{ "best_global_step": null, "best_metric": null, "best_model_checkpoint": null, "epoch": 1.0, "eval_steps": 200, "global_step": 681, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "KL/chosen_KL_mean": 0.00527191162109375, "KL/mean": 0.016706019639968872, "KL/rejected_KL_mean": 0.028141021728515625, "KL/std": 0.272699236869812, "epoch": 0.0014684287812041115, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.02287006378173828, "fcm_dpo/q_t": 0.5005706548690796, "grad_norm": 83.52790069580078, "learning_rate": 0.0, "logits/chosen": -0.4974287748336792, "logits/rejected": -0.43299180269241333, "logps/chosen": -50.1435661315918, "logps/ref_chosen": -50.14883804321289, "logps/ref_rejected": -74.1280517578125, "logps/rejected": -74.09991455078125, "loss": 1.389, "margin_dpo/margin_mean": -0.02287048101425171, "margin_dpo/margin_std": 0.41920793056488037, "step": 1 }, { "KL/chosen_KL_mean": -0.03498649597167969, "KL/mean": -0.00212840735912323, "KL/rejected_KL_mean": 0.030735015869140625, "KL/std": 0.24797174334526062, "epoch": 0.002936857562408223, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.06572261452674866, "fcm_dpo/q_t": 0.5016425848007202, "grad_norm": 72.21088409423828, "learning_rate": 7.246376811594203e-09, "logits/chosen": -0.49536412954330444, "logits/rejected": -0.4594460427761078, "logps/chosen": -52.65568923950195, "logps/ref_chosen": -52.620704650878906, "logps/ref_rejected": -75.30413818359375, "logps/rejected": -75.27340698242188, "loss": 1.3932, "margin_dpo/margin_mean": -0.06572240591049194, "margin_dpo/margin_std": 0.35048407316207886, "step": 2 }, { "KL/chosen_KL_mean": 0.017366409301757812, "KL/mean": 0.014461815357208252, "KL/rejected_KL_mean": 0.01155853271484375, "KL/std": 0.2480686902999878, "epoch": 0.004405286343612335, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.005803465843200684, "fcm_dpo/q_t": 0.499855101108551, "grad_norm": 70.91714477539062, "learning_rate": 1.4492753623188406e-08, "logits/chosen": -0.4816860556602478, "logits/rejected": -0.4422453045845032, "logps/chosen": -60.964229583740234, "logps/ref_chosen": -60.981597900390625, "logps/ref_rejected": -68.67259216308594, "logps/rejected": -68.66104125976562, "loss": 1.386, "margin_dpo/margin_mean": 0.0058036744594573975, "margin_dpo/margin_std": 0.34888899326324463, "step": 3 }, { "KL/chosen_KL_mean": -0.053741455078125, "KL/mean": 0.010254994034767151, "KL/rejected_KL_mean": 0.074249267578125, "KL/std": 0.3106370270252228, "epoch": 0.005873715124816446, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.12799236178398132, "fcm_dpo/q_t": 0.5031979084014893, "grad_norm": 72.49424743652344, "learning_rate": 2.1739130434782606e-08, "logits/chosen": -0.46865934133529663, "logits/rejected": -0.4410114884376526, "logps/chosen": -56.82145309448242, "logps/ref_chosen": -56.7677116394043, "logps/ref_rejected": -86.64710998535156, "logps/rejected": -86.57286071777344, "loss": 1.3996, "margin_dpo/margin_mean": -0.12799254059791565, "margin_dpo/margin_std": 0.42509347200393677, "step": 4 }, { "KL/chosen_KL_mean": 0.016374588012695312, "KL/mean": 0.007320374250411987, "KL/rejected_KL_mean": -0.00173187255859375, "KL/std": 0.25842905044555664, "epoch": 0.007342143906020558, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.018107712268829346, "fcm_dpo/q_t": 0.4995471239089966, "grad_norm": 89.32853698730469, "learning_rate": 2.898550724637681e-08, "logits/chosen": -0.5146687030792236, "logits/rejected": -0.4709242582321167, "logps/chosen": -53.84300231933594, "logps/ref_chosen": -53.859375, "logps/ref_rejected": -84.14918518066406, "logps/rejected": -84.15091705322266, "loss": 1.3848, "margin_dpo/margin_mean": 0.01810815930366516, "margin_dpo/margin_std": 0.3279362916946411, "step": 5 }, { "KL/chosen_KL_mean": -0.033203125, "KL/mean": -0.029030516743659973, "KL/rejected_KL_mean": -0.0248565673828125, "KL/std": 0.2704921364784241, "epoch": 0.00881057268722467, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.008346349000930786, "fcm_dpo/q_t": 0.5002087354660034, "grad_norm": 92.05367279052734, "learning_rate": 3.6231884057971014e-08, "logits/chosen": -0.5008316040039062, "logits/rejected": -0.45828235149383545, "logps/chosen": -63.040687561035156, "logps/ref_chosen": -63.007484436035156, "logps/ref_rejected": -92.64534759521484, "logps/rejected": -92.67020416259766, "loss": 1.3876, "margin_dpo/margin_mean": -0.008346527814865112, "margin_dpo/margin_std": 0.41183364391326904, "step": 6 }, { "KL/chosen_KL_mean": -0.0074214935302734375, "KL/mean": 0.024099931120872498, "KL/rejected_KL_mean": 0.055622100830078125, "KL/std": 0.3060908317565918, "epoch": 0.010279001468428781, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.06303969025611877, "fcm_dpo/q_t": 0.5015759468078613, "grad_norm": 82.6351318359375, "learning_rate": 4.347826086956521e-08, "logits/chosen": -0.4966989755630493, "logits/rejected": -0.4628666639328003, "logps/chosen": -57.78224182128906, "logps/ref_chosen": -57.774818420410156, "logps/ref_rejected": -103.92059326171875, "logps/rejected": -103.86497497558594, "loss": 1.3931, "margin_dpo/margin_mean": -0.06303924322128296, "margin_dpo/margin_std": 0.4346626400947571, "step": 7 }, { "KL/chosen_KL_mean": 0.03885078430175781, "KL/mean": 0.04787926375865936, "KL/rejected_KL_mean": 0.056911468505859375, "KL/std": 0.3092794120311737, "epoch": 0.011747430249632892, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.018061310052871704, "fcm_dpo/q_t": 0.5004498958587646, "grad_norm": 78.8070068359375, "learning_rate": 5.0724637681159424e-08, "logits/chosen": -0.4943142533302307, "logits/rejected": -0.466851145029068, "logps/chosen": -58.67718505859375, "logps/ref_chosen": -58.716033935546875, "logps/ref_rejected": -79.3114242553711, "logps/rejected": -79.2545166015625, "loss": 1.3885, "margin_dpo/margin_mean": -0.01806098222732544, "margin_dpo/margin_std": 0.4174906015396118, "step": 8 }, { "KL/chosen_KL_mean": 0.028049468994140625, "KL/mean": 0.013820052146911621, "KL/rejected_KL_mean": -0.000408172607421875, "KL/std": 0.29441195726394653, "epoch": 0.013215859030837005, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.028452813625335693, "fcm_dpo/q_t": 0.49928951263427734, "grad_norm": 85.16717529296875, "learning_rate": 5.797101449275362e-08, "logits/chosen": -0.5045357346534729, "logits/rejected": -0.4584883451461792, "logps/chosen": -69.83879089355469, "logps/ref_chosen": -69.8668441772461, "logps/ref_rejected": -99.6026611328125, "logps/rejected": -99.60307312011719, "loss": 1.384, "margin_dpo/margin_mean": 0.02845209836959839, "margin_dpo/margin_std": 0.4499683976173401, "step": 9 }, { "KL/chosen_KL_mean": 0.000568389892578125, "KL/mean": -0.016893252730369568, "KL/rejected_KL_mean": -0.034351348876953125, "KL/std": 0.25498026609420776, "epoch": 0.014684287812041116, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.034921616315841675, "fcm_dpo/q_t": 0.4991276264190674, "grad_norm": 70.64812469482422, "learning_rate": 6.521739130434782e-08, "logits/chosen": -0.48759734630584717, "logits/rejected": -0.44404521584510803, "logps/chosen": -48.35711669921875, "logps/ref_chosen": -48.35768508911133, "logps/ref_rejected": -80.37206268310547, "logps/rejected": -80.40641784667969, "loss": 1.3832, "margin_dpo/margin_mean": 0.034921616315841675, "margin_dpo/margin_std": 0.374549925327301, "step": 10 }, { "KL/chosen_KL_mean": 0.018220901489257812, "KL/mean": 0.0023308396339416504, "KL/rejected_KL_mean": -0.01355743408203125, "KL/std": 0.2806966304779053, "epoch": 0.016152716593245228, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.031774938106536865, "fcm_dpo/q_t": 0.49920594692230225, "grad_norm": 68.4999008178711, "learning_rate": 7.246376811594203e-08, "logits/chosen": -0.4650164842605591, "logits/rejected": -0.4402990937232971, "logps/chosen": -52.998634338378906, "logps/ref_chosen": -53.01685333251953, "logps/ref_rejected": -87.78038024902344, "logps/rejected": -87.79393005371094, "loss": 1.3834, "margin_dpo/margin_mean": 0.03177538514137268, "margin_dpo/margin_std": 0.34445643424987793, "step": 11 }, { "KL/chosen_KL_mean": 0.030973434448242188, "KL/mean": -0.03026643395423889, "KL/rejected_KL_mean": -0.09149932861328125, "KL/std": 0.2564256191253662, "epoch": 0.01762114537444934, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.12247461080551147, "fcm_dpo/q_t": 0.49693846702575684, "grad_norm": 94.00206756591797, "learning_rate": 7.971014492753623e-08, "logits/chosen": -0.5384472608566284, "logits/rejected": -0.5022361278533936, "logps/chosen": -61.77446365356445, "logps/ref_chosen": -61.80543518066406, "logps/ref_rejected": -104.8582763671875, "logps/rejected": -104.94976806640625, "loss": 1.3744, "margin_dpo/margin_mean": 0.12247374653816223, "margin_dpo/margin_std": 0.3362644910812378, "step": 12 }, { "KL/chosen_KL_mean": -0.03291130065917969, "KL/mean": -0.008768260478973389, "KL/rejected_KL_mean": 0.015377044677734375, "KL/std": 0.2577245235443115, "epoch": 0.01908957415565345, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.04828256368637085, "fcm_dpo/q_t": 0.50120609998703, "grad_norm": 79.38003540039062, "learning_rate": 8.695652173913042e-08, "logits/chosen": -0.4910910725593567, "logits/rejected": -0.4638311266899109, "logps/chosen": -64.29326629638672, "logps/ref_chosen": -64.2603530883789, "logps/ref_rejected": -87.20307922363281, "logps/rejected": -87.18770599365234, "loss": 1.3915, "margin_dpo/margin_mean": -0.04828298091888428, "margin_dpo/margin_std": 0.3570724129676819, "step": 13 }, { "KL/chosen_KL_mean": -0.013036727905273438, "KL/mean": 0.009792491793632507, "KL/rejected_KL_mean": 0.032619476318359375, "KL/std": 0.2999943494796753, "epoch": 0.020558002936857563, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.04565081000328064, "fcm_dpo/q_t": 0.5011403560638428, "grad_norm": 85.56097412109375, "learning_rate": 9.420289855072464e-08, "logits/chosen": -0.4799395501613617, "logits/rejected": -0.440725177526474, "logps/chosen": -58.12324523925781, "logps/ref_chosen": -58.11021041870117, "logps/ref_rejected": -104.04708099365234, "logps/rejected": -104.01446533203125, "loss": 1.3913, "margin_dpo/margin_mean": -0.045650362968444824, "margin_dpo/margin_std": 0.404807984828949, "step": 14 }, { "KL/chosen_KL_mean": -0.06286811828613281, "KL/mean": -0.05339600145816803, "KL/rejected_KL_mean": -0.0439300537109375, "KL/std": 0.28460076451301575, "epoch": 0.022026431718061675, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.01893576979637146, "fcm_dpo/q_t": 0.5004709959030151, "grad_norm": 64.50666809082031, "learning_rate": 1.0144927536231885e-07, "logits/chosen": -0.5097112655639648, "logits/rejected": -0.49188485741615295, "logps/chosen": -57.02977752685547, "logps/ref_chosen": -56.96691131591797, "logps/ref_rejected": -80.80863952636719, "logps/rejected": -80.85256958007812, "loss": 1.3887, "margin_dpo/margin_mean": -0.0189359188079834, "margin_dpo/margin_std": 0.4248620867729187, "step": 15 }, { "KL/chosen_KL_mean": -0.011941909790039062, "KL/mean": -0.038836658000946045, "KL/rejected_KL_mean": -0.06572723388671875, "KL/std": 0.2758028507232666, "epoch": 0.023494860499265784, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.0537833571434021, "fcm_dpo/q_t": 0.49865537881851196, "grad_norm": 83.70243072509766, "learning_rate": 1.0869565217391303e-07, "logits/chosen": -0.5452868342399597, "logits/rejected": -0.5062515735626221, "logps/chosen": -61.751834869384766, "logps/ref_chosen": -61.739891052246094, "logps/ref_rejected": -84.36947631835938, "logps/rejected": -84.43519592285156, "loss": 1.3813, "margin_dpo/margin_mean": 0.053783416748046875, "margin_dpo/margin_std": 0.3853508234024048, "step": 16 }, { "KL/chosen_KL_mean": -0.00653839111328125, "KL/mean": -0.028867706656455994, "KL/rejected_KL_mean": -0.05120086669921875, "KL/std": 0.2653309106826782, "epoch": 0.024963289280469897, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.04466333985328674, "fcm_dpo/q_t": 0.4988834261894226, "grad_norm": 79.01133728027344, "learning_rate": 1.1594202898550725e-07, "logits/chosen": -0.488353431224823, "logits/rejected": -0.4496167302131653, "logps/chosen": -67.71687316894531, "logps/ref_chosen": -67.71033477783203, "logps/ref_rejected": -85.37865447998047, "logps/rejected": -85.42985534667969, "loss": 1.3822, "margin_dpo/margin_mean": 0.04466351866722107, "margin_dpo/margin_std": 0.36588189005851746, "step": 17 }, { "KL/chosen_KL_mean": 0.013227462768554688, "KL/mean": -0.008445173501968384, "KL/rejected_KL_mean": -0.030117034912109375, "KL/std": 0.26379337906837463, "epoch": 0.02643171806167401, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.04334592819213867, "fcm_dpo/q_t": 0.49891626834869385, "grad_norm": 82.08016967773438, "learning_rate": 1.2318840579710146e-07, "logits/chosen": -0.5007514357566833, "logits/rejected": -0.44611847400665283, "logps/chosen": -47.72626495361328, "logps/ref_chosen": -47.7394905090332, "logps/ref_rejected": -75.4722900390625, "logps/rejected": -75.50240325927734, "loss": 1.3823, "margin_dpo/margin_mean": 0.04334595799446106, "margin_dpo/margin_std": 0.3786376118659973, "step": 18 }, { "KL/chosen_KL_mean": 0.01825714111328125, "KL/mean": -0.03181253373622894, "KL/rejected_KL_mean": -0.081878662109375, "KL/std": 0.24455955624580383, "epoch": 0.027900146842878122, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.1001291573047638, "fcm_dpo/q_t": 0.49749767780303955, "grad_norm": 73.14253997802734, "learning_rate": 1.3043478260869563e-07, "logits/chosen": -0.4779819846153259, "logits/rejected": -0.42704999446868896, "logps/chosen": -70.18710327148438, "logps/ref_chosen": -70.20536041259766, "logps/ref_rejected": -89.7575912475586, "logps/rejected": -89.83946990966797, "loss": 1.3766, "margin_dpo/margin_mean": 0.10012984275817871, "margin_dpo/margin_std": 0.33631211519241333, "step": 19 }, { "KL/chosen_KL_mean": -0.012134552001953125, "KL/mean": -0.04402177035808563, "KL/rejected_KL_mean": -0.07590484619140625, "KL/std": 0.22419436275959015, "epoch": 0.02936857562408223, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.06377437710762024, "fcm_dpo/q_t": 0.4984060525894165, "grad_norm": 73.54801940917969, "learning_rate": 1.3768115942028986e-07, "logits/chosen": -0.5527380108833313, "logits/rejected": -0.49656909704208374, "logps/chosen": -50.81537628173828, "logps/ref_chosen": -50.80324172973633, "logps/ref_rejected": -78.82334899902344, "logps/rejected": -78.89925384521484, "loss": 1.3802, "margin_dpo/margin_mean": 0.06377455592155457, "margin_dpo/margin_std": 0.3116491734981537, "step": 20 }, { "KL/chosen_KL_mean": 0.03291130065917969, "KL/mean": -0.04309450834989548, "KL/rejected_KL_mean": -0.11909866333007812, "KL/std": 0.277535617351532, "epoch": 0.030837004405286344, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.1520112305879593, "fcm_dpo/q_t": 0.49620136618614197, "grad_norm": 77.38446044921875, "learning_rate": 1.4492753623188405e-07, "logits/chosen": -0.5033525824546814, "logits/rejected": -0.48031163215637207, "logps/chosen": -50.03010559082031, "logps/ref_chosen": -50.063018798828125, "logps/ref_rejected": -77.86878967285156, "logps/rejected": -77.9878921508789, "loss": 1.3714, "margin_dpo/margin_mean": 0.1520119160413742, "margin_dpo/margin_std": 0.33785581588745117, "step": 21 }, { "KL/chosen_KL_mean": 0.049625396728515625, "KL/mean": -0.0646744817495346, "KL/rejected_KL_mean": -0.17897415161132812, "KL/std": 0.2905516028404236, "epoch": 0.032305433186490456, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.2286020815372467, "fcm_dpo/q_t": 0.4942886233329773, "grad_norm": 84.13846588134766, "learning_rate": 1.5217391304347825e-07, "logits/chosen": -0.47962144017219543, "logits/rejected": -0.43577441573143005, "logps/chosen": -59.00801086425781, "logps/ref_chosen": -59.05763626098633, "logps/ref_rejected": -97.50466918945312, "logps/rejected": -97.68364715576172, "loss": 1.364, "margin_dpo/margin_mean": 0.22860169410705566, "margin_dpo/margin_std": 0.4178020656108856, "step": 22 }, { "KL/chosen_KL_mean": 0.03469657897949219, "KL/mean": -0.05604839324951172, "KL/rejected_KL_mean": -0.14679336547851562, "KL/std": 0.29135066270828247, "epoch": 0.033773861967694566, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.1814877986907959, "fcm_dpo/q_t": 0.4954661726951599, "grad_norm": 80.80692291259766, "learning_rate": 1.5942028985507245e-07, "logits/chosen": -0.5033831000328064, "logits/rejected": -0.48167192935943604, "logps/chosen": -60.042999267578125, "logps/ref_chosen": -60.07769775390625, "logps/ref_rejected": -81.13955688476562, "logps/rejected": -81.28634643554688, "loss": 1.3686, "margin_dpo/margin_mean": 0.18148785829544067, "margin_dpo/margin_std": 0.37989723682403564, "step": 23 }, { "KL/chosen_KL_mean": 0.027399063110351562, "KL/mean": -0.1130581721663475, "KL/rejected_KL_mean": -0.25351715087890625, "KL/std": 0.2865855097770691, "epoch": 0.03524229074889868, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.2809140086174011, "fcm_dpo/q_t": 0.4929804801940918, "grad_norm": 84.92632293701172, "learning_rate": 1.6666666666666665e-07, "logits/chosen": -0.5087974071502686, "logits/rejected": -0.49225664138793945, "logps/chosen": -44.263633728027344, "logps/ref_chosen": -44.29103469848633, "logps/ref_rejected": -99.12521362304688, "logps/rejected": -99.37873077392578, "loss": 1.3587, "margin_dpo/margin_mean": 0.28091397881507874, "margin_dpo/margin_std": 0.367694616317749, "step": 24 }, { "KL/chosen_KL_mean": -0.0003814697265625, "KL/mean": -0.09033405780792236, "KL/rejected_KL_mean": -0.1802825927734375, "KL/std": 0.3605647385120392, "epoch": 0.03671071953010279, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.17990106344223022, "fcm_dpo/q_t": 0.4955053925514221, "grad_norm": 74.13409423828125, "learning_rate": 1.7391304347826085e-07, "logits/chosen": -0.5078995823860168, "logits/rejected": -0.47833961248397827, "logps/chosen": -52.53743362426758, "logps/ref_chosen": -52.537052154541016, "logps/ref_rejected": -89.34219360351562, "logps/rejected": -89.52247619628906, "loss": 1.3689, "margin_dpo/margin_mean": 0.17990124225616455, "margin_dpo/margin_std": 0.45145928859710693, "step": 25 }, { "KL/chosen_KL_mean": 0.06395149230957031, "KL/mean": -0.13807255029678345, "KL/rejected_KL_mean": -0.3400993347167969, "KL/std": 0.4370245337486267, "epoch": 0.0381791483113069, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.40405547618865967, "fcm_dpo/q_t": 0.4899117946624756, "grad_norm": 86.72003173828125, "learning_rate": 1.8115942028985507e-07, "logits/chosen": -0.5410639047622681, "logits/rejected": -0.5095586776733398, "logps/chosen": -53.858856201171875, "logps/ref_chosen": -53.92280578613281, "logps/ref_rejected": -103.35971069335938, "logps/rejected": -103.69981384277344, "loss": 1.3471, "margin_dpo/margin_mean": 0.4040555953979492, "margin_dpo/margin_std": 0.5661390423774719, "step": 26 }, { "KL/chosen_KL_mean": 0.1539630889892578, "KL/mean": -0.13156871497631073, "KL/rejected_KL_mean": -0.4171028137207031, "KL/std": 0.48169609904289246, "epoch": 0.039647577092511016, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.571062445640564, "fcm_dpo/q_t": 0.48574382066726685, "grad_norm": 93.52059936523438, "learning_rate": 1.8840579710144927e-07, "logits/chosen": -0.526579737663269, "logits/rejected": -0.49027207493782043, "logps/chosen": -42.74456787109375, "logps/ref_chosen": -42.898529052734375, "logps/ref_rejected": -98.72419738769531, "logps/rejected": -99.14130401611328, "loss": 1.3308, "margin_dpo/margin_mean": 0.5710620880126953, "margin_dpo/margin_std": 0.5639763474464417, "step": 27 }, { "KL/chosen_KL_mean": -0.003818511962890625, "KL/mean": -0.1476629376411438, "KL/rejected_KL_mean": -0.29151153564453125, "KL/std": 0.4086608588695526, "epoch": 0.041116005873715125, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.28768980503082275, "fcm_dpo/q_t": 0.4928103983402252, "grad_norm": 77.30921936035156, "learning_rate": 1.9565217391304347e-07, "logits/chosen": -0.5286385416984558, "logits/rejected": -0.4755781292915344, "logps/chosen": -60.56031799316406, "logps/ref_chosen": -60.55650329589844, "logps/ref_rejected": -91.40111541748047, "logps/rejected": -91.692626953125, "loss": 1.3587, "margin_dpo/margin_mean": 0.287690132856369, "margin_dpo/margin_std": 0.6202329993247986, "step": 28 }, { "KL/chosen_KL_mean": 0.11627197265625, "KL/mean": -0.15424633026123047, "KL/rejected_KL_mean": -0.4247627258300781, "KL/std": 0.48740673065185547, "epoch": 0.042584434654919234, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.5410360097885132, "fcm_dpo/q_t": 0.4864945113658905, "grad_norm": 91.11680603027344, "learning_rate": 2.028985507246377e-07, "logits/chosen": -0.5516057014465332, "logits/rejected": -0.5051902532577515, "logps/chosen": -57.69151306152344, "logps/ref_chosen": -57.80778503417969, "logps/ref_rejected": -97.39434814453125, "logps/rejected": -97.8191146850586, "loss": 1.3339, "margin_dpo/margin_mean": 0.5410366654396057, "margin_dpo/margin_std": 0.6225091218948364, "step": 29 }, { "KL/chosen_KL_mean": 0.14649200439453125, "KL/mean": -0.21738092601299286, "KL/rejected_KL_mean": -0.5812568664550781, "KL/std": 0.6298277378082275, "epoch": 0.04405286343612335, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.7277464866638184, "fcm_dpo/q_t": 0.48183703422546387, "grad_norm": 87.24919128417969, "learning_rate": 2.1014492753623187e-07, "logits/chosen": -0.5027915835380554, "logits/rejected": -0.47236695885658264, "logps/chosen": -52.430877685546875, "logps/ref_chosen": -52.577369689941406, "logps/ref_rejected": -98.48920440673828, "logps/rejected": -99.07046508789062, "loss": 1.3159, "margin_dpo/margin_mean": 0.7277467846870422, "margin_dpo/margin_std": 0.6605062484741211, "step": 30 }, { "KL/chosen_KL_mean": 0.08755111694335938, "KL/mean": -0.1343349814414978, "KL/rejected_KL_mean": -0.35622406005859375, "KL/std": 0.5193805694580078, "epoch": 0.04552129221732746, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.4437781572341919, "fcm_dpo/q_t": 0.4889264404773712, "grad_norm": 68.48614501953125, "learning_rate": 