": 11.503626823425293, "learning_rate": 1.1730258863039347e-07, "logits/chosen": -1.5688313245773315, "logits/rejected": -1.1244767904281616, "logps/chosen": -206.88958740234375, "logps/ref_chosen": -94.05874633789062, "logps/ref_rejected": -108.56298828125, "logps/rejected": -309.69354248046875, "loss": 0.7495, "margin_dpo/margin_mean": 88.29972839355469, "margin_dpo/margin_std": 75.29219818115234, "step": 485 }, { "epoch": 0.7136563876651982, "fcm_dpo/beta": 0.011598912999033928, "fcm_dpo/delta": -0.06779532879590988, "fcm_dpo/margin": 78.58279418945312, "fcm_dpo/q_t": 0.31678569316864014, "grad_norm": 13.836838722229004, "learning_rate": 1.1621671468032493e-07, "logits/chosen": -1.9958369731903076, "logits/rejected": -1.2726597785949707, "logps/chosen": -223.26283264160156, "logps/ref_chosen": -93.74588012695312, "logps/ref_rejected": -98.07064819335938, "logps/rejected": -306.17041015625, "loss": 0.8536, "margin_dpo/margin_mean": 78.58279418945312, "margin_dpo/margin_std": 80.64280700683594, "step": 486 }, { "epoch": 0.7151248164464024, "fcm_dpo/beta": 0.011509329080581665, "fcm_dpo/delta": 0.025798970833420753, "fcm_dpo/margin": 71.78614807128906, "fcm_dpo/q_t": 0.3311973214149475, "grad_norm": 12.209793090820312, "learning_rate": 1.1513436604424378e-07, "logits/chosen": -2.163451910018921, "logits/rejected": -1.856655240058899, "logps/chosen": -215.14894104003906, "logps/ref_chosen": -88.0335693359375, "logps/ref_rejected": -98.47209930419922, "logps/rejected": -297.3736267089844, "loss": 0.8644, "margin_dpo/margin_mean": 71.78614807128906, "margin_dpo/margin_std": 73.53408813476562, "step": 487 }, { "epoch": 0.7165932452276065, "fcm_dpo/beta": 0.011546431109309196, "fcm_dpo/delta": -0.007139893714338541, "fcm_dpo/margin": 74.18093872070312, "fcm_dpo/q_t": 0.31756943464279175, "grad_norm": 13.333039283752441, "learning_rate": 1.1405557124304335e-07, "logits/chosen": -1.5974540710449219, "logits/rejected": -1.347166657447815, "logps/chosen": -205.8610076904297, "logps/ref_chosen": -84.78964233398438, "logps/ref_rejected": -90.2734603881836, "logps/rejected": -285.5257568359375, "loss": 0.8218, "margin_dpo/margin_mean": 74.18093872070312, "margin_dpo/margin_std": 63.89453887939453, "step": 488 }, { "epoch": 0.7180616740088106, "fcm_dpo/beta": 0.011623598635196686, "fcm_dpo/delta": 0.0482901856303215, "fcm_dpo/margin": 69.30853271484375, "fcm_dpo/q_t": 0.3355333209037781, "grad_norm": 12.906960487365723, "learning_rate": 1.1298035870396985e-07, "logits/chosen": -1.6089611053466797, "logits/rejected": -1.1197456121444702, "logps/chosen": -221.5333709716797, "logps/ref_chosen": -90.46929931640625, "logps/ref_rejected": -86.39761352539062, "logps/rejected": -286.77020263671875, "loss": 0.8827, "margin_dpo/margin_mean": 69.30853271484375, "margin_dpo/margin_std": 73.18409729003906, "step": 489 }, { "epoch": 0.7195301027900147, "fcm_dpo/beta": 0.012023422867059708, "fcm_dpo/delta": 0.13658013939857483, "fcm_dpo/margin": 59.857852935791016, "fcm_dpo/q_t": 0.3625268340110779, "grad_norm": 13.699275016784668, "learning_rate": 1.1190875675987355e-07, "logits/chosen": -2.1816554069519043, "logits/rejected": -2.236203193664551, "logps/chosen": -222.6205291748047, "logps/ref_chosen": -85.32012939453125, "logps/ref_rejected": -115.99385070800781, "logps/rejected": -313.152099609375, "loss": 1.0544, "margin_dpo/margin_mean": 59.857852935791016, "margin_dpo/margin_std": 92.6108627319336, "step": 490 }, { "epoch": 0.7209985315712188, "fcm_dpo/beta": 0.012138845399022102, "fcm_dpo/delta": 0.09064659476280212, "fcm_dpo/margin": 63.129661560058594, "fcm_dpo/q_t": 0.34301620721817017, "grad_norm": 13.435876846313477, "learning_rate": 1.1084079364846241e-07, "logits/chosen": -2.065945625305176, "logits/rejected": -1.585937738418579, "logps/chosen": -220.5257568359375, "logps/ref_chosen": -86.14352416992188, "logps/ref_rejected": -80.67945861816406, "logps/rejected": -278.19134521484375, "loss": 0.9492, "margin_dpo/margin_mean": 63.129661560058594, "margin_dpo/margin_std": 78.26299285888672, "step": 491 }, { "epoch": 0.7224669603524229, "fcm_dpo/beta": 0.012580521404743195, "fcm_dpo/delta": 0.20928707718849182, "fcm_dpo/margin": 52.022743225097656, "fcm_dpo/q_t": 0.3731645345687866, "grad_norm": 17.193553924560547, "learning_rate": 1.097764975115576e-07, "logits/chosen": -2.168747901916504, "logits/rejected": -1.7894206047058105, "logps/chosen": -206.310791015625, "logps/ref_chosen": -81.10757446289062, "logps/ref_rejected": -80.75199890136719, "logps/rejected": -257.97796630859375, "loss": 1.1032, "margin_dpo/margin_mean": 52.02274703979492, "margin_dpo/margin_std": 90.33053588867188, "step": 492 }, { "epoch": 0.723935389133627, "fcm_dpo/beta": 0.012747792527079582, "fcm_dpo/delta": 0.06546922028064728, "fcm_dpo/margin": 61.89603805541992, "fcm_dpo/q_t": 0.3345375061035156, "grad_norm": 15.1096773147583, "learning_rate": 1.0871589639435203e-07, "logits/chosen": -2.51515793800354, "logits/rejected": -1.8059933185577393, "logps/chosen": -238.19631958007812, "logps/ref_chosen": -112.20733642578125, "logps/ref_rejected": -93.60719299316406, "logps/rejected": -281.4921875, "loss": 0.8945, "margin_dpo/margin_mean": 61.89603805541992, "margin_dpo/margin_std": 64.05390167236328, "step": 493 }, { "epoch": 0.7254038179148311, "fcm_dpo/beta": 0.012348496355116367, "fcm_dpo/delta": -0.22160913050174713, "fcm_dpo/margin": 84.58407592773438, "fcm_dpo/q_t": 0.2897966206073761, "grad_norm": 13.276835441589355, "learning_rate": 1.0765901824467166e-07, "logits/chosen": -1.4075872898101807, "logits/rejected": -1.3157413005828857, "logps/chosen": -194.07228088378906, "logps/ref_chosen": -73.11489868164062, "logps/ref_rejected": -92.16300201416016, "logps/rejected": -297.7044677734375, "loss": 0.7764, "margin_dpo/margin_mean": 84.58407592773438, "margin_dpo/margin_std": 76.84968566894531, "step": 494 }, { "epoch": 0.7268722466960352, "fcm_dpo/beta": 0.01227477751672268, "fcm_dpo/delta": -0.024101722985506058, "fcm_dpo/margin": 71.03709411621094, "fcm_dpo/q_t": 0.3242284059524536, "grad_norm": 13.635724067687988, "learning_rate": 1.0660589091223854e-07, "logits/chosen": -2.0604519844055176, "logits/rejected": -1.5982627868652344, "logps/chosen": -220.474853515625, "logps/ref_chosen": -99.52031707763672, "logps/ref_rejected": -97.93089294433594, "logps/rejected": -289.92254638671875, "loss": 0.8962, "margin_dpo/margin_mean": 71.03709411621094, "margin_dpo/margin_std": 81.31465911865234, "step": 495 }, { "epoch": 0.7283406754772394, "fcm_dpo/beta": 0.012515192851424217, "fcm_dpo/delta": 0.1012970507144928, "fcm_dpo/margin": 60.364662170410156, "fcm_dpo/q_t": 0.3426893353462219, "grad_norm": 16.073570251464844, "learning_rate": 1.0555654214793722e-07, "logits/chosen": -2.1137051582336426, "logits/rejected": -1.4734668731689453, "logps/chosen": -240.74362182617188, "logps/ref_chosen": -107.85675048828125, "logps/ref_rejected": -92.77056121826172, "logps/rejected": -286.0220947265625, "loss": 0.9345, "margin_dpo/margin_mean": 60.36465835571289, "margin_dpo/margin_std": 70.0220947265625, "step": 496 }, { "epoch": 0.7298091042584435, "fcm_dpo/beta": 0.013011728413403034, "fcm_dpo/delta": 0.278378963470459, "fcm_dpo/margin": 45.2384147644043, "fcm_dpo/q_t": 0.38457757234573364, "grad_norm": 16.124645233154297, "learning_rate": 1.0451099960308374e-07, "logits/chosen": -1.7485501766204834, "logits/rejected": -1.3026444911956787, "logps/chosen": -238.62753295898438, "logps/ref_chosen": -92.08322143554688, "logps/ref_rejected": -81.79503631591797, "logps/rejected": -273.5777587890625, "loss": 1.0902, "margin_dpo/margin_mean": 45.23841094970703, "margin_dpo/margin_std": 75.66600799560547, "step": 497 }, { "epoch": 0.7312775330396476, "fcm_dpo/beta": 0.0131129901856184, "fcm_dpo/delta": -0.06823568046092987, "fcm_dpo/margin": 69.55242156982422, "fcm_dpo/q_t": 0.31473374366760254, "grad_norm": 14.976054191589355, "learning_rate": 1.0346929082869641e-07, "logits/chosen": -1.6781526803970337, "logits/rejected": -1.1911282539367676, "logps/chosen": -218.24331665039062, "logps/ref_chosen": -98.19437408447266, "logps/ref_rejected": -90.68745422363281, "logps/rejected": -280.288818359375, "loss": 0.8364, "margin_dpo/margin_mean": 69.55242919921875, "margin_dpo/margin_std": 69.84315490722656, "step": 498 }, { "epoch": 0.7327459618208517, "fcm_dpo/beta": 0.01302518229931593, "fcm_dpo/delta": -0.05266323313117027, "fcm_dpo/margin": 68.91853332519531, "fcm_dpo/q_t": 0.32476356625556946, "grad_norm": 18.89053726196289, "learning_rate": 1.0243144327477013e-07, "logits/chosen": -2.098900318145752, "logits/rejected": -2.0415899753570557, "logps/chosen": -211.9065704345703, "logps/ref_chosen": -81.0399169921875, "logps/ref_rejected": -106.92170715332031, "logps/rejected": -306.7069091796875, "loss": 0.9069, "margin_dpo/margin_mean": 68.91853332519531, "margin_dpo/margin_std": 82.49652099609375, "step": 499 }, { "epoch": 0.7342143906020558, "fcm_dpo/beta": 0.01297258585691452, "fcm_dpo/delta": 0.03623470664024353, "fcm_dpo/margin": 62.95831298828125, "fcm_dpo/q_t": 0.3484196364879608, "grad_norm": 16.346345901489258, "learning_rate": 1.0139748428955333e-07, "logits/chosen": -1.630037546157837, "logits/rejected": -1.4560880661010742, "logps/chosen": -227.47012329101562, "logps/ref_chosen": -89.248046875, "logps/ref_rejected": -100.41021728515625, "logps/rejected": -301.590576171875, "loss": 0.9588, "margin_dpo/margin_mean": 62.95831298828125, "margin_dpo/margin_std": 83.92835235595703, "step": 500 }, { "epoch": 0.7342143906020558, "eval_fcm_dpo/beta": 0.013016212731599808, "eval_logits/chosen": -1.8269517421722412, "eval_logits/rejected": -1.456815481185913, "eval_logps/chosen": -233.72581481933594, "eval_logps/ref_chosen": -100.49356842041016, "eval_logps/ref_rejected": -94.06775665283203, "eval_logps/rejected": -285.7688903808594, "eval_loss": 0.4848414361476898, "eval_margin_dpo/margin_mean": 58.46889877319336, "eval_margin_dpo/margin_std": 75.4513168334961, "eval_runtime": 43.7742, "eval_samples_per_second": 53.433, "eval_steps_per_second": 1.69, "step": 500 }, { "epoch": 0.73568281938326, "fcm_dpo/beta": 0.013117056339979172, "fcm_dpo/delta": -0.04707694798707962, "fcm_dpo/margin": 67.81106567382812, "fcm_dpo/q_t": 0.34009435772895813, "grad_norm": 17.60281753540039, "learning_rate": 1.0036744111882672e-07, "logits/chosen": -1.5807795524597168, "logits/rejected": -1.1496039628982544, "logps/chosen": -212.7261199951172, "logps/ref_chosen": -90.596923828125, "logps/ref_rejected": -85.2972640991211, "logps/rejected": -275.237548828125, "loss": 0.9848, "margin_dpo/margin_mean": 67.81106567382812, "margin_dpo/margin_std": 93.16748809814453, "step": 501 }, { "epoch": 0.737151248164464, "fcm_dpo/beta": 0.013137854635715485, "fcm_dpo/delta": 0.034719593822956085, "fcm_dpo/margin": 61.971439361572266, "fcm_dpo/q_t": 0.34164291620254517, "grad_norm": 16.562210083007812, "learning_rate": 9.934134090518592e-08, "logits/chosen": -1.7543015480041504, "logits/rejected": -1.0072808265686035, "logps/chosen": -227.01095581054688, "logps/ref_chosen": -103.47073364257812, "logps/ref_rejected": -89.04415893554688, "logps/rejected": -274.55584716796875, "loss": 0.9784, "margin_dpo/margin_mean": 61.971435546875, "margin_dpo/margin_std": 80.59944152832031, "step": 502 }, { "epoch": 0.7386196769456681, "fcm_dpo/beta": 0.012945367023348808, "fcm_dpo/delta": -0.0793260931968689, "fcm_dpo/margin": 71.16193389892578, "fcm_dpo/q_t": 0.3206115961074829, "grad_norm": 14.09860610961914, "learning_rate": 9.831921068732571e-08, "logits/chosen": -1.8249884843826294, "logits/rejected": -1.3317406177520752, "logps/chosen": -212.64935302734375, "logps/ref_chosen": -93.72998809814453, "logps/ref_rejected": -87.60896301269531, "logps/rejected": -277.69024658203125, "loss": 0.8949, "margin_dpo/margin_mean": 71.16192626953125, "margin_dpo/margin_std": 81.064208984375, "step": 503 }, { "epoch": 0.7400881057268722, "fcm_dpo/beta": 0.01244557648897171, "fcm_dpo/delta": -0.20257821679115295, "fcm_dpo/margin": 82.92597961425781, "fcm_dpo/q_t": 0.29116249084472656, "grad_norm": 16.30722999572754, "learning_rate": 9.730107739932805e-08, "logits/chosen": -1.9234204292297363, "logits/rejected": -1.5795867443084717, "logps/chosen": -212.29249572753906, "logps/ref_chosen": -92.2918701171875, "logps/ref_rejected": -109.03765106201172, "logps/rejected": -311.9642333984375, "loss": 0.7785, "margin_dpo/margin_mean": 82.92598724365234, "margin_dpo/margin_std": 76.13577270507812, "step": 504 }, { "epoch": 0.7415565345080763, "fcm_dpo/beta": 0.012345531024038792, "fcm_dpo/delta": 0.030347377061843872, "fcm_dpo/margin": 66.58810424804688, "fcm_dpo/q_t": 0.33173486590385437, "grad_norm": 13.986702919006348, "learning_rate": 9.628696786995188e-08, "logits/chosen": -2.3997888565063477, "logits/rejected": -1.855478048324585, "logps/chosen": -240.9890594482422, "logps/ref_chosen": -106.27693176269531, "logps/ref_rejected": -94.20611572265625, "logps/rejected": -295.50634765625, "loss": 0.8808, "margin_dpo/margin_mean": 66.58810424804688, "margin_dpo/margin_std": 69.91239929199219, "step": 505 }, { "epoch": 0.7430249632892805, "fcm_dpo/beta": 0.012313488870859146, "fcm_dpo/delta": 0.00921265035867691, "fcm_dpo/margin": 68.321533203125, "fcm_dpo/q_t": 0.32348087430000305, "grad_norm": 13.006158828735352, "learning_rate": 9.527690882192635e-08, "logits/chosen": -1.9842698574066162, "logits/rejected": -1.6244713068008423, "logps/chosen": -212.46749877929688, "logps/ref_chosen": -81.44086456298828, "logps/ref_rejected": -85.66439819335938, "logps/rejected": -285.0125427246094, "loss": 0.8633, "margin_dpo/margin_mean": 68.321533203125, "margin_dpo/margin_std": 67.43982696533203, "step": 506 }, { "epoch": 0.7444933920704846, "fcm_dpo/beta": 0.01229831762611866, "fcm_dpo/delta": -0.052599359303712845, "fcm_dpo/margin": 73.00973510742188, "fcm_dpo/q_t": 0.3235676884651184, "grad_norm": 13.053921699523926, "learning_rate": 9.427092687124691e-08, "logits/chosen": -2.104893207550049, "logits/rejected": -1.6638824939727783, "logps/chosen": -234.46234130859375, "logps/ref_chosen": -99.42694854736328, "logps/ref_rejected": -102.25296020507812, "logps/rejected": -310.298095703125, "loss": 0.8921, "margin_dpo/margin_mean": 73.00973510742188, "margin_dpo/margin_std": 84.01727294921875, "step": 507 }, { "epoch": 0.7459618208516887, "fcm_dpo/beta": 0.012360332533717155, "fcm_dpo/delta": 0.08519750088453293, "fcm_dpo/margin": 62.41271209716797, "fcm_dpo/q_t": 0.35167741775512695, "grad_norm": 14.7086763381958, "learning_rate": 9.326904852647344e-08, "logits/chosen": -2.081237554550171, "logits/rejected": -1.6077560186386108, "logps/chosen": -230.1769256591797, "logps/ref_chosen": -97.9156494140625, "logps/ref_rejected": -103.79782104492188, "logps/rejected": -298.4718017578125, "loss": 0.9813, "margin_dpo/margin_mean": 62.4127197265625, "margin_dpo/margin_std": 83.03570556640625, "step": 508 }, { "epoch": 0.7474302496328928, "fcm_dpo/beta": 0.012669201008975506, "fcm_dpo/delta": 0.0940362885594368, "fcm_dpo/margin": 60.193206787109375, "fcm_dpo/q_t": 0.3435523509979248, "grad_norm": 17.660673141479492, "learning_rate": 9.227130018803195e-08, "logits/chosen": -1.9336552619934082, "logits/rejected": -1.2759418487548828, "logps/chosen": -226.89523315429688, "logps/ref_chosen": -100.72984313964844, "logps/ref_rejected": -86.943359375, "logps/rejected": -273.3019714355469, "loss": 0.9661, "margin_dpo/margin_mean": 60.193206787109375, "margin_dpo/margin_std": 76.53768920898438, "step": 509 }, { "epoch": 0.748898678414097, "fcm_dpo/beta": 0.012600088492035866, "fcm_dpo/delta": -0.0554606169462204, "fcm_dpo/margin": 71.46772766113281, "fcm_dpo/q_t": 0.3216525912284851, "grad_norm": 15.701079368591309, "learning_rate": 9.127770814751932e-08, "logits/chosen": -1.5974844694137573, "logits/rejected": -1.398494005203247, "logps/chosen": -234.45346069335938, "logps/ref_chosen": -89.16938781738281, "logps/ref_rejected": -107.91939544677734, "logps/rejected": -324.67120361328125, "loss": 0.8857, "margin_dpo/margin_mean": 71.46772766113281, "margin_dpo/margin_std": 79.08045196533203, "step": 510 }, { "epoch": 0.750367107195301, "fcm_dpo/beta": 0.012481224723160267, "fcm_dpo/delta": -0.06006241962313652, "fcm_dpo/margin": 72.47528076171875, "fcm_dpo/q_t": 0.32413989305496216, "grad_norm": 13.066450119018555, "learning_rate": 9.028829858700973e-08, "logits/chosen": -2.066761016845703, "logits/rejected": -1.504467248916626, "logps/chosen": -223.00457763671875, "logps/ref_chosen": -89.74979400634766, "logps/ref_rejected": -97.73127746582031, "logps/rejected": -303.4613342285156, "loss": 0.863, "margin_dpo/margin_mean": 72.47528076171875, "margin_dpo/margin_std": 80.30681610107422, "step": 511 }, { "epoch": 0.7518355359765051, "fcm_dpo/beta": 0.01244446448981762, "fcm_dpo/delta": -0.025120899081230164, "fcm_dpo/margin": 70.07406616210938, "fcm_dpo/q_t": 0.3278118968009949, "grad_norm": 15.821978569030762, "learning_rate": 8.930309757836516e-08, "logits/chosen": -1.7284451723098755, "logits/rejected": -1.2756907939910889, "logps/chosen": -213.42648315429688, "logps/ref_chosen": -85.73515319824219, "logps/ref_rejected": -87.6248779296875, "logps/rejected": -285.3902587890625, "loss": 0.9145, "margin_dpo/margin_mean": 70.07405853271484, "margin_dpo/margin_std": 83.03077697753906, "step": 512 }, { "epoch": 0.7533039647577092, "fcm_dpo/beta": 0.012370715849101543, "fcm_dpo/delta": 0.07577605545520782, "fcm_dpo/margin": 63.00016403198242, "fcm_dpo/q_t": 0.3346492052078247, "grad_norm": 14.578227043151855, "learning_rate": 8.832213108254863e-08, "logits/chosen": -2.4033966064453125, "logits/rejected": -1.7322185039520264, "logps/chosen": -229.25015258789062, "logps/ref_chosen": -96.19050598144531, "logps/ref_rejected": -82.91773986816406, "logps/rejected": -278.9775390625, "loss": 0.919, "margin_dpo/margin_mean": 63.00016403198242, "margin_dpo/margin_std": 69.4985580444336, "step": 513 }, { "epoch": 0.7547723935389133, "fcm_dpo/beta": 0.012548575177788734, "fcm_dpo/delta": 0.03780820220708847, "fcm_dpo/margin": 64.95950317382812, "fcm_dpo/q_t": 0.3395010828971863, "grad_norm": 15.710688591003418, "learning_rate": 8.734542494893954e-08, "logits/chosen": -1.7342252731323242, "logits/rejected": -1.1726717948913574, "logps/chosen": -225.5660858154297, "logps/ref_chosen": -97.82363891601562, "logps/ref_rejected": -85.36982727050781, "logps/rejected": -278.07177734375, "loss": 0.9158, "margin_dpo/margin_mean": 64.95950317382812, "margin_dpo/margin_std": 76.12434387207031, "step": 514 }, { "epoch": 0.7562408223201175, "fcm_dpo/beta": 0.012832360342144966, "fcm_dpo/delta": 0.13085728883743286, "fcm_dpo/margin": 56.79905319213867, "fcm_dpo/q_t": 0.35357731580734253, "grad_norm": 19.61615753173828, "learning_rate": 8.637300491465272e-08, "logits/chosen": -1.9679467678070068, "logits/rejected": -1.7441508769989014, "logps/chosen": -210.46856689453125, "logps/ref_chosen": -81.93355560302734, "logps/ref_rejected": -93.57656860351562, "logps/rejected": -278.91064453125, "loss": 0.9894, "margin_dpo/margin_mean": 56.79905319213867, "margin_dpo/margin_std": 76.72676086425781, "step": 515 }, { "epoch": 0.7577092511013216, "fcm_dpo/beta": 0.01302811224013567, "fcm_dpo/delta": 0.023421771824359894, "fcm_dpo/margin": 63.583045959472656, "fcm_dpo/q_t": 0.34561917185783386, "grad_norm": 15.823124885559082, "learning_rate": 8.540489660386064e-08, "logits/chosen": -2.2833476066589355, "logits/rejected": -2.0779836177825928, "logps/chosen": -224.8201141357422, "logps/ref_chosen": -97.09166717529297, "logps/ref_rejected": -118.55667114257812, "logps/rejected": -309.8681640625, "loss": 1.0038, "margin_dpo/margin_mean": 63.583045959472656, "margin_dpo/margin_std": 91.61796569824219, "step": 516 }, { "epoch": 0.7591776798825257, "fcm_dpo/beta": 0.012637123465538025, "fcm_dpo/delta": -0.1259511113166809, "fcm_dpo/margin": 75.99790954589844, "fcm_dpo/q_t": 0.31130844354629517, "grad_norm": 14.500532150268555, "learning_rate": 8.444112552711752e-08, "logits/chosen": -1.8676999807357788, "logits/rejected": -1.4454941749572754, "logps/chosen": -229.08517456054688, "logps/ref_chosen": -93.18656921386719, "logps/ref_rejected": -95.5927963256836, "logps/rejected": -307.48931884765625, "loss": 0.8434, "margin_dpo/margin_mean": 75.99791717529297, "margin_dpo/margin_std": 77.92581176757812, "step": 517 }, { "epoch": 0.7606461086637298, "fcm_dpo/beta": 0.012939132750034332, "fcm_dpo/delta": 0.1308937817811966, "fcm_dpo/margin": 56.31306838989258, "fcm_dpo/q_t": 0.35669273138046265, "grad_norm": 17.170347213745117, "learning_rate": 8.348171708068747e-08, "logits/chosen": -1.8192179203033447, "logits/rejected": -1.4760829210281372, "logps/chosen": -229.3511962890625, "logps/ref_chosen": -87.0213623046875, "logps/ref_rejected": -91.24456024169922, "logps/rejected": -289.887451171875, "loss": 0.9965, "margin_dpo/margin_mean": 56.313072204589844, "margin_dpo/margin_std": 78.62522888183594, "step": 518 }, { "epoch": 0.762114537444934, "fcm_dpo/beta": 0.012952839955687523, "fcm_dpo/delta": 0.02188480645418167, "fcm_dpo/margin": 63.96747970581055, "fcm_dpo/q_t": 0.33184218406677246, "grad_norm": 13.27155590057373, "learning_rate": 8.25266965458755e-08, "logits/chosen": -1.8997950553894043, "logits/rejected": -1.5174694061279297, "logps/chosen": -232.2674560546875, "logps/ref_chosen": -100.68755340576172, "logps/ref_rejected": -109.03042602539062, "logps/rejected": -304.5777893066406, "loss": 0.8681, "margin_dpo/margin_mean": 63.96747589111328, "margin_dpo/margin_std": 64.88044738769531, "step": 519 }, { "epoch": 0.7635829662261381, "fcm_dpo/beta": 0.012810420244932175, "fcm_dpo/delta": -0.10472933202981949, "fcm_dpo/margin": 73.66597747802734, "fcm_dpo/q_t": 0.3046000599861145, "grad_norm": 14.583555221557617, "learning_rate": 8.15760890883607e-08, "logits/chosen": -1.6726075410842896, "logits/rejected": -1.24037504196167, "logps/chosen": -218.2542724609375, "logps/ref_chosen": -98.77762603759766, "logps/ref_rejected": -106.9457015991211, "logps/rejected": -300.08831787109375, "loss": 0.7999, "margin_dpo/margin_mean": 73.66596984863281, "margin_dpo/margin_std": 66.00288391113281, "step": 520 }, { "epoch": 0.7650513950073421, "fcm_dpo/beta": 0.012788889929652214, "fcm_dpo/delta": -0.056156057864427567, "fcm_dpo/margin": 70.42420959472656, "fcm_dpo/q_t": 0.3153071999549866, "grad_norm": 14.226971626281738, "learning_rate": 8.062991975753378e-08, "logits/chosen": -2.0882833003997803, "logits/rejected": -1.7669947147369385, "logps/chosen": -214.2764129638672, "logps/ref_chosen": -86.35108947753906, "logps/ref_rejected": -90.39317321777344, "logps/rejected": -288.7427062988281, "loss": 0.8338, "margin_dpo/margin_mean": 70.42420959472656, "margin_dpo/margin_std": 67.74490356445312, "step": 521 }, { "epoch": 0.7665198237885462, "fcm_dpo/beta": 0.012705507688224316, "fcm_dpo/delta": -0.01384912058711052, "fcm_dpo/margin": 67.86552429199219, "fcm_dpo/q_t": 0.3353983163833618, "grad_norm": 14.730003356933594, "learning_rate": 7.968821348583643e-08, "logits/chosen": -1.7902132272720337, "logits/rejected": -1.4231998920440674, "logps/chosen": -204.54205322265625, "logps/ref_chosen": -76.008056640625, "logps/ref_rejected": -74.44125366210938, "logps/rejected": -270.84075927734375, "loss": 0.9099, "margin_dpo/margin_mean": 67.86552429199219, "margin_dpo/margin_std": 81.35099792480469, "step": 522 }, { "epoch": 0.7679882525697503, "fcm_dpo/beta": 0.013018842786550522, "fcm_dpo/delta": 0.22799676656723022, "fcm_dpo/margin": 48.934547424316406, "fcm_dpo/q_t": 0.3743229806423187, "grad_norm": 20.326669692993164, "learning_rate": 7.875099508810484e-08, "logits/chosen": -2.163761615753174, "logits/rejected": -1.4758130311965942, "logps/chosen": -242.2533721923828, "logps/ref_chosen": -96.49627685546875, "logps/ref_rejected": -91.0152816772461, "logps/rejected": -285.7069396972656, "loss": 1.0867, "margin_dpo/margin_mean": 48.934547424316406, "margin_dpo/margin_std": 81.15193176269531, "step": 523 }, { "epoch": 0.7694566813509545, "fcm_dpo/beta": 0.013128972612321377, "fcm_dpo/delta": -0.031645577400922775, "fcm_dpo/margin": 66.92581176757812, "fcm_dpo/q_t": 0.3286508023738861, "grad_norm": 15.882726669311523, "learning_rate": 7.781828926091535e-08, "logits/chosen": -2.137564182281494, "logits/rejected": -1.4685100317001343, "logps/chosen": -236.1751708984375, "logps/ref_chosen": -103.40550994873047, "logps/ref_rejected": -87.91015625, "logps/rejected": -287.60565185546875, "loss": 0.8876, "margin_dpo/margin_mean": 66.92581939697266, "margin_dpo/margin_std": 76.40330505371094, "step": 524 }, { "epoch": 0.7709251101321586, "fcm_dpo/beta": 0.012723056599497795, "fcm_dpo/delta": -0.09891921281814575, "fcm_dpo/margin": 73.25902557373047, "fcm_dpo/q_t": 0.3239433765411377, "grad_norm": 13.863753318786621, "learning_rate": 7.689012058193384e-08, "logits/chosen": -1.6756742000579834, "logits/rejected": -1.5884668827056885, "logps/chosen": -208.05990600585938, "logps/ref_chosen": -84.20648193359375, "logps/ref_rejected": -106.26158905029297, "logps/rejected": -303.3740234375, "loss": 0.8701, "margin_dpo/margin_mean": 73.25902557373047, "margin_dpo/margin_std": 79.67434692382812, "step": 525 }, { "epoch": 0.7723935389133627, "fcm_dpo/beta": 0.012752903625369072, "fcm_dpo/delta": -0.08910477161407471, "fcm_dpo/margin": 72.99378967285156, "fcm_dpo/q_t": 0.3187520205974579, "grad_norm": 12.174683570861816, "learning_rate": 7.596651350926836e-08, "logits/chosen": -1.6931428909301758, "logits/rejected": -1.0449802875518799, "logps/chosen": -234.44082641601562, "logps/ref_chosen": -102.85411071777344, "logps/ref_rejected": -93.65145874023438, "logps/rejected": -298.23199462890625, "loss": 0.8781, "margin_dpo/margin_mean": 72.99378967285156, "margin_dpo/margin_std": 82.09791564941406, "step": 526 }, { "epoch": 0.7738619676945668, "fcm_dpo/beta": 0.012565705925226212, "fcm_dpo/delta": -0.046583615243434906, "fcm_dpo/margin": 71.01980590820312, "fcm_dpo/q_t": 0.317333459854126, "grad_norm": 12.275128364562988, "learning_rate": 7.504749238082414e-08, "logits/chosen": -3.2490463256835938, "logits/rejected": -2.4732651710510254, "logps/chosen": -226.13568115234375, "logps/ref_chosen": -96.76543426513672, "logps/ref_rejected": -85.39518737792969, "logps/rejected": -285.78521728515625, "loss": 0.8292, "margin_dpo/margin_mean": 71.01980590820312, "margin_dpo/margin_std": 68.41850280761719, "step": 527 }, { "epoch": 0.775330396475771, "fcm_dpo/beta": 0.012453284114599228, "fcm_dpo/delta": -0.058921098709106445, "fcm_dpo/margin": 72.56159973144531, "fcm_dpo/q_t": 0.31946584582328796, "grad_norm": 13.592493057250977, "learning_rate": 7.413308141366254e-08, "logits/chosen": -2.2505486011505127, "logits/rejected": -1.7673816680908203, "logps/chosen": -215.16543579101562, "logps/ref_chosen": -94.1649169921875, "logps/ref_rejected": -100.372314453125, "logps/rejected": -293.9344482421875, "loss": 0.8619, "margin_dpo/margin_mean": 72.56159973144531, "margin_dpo/margin_std": 77.28302764892578, "step": 528 }, { "epoch": 0.7767988252569751, "fcm_dpo/beta": 0.012501591816544533, "fcm_dpo/delta": 0.05900266766548157, "fcm_dpo/margin": 63.64674758911133, "fcm_dpo/q_t": 0.34856054186820984, "grad_norm": 15.007195472717285, "learning_rate": 7.322330470336313e-08, "logits/chosen": -1.9466891288757324, "logits/rejected": -1.7657006978988647, "logps/chosen": -221.71871948242188, "logps/ref_chosen": -80.77680206298828, "logps/ref_rejected": -97.191162109375, "logps/rejected": -301.77984619140625, "loss": 0.9653, "margin_dpo/margin_mean": 63.64674758911133, "margin_dpo/margin_std": 85.80581665039062, "step": 529 }, { "epoch": 0.7782672540381792, "fcm_dpo/beta": 0.012481624260544777, "fcm_dpo/delta": 0.0072569213807582855, "fcm_dpo/margin": 67.52183532714844, "fcm_dpo/q_t": 0.3254541754722595, "grad_norm": 12.42430305480957, "learning_rate": 7.231818622338822e-08, "logits/chosen": -1.5578473806381226, "logits/rejected": -1.227747917175293, "logps/chosen": -209.4708251953125, "logps/ref_chosen": -79.7757339477539, "logps/ref_rejected": -92.05390167236328, "logps/rejected": -289.27081298828125, "loss": 0.8365, "margin_dpo/margin_mean": 67.5218276977539, "margin_dpo/margin_std": 62.72408676147461, "step": 530 }, { "epoch": 0.7797356828193832, "fcm_dpo/beta": 0.012329411692917347, "fcm_dpo/delta": -0.1382937878370285, "fcm_dpo/margin": 79.07818603515625, "fcm_dpo/q_t": 0.2973644733428955, "grad_norm": 11.442774772644043, "learning_rate": 7.141774982445147e-08, "logits/chosen": -1.8888896703720093, "logits/rejected": -1.2946207523345947, "logps/chosen": -206.50308227539062, "logps/ref_chosen": -79.31919860839844, "logps/ref_rejected": -78.10832214355469, "logps/rejected": -284.37042236328125, "loss": 0.7856, "margin_dpo/margin_mean": 79.07820129394531, "margin_dpo/margin_std": 69.38872528076172, "step": 531 }, { "epoch": 0.7812041116005873, "fcm_dpo/beta": 0.01183237787336111, "fcm_dpo/delta": -0.23072879016399384, "fcm_dpo/margin": 89.30570983886719, "fcm_dpo/q_t": 0.28500646352767944, "grad_norm": 12.133696556091309, "learning_rate": 7.052201923388953e-08, "logits/chosen": -1.5650770664215088, "logits/rejected": -0.9464230537414551, "logps/chosen": -235.76779174804688, "logps/ref_chosen": -101.63691711425781, "logps/ref_rejected": -93.34539031982422, "logps/rejected": -316.781982421875, "loss": 0.7458, "margin_dpo/margin_mean": 89.30570983886719, "margin_dpo/margin_std": 74.34683227539062, "step": 532 }, { "epoch": 0.7826725403817915, "fcm_dpo/beta": 0.012062860652804375, "fcm_dpo/delta": 0.2239835560321808, "fcm_dpo/margin": 53.07807159423828, "fcm_dpo/q_t": 0.372709721326828, "grad_norm": 14.061158180236816, "learning_rate": 6.963101805503646e-08, "logits/chosen": -2.508441925048828, "logits/rejected": -1.9440903663635254, "logps/chosen": -220.5234832763672, "logps/ref_chosen": -90.74664306640625, "logps/ref_rejected": -82.88848876953125, "logps/rejected": -265.743408203125, "loss": 1.0218, "margin_dpo/margin_mean": 53.07807540893555, "margin_dpo/margin_std": 76.28727722167969, "step": 533 }, { "epoch": 0.7841409691629956, "fcm_dpo/beta": 