2.1739130434782607e-07, "logits/chosen": -0.5168442726135254, "logits/rejected": -0.4734218716621399, "logps/chosen": -63.71936798095703, "logps/ref_chosen": -63.806922912597656, "logps/ref_rejected": -72.89400482177734, "logps/rejected": -73.25022888183594, "loss": 1.3435, "margin_dpo/margin_mean": 0.44377774000167847, "margin_dpo/margin_std": 0.6583547592163086, "step": 31 }, { "KL/chosen_KL_mean": 0.18407058715820312, "KL/mean": -0.19383396208286285, "KL/rejected_KL_mean": -0.5717430114746094, "KL/std": 0.7323002815246582, "epoch": 0.04698972099853157, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.755810022354126, "fcm_dpo/q_t": 0.4811716079711914, "grad_norm": 83.46903991699219, "learning_rate": 2.2463768115942027e-07, "logits/chosen": -0.5339919328689575, "logits/rejected": -0.49401533603668213, "logps/chosen": -62.55545425415039, "logps/ref_chosen": -62.739524841308594, "logps/ref_rejected": -89.3175048828125, "logps/rejected": -89.88924407958984, "loss": 1.3141, "margin_dpo/margin_mean": 0.7558099627494812, "margin_dpo/margin_std": 0.8941420316696167, "step": 32 }, { "KL/chosen_KL_mean": 0.10718154907226562, "KL/mean": -0.2103760838508606, "KL/rejected_KL_mean": -0.5279350280761719, "KL/std": 0.5602357387542725, "epoch": 0.048458149779735685, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.6351122260093689, "fcm_dpo/q_t": 0.48414498567581177, "grad_norm": 72.1122817993164, "learning_rate": 2.318840579710145e-07, "logits/chosen": -0.489587664604187, "logits/rejected": -0.46343374252319336, "logps/chosen": -53.15379333496094, "logps/ref_chosen": -53.26097106933594, "logps/ref_rejected": -87.8851318359375, "logps/rejected": -88.41307067871094, "loss": 1.3248, "margin_dpo/margin_mean": 0.635112464427948, "margin_dpo/margin_std": 0.6338217854499817, "step": 33 }, { "KL/chosen_KL_mean": 0.08233070373535156, "KL/mean": -0.30296576023101807, "KL/rejected_KL_mean": -0.6882667541503906, "KL/std": 0.7120438814163208, "epoch": 0.049926578560939794, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.7705912590026855, "fcm_dpo/q_t": 0.4807863235473633, "grad_norm": 78.05076599121094, "learning_rate": 2.391304347826087e-07, "logits/chosen": -0.5040176510810852, "logits/rejected": -0.4865629971027374, "logps/chosen": -50.7349967956543, "logps/ref_chosen": -50.81732940673828, "logps/ref_rejected": -101.92184448242188, "logps/rejected": -102.610107421875, "loss": 1.3124, "margin_dpo/margin_mean": 0.7705909013748169, "margin_dpo/margin_std": 0.8183022141456604, "step": 34 }, { "KL/chosen_KL_mean": 0.13801002502441406, "KL/mean": -0.4717918634414673, "KL/rejected_KL_mean": -1.0815925598144531, "KL/std": 1.019068956375122, "epoch": 0.0513950073421439, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.2196067571640015, "fcm_dpo/q_t": 0.46972382068634033, "grad_norm": 82.06009674072266, "learning_rate": 2.463768115942029e-07, "logits/chosen": -0.5100936889648438, "logits/rejected": -0.4733096957206726, "logps/chosen": -50.88648223876953, "logps/ref_chosen": -51.02449035644531, "logps/ref_rejected": -106.82443237304688, "logps/rejected": -107.90602111816406, "loss": 1.2714, "margin_dpo/margin_mean": 1.219606637954712, "margin_dpo/margin_std": 1.1478991508483887, "step": 35 }, { "KL/chosen_KL_mean": 0.07729721069335938, "KL/mean": -0.5002259016036987, "KL/rejected_KL_mean": -1.0777473449707031, "KL/std": 1.079554796218872, "epoch": 0.05286343612334802, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.1550471782684326, "fcm_dpo/q_t": 0.47128745913505554, "grad_norm": 73.2303237915039, "learning_rate": 2.536231884057971e-07, "logits/chosen": -0.5755394697189331, "logits/rejected": -0.5400795936584473, "logps/chosen": -51.91419219970703, "logps/ref_chosen": -51.991493225097656, "logps/ref_rejected": -86.0406265258789, "logps/rejected": -87.11837768554688, "loss": 1.2782, "margin_dpo/margin_mean": 1.1550464630126953, "margin_dpo/margin_std": 1.223704218864441, "step": 36 }, { "KL/chosen_KL_mean": 0.012973785400390625, "KL/mean": -0.5021691918373108, "KL/rejected_KL_mean": -1.0173149108886719, "KL/std": 1.0568256378173828, "epoch": 0.05433186490455213, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.0302867889404297, "fcm_dpo/q_t": 0.474433034658432, "grad_norm": 61.57312774658203, "learning_rate": 2.6086956521739126e-07, "logits/chosen": -0.5395572781562805, "logits/rejected": -0.4981452226638794, "logps/chosen": -62.79413604736328, "logps/ref_chosen": -62.807106018066406, "logps/ref_rejected": -77.89507293701172, "logps/rejected": -78.91238403320312, "loss": 1.2906, "margin_dpo/margin_mean": 1.0302867889404297, "margin_dpo/margin_std": 1.334470510482788, "step": 37 }, { "KL/chosen_KL_mean": 0.13834381103515625, "KL/mean": -0.5442078113555908, "KL/rejected_KL_mean": -1.2267570495605469, "KL/std": 1.350521206855774, "epoch": 0.055800293685756244, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.3651026487350464, "fcm_dpo/q_t": 0.4663234353065491, "grad_norm": 69.9059829711914, "learning_rate": 2.681159420289855e-07, "logits/chosen": -0.5119627714157104, "logits/rejected": -0.4792238473892212, "logps/chosen": -48.252174377441406, "logps/ref_chosen": -48.39051818847656, "logps/ref_rejected": -97.91244506835938, "logps/rejected": -99.13919830322266, "loss": 1.2613, "margin_dpo/margin_mean": 1.3651025295257568, "margin_dpo/margin_std": 1.6618046760559082, "step": 38 }, { "KL/chosen_KL_mean": 0.09346199035644531, "KL/mean": -0.7377809882164001, "KL/rejected_KL_mean": -1.5690231323242188, "KL/std": 1.284562349319458, "epoch": 0.05726872246696035, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.6624829769134521, "fcm_dpo/q_t": 0.4587351083755493, "grad_norm": 76.24556732177734, "learning_rate": 2.753623188405797e-07, "logits/chosen": -0.5892033576965332, "logits/rejected": -0.5514540076255798, "logps/chosen": -50.657012939453125, "logps/ref_chosen": -50.75047302246094, "logps/ref_rejected": -78.56951141357422, "logps/rejected": -80.13853454589844, "loss": 1.2315, "margin_dpo/margin_mean": 1.6624832153320312, "margin_dpo/margin_std": 1.3248234987258911, "step": 39 }, { "KL/chosen_KL_mean": 0.2594871520996094, "KL/mean": -0.5624086856842041, "KL/rejected_KL_mean": -1.3843040466308594, "KL/std": 1.4190328121185303, "epoch": 0.05873715124816446, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.6437921524047852, "fcm_dpo/q_t": 0.4594607949256897, "grad_norm": 60.083221435546875, "learning_rate": 2.8260869565217386e-07, "logits/chosen": -0.502472996711731, "logits/rejected": -0.47134459018707275, "logps/chosen": -57.72557830810547, "logps/ref_chosen": -57.985069274902344, "logps/ref_rejected": -74.3000717163086, "logps/rejected": -75.68437194824219, "loss": 1.2362, "margin_dpo/margin_mean": 1.643792986869812, "margin_dpo/margin_std": 1.6722097396850586, "step": 40 }, { "KL/chosen_KL_mean": -0.000926971435546875, "KL/mean": -0.9511067867279053, "KL/rejected_KL_mean": -1.9012832641601562, "KL/std": 1.8827296495437622, "epoch": 0.06020558002936858, "fcm_dpo/beta": 0.10071444511413574, "fcm_dpo/delta": 0.0709388256072998, "fcm_dpo/margin": 1.9003576040267944, "fcm_dpo/q_t": 0.45327073335647583, "grad_norm": 68.3230209350586, "learning_rate": 2.898550724637681e-07, "logits/chosen": -0.5602806806564331, "logits/rejected": -0.5249433517456055, "logps/chosen": -62.69674301147461, "logps/ref_chosen": -62.69581604003906, "logps/ref_rejected": -97.02352905273438, "logps/rejected": -98.92481231689453, "loss": 1.2148, "margin_dpo/margin_mean": 1.9003586769104004, "margin_dpo/margin_std": 2.0421926975250244, "step": 41 }, { "KL/chosen_KL_mean": 0.2127838134765625, "KL/mean": -1.085558295249939, "KL/rejected_KL_mean": -2.383892059326172, "KL/std": 2.2120487689971924, "epoch": 0.06167400881057269, "fcm_dpo/beta": 0.10279597342014313, "fcm_dpo/delta": 0.08528625965118408, "fcm_dpo/margin": 2.5966811180114746, "fcm_dpo/q_t": 0.4356800317764282, "grad_norm": 80.10550689697266, "learning_rate": 2.971014492753623e-07, "logits/chosen": -0.5423730611801147, "logits/rejected": -0.4957752823829651, "logps/chosen": -58.75364303588867, "logps/ref_chosen": -58.966426849365234, "logps/ref_rejected": -109.90837097167969, "logps/rejected": -112.29226684570312, "loss": 1.1536, "margin_dpo/margin_mean": 2.596680164337158, "margin_dpo/margin_std": 2.525965690612793, "step": 42 }, { "KL/chosen_KL_mean": 0.5627727508544922, "KL/mean": -0.6873226165771484, "KL/rejected_KL_mean": -1.9374237060546875, "KL/std": 1.8489782810211182, "epoch": 0.0631424375917768, "fcm_dpo/beta": 0.10443605482578278, "fcm_dpo/delta": 0.09136471152305603, "fcm_dpo/margin": 2.5001907348632812, "fcm_dpo/q_t": 0.4365711510181427, "grad_norm": 74.09246826171875, "learning_rate": 3.043478260869565e-07, "logits/chosen": -0.5289781093597412, "logits/rejected": -0.5033968687057495, "logps/chosen": -53.593223571777344, "logps/ref_chosen": -54.15599822998047, "logps/ref_rejected": -96.48019409179688, "logps/rejected": -98.41761779785156, "loss": 1.1533, "margin_dpo/margin_mean": 2.5001907348632812, "margin_dpo/margin_std": 2.0130953788757324, "step": 43 }, { "KL/chosen_KL_mean": 0.2562217712402344, "KL/mean": -1.1680686473846436, "KL/rejected_KL_mean": -2.5923614501953125, "KL/std": 2.291693687438965, "epoch": 0.06461086637298091, "fcm_dpo/beta": 0.1057717502117157, "fcm_dpo/delta": 0.0502425879240036, "fcm_dpo/margin": 2.8485794067382812, "fcm_dpo/q_t": 0.42686349153518677, "grad_norm": 81.92259216308594, "learning_rate": 3.115942028985507e-07, "logits/chosen": -0.48028355836868286, "logits/rejected": -0.46020689606666565, "logps/chosen": -49.8222770690918, "logps/ref_chosen": -50.07849884033203, "logps/ref_rejected": -108.78376007080078, "logps/rejected": -111.3761215209961, "loss": 1.1213, "margin_dpo/margin_mean": 2.848578453063965, "margin_dpo/margin_std": 2.2468016147613525, "step": 44 }, { "KL/chosen_KL_mean": 0.1484088897705078, "KL/mean": -0.985015869140625, "KL/rejected_KL_mean": -2.118438720703125, "KL/std": 2.0992965698242188, "epoch": 0.06607929515418502, "fcm_dpo/beta": 0.10800284147262573, "fcm_dpo/delta": 0.10763342678546906, "fcm_dpo/margin": 2.2668519020080566, "fcm_dpo/q_t": 0.4414454698562622, "grad_norm": 65.85893249511719, "learning_rate": 3.188405797101449e-07, "logits/chosen": -0.45797061920166016, "logits/rejected": -0.4448603093624115, "logps/chosen": -48.266517639160156, "logps/ref_chosen": -48.4149284362793, "logps/ref_rejected": -77.93643188476562, "logps/rejected": -80.05487060546875, "loss": 1.1763, "margin_dpo/margin_mean": 2.2668511867523193, "margin_dpo/margin_std": 2.624885082244873, "step": 45 }, { "KL/chosen_KL_mean": 0.2292613983154297, "KL/mean": -1.2564077377319336, "KL/rejected_KL_mean": -2.742076873779297, "KL/std": 2.657705783843994, "epoch": 0.06754772393538913, "fcm_dpo/beta": 0.1090465784072876, "fcm_dpo/delta": 0.026768168434500694, "fcm_dpo/margin": 2.9713401794433594, "fcm_dpo/q_t": 0.4231029450893402, "grad_norm": 73.93266296386719, "learning_rate": 3.260869565217391e-07, "logits/chosen": -0.5324057340621948, "logits/rejected": -0.48163044452667236, "logps/chosen": -55.770164489746094, "logps/ref_chosen": -55.999427795410156, "logps/ref_rejected": -95.652587890625, "logps/rejected": -98.39466857910156, "loss": 1.1181, "margin_dpo/margin_mean": 2.9713401794433594, "margin_dpo/margin_std": 3.2657053470611572, "step": 46 }, { "KL/chosen_KL_mean": 0.3735790252685547, "KL/mean": -1.1065313816070557, "KL/rejected_KL_mean": -2.5866432189941406, "KL/std": 2.6243019104003906, "epoch": 0.06901615271659324, "fcm_dpo/beta": 0.10978727042675018, "fcm_dpo/delta": 0.02556876838207245, "fcm_dpo/margin": 2.960222005844116, "fcm_dpo/q_t": 0.4216909408569336, "grad_norm": 70.09626007080078, "learning_rate": 3.333333333333333e-07, "logits/chosen": -0.5489112734794617, "logits/rejected": -0.49465492367744446, "logps/chosen": -57.55249786376953, "logps/ref_chosen": -57.92607879638672, "logps/ref_rejected": -94.67920684814453, "logps/rejected": -97.26585388183594, "loss": 1.1101, "margin_dpo/margin_mean": 2.9602224826812744, "margin_dpo/margin_std": 2.731231927871704, "step": 47 }, { "KL/chosen_KL_mean": 0.04618263244628906, "KL/mean": -1.5137662887573242, "KL/rejected_KL_mean": -3.0737152099609375, "KL/std": 2.6629862785339355, "epoch": 0.07048458149779736, "fcm_dpo/beta": 0.1094270646572113, "fcm_dpo/delta": 0.00876920111477375, "fcm_dpo/margin": 3.1198978424072266, "fcm_dpo/q_t": 0.4179936647415161, "grad_norm": 76.3647689819336, "learning_rate": 3.4057971014492755e-07, "logits/chosen": -0.6007370948791504, "logits/rejected": -0.5431762933731079, "logps/chosen": -57.14189147949219, "logps/ref_chosen": -57.188072204589844, "logps/ref_rejected": -88.0166015625, "logps/rejected": -91.09031677246094, "loss": 1.1026, "margin_dpo/margin_mean": 3.1198978424072266, "margin_dpo/margin_std": 2.9437079429626465, "step": 48 }, { "KL/chosen_KL_mean": 0.3652973175048828, "KL/mean": -1.5648986101150513, "KL/rejected_KL_mean": -3.4950904846191406, "KL/std": 3.375471591949463, "epoch": 0.07195301027900147, "fcm_dpo/beta": 0.1089581847190857, "fcm_dpo/delta": -0.07370879501104355, "fcm_dpo/margin": 3.8603830337524414, "fcm_dpo/q_t": 0.4010041654109955, "grad_norm": 74.53569030761719, "learning_rate": 3.478260869565217e-07, "logits/chosen": -0.5346788763999939, "logits/rejected": -0.4745950698852539, "logps/chosen": -61.31997299194336, "logps/ref_chosen": -61.685272216796875, "logps/ref_rejected": -83.76747131347656, "logps/rejected": -87.26256561279297, "loss": 1.0501, "margin_dpo/margin_mean": 3.8603830337524414, "margin_dpo/margin_std": 3.772658348083496, "step": 49 }, { "KL/chosen_KL_mean": -0.15436363220214844, "KL/mean": -2.2124156951904297, "KL/rejected_KL_mean": -4.2704620361328125, "KL/std": 3.501716136932373, "epoch": 0.07342143906020558, "fcm_dpo/beta": 0.10655629634857178, "fcm_dpo/delta": -0.09280845522880554, "fcm_dpo/margin": 4.116090297698975, "fcm_dpo/q_t": 0.3962603211402893, "grad_norm": 65.80304718017578, "learning_rate": 3.5507246376811595e-07, "logits/chosen": -0.5511945486068726, "logits/rejected": -0.5155908465385437, "logps/chosen": -58.878501892089844, "logps/ref_chosen": -58.72413635253906, "logps/ref_rejected": -96.35814666748047, "logps/rejected": -100.62860870361328, "loss": 1.0377, "margin_dpo/margin_mean": 4.116090297698975, "margin_dpo/margin_std": 4.020236015319824, "step": 50 }, { "KL/chosen_KL_mean": -0.3026409149169922, "KL/mean": -2.3341808319091797, "KL/rejected_KL_mean": -4.365726470947266, "KL/std": 4.1528143882751465, "epoch": 0.07488986784140969, "fcm_dpo/beta": 0.10447999835014343, "fcm_dpo/delta": -0.07857784628868103, "fcm_dpo/margin": 4.063086032867432, "fcm_dpo/q_t": 0.40330058336257935, "grad_norm": 53.87831115722656, "learning_rate": 3.6231884057971015e-07, "logits/chosen": -0.5541812181472778, "logits/rejected": -0.5227280259132385, "logps/chosen": -61.676307678222656, "logps/ref_chosen": -61.3736686706543, "logps/ref_rejected": -76.00199890136719, "logps/rejected": -80.36772918701172, "loss": 1.0713, "margin_dpo/margin_mean": 4.063085556030273, "margin_dpo/margin_std": 5.101934432983398, "step": 51 }, { "KL/chosen_KL_mean": 0.32053184509277344, "KL/mean": -2.7178292274475098, "KL/rejected_KL_mean": -5.756187438964844, "KL/std": 4.713332176208496, "epoch": 0.0763582966226138, "fcm_dpo/beta": 0.10033614188432693, "fcm_dpo/delta": -0.2765745520591736, "fcm_dpo/margin": 6.076724052429199, "fcm_dpo/q_t": 0.3581005930900574, "grad_norm": 59.589393615722656, "learning_rate": 3.695652173913043e-07, "logits/chosen": -0.5497130751609802, "logits/rejected": -0.49386975169181824, "logps/chosen": -52.016822814941406, "logps/ref_chosen": -52.33735656738281, "logps/ref_rejected": -79.97391510009766, "logps/rejected": -85.7301025390625, "loss": 0.9249, "margin_dpo/margin_mean": 6.076724052429199, "margin_dpo/margin_std": 5.1023430824279785, "step": 52 }, { "KL/chosen_KL_mean": -0.2184162139892578, "KL/mean": -3.311248302459717, "KL/rejected_KL_mean": -6.404079437255859, "KL/std": 5.3616042137146, "epoch": 0.07782672540381791, "fcm_dpo/beta": 0.09587851911783218, "fcm_dpo/delta": -0.2576024532318115, "fcm_dpo/margin": 6.185666561126709, "fcm_dpo/q_t": 0.3689725399017334, "grad_norm": 57.90284729003906, "learning_rate": 3.7681159420289855e-07, "logits/chosen": -0.6244029998779297, "logits/rejected": -0.6030861139297485, "logps/chosen": -53.53306579589844, "logps/ref_chosen": -53.31465148925781, "logps/ref_rejected": -91.78359985351562, "logps/rejected": -98.18767547607422, "loss": 0.9643, "margin_dpo/margin_mean": 6.185666084289551, "margin_dpo/margin_std": 6.1749587059021, "step": 53 }, { "KL/chosen_KL_mean": -0.3918323516845703, "KL/mean": -3.089035987854004, "KL/rejected_KL_mean": -5.7862396240234375, "KL/std": 5.119150161743164, "epoch": 0.07929515418502203, "fcm_dpo/beta": 0.09195560961961746, "fcm_dpo/delta": -0.1534874439239502, "fcm_dpo/margin": 5.394400119781494, "fcm_dpo/q_t": 0.3844445049762726, "grad_norm": 52.45349884033203, "learning_rate": 3.8405797101449274e-07, "logits/chosen": -0.5996332764625549, "logits/rejected": -0.5465092062950134, "logps/chosen": -51.08049392700195, "logps/ref_chosen": -50.68865966796875, "logps/ref_rejected": -91.71539306640625, "logps/rejected": -97.50163269042969, "loss": 0.9999, "margin_dpo/margin_mean": 5.394399642944336, "margin_dpo/margin_std": 5.113949775695801, "step": 54 }, { "KL/chosen_KL_mean": -0.9258518218994141, "KL/mean": -4.1504225730896, "KL/rejected_KL_mean": -7.374988555908203, "KL/std": 6.164496421813965, "epoch": 0.08076358296622614, "fcm_dpo/beta": 0.08832551538944244, "fcm_dpo/delta": -0.23232939839363098, "fcm_dpo/margin": 6.4491424560546875, "fcm_dpo/q_t": 0.3728960454463959, "grad_norm": 50.09782028198242, "learning_rate": 3.9130434782608694e-07, "logits/chosen": -0.6301623582839966, "logits/rejected": -0.5663325786590576, "logps/chosen": -63.54108428955078, "logps/ref_chosen": -62.615234375, "logps/ref_rejected": -88.99349975585938, "logps/rejected": -96.36849212646484, "loss": 0.991, "margin_dpo/margin_mean": 6.449141502380371, "margin_dpo/margin_std": 7.437203407287598, "step": 55 }, { "KL/chosen_KL_mean": -0.7239055633544922, "KL/mean": -3.78607177734375, "KL/rejected_KL_mean": -6.848243713378906, "KL/std": 6.047898292541504, "epoch": 0.08223201174743025, "fcm_dpo/beta": 0.08475831151008606, "fcm_dpo/delta": -0.17821934819221497, "fcm_dpo/margin": 6.124341011047363, "fcm_dpo/q_t": 0.38335487246513367, "grad_norm": 44.37767791748047, "learning_rate": 3.9855072463768114e-07, "logits/chosen": -0.619949460029602, "logits/rejected": -0.5779317021369934, "logps/chosen": -58.65663146972656, "logps/ref_chosen": -57.9327278137207, "logps/ref_rejected": -94.1744384765625, "logps/rejected": -101.0226821899414, "loss": 1.0199, "margin_dpo/margin_mean": 6.124341011047363, "margin_dpo/margin_std": 7.334414482116699, "step": 56 }, { "KL/chosen_KL_mean": -0.7689876556396484, "KL/mean": -4.241398811340332, "KL/rejected_KL_mean": -7.71380615234375, "KL/std": 6.030036449432373, "epoch": 0.08370044052863436, "fcm_dpo/beta": 0.08145327121019363, "fcm_dpo/delta": -0.2280295044183731, "fcm_dpo/margin": 6.944817066192627, "fcm_dpo/q_t": 0.3689224123954773, "grad_norm": 48.75211715698242, "learning_rate": 4.057971014492754e-07, "logits/chosen": -0.5575429797172546, "logits/rejected": -0.5283694267272949, "logps/chosen": -71.26427459716797, "logps/ref_chosen": -70.49528503417969, "logps/ref_rejected": -95.56546020507812, "logps/rejected": -103.27926635742188, "loss": 0.968, "margin_dpo/margin_mean": 6.944817543029785, "margin_dpo/margin_std": 6.705440044403076, "step": 57 }, { "KL/chosen_KL_mean": -1.0128746032714844, "KL/mean": -4.868136405944824, "KL/rejected_KL_mean": -8.723396301269531, "KL/std": 6.835504531860352, "epoch": 0.08516886930983847, "fcm_dpo/beta": 0.07740665972232819, "fcm_dpo/delta": -0.2618388831615448, "fcm_dpo/margin": 7.710516929626465, "fcm_dpo/q_t": 0.36622318625450134, "grad_norm": 50.176239013671875, "learning_rate": 4.1304347826086954e-07, "logits/chosen": -0.5943198204040527, "logits/rejected": -0.5183029174804688, "logps/chosen": -63.14581298828125, "logps/ref_chosen": -62.13294219970703, "logps/ref_rejected": -84.61729431152344, "logps/rejected": -93.34069061279297, "loss": 0.9646, "margin_dpo/margin_mean": 7.710516929626465, "margin_dpo/margin_std": 8.16922378540039, "step": 58 }, { "KL/chosen_KL_mean": -1.5172653198242188, "KL/mean": -5.735530853271484, "KL/rejected_KL_mean": -9.95379638671875, "KL/std": 7.462213516235352, "epoch": 0.08663729809104258, "fcm_dpo/beta": 0.07241758704185486, "fcm_dpo/delta": -0.28130438923835754, "fcm_dpo/margin": 8.436534881591797, "fcm_dpo/q_t": 0.3623542785644531, "grad_norm": 47.03469467163086, "learning_rate": 4.2028985507246374e-07, "logits/chosen": -0.6317316293716431, "logits/rejected": -0.5917737483978271, "logps/chosen": -53.449790954589844, "logps/ref_chosen": -51.932525634765625, "logps/ref_rejected": -88.88520050048828, "logps/rejected": -98.83899688720703, "loss": 0.9528, "margin_dpo/margin_mean": 8.436534881591797, "margin_dpo/margin_std": 8.641322135925293, "step": 59 }, { "KL/chosen_KL_mean": -2.6389827728271484, "KL/mean": -6.052818298339844, "KL/rejected_KL_mean": -9.46664810180664, "KL/std": 6.807157039642334, "epoch": 0.0881057268722467, "fcm_dpo/beta": 0.07023543119430542, "fcm_dpo/delta": -0.13676942884922028, "fcm_dpo/margin": 6.827664375305176, "fcm_dpo/q_t": 0.3875412940979004, "grad_norm": 48.9000244140625, "learning_rate": 4.2753623188405794e-07, "logits/chosen": -0.5942388772964478, "logits/rejected": -0.5327954292297363, "logps/chosen": -63.58116912841797, "logps/ref_chosen": -60.94218826293945, "logps/ref_rejected": -85.39340209960938, "logps/rejected": -94.86004638671875, "loss": 1.0321, "margin_dpo/margin_mean": 6.827664375305176, "margin_dpo/margin_std": 7.711092948913574, "step": 60 }, { "KL/chosen_KL_mean": -1.5054092407226562, "KL/mean": -5.698062896728516, "KL/rejected_KL_mean": -9.890716552734375, "KL/std": 9.148716926574707, "epoch": 0.08957415565345081, "fcm_dpo/beta": 0.06738986074924469, "fcm_dpo/delta": -0.22911055386066437, "fcm_dpo/margin": 8.385308265686035, "fcm_dpo/q_t": 0.37978309392929077, "grad_norm": 41.77968978881836, "learning_rate": 4.3478260869565214e-07, "logits/chosen": -0.6104527115821838, "logits/rejected": -0.5764514207839966, "logps/chosen": -62.13893127441406, "logps/ref_chosen": -60.633522033691406, "logps/ref_rejected": -89.85249328613281, "logps/rejected": -99.74320983886719, "loss": 1.0155, "margin_dpo/margin_mean": 8.385307312011719, "margin_dpo/margin_std": 11.387674331665039, "step": 61 }, { "KL/chosen_KL_mean": -1.747161865234375, "KL/mean": -4.856363773345947, "KL/rejected_KL_mean": -7.965564727783203, "KL/std": 6.7253522872924805, "epoch": 0.09104258443465492, "fcm_dpo/beta": 0.06641829758882523, "fcm_dpo/delta": -0.06573815643787384, "fcm_dpo/margin": 6.218404293060303, "fcm_dpo/q_t": 0.40589573979377747, "grad_norm": 40.88990020751953, "learning_rate": 4.420289855072464e-07, "logits/chosen": -0.5971473455429077, "logits/rejected": -0.5622413158416748, "logps/chosen": -57.89793395996094, "logps/ref_chosen": -56.15077209472656, "logps/ref_rejected": -75.56619262695312, "logps/rejected": -83.5317611694336, "loss": 1.0859, "margin_dpo/margin_mean": 6.218404769897461, "margin_dpo/margin_std": 8.258329391479492, "step": 62 }, { "KL/chosen_KL_mean": -2.8008480072021484, "KL/mean": -7.036721229553223, "KL/rejected_KL_mean": -11.272594451904297, "KL/std": 8.544384956359863, "epoch": 0.09251101321585903, "fcm_dpo/beta": 0.06367187201976776, "fcm_dpo/delta": -0.20195239782333374, "fcm_dpo/margin": 8.4717435836792, "fcm_dpo/q_t": 0.37538081407546997, "grad_norm": 42.83129119873047, "learning_rate": 4.4927536231884053e-07, "logits/chosen": -0.583976686000824, "logits/rejected": -0.5383748412132263, "logps/chosen": -75.9482421875, "logps/ref_chosen": -73.14739227294922, "logps/ref_rejected": -97.61006164550781, "logps/rejected": -108.88265991210938, "loss": 0.9938, "margin_dpo/margin_mean": 8.471742630004883, "margin_dpo/margin_std": 8.990781784057617, "step": 63 }, { "KL/chosen_KL_mean": -1.124542236328125, "KL/mean": -6.140851020812988, "KL/rejected_KL_mean": -11.157161712646484, "KL/std": 9.27166748046875, "epoch": 0.09397944199706314, "fcm_dpo/beta": 0.06026393920183182, "fcm_dpo/delta": -0.2753208875656128, "fcm_dpo/margin": 10.03261661529541, "fcm_dpo/q_t": 0.36537328362464905, "grad_norm": 38.817970275878906, "learning_rate": 4.5652173913043473e-07, "logits/chosen": -0.5995528697967529, "logits/rejected": -0.5680242776870728, "logps/chosen": -55.12314224243164, "logps/ref_chosen": -53.998600006103516, "logps/ref_rejected": -93.53019714355469, "logps/rejected": -104.68736267089844, "loss": 0.9729, "margin_dpo/margin_mean": 10.032617568969727, "margin_dpo/margin_std": 11.401942253112793, "step": 64 }, { "KL/chosen_KL_mean": -3.4033946990966797, "KL/mean": -8.52043342590332, "KL/rejected_KL_mean": -13.637474060058594, "KL/std": 10.473337173461914, "epoch": 0.09544787077826726, "fcm_dpo/beta": 0.0574415922164917, "fcm_dpo/delta": -0.2546284794807434, "fcm_dpo/margin": 10.234081268310547, "fcm_dpo/q_t": 0.3663010597229004, "grad_norm": 39.211761474609375, "learning_rate": 4.63768115942029e-07, "logits/chosen": -0.6607149839401245, "logits/rejected": -0.6480229496955872, "logps/chosen": -68.23939514160156, "logps/ref_chosen": -64.83599853515625, "logps/ref_rejected": -109.94645690917969, "logps/rejected": -123.58393859863281, "loss": 0.9811, "margin_dpo/margin_mean": 10.234081268310547, "margin_dpo/margin_std": 11.489011764526367, "step": 65 }, { "KL/chosen_KL_mean": -3.2082653045654297, "KL/mean": -7.8406548500061035, "KL/rejected_KL_mean": -12.473045349121094, "KL/std": 9.421492576599121, "epoch": 0.09691629955947137, "fcm_dpo/beta": 0.055195845663547516, "fcm_dpo/delta": -0.17202530801296234, "fcm_dpo/margin": 9.264777183532715, "fcm_dpo/q_t": 0.38581931591033936, "grad_norm": 34.907039642333984, "learning_rate": 4.7101449275362313e-07, "logits/chosen": -0.6280045509338379, "logits/rejected": -0.5945169925689697, "logps/chosen": -54.651790618896484, "logps/ref_chosen": -51.44352722167969, "logps/ref_rejected": -75.63629913330078, "logps/rejected": -88.10934448242188, "loss": 1.025, "margin_dpo/margin_mean": 9.264778137207031, "margin_dpo/margin_std": 11.395795822143555, "step": 66 }, { "KL/chosen_KL_mean": -2.8829917907714844, "KL/mean": -7.654372215270996, "KL/rejected_KL_mean": -12.42575454711914, "KL/std": 9.585004806518555, "epoch": 0.09838472834067548, "fcm_dpo/beta": 0.053811218589544296, "fcm_dpo/delta": -0.1720547378063202, "fcm_dpo/margin": 9.542757987976074, "fcm_dpo/q_t": 0.38516417145729065, "grad_norm": 35.49789047241211, "learning_rate": 4.782608695652174e-07, "logits/chosen": -0.599655032157898, "logits/rejected": -0.5566652417182922, "logps/chosen": -62.22379684448242, "logps/ref_chosen": -59.34080505371094, "logps/ref_rejected": -72.78728485107422, "logps/rejected": -85.21304321289062, "loss": 1.0218, "margin_dpo/margin_mean": 9.542757987976074, "margin_dpo/margin_std": 11.59894847869873, "step": 67 }, { "KL/chosen_KL_mean": -3.4374046325683594, "KL/mean": -7.992835521697998, "KL/rejected_KL_mean": -12.548263549804688, "KL/std": 8.550180435180664, "epoch": 0.09985315712187959, "fcm_dpo/beta": 0.05226890742778778, "fcm_dpo/delta": -0.13234364986419678, "fcm_dpo/margin": 9.110858917236328, "fcm_dpo/q_t": 0.38975322246551514, "grad_norm": 34.271942138671875, "learning_rate": 4.855072463768116e-07, "logits/chosen": -0.6388707756996155, "logits/rejected": -0.5778865814208984, "logps/chosen": -68.64323425292969, "logps/ref_chosen": -65.2058334350586, "logps/ref_rejected": -77.20724487304688, "logps/rejected": -89.75550842285156, "loss": 1.0236, "margin_dpo/margin_mean": 9.110858917236328, "margin_dpo/margin_std": 9.605062484741211, "step": 68 }, { "KL/chosen_KL_mean": -4.114860534667969, "KL/mean": -9.939827919006348, "KL/rejected_KL_mean": -15.764793395996094, "KL/std": 10.603840827941895, "epoch": 0.1013215859030837, "fcm_dpo/beta": 0.04974665865302086, "fcm_dpo/delta": -0.24509744346141815, "fcm_dpo/margin": 11.649931907653809, "fcm_dpo/q_t": 0.3687508702278137, "grad_norm": 36.500450134277344, "learning_rate": 4.927536231884058e-07, "logits/chosen": -0.6228272318840027, "logits/rejected": -0.5997121930122375, "logps/chosen": -63.93410110473633, "logps/ref_chosen": -59.81924057006836, "logps/ref_rejected": -103.38886260986328, "logps/rejected": -119.15365600585938, "loss": 0.9628, "margin_dpo/margin_mean": 11.649931907653809, "margin_dpo/margin_std": 11.750423431396484, "step": 69 }, { "KL/chosen_KL_mean": -6.090396881103516, "KL/mean": -12.428565979003906, "KL/rejected_KL_mean": -18.76673126220703, "KL/std": 13.280801773071289, "epoch": 0.1027900146842878, "fcm_dpo/beta": 0.047557003796100616, "fcm_dpo/delta": -0.2684404253959656, "fcm_dpo/margin": 12.676331520080566, "fcm_dpo/q_t": 0.36635735630989075, "grad_norm": 36.634315490722656, "learning_rate": 5e-07, "logits/chosen": -0.6058599948883057, "logits/rejected": -0.5707762241363525, "logps/chosen": -68.02104187011719, "logps/ref_chosen": -61.930641174316406, "logps/ref_rejected": -91.06078338623047, "logps/rejected": -109.8275146484375, "loss": 0.9698, "margin_dpo/margin_mean": 12.676332473754883, "margin_dpo/margin_std": 14.070359230041504, "step": 70 }, { "KL/chosen_KL_mean": -6.061298370361328, "KL/mean": -13.633516311645508, "KL/rejected_KL_mean": -21.205734252929688, "KL/std": 14.106433868408203, "epoch": 0.10425844346549193, "fcm_dpo/beta": 0.044265132397413254, "fcm_dpo/delta": -0.34492480754852295, "fcm_dpo/margin": 15.14443588256836, "fcm_dpo/q_t": 0.3534585237503052, "grad_norm": 34.083553314208984, "learning_rate": 4.999967061337492e-07, "logits/chosen": -0.6632341146469116, "logits/rejected": -0.6252255439758301, "logps/chosen": -67.81163787841797, "logps/ref_chosen": -61.750335693359375, "logps/ref_rejected": -97.33662414550781, "logps/rejected": -118.5423583984375, "loss": 0.923, "margin_dpo/margin_mean": 15.14443588256836, "margin_dpo/margin_std": 15.41769790649414, "step": 71 }, { "KL/chosen_KL_mean": -7.117975234985352, "KL/mean": -15.00790786743164, "KL/rejected_KL_mean": -22.897842407226562, "KL/std": 15.219398498535156, "epoch": 0.10572687224669604, "fcm_dpo/beta": 0.04147840291261673, "fcm_dpo/delta": -0.3267444372177124, "fcm_dpo/margin": 15.779868125915527, "fcm_dpo/q_t": 0.35581284761428833, "grad_norm": 33.533538818359375, "learning_rate": 4.999868246217933e-07, "logits/chosen": -0.659099817276001, "logits/rejected": -0.6226309537887573, "logps/chosen": -73.17138671875, "logps/ref_chosen": -66.05341339111328, "logps/ref_rejected": -95.2869873046875, "logps/rejected": -118.18482971191406, "loss": 0.9517, "margin_dpo/margin_mean": 15.779868125915527, "margin_dpo/margin_std": 18.023799896240234, "step": 72 }, { "KL/chosen_KL_mean": -9.201299667358398, "KL/mean": -17.138395309448242, "KL/rejected_KL_mean": -25.07549285888672, "KL/std": 19.838104248046875, "epoch": 0.10719530102790015, "fcm_dpo/beta": 0.03898513317108154, "fcm_dpo/delta": -0.28837209939956665, "fcm_dpo/margin": 15.87419319152832, "fcm_dpo/q_t": 0.37557166814804077, "grad_norm": 34.42801284790039, "learning_rate": 4.999703557245192e-07, "logits/chosen": -0.690470814704895, "logits/rejected": -0.6502715349197388, "logps/chosen": -75.45757293701172, "logps/ref_chosen": -66.25627136230469, "logps/ref_rejected": -90.45613098144531, "logps/rejected": -115.53163146972656, "loss": 1.0506, "margin_dpo/margin_mean": 15.87419319152832, "margin_dpo/margin_std": 24.97110366821289, "step": 73 }, { "KL/chosen_KL_mean": -9.636629104614258, "KL/mean": -18.45030403137207, "KL/rejected_KL_mean": -27.263980865478516, "KL/std": 21.931224822998047, "epoch": 0.10866372980910426, "fcm_dpo/beta": 0.036717548966407776, "fcm_dpo/delta": -0.3187072277069092, "fcm_dpo/margin": 17.62735366821289, "fcm_dpo/q_t": 0.3689573109149933, "grad_norm": 35.52349853515625, "learning_rate": 4.999472998758977e-07, "logits/chosen": -0.6715037822723389, "logits/rejected": -0.6643733978271484, "logps/chosen": -63.0615119934082, "logps/ref_chosen": -53.42488098144531, "logps/ref_rejected": -95.94693756103516, "logps/rejected": -123.21092224121094, "loss": 1.007, "margin_dpo/margin_mean": 17.627355575561523, "margin_dpo/margin_std": 27.311134338378906, "step": 74 }, { "KL/chosen_KL_mean": -9.528879165649414, "KL/mean": -21.829769134521484, "KL/rejected_KL_mean": -34.130653381347656, "KL/std": 22.312917709350586, "epoch": 0.11013215859030837, "fcm_dpo/beta": 0.03375118225812912, "fcm_dpo/delta": -0.5213460922241211, "fcm_dpo/margin": 24.601776123046875, "fcm_dpo/q_t": 0.3274177014827728, "grad_norm": 29.952869415283203, "learning_rate": 4.999176576834721e-07, "logits/chosen": -0.6653124094009399, "logits/rejected": -0.6586759090423584, "logps/chosen": -61.390541076660156, "logps/ref_chosen": -51.861663818359375, "logps/ref_rejected": -111.25398254394531, "logps/rejected": -145.3846435546875, "loss": 0.8665, "margin_dpo/margin_mean": 24.601776123046875, "margin_dpo/margin_std": 25.416664123535156, "step": 75 }, { "KL/chosen_KL_mean": -11.458761215209961, "KL/mean": -19.180240631103516, "KL/rejected_KL_mean": -26.901718139648438, "KL/std": 18.074440002441406, "epoch": 0.11160058737151249, "fcm_dpo/beta": 0.031918227672576904, "fcm_dpo/delta": -0.15023696422576904, "fcm_dpo/margin": 15.442970275878906, "fcm_dpo/q_t": 0.3866853415966034, "grad_norm": 30.53144645690918, "learning_rate": 4.998814299283415e-07, "logits/chosen": -0.6922107934951782, "logits/rejected": -0.6505155563354492, "logps/chosen": -64.72480010986328, "logps/ref_chosen": -53.26603698730469, "logps/ref_rejected": -78.21662902832031, "logps/rejected": -105.11834716796875, "loss": 1.0424, "margin_dpo/margin_mean": 15.442970275878906, "margin_dpo/margin_std": 19.84844207763672, "step": 76 }, { "KL/chosen_KL_mean": -10.305963516235352, "KL/mean": -22.921131134033203, "KL/rejected_KL_mean": -35.53630828857422, "KL/std": 24.69320297241211, "epoch": 0.1130690161527166, "fcm_dpo/beta": 0.029510796070098877, "fcm_dpo/delta": -0.4299943149089813, "fcm_dpo/margin": 25.23033905029297, "fcm_dpo/q_t": 0.3367495834827423, "grad_norm": 31.526256561279297, "learning_rate": 4.998386175651409e-07, "logits/chosen": -0.6924519538879395, "logits/rejected": -0.6561889052391052, "logps/chosen": -68.40264129638672, "logps/ref_chosen": -58.0966796875, "logps/ref_rejected": -93.77361297607422, "logps/rejected": -129.30992126464844, "loss": 0.913, "margin_dpo/margin_mean": 25.23033905029297, "margin_dpo/margin_std": 27.491090774536133, "step": 77 }, { "KL/chosen_KL_mean": -10.388359069824219, "KL/mean": -19.755765914916992, "KL/rejected_KL_mean": -29.123172760009766, "KL/std": 19.83435821533203, "epoch": 0.1145374449339207, "fcm_dpo/beta": 0.028303634375333786, "fcm_dpo/delta": -0.19002512097358704, "fcm_dpo/margin": 18.73480987548828, "fcm_dpo/q_t": 0.3818398714065552, "grad_norm": 28.295921325683594, "learning_rate": 4.997892217220159e-07, "logits/chosen": -0.6519057154655457, "logits/rejected": -0.628269374370575, "logps/chosen": -66.00214385986328, "logps/ref_chosen": -55.61378479003906, "logps/ref_rejected": -84.93436431884766, "logps/rejected": -114.05754089355469, "loss": 1.0187, "margin_dpo/margin_mean": 18.73480987548828, "margin_dpo/margin_std": 22.818925857543945, "step": 78 }, { "KL/chosen_KL_mean": -10.710470199584961, "KL/mean": -20.990741729736328, "KL/rejected_KL_mean": -31.271018981933594, "KL/std": 22.39347267150879, "epoch": 0.11600587371512482, "fcm_dpo/beta": 0.027051560580730438, "fcm_dpo/delta": -0.21849405765533447, "fcm_dpo/margin": 20.56053924560547, "fcm_dpo/q_t": 0.3806380033493042, "grad_norm": 25.06480598449707, "learning_rate": 4.997332437005931e-07, "logits/chosen": -0.6494733095169067, "logits/rejected": -0.6198798418045044, "logps/chosen": -66.16095733642578, "logps/ref_chosen": -55.45048522949219, "logps/ref_rejected": -87.64756774902344, "logps/rejected": -118.9185791015625, "loss": 1.0248, "margin_dpo/margin_mean": 20.56053924560547, "margin_dpo/margin_std": 27.430313110351562, "step": 79 }, { "KL/chosen_KL_mean": -13.9652099609375, "KL/mean": -23.680742263793945, "KL/rejected_KL_mean": -33.396278381347656, "KL/std": 23.99298095703125, "epoch": 0.11747430249632893, "fcm_dpo/beta": 0.02592482417821884, "fcm_dpo/delta": -0.16428518295288086, "fcm_dpo/margin": 19.431066513061523, "fcm_dpo/q_t": 0.3933182656764984, "grad_norm": 26.278663635253906, "learning_rate": 4.996706849759452e-07, "logits/chosen": -0.7399425506591797, "logits/rejected": -0.7012728452682495, "logps/chosen": -72.4844970703125, "logps/ref_chosen": -58.519290924072266, "logps/ref_rejected": -87.54750061035156, "logps/rejected": -120.94377899169922, "loss": 1.0634, "margin_dpo/margin_mean": 19.43106460571289, "margin_dpo/margin_std": 27.870834350585938, "step": 80 }, { "KL/chosen_KL_mean": -13.097003936767578, "KL/mean": -26.419586181640625, "KL/rejected_KL_mean": -39.74217224121094, "KL/std": 28.39004898071289, "epoch": 0.11894273127753303, "fcm_dpo/beta": 0.02453438937664032, "fcm_dpo/delta": -0.3274933397769928, "fcm_dpo/margin": 26.645172119140625, "fcm_dpo/q_t": 0.36159569025039673, "grad_norm": 27.374237060546875, "learning_rate": 4.996015471965529e-07, "logits/chosen": -0.7171902656555176, "logits/rejected": -0.6900858879089355, "logps/chosen": -79.54586791992188, "logps/ref_chosen": -66.44886779785156, "logps/ref_rejected": -129.66270446777344, "logps/rejected": -169.40487670898438, "loss": 0.9726, "margin_dpo/margin_mean": 26.645172119140625, "margin_dpo/margin_std": 33.81873321533203, "step": 81 }, { "KL/chosen_KL_mean": -15.472373962402344, "KL/mean": -26.10114288330078, "KL/rejected_KL_mean": -36.729915618896484, "KL/std": 25.81859588623047, "epoch": 0.12041116005873716, "fcm_dpo/beta": 0.02389688789844513, "fcm_dpo/delta": -0.16673710942268372, "fcm_dpo/margin": 21.257539749145508, "fcm_dpo/q_t": 0.39267510175704956, "grad_norm": 29.728790283203125, "learning_rate": 4.995258321842611e-07, "logits/chosen": -0.6627001166343689, "logits/rejected": -0.6532809734344482, "logps/chosen": -67.70475769042969, "logps/ref_chosen": -52.232383728027344, "logps/ref_rejected": -90.74325561523438, "logps/rejected": -127.4731674194336, "loss": 1.1015, "margin_dpo/margin_mean": 21.257539749145508, "margin_dpo/margin_std": 35.67451095581055, "step": 82 }, { "KL/chosen_KL_mean": -15.941610336303711, "KL/mean": -28.5517635345459, "KL/rejected_KL_mean": -41.16191482543945, "KL/std": 26.822208404541016, "epoch": 0.12187958883994127, "fcm_dpo/beta": 0.02258511260151863, "fcm_dpo/delta": -0.23390883207321167, "fcm_dpo/margin": 25.22031021118164, "fcm_dpo/q_t": 0.37505269050598145, "grad_norm": 25.695941925048828, "learning_rate": 4.994435419342304e-07, "logits/chosen": -0.6607187986373901, "logits/rejected": -0.6281884908676147, "logps/chosen": -71.76899719238281, "logps/ref_chosen": -55.82738494873047, "logps/ref_rejected": -103.71589660644531, "logps/rejected": -144.8778076171875, "loss": 1.0091, "margin_dpo/margin_mean": 25.220308303833008, "margin_dpo/margin_std": 31.849729537963867, "step": 83 }, { "KL/chosen_KL_mean": -15.136430740356445, "KL/mean": -25.702499389648438, "KL/rejected_KL_mean": -36.26857376098633, "KL/std": 22.970796585083008, "epoch": 0.12334801762114538, "fcm_dpo/beta": 0.021835383027791977, "fcm_dpo/delta": -0.11882907897233963, "fcm_dpo/margin": 21.132139205932617, "fcm_dpo/q_t": 0.3920828700065613, "grad_norm": 24.43920135498047, "learning_rate": 4.993546786148857e-07, "logits/chosen": -0.6809099912643433, "logits/rejected": -0.6443426609039307, "logps/chosen": -82.3125991821289, "logps/ref_chosen": -67.1761703491211, "logps/ref_rejected": -87.29859924316406, "logps/rejected": -123.56717681884766, "loss": 1.0414, "margin_dpo/margin_mean": 21.132139205932617, "margin_dpo/margin_std": 23.453916549682617, "step": 84 }, { "KL/chosen_KL_mean": -15.309465408325195, "KL/mean": -26.37848663330078, "KL/rejected_KL_mean": -37.447505950927734, "KL/std": 23.634626388549805, "epoch": 0.12481644640234948, "fcm_dpo/beta": 0.02158314362168312, "fcm_dpo/delta": -0.13427422940731049, "fcm_dpo/margin": 22.138042449951172, "fcm_dpo/q_t": 0.3926207423210144, "grad_norm": 24.075742721557617, "learning_rate": 4.992592445678582e-07, "logits/chosen": -0.6298732757568359, "logits/rejected": -0.5962429046630859, "logps/chosen": -73.71607971191406, "logps/ref_chosen": -58.4066162109375, "logps/ref_rejected": -78.63880157470703, "logps/rejected": -116.0863037109375, "loss": 1.0553, "margin_dpo/margin_mean": 22.138044357299805, "margin_dpo/margin_std": 28.41199493408203, "step": 85 }, { "KL/chosen_KL_mean": -19.839567184448242, "KL/mean": -31.57115364074707, "KL/rejected_KL_mean": -43.302734375, "KL/std": 32.33202362060547, "epoch": 0.1262848751835536, "fcm_dpo/beta": 0.020907826721668243, "fcm_dpo/delta": -0.14755286276340485, "fcm_dpo/margin": 23.463176727294922, "fcm_dpo/q_t": 0.4030860960483551, "grad_norm": 28.157058715820312, "learning_rate": 4.991572423079235e-07, "logits/chosen": -0.6940969824790955, "logits/rejected": -0.6837495565414429, "logps/chosen": -75.97703552246094, "logps/ref_chosen": -56.13746643066406, "logps/ref_rejected": -88.12165069580078, "logps/rejected": -131.42437744140625, "loss": 1.1321, "margin_dpo/margin_mean": 23.463176727294922, "margin_dpo/margin_std": 44.730472564697266, "step": 86 }, { "KL/chosen_KL_mean": -19.493261337280273, "KL/mean": -33.350990295410156, "KL/rejected_KL_mean": -47.20872116088867, "KL/std": 32.10251998901367, "epoch": 0.1277533039647577, "fcm_dpo/beta": 0.019912412390112877, "fcm_dpo/delta": -0.2153429239988327, "fcm_dpo/margin": 27.715457916259766, "fcm_dpo/q_t": 0.37868428230285645, "grad_norm": 24.050752639770508, "learning_rate": 4.990486745229364e-07, "logits/chosen": -0.7316626310348511, "logits/rejected": -0.7104564905166626, "logps/chosen": -75.12935638427734, "logps/ref_chosen": -55.63609313964844, "logps/ref_rejected": -95.46757507324219, "logps/rejected": -142.67630004882812, "loss": 1.0449, "margin_dpo/margin_mean": 27.715457916259766, "margin_dpo/margin_std": 39.20777893066406, "step": 87 }, { "KL/chosen_KL_mean": -24.775909423828125, "KL/mean": -35.22772216796875, "KL/rejected_KL_mean": -45.679527282714844, "KL/std": 33.75319290161133, "epoch": 0.12922173274596183, "fcm_dpo/beta": 0.019512329250574112, "fcm_dpo/delta": -0.061230212450027466, "fcm_dpo/margin": 20.903621673583984, "fcm_dpo/q_t": 0.41208845376968384, "grad_norm": 25.076162338256836, "learning_rate": 4.989335440737586e-07, "logits/chosen": -0.6718587279319763, "logits/rejected": -0.6672728061676025, "logps/chosen": -98.44705963134766, "logps/ref_chosen": -73.67115020751953, "logps/ref_rejected": -106.70849609375, "logps/rejected": -152.38803100585938, "loss": 1.1436, "margin_dpo/margin_mean": 20.903621673583984, "margin_dpo/margin_std": 35.89170837402344, "step": 88 }, { "KL/chosen_KL_mean": -15.323110580444336, "KL/mean": -27.0899715423584, "KL/rejected_KL_mean": -38.856834411621094, "KL/std": 28.511974334716797, "epoch": 0.13069016152716592, "fcm_dpo/beta": 0.01931881159543991, "fcm_dpo/delta": -0.10952495038509369, "fcm_dpo/margin": 23.533710479736328, "fcm_dpo/q_t": 0.39783668518066406, "grad_norm": 22.88800048828125, "learning_rate": 4.988118539941847e-07, "logits/chosen": -0.7127549648284912, "logits/rejected": -0.6814610958099365, "logps/chosen": -75.94802856445312, "logps/ref_chosen": -60.624916076660156, "logps/ref_rejected": -82.08354949951172, "logps/rejected": -120.94038391113281, "loss": 1.0703, "margin_dpo/margin_mean": 23.533714294433594, "margin_dpo/margin_std": 32.75722122192383, "step": 89 }, { "KL/chosen_KL_mean": -19.520963668823242, "KL/mean": -35.704227447509766, "KL/rejected_KL_mean": -51.88749313354492, "KL/std": 38.83220672607422, "epoch": 0.13215859030837004, "fcm_dpo/beta": 0.018566761165857315, "fcm_dpo/delta": -0.26629799604415894, "fcm_dpo/margin": 32.36653137207031, "fcm_dpo/q_t": 0.3788330554962158, "grad_norm": 23.77947235107422, "learning_rate": 4.986836074908615e-07, "logits/chosen": -0.656635046005249, "logits/rejected": -0.6720539331436157, "logps/chosen": -72.8062744140625, "logps/ref_chosen": -53.285308837890625, "logps/ref_rejected": -111.54470825195312, "logps/rejected": -163.4322052001953, "loss": 