0.012088856659829617, "fcm_dpo/delta": -0.07449840009212494, "fcm_dpo/margin": 75.91458129882812, "fcm_dpo/q_t": 0.3064717650413513, "grad_norm": 11.550370216369629, "learning_rate": 6.874476976660184e-08, "logits/chosen": -1.9885363578796387, "logits/rejected": -1.613828182220459, "logps/chosen": -205.0845947265625, "logps/ref_chosen": -86.6308364868164, "logps/ref_rejected": -84.87196350097656, "logps/rejected": -279.24029541015625, "loss": 0.7894, "margin_dpo/margin_mean": 75.91458129882812, "margin_dpo/margin_std": 63.02655029296875, "step": 534 }, { "epoch": 0.7856093979441997, "fcm_dpo/beta": 0.011910516768693924, "fcm_dpo/delta": -0.1172776147723198, "fcm_dpo/margin": 80.23560333251953, "fcm_dpo/q_t": 0.3084285855293274, "grad_norm": 12.470861434936523, "learning_rate": 6.786329772205246e-08, "logits/chosen": -1.8749089241027832, "logits/rejected": -1.6865028142929077, "logps/chosen": -207.06907653808594, "logps/ref_chosen": -90.34539031982422, "logps/ref_rejected": -100.7779769897461, "logps/rejected": -297.7372741699219, "loss": 0.826, "margin_dpo/margin_mean": 80.23561096191406, "margin_dpo/margin_std": 79.02330017089844, "step": 535 }, { "epoch": 0.7870778267254038, "fcm_dpo/beta": 0.011703304946422577, "fcm_dpo/delta": -0.013192192651331425, "fcm_dpo/margin": 73.65609741210938, "fcm_dpo/q_t": 0.32144200801849365, "grad_norm": 12.106589317321777, "learning_rate": 6.698662514899638e-08, "logits/chosen": -2.0477590560913086, "logits/rejected": -1.803931713104248, "logps/chosen": -203.61679077148438, "logps/ref_chosen": -78.95956420898438, "logps/ref_rejected": -93.08779907226562, "logps/rejected": -291.401123046875, "loss": 0.8582, "margin_dpo/margin_mean": 73.65609741210938, "margin_dpo/margin_std": 74.89938354492188, "step": 536 }, { "epoch": 0.788546255506608, "fcm_dpo/beta": 0.011510086245834827, "fcm_dpo/delta": -0.182314932346344, "fcm_dpo/margin": 88.06642150878906, "fcm_dpo/q_t": 0.2944716215133667, "grad_norm": 9.836989402770996, "learning_rate": 6.611477514857114e-08, "logits/chosen": -1.6331989765167236, "logits/rejected": -0.941495418548584, "logps/chosen": -192.248779296875, "logps/ref_chosen": -86.70939636230469, "logps/ref_rejected": -78.90766906738281, "logps/rejected": -272.51348876953125, "loss": 0.7689, "margin_dpo/margin_mean": 88.06642150878906, "margin_dpo/margin_std": 75.45658874511719, "step": 537 }, { "epoch": 0.7900146842878121, "fcm_dpo/beta": 0.011464505456387997, "fcm_dpo/delta": 0.08413688838481903, "fcm_dpo/margin": 67.333984375, "fcm_dpo/q_t": 0.33640989661216736, "grad_norm": 13.633325576782227, "learning_rate": 6.524777069483525e-08, "logits/chosen": -1.8363103866577148, "logits/rejected": -1.3120193481445312, "logps/chosen": -238.7105712890625, "logps/ref_chosen": -98.44214630126953, "logps/ref_rejected": -91.09465026855469, "logps/rejected": -298.69708251953125, "loss": 0.874, "margin_dpo/margin_mean": 67.333984375, "margin_dpo/margin_std": 65.11286926269531, "step": 538 }, { "epoch": 0.7914831130690162, "fcm_dpo/beta": 0.01155300997197628, "fcm_dpo/delta": 0.028006993234157562, "fcm_dpo/margin": 71.33673095703125, "fcm_dpo/q_t": 0.33324581384658813, "grad_norm": 11.587471961975098, "learning_rate": 6.438563463416221e-08, "logits/chosen": -2.5877199172973633, "logits/rejected": -2.1347360610961914, "logps/chosen": -217.88844299316406, "logps/ref_chosen": -90.93211364746094, "logps/ref_rejected": -97.44546508789062, "logps/rejected": -295.738525390625, "loss": 0.9109, "margin_dpo/margin_mean": 71.33673095703125, "margin_dpo/margin_std": 82.09566497802734, "step": 539 }, { "epoch": 0.7929515418502202, "fcm_dpo/beta": 0.011464366689324379, "fcm_dpo/delta": -0.09363038092851639, "fcm_dpo/margin": 81.52059936523438, "fcm_dpo/q_t": 0.3140173852443695, "grad_norm": 13.714081764221191, "learning_rate": 6.352838968463919e-08, "logits/chosen": -2.0992918014526367, "logits/rejected": -1.8551020622253418, "logps/chosen": -236.30484008789062, "logps/ref_chosen": -107.86990356445312, "logps/ref_rejected": -121.39350128173828, "logps/rejected": -331.3490295410156, "loss": 0.8469, "margin_dpo/margin_mean": 81.52059936523438, "margin_dpo/margin_std": 84.92539978027344, "step": 540 }, { "epoch": 0.7944199706314243, "fcm_dpo/beta": 0.01155078411102295, "fcm_dpo/delta": 0.18257349729537964, "fcm_dpo/margin": 58.86783981323242, "fcm_dpo/q_t": 0.35946542024612427, "grad_norm": 13.322444915771484, "learning_rate": 6.267605843546767e-08, "logits/chosen": -2.391913414001465, "logits/rejected": -1.9251333475112915, "logps/chosen": -250.4494171142578, "logps/ref_chosen": -106.56748962402344, "logps/ref_rejected": -109.06633758544922, "logps/rejected": -311.81610107421875, "loss": 0.96, "margin_dpo/margin_mean": 58.86784362792969, "margin_dpo/margin_std": 69.12492370605469, "step": 541 }, { "epoch": 0.7958883994126285, "fcm_dpo/beta": 0.011575637385249138, "fcm_dpo/delta": -0.11270305514335632, "fcm_dpo/margin": 82.24880981445312, "fcm_dpo/q_t": 0.30517640709877014, "grad_norm": 11.769819259643555, "learning_rate": 6.182866334636888e-08, "logits/chosen": -2.314541816711426, "logits/rejected": -2.1913418769836426, "logps/chosen": -210.772705078125, "logps/ref_chosen": -87.36929321289062, "logps/ref_rejected": -102.82034301757812, "logps/rejected": -308.47259521484375, "loss": 0.8087, "margin_dpo/margin_mean": 82.24880981445312, "margin_dpo/margin_std": 76.32795715332031, "step": 542 }, { "epoch": 0.7973568281938326, "fcm_dpo/beta": 0.011980516836047173, "fcm_dpo/delta": 0.3275827467441559, "fcm_dpo/margin": 45.187530517578125, "fcm_dpo/q_t": 0.39750808477401733, "grad_norm": 16.94487762451172, "learning_rate": 6.098622674699147e-08, "logits/chosen": -1.8013958930969238, "logits/rejected": -1.8136345148086548, "logps/chosen": -229.92324829101562, "logps/ref_chosen": -87.152587890625, "logps/ref_rejected": -111.53172302246094, "logps/rejected": -299.48992919921875, "loss": 1.15, "margin_dpo/margin_mean": 45.187530517578125, "margin_dpo/margin_std": 86.8045654296875, "step": 543 }, { "epoch": 0.7988252569750367, "fcm_dpo/beta": 0.012570763006806374, "fcm_dpo/delta": 0.166754812002182, "fcm_dpo/margin": 55.278053283691406, "fcm_dpo/q_t": 0.36193686723709106, "grad_norm": 16.597082138061523, "learning_rate": 6.01487708363232e-08, "logits/chosen": -1.80258309841156, "logits/rejected": -1.694300651550293, "logps/chosen": -230.8009033203125, "logps/ref_chosen": -86.14691162109375, "logps/ref_rejected": -106.3939208984375, "logps/rejected": -306.32598876953125, "loss": 0.9627, "margin_dpo/margin_mean": 55.278053283691406, "margin_dpo/margin_std": 71.11997985839844, "step": 544 }, { "epoch": 0.8002936857562408, "fcm_dpo/beta": 0.01272234320640564, "fcm_dpo/delta": -0.014371376484632492, "fcm_dpo/margin": 67.8095474243164, "fcm_dpo/q_t": 0.33010321855545044, "grad_norm": 15.109139442443848, "learning_rate": 5.9316317682106294e-08, "logits/chosen": -1.6739213466644287, "logits/rejected": -1.2589714527130127, "logps/chosen": -229.88116455078125, "logps/ref_chosen": -96.42424011230469, "logps/ref_rejected": -100.73405456542969, "logps/rejected": -302.0005187988281, "loss": 0.9374, "margin_dpo/margin_mean": 67.8095474243164, "margin_dpo/margin_std": 83.960205078125, "step": 545 }, { "epoch": 0.801762114537445, "fcm_dpo/beta": 0.012656424194574356, "fcm_dpo/delta": 0.03976099565625191, "fcm_dpo/margin": 64.24735260009766, "fcm_dpo/q_t": 0.334933340549469, "grad_norm": 14.526140213012695, "learning_rate": 5.848888922025552e-08, "logits/chosen": -1.5082588195800781, "logits/rejected": -1.1441333293914795, "logps/chosen": -222.2100830078125, "logps/ref_chosen": -86.02112579345703, "logps/ref_rejected": -87.37263488769531, "logps/rejected": -287.8089599609375, "loss": 0.9323, "margin_dpo/margin_mean": 64.24735260009766, "margin_dpo/margin_std": 77.9768295288086, "step": 546 }, { "epoch": 0.8032305433186491, "fcm_dpo/beta": 0.012681668624281883, "fcm_dpo/delta": -0.08525756001472473, "fcm_dpo/margin": 73.09080505371094, "fcm_dpo/q_t": 0.3053755462169647, "grad_norm": 13.63310432434082, "learning_rate": 5.7666507254280265e-08, "logits/chosen": -1.8355967998504639, "logits/rejected": -1.389392375946045, "logps/chosen": -232.23291015625, "logps/ref_chosen": -100.6510009765625, "logps/ref_rejected": -97.9117431640625, "logps/rejected": -302.58447265625, "loss": 0.7958, "margin_dpo/margin_mean": 73.09080505371094, "margin_dpo/margin_std": 62.29001235961914, "step": 547 }, { "epoch": 0.8046989720998532, "fcm_dpo/beta": 0.012857835739850998, "fcm_dpo/delta": 0.1577250212430954, "fcm_dpo/margin": 54.670692443847656, "fcm_dpo/q_t": 0.3624279201030731, "grad_norm": 16.492679595947266, "learning_rate": 5.684919345471029e-08, "logits/chosen": -2.2445602416992188, "logits/rejected": -1.89778733253479, "logps/chosen": -231.13241577148438, "logps/ref_chosen": -96.03010559082031, "logps/ref_rejected": -98.41322326660156, "logps/rejected": -288.18621826171875, "loss": 1.0519, "margin_dpo/margin_mean": 54.67068862915039, "margin_dpo/margin_std": 85.29281616210938, "step": 548 }, { "epoch": 0.8061674008810573, "fcm_dpo/beta": 0.012982659041881561, "fcm_dpo/delta": -0.01578594744205475, "fcm_dpo/margin": 66.53627014160156, "fcm_dpo/q_t": 0.33002233505249023, "grad_norm": 15.831024169921875, "learning_rate": 5.603696935852426e-08, "logits/chosen": -1.904597520828247, "logits/rejected": -1.4875590801239014, "logps/chosen": -207.3582305908203, "logps/ref_chosen": -83.45095825195312, "logps/ref_rejected": -81.38023376464844, "logps/rejected": -271.8237609863281, "loss": 0.9136, "margin_dpo/margin_mean": 66.53626251220703, "margin_dpo/margin_std": 78.87059020996094, "step": 549 }, { "epoch": 0.8076358296622613, "fcm_dpo/beta": 0.012999649159610271, "fcm_dpo/delta": 0.010091569274663925, "fcm_dpo/margin": 64.60523986816406, "fcm_dpo/q_t": 0.32790806889533997, "grad_norm": 14.324188232421875, "learning_rate": 5.5229856368582376e-08, "logits/chosen": -1.564119815826416, "logits/rejected": -1.3338861465454102, "logps/chosen": -226.260498046875, "logps/ref_chosen": -89.59434509277344, "logps/ref_rejected": -101.31549835205078, "logps/rejected": -302.5868835449219, "loss": 0.8718, "margin_dpo/margin_mean": 64.60523986816406, "margin_dpo/margin_std": 65.74769592285156, "step": 550 }, { "epoch": 0.8091042584434655, "fcm_dpo/beta": 0.01258097868412733, "fcm_dpo/delta": -0.110738605260849, "fcm_dpo/margin": 75.28948974609375, "fcm_dpo/q_t": 0.31237906217575073, "grad_norm": 16.70378875732422, "learning_rate": 5.4427875753062734e-08, "logits/chosen": -1.785810947418213, "logits/rejected": -1.634450912475586, "logps/chosen": -216.82296752929688, "logps/ref_chosen": -95.08216094970703, "logps/ref_rejected": -117.7192611694336, "logps/rejected": -314.74957275390625, "loss": 0.8109, "margin_dpo/margin_mean": 75.28948974609375, "margin_dpo/margin_std": 73.0545883178711, "step": 551 }, { "epoch": 0.8105726872246696, "fcm_dpo/beta": 0.011867163702845573, "fcm_dpo/delta": -0.32839635014533997, "fcm_dpo/margin": 95.468994140625, "fcm_dpo/q_t": 0.2767210304737091, "grad_norm": 11.077062606811523, "learning_rate": 5.363104864490034e-08, "logits/chosen": -2.2213876247406006, "logits/rejected": -1.8646334409713745, "logps/chosen": -221.26583862304688, "logps/ref_chosen": -101.40918731689453, "logps/ref_rejected": -110.83843994140625, "logps/rejected": -326.1640625, "loss": 0.7468, "margin_dpo/margin_mean": 95.46898651123047, "margin_dpo/margin_std": 84.19966125488281, "step": 552 }, { "epoch": 0.8120411160058737, "fcm_dpo/beta": 0.012067961506545544, "fcm_dpo/delta": 0.11354109644889832, "fcm_dpo/margin": 61.68964385986328, "fcm_dpo/q_t": 0.3463653326034546, "grad_norm": 14.573863983154297, "learning_rate": 5.2839396041230415e-08, "logits/chosen": -1.5574626922607422, "logits/rejected": -1.2425198554992676, "logps/chosen": -228.42669677734375, "logps/ref_chosen": -97.96388244628906, "logps/ref_rejected": -103.57719421386719, "logps/rejected": -295.72967529296875, "loss": 0.9434, "margin_dpo/margin_mean": 61.68965148925781, "margin_dpo/margin_std": 73.47972106933594, "step": 553 }, { "epoch": 0.8135095447870778, "fcm_dpo/beta": 0.012185090221464634, "fcm_dpo/delta": -0.0028081685304641724, "fcm_dpo/margin": 69.8938217163086, "fcm_dpo/q_t": 0.33668839931488037, "grad_norm": 14.112452507019043, "learning_rate": 5.205293880283551e-08, "logits/chosen": -1.8471527099609375, "logits/rejected": -1.0901433229446411, "logps/chosen": -237.7957000732422, "logps/ref_chosen": -103.68680572509766, "logps/ref_rejected": -92.99210357666016, "logps/rejected": -296.99481201171875, "loss": 0.9633, "margin_dpo/margin_mean": 69.8938217163086, "margin_dpo/margin_std": 92.53191375732422, "step": 554 }, { "epoch": 0.8149779735682819, "fcm_dpo/beta": 0.012208528816699982, "fcm_dpo/delta": 0.06300178170204163, "fcm_dpo/margin": 64.8729248046875, "fcm_dpo/q_t": 0.3472994565963745, "grad_norm": 14.670280456542969, "learning_rate": 5.127169765359515e-08, "logits/chosen": -1.7789452075958252, "logits/rejected": -1.679337978363037, "logps/chosen": -227.99090576171875, "logps/ref_chosen": -86.20362854003906, "logps/ref_rejected": -114.54719543457031, "logps/rejected": -321.2073974609375, "loss": 0.962, "margin_dpo/margin_mean": 64.8729248046875, "margin_dpo/margin_std": 85.02949523925781, "step": 555 }, { "epoch": 0.8164464023494861, "fcm_dpo/beta": 0.012215223163366318, "fcm_dpo/delta": -0.005190940573811531, "fcm_dpo/margin": 69.96516418457031, "fcm_dpo/q_t": 0.326467365026474, "grad_norm": 13.305697441101074, "learning_rate": 5.049569317994012e-08, "logits/chosen": -1.7430949211120605, "logits/rejected": -1.4270676374435425, "logps/chosen": -215.5836639404297, "logps/ref_chosen": -88.40348815917969, "logps/ref_rejected": -106.63911437988281, "logps/rejected": -303.7844543457031, "loss": 0.8898, "margin_dpo/margin_mean": 69.96516418457031, "margin_dpo/margin_std": 77.99856567382812, "step": 556 }, { "epoch": 0.8179148311306902, "fcm_dpo/beta": 0.011995375156402588, "fcm_dpo/delta": -0.07914155721664429, "fcm_dpo/margin": 76.68682861328125, "fcm_dpo/q_t": 0.3210620880126953, "grad_norm": 11.786112785339355, "learning_rate": 4.9724945830310144e-08, "logits/chosen": -2.0711090564727783, "logits/rejected": -1.9035547971725464, "logps/chosen": -234.16250610351562, "logps/ref_chosen": -98.83523559570312, "logps/ref_rejected": -115.64888763427734, "logps/rejected": -327.6629943847656, "loss": 0.8671, "margin_dpo/margin_mean": 76.68682861328125, "margin_dpo/margin_std": 83.49786376953125, "step": 557 }, { "epoch": 0.8193832599118943, "fcm_dpo/beta": 0.011821841821074486, "fcm_dpo/delta": -0.09789042174816132, "fcm_dpo/margin": 79.33963012695312, "fcm_dpo/q_t": 0.3074336647987366, "grad_norm": 12.818613052368164, "learning_rate": 4.8959475914614554e-08, "logits/chosen": -2.476902961730957, "logits/rejected": -1.8757622241973877, "logps/chosen": -243.42117309570312, "logps/ref_chosen": -106.01183319091797, "logps/ref_rejected": -108.39131927490234, "logps/rejected": -325.1402893066406, "loss": 0.8074, "margin_dpo/margin_mean": 79.33963012695312, "margin_dpo/margin_std": 73.38945770263672, "step": 558 }, { "epoch": 0.8208516886930984, "fcm_dpo/beta": 0.01184871792793274, "fcm_dpo/delta": 0.024944249540567398, "fcm_dpo/margin": 69.80717468261719, "fcm_dpo/q_t": 0.32996243238449097, "grad_norm": 14.529841423034668, "learning_rate": 4.8199303603697614e-08, "logits/chosen": -2.562124252319336, "logits/rejected": -2.099799871444702, "logps/chosen": -242.94049072265625, "logps/ref_chosen": -100.42474365234375, "logps/ref_rejected": -100.41270446777344, "logps/rejected": -312.7356262207031, "loss": 0.9057, "margin_dpo/margin_mean": 69.80717468261719, "margin_dpo/margin_std": 79.37966918945312, "step": 559 }, { "epoch": 0.8223201174743024, "fcm_dpo/beta": 0.011990891769528389, "fcm_dpo/delta": 0.07405167818069458, "fcm_dpo/margin": 65.19851684570312, "fcm_dpo/q_t": 0.34072330594062805, "grad_norm": 13.63769245147705, "learning_rate": 4.7444448928806615e-08, "logits/chosen": -1.3843114376068115, "logits/rejected": -0.9339506030082703, "logps/chosen": -229.32911682128906, "logps/ref_chosen": -87.63551330566406, "logps/ref_rejected": -87.37985229492188, "logps/rejected": -294.27197265625, "loss": 0.9346, "margin_dpo/margin_mean": 65.1985092163086, "margin_dpo/margin_std": 78.38851928710938, "step": 560 }, { "epoch": 0.8237885462555066, "fcm_dpo/beta": 0.012294452637434006, "fcm_dpo/delta": 0.0976395532488823, "fcm_dpo/margin": 61.706260681152344, "fcm_dpo/q_t": 0.34566307067871094, "grad_norm": 13.019865989685059, "learning_rate": 4.669493178106432e-08, "logits/chosen": -1.9657562971115112, "logits/rejected": -1.9275360107421875, "logps/chosen": -221.61062622070312, "logps/ref_chosen": -74.88140106201172, "logps/ref_rejected": -106.04796600341797, "logps/rejected": -314.48345947265625, "loss": 0.9111, "margin_dpo/margin_mean": 61.706260681152344, "margin_dpo/margin_std": 68.51325225830078, "step": 561 }, { "epoch": 0.8252569750367107, "fcm_dpo/beta": 0.012495906092226505, "fcm_dpo/delta": 0.11548486351966858, "fcm_dpo/margin": 59.469627380371094, "fcm_dpo/q_t": 0.35613298416137695, "grad_norm": 16.456663131713867, "learning_rate": 4.5950771910944596e-08, "logits/chosen": -2.2097997665405273, "logits/rejected": -1.7643963098526, "logps/chosen": -233.54342651367188, "logps/ref_chosen": -100.66322326660156, "logps/ref_rejected": -102.24087524414062, "logps/rejected": -294.5906982421875, "loss": 1.0184, "margin_dpo/margin_mean": 59.469627380371094, "margin_dpo/margin_std": 88.21414184570312, "step": 562 }, { "epoch": 0.8267254038179148, "fcm_dpo/beta": 0.012950560078024864, "fcm_dpo/delta": 0.14667396247386932, "fcm_dpo/margin": 54.94389343261719, "fcm_dpo/q_t": 0.35650813579559326, "grad_norm": 15.449488639831543, "learning_rate": 4.521198892775202e-08, "logits/chosen": -1.6731395721435547, "logits/rejected": -1.3347598314285278, "logps/chosen": -239.73776245117188, "logps/ref_chosen": -96.25921630859375, "logps/ref_rejected": -101.20503997802734, "logps/rejected": -299.6274719238281, "loss": 0.9925, "margin_dpo/margin_mean": 54.94389343261719, "margin_dpo/margin_std": 74.90940856933594, "step": 563 }, { "epoch": 0.8281938325991189, "fcm_dpo/beta": 0.01319155003875494, "fcm_dpo/delta": 0.06683439761400223, "fcm_dpo/margin": 59.61730194091797, "fcm_dpo/q_t": 0.3362036943435669, "grad_norm": 13.23291015625, "learning_rate": 4.447860229910544e-08, "logits/chosen": -2.3670005798339844, "logits/rejected": -1.8159103393554688, "logps/chosen": -251.44268798828125, "logps/ref_chosen": -112.96040344238281, "logps/ref_rejected": -99.52803802490234, "logps/rejected": -297.6276550292969, "loss": 0.8982, "margin_dpo/margin_mean": 59.6172981262207, "margin_dpo/margin_std": 62.563438415527344, "step": 564 }, { "epoch": 0.8296622613803231, "fcm_dpo/beta": 0.012753048911690712, "fcm_dpo/delta": -0.17041520774364471, "fcm_dpo/margin": 78.61331176757812, "fcm_dpo/q_t": 0.30064839124679565, "grad_norm": 12.86750316619873, "learning_rate": 4.375063135042445e-08, "logits/chosen": -1.7283689975738525, "logits/rejected": -1.225877285003662, "logps/chosen": -224.79379272460938, "logps/ref_chosen": -95.36558532714844, "logps/ref_rejected": -92.21737670898438, "logps/rejected": -300.2589111328125, "loss": 0.803, "margin_dpo/margin_mean": 78.61331176757812, "margin_dpo/margin_std": 75.22626495361328, "step": 565 }, { "epoch": 0.8311306901615272, "fcm_dpo/beta": 0.012551356106996536, "fcm_dpo/delta": -0.08185572922229767, "fcm_dpo/margin": 73.64617919921875, "fcm_dpo/q_t": 0.31478655338287354, "grad_norm": 12.691439628601074, "learning_rate": 4.3028095264420525e-08, "logits/chosen": -1.8287672996520996, "logits/rejected": -1.3916640281677246, "logps/chosen": -235.4650421142578, "logps/ref_chosen": -98.97320556640625, "logps/ref_rejected": -110.82090759277344, "logps/rejected": -320.95892333984375, "loss": 0.8519, "margin_dpo/margin_mean": 73.64617156982422, "margin_dpo/margin_std": 77.48297119140625, "step": 566 }, { "epoch": 0.8325991189427313, "fcm_dpo/beta": 0.012515963986515999, "fcm_dpo/delta": 0.07364558428525925, "fcm_dpo/margin": 62.43628692626953, "fcm_dpo/q_t": 0.3408878445625305, "grad_norm": 15.552478790283203, "learning_rate": 4.231101308059165e-08, "logits/chosen": -2.209388256072998, "logits/rejected": -1.5480117797851562, "logps/chosen": -234.72930908203125, "logps/ref_chosen": -92.7035903930664, "logps/ref_rejected": -91.22431945800781, "logps/rejected": -295.6863098144531, "loss": 0.9282, "margin_dpo/margin_mean": 62.43628692626953, "margin_dpo/margin_std": 72.3102035522461, "step": 567 }, { "epoch": 0.8340675477239354, "fcm_dpo/beta": 0.012427978217601776, "fcm_dpo/delta": -0.09094738960266113, "fcm_dpo/margin": 74.95256805419922, "fcm_dpo/q_t": 0.3132409453392029, "grad_norm": 12.722140312194824, "learning_rate": 4.1599403694720145e-08, "logits/chosen": -1.535170316696167, "logits/rejected": -1.3898873329162598, "logps/chosen": -206.69818115234375, "logps/ref_chosen": -73.13226318359375, "logps/ref_rejected": -94.95265197753906, "logps/rejected": -303.47113037109375, "loss": 0.8189, "margin_dpo/margin_mean": 74.95256805419922, "margin_dpo/margin_std": 72.48225402832031, "step": 568 }, { "epoch": 0.8355359765051396, "fcm_dpo/beta": 0.012630300596356392, "fcm_dpo/delta": 0.0689060166478157, "fcm_dpo/margin": 62.19904327392578, "fcm_dpo/q_t": 0.3414260149002075, "grad_norm": 16.49064064025879, "learning_rate": 4.089328585837512e-08, "logits/chosen": -2.0152597427368164, "logits/rejected": -1.65643310546875, "logps/chosen": -234.0463409423828, "logps/ref_chosen": -89.21708679199219, "logps/ref_rejected": -86.41105651855469, "logps/rejected": -293.4393615722656, "loss": 0.9867, "margin_dpo/margin_mean": 62.19904327392578, "margin_dpo/margin_std": 83.06907653808594, "step": 569 }, { "epoch": 0.8370044052863436, "fcm_dpo/beta": 0.012341935187578201, "fcm_dpo/delta": -0.1500765085220337, "fcm_dpo/margin": 79.84074401855469, "fcm_dpo/q_t": 0.29707303643226624, "grad_norm": 12.651300430297852, "learning_rate": 4.019267817841834e-08, "logits/chosen": -2.3013365268707275, "logits/rejected": -1.6241872310638428, "logps/chosen": -225.6119384765625, "logps/ref_chosen": -94.76277160644531, "logps/ref_rejected": -88.846923828125, "logps/rejected": -299.5368347167969, "loss": 0.7799, "margin_dpo/margin_mean": 79.84074401855469, "margin_dpo/margin_std": 70.12899780273438, "step": 570 }, { "epoch": 0.8384728340675477, "fcm_dpo/beta": 0.012180352583527565, "fcm_dpo/delta": -0.05313785746693611, "fcm_dpo/margin": 73.75041961669922, "fcm_dpo/q_t": 0.32369863986968994, "grad_norm": 15.93890380859375, "learning_rate": 3.9497599116513705e-08, "logits/chosen": -1.891650676727295, "logits/rejected": -1.7370983362197876, "logps/chosen": -211.00099182128906, "logps/ref_chosen": -80.8212890625, "logps/ref_rejected": -97.01528930664062, "logps/rejected": -300.9454345703125, "loss": 0.8998, "margin_dpo/margin_mean": 73.75041198730469, "margin_dpo/margin_std": 85.5877914428711, "step": 571 }, { "epoch": 0.8399412628487518, "fcm_dpo/beta": 0.011995727196335793, "fcm_dpo/delta": -0.03884110599756241, "fcm_dpo/margin": 73.78263854980469, "fcm_dpo/q_t": 0.328910231590271, "grad_norm": 12.733126640319824, "learning_rate": 3.880806698864086e-08, "logits/chosen": -1.7663264274597168, "logits/rejected": -1.6016298532485962, "logps/chosen": -211.55044555664062, "logps/ref_chosen": -77.37992858886719, "logps/ref_rejected": -90.56597900390625, "logps/rejected": -298.5191650390625, "loss": 0.8762, "margin_dpo/margin_mean": 73.78263854980469, "margin_dpo/margin_std": 83.24708557128906, "step": 572 }, { "epoch": 0.8414096916299559, "fcm_dpo/beta": 0.01217666082084179, "fcm_dpo/delta": -0.011429935693740845, "fcm_dpo/margin": 70.3113021850586, "fcm_dpo/q_t": 0.32651984691619873, "grad_norm": 13.776387214660645, "learning_rate": 3.812409996461275e-08, "logits/chosen": -2.3168587684631348, "logits/rejected": -2.1058921813964844, "logps/chosen": -215.7606201171875, "logps/ref_chosen": -83.7060775756836, "logps/ref_rejected": -90.69746398925781, "logps/rejected": -293.06329345703125, "loss": 0.9059, "margin_dpo/margin_mean": 70.3113021850586, "margin_dpo/margin_std": 77.29948425292969, "step": 573 }, { "epoch": 0.8428781204111601, "fcm_dpo/beta": 0.012046435847878456, "fcm_dpo/delta": 0.0020178817212581635, "fcm_dpo/margin": 70.36782836914062, "fcm_dpo/q_t": 0.33117565512657166, "grad_norm": 14.58945369720459, "learning_rate": 3.74457160675965e-08, "logits/chosen": -2.1412198543548584, "logits/rejected": -1.8174076080322266, "logps/chosen": -223.34002685546875, "logps/ref_chosen": -85.12948608398438, "logps/ref_rejected": -99.05403137207031, "logps/rejected": -307.63238525390625, "loss": 0.9014, "margin_dpo/margin_mean": 70.36782836914062, "margin_dpo/margin_std": 80.83290100097656, "step": 574 }, { "epoch": 0.8443465491923642, "fcm_dpo/beta": 0.011899670585989952, "fcm_dpo/delta": -0.02533550187945366, "fcm_dpo/margin": 73.34465789794922, "fcm_dpo/q_t": 0.32771629095077515, "grad_norm": 17.919416427612305, "learning_rate": 3.677293317363864e-08, "logits/chosen": -1.301839828491211, "logits/rejected": -0.845508873462677, "logps/chosen": -246.7624969482422, "logps/ref_chosen": -105.77084350585938, "logps/ref_rejected": -100.94944763183594, "logps/rejected": -315.2857666015625, "loss": 0.9172, "margin_dpo/margin_mean": 73.34465789794922, "margin_dpo/margin_std": 88.70787048339844, "step": 575 }, { "epoch": 0.8458149779735683, "fcm_dpo/beta": 0.012019166722893715, "fcm_dpo/delta": 0.01799788698554039, "fcm_dpo/margin": 69.30175018310547, "fcm_dpo/q_t": 0.33523285388946533, "grad_norm": 13.958468437194824, "learning_rate": 3.6105769011194224e-08, "logits/chosen": -2.088580846786499, "logits/rejected": -2.000969886779785, "logps/chosen": -223.76405334472656, "logps/ref_chosen": -81.20452880859375, "logps/ref_rejected": -105.40940856933594, "logps/rejected": -317.2706604003906, "loss": 0.9374, "margin_dpo/margin_mean": 69.30174255371094, "margin_dpo/margin_std": 85.55188751220703, "step": 576 }, { "epoch": 0.8472834067547724, "fcm_dpo/beta": 0.011663603596389294, "fcm_dpo/delta": -0.13514071702957153, "fcm_dpo/margin": 83.21434783935547, "fcm_dpo/q_t": 0.30806174874305725, "grad_norm": 11.793753623962402, "learning_rate": 3.5444241160659304e-08, "logits/chosen": -2.0075855255126953, "logits/rejected": -1.3254092931747437, "logps/chosen": -228.4785919189453, "logps/ref_chosen": -101.46419525146484, "logps/ref_rejected": -89.32637023925781, "logps/rejected": -299.55511474609375, "loss": 0.8493, "margin_dpo/margin_mean": 83.21435546875, "margin_dpo/margin_std": 87.91677856445312, "step": 577 }, { "epoch": 0.8487518355359766, "fcm_dpo/beta": 0.011880055069923401, "fcm_dpo/delta": 0.11570513248443604, "fcm_dpo/margin": 62.5034294128418, "fcm_dpo/q_t": 0.34822767972946167, "grad_norm": 15.396873474121094, "learning_rate": 3.478836705390808e-08, "logits/chosen": -1.3165785074234009, "logits/rejected": -1.0804065465927124, "logps/chosen": -223.2236328125, "logps/ref_chosen": -77.29241943359375, "logps/ref_rejected": -88.24701690673828, "logps/rejected": -296.681640625, "loss": 0.9618, "margin_dpo/margin_mean": 62.50342559814453, "margin_dpo/margin_std": 78.88618469238281, "step": 578 }, { "epoch": 0.8502202643171806, "fcm_dpo/beta": 0.012005902826786041, "fcm_dpo/delta": 0.08513668179512024, "fcm_dpo/margin": 64.23765563964844, "fcm_dpo/q_t": 0.34554487466812134, "grad_norm": 13.368642807006836, "learning_rate": 3.41381639738331e-08, "logits/chosen": -1.716012954711914, "logits/rejected": -1.379181146621704, "logps/chosen": -222.61764526367188, "logps/ref_chosen": -87.34305572509766, "logps/ref_rejected": -102.12100982666016, "logps/rejected": -301.6332702636719, "loss": 0.936, "margin_dpo/margin_mean": 64.23765563964844, "margin_dpo/margin_std": 75.44595336914062, "step": 579 }, { "epoch": 0.8516886930983847, "fcm_dpo/beta": 0.011975444853305817, "fcm_dpo/delta": -0.028606295585632324, "fcm_dpo/margin": 73.02272033691406, "fcm_dpo/q_t": 0.32629266381263733, "grad_norm": 13.48125171661377, "learning_rate": 3.349364905389032e-08, "logits/chosen": -1.2470667362213135, "logits/rejected": -0.930492639541626, "logps/chosen": -220.02879333496094, "logps/ref_chosen": -88.92879486083984, "logps/ref_rejected": -88.69661712646484, "logps/rejected": -292.8193359375, "loss": 0.8775, "margin_dpo/margin_mean": 73.02272033691406, "margin_dpo/margin_std": 78.72087860107422, "step": 580 }, { "epoch": 0.8531571218795888, "fcm_dpo/beta": 0.012202093377709389, "fcm_dpo/delta": 0.066944919526577, "fcm_dpo/margin": 64.5920181274414, "fcm_dpo/q_t": 0.34423375129699707, "grad_norm": 15.27383041381836, "learning_rate": 3.285483927764726e-08, "logits/chosen": -2.02988338470459, "logits/rejected": -1.6924233436584473, "logps/chosen": -231.8140869140625, "logps/ref_chosen": -92.22323608398438, "logps/ref_rejected": -99.38943481445312, "logps/rejected": -303.57232666015625, "loss": 0.9915, "margin_dpo/margin_mean": 64.5920181274414, "margin_dpo/margin_std": 88.16921997070312, "step": 581 }, { "epoch": 0.8546255506607929, "fcm_dpo/beta": 0.012522203847765923, "fcm_dpo/delta": 0.06351852416992188, "fcm_dpo/margin": 62.802978515625, "fcm_dpo/q_t": 0.3435629606246948, "grad_norm": 13.179994583129883, "learning_rate": 3.222175147833556e-08, "logits/chosen": -2.2159390449523926, "logits/rejected": -2.1951913833618164, "logps/chosen": -209.8260040283203, "logps/ref_chosen": -87.33561706542969, "logps/ref_rejected": -115.1544189453125, "logps/rejected": -300.4477844238281, "loss": 0.9489, "margin_dpo/margin_mean": 62.802974700927734, "margin_dpo/margin_std": 77.04085540771484, "step": 582 }, { "epoch": 0.856093979441997, "fcm_dpo/beta": 0.01241613645106554, "fcm_dpo/delta": 0.048651568591594696, "fcm_dpo/margin": 64.84408569335938, "fcm_dpo/q_t": 0.34346115589141846, "grad_norm": 14.555140495300293, "learning_rate": 3.159440233840763e-08, "logits/chosen": -1.4759223461151123, "logits/rejected": -1.245713233947754, "logps/chosen": -229.76559448242188, "logps/ref_chosen": -83.71707153320312, "logps/ref_rejected": -94.58465576171875, "logps/rejected": -305.477294921875, "loss": 0.9223, "margin_dpo/margin_mean": 