1.0449, "margin_dpo/margin_mean": 32.36653137207031, "margin_dpo/margin_std": 48.41737365722656, "step": 90 }, { "KL/chosen_KL_mean": -20.428979873657227, "KL/mean": -33.47758483886719, "KL/rejected_KL_mean": -46.52619171142578, "KL/std": 30.834447860717773, "epoch": 0.13362701908957417, "fcm_dpo/beta": 0.01787046156823635, "fcm_dpo/delta": -0.12188145518302917, "fcm_dpo/margin": 26.097206115722656, "fcm_dpo/q_t": 0.39908722043037415, "grad_norm": 22.826419830322266, "learning_rate": 4.985488079432037e-07, "logits/chosen": -0.6783984899520874, "logits/rejected": -0.646409273147583, "logps/chosen": -82.23193359375, "logps/ref_chosen": -61.802955627441406, "logps/ref_rejected": -87.87395477294922, "logps/rejected": -134.400146484375, "loss": 1.089, "margin_dpo/margin_mean": 26.09720230102539, "margin_dpo/margin_std": 40.38761901855469, "step": 91 }, { "KL/chosen_KL_mean": -18.27140235900879, "KL/mean": -31.61359977722168, "KL/rejected_KL_mean": -44.9557991027832, "KL/std": 31.74585723876953, "epoch": 0.13509544787077826, "fcm_dpo/beta": 0.017471997067332268, "fcm_dpo/delta": -0.12181757390499115, "fcm_dpo/margin": 26.68439483642578, "fcm_dpo/q_t": 0.39748257398605347, "grad_norm": 21.23597526550293, "learning_rate": 4.984074589033043e-07, "logits/chosen": -0.7334482669830322, "logits/rejected": -0.714074969291687, "logps/chosen": -69.91217041015625, "logps/ref_chosen": -51.640769958496094, "logps/ref_rejected": -77.88117980957031, "logps/rejected": -122.83698272705078, "loss": 1.076, "margin_dpo/margin_mean": 26.68439483642578, "margin_dpo/margin_std": 38.410037994384766, "step": 92 }, { "KL/chosen_KL_mean": -20.64739227294922, "KL/mean": -33.98801803588867, "KL/rejected_KL_mean": -47.32863998413086, "KL/std": 29.834144592285156, "epoch": 0.13656387665198239, "fcm_dpo/beta": 0.017019839957356453, "fcm_dpo/delta": -0.10895700007677078, "fcm_dpo/margin": 26.681251525878906, "fcm_dpo/q_t": 0.3978961706161499, "grad_norm": 22.11734390258789, "learning_rate": 4.982595640958425e-07, "logits/chosen": -0.7326809167861938, "logits/rejected": -0.6875460147857666, "logps/chosen": -73.1766357421875, "logps/ref_chosen": -52.529239654541016, "logps/ref_rejected": -77.16075134277344, "logps/rejected": -124.48938751220703, "loss": 1.0595, "margin_dpo/margin_mean": 26.681249618530273, "margin_dpo/margin_std": 34.84459686279297, "step": 93 }, { "KL/chosen_KL_mean": -22.205469131469727, "KL/mean": -37.707279205322266, "KL/rejected_KL_mean": -53.20908737182617, "KL/std": 34.52550506591797, "epoch": 0.13803230543318648, "fcm_dpo/beta": 0.016398325562477112, "fcm_dpo/delta": -0.1683567464351654, "fcm_dpo/margin": 31.003623962402344, "fcm_dpo/q_t": 0.3850182890892029, "grad_norm": 21.626747131347656, "learning_rate": 4.98105127417984e-07, "logits/chosen": -0.6754692792892456, "logits/rejected": -0.6637367010116577, "logps/chosen": -83.42808532714844, "logps/ref_chosen": -61.22261047363281, "logps/ref_rejected": -99.59902954101562, "logps/rejected": -152.80812072753906, "loss": 1.0229, "margin_dpo/margin_mean": 31.00362205505371, "margin_dpo/margin_std": 36.4665641784668, "step": 94 }, { "KL/chosen_KL_mean": -21.489198684692383, "KL/mean": -33.98937225341797, "KL/rejected_KL_mean": -46.48954772949219, "KL/std": 33.66114807128906, "epoch": 0.1395007342143906, "fcm_dpo/beta": 0.016146240755915642, "fcm_dpo/delta": -0.05641428753733635, "fcm_dpo/margin": 25.000350952148438, "fcm_dpo/q_t": 0.40661683678627014, "grad_norm": 20.469913482666016, "learning_rate": 4.979441529392784e-07, "logits/chosen": -0.6882494688034058, "logits/rejected": -0.6622629761695862, "logps/chosen": -74.01284790039062, "logps/ref_chosen": -52.523643493652344, "logps/ref_rejected": -75.8803482055664, "logps/rejected": -122.3698959350586, "loss": 1.0993, "margin_dpo/margin_mean": 25.000349044799805, "margin_dpo/margin_std": 35.28263854980469, "step": 95 }, { "KL/chosen_KL_mean": -20.916810989379883, "KL/mean": -37.918636322021484, "KL/rejected_KL_mean": -54.92045593261719, "KL/std": 37.27104187011719, "epoch": 0.14096916299559473, "fcm_dpo/beta": 0.015592942014336586, "fcm_dpo/delta": -0.19270357489585876, "fcm_dpo/margin": 34.0036506652832, "fcm_dpo/q_t": 0.38242608308792114, "grad_norm": 20.710969924926758, "learning_rate": 4.977766449015534e-07, "logits/chosen": -0.707472562789917, "logits/rejected": -0.6825254559516907, "logps/chosen": -83.07378387451172, "logps/ref_chosen": -62.15697479248047, "logps/ref_rejected": -96.59601593017578, "logps/rejected": -151.5164794921875, "loss": 1.0147, "margin_dpo/margin_mean": 34.0036506652832, "margin_dpo/margin_std": 41.2987060546875, "step": 96 }, { "KL/chosen_KL_mean": -22.15590476989746, "KL/mean": -35.208744049072266, "KL/rejected_KL_mean": -48.26157760620117, "KL/std": 31.424772262573242, "epoch": 0.14243759177679882, "fcm_dpo/beta": 0.015535632148385048, "fcm_dpo/delta": -0.05870715156197548, "fcm_dpo/margin": 26.10567855834961, "fcm_dpo/q_t": 0.40512728691101074, "grad_norm": 21.590152740478516, "learning_rate": 4.976026077188012e-07, "logits/chosen": -0.5965183973312378, "logits/rejected": -0.5535776019096375, "logps/chosen": -76.8022689819336, "logps/ref_chosen": -54.646366119384766, "logps/ref_rejected": -76.96475219726562, "logps/rejected": -125.22633361816406, "loss": 1.0831, "margin_dpo/margin_mean": 26.105682373046875, "margin_dpo/margin_std": 31.313583374023438, "step": 97 }, { "KL/chosen_KL_mean": -27.724029541015625, "KL/mean": -42.68238830566406, "KL/rejected_KL_mean": -57.64073944091797, "KL/std": 36.01885986328125, "epoch": 0.14390602055800295, "fcm_dpo/beta": 0.015098905190825462, "fcm_dpo/delta": -0.10677133500576019, "fcm_dpo/margin": 29.916717529296875, "fcm_dpo/q_t": 0.39555469155311584, "grad_norm": 22.497968673706055, "learning_rate": 4.974220459770639e-07, "logits/chosen": -0.6522549390792847, "logits/rejected": -0.6376872062683105, "logps/chosen": -92.98265838623047, "logps/ref_chosen": -65.25862884521484, "logps/ref_rejected": -96.5274887084961, "logps/rejected": -154.16822814941406, "loss": 1.0859, "margin_dpo/margin_mean": 29.916717529296875, "margin_dpo/margin_std": 43.398094177246094, "step": 98 }, { "KL/chosen_KL_mean": -21.62339210510254, "KL/mean": -40.180118560791016, "KL/rejected_KL_mean": -58.736846923828125, "KL/std": 38.713218688964844, "epoch": 0.14537444933920704, "fcm_dpo/beta": 0.014553403481841087, "fcm_dpo/delta": -0.20181076228618622, "fcm_dpo/margin": 37.11345672607422, "fcm_dpo/q_t": 0.3828354477882385, "grad_norm": 19.82951545715332, "learning_rate": 4.972349644343108e-07, "logits/chosen": -0.6461166739463806, "logits/rejected": -0.6486799716949463, "logps/chosen": -67.26187896728516, "logps/ref_chosen": -45.638484954833984, "logps/ref_rejected": -86.43793487548828, "logps/rejected": -145.17477416992188, "loss": 1.0199, "margin_dpo/margin_mean": 37.11345672607422, "margin_dpo/margin_std": 48.379791259765625, "step": 99 }, { "KL/chosen_KL_mean": -24.61417579650879, "KL/mean": -35.05047607421875, "KL/rejected_KL_mean": -45.48677444458008, "KL/std": 31.448835372924805, "epoch": 0.14684287812041116, "fcm_dpo/beta": 0.01442507840692997, "fcm_dpo/delta": -0.015769226476550102, "fcm_dpo/margin": 20.872608184814453, "fcm_dpo/q_t": 0.43076425790786743, "grad_norm": 21.535491943359375, "learning_rate": 4.970413680203148e-07, "logits/chosen": -0.6545113325119019, "logits/rejected": -0.6143167018890381, "logps/chosen": -82.20816040039062, "logps/ref_chosen": -57.59397888183594, "logps/ref_rejected": -74.06021118164062, "logps/rejected": -119.54698181152344, "loss": 1.1936, "margin_dpo/margin_mean": 20.872608184814453, "margin_dpo/margin_std": 40.602088928222656, "step": 100 }, { "KL/chosen_KL_mean": -30.125629425048828, "KL/mean": -43.047637939453125, "KL/rejected_KL_mean": -55.96964645385742, "KL/std": 38.88969421386719, "epoch": 0.14831130690161526, "fcm_dpo/beta": 0.014283875934779644, "fcm_dpo/delta": -0.020545855164527893, "fcm_dpo/margin": 25.844009399414062, "fcm_dpo/q_t": 0.41966480016708374, "grad_norm": 21.14192771911621, "learning_rate": 4.968412618365215e-07, "logits/chosen": -0.6575559377670288, "logits/rejected": -0.6289372444152832, "logps/chosen": -91.77448272705078, "logps/ref_chosen": -61.64885330200195, "logps/ref_rejected": -83.18968200683594, "logps/rejected": -139.15933227539062, "loss": 1.1538, "margin_dpo/margin_mean": 25.844009399414062, "margin_dpo/margin_std": 46.674888610839844, "step": 101 }, { "KL/chosen_KL_mean": -32.28032684326172, "KL/mean": -41.69635009765625, "KL/rejected_KL_mean": -51.112369537353516, "KL/std": 36.33782958984375, "epoch": 0.14977973568281938, "fcm_dpo/beta": 0.014304671436548233, "fcm_dpo/delta": -0.026983171701431274, "fcm_dpo/margin": 18.832042694091797, "fcm_dpo/q_t": 0.4389786124229431, "grad_norm": 23.982179641723633, "learning_rate": 4.966346511559149e-07, "logits/chosen": -0.6497610211372375, "logits/rejected": -0.6050263047218323, "logps/chosen": -96.35920715332031, "logps/ref_chosen": -64.0788803100586, "logps/ref_rejected": -68.18707275390625, "logps/rejected": -119.29944610595703, "loss": 1.2298, "margin_dpo/margin_mean": 18.832042694091797, "margin_dpo/margin_std": 43.388916015625, "step": 102 }, { "KL/chosen_KL_mean": -26.217662811279297, "KL/mean": -45.513275146484375, "KL/rejected_KL_mean": -64.80888366699219, "KL/std": 40.776275634765625, "epoch": 0.1512481644640235, "fcm_dpo/beta": 0.013833096250891685, "fcm_dpo/delta": -0.1946529746055603, "fcm_dpo/margin": 38.59122085571289, "fcm_dpo/q_t": 0.38289010524749756, "grad_norm": 20.647781372070312, "learning_rate": 4.964215414228785e-07, "logits/chosen": -0.678255021572113, "logits/rejected": -0.6474459767341614, "logps/chosen": -87.51693725585938, "logps/ref_chosen": -61.299278259277344, "logps/ref_rejected": -93.57270812988281, "logps/rejected": -158.381591796875, "loss": 1.0221, "margin_dpo/margin_mean": 38.591217041015625, "margin_dpo/margin_std": 49.838680267333984, "step": 103 }, { "KL/chosen_KL_mean": -28.221818923950195, "KL/mean": -46.5230712890625, "KL/rejected_KL_mean": -64.8243179321289, "KL/std": 44.71335983276367, "epoch": 0.1527165932452276, "fcm_dpo/beta": 0.013438082300126553, "fcm_dpo/delta": -0.1490502506494522, "fcm_dpo/margin": 36.602508544921875, "fcm_dpo/q_t": 0.3954700827598572, "grad_norm": 20.52446937561035, "learning_rate": 4.96201938253052e-07, "logits/chosen": -0.6646419763565063, "logits/rejected": -0.6363682746887207, "logps/chosen": -82.59458923339844, "logps/ref_chosen": -54.372772216796875, "logps/ref_rejected": -89.5647201538086, "logps/rejected": -154.3890380859375, "loss": 1.0727, "margin_dpo/margin_mean": 36.602508544921875, "margin_dpo/margin_std": 54.992401123046875, "step": 104 }, { "KL/chosen_KL_mean": -26.766700744628906, "KL/mean": -52.79180908203125, "KL/rejected_KL_mean": -78.81692504882812, "KL/std": 43.03718185424805, "epoch": 0.15418502202643172, "fcm_dpo/beta": 0.012692511081695557, "fcm_dpo/delta": -0.3323138356208801, "fcm_dpo/margin": 52.05022430419922, "fcm_dpo/q_t": 0.348787784576416, "grad_norm": 20.9357852935791, "learning_rate": 4.959758474331832e-07, "logits/chosen": -0.6385953426361084, "logits/rejected": -0.6172483563423157, "logps/chosen": -81.40564727783203, "logps/ref_chosen": -54.638946533203125, "logps/ref_rejected": -97.97351837158203, "logps/rejected": -176.79043579101562, "loss": 0.9122, "margin_dpo/margin_mean": 52.05022430419922, "margin_dpo/margin_std": 47.93716049194336, "step": 105 }, { "KL/chosen_KL_mean": -30.216970443725586, "KL/mean": -46.443695068359375, "KL/rejected_KL_mean": -62.67041778564453, "KL/std": 37.993072509765625, "epoch": 0.15565345080763582, "fcm_dpo/beta": 0.012338995933532715, "fcm_dpo/delta": -0.052742164582014084, "fcm_dpo/margin": 32.45343780517578, "fcm_dpo/q_t": 0.40674251317977905, "grad_norm": 19.48843765258789, "learning_rate": 4.957432749209755e-07, "logits/chosen": -0.6057268381118774, "logits/rejected": -0.5791513919830322, "logps/chosen": -85.04986572265625, "logps/ref_chosen": -54.83289337158203, "logps/ref_rejected": -85.22461700439453, "logps/rejected": -147.89503479003906, "loss": 1.0871, "margin_dpo/margin_mean": 32.45343780517578, "margin_dpo/margin_std": 41.20294189453125, "step": 106 }, { "KL/chosen_KL_mean": -35.853370666503906, "KL/mean": -54.165565490722656, "KL/rejected_KL_mean": -72.4777603149414, "KL/std": 46.64228439331055, "epoch": 0.15712187958883994, "fcm_dpo/beta": 0.01210312731564045, "fcm_dpo/delta": -0.09832476079463959, "fcm_dpo/margin": 36.62439727783203, "fcm_dpo/q_t": 0.3984289765357971, "grad_norm": 19.72745132446289, "learning_rate": 4.955042268449307e-07, "logits/chosen": -0.6482241153717041, "logits/rejected": -0.6045652627944946, "logps/chosen": -105.56118774414062, "logps/ref_chosen": -69.70780944824219, "logps/ref_rejected": -94.73950958251953, "logps/rejected": -167.21726989746094, "loss": 1.0757, "margin_dpo/margin_mean": 36.62439727783203, "margin_dpo/margin_std": 49.4898681640625, "step": 107 }, { "KL/chosen_KL_mean": -31.727853775024414, "KL/mean": -54.42365646362305, "KL/rejected_KL_mean": -77.11946105957031, "KL/std": 51.947845458984375, "epoch": 0.15859030837004406, "fcm_dpo/beta": 0.011752867139875889, "fcm_dpo/delta": -0.193751722574234, "fcm_dpo/margin": 45.3916130065918, "fcm_dpo/q_t": 0.38766244053840637, "grad_norm": 19.59799575805664, "learning_rate": 4.952587095041881e-07, "logits/chosen": -0.6367114782333374, "logits/rejected": -0.6192809343338013, "logps/chosen": -87.73773956298828, "logps/ref_chosen": -56.0098876953125, "logps/ref_rejected": -95.79601287841797, "logps/rejected": -172.91546630859375, "loss": 1.0566, "margin_dpo/margin_mean": 45.39160919189453, "margin_dpo/margin_std": 67.23835754394531, "step": 108 }, { "KL/chosen_KL_mean": -29.637109756469727, "KL/mean": -53.46610641479492, "KL/rejected_KL_mean": -77.29510498046875, "KL/std": 49.2529182434082, "epoch": 0.16005873715124816, "fcm_dpo/beta": 0.01125399675220251, "fcm_dpo/delta": -0.19788390398025513, "fcm_dpo/margin": 47.657989501953125, "fcm_dpo/q_t": 0.379015177488327, "grad_norm": 20.628564834594727, "learning_rate": 4.95006729368358e-07, "logits/chosen": -0.5834232568740845, "logits/rejected": -0.5629913806915283, "logps/chosen": -92.52259826660156, "logps/ref_chosen": -62.88549041748047, "logps/ref_rejected": -98.68573760986328, "logps/rejected": -175.9808349609375, "loss": 1.0206, "margin_dpo/margin_mean": 47.65799331665039, "margin_dpo/margin_std": 59.39226531982422, "step": 109 }, { "KL/chosen_KL_mean": -31.1754150390625, "KL/mean": -52.55781555175781, "KL/rejected_KL_mean": -73.94021606445312, "KL/std": 49.422637939453125, "epoch": 0.16152716593245228, "fcm_dpo/beta": 0.01084812544286251, "fcm_dpo/delta": -0.12373995780944824, "fcm_dpo/margin": 42.764793395996094, "fcm_dpo/q_t": 0.3965584337711334, "grad_norm": 17.69667625427246, "learning_rate": 4.947482930773511e-07, "logits/chosen": -0.5745487213134766, "logits/rejected": -0.5401524305343628, "logps/chosen": -89.9291000366211, "logps/ref_chosen": -58.753684997558594, "logps/ref_rejected": -79.75001525878906, "logps/rejected": -153.6902313232422, "loss": 1.0805, "margin_dpo/margin_mean": 42.764793395996094, "margin_dpo/margin_std": 59.920650482177734, "step": 110 }, { "KL/chosen_KL_mean": -35.113006591796875, "KL/mean": -59.724143981933594, "KL/rejected_KL_mean": -84.33528137207031, "KL/std": 56.133522033691406, "epoch": 0.16299559471365638, "fcm_dpo/beta": 0.010561220347881317, "fcm_dpo/delta": -0.18071341514587402, "fcm_dpo/margin": 49.22227478027344, "fcm_dpo/q_t": 0.38506948947906494, "grad_norm": 19.484743118286133, "learning_rate": 4.944834074412042e-07, "logits/chosen": -0.6309810876846313, "logits/rejected": -0.6101324558258057, "logps/chosen": -103.73710632324219, "logps/ref_chosen": -68.62410736083984, "logps/ref_rejected": -98.42886352539062, "logps/rejected": -182.76414489746094, "loss": 1.0576, "margin_dpo/margin_mean": 49.22227478027344, "margin_dpo/margin_std": 70.08132934570312, "step": 111 }, { "KL/chosen_KL_mean": -32.352989196777344, "KL/mean": -47.761932373046875, "KL/rejected_KL_mean": -63.17087936401367, "KL/std": 38.15313720703125, "epoch": 0.1644640234948605, "fcm_dpo/beta": 0.010518467053771019, "fcm_dpo/delta": 0.02647808939218521, "fcm_dpo/margin": 30.817890167236328, "fcm_dpo/q_t": 0.4247877597808838, "grad_norm": 18.049718856811523, "learning_rate": 4.942120794399002e-07, "logits/chosen": -0.5829577445983887, "logits/rejected": -0.5457116365432739, "logps/chosen": -82.60263061523438, "logps/ref_chosen": -50.24964141845703, "logps/ref_rejected": -64.77442932128906, "logps/rejected": -127.9453125, "loss": 1.1474, "margin_dpo/margin_mean": 30.817893981933594, "margin_dpo/margin_std": 46.761627197265625, "step": 112 }, { "KL/chosen_KL_mean": -41.24876403808594, "KL/mean": -57.47443389892578, "KL/rejected_KL_mean": -73.70010375976562, "KL/std": 40.94397735595703, "epoch": 0.16593245227606462, "fcm_dpo/beta": 0.010588676668703556, "fcm_dpo/delta": 0.00660739466547966, "fcm_dpo/margin": 32.45133972167969, "fcm_dpo/q_t": 0.4201717972755432, "grad_norm": 18.475440979003906, "learning_rate": 4.939343162231841e-07, "logits/chosen": -0.5695180892944336, "logits/rejected": -0.5281089544296265, "logps/chosen": -107.96171569824219, "logps/ref_chosen": -66.71295166015625, "logps/ref_rejected": -77.96870422363281, "logps/rejected": -151.66880798339844, "loss": 1.126, "margin_dpo/margin_mean": 32.45133972167969, "margin_dpo/margin_std": 45.22571563720703, "step": 113 }, { "KL/chosen_KL_mean": -37.32588195800781, "KL/mean": -63.44007873535156, "KL/rejected_KL_mean": -89.55429077148438, "KL/std": 60.05897903442383, "epoch": 0.16740088105726872, "fcm_dpo/beta": 0.010224482975900173, "fcm_dpo/delta": -0.19627024233341217, "fcm_dpo/margin": 52.22840881347656, "fcm_dpo/q_t": 0.38643473386764526, "grad_norm": 20.448719024658203, "learning_rate": 4.936501251103751e-07, "logits/chosen": -0.5776060819625854, "logits/rejected": -0.5475910902023315, "logps/chosen": -95.1109619140625, "logps/ref_chosen": -57.78507995605469, "logps/ref_rejected": -87.10966491699219, "logps/rejected": -176.6639404296875, "loss": 1.0341, "margin_dpo/margin_mean": 52.2284049987793, "margin_dpo/margin_std": 73.12133026123047, "step": 114 }, { "KL/chosen_KL_mean": -48.408294677734375, "KL/mean": -66.1071548461914, "KL/rejected_KL_mean": -83.80602264404297, "KL/std": 57.53717803955078, "epoch": 0.16886930983847284, "fcm_dpo/beta": 0.010193654336035252, "fcm_dpo/delta": -0.011369943618774414, "fcm_dpo/margin": 35.397727966308594, "fcm_dpo/q_t": 0.42307719588279724, "grad_norm": 24.033288955688477, "learning_rate": 4.933595135901732e-07, "logits/chosen": -0.6035769581794739, "logits/rejected": -0.587733268737793, "logps/chosen": -113.99093627929688, "logps/ref_chosen": -65.5826416015625, "logps/ref_rejected": -98.56552124023438, "logps/rejected": -182.37155151367188, "loss": 1.1912, "margin_dpo/margin_mean": 35.397727966308594, "margin_dpo/margin_std": 76.34100341796875, "step": 115 }, { "KL/chosen_KL_mean": -35.714500427246094, "KL/mean": -55.78123474121094, "KL/rejected_KL_mean": -75.84798431396484, "KL/std": 48.34644317626953, "epoch": 0.17033773861967694, "fcm_dpo/beta": 0.01011539064347744, "fcm_dpo/delta": -0.05847224220633507, "fcm_dpo/margin": 40.13346862792969, "fcm_dpo/q_t": 0.4071720540523529, "grad_norm": 19.61532974243164, "learning_rate": 4.930624893204624e-07, "logits/chosen": -0.5785757899284363, "logits/rejected": -0.5719596147537231, "logps/chosen": -87.11480712890625, "logps/ref_chosen": -51.40031433105469, "logps/ref_rejected": -80.5218505859375, "logps/rejected": -156.36984252929688, "loss": 1.0872, "margin_dpo/margin_mean": 40.13346862792969, "margin_dpo/margin_std": 52.50077819824219, "step": 116 }, { "KL/chosen_KL_mean": -46.457366943359375, "KL/mean": -62.5472526550293, "KL/rejected_KL_mean": -78.63714599609375, "KL/std": 51.26929473876953, "epoch": 0.17180616740088106, "fcm_dpo/beta": 0.010081902146339417, "fcm_dpo/delta": 0.026419375091791153, "fcm_dpo/margin": 32.179771423339844, "fcm_dpo/q_t": 0.4264397621154785, "grad_norm": 24.782325744628906, "learning_rate": 4.927590601281083e-07, "logits/chosen": -0.5525811910629272, "logits/rejected": -0.5172309875488281, "logps/chosen": -115.75577545166016, "logps/ref_chosen": -69.29840850830078, "logps/ref_rejected": -66.583984375, "logps/rejected": -145.22113037109375, "loss": 1.1688, "margin_dpo/margin_mean": 32.17976760864258, "margin_dpo/margin_std": 58.62355041503906, "step": 117 }, { "KL/chosen_KL_mean": -36.07121276855469, "KL/mean": -55.96504211425781, "KL/rejected_KL_mean": -75.85887145996094, "KL/std": 46.31214904785156, "epoch": 0.17327459618208516, "fcm_dpo/beta": 0.010025271214544773, "fcm_dpo/delta": -0.050886109471321106, "fcm_dpo/margin": 39.787662506103516, "fcm_dpo/q_t": 0.40897485613822937, "grad_norm": 19.208255767822266, "learning_rate": 4.924492340087524e-07, "logits/chosen": -0.6033087968826294, "logits/rejected": -0.5862714052200317, "logps/chosen": -91.71218872070312, "logps/ref_chosen": -55.6409797668457, "logps/ref_rejected": -75.66905975341797, "logps/rejected": -151.52792358398438, "loss": 1.0925, "margin_dpo/margin_mean": 39.78766632080078, "margin_dpo/margin_std": 53.461585998535156, "step": 118 }, { "KL/chosen_KL_mean": -49.75245666503906, "KL/mean": -69.50601196289062, "KL/rejected_KL_mean": -89.25955200195312, "KL/std": 52.264060974121094, "epoch": 0.17474302496328928, "fcm_dpo/beta": 0.009865409694612026, "fcm_dpo/delta": -0.042472440749406815, "fcm_dpo/margin": 39.50709533691406, "fcm_dpo/q_t": 0.413967102766037, "grad_norm": 20.008079528808594, "learning_rate": 4.92133019126601e-07, "logits/chosen": -0.5890421271324158, "logits/rejected": -0.5784463882446289, "logps/chosen": -123.26264953613281, "logps/ref_chosen": -73.51019287109375, "logps/ref_rejected": -102.977294921875, "logps/rejected": -192.23684692382812, "loss": 1.1279, "margin_dpo/margin_mean": 39.50709533691406, "margin_dpo/margin_std": 64.39222717285156, "step": 119 }, { "KL/chosen_KL_mean": -50.61544418334961, "KL/mean": -77.9512939453125, "KL/rejected_KL_mean": -105.28715515136719, "KL/std": 65.06802368164062, "epoch": 0.1762114537444934, "fcm_dpo/beta": 0.00962957926094532, "fcm_dpo/delta": -0.18642207980155945, "fcm_dpo/margin": 54.67171096801758, "fcm_dpo/q_t": 0.382351815700531, "grad_norm": 20.747541427612305, "learning_rate": 4.918104238142103e-07, "logits/chosen": -0.6031701564788818, "logits/rejected": -0.575164794921875, "logps/chosen": -127.39627838134766, "logps/ref_chosen": -76.78083801269531, "logps/ref_rejected": -108.02374267578125, "logps/rejected": -213.31089782714844, "loss": 1.0242, "margin_dpo/margin_mean": 54.67170715332031, "margin_dpo/margin_std": 68.70195007324219, "step": 120 }, { "KL/chosen_KL_mean": -49.010231018066406, "KL/mean": -79.25602722167969, "KL/rejected_KL_mean": -109.5018310546875, "KL/std": 63.33201217651367, "epoch": 0.1776798825256975, "fcm_dpo/beta": 0.009193992242217064, "fcm_dpo/delta": -0.21982212364673615, "fcm_dpo/margin": 60.49159240722656, "fcm_dpo/q_t": 0.3774503469467163, "grad_norm": 23.35782814025879, "learning_rate": 4.91481456572267e-07, "logits/chosen": -0.5464009642601013, "logits/rejected": -0.543734073638916, "logps/chosen": -110.80012512207031, "logps/ref_chosen": -61.789894104003906, "logps/ref_rejected": -109.99456787109375, "logps/rejected": -219.49639892578125, "loss": 1.0224, "margin_dpo/margin_mean": 60.49159240722656, "margin_dpo/margin_std": 79.29399108886719, "step": 121 }, { "KL/chosen_KL_mean": -44.47397232055664, "KL/mean": -81.9617919921875, "KL/rejected_KL_mean": -119.44963073730469, "KL/std": 71.8316650390625, "epoch": 0.17914831130690162, "fcm_dpo/beta": 0.008711527101695538, "fcm_dpo/delta": -0.32415759563446045, "fcm_dpo/margin": 74.97564697265625, "fcm_dpo/q_t": 0.3540286719799042, "grad_norm": 21.536544799804688, "learning_rate": 4.911461260693638e-07, "logits/chosen": -0.535602331161499, "logits/rejected": -0.5521268844604492, "logps/chosen": -91.37618255615234, "logps/ref_chosen": -46.9022102355957, "logps/ref_rejected": -106.71418762207031, "logps/rejected": -226.163818359375, "loss": 0.9307, "margin_dpo/margin_mean": 74.97564697265625, "margin_dpo/margin_std": 76.03938293457031, "step": 122 }, { "KL/chosen_KL_mean": -49.342002868652344, "KL/mean": -73.47360229492188, "KL/rejected_KL_mean": -97.60520935058594, "KL/std": 60.79751205444336, "epoch": 0.18061674008810572, "fcm_dpo/beta": 0.008437592536211014, "fcm_dpo/delta": -0.06059284880757332, "fcm_dpo/margin": 48.26319122314453, "fcm_dpo/q_t": 0.410540372133255, "grad_norm": 19.794395446777344, "learning_rate": 4.908044411417711e-07, "logits/chosen": -0.5162737369537354, "logits/rejected": -0.5014143586158752, "logps/chosen": -110.68064880371094, "logps/ref_chosen": -61.33863830566406, "logps/ref_rejected": -87.775390625, "logps/rejected": -185.38059997558594, "loss": 1.1306, "margin_dpo/margin_mean": 48.263187408447266, "margin_dpo/margin_std": 82.03060913085938, "step": 123 }, { "KL/chosen_KL_mean": -57.09165954589844, "KL/mean": -93.51963806152344, "KL/rejected_KL_mean": -129.94761657714844, "KL/std": 84.41590881347656, "epoch": 0.18208516886930984, "fcm_dpo/beta": 0.008119284175336361, "fcm_dpo/delta": -0.2575443387031555, "fcm_dpo/margin": 72.85594177246094, "fcm_dpo/q_t": 0.37762200832366943, "grad_norm": 19.897216796875, "learning_rate": 4.904564107932048e-07, "logits/chosen": -0.5154942274093628, "logits/rejected": -0.5195925235748291, "logps/chosen": -128.5399932861328, "logps/ref_chosen": -71.44833374023438, "logps/ref_rejected": -117.58056640625, "logps/rejected": -247.52818298339844, "loss": 1.0371, "margin_dpo/margin_mean": 72.85594177246094, "margin_dpo/margin_std": 106.19161987304688, "step": 124 }, { "KL/chosen_KL_mean": -46.80385208129883, "KL/mean": -78.55594635009766, "KL/rejected_KL_mean": -110.30804443359375, "KL/std": 71.42005157470703, "epoch": 0.18355359765051396, "fcm_dpo/beta": 0.007860703393816948, "fcm_dpo/delta": -0.1568271368741989, "fcm_dpo/margin": 63.504180908203125, "fcm_dpo/q_t": 0.3909478783607483, "grad_norm": 17.979494094848633, "learning_rate": 4.90102044194588e-07, "logits/chosen": -0.47995638847351074, "logits/rejected": -0.48471179604530334, "logps/chosen": -96.9407958984375, "logps/ref_chosen": -50.136940002441406, "logps/ref_rejected": -83.98861694335938, "logps/rejected": -194.29666137695312, "loss": 1.0597, "margin_dpo/margin_mean": 63.504180908203125, "margin_dpo/margin_std": 89.72810363769531, "step": 125 }, { "KL/chosen_KL_mean": -50.658485412597656, "KL/mean": -80.66192626953125, "KL/rejected_KL_mean": -110.66535949707031, "KL/std": 64.82109832763672, "epoch": 0.18502202643171806, "fcm_dpo/beta": 0.007639255840331316, "fcm_dpo/delta": -0.1139136403799057, "fcm_dpo/margin": 60.00688552856445, "fcm_dpo/q_t": 0.39801329374313354, "grad_norm": 19.024023056030273, "learning_rate": 4.897413506838102e-07, "logits/chosen": -0.48899418115615845, "logits/rejected": -0.4869902729988098, "logps/chosen": -106.32555389404297, "logps/ref_chosen": -55.66706848144531, "logps/ref_rejected": -98.1297607421875, "logps/rejected": -208.7951202392578, "loss": 1.0697, "margin_dpo/margin_mean": 60.00688552856445, "margin_dpo/margin_std": 83.00294494628906, "step": 126 }, { "KL/chosen_KL_mean": -49.5046272277832, "KL/mean": -70.91242980957031, "KL/rejected_KL_mean": -92.32023620605469, "KL/std": 56.90325164794922, "epoch": 0.18649045521292218, "fcm_dpo/beta": 0.007639319635927677, "fcm_dpo/delta": 0.02368098497390747, "fcm_dpo/margin": 42.81562042236328, "fcm_dpo/q_t": 0.42351189255714417, "grad_norm": 18.42085075378418, "learning_rate": 4.89374339765481e-07, "logits/chosen": -0.5007494688034058, "logits/rejected": -0.48394691944122314, "logps/chosen": -106.05931091308594, "logps/ref_chosen": -56.55467987060547, "logps/ref_rejected": -76.7957763671875, "logps/rejected": -169.11602783203125, "loss": 1.1575, "margin_dpo/margin_mean": 42.815616607666016, "margin_dpo/margin_std": 71.940673828125, "step": 127 }, { "KL/chosen_KL_mean": -52.58698272705078, "KL/mean": -75.38080596923828, "KL/rejected_KL_mean": -98.17463684082031, "KL/std": 66.4472885131836, "epoch": 0.18795888399412627, "fcm_dpo/beta": 0.007676831912249327, "fcm_dpo/delta": -0.000309869647026062, "fcm_dpo/margin": 45.58763885498047, "fcm_dpo/q_t": 0.4211137592792511, "grad_norm": 22.759822845458984, "learning_rate": 4.890010211106795e-07, "logits/chosen": -0.46261486411094666, "logits/rejected": -0.44473135471343994, "logps/chosen": -110.70793914794922, "logps/ref_chosen": -58.12095642089844, "logps/ref_rejected": -76.43896484375, "logps/rejected": -174.61358642578125, "loss": 1.1643, "margin_dpo/margin_mean": 45.587646484375, "margin_dpo/margin_std": 83.7347183227539, "step": 128 }, { "KL/chosen_KL_mean": -63.59666442871094, "KL/mean": -87.52311706542969, "KL/rejected_KL_mean": -111.44956970214844, "KL/std": 74.28958892822266, "epoch": 0.1894273127753304, "fcm_dpo/beta": 0.007631917484104633, "fcm_dpo/delta": -0.015783965587615967, "fcm_dpo/margin": 47.8529052734375, "fcm_dpo/q_t": 0.42259424924850464, "grad_norm": 18.812211990356445, "learning_rate": 4.88621404556699e-07, "logits/chosen": -0.4808692932128906, "logits/rejected": -0.47164222598075867, "logps/chosen": -130.51303100585938, "logps/ref_chosen": -66.91637420654297, "logps/ref_rejected": -96.6422119140625, "logps/rejected": -208.09178161621094, "loss": 1.1733, "margin_dpo/margin_mean": 47.8529052734375, "margin_dpo/margin_std": 97.08119201660156, "step": 129 }, { "KL/chosen_KL_mean": -46.903900146484375, "KL/mean": -83.26931762695312, "KL/rejected_KL_mean": -119.6347427368164, "KL/std": 74.58981323242188, "epoch": 0.19089574155653452, "fcm_dpo/beta": 0.007464664056897163, "fcm_dpo/delta": -0.20387914776802063, "fcm_dpo/margin": 72.7308349609375, "fcm_dpo/q_t": 0.381303608417511, "grad_norm": 19.38671112060547, "learning_rate": 4.882355001067891e-07, "logits/chosen": -0.4573250412940979, "logits/rejected": -0.45176631212234497, "logps/chosen": -91.57075500488281, "logps/ref_chosen": -44.66685104370117, "logps/ref_rejected": -82.78165435791016, "logps/rejected": -202.41639709472656, "loss": 1.0269, "margin_dpo/margin_mean": 72.7308349609375, "margin_dpo/margin_std": 91.81611633300781, "step": 130 }, { "KL/chosen_KL_mean": -42.863670349121094, "KL/mean": -79.78742218017578, "KL/rejected_KL_mean": -116.711181640625, "KL/std": 76.12089538574219, "epoch": 0.19236417033773862, "fcm_dpo/beta": 0.007108722813427448, "fcm_dpo/delta": -0.18494439125061035, "fcm_dpo/margin": 73.84750366210938, "fcm_dpo/q_t": 0.3789236545562744, "grad_norm": 25.050138473510742, "learning_rate": 4.878433179298909e-07, "logits/chosen": -0.41149088740348816, "logits/rejected": -0.4173048436641693, "logps/chosen": -87.78825378417969, "logps/ref_chosen": -44.924591064453125, "logps/ref_rejected": -88.44401550292969, "logps/rejected": -205.15518188476562, "loss": 1.0083, "margin_dpo/margin_mean": 73.84750366210938, "margin_dpo/margin_std": 82.75117492675781, "step": 131 }, { "KL/chosen_KL_mean": -56.301605224609375, "KL/mean": -86.18394470214844, "KL/rejected_KL_mean": -116.06629943847656, "KL/std": 74.9560546875, "epoch": 0.19383259911894274, "fcm_dpo/beta": 0.006964902393519878, "fcm_dpo/delta": -0.06934472173452377, "fcm_dpo/margin": 59.764686584472656, "fcm_dpo/q_t": 0.41001617908477783, "grad_norm": 18.39053726196289, "learning_rate": 4.874448683603694e-07, "logits/chosen": -0.4645339548587799, "logits/rejected": -0.46584564447402954, "logps/chosen": -115.30268859863281, "logps/ref_chosen": -59.00108337402344, "logps/ref_rejected": -87.89215087890625, "logps/rejected": -203.9584503173828, "loss": 1.1137, "margin_dpo/margin_mean": 59.76469421386719, "margin_dpo/margin_std": 97.89826965332031, "step": 132 }, { "KL/chosen_KL_mean": -66.9070053100586, "KL/mean": -93.52379608154297, "KL/rejected_KL_mean": -120.14059448242188, "KL/std": 67.77723693847656, "epoch": 0.19530102790014683, "fcm_dpo/beta": 0.006921333260834217, "fcm_dpo/delta": -0.01925438642501831, "fcm_dpo/margin": 53.23359680175781, "fcm_dpo/q_t": 0.4178709387779236, "grad_norm": 24.08075523376465, "learning_rate": 4.870401618977415e-07, "logits/chosen": -0.4671233296394348, "logits/rejected": -0.455013632774353, "logps/chosen": -133.51150512695312, "logps/ref_chosen": -66.60449981689453, "logps/ref_rejected": -96.33355712890625, "logps/rejected": -216.47415161132812, "loss": 1.1328, "margin_dpo/margin_mean": 53.23359680175781, "margin_dpo/margin_std": 86.62762451171875, "step": 133 }, { "KL/chosen_KL_mean": -52.65304946899414, "KL/mean": -80.64926147460938, "KL/rejected_KL_mean": -108.64547729492188, "KL/std": 65.32865142822266, "epoch": 0.19676945668135096, "fcm_dpo/beta": 0.0069086989387869835, "fcm_dpo/delta": -0.03839034587144852, "fcm_dpo/margin": 55.992435455322266, "fcm_dpo/q_t": 0.4121500253677368, "grad_norm": 16.892353057861328, "learning_rate": 4.866292092063986e-07, "logits/chosen": -0.432298481464386, "logits/rejected": -0.4217216372489929, "logps/chosen": -104.72230529785156, "logps/ref_chosen": -52.06925582885742, "logps/ref_rejected": -87.6545181274414, "logps/rejected": -196.29998779296875, "loss": 1.0992, "margin_dpo/margin_mean": 55.992431640625, "margin_dpo/margin_std": 76.5130615234375, "step": 134 }, { "KL/chosen_KL_mean": -57.416786193847656, "KL/mean": -98.86827087402344, "KL/rejected_KL_mean": -140.31976318359375, "KL/std": 87.58612823486328, "epoch": 0.19823788546255505, "fcm_dpo/beta": 0.006657840684056282, "fcm_dpo/delta": -0.2138073444366455, "fcm_dpo/margin": 82.90298461914062, "fcm_dpo/q_t": 0.37990349531173706, "grad_norm": 19.722549438476562, "learning_rate": 4.862120211153265e-07, "logits/chosen": -0.44293975830078125, "logits/rejected": -0.47638678550720215, "logps/chosen": -107.77064514160156, "logps/ref_chosen": -50.353858947753906, "logps/ref_rejected": -115.97975158691406, "logps/rejected": -256.29949951171875, "loss": 1.0199, "margin_dpo/margin_mean": 82.90296936035156, "margin_dpo/margin_std": 106.70145416259766, "step": 135 }, { "KL/chosen_KL_mean": -68.58837890625, "KL/mean": -96.16827392578125, "KL/rejected_KL_mean": -123.74815368652344, "KL/std": 80.87139892578125, "epoch": 0.19970631424375918, "fcm_dpo/beta": 0.0065402439795434475, "fcm_dpo/delta": -0.011789493262767792, "fcm_dpo/margin": 55.159759521484375, "fcm_dpo/q_t": 0.4252905547618866, "grad_norm": 18.383068084716797, "learning_rate": 4.857886086178193e-07, "logits/chosen": -0.4794670045375824, "logits/rejected": -0.4755587577819824, "logps/chosen": -133.660888671875, "logps/ref_chosen": -65.072509765625, "logps/ref_rejected": -96.32122802734375, "logps/rejected": -220.0693817138672, "loss": 1.1626, "margin_dpo/margin_mean": 55.159759521484375, "margin_dpo/margin_std": 104.8607177734375, "step": 136 }, { "KL/chosen_KL_mean": -64.34596252441406, "KL/mean": -107.5447998046875, "KL/rejected_KL_mean": -150.74363708496094, "KL/std": 105.28067016601562, "epoch": 0.2011747430249633, "fcm_dpo/beta": 0.006361949257552624, "fcm_dpo/delta": -0.21148346364498138, "fcm_dpo/margin": 86.39767456054688, "fcm_dpo/q_t": 0.3858060836791992, "grad_norm": 16.011646270751953, "learning_rate": 4.853589828711902e-07, "logits/chosen": -0.39821356534957886, "logits/rejected": -0.424802303314209, "logps/chosen": -113.10507202148438, "logps/ref_chosen": -48.759117126464844, "logps/ref_rejected": -113.86376953125, "logps/rejected": -264.607421875, "loss": 1.0481, "margin_dpo/margin_mean": 86.39767456054688, "margin_dpo/margin_std": 127.21504211425781, "step": 137 }, { "KL/chosen_KL_mean": -67.67312622070312, "KL/mean": -100.0539321899414, "KL/rejected_KL_mean": -132.4347381591797, "KL/std": 79.61257934570312, "epoch": 0.2026431718061674, "fcm_dpo/beta": 0.006255497224628925, "fcm_dpo/delta": -0.05740416795015335, "fcm_dpo/margin": 64.76160430908203, "fcm_dpo/q_t": 0.4059930443763733, "grad_norm": 18.952259063720703, "learning_rate": 4.849231551964771e-07, "logits/chosen": -0.41318434476852417, "logits/rejected": -0.40368473529815674, "logps/chosen": -128.19277954101562, "logps/ref_chosen": -60.519649505615234, "logps/ref_rejected": -93.19694519042969, "logps/rejected": -225.63168334960938, "loss": 1.0821, "margin_dpo/margin_mean": 64.76160430908203, "margin_dpo/margin_std": 81.33100891113281, "step": 138 }, { "KL/chosen_KL_mean": -56.83777618408203, "KL/mean": -94.96588897705078, "KL/rejected_KL_mean": -133.093994140625, "KL/std": 75.09423065185547, "epoch": 0.20411160058737152, "fcm_dpo/beta": 0.006126572377979755, "fcm_dpo/delta": -0.12276843190193176, "fcm_dpo/margin": 76.2562255859375, "fcm_dpo/q_t": 0.39481863379478455, "grad_norm": 18.048723220825195, "learning_rate": 4.844811370781446e-07, "logits/chosen": -0.4068523049354553, "logits/rejected": -0.399114191532135, "logps/chosen": -103.72915649414062, "logps/ref_chosen": -46.89138412475586, "logps/ref_rejected": -79.72798156738281, "logps/rejected": -212.82199096679688, "loss": 1.0533, "margin_dpo/margin_mean": 76.25623321533203, "margin_dpo/margin_std": 98.57319641113281, "step": 139 }, { "KL/chosen_KL_mean": -68.70712280273438, "KL/mean": -102.94752502441406, "KL/rejected_KL_mean": -137.18792724609375, "KL/std": 82.87702941894531, "epoch": 0.2055800293685756, "fcm_dpo/beta": 0.006015871651470661, "fcm_dpo/delta": -0.06465409696102142, "fcm_dpo/margin": 68.48081970214844, "fcm_dpo/q_t": 0.4066656231880188, "grad_norm": 19.090078353881836, "learning_rate": 4.840329401637809e-07, "logits/chosen": -0.39976561069488525, "logits/rejected": -0.3883041739463806, "logps/chosen": -127.68183135986328, "logps/ref_chosen": -58.97471618652344, "logps/ref_rejected": -83.28410339355469, "logps/rejected": -220.4720458984375, "loss": 1.0998, "margin_dpo/margin_mean": 68.4808120727539, "margin_dpo/margin_std": 100.59109497070312, "step": 140 }, { "KL/chosen_KL_mean": -78.09503173828125, "KL/mean": -110.78746032714844, "KL/rejected_KL_mean": -143.47988891601562, "KL/std": 93.90158081054688, "epoch": 0.20704845814977973, "fcm_dpo/beta": 0.005964947864413261, "fcm_dpo/delta": -0.041621994227170944, "fcm_dpo/margin": 65.38487243652344, "fcm_dpo/q_t": 0.4113631248474121, "grad_norm": 23.17949104309082, "learning_rate": 4.83578576263792e-07, "logits/chosen": -0.38903534412384033, "logits/rejected": -0.37867170572280884, "logps/chosen": -153.17068481445312, "logps/ref_chosen": -75.07566833496094, "logps/ref_rejected": -98.1922607421875, "logps/rejected": -241.67214965820312, "loss": 1.1311, "margin_dpo/margin_mean": 65.3848648071289, "margin_dpo/margin_std": 109.24861145019531, "step": 141 }, { "KL/chosen_KL_mean": -78.31304931640625, "KL/mean": -115.74881744384766, "KL/rejected_KL_mean": -153.18460083007812, "KL/std": 100.52348327636719, "epoch": 0.20851688693098386, "fcm_dpo/beta": 0.005887279286980629, "fcm_dpo/delta": -0.09491311013698578, "fcm_dpo/margin": 74.87155151367188, "fcm_dpo/q_t": 0.4024774432182312, "grad_norm": 22.61551284790039, "learning_rate": 4.83118057351089e-07, "logits/chosen": -0.34859800338745117, "logits/rejected": -0.3484548330307007, "logps/chosen": -136.34097290039062, "logps/ref_chosen": -58.027931213378906, "logps/ref_rejected": -94.58222961425781, "logps/rejected": -247.76681518554688, "loss": 1.108, "margin_dpo/margin_mean": 74.87155151367188, "margin_dpo/margin_std": 118.29385375976562, "step": 142 }, { "KL/chosen_KL_mean": -82.15674591064453, "KL/mean": -106.3138427734375, "KL/rejected_KL_mean": -130.470947265625, "KL/std": 88.07025146484375, "epoch": 0.20998531571218795, "fcm_dpo/beta": 0.005829214118421078, "fcm_dpo/delta": -0.005923248827457428, "fcm_dpo/margin": 48.31420135498047, "fcm_dpo/q_t": 0.4382452368736267, "grad_norm": 19.985225677490234, "learning_rate": 4.826513955607734e-07, "logits/chosen": -0.3621300458908081, "logits/rejected": -0.35844361782073975, "logps/chosen": -139.75320434570312, "logps/ref_chosen": -57.59645080566406, "logps/ref_rejected": -78.99957275390625, "logps/rejected": -209.47052001953125, "loss": 1.2106, "margin_dpo/margin_mean": 48.314205169677734, "margin_dpo/margin_std": 103.77206420898438, "step": 143 }, { "KL/chosen_KL_mean": -74.73379516601562, "KL/mean": -103.55740356445312, "KL/rejected_KL_mean": -132.38101196289062, "KL/std": 75.34092712402344, "epoch": 0.21145374449339208, "fcm_dpo/beta": 0.005838489159941673, "fcm_dpo/delta": 0.013897043652832508, "fcm_dpo/margin": 57.6472053527832, "fcm_dpo/q_t": 0.4226115942001343, "grad_norm": 18.863821029663086, "learning_rate": 4.821786031898176e-07, "logits/chosen": -0.3513190448284149, "logits/rejected": -0.3433167338371277, "logps/chosen": -134.64016723632812, "logps/ref_chosen": -59.90636444091797, "logps/ref_rejected": -82.00025939941406, "logps/rejected": -214.38125610351562, "loss": 1.1412, "margin_dpo/margin_mean": 57.6472053527832, "margin_dpo/margin_std": 88.47941589355469, "step": 144 }, { "KL/chosen_KL_mean": -71.26634979248047, "KL/mean": -102.69514465332031, "KL/rejected_KL_mean": -134.12393188476562, "KL/std": 74.58995056152344, "epoch": 0.21292217327459617, "fcm_dpo/beta": 0.005827681161463261, "fcm_dpo/delta": -0.01694849133491516, "fcm_dpo/margin": 62.85758972167969, "fcm_dpo/q_t": 0.4155125617980957, "grad_norm": 20.758092880249023, "learning_rate": 4.816996926967401e-07, "logits/chosen": -0.4076671600341797, "logits/rejected": -0.398476779460907, "logps/chosen": -127.86701965332031, "logps/ref_chosen": -56.60066604614258, "logps/ref_rejected": -77.86631774902344, "logps/rejected": -211.99024963378906, "loss": 1.1196, "margin_dpo/margin_mean": 62.85758972167969, "margin_dpo/margin_std": 91.47295379638672, "step": 145 }, { "KL/chosen_KL_mean": -97.09615325927734, "KL/mean": -121.29571533203125, "KL/rejected_KL_mean": -145.49526977539062, "KL/std": 82.2374038696289, "epoch": 0.2143906020558003, "fcm_dpo/beta": 0.0058697364293038845, "fcm_dpo/delta": 0.06787579506635666, "fcm_dpo/margin": 48.399139404296875, "fcm_dpo/q_t": 0.43450281023979187, "grad_norm": 23.965456008911133, "learning_rate": 4.812146767012779e-07, "logits/chosen": -0.357689768075943, "logits/rejected": -0.33603012561798096, "logps/chosen": -163.0966033935547, "logps/ref_chosen": -66.00045013427734, "logps/ref_rejected": -81.70278930664062, "logps/rejected": -227.19805908203125, "loss": 1.2061, "margin_dpo/margin_mean": 48.39913558959961, "margin_dpo/margin_std": 99.55833435058594, "step": 146 }, { "KL/chosen_KL_mean": -72.37018585205078, "KL/mean": -105.23790740966797, "KL/rejected_KL_mean": -138.10562133789062, "KL/std": 82.96342468261719, "epoch": 0.21585903083700442, "fcm_dpo/beta": 0.0058630527928471565, "fcm_dpo/delta": -0.036948516964912415, "fcm_dpo/margin": 65.73544311523438, "fcm_dpo/q_t": 0.41322061419487, "grad_norm": 19.186176300048828, "learning_rate": 4.807235679840536e-07, "logits/chosen": -0.3486284017562866, "logits/rejected": -0.3298673927783966, "logps/chosen": -125.77567291259766, "logps/ref_chosen": -53.405487060546875, "logps/ref_rejected": -71.39060974121094, "logps/rejected": -209.49624633789062, "loss": 1.1224, "margin_dpo/margin_mean": 65.73544311523438, "margin_dpo/margin_std": 103.400634765625, "step": 147 }, { "KL/chosen_KL_mean": -71.82711029052734, "KL/mean": -101.01795959472656, "KL/rejected_KL_mean": -130.2088165283203, "KL/std": 84.29672241210938, "epoch": 0.2173274596182085, "fcm_dpo/beta": 0.005813123658299446, "fcm_dpo/delta": -0.08374335616827011, "fcm_dpo/margin": 58.3817138671875, "fcm_dpo/q_t": 0.4239063560962677, "grad_norm": 17.18868064880371, "learning_rate": 4.802263794862384e-07, "logits/chosen": -0.45455917716026306, "logits/rejected": -0.450608491897583, "logps/chosen": -136.76419067382812, "logps/ref_chosen": -64.93708038330078, "logps/ref_rejected": -103.09384155273438, "logps/rejected": -233.3026580810547, "loss": 1.1465, "margin_dpo/margin_mean": 58.38170623779297, "margin_dpo/margin_std": 89.03665161132812, "step": 148 }, { "KL/chosen_KL_mean": -68.49336242675781, "KL/mean": -104.46226501464844, "KL/rejected_KL_mean": -140.43118286132812, "KL/std": 75.27666473388672, "epoch": 0.21879588839941264, "fcm_dpo/beta": 0.005668168887495995, "fcm_dpo/delta": -0.0618739016354084, "fcm_dpo/margin": 71.93780517578125, "fcm_dpo/q_t": 0.4055221676826477, "grad_norm": 18.368350982666016, "learning_rate": 