64.84408569335938, "margin_dpo/margin_std": 78.4787368774414, "step": 583 }, { "epoch": 0.8575624082232012, "fcm_dpo/beta": 0.012546549551188946, "fcm_dpo/delta": -0.057240843772888184, "fcm_dpo/margin": 71.72808837890625, "fcm_dpo/q_t": 0.3223274052143097, "grad_norm": 14.367459297180176, "learning_rate": 3.0972808389096635e-08, "logits/chosen": -1.9238438606262207, "logits/rejected": -1.4491448402404785, "logps/chosen": -223.58079528808594, "logps/ref_chosen": -94.25955963134766, "logps/ref_rejected": -104.48269653320312, "logps/rejected": -305.5320129394531, "loss": 0.8921, "margin_dpo/margin_mean": 71.72808837890625, "margin_dpo/margin_std": 80.05842590332031, "step": 584 }, { "epoch": 0.8590308370044053, "fcm_dpo/beta": 0.012124446220695972, "fcm_dpo/delta": -0.1827441304922104, "fcm_dpo/margin": 83.6611557006836, "fcm_dpo/q_t": 0.30458420515060425, "grad_norm": 12.921794891357422, "learning_rate": 3.035698600998121e-08, "logits/chosen": -1.9596983194351196, "logits/rejected": -1.6787914037704468, "logps/chosen": -222.28982543945312, "logps/ref_chosen": -85.63676452636719, "logps/ref_rejected": -90.28547668457031, "logps/rejected": -310.5997314453125, "loss": 0.8384, "margin_dpo/margin_mean": 83.6611557006836, "margin_dpo/margin_std": 88.93515014648438, "step": 585 }, { "epoch": 0.8604992657856094, "fcm_dpo/beta": 0.012192027643322945, "fcm_dpo/delta": 0.14729395508766174, "fcm_dpo/margin": 58.48194885253906, "fcm_dpo/q_t": 0.3539287745952606, "grad_norm": 13.725818634033203, "learning_rate": 2.974695142855388e-08, "logits/chosen": -1.8884037733078003, "logits/rejected": -1.6240513324737549, "logps/chosen": -231.93746948242188, "logps/ref_chosen": -83.68114471435547, "logps/ref_rejected": -98.35916137695312, "logps/rejected": -305.097412109375, "loss": 0.9459, "margin_dpo/margin_mean": 58.48194885253906, "margin_dpo/margin_std": 69.6341552734375, "step": 586 }, { "epoch": 0.8619676945668135, "fcm_dpo/beta": 0.012489515356719494, "fcm_dpo/delta": 0.07789243757724762, "fcm_dpo/margin": 62.21076583862305, "fcm_dpo/q_t": 0.3466240167617798, "grad_norm": 12.888617515563965, "learning_rate": 2.9142720719793122e-08, "logits/chosen": -2.3818912506103516, "logits/rejected": -2.0904064178466797, "logps/chosen": -201.2530975341797, "logps/ref_chosen": -77.57750701904297, "logps/ref_rejected": -89.32032775878906, "logps/rejected": -275.2066650390625, "loss": 0.9453, "margin_dpo/margin_mean": 62.21076202392578, "margin_dpo/margin_std": 77.28038024902344, "step": 587 }, { "epoch": 0.8634361233480177, "fcm_dpo/beta": 0.012591280043125153, "fcm_dpo/delta": 0.11278443038463593, "fcm_dpo/margin": 59.20628356933594, "fcm_dpo/q_t": 0.35483455657958984, "grad_norm": 14.562524795532227, "learning_rate": 2.8544309805740018e-08, "logits/chosen": -1.7069602012634277, "logits/rejected": -1.6397366523742676, "logps/chosen": -215.88272094726562, "logps/ref_chosen": -83.80426025390625, "logps/ref_rejected": -113.11943054199219, "logps/rejected": -304.4042053222656, "loss": 1.0127, "margin_dpo/margin_mean": 59.20628356933594, "margin_dpo/margin_std": 84.50772857666016, "step": 588 }, { "epoch": 0.8649045521292217, "fcm_dpo/beta": 0.01244035642594099, "fcm_dpo/delta": -0.14263486862182617, "fcm_dpo/margin": 78.6182861328125, "fcm_dpo/q_t": 0.2986539602279663, "grad_norm": 15.344147682189941, "learning_rate": 2.7951734455078786e-08, "logits/chosen": -1.6551861763000488, "logits/rejected": -1.247209072113037, "logps/chosen": -235.73348999023438, "logps/ref_chosen": -96.46501922607422, "logps/ref_rejected": -116.03719329833984, "logps/rejected": -333.9239501953125, "loss": 0.7867, "margin_dpo/margin_mean": 78.6182861328125, "margin_dpo/margin_std": 69.69014739990234, "step": 589 }, { "epoch": 0.8663729809104258, "fcm_dpo/beta": 0.012330231256783009, "fcm_dpo/delta": -0.08336605131626129, "fcm_dpo/margin": 75.031005859375, "fcm_dpo/q_t": 0.3123021721839905, "grad_norm": 12.972137451171875, "learning_rate": 2.736501028272095e-08, "logits/chosen": -1.9837596416473389, "logits/rejected": -1.8040398359298706, "logps/chosen": -217.9958038330078, "logps/ref_chosen": -88.4278564453125, "logps/ref_rejected": -110.19876861572266, "logps/rejected": -314.7977294921875, "loss": 0.8127, "margin_dpo/margin_mean": 75.031005859375, "margin_dpo/margin_std": 70.29672241210938, "step": 590 }, { "epoch": 0.8678414096916299, "fcm_dpo/beta": 0.011928400956094265, "fcm_dpo/delta": -0.12863200902938843, "fcm_dpo/margin": 80.97084045410156, "fcm_dpo/q_t": 0.30411410331726074, "grad_norm": 11.441441535949707, "learning_rate": 2.678415274939408e-08, "logits/chosen": -2.0621907711029053, "logits/rejected": -1.4476604461669922, "logps/chosen": -214.27345275878906, "logps/ref_chosen": -87.84968566894531, "logps/ref_rejected": -89.57919311523438, "logps/rejected": -296.97381591796875, "loss": 0.7947, "margin_dpo/margin_mean": 80.97084045410156, "margin_dpo/margin_std": 74.34172058105469, "step": 591 }, { "epoch": 0.869309838472834, "fcm_dpo/beta": 0.012076403945684433, "fcm_dpo/delta": 0.10909491032361984, "fcm_dpo/margin": 62.0214958190918, "fcm_dpo/q_t": 0.34997737407684326, "grad_norm": 14.482314109802246, "learning_rate": 2.6209177161234442e-08, "logits/chosen": -1.8271369934082031, "logits/rejected": -1.5044286251068115, "logps/chosen": -223.06594848632812, "logps/ref_chosen": -74.97573852539062, "logps/ref_rejected": -81.55262756347656, "logps/rejected": -291.66436767578125, "loss": 0.9767, "margin_dpo/margin_mean": 62.02149963378906, "margin_dpo/margin_std": 82.1724853515625, "step": 592 }, { "epoch": 0.8707782672540382, "fcm_dpo/beta": 0.012362951412796974, "fcm_dpo/delta": 0.17887909710407257, "fcm_dpo/margin": 55.27935791015625, "fcm_dpo/q_t": 0.36694955825805664, "grad_norm": 19.68903160095215, "learning_rate": 2.564009866938349e-08, "logits/chosen": -0.6591614484786987, "logits/rejected": -0.34581655263900757, "logps/chosen": -211.53347778320312, "logps/ref_chosen": -72.23902893066406, "logps/ref_rejected": -70.09652709960938, "logps/rejected": -264.67034912109375, "loss": 1.0278, "margin_dpo/margin_mean": 55.27935791015625, "margin_dpo/margin_std": 81.17341613769531, "step": 593 }, { "epoch": 0.8722466960352423, "fcm_dpo/beta": 0.0123524721711874, "fcm_dpo/delta": -0.1632692813873291, "fcm_dpo/margin": 80.72880554199219, "fcm_dpo/q_t": 0.2972342073917389, "grad_norm": 11.561514854431152, "learning_rate": 2.5076932269588708e-08, "logits/chosen": -1.9211974143981934, "logits/rejected": -1.4067270755767822, "logps/chosen": -216.4187774658203, "logps/ref_chosen": -90.43771362304688, "logps/ref_rejected": -92.267578125, "logps/rejected": -298.9774475097656, "loss": 0.7671, "margin_dpo/margin_mean": 80.72880554199219, "margin_dpo/margin_std": 69.18890380859375, "step": 594 }, { "epoch": 0.8737151248164464, "fcm_dpo/beta": 0.012263715267181396, "fcm_dpo/delta": 0.021671980619430542, "fcm_dpo/margin": 67.65665435791016, "fcm_dpo/q_t": 0.3382018506526947, "grad_norm": 14.33916187286377, "learning_rate": 2.451969280180849e-08, "logits/chosen": -1.8135511875152588, "logits/rejected": -1.3942694664001465, "logps/chosen": -212.7932891845703, "logps/ref_chosen": -81.51480102539062, "logps/ref_rejected": -86.52047729492188, "logps/rejected": -285.45562744140625, "loss": 0.9179, "margin_dpo/margin_mean": 67.65666198730469, "margin_dpo/margin_std": 80.72854614257812, "step": 595 }, { "epoch": 0.8751835535976505, "fcm_dpo/beta": 0.012409276328980923, "fcm_dpo/delta": 0.08794566988945007, "fcm_dpo/margin": 61.938507080078125, "fcm_dpo/q_t": 0.3473287522792816, "grad_norm": 16.3719425201416, "learning_rate": 2.396839494982103e-08, "logits/chosen": -1.658387541770935, "logits/rejected": -1.0812584161758423, "logps/chosen": -223.06051635742188, "logps/ref_chosen": -88.20805358886719, "logps/ref_rejected": -86.09375, "logps/rejected": -282.88470458984375, "loss": 0.9722, "margin_dpo/margin_mean": 61.938507080078125, "margin_dpo/margin_std": 80.60945129394531, "step": 596 }, { "epoch": 0.8766519823788547, "fcm_dpo/beta": 0.012180918827652931, "fcm_dpo/delta": -0.11607787013053894, "fcm_dpo/margin": 78.33621215820312, "fcm_dpo/q_t": 0.31094270944595337, "grad_norm": 11.810127258300781, "learning_rate": 2.3423053240837514e-08, "logits/chosen": -1.5121232271194458, "logits/rejected": -1.3904181718826294, "logps/chosen": -216.66140747070312, "logps/ref_chosen": -83.58998107910156, "logps/ref_rejected": -106.94439697265625, "logps/rejected": -318.35205078125, "loss": 0.8322, "margin_dpo/margin_mean": 78.33621215820312, "margin_dpo/margin_std": 80.11341857910156, "step": 597 }, { "epoch": 0.8781204111600588, "fcm_dpo/beta": 0.012269245460629463, "fcm_dpo/delta": 0.11151318997144699, "fcm_dpo/margin": 60.842613220214844, "fcm_dpo/q_t": 0.34948280453681946, "grad_norm": 13.330480575561523, "learning_rate": 2.2883682045119062e-08, "logits/chosen": -2.012774705886841, "logits/rejected": -1.704032301902771, "logps/chosen": -219.08596801757812, "logps/ref_chosen": -90.18262481689453, "logps/ref_rejected": -94.79856872558594, "logps/rejected": -284.5445251464844, "loss": 0.9748, "margin_dpo/margin_mean": 60.842613220214844, "margin_dpo/margin_std": 78.31558227539062, "step": 598 }, { "epoch": 0.8795888399412628, "fcm_dpo/beta": 0.01254078559577465, "fcm_dpo/delta": 0.1279219090938568, "fcm_dpo/margin": 58.25449752807617, "fcm_dpo/q_t": 0.35618844628334045, "grad_norm": 16.062721252441406, "learning_rate": 2.2350295575598367e-08, "logits/chosen": -1.863084316253662, "logits/rejected": -1.7345049381256104, "logps/chosen": -209.37547302246094, "logps/ref_chosen": -77.84226989746094, "logps/ref_rejected": -89.1976318359375, "logps/rejected": -278.9853210449219, "loss": 0.9903, "margin_dpo/margin_mean": 58.25449752807617, "margin_dpo/margin_std": 76.904541015625, "step": 599 }, { "epoch": 0.8810572687224669, "fcm_dpo/beta": 0.012659793719649315, "fcm_dpo/delta": -0.05752100050449371, "fcm_dpo/margin": 71.27790832519531, "fcm_dpo/q_t": 0.3214208483695984, "grad_norm": 12.589545249938965, "learning_rate": 2.1822907887504932e-08, "logits/chosen": -2.3141098022460938, "logits/rejected": -1.9681317806243896, "logps/chosen": -229.8334197998047, "logps/ref_chosen": -92.18781280517578, "logps/ref_rejected": -92.95429992675781, "logps/rejected": -301.8778076171875, "loss": 0.8671, "margin_dpo/margin_mean": 71.27790832519531, "margin_dpo/margin_std": 76.77690887451172, "step": 600 }, { "epoch": 0.8810572687224669, "eval_fcm_dpo/beta": 0.012609176337718964, "eval_logits/chosen": -1.6990704536437988, "eval_logits/rejected": -1.3195072412490845, "eval_logps/chosen": -239.63792419433594, "eval_logps/ref_chosen": -100.49356842041016, "eval_logps/ref_rejected": -94.06775665283203, "eval_logps/rejected": -292.5469055175781, "eval_loss": 0.48429349064826965, "eval_margin_dpo/margin_mean": 59.33479309082031, "eval_margin_dpo/margin_std": 75.8063735961914, "eval_runtime": 43.7607, "eval_samples_per_second": 53.45, "eval_steps_per_second": 1.691, "step": 600 }, { "epoch": 0.882525697503671, "fcm_dpo/beta": 0.012439209967851639, "fcm_dpo/delta": -0.04578261449933052, "fcm_dpo/margin": 71.59075927734375, "fcm_dpo/q_t": 0.3206157386302948, "grad_norm": 13.716604232788086, "learning_rate": 2.1301532877994742e-08, "logits/chosen": -1.623553991317749, "logits/rejected": -1.2954704761505127, "logps/chosen": -231.75643920898438, "logps/ref_chosen": -89.21614074707031, "logps/ref_rejected": -100.17054748535156, "logps/rejected": -314.3016052246094, "loss": 0.8497, "margin_dpo/margin_mean": 71.59075927734375, "margin_dpo/margin_std": 73.1295394897461, "step": 601 }, { "epoch": 0.8839941262848752, "fcm_dpo/beta": 0.012094026431441307, "fcm_dpo/delta": -0.21615482866764069, "fcm_dpo/margin": 86.27711486816406, "fcm_dpo/q_t": 0.29506054520606995, "grad_norm": 12.191177368164062, "learning_rate": 2.0786184285784298e-08, "logits/chosen": -2.171415090560913, "logits/rejected": -1.9284133911132812, "logps/chosen": -195.90830993652344, "logps/ref_chosen": -80.05760192871094, "logps/ref_rejected": -93.197509765625, "logps/rejected": -295.3253173828125, "loss": 0.793, "margin_dpo/margin_mean": 86.27711486816406, "margin_dpo/margin_std": 84.91157531738281, "step": 602 }, { "epoch": 0.8854625550660793, "fcm_dpo/beta": 0.011891430243849754, "fcm_dpo/delta": -0.06066790223121643, "fcm_dpo/margin": 76.12080383300781, "fcm_dpo/q_t": 0.31821417808532715, "grad_norm": 13.448278427124023, "learning_rate": 2.0276875690788204e-08, "logits/chosen": -1.967059850692749, "logits/rejected": -1.3754029273986816, "logps/chosen": -224.9002227783203, "logps/ref_chosen": -102.30958557128906, "logps/ref_rejected": -108.06884765625, "logps/rejected": -306.7803039550781, "loss": 0.847, "margin_dpo/margin_mean": 76.12080383300781, "margin_dpo/margin_std": 80.23945617675781, "step": 603 }, { "epoch": 0.8869309838472834, "fcm_dpo/beta": 0.011719077825546265, "fcm_dpo/delta": -0.028855059295892715, "fcm_dpo/margin": 74.73594665527344, "fcm_dpo/q_t": 0.31438958644866943, "grad_norm": 11.16280460357666, "learning_rate": 1.977362051376158e-08, "logits/chosen": -1.7443255186080933, "logits/rejected": -1.5849244594573975, "logps/chosen": -206.21484375, "logps/ref_chosen": -78.17408752441406, "logps/ref_rejected": -99.49610900878906, "logps/rejected": -302.2728271484375, "loss": 0.8316, "margin_dpo/margin_mean": 74.73595428466797, "margin_dpo/margin_std": 68.67886352539062, "step": 604 }, { "epoch": 0.8883994126284875, "fcm_dpo/beta": 0.01191855501383543, "fcm_dpo/delta": 0.10982121527194977, "fcm_dpo/margin": 62.803863525390625, "fcm_dpo/q_t": 0.3468341827392578, "grad_norm": 12.669288635253906, "learning_rate": 1.9276432015946446e-08, "logits/chosen": -1.9019720554351807, "logits/rejected": -1.4060518741607666, "logps/chosen": -232.10272216796875, "logps/ref_chosen": -94.77332305908203, "logps/ref_rejected": -107.30490112304688, "logps/rejected": -307.43817138671875, "loss": 0.9308, "margin_dpo/margin_mean": 62.80385971069336, "margin_dpo/margin_std": 72.63813781738281, "step": 605 }, { "epoch": 0.8898678414096917, "fcm_dpo/beta": 0.01211099699139595, "fcm_dpo/delta": -0.012885570526123047, "fcm_dpo/margin": 71.01239013671875, "fcm_dpo/q_t": 0.3213347792625427, "grad_norm": 13.087188720703125, "learning_rate": 1.8785323298722093e-08, "logits/chosen": -2.011061191558838, "logits/rejected": -1.5981237888336182, "logps/chosen": -220.9056396484375, "logps/ref_chosen": -87.7533950805664, "logps/ref_rejected": -104.2422103881836, "logps/rejected": -308.4068603515625, "loss": 0.8646, "margin_dpo/margin_mean": 71.01239013671875, "margin_dpo/margin_std": 71.47808837890625, "step": 606 }, { "epoch": 0.8913362701908958, "fcm_dpo/beta": 0.011857230216264725, "fcm_dpo/delta": -0.05570369213819504, "fcm_dpo/margin": 75.94203186035156, "fcm_dpo/q_t": 0.3244081139564514, "grad_norm": 13.238591194152832, "learning_rate": 1.8300307303259904e-08, "logits/chosen": -1.8522627353668213, "logits/rejected": -1.3280799388885498, "logps/chosen": -220.33203125, "logps/ref_chosen": -88.32904052734375, "logps/ref_rejected": -86.76810455322266, "logps/rejected": -294.713134765625, "loss": 0.9333, "margin_dpo/margin_mean": 75.94203186035156, "margin_dpo/margin_std": 94.53721618652344, "step": 607 }, { "epoch": 0.8928046989720999, "fcm_dpo/beta": 0.011611446738243103, "fcm_dpo/delta": -0.07125182449817657, "fcm_dpo/margin": 78.56434631347656, "fcm_dpo/q_t": 0.3104247748851776, "grad_norm": 14.735077857971191, "learning_rate": 1.7821396810182437e-08, "logits/chosen": -1.979656457901001, "logits/rejected": -1.6973553895950317, "logps/chosen": -205.81634521484375, "logps/ref_chosen": -85.76937866210938, "logps/ref_rejected": -100.23281860351562, "logps/rejected": -298.8441162109375, "loss": 0.8636, "margin_dpo/margin_mean": 78.56434631347656, "margin_dpo/margin_std": 80.66757202148438, "step": 608 }, { "epoch": 0.8942731277533039, "fcm_dpo/beta": 0.011500250548124313, "fcm_dpo/delta": -0.07149754464626312, "fcm_dpo/margin": 79.48577117919922, "fcm_dpo/q_t": 0.3168375492095947, "grad_norm": 13.526650428771973, "learning_rate": 1.7348604439226617e-08, "logits/chosen": -2.310915946960449, "logits/rejected": -1.89617919921875, "logps/chosen": -219.12498474121094, "logps/ref_chosen": -92.9665756225586, "logps/ref_rejected": -95.91818237304688, "logps/rejected": -301.5623474121094, "loss": 0.845, "margin_dpo/margin_mean": 79.48577117919922, "margin_dpo/margin_std": 80.12174224853516, "step": 609 }, { "epoch": 0.895741556534508, "fcm_dpo/beta": 0.011400583200156689, "fcm_dpo/delta": -0.05915520340204239, "fcm_dpo/margin": 79.26727294921875, "fcm_dpo/q_t": 0.3089755177497864, "grad_norm": 11.487713813781738, "learning_rate": 1.6881942648911074e-08, "logits/chosen": -1.8207099437713623, "logits/rejected": -1.2872872352600098, "logps/chosen": -221.53671264648438, "logps/ref_chosen": -94.70028686523438, "logps/ref_rejected": -89.68739318847656, "logps/rejected": -295.7911071777344, "loss": 0.8194, "margin_dpo/margin_mean": 79.26727294921875, "margin_dpo/margin_std": 72.37725830078125, "step": 610 }, { "epoch": 0.8972099853157122, "fcm_dpo/beta": 0.011329730972647667, "fcm_dpo/delta": -0.07637893408536911, "fcm_dpo/margin": 81.11207580566406, "fcm_dpo/q_t": 0.3176420331001282, "grad_norm": 12.095680236816406, "learning_rate": 1.6421423736208e-08, "logits/chosen": -1.9744651317596436, "logits/rejected": -1.6172444820404053, "logps/chosen": -213.77818298339844, "logps/ref_chosen": -86.78334045410156, "logps/ref_rejected": -89.84307861328125, "logps/rejected": -297.95001220703125, "loss": 0.8324, "margin_dpo/margin_mean": 81.11207580566406, "margin_dpo/margin_std": 81.14633178710938, "step": 611 }, { "epoch": 0.8986784140969163, "fcm_dpo/beta": 0.011108027771115303, "fcm_dpo/delta": -0.03341337665915489, "fcm_dpo/margin": 79.22400665283203, "fcm_dpo/q_t": 0.3166874051094055, "grad_norm": 12.299027442932129, "learning_rate": 1.5967059836219042e-08, "logits/chosen": -2.1468470096588135, "logits/rejected": -1.3182777166366577, "logps/chosen": -237.84152221679688, "logps/ref_chosen": -101.02015686035156, "logps/ref_rejected": -93.78302764892578, "logps/rejected": -309.82843017578125, "loss": 0.8232, "margin_dpo/margin_mean": 79.22400665283203, "margin_dpo/margin_std": 73.36766052246094, "step": 612 }, { "epoch": 0.9001468428781204, "fcm_dpo/beta": 0.011010472662746906, "fcm_dpo/delta": -0.06952960789203644, "fcm_dpo/margin": 82.93061828613281, "fcm_dpo/q_t": 0.3106223940849304, "grad_norm": 12.303360939025879, "learning_rate": 1.551886292185553e-08, "logits/chosen": -1.7026329040527344, "logits/rejected": -1.4959499835968018, "logps/chosen": -206.756591796875, "logps/ref_chosen": -88.98861694335938, "logps/ref_rejected": -109.99551391601562, "logps/rejected": -310.694091796875, "loss": 0.7956, "margin_dpo/margin_mean": 82.93061828613281, "margin_dpo/margin_std": 72.88465881347656, "step": 613 }, { "epoch": 0.9016152716593245, "fcm_dpo/beta": 0.011123564094305038, "fcm_dpo/delta": 0.09568654745817184, "fcm_dpo/margin": 68.47465515136719, "fcm_dpo/q_t": 0.3426240086555481, "grad_norm": 14.612199783325195, "learning_rate": 1.507684480352292e-08, "logits/chosen": -1.474619746208191, "logits/rejected": -1.4545549154281616, "logps/chosen": -220.43658447265625, "logps/ref_chosen": -80.20005798339844, "logps/ref_rejected": -109.86239624023438, "logps/rejected": -318.57354736328125, "loss": 0.9322, "margin_dpo/margin_mean": 68.47465515136719, "margin_dpo/margin_std": 78.6995849609375, "step": 614 }, { "epoch": 0.9030837004405287, "fcm_dpo/beta": 0.01152871549129486, "fcm_dpo/delta": 0.1148013100028038, "fcm_dpo/margin": 64.1284408569336, "fcm_dpo/q_t": 0.3421385884284973, "grad_norm": 12.715716361999512, "learning_rate": 1.4641017128809801e-08, "logits/chosen": -1.756435751914978, "logits/rejected": -1.351597785949707, "logps/chosen": -231.73660278320312, "logps/ref_chosen": -100.43525695800781, "logps/ref_rejected": -101.1800537109375, "logps/rejected": -296.60986328125, "loss": 0.9135, "margin_dpo/margin_mean": 64.1284408569336, "margin_dpo/margin_std": 64.88854217529297, "step": 615 }, { "epoch": 0.9045521292217328, "fcm_dpo/beta": 0.01139888633042574, "fcm_dpo/delta": -0.07935352623462677, "fcm_dpo/margin": 80.85199737548828, "fcm_dpo/q_t": 0.31489297747612, "grad_norm": 11.985855102539062, "learning_rate": 1.4211391382180637e-08, "logits/chosen": -1.806459665298462, "logits/rejected": -1.322357177734375, "logps/chosen": -227.93759155273438, "logps/ref_chosen": -92.49292755126953, "logps/ref_rejected": -82.06065368652344, "logps/rejected": -298.3572998046875, "loss": 0.8462, "margin_dpo/margin_mean": 80.85198974609375, "margin_dpo/margin_std": 83.83363342285156, "step": 616 }, { "epoch": 0.9060205580029369, "fcm_dpo/beta": 0.011469890363514423, "fcm_dpo/delta": 0.11235801875591278, "fcm_dpo/margin": 65.0345230102539, "fcm_dpo/q_t": 0.3446882367134094, "grad_norm": 13.72085952758789, "learning_rate": 1.378797888467345e-08, "logits/chosen": -1.5556066036224365, "logits/rejected": -0.915245771408081, "logps/chosen": -228.89675903320312, "logps/ref_chosen": -91.09699249267578, "logps/ref_rejected": -70.41004943847656, "logps/rejected": -273.24432373046875, "loss": 0.9388, "margin_dpo/margin_mean": 65.0345230102539, "margin_dpo/margin_std": 76.74852752685547, "step": 617 }, { "epoch": 0.9074889867841409, "fcm_dpo/beta": 0.011595992371439934, "fcm_dpo/delta": 0.06170836463570595, "fcm_dpo/margin": 68.39972686767578, "fcm_dpo/q_t": 0.33479219675064087, "grad_norm": 15.96437931060791, "learning_rate": 1.3370790793601371e-08, "logits/chosen": -2.159393310546875, "logits/rejected": -1.8255435228347778, "logps/chosen": -240.608642578125, "logps/ref_chosen": -102.02059936523438, "logps/ref_rejected": -99.80119323730469, "logps/rejected": -306.7889709472656, "loss": 0.931, "margin_dpo/margin_mean": 68.39972686767578, "margin_dpo/margin_std": 80.0145263671875, "step": 618 }, { "epoch": 0.908957415565345, "fcm_dpo/beta": 0.01181875355541706, "fcm_dpo/delta": 0.07929763197898865, "fcm_dpo/margin": 65.72555541992188, "fcm_dpo/q_t": 0.34781360626220703, "grad_norm": 12.926453590393066, "learning_rate": 1.2959838102258535e-08, "logits/chosen": -1.6464307308197021, "logits/rejected": -1.2737534046173096, "logps/chosen": -222.141845703125, "logps/ref_chosen": -89.74136352539062, "logps/ref_rejected": -99.90138244628906, "logps/rejected": -298.02740478515625, "loss": 0.9538, "margin_dpo/margin_mean": 65.72555541992188, "margin_dpo/margin_std": 84.30938720703125, "step": 619 }, { "epoch": 0.9104258443465492, "fcm_dpo/beta": 0.01165106613188982, "fcm_dpo/delta": -0.0010028481483459473, "fcm_dpo/margin": 72.52044677734375, "fcm_dpo/q_t": 0.3334208130836487, "grad_norm": 13.063641548156738, "learning_rate": 1.2555131639630567e-08, "logits/chosen": -2.2908899784088135, "logits/rejected": -1.7881639003753662, "logps/chosen": -214.04869079589844, "logps/ref_chosen": -85.12431335449219, "logps/ref_rejected": -85.41253662109375, "logps/rejected": -286.85736083984375, "loss": 0.9281, "margin_dpo/margin_mean": 72.52044677734375, "margin_dpo/margin_std": 82.46279907226562, "step": 620 }, { "epoch": 0.9118942731277533, "fcm_dpo/beta": 0.012033706530928612, "fcm_dpo/delta": 0.00012670457363128662, "fcm_dpo/margin": 70.27881622314453, "fcm_dpo/q_t": 0.3351132273674011, "grad_norm": 13.528855323791504, "learning_rate": 1.2156682070109086e-08, "logits/chosen": -1.9692211151123047, "logits/rejected": -1.7110702991485596, "logps/chosen": -219.129638671875, "logps/ref_chosen": -89.24842071533203, "logps/ref_rejected": -95.46463775634766, "logps/rejected": -295.6246643066406, "loss": 0.9452, "margin_dpo/margin_mean": 70.27881622314453, "margin_dpo/margin_std": 86.90354919433594, "step": 621 }, { "epoch": 0.9133627019089574, "fcm_dpo/beta": 0.011807737872004509, "fcm_dpo/delta": -0.022110551595687866, "fcm_dpo/margin": 73.69373321533203, "fcm_dpo/q_t": 0.32350027561187744, "grad_norm": 15.262984275817871, "learning_rate": 1.1764499893210878e-08, "logits/chosen": -1.2746450901031494, "logits/rejected": -0.5793788433074951, "logps/chosen": -229.22116088867188, "logps/ref_chosen": -99.79412841796875, "logps/ref_rejected": -90.82821655273438, "logps/rejected": -293.948974609375, "loss": 0.8526, "margin_dpo/margin_mean": 73.6937255859375, "margin_dpo/margin_std": 74.19442749023438, "step": 622 }, { "epoch": 0.9148311306901615, "fcm_dpo/beta": 0.012104466557502747, "fcm_dpo/delta": 0.20468077063560486, "fcm_dpo/margin": 54.466007232666016, "fcm_dpo/q_t": 0.37531763315200806, "grad_norm": 16.303054809570312, "learning_rate": 1.1378595443300998e-08, "logits/chosen": -2.4043192863464355, "logits/rejected": -1.9884238243103027, "logps/chosen": -225.4462432861328, "logps/ref_chosen": -90.45555114746094, "logps/ref_rejected": -91.32276153564453, "logps/rejected": -280.7794494628906, "loss": 1.0759, "margin_dpo/margin_mean": 54.466007232666016, "margin_dpo/margin_std": 88.74128723144531, "step": 623 }, { "epoch": 0.9162995594713657, "fcm_dpo/beta": 0.01205630972981453, "fcm_dpo/delta": -0.1360035389661789, "fcm_dpo/margin": 80.68819427490234, "fcm_dpo/q_t": 0.3046037554740906, "grad_norm": 13.448029518127441, "learning_rate": 1.0998978889320582e-08, "logits/chosen": -2.6992523670196533, "logits/rejected": -1.788914442062378, "logps/chosen": -244.6612548828125, "logps/ref_chosen": -109.87522888183594, "logps/ref_rejected": -104.77320861816406, "logps/rejected": -320.2474365234375, "loss": 0.8112, "margin_dpo/margin_mean": 80.68818664550781, "margin_dpo/margin_std": 78.43266296386719, "step": 624 }, { "epoch": 0.9177679882525698, "fcm_dpo/beta": 0.012182200327515602, "fcm_dpo/delta": 0.10253170877695084, "fcm_dpo/margin": 61.96144104003906, "fcm_dpo/q_t": 0.3474324345588684, "grad_norm": 13.418804168701172, "learning_rate": 1.0625660234518913e-08, "logits/chosen": -1.6345865726470947, "logits/rejected": -1.2501837015151978, "logps/chosen": -230.34454345703125, "logps/ref_chosen": -87.16815948486328, "logps/ref_rejected": -91.86148071289062, "logps/rejected": -296.99932861328125, "loss": 0.9269, "margin_dpo/margin_mean": 61.96144104003906, "margin_dpo/margin_std": 72.25453186035156, "step": 625 }, { "epoch": 0.9192364170337739, "fcm_dpo/beta": 0.012475828640162945, "fcm_dpo/delta": 0.10793514549732208, "fcm_dpo/margin": 60.04652404785156, "fcm_dpo/q_t": 0.3585766553878784, "grad_norm": 15.548441886901855, "learning_rate": 1.0258649316189721e-08, "logits/chosen": -1.4249701499938965, "logits/rejected": -0.9961116313934326, "logps/chosen": -241.67068481445312, "logps/ref_chosen": -104.22421264648438, "logps/ref_rejected": -104.1774673461914, "logps/rejected": -301.67047119140625, "loss": 1.0141, "margin_dpo/margin_mean": 60.04652404785156, "margin_dpo/margin_std": 88.77542877197266, "step": 626 }, { "epoch": 0.920704845814978, "fcm_dpo/beta": 0.012376347556710243, "fcm_dpo/delta": -0.04472092539072037, "fcm_dpo/margin": 71.95941162109375, "fcm_dpo/q_t": 0.32779771089553833, "grad_norm": 11.651665687561035, "learning_rate": 9.897955805412e-09, "logits/chosen": -1.5501372814178467, "logits/rejected": -1.6814649105072021, "logps/chosen": -200.94879150390625, "logps/ref_chosen": -74.93461608886719, "logps/ref_rejected": -112.57289123535156, "logps/rejected": -310.5464782714844, "loss": 0.9017, "margin_dpo/margin_mean": 71.95941162109375, "margin_dpo/margin_std": 85.76150512695312, "step": 627 }, { "epoch": 0.922173274596182, "fcm_dpo/beta": 0.012193206697702408, "fcm_dpo/delta": -0.09546506404876709, "fcm_dpo/margin": 76.80429077148438, "fcm_dpo/q_t": 0.308502733707428, "grad_norm": 14.485041618347168, "learning_rate": 9.543589206795238e-09, "logits/chosen": -2.234900712966919, "logits/rejected": -1.9763065576553345, "logps/chosen": -223.16265869140625, "logps/ref_chosen": -93.69107818603516, "logps/ref_rejected": -107.34395599365234, "logps/rejected": -313.61981201171875, "loss": 0.8052, "margin_dpo/margin_mean": 76.80429077148438, "margin_dpo/margin_std": 70.0140609741211, "step": 628 }, { "epoch": 0.9236417033773862, "fcm_dpo/beta": 0.012154782190918922, "fcm_dpo/delta": 0.039435502141714096, "fcm_dpo/margin": 66.93384552001953, "fcm_dpo/q_t": 0.33409008383750916, "grad_norm": 14.45174789428711, "learning_rate": 9.19555885822887e-09, "logits/chosen": -2.140451431274414, "logits/rejected": -1.7289149761199951, "logps/chosen": -241.05328369140625, "logps/ref_chosen": -103.23037719726562, "logps/ref_rejected": -97.16841888427734, "logps/rejected": -301.9251708984375, "loss": 0.924, "margin_dpo/margin_mean": 66.93384552001953, "margin_dpo/margin_std": 78.18603515625, "step": 629 }, { "epoch": 0.9251101321585903, "fcm_dpo/beta": 0.012714162468910217, "fcm_dpo/delta": 0.19320565462112427, "fcm_dpo/margin": 52.378536224365234, "fcm_dpo/q_t": 0.3655824065208435, "grad_norm": 17.370262145996094, "learning_rate": 8.85387393063622e-09, "logits/chosen": -2.005028247833252, "logits/rejected": -1.4612559080123901, "logps/chosen": -221.4185333251953, "logps/ref_chosen": -93.89755249023438, "logps/ref_rejected": -89.3743896484375, "logps/rejected": -269.27392578125, "loss": 1.0296, "margin_dpo/margin_mean": 52.37853240966797, "margin_dpo/margin_std": 74.43370819091797, "step": 630 }, { "epoch": 0.9265785609397944, "fcm_dpo/beta": 0.012599373236298561, "fcm_dpo/delta": -0.037173446267843246, "fcm_dpo/margin": 70.12379455566406, "fcm_dpo/q_t": 0.3170943558216095, "grad_norm": 13.337695121765137, "learning_rate": 8.518543427732949e-09, "logits/chosen": -2.624936580657959, "logits/rejected": -2.2076611518859863, "logps/chosen": -211.30435180664062, "logps/ref_chosen": -87.77082061767578, "logps/ref_rejected": -88.68241882324219, "logps/rejected": -282.3397521972656, "loss": 0.8397, "margin_dpo/margin_mean": 70.12379455566406, "margin_dpo/margin_std": 68.069580078125, "step": 631 }, { "epoch": 0.9280469897209985, "fcm_dpo/beta": 0.012279413640499115, "fcm_dpo/delta": -0.10962569713592529, "fcm_dpo/margin": 77.01805114746094, "fcm_dpo/q_t": 0.31531599164009094, "grad_norm": 13.519450187683105, "learning_rate": 8.189576185789637e-09, "logits/chosen": -1.8649694919586182, "logits/rejected": -1.4337584972381592, "logps/chosen": -212.1048583984375, "logps/ref_chosen": -88.62652587890625, "logps/ref_rejected": -91.45091247558594, "logps/rejected": -291.9472961425781, "loss": 