4.797231243092118e-07, "logits/chosen": -0.4379425644874573, "logits/rejected": -0.42654889822006226, "logps/chosen": -126.96712493896484, "logps/ref_chosen": -58.47376251220703, "logps/ref_rejected": -99.31474304199219, "logps/rejected": -239.74591064453125, "loss": 1.0855, "margin_dpo/margin_mean": 71.93780517578125, "margin_dpo/margin_std": 89.95822143554688, "step": 149 }, { "KL/chosen_KL_mean": -60.557098388671875, "KL/mean": -96.39286804199219, "KL/rejected_KL_mean": -132.2286376953125, "KL/std": 88.98016357421875, "epoch": 0.22026431718061673, "fcm_dpo/beta": 0.005610051564872265, "fcm_dpo/delta": -0.05549062415957451, "fcm_dpo/margin": 71.67154693603516, "fcm_dpo/q_t": 0.41181066632270813, "grad_norm": 17.013290405273438, "learning_rate": 4.792138157142157e-07, "logits/chosen": -0.3922620415687561, "logits/rejected": -0.3972320556640625, "logps/chosen": -106.26290893554688, "logps/ref_chosen": -45.705810546875, "logps/ref_rejected": -83.34759521484375, "logps/rejected": -215.57623291015625, "loss": 1.1032, "margin_dpo/margin_mean": 71.67154693603516, "margin_dpo/margin_std": 106.37786865234375, "step": 150 }, { "KL/chosen_KL_mean": -76.32244873046875, "KL/mean": -111.68994140625, "KL/rejected_KL_mean": -147.0574493408203, "KL/std": 84.279541015625, "epoch": 0.22173274596182085, "fcm_dpo/beta": 0.005587285850197077, "fcm_dpo/delta": -0.04712191969156265, "fcm_dpo/margin": 70.73500061035156, "fcm_dpo/q_t": 0.40847840905189514, "grad_norm": 17.660154342651367, "learning_rate": 4.786984671220053e-07, "logits/chosen": -0.46673595905303955, "logits/rejected": -0.4433782398700714, "logps/chosen": -146.89328002929688, "logps/ref_chosen": -70.57083129882812, "logps/ref_rejected": -100.46382141113281, "logps/rejected": -247.52127075195312, "loss": 1.0921, "margin_dpo/margin_mean": 70.73500061035156, "margin_dpo/margin_std": 92.79745483398438, "step": 151 }, { "KL/chosen_KL_mean": -66.9847412109375, "KL/mean": -110.66506958007812, "KL/rejected_KL_mean": -154.34539794921875, "KL/std": 86.341552734375, "epoch": 0.22320117474302498, "fcm_dpo/beta": 0.005478001199662685, "fcm_dpo/delta": -0.13483843207359314, "fcm_dpo/margin": 87.36067199707031, "fcm_dpo/q_t": 0.39175742864608765, "grad_norm": 18.47982406616211, "learning_rate": 4.78177092112495e-07, "logits/chosen": -0.4129364788532257, "logits/rejected": -0.4130780100822449, "logps/chosen": -127.14912414550781, "logps/ref_chosen": -60.16438674926758, "logps/ref_rejected": -106.14045715332031, "logps/rejected": -260.48583984375, "loss": 1.0436, "margin_dpo/margin_mean": 87.36066436767578, "margin_dpo/margin_std": 106.23178100585938, "step": 152 }, { "KL/chosen_KL_mean": -67.83352661132812, "KL/mean": -104.77737426757812, "KL/rejected_KL_mean": -141.72122192382812, "KL/std": 95.0350341796875, "epoch": 0.22466960352422907, "fcm_dpo/beta": 0.005398088600486517, "fcm_dpo/delta": -0.050855688750743866, "fcm_dpo/margin": 73.8876953125, "fcm_dpo/q_t": 0.41223540902137756, "grad_norm": 13.99528694152832, "learning_rate": 4.776497044244016e-07, "logits/chosen": -0.4175063967704773, "logits/rejected": -0.415975421667099, "logps/chosen": -124.1488037109375, "logps/ref_chosen": -56.315277099609375, "logps/ref_rejected": -85.65583801269531, "logps/rejected": -227.37704467773438, "loss": 1.1165, "margin_dpo/margin_mean": 73.8876953125, "margin_dpo/margin_std": 117.51593780517578, "step": 153 }, { "KL/chosen_KL_mean": -81.009765625, "KL/mean": -115.98365783691406, "KL/rejected_KL_mean": -150.95755004882812, "KL/std": 94.65357208251953, "epoch": 0.2261380323054332, "fcm_dpo/beta": 0.005372250452637672, "fcm_dpo/delta": -0.026948172599077225, "fcm_dpo/margin": 69.94779205322266, "fcm_dpo/q_t": 0.41602611541748047, "grad_norm": 17.572832107543945, "learning_rate": 4.771163179548808e-07, "logits/chosen": -0.4119012653827667, "logits/rejected": -0.4173479378223419, "logps/chosen": -143.75233459472656, "logps/ref_chosen": -62.74256896972656, "logps/ref_rejected": -104.24420166015625, "logps/rejected": -255.20175170898438, "loss": 1.1465, "margin_dpo/margin_mean": 69.94779205322266, "margin_dpo/margin_std": 122.20135498046875, "step": 154 }, { "KL/chosen_KL_mean": -72.86953735351562, "KL/mean": -107.74058532714844, "KL/rejected_KL_mean": -142.6116180419922, "KL/std": 85.74209594726562, "epoch": 0.2276064610866373, "fcm_dpo/beta": 0.005339436233043671, "fcm_dpo/delta": -0.023303672671318054, "fcm_dpo/margin": 69.74208068847656, "fcm_dpo/q_t": 0.41487759351730347, "grad_norm": 17.113283157348633, "learning_rate": 4.7657694675916247e-07, "logits/chosen": -0.43295127153396606, "logits/rejected": -0.41761913895606995, "logps/chosen": -133.52272033691406, "logps/ref_chosen": -60.65318298339844, "logps/ref_rejected": -77.49220275878906, "logps/rejected": -220.10382080078125, "loss": 1.1242, "margin_dpo/margin_mean": 69.74208068847656, "margin_dpo/margin_std": 107.39170837402344, "step": 155 }, { "KL/chosen_KL_mean": -102.05209350585938, "KL/mean": -121.50769805908203, "KL/rejected_KL_mean": -140.9633026123047, "KL/std": 92.87469482421875, "epoch": 0.2290748898678414, "fcm_dpo/beta": 0.005365746095776558, "fcm_dpo/delta": 0.042066704481840134, "fcm_dpo/margin": 38.91120910644531, "fcm_dpo/q_t": 0.45207709074020386, "grad_norm": 26.29570198059082, "learning_rate": 4.7603160505017893e-07, "logits/chosen": -0.3665209412574768, "logits/rejected": -0.3663702607154846, "logps/chosen": -171.54397583007812, "logps/ref_chosen": -69.49188232421875, "logps/ref_rejected": -77.16929626464844, "logps/rejected": -218.13259887695312, "loss": 1.2896, "margin_dpo/margin_mean": 38.91120910644531, "margin_dpo/margin_std": 121.42982482910156, "step": 156 }, { "KL/chosen_KL_mean": -90.3784408569336, "KL/mean": -135.53335571289062, "KL/rejected_KL_mean": -180.68826293945312, "KL/std": 100.72486877441406, "epoch": 0.2305433186490455, "fcm_dpo/beta": 0.005238070152699947, "fcm_dpo/delta": -0.13022619485855103, "fcm_dpo/margin": 90.30982208251953, "fcm_dpo/q_t": 0.3903403580188751, "grad_norm": 22.826868057250977, "learning_rate": 4.7548030719819154e-07, "logits/chosen": -0.32157012820243835, "logits/rejected": -0.32990676164627075, "logps/chosen": -151.74688720703125, "logps/ref_chosen": -61.368438720703125, "logps/ref_rejected": -107.64636993408203, "logps/rejected": -288.33465576171875, "loss": 1.0571, "margin_dpo/margin_mean": 90.30982971191406, "margin_dpo/margin_std": 114.6199951171875, "step": 157 }, { "KL/chosen_KL_mean": -92.16410827636719, "KL/mean": -140.3941192626953, "KL/rejected_KL_mean": -188.62411499023438, "KL/std": 124.57286834716797, "epoch": 0.23201174743024963, "fcm_dpo/beta": 0.005113640800118446, "fcm_dpo/delta": -0.1506519466638565, "fcm_dpo/margin": 96.45999908447266, "fcm_dpo/q_t": 0.3960997462272644, "grad_norm": 16.963441848754883, "learning_rate": 4.7492306773041136e-07, "logits/chosen": -0.31055477261543274, "logits/rejected": -0.3295873701572418, "logps/chosen": -149.77703857421875, "logps/ref_chosen": -57.612918853759766, "logps/ref_rejected": -113.6946792602539, "logps/rejected": -302.31878662109375, "loss": 1.0799, "margin_dpo/margin_mean": 96.45999908447266, "margin_dpo/margin_std": 151.36740112304688, "step": 158 }, { "KL/chosen_KL_mean": -104.50662231445312, "KL/mean": -136.48501586914062, "KL/rejected_KL_mean": -168.46337890625, "KL/std": 109.98271179199219, "epoch": 0.23348017621145375, "fcm_dpo/beta": 0.0051115090027451515, "fcm_dpo/delta": 0.023419687524437904, "fcm_dpo/margin": 63.956756591796875, "fcm_dpo/q_t": 0.42578351497650146, "grad_norm": 19.77988624572754, "learning_rate": 4.743599013306165e-07, "logits/chosen": -0.33859413862228394, "logits/rejected": -0.3108740448951721, "logps/chosen": -186.0669708251953, "logps/ref_chosen": -81.56034851074219, "logps/ref_rejected": -88.89871215820312, "logps/rejected": -257.3620910644531, "loss": 1.1747, "margin_dpo/margin_mean": 63.956748962402344, "margin_dpo/margin_std": 119.0658950805664, "step": 159 }, { "KL/chosen_KL_mean": -105.49040985107422, "KL/mean": -149.2151336669922, "KL/rejected_KL_mean": -192.93984985351562, "KL/std": 116.9592514038086, "epoch": 0.23494860499265785, "fcm_dpo/beta": 0.005003074184060097, "fcm_dpo/delta": -0.09241557866334915, "fcm_dpo/margin": 87.44945526123047, "fcm_dpo/q_t": 0.4059435725212097, "grad_norm": 22.12449073791504, "learning_rate": 4.737908228387656e-07, "logits/chosen": -0.31422072649002075, "logits/rejected": -0.3084676265716553, "logps/chosen": -171.22128295898438, "logps/ref_chosen": -65.73088073730469, "logps/ref_rejected": -97.21781921386719, "logps/rejected": -290.15765380859375, "loss": 1.1166, "margin_dpo/margin_mean": 87.44944763183594, "margin_dpo/margin_std": 146.5648193359375, "step": 160 }, { "KL/chosen_KL_mean": -89.08880615234375, "KL/mean": -127.54766845703125, "KL/rejected_KL_mean": -166.0065460205078, "KL/std": 91.42471313476562, "epoch": 0.23641703377386197, "fcm_dpo/beta": 0.004976830445230007, "fcm_dpo/delta": -0.03411930426955223, "fcm_dpo/margin": 76.917724609375, "fcm_dpo/q_t": 0.4135284423828125, "grad_norm": 19.238542556762695, "learning_rate": 4.7321584725060594e-07, "logits/chosen": -0.319388210773468, "logits/rejected": -0.3227710723876953, "logps/chosen": -141.52528381347656, "logps/ref_chosen": -52.43647003173828, "logps/ref_rejected": -83.43095397949219, "logps/rejected": -249.4375, "loss": 1.1164, "margin_dpo/margin_mean": 76.917724609375, "margin_dpo/margin_std": 115.2650146484375, "step": 161 }, { "KL/chosen_KL_mean": -87.23893737792969, "KL/mean": -124.27473449707031, "KL/rejected_KL_mean": -161.310546875, "KL/std": 101.31791687011719, "epoch": 0.23788546255506607, "fcm_dpo/beta": 0.0048839072696864605, "fcm_dpo/delta": -0.086455799639225, "fcm_dpo/margin": 74.07159423828125, "fcm_dpo/q_t": 0.4182526171207428, "grad_norm": 19.80852508544922, "learning_rate": 4.7263498971727905e-07, "logits/chosen": -0.3837524354457855, "logits/rejected": -0.37143242359161377, "logps/chosen": -149.84951782226562, "logps/ref_chosen": -62.6105842590332, "logps/ref_rejected": -89.39057922363281, "logps/rejected": -250.70111083984375, "loss": 1.141, "margin_dpo/margin_mean": 74.07159423828125, "margin_dpo/margin_std": 120.83529663085938, "step": 162 }, { "KL/chosen_KL_mean": -96.60433959960938, "KL/mean": -132.7714080810547, "KL/rejected_KL_mean": -168.9384765625, "KL/std": 101.78199768066406, "epoch": 0.2393538913362702, "fcm_dpo/beta": 0.004882782697677612, "fcm_dpo/delta": -0.0033093392848968506, "fcm_dpo/margin": 72.33414459228516, "fcm_dpo/q_t": 0.4210611581802368, "grad_norm": 18.05702018737793, "learning_rate": 4.720482655449212e-07, "logits/chosen": -0.2983933985233307, "logits/rejected": -0.28219735622406006, "logps/chosen": -151.6259765625, "logps/ref_chosen": -55.021629333496094, "logps/ref_rejected": -75.418212890625, "logps/rejected": -244.356689453125, "loss": 1.1471, "margin_dpo/margin_mean": 72.33415222167969, "margin_dpo/margin_std": 123.11458587646484, "step": 163 }, { "KL/chosen_KL_mean": -88.20658111572266, "KL/mean": -133.68885803222656, "KL/rejected_KL_mean": -179.17112731933594, "KL/std": 99.22830963134766, "epoch": 0.24082232011747431, "fcm_dpo/beta": 0.004785895813256502, "fcm_dpo/delta": -0.09055467694997787, "fcm_dpo/margin": 90.96453857421875, "fcm_dpo/q_t": 0.39980244636535645, "grad_norm": 19.5128231048584, "learning_rate": 4.714556901942599e-07, "logits/chosen": -0.3039168119430542, "logits/rejected": -0.2934049963951111, "logps/chosen": -143.84725952148438, "logps/ref_chosen": -55.64066696166992, "logps/ref_rejected": -79.66463470458984, "logps/rejected": -258.83575439453125, "loss": 1.068, "margin_dpo/margin_mean": 90.96453857421875, "margin_dpo/margin_std": 114.03044128417969, "step": 164 }, { "KL/chosen_KL_mean": -94.5549087524414, "KL/mean": -124.16766357421875, "KL/rejected_KL_mean": -153.7803955078125, "KL/std": 84.79214477539062, "epoch": 0.2422907488986784, "fcm_dpo/beta": 0.004827073775231838, "fcm_dpo/delta": 0.06602644920349121, "fcm_dpo/margin": 59.225494384765625, "fcm_dpo/q_t": 0.4348486065864563, "grad_norm": 18.897262573242188, "learning_rate": 4.708572792802069e-07, "logits/chosen": -0.32065290212631226, "logits/rejected": -0.2971528172492981, "logps/chosen": -155.8656005859375, "logps/ref_chosen": -61.310691833496094, "logps/ref_rejected": -73.67060852050781, "logps/rejected": -227.45101928710938, "loss": 1.1927, "margin_dpo/margin_mean": 59.225494384765625, "margin_dpo/margin_std": 113.44796752929688, "step": 165 }, { "KL/chosen_KL_mean": -84.27728271484375, "KL/mean": -139.05355834960938, "KL/rejected_KL_mean": -193.82984924316406, "KL/std": 123.94483947753906, "epoch": 0.24375917767988253, "fcm_dpo/beta": 0.004713358357548714, "fcm_dpo/delta": -0.17637449502944946, "fcm_dpo/margin": 109.55255889892578, "fcm_dpo/q_t": 0.39149314165115356, "grad_norm": 16.227643966674805, "learning_rate": 4.702530485714461e-07, "logits/chosen": -0.30493152141571045, "logits/rejected": -0.31587445735931396, "logps/chosen": -135.26087951660156, "logps/ref_chosen": -50.98360061645508, "logps/ref_rejected": -98.09512329101562, "logps/rejected": -291.92498779296875, "loss": 1.049, "margin_dpo/margin_mean": 109.55255889892578, "margin_dpo/margin_std": 156.71432495117188, "step": 166 }, { "KL/chosen_KL_mean": -84.61183166503906, "KL/mean": -142.53759765625, "KL/rejected_KL_mean": -200.4633331298828, "KL/std": 112.2884521484375, "epoch": 0.24522760646108663, "fcm_dpo/beta": 0.00455870758742094, "fcm_dpo/delta": -0.18809391558170319, "fcm_dpo/margin": 115.85150909423828, "fcm_dpo/q_t": 0.380374550819397, "grad_norm": 17.7374267578125, "learning_rate": 4.6964301399001877e-07, "logits/chosen": -0.29031479358673096, "logits/rejected": -0.2939522862434387, "logps/chosen": -135.03591918945312, "logps/ref_chosen": -50.424095153808594, "logps/ref_rejected": -96.03042602539062, "logps/rejected": -296.4937744140625, "loss": 1.0076, "margin_dpo/margin_mean": 115.85150909423828, "margin_dpo/margin_std": 131.866943359375, "step": 167 }, { "KL/chosen_KL_mean": -92.86872863769531, "KL/mean": -134.5798797607422, "KL/rejected_KL_mean": -176.29103088378906, "KL/std": 104.45317077636719, "epoch": 0.24669603524229075, "fcm_dpo/beta": 0.00448962114751339, "fcm_dpo/delta": -0.0256113912910223, "fcm_dpo/margin": 83.42231750488281, "fcm_dpo/q_t": 0.41462987661361694, "grad_norm": 19.607961654663086, "learning_rate": 4.690271916109034e-07, "logits/chosen": -0.2678576111793518, "logits/rejected": -0.259553462266922, "logps/chosen": -142.33154296875, "logps/ref_chosen": -49.462825775146484, "logps/ref_rejected": -75.30855560302734, "logps/rejected": -251.59959411621094, "loss": 1.1117, "margin_dpo/margin_mean": 83.42231750488281, "margin_dpo/margin_std": 118.4161376953125, "step": 168 }, { "KL/chosen_KL_mean": -97.21165466308594, "KL/mean": -133.64254760742188, "KL/rejected_KL_mean": -170.0734405517578, "KL/std": 102.94970703125, "epoch": 0.24816446402349487, "fcm_dpo/beta": 0.0044325110502541065, "fcm_dpo/delta": -0.06697965413331985, "fcm_dpo/margin": 72.86180114746094, "fcm_dpo/q_t": 0.4287607669830322, "grad_norm": 18.7218017578125, "learning_rate": 4.6840559766159235e-07, "logits/chosen": -0.28774628043174744, "logits/rejected": -0.2726348638534546, "logps/chosen": -157.01510620117188, "logps/ref_chosen": -59.803443908691406, "logps/ref_rejected": -83.34574890136719, "logps/rejected": -253.419189453125, "loss": 1.1829, "margin_dpo/margin_mean": 72.86180114746094, "margin_dpo/margin_std": 142.5083465576172, "step": 169 }, { "KL/chosen_KL_mean": -86.51754760742188, "KL/mean": -129.5677947998047, "KL/rejected_KL_mean": -172.6180419921875, "KL/std": 98.91424560546875, "epoch": 0.24963289280469897, "fcm_dpo/beta": 0.0043886760249733925, "fcm_dpo/delta": -0.02974541112780571, "fcm_dpo/margin": 86.10049438476562, "fcm_dpo/q_t": 0.41227343678474426, "grad_norm": 15.836402893066406, "learning_rate": 4.6777824852166437e-07, "logits/chosen": -0.2773169279098511, "logits/rejected": -0.26892268657684326, "logps/chosen": -135.98931884765625, "logps/ref_chosen": -49.471771240234375, "logps/ref_rejected": -75.91734313964844, "logps/rejected": -248.53538513183594, "loss": 1.1099, "margin_dpo/margin_mean": 86.10049438476562, "margin_dpo/margin_std": 118.6636734008789, "step": 170 }, { "KL/chosen_KL_mean": -125.79786682128906, "KL/mean": -160.0425262451172, "KL/rejected_KL_mean": -194.28720092773438, "KL/std": 116.10159301757812, "epoch": 0.2511013215859031, "fcm_dpo/beta": 0.004434281028807163, "fcm_dpo/delta": 0.04777422547340393, "fcm_dpo/margin": 68.48933410644531, "fcm_dpo/q_t": 0.4349101185798645, "grad_norm": 22.20769691467285, "learning_rate": 4.6714516072235273e-07, "logits/chosen": -0.3094336986541748, "logits/rejected": -0.2986791431903839, "logps/chosen": -210.29718017578125, "logps/ref_chosen": -84.49931335449219, "logps/ref_rejected": -109.38209533691406, "logps/rejected": -303.6693115234375, "loss": 1.2027, "margin_dpo/margin_mean": 68.48932647705078, "margin_dpo/margin_std": 149.46047973632812, "step": 171 }, { "KL/chosen_KL_mean": -109.93336486816406, "KL/mean": -146.34115600585938, "KL/rejected_KL_mean": -182.74893188476562, "KL/std": 112.77561950683594, "epoch": 0.2525697503671072, "fcm_dpo/beta": 0.0044615124352276325, "fcm_dpo/delta": 0.025982577353715897, "fcm_dpo/margin": 72.8155517578125, "fcm_dpo/q_t": 0.4260951280593872, "grad_norm": 16.35190773010254, "learning_rate": 4.6650635094610966e-07, "logits/chosen": -0.33784881234169006, "logits/rejected": -0.32723069190979004, "logps/chosen": -178.5872802734375, "logps/ref_chosen": -68.65391540527344, "logps/ref_rejected": -85.43667602539062, "logps/rejected": -268.18560791015625, "loss": 1.1606, "margin_dpo/margin_mean": 72.8155517578125, "margin_dpo/margin_std": 125.91693115234375, "step": 172 }, { "KL/chosen_KL_mean": -99.22401428222656, "KL/mean": -135.5485076904297, "KL/rejected_KL_mean": -171.87301635742188, "KL/std": 104.18135070800781, "epoch": 0.2540381791483113, "fcm_dpo/beta": 0.0045021940022706985, "fcm_dpo/delta": 0.02339586801826954, "fcm_dpo/margin": 72.64897918701172, "fcm_dpo/q_t": 0.42472150921821594, "grad_norm": 18.52365493774414, "learning_rate": 4.6586183602616687e-07, "logits/chosen": -0.3810299038887024, "logits/rejected": -0.35811132192611694, "logps/chosen": -162.27490234375, "logps/ref_chosen": -63.050880432128906, "logps/ref_rejected": -78.68392181396484, "logps/rejected": -250.5569305419922, "loss": 1.1417, "margin_dpo/margin_mean": 72.64897918701172, "margin_dpo/margin_std": 108.23287963867188, "step": 173 }, { "KL/chosen_KL_mean": -92.88302612304688, "KL/mean": -136.38131713867188, "KL/rejected_KL_mean": -179.87962341308594, "KL/std": 106.6904296875, "epoch": 0.2555066079295154, "fcm_dpo/beta": 0.004488729871809483, "fcm_dpo/delta": -0.04252074658870697, "fcm_dpo/margin": 86.99659729003906, "fcm_dpo/q_t": 0.41205543279647827, "grad_norm": 18.965898513793945, "learning_rate": 4.652116329460919e-07, "logits/chosen": -0.3001505732536316, "logits/rejected": -0.318511426448822, "logps/chosen": -146.24600219726562, "logps/ref_chosen": -53.36296844482422, "logps/ref_rejected": -101.91120910644531, "logps/rejected": -281.79083251953125, "loss": 1.1136, "margin_dpo/margin_mean": 86.99659729003906, "margin_dpo/margin_std": 129.03590393066406, "step": 174 }, { "KL/chosen_KL_mean": -84.55975341796875, "KL/mean": -143.66534423828125, "KL/rejected_KL_mean": -202.77093505859375, "KL/std": 113.07514953613281, "epoch": 0.25697503671071953, "fcm_dpo/beta": 0.004336726851761341, "fcm_dpo/delta": -0.17170114815235138, "fcm_dpo/margin": 118.21118927001953, "fcm_dpo/q_t": 0.3807182312011719, "grad_norm": 26.023164749145508, "learning_rate": 4.645557588393406e-07, "logits/chosen": -0.29982995986938477, "logits/rejected": -0.28770336508750916, "logps/chosen": -129.97752380371094, "logps/ref_chosen": -45.417762756347656, "logps/ref_rejected": -89.50579833984375, "logps/rejected": -292.2767333984375, "loss": 0.9978, "margin_dpo/margin_mean": 118.21118927001953, "margin_dpo/margin_std": 118.14212036132812, "step": 175 }, { "KL/chosen_KL_mean": -91.14138793945312, "KL/mean": -141.39892578125, "KL/rejected_KL_mean": -191.65646362304688, "KL/std": 113.75053405761719, "epoch": 0.25844346549192365, "fcm_dpo/beta": 0.004258063621819019, "fcm_dpo/delta": -0.08141526579856873, "fcm_dpo/margin": 100.51506042480469, "fcm_dpo/q_t": 0.40386536717414856, "grad_norm": 17.673473358154297, "learning_rate": 4.638942309888058e-07, "logits/chosen": -0.2949311137199402, "logits/rejected": -0.31282657384872437, "logps/chosen": -141.59423828125, "logps/ref_chosen": -50.452842712402344, "logps/ref_rejected": -95.5589599609375, "logps/rejected": -287.2154235839844, "loss": 1.0742, "margin_dpo/margin_mean": 100.51506042480469, "margin_dpo/margin_std": 131.3572998046875, "step": 176 }, { "KL/chosen_KL_mean": -106.9547119140625, "KL/mean": -156.6512451171875, "KL/rejected_KL_mean": -206.34779357910156, "KL/std": 124.73439025878906, "epoch": 0.2599118942731278, "fcm_dpo/beta": 0.004201659932732582, "fcm_dpo/delta": -0.07051907479763031, "fcm_dpo/margin": 99.39309692382812, "fcm_dpo/q_t": 0.40569955110549927, "grad_norm": 22.143009185791016, "learning_rate": 4.6322706682636137e-07, "logits/chosen": -0.34562036395072937, "logits/rejected": -0.34030306339263916, "logps/chosen": -168.17117309570312, "logps/ref_chosen": -61.216468811035156, "logps/ref_rejected": -95.89378356933594, "logps/rejected": -302.2415771484375, "loss": 1.0813, "margin_dpo/margin_mean": 99.39309692382812, "margin_dpo/margin_std": 132.520263671875, "step": 177 }, { "KL/chosen_KL_mean": -119.19122314453125, "KL/mean": -181.4816131591797, "KL/rejected_KL_mean": -243.77198791503906, "KL/std": 145.41358947753906, "epoch": 0.26138032305433184, "fcm_dpo/beta": 0.004061376675963402, "fcm_dpo/delta": -0.16501104831695557, "fcm_dpo/margin": 124.58078002929688, "fcm_dpo/q_t": 0.38797473907470703, "grad_norm": 22.72697639465332, "learning_rate": 4.6255428393240354e-07, "logits/chosen": -0.19636118412017822, "logits/rejected": -0.18638736009597778, "logps/chosen": -177.4560089111328, "logps/ref_chosen": -58.26478958129883, "logps/ref_rejected": -105.3653335571289, "logps/rejected": -349.1373291015625, "loss": 1.0348, "margin_dpo/margin_mean": 124.58078002929688, "margin_dpo/margin_std": 159.75628662109375, "step": 178 }, { "KL/chosen_KL_mean": -124.12211608886719, "KL/mean": -167.00396728515625, "KL/rejected_KL_mean": -209.88583374023438, "KL/std": 125.32534790039062, "epoch": 0.26284875183553597, "fcm_dpo/beta": 0.00402455497533083, "fcm_dpo/delta": 0.004620734602212906, "fcm_dpo/margin": 85.76371765136719, "fcm_dpo/q_t": 0.4227069020271301, "grad_norm": 29.680816650390625, "learning_rate": 4.6187590003538724e-07, "logits/chosen": -0.26940906047821045, "logits/rejected": -0.2804524302482605, "logps/chosen": -185.18045043945312, "logps/ref_chosen": -61.05832290649414, "logps/ref_rejected": -90.52782440185547, "logps/rejected": -300.41363525390625, "loss": 1.1654, "margin_dpo/margin_mean": 85.76371002197266, "margin_dpo/margin_std": 157.05172729492188, "step": 179 }, { "KL/chosen_KL_mean": -109.26609802246094, "KL/mean": -166.31593322753906, "KL/rejected_KL_mean": -223.3657684326172, "KL/std": 112.9661865234375, "epoch": 0.2643171806167401, "fcm_dpo/beta": 0.003975175321102142, "fcm_dpo/delta": -0.10860335826873779, "fcm_dpo/margin": 114.09965515136719, "fcm_dpo/q_t": 0.39550089836120605, "grad_norm": 17.493207931518555, "learning_rate": 4.611919330113591e-07, "logits/chosen": -0.24312031269073486, "logits/rejected": -0.2381986677646637, "logps/chosen": -163.60882568359375, "logps/ref_chosen": -54.34272003173828, "logps/ref_rejected": -98.21183776855469, "logps/rejected": -321.5776062011719, "loss": 1.0554, "margin_dpo/margin_mean": 114.09966278076172, "margin_dpo/margin_std": 141.2296142578125, "step": 180 }, { "KL/chosen_KL_mean": -95.30699920654297, "KL/mean": -130.60113525390625, "KL/rejected_KL_mean": -165.895263671875, "KL/std": 105.03799438476562, "epoch": 0.2657856093979442, "fcm_dpo/beta": 0.004001600667834282, "fcm_dpo/delta": 0.06941692531108856, "fcm_dpo/margin": 70.58827209472656, "fcm_dpo/q_t": 0.4350418150424957, "grad_norm": 17.06695556640625, "learning_rate": 4.605024008834863e-07, "logits/chosen": -0.2419101893901825, "logits/rejected": -0.21502138674259186, "logps/chosen": -150.30746459960938, "logps/ref_chosen": -55.000457763671875, "logps/ref_rejected": -61.656166076660156, "logps/rejected": -227.55142211914062, "loss": 1.1912, "margin_dpo/margin_mean": 70.58827209472656, "margin_dpo/margin_std": 133.7356719970703, "step": 181 }, { "KL/chosen_KL_mean": -91.21590423583984, "KL/mean": -153.8636016845703, "KL/rejected_KL_mean": -216.51129150390625, "KL/std": 126.58869934082031, "epoch": 0.26725403817914833, "fcm_dpo/beta": 0.003911532927304506, "fcm_dpo/delta": -0.14804580807685852, "fcm_dpo/margin": 125.29539489746094, "fcm_dpo/q_t": 0.3870447278022766, "grad_norm": 19.177839279174805, "learning_rate": 4.598073218215817e-07, "logits/chosen": -0.24630862474441528, "logits/rejected": -0.257821261882782, "logps/chosen": -132.32376098632812, "logps/ref_chosen": -41.107852935791016, "logps/ref_rejected": -89.5215835571289, "logps/rejected": -306.03289794921875, "loss": 1.0378, "margin_dpo/margin_mean": 125.29539489746094, "margin_dpo/margin_std": 152.4662322998047, "step": 182 }, { "KL/chosen_KL_mean": -132.7602996826172, "KL/mean": -165.40869140625, "KL/rejected_KL_mean": -198.0570831298828, "KL/std": 105.83728790283203, "epoch": 0.2687224669603524, "fcm_dpo/beta": 0.003861016361042857, "fcm_dpo/delta": -0.043808713555336, "fcm_dpo/margin": 65.29679107666016, "fcm_dpo/q_t": 0.44055789709091187, "grad_norm": 19.30330467224121, "learning_rate": 4.5910671414162484e-07, "logits/chosen": -0.26778051257133484, "logits/rejected": -0.2602764964103699, "logps/chosen": -190.28485107421875, "logps/ref_chosen": -57.52456283569336, "logps/ref_rejected": -75.97572326660156, "logps/rejected": -274.0328063964844, "loss": 1.205, "margin_dpo/margin_mean": 65.29679107666016, "margin_dpo/margin_std": 115.09003448486328, "step": 183 }, { "KL/chosen_KL_mean": -112.4358139038086, "KL/mean": -147.748291015625, "KL/rejected_KL_mean": -183.06076049804688, "KL/std": 100.15349578857422, "epoch": 0.2701908957415565, "fcm_dpo/beta": 0.00385840330272913, "fcm_dpo/delta": -0.0067726364359259605, "fcm_dpo/margin": 70.62494659423828, "fcm_dpo/q_t": 0.4372936487197876, "grad_norm": 19.006750106811523, "learning_rate": 4.5840059630527985e-07, "logits/chosen": -0.3127518892288208, "logits/rejected": -0.30516594648361206, "logps/chosen": -170.98077392578125, "logps/ref_chosen": -58.544952392578125, "logps/ref_rejected": -76.63406372070312, "logps/rejected": -259.69482421875, "loss": 1.1911, "margin_dpo/margin_mean": 70.62494659423828, "margin_dpo/margin_std": 126.89500427246094, "step": 184 }, { "KL/chosen_KL_mean": -119.8270492553711, "KL/mean": -147.22073364257812, "KL/rejected_KL_mean": -174.61444091796875, "KL/std": 113.41376495361328, "epoch": 0.27165932452276065, "fcm_dpo/beta": 0.0038557900115847588, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 54.78740310668945, "fcm_dpo/q_t": 0.45315247774124146, "grad_norm": 18.04941749572754, "learning_rate": 4.5768898691940836e-07, "logits/chosen": -0.27935075759887695, "logits/rejected": -0.2586449682712555, "logps/chosen": -181.8529052734375, "logps/ref_chosen": -62.025848388671875, "logps/ref_rejected": -73.7625961303711, "logps/rejected": -248.37704467773438, "loss": 1.2524, "margin_dpo/margin_mean": 54.78739929199219, "margin_dpo/margin_std": 138.59088134765625, "step": 185 }, { "KL/chosen_KL_mean": -109.99482727050781, "KL/mean": -162.43775939941406, "KL/rejected_KL_mean": -214.88067626953125, "KL/std": 113.16603088378906, "epoch": 0.27312775330396477, "fcm_dpo/beta": 0.0038217236287891865, "fcm_dpo/delta": -0.05303904414176941, "fcm_dpo/margin": 104.88587188720703, "fcm_dpo/q_t": 0.4073418080806732, "grad_norm": 22.06662940979004, "learning_rate": 4.5697190473557947e-07, "logits/chosen": -0.3389773666858673, "logits/rejected": -0.32196861505508423, "logps/chosen": -179.34829711914062, "logps/ref_chosen": -69.35346984863281, "logps/ref_rejected": -88.07244873046875, "logps/rejected": -302.953125, "loss": 1.0831, "margin_dpo/margin_mean": 104.8858642578125, "margin_dpo/margin_std": 130.91024780273438, "step": 186 }, { "KL/chosen_KL_mean": -101.02510070800781, "KL/mean": -145.2807159423828, "KL/rejected_KL_mean": -189.5363311767578, "KL/std": 107.92250061035156, "epoch": 0.2745961820851689, "fcm_dpo/beta": 0.0038345775101333857, "fcm_dpo/delta": 0.010683363303542137, "fcm_dpo/margin": 88.51124572753906, "fcm_dpo/q_t": 0.4206673204898834, "grad_norm": 21.087003707885742, "learning_rate": 4.5624936864957555e-07, "logits/chosen": -0.2874957323074341, "logits/rejected": -0.281036376953125, "logps/chosen": -153.78155517578125, "logps/ref_chosen": -52.7564582824707, "logps/ref_rejected": -81.96910095214844, "logps/rejected": -271.50543212890625, "loss": 1.124, "margin_dpo/margin_mean": 88.51124572753906, "margin_dpo/margin_std": 116.87910461425781, "step": 187 }, { "KL/chosen_KL_mean": -93.75711059570312, "KL/mean": -147.31398010253906, "KL/rejected_KL_mean": -200.870849609375, "KL/std": 119.8453598022461, "epoch": 0.27606461086637296, "fcm_dpo/beta": 0.0037920954637229443, "fcm_dpo/delta": -0.05853914096951485, "fcm_dpo/margin": 107.1137466430664, "fcm_dpo/q_t": 0.4064713716506958, "grad_norm": 26.782394409179688, "learning_rate": 4.5552139770089454e-07, "logits/chosen": -0.288492351770401, "logits/rejected": -0.29583364725112915, "logps/chosen": -143.172607421875, "logps/ref_chosen": -49.415489196777344, "logps/ref_rejected": -89.54043579101562, "logps/rejected": -290.4112854003906, "loss": 1.0786, "margin_dpo/margin_mean": 107.1137466430664, "margin_dpo/margin_std": 131.83758544921875, "step": 188 }, { "KL/chosen_KL_mean": -110.248291015625, "KL/mean": -153.5526885986328, "KL/rejected_KL_mean": -196.85708618164062, "KL/std": 120.16261291503906, "epoch": 0.2775330396475771, "fcm_dpo/beta": 0.0037943553179502487, "fcm_dpo/delta": 0.022069651633501053, "fcm_dpo/margin": 86.60880279541016, "fcm_dpo/q_t": 0.4262270927429199, "grad_norm": 19.686960220336914, "learning_rate": 4.5478801107224794e-07, "logits/chosen": -0.30030959844589233, "logits/rejected": -0.28199464082717896, "logps/chosen": -162.64724731445312, "logps/ref_chosen": -52.39896011352539, "logps/ref_rejected": -72.16735076904297, "logps/rejected": -269.0244445800781, "loss": 1.1554, "margin_dpo/margin_mean": 86.60879516601562, "margin_dpo/margin_std": 148.0962371826172, "step": 189 }, { "KL/chosen_KL_mean": -117.99136352539062, "KL/mean": -169.86740112304688, "KL/rejected_KL_mean": -221.74343872070312, "KL/std": 129.9977264404297, "epoch": 0.2790014684287812, "fcm_dpo/beta": 0.0037976130843162537, "fcm_dpo/delta": -0.04660265892744064, "fcm_dpo/margin": 103.75205993652344, "fcm_dpo/q_t": 0.41064584255218506, "grad_norm": 17.065195083618164, "learning_rate": 4.5404922808905543e-07, "logits/chosen": -0.34390515089035034, "logits/rejected": -0.3343849182128906, "logps/chosen": -182.6744384765625, "logps/ref_chosen": -64.68305969238281, "logps/ref_rejected": -102.55052185058594, "logps/rejected": -324.2939453125, "loss": 1.1107, "margin_dpo/margin_mean": 103.75205993652344, "margin_dpo/margin_std": 149.59950256347656, "step": 190 }, { "KL/chosen_KL_mean": -109.85092163085938, "KL/mean": -184.33343505859375, "KL/rejected_KL_mean": -258.8159484863281, "KL/std": 148.56558227539062, "epoch": 0.28046989720998533, "fcm_dpo/beta": 0.0036355718038976192, "fcm_dpo/delta": -0.2035076916217804, "fcm_dpo/margin": 148.96502685546875, "fcm_dpo/q_t": 0.3776732087135315, "grad_norm": 19.175321578979492, "learning_rate": 4.5330506821893565e-07, "logits/chosen": -0.3079345226287842, "logits/rejected": -0.2833190858364105, "logps/chosen": -178.50979614257812, "logps/ref_chosen": -68.65887451171875, "logps/ref_rejected": -110.1396713256836, "logps/rejected": -368.95562744140625, "loss": 0.9954, "margin_dpo/margin_mean": 148.96502685546875, "margin_dpo/margin_std": 165.2587432861328, "step": 191 }, { "KL/chosen_KL_mean": -141.1156463623047, "KL/mean": -190.08807373046875, "KL/rejected_KL_mean": -239.06048583984375, "KL/std": 129.36810302734375, "epoch": 0.28193832599118945, "fcm_dpo/beta": 0.003615614492446184, "fcm_dpo/delta": -0.0043648164719343185, "fcm_dpo/margin": 97.9448471069336, "fcm_dpo/q_t": 0.42013078927993774, "grad_norm": 20.97152328491211, "learning_rate": 4.5255555107119336e-07, "logits/chosen": -0.274461030960083, "logits/rejected": -0.2736404240131378, "logps/chosen": -210.84255981445312, "logps/ref_chosen": -69.72691345214844, "logps/ref_rejected": -103.32135009765625, "logps/rejected": -342.3818359375, "loss": 1.1403, "margin_dpo/margin_mean": 97.9448471069336, "margin_dpo/margin_std": 160.0266876220703, "step": 192 }, { "KL/chosen_KL_mean": -140.22830200195312, "KL/mean": -168.90858459472656, "KL/rejected_KL_mean": -197.5888671875, "KL/std": 121.005859375, "epoch": 0.2834067547723935, "fcm_dpo/beta": 0.0036203060299158096, "fcm_dpo/delta": 0.0363609679043293, "fcm_dpo/margin": 57.36055374145508, "fcm_dpo/q_t": 0.4501846432685852, "grad_norm": 24.104490280151367, "learning_rate": 4.5180069639630236e-07, "logits/chosen": -0.3034767210483551, "logits/rejected": -0.29830050468444824, "logps/chosen": -200.41879272460938, "logps/ref_chosen": -60.19049835205078, "logps/ref_rejected": -76.40755462646484, "logps/rejected": -273.9964294433594, "loss": 1.2644, "margin_dpo/margin_mean": 57.36054992675781, "margin_dpo/margin_std": 154.25485229492188, "step": 193 }, { "KL/chosen_KL_mean": -87.32644653320312, "KL/mean": -133.92543029785156, "KL/rejected_KL_mean": -180.52444458007812, "KL/std": 99.52772521972656, "epoch": 0.28487518355359764, "fcm_dpo/beta": 0.0036298225168138742, "fcm_dpo/delta": 0.011991962790489197, "fcm_dpo/margin": 93.19798278808594, "fcm_dpo/q_t": 0.41966894268989563, "grad_norm": 15.164206504821777, "learning_rate": 4.510405240853854e-07, "logits/chosen": -0.19916538894176483, "logits/rejected": -0.18301814794540405, "logps/chosen": -125.16681671142578, "logps/ref_chosen": -37.84037399291992, "logps/ref_rejected": -60.684783935546875, "logps/rejected": -241.20921325683594, "loss": 1.115, "margin_dpo/margin_mean": 93.19798278808594, "margin_dpo/margin_std": 109.33871459960938, "step": 194 }, { "KL/chosen_KL_mean": -137.87490844726562, "KL/mean": -187.89657592773438, "KL/rejected_KL_mean": -237.91827392578125, "KL/std": 124.3742904663086, "epoch": 0.28634361233480177, "fcm_dpo/beta": 0.0036336728371679783, "fcm_dpo/delta": -0.014039820060133934, "fcm_dpo/margin": 100.0433578491211, "fcm_dpo/q_t": 0.41508179903030396, "grad_norm": 24.048799514770508, "learning_rate": 4.5027505416968985e-07, "logits/chosen": -0.23971587419509888, "logits/rejected": -0.25928401947021484, "logps/chosen": -192.7664794921875, "logps/ref_chosen": -54.891571044921875, "logps/ref_rejected": -96.77095794677734, "logps/rejected": -334.689208984375, "loss": 1.1062, "margin_dpo/margin_mean": 100.0433578491211, "margin_dpo/margin_std": 127.86747741699219, "step": 195 }, { "KL/chosen_KL_mean": -107.48291015625, "KL/mean": -164.11669921875, "KL/rejected_KL_mean": -220.75048828125, "KL/std": 124.06842041015625, "epoch": 0.2878120411160059, "fcm_dpo/beta": 0.0035826051607728004, "fcm_dpo/delta": -0.05893465504050255, "fcm_dpo/margin": 113.26756286621094, "fcm_dpo/q_t": 0.4072193205356598, "grad_norm": 15.825559616088867, "learning_rate": 4.495043068200599e-07, "logits/chosen": -0.28464287519454956, "logits/rejected": -0.2717984616756439, "logps/chosen": -160.7281494140625, "logps/ref_chosen": -53.245243072509766, "logps/ref_rejected": -76.05294799804688, "logps/rejected": -296.8034362792969, "loss": 1.0916, "margin_dpo/margin_mean": 113.2675552368164, "margin_dpo/margin_std": 148.8846435546875, "step": 196 }, { "KL/chosen_KL_mean": -110.76636505126953, "KL/mean": -154.2803955078125, "KL/rejected_KL_mean": -197.79441833496094, "KL/std": 108.77426147460938, "epoch": 0.28928046989721, "fcm_dpo/beta": 0.003618548158556223, "fcm_dpo/delta": 0.03597838059067726, "fcm_dpo/margin": 87.02806091308594, "fcm_dpo/q_t": 0.4268850088119507, "grad_norm": 17.07707977294922, "learning_rate": 4.4872830234640493e-07, "logits/chosen": -0.2548871636390686, "logits/rejected": -0.2504875063896179, "logps/chosen": -171.18670654296875, "logps/ref_chosen": -60.42033386230469, "logps/ref_rejected": -77.20890808105469, "logps/rejected": -275.0033264160156, "loss": 1.1446, "margin_dpo/margin_mean": 87.02806091308594, "margin_dpo/margin_std": 124.56130981445312, "step": 197 }, { "KL/chosen_KL_mean": -123.58761596679688, "KL/mean": -179.9723663330078, "KL/rejected_KL_mean": -236.3571319580078, "KL/std": 136.68484497070312, "epoch": 0.2907488986784141, "fcm_dpo/beta": 0.0035852163564413786, "fcm_dpo/delta": -0.056587982922792435, "fcm_dpo/margin": 112.76952362060547, "fcm_dpo/q_t": 0.40880346298217773, "grad_norm": 18.692174911499023, "learning_rate": 4.479470611971645e-07, "logits/chosen": -0.31638431549072266, "logits/rejected": -0.3171602487564087, "logps/chosen": -178.62380981445312, "logps/ref_chosen": -55.03618621826172, "logps/ref_rejected": -97.24325561523438, "logps/rejected": -333.60040283203125, "loss": 1.0918, "margin_dpo/margin_mean": 112.76952362060547, "margin_dpo/margin_std": 154.90292358398438, "step": 198 }, { "KL/chosen_KL_mean": -118.904541015625, "KL/mean": -175.37063598632812, "KL/rejected_KL_mean": -231.8367156982422, "KL/std": 124.37785339355469, "epoch": 0.2922173274596182, "fcm_dpo/beta": 0.003529233392328024, "fcm_dpo/delta": -0.051382582634687424, "fcm_dpo/margin": 112.93217468261719, "fcm_dpo/q_t": 0.40841788053512573, "grad_norm": 21.199235916137695, "learning_rate": 4.471606039587695e-07, "logits/chosen": -0.28893113136291504, "logits/rejected": -0.27144867181777954, "logps/chosen": -175.73336791992188, "logps/ref_chosen": -56.828826904296875, "logps/ref_rejected": -84.64820861816406, "logps/rejected": -316.48492431640625, "loss": 1.0975, "margin_dpo/margin_mean": 112.93217468261719, "margin_dpo/margin_std": 153.85972595214844, "step": 199 }, { "KL/chosen_KL_mean": -116.62962341308594, "KL/mean": -172.3009033203125, "KL/rejected_KL_mean": -227.97215270996094, "KL/std": 133.75917053222656, "epoch": 0.2936857562408223, "fcm_dpo/beta": 0.0035058055073022842, "fcm_dpo/delta": -0.04227417707443237, "fcm_dpo/margin": 111.34254455566406, "fcm_dpo/q_t": 0.41175031661987305, "grad_norm": 18.752197265625, "learning_rate": 4.4636895135509966e-07, "logits/chosen": -0.24854665994644165, "logits/rejected": -0.23521234095096588, "logps/chosen": -169.69668579101562, "logps/ref_chosen": -53.06706237792969, "logps/ref_rejected": -80.60843658447266, "logps/rejected": -308.5805969238281, "loss": 1.1174, "margin_dpo/margin_mean": 111.34254455566406, "margin_dpo/margin_std": 172.01788330078125, "step": 200 }, { "KL/chosen_KL_mean": -122.89008331298828, "KL/mean": -179.06658935546875, "KL/rejected_KL_mean": -235.24310302734375, "KL/std": 143.4940948486328, "epoch": 0.29515418502202645, "fcm_dpo/beta": 0.003493840806186199, "fcm_dpo/delta": -0.04429582878947258, "fcm_dpo/margin": 112.35299682617188, "fcm_dpo/q_t": 0.41010570526123047, "grad_norm": 18.05536651611328, "learning_rate": 4.455721242469372e-07, "logits/chosen": -0.3345472812652588, "logits/rejected": -0.3335646986961365, "logps/chosen": -198.29229736328125, "logps/ref_chosen": -75.4022216796875, "logps/ref_rejected": -114.80821990966797, "logps/rejected": -350.05133056640625, "loss": 1.107, "margin_dpo/margin_mean": 112.35299682617188, "margin_dpo/margin_std": 162.2955322265625, "step": 201 }, { "KL/chosen_KL_mean": -127.64757537841797, "KL/mean": -166.51829528808594, "KL/rejected_KL_mean": -205.38900756835938, "KL/std": 123.735107421875, "epoch": 0.2966226138032305, "fcm_dpo/beta": 0.0035203341394662857, "fcm_dpo/delta": 0.07842127978801727, "fcm_dpo/margin": 77.74143981933594, "fcm_dpo/q_t": 0.4383637309074402, "grad_norm": 18.04079246520996, "learning_rate": 4.4477014363141755e-07, "logits/chosen": -0.2855903208255768, "logits/rejected": -0.3018990755081177, "logps/chosen": -177.7489013671875, "logps/ref_chosen": -50.101318359375, "logps/ref_rejected": -86.98503112792969, "logps/rejected": -292.3740234375, "loss": 1.207, "margin_dpo/margin_mean": 77.74143981933594, "margin_dpo/margin_std": 161.36936950683594, "step": 202 }, { "KL/chosen_KL_mean": -126.76716613769531, "KL/mean": -176.1123504638672, "KL/rejected_KL_mean": -225.45751953125, "KL/std": 128.75897216796875, "epoch": 0.29809104258443464, "fcm_dpo/beta": 0.0035307754296809435, "fcm_dpo/delta": 0.0015939096920192242, "fcm_dpo/margin": 98.69035339355469, "fcm_dpo/q_t": 0.4193652272224426, "grad_norm": 20.859642028808594, "learning_rate": 4.439630306414758e-07, "logits/chosen": -0.292889267206192, "logits/rejected": -0.28478172421455383, "logps/chosen": -187.37686157226562, "logps/ref_chosen": -60.60969543457031, "logps/ref_rejected": -85.89596557617188, "logps/rejected": -311.353515625, "loss": 1.1261, "margin_dpo/margin_mean": 98.69036102294922, "margin_dpo/margin_std": 141.41165161132812, "step": 203 }, { "KL/chosen_KL_mean": -141.1234130859375, "KL/mean": -186.42263793945312, "KL/rejected_KL_mean": -231.7218475341797, "KL/std": 140.6131591796875, "epoch": 0.29955947136563876, "fcm_dpo/beta": 0.0035480433143675327, "fcm_dpo/delta": 0.029507920145988464, "fcm_dpo/margin": 90.59844970703125, "fcm_dpo/q_t": 0.4280925393104553, "grad_norm": 23.841333389282227, "learning_rate": 4.431508065452897e-07, "logits/chosen": -0.39130985736846924, "logits/rejected": -0.3560817539691925, "logps/chosen": -221.28836059570312, "logps/ref_chosen": -80.16496276855469, "logps/ref_rejected": -87.69590759277344, "logps/rejected": -319.4177551269531, "loss": 1.1717, "margin_dpo/margin_mean": 90.59844970703125, "margin_dpo/margin_std": 165.41769409179688, "step": 204 }, { "KL/chosen_KL_mean": -137.06591796875, "KL/mean": -196.98837280273438, "KL/rejected_KL_mean": -256.910888671875, "KL/std": 140.85963439941406, "epoch": 0.3010279001468429, "fcm_dpo/beta": 0.0034892673138529062, "fcm_dpo/delta": -0.07289686799049377, "fcm_dpo/margin": 119.84494018554688, "fcm_dpo/q_t": 0.4026746153831482, "grad_norm": 19.944299697875977, "learning_rate": 4.4233349274571974e-07, "logits/chosen": -0.284223735332489, "logits/rejected": -0.25634074211120605, "logps/chosen": -196.45065307617188, "logps/ref_chosen": -59.384735107421875, "logps/ref_rejected": -85.12505340576172, "logps/rejected": -342.0359191894531, "loss": 1.0855, "margin_dpo/margin_mean": 119.84494018554688, "margin_dpo/margin_std": 155.21694946289062, "step": 205 }, { "KL/chosen_KL_mean": -124.25179290771484, "KL/mean": -189.01766967773438, "KL/rejected_KL_mean": -253.78358459472656, "KL/std": 130.41976928710938, "epoch": 0.302496328928047, "fcm_dpo/beta": 0.003440842032432556, "fcm_dpo/delta": -0.10062892735004425, "fcm_dpo/margin": 129.5317840576172, "fcm_dpo/q_t": 0.39475929737091064, "grad_norm": 21.66822624206543, "learning_rate": 4.415111107797445e-07, "logits/chosen": -0.2068089246749878, "logits/rejected": -0.20935988426208496, "logps/chosen": -171.21629333496094, "logps/ref_chosen": -46.964500427246094, "logps/ref_rejected": -98.9534912109375, "logps/rejected": -352.737060546875, "loss": 1.0468, "margin_dpo/margin_mean": 129.5317840576172, "margin_dpo/margin_std": 142.85910034179688, "step": 206 }, { "KL/chosen_KL_mean": -115.6485595703125, "KL/mean": -189.80975341796875, "KL/rejected_KL_mean": -263.9709777832031, "KL/std": 151.3675537109375, "epoch": 0.3039647577092511, "fcm_dpo/beta": 0.0033577182330191135, "fcm_dpo/delta": -0.15566033124923706, "fcm_dpo/margin": 148.32240295410156, "fcm_dpo/q_t": 0.3879104256629944, "grad_norm": 23.90052604675293, "learning_rate": 4.4068368231789365e-07, "logits/chosen": -0.3293609023094177, "logits/rejected": -0.30513134598731995, "logps/chosen": -171.70481872558594, "logps/ref_chosen": -56.05625915527344, "logps/ref_rejected": -84.44779968261719, "logps/rejected": -348.41876220703125, "loss": 1.0267, "margin_dpo/margin_mean": 148.32241821289062, "margin_dpo/margin_std": 176.73928833007812, "step": 207 }, { "KL/chosen_KL_mean": -176.27822875976562, "KL/mean": -233.8270263671875, "KL/rejected_KL_mean": -291.3758544921875, "KL/std": 