0.8654, "margin_dpo/margin_mean": 77.01805114746094, "margin_dpo/margin_std": 84.19517517089844, "step": 632 }, { "epoch": 0.9295154185022027, "fcm_dpo/beta": 0.01287509873509407, "fcm_dpo/delta": 0.20589163899421692, "fcm_dpo/margin": 50.70753479003906, "fcm_dpo/q_t": 0.38306355476379395, "grad_norm": 18.226625442504883, "learning_rate": 7.866980873399015e-09, "logits/chosen": -2.24927020072937, "logits/rejected": -2.0586564540863037, "logps/chosen": -229.56752014160156, "logps/ref_chosen": -81.37442016601562, "logps/ref_rejected": -98.62571716308594, "logps/rejected": -297.5263671875, "loss": 1.1293, "margin_dpo/margin_mean": 50.70753479003906, "margin_dpo/margin_std": 94.33064270019531, "step": 633 }, { "epoch": 0.9309838472834068, "fcm_dpo/beta": 0.013133701868355274, "fcm_dpo/delta": 0.03609214723110199, "fcm_dpo/margin": 61.819000244140625, "fcm_dpo/q_t": 0.33319538831710815, "grad_norm": 15.686573028564453, "learning_rate": 7.550765991247654e-09, "logits/chosen": -1.420647382736206, "logits/rejected": -1.2333729267120361, "logps/chosen": -237.35031127929688, "logps/ref_chosen": -96.12284851074219, "logps/ref_rejected": -112.84780883789062, "logps/rejected": -315.8942565917969, "loss": 0.8899, "margin_dpo/margin_mean": 61.819000244140625, "margin_dpo/margin_std": 63.20722198486328, "step": 634 }, { "epoch": 0.9324522760646109, "fcm_dpo/beta": 0.013437749817967415, "fcm_dpo/delta": 0.19897405803203583, "fcm_dpo/margin": 49.268463134765625, "fcm_dpo/q_t": 0.3701525330543518, "grad_norm": 18.140663146972656, "learning_rate": 7.240939871891699e-09, "logits/chosen": -1.8061546087265015, "logits/rejected": -1.2960937023162842, "logps/chosen": -240.75674438476562, "logps/ref_chosen": -98.68411254882812, "logps/ref_rejected": -89.89920043945312, "logps/rejected": -281.2402648925781, "loss": 1.0481, "margin_dpo/margin_mean": 49.26845932006836, "margin_dpo/margin_std": 77.19035339355469, "step": 635 }, { "epoch": 0.933920704845815, "fcm_dpo/beta": 0.013280387967824936, "fcm_dpo/delta": -0.12279319763183594, "fcm_dpo/margin": 72.37382507324219, "fcm_dpo/q_t": 0.3140905499458313, "grad_norm": 13.60958194732666, "learning_rate": 6.937510679537628e-09, "logits/chosen": -1.3869647979736328, "logits/rejected": -0.9099713563919067, "logps/chosen": -223.68519592285156, "logps/ref_chosen": -90.41796112060547, "logps/ref_rejected": -87.70687866210938, "logps/rejected": -293.34796142578125, "loss": 0.8487, "margin_dpo/margin_mean": 72.37382507324219, "margin_dpo/margin_std": 77.6722640991211, "step": 636 }, { "epoch": 0.9353891336270191, "fcm_dpo/beta": 0.012870658189058304, "fcm_dpo/delta": -0.16235065460205078, "fcm_dpo/margin": 77.41453552246094, "fcm_dpo/q_t": 0.30641061067581177, "grad_norm": 13.35957145690918, "learning_rate": 6.640486409826785e-09, "logits/chosen": -2.052138328552246, "logits/rejected": -1.9478278160095215, "logps/chosen": -216.7566680908203, "logps/ref_chosen": -82.44971466064453, "logps/ref_rejected": -104.02860260009766, "logps/rejected": -315.7500915527344, "loss": 0.838, "margin_dpo/margin_mean": 77.41453552246094, "margin_dpo/margin_std": 81.38263702392578, "step": 637 }, { "epoch": 0.9368575624082232, "fcm_dpo/beta": 0.012524936348199844, "fcm_dpo/delta": -0.18407437205314636, "fcm_dpo/margin": 81.04855346679688, "fcm_dpo/q_t": 0.2905474603176117, "grad_norm": 11.3641357421875, "learning_rate": 6.349874889624962e-09, "logits/chosen": -1.6746790409088135, "logits/rejected": -1.0038890838623047, "logps/chosen": -213.70997619628906, "logps/ref_chosen": -91.92498779296875, "logps/ref_rejected": -86.28703308105469, "logps/rejected": -289.1205749511719, "loss": 0.7404, "margin_dpo/margin_mean": 81.04855346679688, "margin_dpo/margin_std": 61.875587463378906, "step": 638 }, { "epoch": 0.9383259911894273, "fcm_dpo/beta": 0.012901335954666138, "fcm_dpo/delta": 0.29382050037384033, "fcm_dpo/margin": 44.23047637939453, "fcm_dpo/q_t": 0.38710546493530273, "grad_norm": 17.964210510253906, "learning_rate": 6.065683776815933e-09, "logits/chosen": -1.2971713542938232, "logits/rejected": -0.5256457328796387, "logps/chosen": -265.3831481933594, "logps/ref_chosen": -104.52755737304688, "logps/ref_rejected": -81.4803466796875, "logps/rejected": -286.56640625, "loss": 1.1298, "margin_dpo/margin_mean": 44.23047637939453, "margin_dpo/margin_std": 80.78390502929688, "step": 639 }, { "epoch": 0.9397944199706314, "fcm_dpo/beta": 0.012978674843907356, "fcm_dpo/delta": -0.14304372668266296, "fcm_dpo/margin": 75.16079711914062, "fcm_dpo/q_t": 0.3036750257015228, "grad_norm": 14.67367935180664, "learning_rate": 5.7879205600998296e-09, "logits/chosen": -1.6546118259429932, "logits/rejected": -1.1096612215042114, "logps/chosen": -232.30517578125, "logps/ref_chosen": -97.88526916503906, "logps/ref_rejected": -112.70502471923828, "logps/rejected": -322.28570556640625, "loss": 0.8363, "margin_dpo/margin_mean": 75.16079711914062, "margin_dpo/margin_std": 71.88284301757812, "step": 640 }, { "epoch": 0.9412628487518355, "fcm_dpo/beta": 0.01246521808207035, "fcm_dpo/delta": -0.10708156228065491, "fcm_dpo/margin": 75.97200012207031, "fcm_dpo/q_t": 0.3164488673210144, "grad_norm": 13.984350204467773, "learning_rate": 5.516592558795746e-09, "logits/chosen": -2.2499518394470215, "logits/rejected": -1.7141563892364502, "logps/chosen": -221.30508422851562, "logps/ref_chosen": -96.4456787109375, "logps/ref_rejected": -95.13568878173828, "logps/rejected": -295.96710205078125, "loss": 0.8656, "margin_dpo/margin_mean": 75.97200775146484, "margin_dpo/margin_std": 85.18379211425781, "step": 641 }, { "epoch": 0.9427312775330396, "fcm_dpo/beta": 0.012517379596829414, "fcm_dpo/delta": 0.03650232404470444, "fcm_dpo/margin": 65.17138671875, "fcm_dpo/q_t": 0.3425823152065277, "grad_norm": 14.213191032409668, "learning_rate": 5.251706922648868e-09, "logits/chosen": -1.5260307788848877, "logits/rejected": -1.1929254531860352, "logps/chosen": -242.54896545410156, "logps/ref_chosen": -100.75984954833984, "logps/ref_rejected": -114.7076416015625, "logps/rejected": -321.66815185546875, "loss": 0.9487, "margin_dpo/margin_mean": 65.17138671875, "margin_dpo/margin_std": 85.03411865234375, "step": 642 }, { "epoch": 0.9441997063142438, "fcm_dpo/beta": 0.012432137504220009, "fcm_dpo/delta": 0.05663251131772995, "fcm_dpo/margin": 64.00656127929688, "fcm_dpo/q_t": 0.33634692430496216, "grad_norm": 14.948612213134766, "learning_rate": 4.993270631642038e-09, "logits/chosen": -2.1365489959716797, "logits/rejected": -1.7338988780975342, "logps/chosen": -218.6114501953125, "logps/ref_chosen": -84.74365997314453, "logps/ref_rejected": -94.31842041015625, "logps/rejected": -292.1927490234375, "loss": 0.9151, "margin_dpo/margin_mean": 64.00656127929688, "margin_dpo/margin_std": 70.07855224609375, "step": 643 }, { "epoch": 0.9456681350954479, "fcm_dpo/beta": 0.012908038683235645, "fcm_dpo/delta": 0.14631298184394836, "fcm_dpo/margin": 55.31208801269531, "fcm_dpo/q_t": 0.35338038206100464, "grad_norm": 17.55486488342285, "learning_rate": 4.741290495811873e-09, "logits/chosen": -1.9027633666992188, "logits/rejected": -1.626082181930542, "logps/chosen": -219.79922485351562, "logps/ref_chosen": -85.32274627685547, "logps/ref_rejected": -94.60861206054688, "logps/rejected": -284.39715576171875, "loss": 0.9612, "margin_dpo/margin_mean": 55.31208801269531, "margin_dpo/margin_std": 68.78119659423828, "step": 644 }, { "epoch": 0.947136563876652, "fcm_dpo/beta": 0.013628844171762466, "fcm_dpo/delta": 0.29724252223968506, "fcm_dpo/margin": 41.70508575439453, "fcm_dpo/q_t": 0.3945990204811096, "grad_norm": 17.644214630126953, "learning_rate": 4.495773155069299e-09, "logits/chosen": -1.4874303340911865, "logits/rejected": -1.2977774143218994, "logps/chosen": -227.33688354492188, "logps/ref_chosen": -82.59024047851562, "logps/ref_rejected": -103.00375366210938, "logps/rejected": -289.45550537109375, "loss": 1.1689, "margin_dpo/margin_mean": 41.70508575439453, "margin_dpo/margin_std": 83.13195037841797, "step": 645 }, { "epoch": 0.9486049926578561, "fcm_dpo/beta": 0.013419127091765404, "fcm_dpo/delta": -0.12888050079345703, "fcm_dpo/margin": 71.85616302490234, "fcm_dpo/q_t": 0.3052224814891815, "grad_norm": 14.331748962402344, "learning_rate": 4.256725079024553e-09, "logits/chosen": -1.8919644355773926, "logits/rejected": -1.2960618734359741, "logps/chosen": -217.77532958984375, "logps/ref_chosen": -91.1920394897461, "logps/ref_rejected": -83.77833557128906, "logps/rejected": -282.2177734375, "loss": 0.8021, "margin_dpo/margin_mean": 71.85616302490234, "margin_dpo/margin_std": 66.11752319335938, "step": 646 }, { "epoch": 0.9500734214390602, "fcm_dpo/beta": 0.01349719800055027, "fcm_dpo/delta": 0.049388255923986435, "fcm_dpo/margin": 59.609161376953125, "fcm_dpo/q_t": 0.3525710701942444, "grad_norm": 16.58894920349121, "learning_rate": 4.024152566816791e-09, "logits/chosen": -1.5568263530731201, "logits/rejected": -1.4186091423034668, "logps/chosen": -230.43569946289062, "logps/ref_chosen": -88.8444595336914, "logps/ref_rejected": -99.49833679199219, "logps/rejected": -300.6987609863281, "loss": 1.0062, "margin_dpo/margin_mean": 59.609169006347656, "margin_dpo/margin_std": 87.35142517089844, "step": 647 }, { "epoch": 0.9515418502202643, "fcm_dpo/beta": 0.01303325779736042, "fcm_dpo/delta": -0.30287304520606995, "fcm_dpo/margin": 85.91302490234375, "fcm_dpo/q_t": 0.283531129360199, "grad_norm": 13.930960655212402, "learning_rate": 3.798061746947995e-09, "logits/chosen": -2.1560635566711426, "logits/rejected": -2.075185537338257, "logps/chosen": -216.96524047851562, "logps/ref_chosen": -87.84810638427734, "logps/ref_rejected": -104.67005920410156, "logps/rejected": -319.7001953125, "loss": 0.7651, "margin_dpo/margin_mean": 85.91301727294922, "margin_dpo/margin_std": 82.10009002685547, "step": 648 }, { "epoch": 0.9530102790014684, "fcm_dpo/beta": 0.012625191360712051, "fcm_dpo/delta": -0.027126122266054153, "fcm_dpo/margin": 69.18255615234375, "fcm_dpo/q_t": 0.3202654719352722, "grad_norm": 13.677901268005371, "learning_rate": 3.5784585771215235e-09, "logits/chosen": -2.4737744331359863, "logits/rejected": -2.0878794193267822, "logps/chosen": -210.0411376953125, "logps/ref_chosen": -89.6925048828125, "logps/ref_rejected": -88.70658111572266, "logps/rejected": -278.2377624511719, "loss": 0.8399, "margin_dpo/margin_mean": 69.18255615234375, "margin_dpo/margin_std": 65.37664794921875, "step": 649 }, { "epoch": 0.9544787077826725, "fcm_dpo/beta": 0.012697836384177208, "fcm_dpo/delta": -0.015169069170951843, "fcm_dpo/margin": 68.02926635742188, "fcm_dpo/q_t": 0.32649314403533936, "grad_norm": 15.16059684753418, "learning_rate": 3.3653488440851253e-09, "logits/chosen": -1.3418605327606201, "logits/rejected": -1.1561648845672607, "logps/chosen": -227.9875030517578, "logps/ref_chosen": -89.93060302734375, "logps/ref_rejected": -102.61282348632812, "logps/rejected": -308.698974609375, "loss": 0.9067, "margin_dpo/margin_mean": 68.02926635742188, "margin_dpo/margin_std": 80.34864807128906, "step": 650 }, { "epoch": 0.9559471365638766, "fcm_dpo/beta": 0.012280041351914406, "fcm_dpo/delta": -0.18812772631645203, "fcm_dpo/margin": 82.9175796508789, "fcm_dpo/q_t": 0.2923458218574524, "grad_norm": 12.51254653930664, "learning_rate": 3.158738163478475e-09, "logits/chosen": -2.165620803833008, "logits/rejected": -2.0002899169921875, "logps/chosen": -200.1165771484375, "logps/ref_chosen": -79.18731689453125, "logps/ref_rejected": -105.93333435058594, "logps/rejected": -309.7801513671875, "loss": 0.7648, "margin_dpo/margin_mean": 82.91757202148438, "margin_dpo/margin_std": 71.92076110839844, "step": 651 }, { "epoch": 0.9574155653450808, "fcm_dpo/beta": 0.012219004333019257, "fcm_dpo/delta": -0.007074171677231789, "fcm_dpo/margin": 70.08258819580078, "fcm_dpo/q_t": 0.3215482532978058, "grad_norm": 12.98484992980957, "learning_rate": 2.9586319796851555e-09, "logits/chosen": -2.567788600921631, "logits/rejected": -2.267869472503662, "logps/chosen": -223.45941162109375, "logps/ref_chosen": -101.79022216796875, "logps/ref_rejected": -116.3245849609375, "logps/rejected": -308.07635498046875, "loss": 0.8286, "margin_dpo/margin_mean": 70.08259582519531, "margin_dpo/margin_std": 64.39094543457031, "step": 652 }, { "epoch": 0.9588839941262849, "fcm_dpo/beta": 0.012271079234778881, "fcm_dpo/delta": 0.017625533044338226, "fcm_dpo/margin": 67.90986633300781, "fcm_dpo/q_t": 0.3277429938316345, "grad_norm": 13.8803071975708, "learning_rate": 2.7650355656892166e-09, "logits/chosen": -2.142395496368408, "logits/rejected": -1.8992211818695068, "logps/chosen": -230.5672607421875, "logps/ref_chosen": -93.35359191894531, "logps/ref_rejected": -109.12324523925781, "logps/rejected": -314.2467956542969, "loss": 0.8909, "margin_dpo/margin_mean": 67.90987396240234, "margin_dpo/margin_std": 73.30245208740234, "step": 653 }, { "epoch": 0.960352422907489, "fcm_dpo/beta": 0.012467766180634499, "fcm_dpo/delta": 0.13704301416873932, "fcm_dpo/margin": 57.98382568359375, "fcm_dpo/q_t": 0.36294329166412354, "grad_norm": 18.976091384887695, "learning_rate": 2.577954022936174e-09, "logits/chosen": -2.1631104946136475, "logits/rejected": -1.970808982849121, "logps/chosen": -229.53439331054688, "logps/ref_chosen": -89.11553955078125, "logps/ref_rejected": -104.91995239257812, "logps/rejected": -303.3226318359375, "loss": 1.0347, "margin_dpo/margin_mean": 57.98382568359375, "margin_dpo/margin_std": 89.45317077636719, "step": 654 }, { "epoch": 0.9618208516886931, "fcm_dpo/beta": 0.01260659471154213, "fcm_dpo/delta": -0.004192095249891281, "fcm_dpo/margin": 67.70552062988281, "fcm_dpo/q_t": 0.327122300863266, "grad_norm": 14.10622787475586, "learning_rate": 2.397392281198729e-09, "logits/chosen": -2.0303776264190674, "logits/rejected": -2.0248003005981445, "logps/chosen": -212.3218231201172, "logps/ref_chosen": -81.03610229492188, "logps/ref_rejected": -102.80233764648438, "logps/rejected": -301.7935791015625, "loss": 0.9029, "margin_dpo/margin_mean": 67.70552825927734, "margin_dpo/margin_std": 76.8497314453125, "step": 655 }, { "epoch": 0.9632892804698973, "fcm_dpo/beta": 0.012112503871321678, "fcm_dpo/delta": -0.2109536975622177, "fcm_dpo/margin": 85.67755126953125, "fcm_dpo/q_t": 0.2922847270965576, "grad_norm": 15.038117408752441, "learning_rate": 2.223355098446622e-09, "logits/chosen": -1.5206151008605957, "logits/rejected": -1.590084195137024, "logps/chosen": -213.54559326171875, "logps/ref_chosen": -85.32534790039062, "logps/ref_rejected": -118.33866119384766, "logps/rejected": -332.2364501953125, "loss": 0.7521, "margin_dpo/margin_mean": 85.67755126953125, "margin_dpo/margin_std": 75.42826843261719, "step": 656 }, { "epoch": 0.9647577092511013, "fcm_dpo/beta": 0.011937540024518967, "fcm_dpo/delta": -0.011723875999450684, "fcm_dpo/margin": 72.03126525878906, "fcm_dpo/q_t": 0.33278656005859375, "grad_norm": 14.324390411376953, "learning_rate": 2.055847060721566e-09, "logits/chosen": -1.9533765316009521, "logits/rejected": -1.7715580463409424, "logps/chosen": -213.4752197265625, "logps/ref_chosen": -80.19772338867188, "logps/ref_rejected": -102.581298828125, "logps/rejected": -307.89007568359375, "loss": 0.9101, "margin_dpo/margin_mean": 72.03126525878906, "margin_dpo/margin_std": 85.60971069335938, "step": 657 }, { "epoch": 0.9662261380323054, "fcm_dpo/beta": 0.012269849888980389, "fcm_dpo/delta": 0.10941661894321442, "fcm_dpo/margin": 60.93087387084961, "fcm_dpo/q_t": 0.3474779427051544, "grad_norm": 14.329385757446289, "learning_rate": 1.8948725820160662e-09, "logits/chosen": -2.008090019226074, "logits/rejected": -1.540211796760559, "logps/chosen": -236.1888427734375, "logps/ref_chosen": -94.634521484375, "logps/ref_rejected": -101.6316146850586, "logps/rejected": -304.1168212890625, "loss": 0.9428, "margin_dpo/margin_mean": 60.930870056152344, "margin_dpo/margin_std": 73.01217651367188, "step": 658 }, { "epoch": 0.9676945668135095, "fcm_dpo/beta": 0.011998293921351433, "fcm_dpo/delta": -0.11588911712169647, "fcm_dpo/margin": 79.48670959472656, "fcm_dpo/q_t": 0.31771883368492126, "grad_norm": 13.859298706054688, "learning_rate": 1.7404359041573723e-09, "logits/chosen": -2.127384662628174, "logits/rejected": -1.1833677291870117, "logps/chosen": -234.04229736328125, "logps/ref_chosen": -112.55587005615234, "logps/ref_rejected": -93.9216079711914, "logps/rejected": -294.8947448730469, "loss": 0.9104, "margin_dpo/margin_mean": 79.48670959472656, "margin_dpo/margin_std": 94.99459838867188, "step": 659 }, { "epoch": 0.9691629955947136, "fcm_dpo/beta": 0.01178102195262909, "fcm_dpo/delta": -0.10657566785812378, "fcm_dpo/margin": 80.31880187988281, "fcm_dpo/q_t": 0.31062257289886475, "grad_norm": 13.350107192993164, "learning_rate": 1.592541096695571e-09, "logits/chosen": -1.8621575832366943, "logits/rejected": -1.2513611316680908, "logps/chosen": -220.84124755859375, "logps/ref_chosen": -93.37742614746094, "logps/ref_rejected": -81.39482116699219, "logps/rejected": -289.17742919921875, "loss": 0.8279, "margin_dpo/margin_mean": 80.31880187988281, "margin_dpo/margin_std": 80.75775909423828, "step": 660 }, { "epoch": 0.9706314243759178, "fcm_dpo/beta": 0.011570056900382042, "fcm_dpo/delta": -0.12267627567052841, "fcm_dpo/margin": 83.06248474121094, "fcm_dpo/q_t": 0.30892378091812134, "grad_norm": 11.488774299621582, "learning_rate": 1.4511920567963908e-09, "logits/chosen": -1.920933723449707, "logits/rejected": -1.4082481861114502, "logps/chosen": -223.60641479492188, "logps/ref_chosen": -87.85516357421875, "logps/ref_rejected": -92.40330505371094, "logps/rejected": -311.217041015625, "loss": 0.8254, "margin_dpo/margin_mean": 83.06248474121094, "margin_dpo/margin_std": 83.68098449707031, "step": 661 }, { "epoch": 0.9720998531571219, "fcm_dpo/beta": 0.011501193046569824, "fcm_dpo/delta": 0.00793197751045227, "fcm_dpo/margin": 73.25682067871094, "fcm_dpo/q_t": 0.3226693272590637, "grad_norm": 11.854416847229004, "learning_rate": 1.3163925091384532e-09, "logits/chosen": -1.7558529376983643, "logits/rejected": -1.2050141096115112, "logps/chosen": -232.50390625, "logps/ref_chosen": -102.77980041503906, "logps/ref_rejected": -95.22531127929688, "logps/rejected": -298.20623779296875, "loss": 0.8323, "margin_dpo/margin_mean": 73.25682067871094, "margin_dpo/margin_std": 65.54705810546875, "step": 662 }, { "epoch": 0.973568281938326, "fcm_dpo/beta": 0.011573947966098785, "fcm_dpo/delta": 0.031047917902469635, "fcm_dpo/margin": 70.93704223632812, "fcm_dpo/q_t": 0.3348495364189148, "grad_norm": 12.32866096496582, "learning_rate": 1.1881460058152382e-09, "logits/chosen": -2.1356537342071533, "logits/rejected": -1.9021480083465576, "logps/chosen": -223.847900390625, "logps/ref_chosen": -96.34658813476562, "logps/ref_rejected": -120.52645874023438, "logps/rejected": -318.9648132324219, "loss": 0.9124, "margin_dpo/margin_mean": 70.93704223632812, "margin_dpo/margin_std": 81.27235412597656, "step": 663 }, { "epoch": 0.9750367107195301, "fcm_dpo/beta": 0.011720478534698486, "fcm_dpo/delta": 0.0888986736536026, "fcm_dpo/margin": 65.49781799316406, "fcm_dpo/q_t": 0.3386794328689575, "grad_norm": 14.023902893066406, "learning_rate": 1.066455926241383e-09, "logits/chosen": -1.7385942935943604, "logits/rejected": -1.5541484355926514, "logps/chosen": -230.68548583984375, "logps/ref_chosen": -91.84242248535156, "logps/ref_rejected": -111.8366928100586, "logps/rejected": -316.17755126953125, "loss": 0.9057, "margin_dpo/margin_mean": 65.49781036376953, "margin_dpo/margin_std": 69.79621124267578, "step": 664 }, { "epoch": 0.9765051395007343, "fcm_dpo/beta": 0.01180583517998457, "fcm_dpo/delta": 0.018876872956752777, "fcm_dpo/margin": 70.52334594726562, "fcm_dpo/q_t": 0.3239971399307251, "grad_norm": 12.435693740844727, "learning_rate": 9.513254770636137e-10, "logits/chosen": -2.437134265899658, "logits/rejected": -2.011138677597046, "logps/chosen": -209.5286865234375, "logps/ref_chosen": -88.18618774414062, "logps/ref_rejected": -91.9120101928711, "logps/rejected": -283.77789306640625, "loss": 0.853, "margin_dpo/margin_mean": 70.52334594726562, "margin_dpo/margin_std": 67.84070587158203, "step": 665 }, { "epoch": 0.9779735682819384, "fcm_dpo/beta": 0.011835319921374321, "fcm_dpo/delta": 0.05988616496324539, "fcm_dpo/margin": 67.1385726928711, "fcm_dpo/q_t": 0.34297046065330505, "grad_norm": 15.745160102844238, "learning_rate": 8.427576920763956e-10, "logits/chosen": -1.4836091995239258, "logits/rejected": -1.1046907901763916, "logps/chosen": -237.0187225341797, "logps/ref_chosen": -100.97459411621094, "logps/ref_rejected": -101.24992370605469, "logps/rejected": -304.4326171875, "loss": 0.949, "margin_dpo/margin_mean": 67.13858032226562, "margin_dpo/margin_std": 83.11888122558594, "step": 666 }, { "epoch": 0.9794419970631424, "fcm_dpo/beta": 0.011725520715117455, "fcm_dpo/delta": -0.0794256329536438, "fcm_dpo/margin": 78.543212890625, "fcm_dpo/q_t": 0.3089987635612488, "grad_norm": 12.336382865905762, "learning_rate": 7.407554321417764e-10, "logits/chosen": -1.6863865852355957, "logits/rejected": -1.050025463104248, "logps/chosen": -240.4290771484375, "logps/ref_chosen": -97.5711669921875, "logps/ref_rejected": -93.58476257324219, "logps/rejected": -314.98590087890625, "loss": 0.8034, "margin_dpo/margin_mean": 78.543212890625, "margin_dpo/margin_std": 69.23985290527344, "step": 667 }, { "epoch": 0.9809104258443465, "fcm_dpo/beta": 0.011529900133609772, "fcm_dpo/delta": -0.15593984723091125, "fcm_dpo/margin": 85.928466796875, "fcm_dpo/q_t": 0.30568087100982666, "grad_norm": 14.262736320495605, "learning_rate": 6.453213851142225e-10, "logits/chosen": -2.211690902709961, "logits/rejected": -1.6778247356414795, "logps/chosen": -231.51792907714844, "logps/ref_chosen": -102.5750503540039, "logps/ref_rejected": -108.81768798828125, "logps/rejected": -323.68902587890625, "loss": 0.7933, "margin_dpo/margin_mean": 85.928466796875, "margin_dpo/margin_std": 82.60166931152344, "step": 668 }, { "epoch": 0.9823788546255506, "fcm_dpo/beta": 0.011247916147112846, "fcm_dpo/delta": -0.08348070085048676, "fcm_dpo/margin": 82.30418395996094, "fcm_dpo/q_t": 0.3088477849960327, "grad_norm": 12.096893310546875, "learning_rate": 5.564580657695939e-10, "logits/chosen": -1.8613781929016113, "logits/rejected": -1.4636726379394531, "logps/chosen": -207.69525146484375, "logps/ref_chosen": -89.49478149414062, "logps/ref_rejected": -82.51950073242188, "logps/rejected": -283.024169921875, "loss": 0.8056, "margin_dpo/margin_mean": 82.30418395996094, "margin_dpo/margin_std": 74.32376098632812, "step": 669 }, { "epoch": 0.9838472834067548, "fcm_dpo/beta": 0.010946569964289665, "fcm_dpo/delta": -0.1527199000120163, "fcm_dpo/margin": 90.21934509277344, "fcm_dpo/q_t": 0.2935888469219208, "grad_norm": 11.206657409667969, "learning_rate": 4.741678157389739e-10, "logits/chosen": -1.7874915599822998, "logits/rejected": -1.2579689025878906, "logps/chosen": -214.5462188720703, "logps/ref_chosen": -95.45459747314453, "logps/ref_rejected": -101.53292846679688, "logps/rejected": -310.8439025878906, "loss": 0.7782, "margin_dpo/margin_mean": 90.21934509277344, "margin_dpo/margin_std": 78.12318420410156, "step": 670 }, { "epoch": 0.9853157121879589, "fcm_dpo/beta": 0.011242946609854698, "fcm_dpo/delta": 0.16490933299064636, "fcm_dpo/margin": 61.64239501953125, "fcm_dpo/q_t": 0.35765212774276733, "grad_norm": 13.497610092163086, "learning_rate": 3.9845280344705245e-10, "logits/chosen": -1.8117475509643555, "logits/rejected": -1.5957112312316895, "logps/chosen": -226.856689453125, "logps/ref_chosen": -82.12312316894531, "logps/ref_rejected": -90.21969604492188, "logps/rejected": -296.59564208984375, "loss": 0.9702, "margin_dpo/margin_mean": 61.64239501953125, "margin_dpo/margin_std": 76.5667495727539, "step": 671 }, { "epoch": 0.986784140969163, "fcm_dpo/beta": 0.011585546657443047, "fcm_dpo/delta": 0.20920199155807495, "fcm_dpo/margin": 56.48442459106445, "fcm_dpo/q_t": 0.36586183309555054, "grad_norm": 13.58948040008545, "learning_rate": 3.293150240547549e-10, "logits/chosen": -1.972328543663025, "logits/rejected": -1.6273032426834106, "logps/chosen": -232.64218139648438, "logps/ref_chosen": -90.0619125366211, "logps/ref_rejected": -100.45323181152344, "logps/rejected": -299.5179443359375, "loss": 0.9928, "margin_dpo/margin_mean": 56.48442840576172, "margin_dpo/margin_std": 72.01486206054688, "step": 672 }, { "epoch": 0.9882525697503671, "fcm_dpo/beta": 0.011844823136925697, "fcm_dpo/delta": 0.054356593638658524, "fcm_dpo/margin": 67.50054168701172, "fcm_dpo/q_t": 0.33461272716522217, "grad_norm": 13.27198314666748, "learning_rate": 2.6675629940689504e-10, "logits/chosen": -1.9633225202560425, "logits/rejected": -1.5942094326019287, "logps/chosen": -204.55674743652344, "logps/ref_chosen": -79.26315307617188, "logps/ref_rejected": -91.34925079345703, "logps/rejected": -284.14337158203125, "loss": 0.8878, "margin_dpo/margin_mean": 67.50054931640625, "margin_dpo/margin_std": 71.98117065429688, "step": 673 }, { "epoch": 0.9897209985315712, "fcm_dpo/beta": 0.011596374213695526, "fcm_dpo/delta": -0.12112964689731598, "fcm_dpo/margin": 82.70986938476562, "fcm_dpo/q_t": 0.3062339425086975, "grad_norm": 13.10261344909668, "learning_rate": 2.1077827798404725e-10, "logits/chosen": -1.6151472330093384, "logits/rejected": -1.2215874195098877, "logps/chosen": -197.07003784179688, "logps/ref_chosen": -75.45831298828125, "logps/ref_rejected": -76.20362854003906, "logps/rejected": -280.5252380371094, "loss": 0.7916, "margin_dpo/margin_mean": 82.70986938476562, "margin_dpo/margin_std": 76.59530639648438, "step": 674 }, { "epoch": 0.9911894273127754, "fcm_dpo/beta": 0.011463642120361328, "fcm_dpo/delta": 0.01827855408191681, "fcm_dpo/margin": 72.52214050292969, "fcm_dpo/q_t": 0.33195608854293823, "grad_norm": 14.91750431060791, "learning_rate": 1.6138243485910863e-10, "logits/chosen": -1.5982558727264404, "logits/rejected": -1.2730377912521362, "logps/chosen": -214.935791015625, "logps/ref_chosen": -79.90953063964844, "logps/ref_rejected": -81.21824645996094, "logps/rejected": -288.76666259765625, "loss": 0.9378, "margin_dpo/margin_mean": 72.52214050292969, "margin_dpo/margin_std": 88.32064819335938, "step": 675 }, { "epoch": 0.9926578560939795, "fcm_dpo/beta": 0.011480691842734814, "fcm_dpo/delta": -0.03261286020278931, "fcm_dpo/margin": 76.57514953613281, "fcm_dpo/q_t": 0.31988048553466797, "grad_norm": 13.69805908203125, "learning_rate": 1.1857007165852472e-10, "logits/chosen": -2.108896493911743, "logits/rejected": -1.548959732055664, "logps/chosen": -237.62142944335938, "logps/ref_chosen": -98.17111206054688, "logps/ref_rejected": -95.024658203125, "logps/rejected": -311.05010986328125, "loss": 0.8634, "margin_dpo/margin_mean": 76.57515716552734, "margin_dpo/margin_std": 78.37229919433594, "step": 676 }, { "epoch": 0.9941262848751835, "fcm_dpo/beta": 0.011398270726203918, "fcm_dpo/delta": -0.053326986730098724, "fcm_dpo/margin": 78.80928039550781, "fcm_dpo/q_t": 0.311850368976593, "grad_norm": 11.52205753326416, "learning_rate": 8.23423165278725e-11, "logits/chosen": -1.9365110397338867, "logits/rejected": -1.2789349555969238, "logps/chosen": -218.03697204589844, "logps/ref_chosen": -91.37928771972656, "logps/ref_rejected": -82.87776947021484, "logps/rejected": -288.3447265625, "loss": 0.802, "margin_dpo/margin_mean": 78.80928039550781, "margin_dpo/margin_std": 67.46434020996094, "step": 677 }, { "epoch": 0.9955947136563876, "fcm_dpo/beta": 0.011227824725210667, "fcm_dpo/delta": -0.1735784113407135, "fcm_dpo/margin": 89.56806945800781, "fcm_dpo/q_t": 0.2927679419517517, "grad_norm": 11.666086196899414, "learning_rate": 5.270012410216185e-11, "logits/chosen": -1.7108724117279053, "logits/rejected": -1.3849983215332031, "logps/chosen": -200.70018005371094, "logps/ref_chosen": -75.6458740234375, "logps/ref_rejected": -86.96611022949219, "logps/rejected": -301.5885009765625, "loss": 0.7627, "margin_dpo/margin_mean": 89.56806945800781, "margin_dpo/margin_std": 73.57870483398438, "step": 678 }, { "epoch": 0.9970631424375918, "fcm_dpo/beta": 0.011059444397687912, "fcm_dpo/delta": 0.03496631607413292, "fcm_dpo/margin": 73.9551773071289, "fcm_dpo/q_t": 0.3249191641807556, "grad_norm": 11.829185485839844, "learning_rate": 2.9644275480772416e-11, "logits/chosen": -2.3330090045928955, "logits/rejected": -1.82859206199646, "logps/chosen": -217.5744171142578, "logps/ref_chosen": -80.77344512939453, "logps/ref_rejected": -82.87850189208984, "logps/rejected": -293.6346435546875, "loss": 0.8263, "margin_dpo/margin_mean": 73.95518493652344, "margin_dpo/margin_std": 62.57131576538086, "step": 679 }, { "epoch": 0.9985315712187959, "fcm_dpo/beta": 0.011266154237091541, "fcm_dpo/delta": 0.01519148051738739, "fcm_dpo/margin": 73.86129760742188, "fcm_dpo/q_t": 0.3306913375854492, "grad_norm": 11.578691482543945, "learning_rate": 1.31753782067201e-11, "logits/chosen": -1.9968955516815186, "logits/rejected": -1.688669204711914, "logps/chosen": -249.28564453125, "logps/ref_chosen": -107.68292999267578, "logps/ref_rejected": -116.0948715209961, "logps/rejected": -331.5588684082031, "loss": 0.889, "margin_dpo/margin_mean": 73.86129760742188, "margin_dpo/margin_std": 76.965576171875, "step": 680 }, { "epoch": 1.0, "fcm_dpo/beta": 0.011270027607679367, "fcm_dpo/delta": 0.07696523517370224, "fcm_dpo/margin": 69.10145568847656, "fcm_dpo/q_t": 0.3385009765625, "grad_norm": 12.040020942687988, "learning_rate": 3.2938662507808745e-12, "logits/chosen": -2.3690905570983887, "logits/rejected": -2.0767529010772705, "logps/chosen": -219.39236450195312, "logps/ref_chosen": -93.01106262207031, "logps/ref_rejected": -94.8221664428711, "logps/rejected": -290.304931640625, "loss": 0.9025, "margin_dpo/margin_mean": 69.10145568847656, "margin_dpo/margin_std": 75.53838348388672, "step": 681 }, { "epoch": 1.0, "step": 681, "total_flos": 0.0, "train_loss": 0.9292974616462438, "train_runtime": 2239.8819, "train_samples_per_second": 19.464, "train_steps_per_second": 0.304 } ], "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 }