145.08538818359375, "epoch": 0.3054331864904552, "fcm_dpo/beta": 0.0033049010671675205, "fcm_dpo/delta": -0.03175726160407066, "fcm_dpo/margin": 115.09761047363281, "fcm_dpo/q_t": 0.41257303953170776, "grad_norm": 28.714868545532227, "learning_rate": 4.398512291636768e-07, "logits/chosen": -0.3144323527812958, "logits/rejected": -0.2976739704608917, "logps/chosen": -243.34584045410156, "logps/ref_chosen": -67.06761169433594, "logps/ref_rejected": -94.28689575195312, "logps/rejected": -385.6627502441406, "loss": 1.1231, "margin_dpo/margin_mean": 115.09761047363281, "margin_dpo/margin_std": 178.362548828125, "step": 208 }, { "KL/chosen_KL_mean": -148.74696350097656, "KL/mean": -200.42440795898438, "KL/rejected_KL_mean": -252.1018524169922, "KL/std": 131.3517303466797, "epoch": 0.3069016152716593, "fcm_dpo/beta": 0.0033130061347037554, "fcm_dpo/delta": 0.007838414050638676, "fcm_dpo/margin": 103.35487365722656, "fcm_dpo/q_t": 0.42111438512802124, "grad_norm": 22.145084381103516, "learning_rate": 4.3901377325300857e-07, "logits/chosen": -0.20952649414539337, "logits/rejected": -0.1976684331893921, "logps/chosen": -204.92864990234375, "logps/ref_chosen": -56.18169403076172, "logps/ref_rejected": -80.94152069091797, "logps/rejected": -333.0433654785156, "loss": 1.1463, "margin_dpo/margin_mean": 103.35487365722656, "margin_dpo/margin_std": 165.46578979492188, "step": 209 }, { "KL/chosen_KL_mean": -131.93817138671875, "KL/mean": -192.3253173828125, "KL/rejected_KL_mean": -252.71246337890625, "KL/std": 135.9736328125, "epoch": 0.30837004405286345, "fcm_dpo/beta": 0.0032999245449900627, "fcm_dpo/delta": -0.05067028850317001, "fcm_dpo/margin": 120.77430725097656, "fcm_dpo/q_t": 0.4087422788143158, "grad_norm": 21.08250617980957, "learning_rate": 4.381713366536311e-07, "logits/chosen": -0.25526225566864014, "logits/rejected": -0.2489766627550125, "logps/chosen": -178.30999755859375, "logps/ref_chosen": -46.371822357177734, "logps/ref_rejected": -76.68162536621094, "logps/rejected": -329.39410400390625, "loss": 1.0971, "margin_dpo/margin_mean": 120.7743148803711, "margin_dpo/margin_std": 164.735595703125, "step": 210 }, { "KL/chosen_KL_mean": -187.6256103515625, "KL/mean": -238.52076721191406, "KL/rejected_KL_mean": -289.4158935546875, "KL/std": 153.42758178710938, "epoch": 0.30983847283406757, "fcm_dpo/beta": 0.0032891465816646814, "fcm_dpo/delta": 0.015694769099354744, "fcm_dpo/margin": 101.79029083251953, "fcm_dpo/q_t": 0.42646709084510803, "grad_norm": 28.087804794311523, "learning_rate": 4.373239415645323e-07, "logits/chosen": -0.31508252024650574, "logits/rejected": -0.28154420852661133, "logps/chosen": -266.5579833984375, "logps/ref_chosen": -78.93235778808594, "logps/ref_rejected": -86.82098388671875, "logps/rejected": -376.23687744140625, "loss": 1.1642, "margin_dpo/margin_mean": 101.79029083251953, "margin_dpo/margin_std": 181.40985107421875, "step": 211 }, { "KL/chosen_KL_mean": -155.69639587402344, "KL/mean": -228.75802612304688, "KL/rejected_KL_mean": -301.81964111328125, "KL/std": 165.00823974609375, "epoch": 0.31130690161527164, "fcm_dpo/beta": 0.003207254456356168, "fcm_dpo/delta": -0.12612880766391754, "fcm_dpo/margin": 146.12322998046875, "fcm_dpo/q_t": 0.39321210980415344, "grad_norm": 25.95430564880371, "learning_rate": 4.3647161031536086e-07, "logits/chosen": -0.2920724153518677, "logits/rejected": -0.2858331799507141, "logps/chosen": -213.89340209960938, "logps/ref_chosen": -58.19701385498047, "logps/ref_rejected": -103.05785369873047, "logps/rejected": -404.87750244140625, "loss": 1.0603, "margin_dpo/margin_mean": 146.12326049804688, "margin_dpo/margin_std": 185.58700561523438, "step": 212 }, { "KL/chosen_KL_mean": -146.40249633789062, "KL/mean": -214.92503356933594, "KL/rejected_KL_mean": -283.44757080078125, "KL/std": 141.89804077148438, "epoch": 0.31277533039647576, "fcm_dpo/beta": 0.0031565751414746046, "fcm_dpo/delta": -0.08675891906023026, "fcm_dpo/margin": 137.04510498046875, "fcm_dpo/q_t": 0.39976924657821655, "grad_norm": 25.628156661987305, "learning_rate": 4.3561436536583774e-07, "logits/chosen": -0.30385902523994446, "logits/rejected": -0.2821686863899231, "logps/chosen": -213.91519165039062, "logps/ref_chosen": -67.51271057128906, "logps/ref_rejected": -93.91471862792969, "logps/rejected": -377.3623046875, "loss": 1.067, "margin_dpo/margin_mean": 137.04510498046875, "margin_dpo/margin_std": 169.2763671875, "step": 213 }, { "KL/chosen_KL_mean": -128.19113159179688, "KL/mean": -190.6093292236328, "KL/rejected_KL_mean": -253.02752685546875, "KL/std": 141.2437744140625, "epoch": 0.3142437591776799, "fcm_dpo/beta": 0.0031324918381869793, "fcm_dpo/delta": -0.04271193593740463, "fcm_dpo/margin": 124.83644104003906, "fcm_dpo/q_t": 0.4102705121040344, "grad_norm": 23.725318908691406, "learning_rate": 4.3475222930516473e-07, "logits/chosen": -0.21322600543498993, "logits/rejected": -0.2185087352991104, "logps/chosen": -169.7960205078125, "logps/ref_chosen": -41.604888916015625, "logps/ref_rejected": -77.51741027832031, "logps/rejected": -330.5449523925781, "loss": 1.0966, "margin_dpo/margin_mean": 124.83644104003906, "margin_dpo/margin_std": 166.6226806640625, "step": 214 }, { "KL/chosen_KL_mean": -150.10049438476562, "KL/mean": -216.70033264160156, "KL/rejected_KL_mean": -283.3001708984375, "KL/std": 148.46713256835938, "epoch": 0.315712187958884, "fcm_dpo/beta": 0.0030872428324073553, "fcm_dpo/delta": -0.06415253132581711, "fcm_dpo/margin": 133.19969177246094, "fcm_dpo/q_t": 0.40401768684387207, "grad_norm": 25.86069107055664, "learning_rate": 4.3388522485142885e-07, "logits/chosen": -0.2457331418991089, "logits/rejected": -0.23719845712184906, "logps/chosen": -203.3797607421875, "logps/ref_chosen": -53.279266357421875, "logps/ref_rejected": -89.96464538574219, "logps/rejected": -373.26483154296875, "loss": 1.0699, "margin_dpo/margin_mean": 133.19969177246094, "margin_dpo/margin_std": 153.16122436523438, "step": 215 }, { "KL/chosen_KL_mean": -155.20941162109375, "KL/mean": -220.2621612548828, "KL/rejected_KL_mean": -285.3149108886719, "KL/std": 152.76821899414062, "epoch": 0.31718061674008813, "fcm_dpo/beta": 0.0030720389913767576, "fcm_dpo/delta": -0.05181992053985596, "fcm_dpo/margin": 130.10549926757812, "fcm_dpo/q_t": 0.40874841809272766, "grad_norm": 24.124347686767578, "learning_rate": 4.330133748510036e-07, "logits/chosen": -0.24108392000198364, "logits/rejected": -0.22588925063610077, "logps/chosen": -204.09719848632812, "logps/ref_chosen": -48.887794494628906, "logps/ref_rejected": -77.19892883300781, "logps/rejected": -362.51385498046875, "loss": 1.105, "margin_dpo/margin_mean": 130.1055145263672, "margin_dpo/margin_std": 188.24560546875, "step": 216 }, { "KL/chosen_KL_mean": -166.11288452148438, "KL/mean": -242.55723571777344, "KL/rejected_KL_mean": -319.0015869140625, "KL/std": 159.12432861328125, "epoch": 0.3186490455212922, "fcm_dpo/beta": 0.002996720839291811, "fcm_dpo/delta": -0.1136351153254509, "fcm_dpo/margin": 152.88873291015625, "fcm_dpo/q_t": 0.39497053623199463, "grad_norm": 21.351091384887695, "learning_rate": 4.3213670227794757e-07, "logits/chosen": -0.20837268233299255, "logits/rejected": -0.2034785896539688, "logps/chosen": -215.95819091796875, "logps/ref_chosen": -49.845306396484375, "logps/ref_rejected": -100.07832336425781, "logps/rejected": -419.0799255371094, "loss": 1.046, "margin_dpo/margin_mean": 152.88873291015625, "margin_dpo/margin_std": 179.6495361328125, "step": 217 }, { "KL/chosen_KL_mean": -172.87002563476562, "KL/mean": -231.294677734375, "KL/rejected_KL_mean": -289.71929931640625, "KL/std": 157.75628662109375, "epoch": 0.3201174743024963, "fcm_dpo/beta": 0.0029808084946125746, "fcm_dpo/delta": 0.0017072586342692375, "fcm_dpo/margin": 116.84925842285156, "fcm_dpo/q_t": 0.42078712582588196, "grad_norm": 20.616666793823242, "learning_rate": 4.3125523023339815e-07, "logits/chosen": -0.23788803815841675, "logits/rejected": -0.23454323410987854, "logps/chosen": -231.44671630859375, "logps/ref_chosen": -58.576683044433594, "logps/ref_rejected": -87.84639739990234, "logps/rejected": -377.5657043457031, "loss": 1.141, "margin_dpo/margin_mean": 116.8492660522461, "margin_dpo/margin_std": 185.65829467773438, "step": 218 }, { "KL/chosen_KL_mean": -183.26531982421875, "KL/mean": -234.03045654296875, "KL/rejected_KL_mean": -284.79559326171875, "KL/std": 163.48797607421875, "epoch": 0.32158590308370044, "fcm_dpo/beta": 0.002978936769068241, "fcm_dpo/delta": -0.023041052743792534, "fcm_dpo/margin": 101.53028869628906, "fcm_dpo/q_t": 0.43124303221702576, "grad_norm": 27.521127700805664, "learning_rate": 4.303689819449636e-07, "logits/chosen": -0.30178552865982056, "logits/rejected": -0.2995387315750122, "logps/chosen": -244.34918212890625, "logps/ref_chosen": -61.083858489990234, "logps/ref_rejected": -85.83042907714844, "logps/rejected": -370.62603759765625, "loss": 1.194, "margin_dpo/margin_mean": 101.53028869628906, "margin_dpo/margin_std": 203.3063507080078, "step": 219 }, { "KL/chosen_KL_mean": -206.34454345703125, "KL/mean": -247.07632446289062, "KL/rejected_KL_mean": -287.80810546875, "KL/std": 141.79627990722656, "epoch": 0.32305433186490456, "fcm_dpo/beta": 0.0029919049702584743, "fcm_dpo/delta": 0.03325273096561432, "fcm_dpo/margin": 81.4636001586914, "fcm_dpo/q_t": 0.44203633069992065, "grad_norm": 26.6117000579834, "learning_rate": 4.2947798076611047e-07, "logits/chosen": -0.2598545551300049, "logits/rejected": -0.23767119646072388, "logps/chosen": -276.3758239746094, "logps/ref_chosen": -70.03128051757812, "logps/ref_rejected": -87.68551635742188, "logps/rejected": -375.49365234375, "loss": 1.2032, "margin_dpo/margin_mean": 81.4636001586914, "margin_dpo/margin_std": 145.33209228515625, "step": 220 }, { "KL/chosen_KL_mean": -169.64871215820312, "KL/mean": -264.8304748535156, "KL/rejected_KL_mean": -360.01226806640625, "KL/std": 177.01751708984375, "epoch": 0.3245227606461087, "fcm_dpo/beta": 0.0029050116427242756, "fcm_dpo/delta": -0.2145058810710907, "fcm_dpo/margin": 190.363525390625, "fcm_dpo/q_t": 0.3714696168899536, "grad_norm": 22.261398315429688, "learning_rate": 4.285822501755485e-07, "logits/chosen": -0.27296262979507446, "logits/rejected": -0.27987274527549744, "logps/chosen": -221.8034210205078, "logps/ref_chosen": -52.15470886230469, "logps/ref_rejected": -106.46768188476562, "logps/rejected": -466.47991943359375, "loss": 0.9744, "margin_dpo/margin_mean": 190.363525390625, "margin_dpo/margin_std": 182.53524780273438, "step": 221 }, { "KL/chosen_KL_mean": -172.16732788085938, "KL/mean": -242.41677856445312, "KL/rejected_KL_mean": -312.6662292480469, "KL/std": 161.145263671875, "epoch": 0.32599118942731276, "fcm_dpo/beta": 0.002854538382962346, "fcm_dpo/delta": -0.05328700318932533, "fcm_dpo/margin": 140.4989013671875, "fcm_dpo/q_t": 0.4074173867702484, "grad_norm": 18.149749755859375, "learning_rate": 4.276818137766118e-07, "logits/chosen": -0.3163185119628906, "logits/rejected": -0.3211904764175415, "logps/chosen": -233.138427734375, "logps/ref_chosen": -60.971099853515625, "logps/ref_rejected": -100.00115203857422, "logps/rejected": -412.6673583984375, "loss": 1.0889, "margin_dpo/margin_mean": 140.4989013671875, "margin_dpo/margin_std": 182.004150390625, "step": 222 }, { "KL/chosen_KL_mean": -180.6187286376953, "KL/mean": -239.54754638671875, "KL/rejected_KL_mean": -298.4763488769531, "KL/std": 154.41107177734375, "epoch": 0.3274596182085169, "fcm_dpo/beta": 0.002839556895196438, "fcm_dpo/delta": 0.01586122065782547, "fcm_dpo/margin": 117.85763549804688, "fcm_dpo/q_t": 0.4231345057487488, "grad_norm": 21.138545989990234, "learning_rate": 4.2677669529663686e-07, "logits/chosen": -0.24469372630119324, "logits/rejected": -0.24137765169143677, "logps/chosen": -233.25930786132812, "logps/ref_chosen": -52.64057540893555, "logps/ref_rejected": -82.82502746582031, "logps/rejected": -381.3013916015625, "loss": 1.1542, "margin_dpo/margin_mean": 117.85763549804688, "margin_dpo/margin_std": 198.8590545654297, "step": 223 }, { "KL/chosen_KL_mean": -159.38333129882812, "KL/mean": -226.15509033203125, "KL/rejected_KL_mean": -292.9268493652344, "KL/std": 172.1931610107422, "epoch": 0.328928046989721, "fcm_dpo/beta": 0.0028151879087090492, "fcm_dpo/delta": -0.02796056494116783, "fcm_dpo/margin": 133.54351806640625, "fcm_dpo/q_t": 0.4151793122291565, "grad_norm": 22.848051071166992, "learning_rate": 4.2586691858633747e-07, "logits/chosen": -0.3147471845149994, "logits/rejected": -0.2979544997215271, "logps/chosen": -207.97874450683594, "logps/ref_chosen": -48.59541320800781, "logps/ref_rejected": -77.11648559570312, "logps/rejected": -370.0433349609375, "loss": 1.1202, "margin_dpo/margin_mean": 133.54351806640625, "margin_dpo/margin_std": 199.27239990234375, "step": 224 }, { "KL/chosen_KL_mean": -179.8327178955078, "KL/mean": -258.5877990722656, "KL/rejected_KL_mean": -337.3428955078125, "KL/std": 164.62838745117188, "epoch": 0.3303964757709251, "fcm_dpo/beta": 0.002788371406495571, "fcm_dpo/delta": -0.09335584938526154, "fcm_dpo/margin": 157.5101776123047, "fcm_dpo/q_t": 0.3994441330432892, "grad_norm": 19.650955200195312, "learning_rate": 4.249525076191759e-07, "logits/chosen": -0.3059282898902893, "logits/rejected": -0.29627132415771484, "logps/chosen": -237.83319091796875, "logps/ref_chosen": -58.000465393066406, "logps/ref_rejected": -99.90291595458984, "logps/rejected": -437.2458190917969, "loss": 1.0678, "margin_dpo/margin_mean": 157.51019287109375, "margin_dpo/margin_std": 203.22439575195312, "step": 225 }, { "KL/chosen_KL_mean": -156.86215209960938, "KL/mean": -220.01309204101562, "KL/rejected_KL_mean": -283.16400146484375, "KL/std": 162.33700561523438, "epoch": 0.33186490455212925, "fcm_dpo/beta": 0.00276209507137537, "fcm_dpo/delta": 0.0005985237658023834, "fcm_dpo/margin": 126.30184936523438, "fcm_dpo/q_t": 0.42073529958724976, "grad_norm": 28.77495765686035, "learning_rate": 4.2403348649073167e-07, "logits/chosen": -0.36503708362579346, "logits/rejected": -0.3280119001865387, "logps/chosen": -215.76095581054688, "logps/ref_chosen": -58.898799896240234, "logps/ref_rejected": -78.68775939941406, "logps/rejected": -361.85174560546875, "loss": 1.1334, "margin_dpo/margin_mean": 126.30184936523438, "margin_dpo/margin_std": 188.94320678710938, "step": 226 }, { "KL/chosen_KL_mean": -186.6765899658203, "KL/mean": -266.1453857421875, "KL/rejected_KL_mean": -345.6142272949219, "KL/std": 186.829833984375, "epoch": 0.3333333333333333, "fcm_dpo/beta": 0.002731041982769966, "fcm_dpo/delta": -0.08844669163227081, "fcm_dpo/margin": 158.93765258789062, "fcm_dpo/q_t": 0.399799644947052, "grad_norm": 20.401033401489258, "learning_rate": 4.2310987941806615e-07, "logits/chosen": -0.35669684410095215, "logits/rejected": -0.34606364369392395, "logps/chosen": -245.74876403808594, "logps/ref_chosen": -59.072181701660156, "logps/ref_rejected": -99.41236877441406, "logps/rejected": -445.026611328125, "loss": 1.0635, "margin_dpo/margin_mean": 158.93765258789062, "margin_dpo/margin_std": 194.41900634765625, "step": 227 }, { "KL/chosen_KL_mean": -194.19752502441406, "KL/mean": -249.16688537597656, "KL/rejected_KL_mean": -304.1362609863281, "KL/std": 154.22523498535156, "epoch": 0.33480176211453744, "fcm_dpo/beta": 0.002752040047198534, "fcm_dpo/delta": 0.048819925636053085, "fcm_dpo/margin": 109.93873596191406, "fcm_dpo/q_t": 0.42963525652885437, "grad_norm": 22.93690299987793, "learning_rate": 4.2218171073908463e-07, "logits/chosen": -0.29892176389694214, "logits/rejected": -0.28179436922073364, "logps/chosen": -260.08880615234375, "logps/ref_chosen": -65.89128875732422, "logps/ref_rejected": -91.04875183105469, "logps/rejected": -395.18499755859375, "loss": 1.1714, "margin_dpo/margin_mean": 109.93873596191406, "margin_dpo/margin_std": 189.5914764404297, "step": 228 }, { "KL/chosen_KL_mean": -204.4345703125, "KL/mean": -262.7249755859375, "KL/rejected_KL_mean": -321.015380859375, "KL/std": 179.88668823242188, "epoch": 0.33627019089574156, "fcm_dpo/beta": 0.0027676185127347708, "fcm_dpo/delta": 0.028249580413103104, "fcm_dpo/margin": 116.58082580566406, "fcm_dpo/q_t": 0.4248887300491333, "grad_norm": 27.2257137298584, "learning_rate": 4.212490049118951e-07, "logits/chosen": -0.41251590847969055, "logits/rejected": -0.38618093729019165, "logps/chosen": -275.14093017578125, "logps/ref_chosen": -70.70637512207031, "logps/ref_rejected": -84.52741241455078, "logps/rejected": -405.54278564453125, "loss": 1.1518, "margin_dpo/margin_mean": 116.58082580566406, "margin_dpo/margin_std": 186.08026123046875, "step": 229 }, { "KL/chosen_KL_mean": -156.17324829101562, "KL/mean": -247.1909942626953, "KL/rejected_KL_mean": -338.2087707519531, "KL/std": 167.25390625, "epoch": 0.3377386196769457, "fcm_dpo/beta": 0.0027067074552178383, "fcm_dpo/delta": -0.1502915769815445, "fcm_dpo/margin": 182.03550720214844, "fcm_dpo/q_t": 0.38420677185058594, "grad_norm": 27.937719345092773, "learning_rate": 4.203117865141635e-07, "logits/chosen": -0.30553507804870605, "logits/rejected": -0.3124315142631531, "logps/chosen": -195.45526123046875, "logps/ref_chosen": -39.282005310058594, "logps/ref_rejected": -85.62191009521484, "logps/rejected": -423.8306884765625, "loss": 1.0113, "margin_dpo/margin_mean": 182.0355224609375, "margin_dpo/margin_std": 184.9110107421875, "step": 230 }, { "KL/chosen_KL_mean": -176.39178466796875, "KL/mean": -238.8753662109375, "KL/rejected_KL_mean": -301.35894775390625, "KL/std": 148.606201171875, "epoch": 0.3392070484581498, "fcm_dpo/beta": 0.0026946812868118286, "fcm_dpo/delta": 0.013711929321289062, "fcm_dpo/margin": 124.96714782714844, "fcm_dpo/q_t": 0.42264679074287415, "grad_norm": 23.682451248168945, "learning_rate": 4.1937008024246625e-07, "logits/chosen": -0.38201549649238586, "logits/rejected": -0.35523295402526855, "logps/chosen": -239.66824340820312, "logps/ref_chosen": -63.27644348144531, "logps/ref_rejected": -74.1239013671875, "logps/rejected": -375.48284912109375, "loss": 1.1305, "margin_dpo/margin_mean": 124.96714782714844, "margin_dpo/margin_std": 175.9566192626953, "step": 231 }, { "KL/chosen_KL_mean": -205.3118133544922, "KL/mean": -255.06503295898438, "KL/rejected_KL_mean": -304.8182678222656, "KL/std": 167.1653289794922, "epoch": 0.3406754772393539, "fcm_dpo/beta": 0.002725243801251054, "fcm_dpo/delta": 0.0810902863740921, "fcm_dpo/margin": 99.50645446777344, "fcm_dpo/q_t": 0.43814921379089355, "grad_norm": 21.643369674682617, "learning_rate": 4.1842391091163933e-07, "logits/chosen": -0.39978110790252686, "logits/rejected": -0.3793877959251404, "logps/chosen": -276.0605773925781, "logps/ref_chosen": -70.74876403808594, "logps/ref_rejected": -83.97706604003906, "logps/rejected": -388.79534912109375, "loss": 1.1853, "margin_dpo/margin_mean": 99.50645446777344, "margin_dpo/margin_std": 172.71176147460938, "step": 232 }, { "KL/chosen_KL_mean": -179.78868103027344, "KL/mean": -259.36474609375, "KL/rejected_KL_mean": -338.9408264160156, "KL/std": 176.54042053222656, "epoch": 0.342143906020558, "fcm_dpo/beta": 0.002710646251216531, "fcm_dpo/delta": -0.08498845994472504, "fcm_dpo/margin": 159.15216064453125, "fcm_dpo/q_t": 0.4036695659160614, "grad_norm": 24.3404483795166, "learning_rate": 4.174733034541245e-07, "logits/chosen": -0.3906532824039459, "logits/rejected": -0.3928545117378235, "logps/chosen": -234.67161560058594, "logps/ref_chosen": -54.8829345703125, "logps/ref_rejected": -107.4800796508789, "logps/rejected": -446.4208984375, "loss": 1.0915, "margin_dpo/margin_mean": 159.15216064453125, "margin_dpo/margin_std": 231.7545166015625, "step": 233 }, { "KL/chosen_KL_mean": -166.94924926757812, "KL/mean": -249.14198303222656, "KL/rejected_KL_mean": -331.334716796875, "KL/std": 158.32025146484375, "epoch": 0.3436123348017621, "fcm_dpo/beta": 0.0026394538581371307, "fcm_dpo/delta": -0.08937112987041473, "fcm_dpo/margin": 164.385498046875, "fcm_dpo/q_t": 0.3987475037574768, "grad_norm": 22.889997482299805, "learning_rate": 4.165182829193126e-07, "logits/chosen": -0.3914503753185272, "logits/rejected": -0.41639894247055054, "logps/chosen": -211.04376220703125, "logps/ref_chosen": -44.094520568847656, "logps/ref_rejected": -100.00663757324219, "logps/rejected": -431.34136962890625, "loss": 1.0537, "margin_dpo/margin_mean": 164.385498046875, "margin_dpo/margin_std": 177.35736083984375, "step": 234 }, { "KL/chosen_KL_mean": -203.8853302001953, "KL/mean": -256.8182067871094, "KL/rejected_KL_mean": -309.75103759765625, "KL/std": 149.91073608398438, "epoch": 0.34508076358296624, "fcm_dpo/beta": 0.0026793205179274082, "fcm_dpo/delta": 0.0680294781923294, "fcm_dpo/margin": 105.86572265625, "fcm_dpo/q_t": 0.43403372168540955, "grad_norm": 19.758920669555664, "learning_rate": 4.1555887447288255e-07, "logits/chosen": -0.40560123324394226, "logits/rejected": -0.38443222641944885, "logps/chosen": -266.12322998046875, "logps/ref_chosen": -62.237911224365234, "logps/ref_rejected": -90.39506530761719, "logps/rejected": -400.1461181640625, "loss": 1.1809, "margin_dpo/margin_mean": 105.86572265625, "margin_dpo/margin_std": 183.34307861328125, "step": 235 }, { "KL/chosen_KL_mean": -155.01666259765625, "KL/mean": -242.4228515625, "KL/rejected_KL_mean": -329.82904052734375, "KL/std": 166.82810974121094, "epoch": 0.3465491923641703, "fcm_dpo/beta": 0.0026379311457276344, "fcm_dpo/delta": -0.11645685136318207, "fcm_dpo/margin": 174.8123779296875, "fcm_dpo/q_t": 0.3907100558280945, "grad_norm": 33.895503997802734, "learning_rate": 4.1459510339613946e-07, "logits/chosen": -0.3970375657081604, "logits/rejected": -0.3981223404407501, "logps/chosen": -204.3580322265625, "logps/ref_chosen": -49.34136199951172, "logps/ref_rejected": -103.51162719726562, "logps/rejected": -433.3406677246094, "loss": 1.0198, "margin_dpo/margin_mean": 174.8123779296875, "margin_dpo/margin_std": 160.76687622070312, "step": 236 }, { "KL/chosen_KL_mean": -225.36773681640625, "KL/mean": -289.1072998046875, "KL/rejected_KL_mean": -352.8468017578125, "KL/std": 175.09231567382812, "epoch": 0.34801762114537443, "fcm_dpo/beta": 0.0026289066299796104, "fcm_dpo/delta": 0.01535837072879076, "fcm_dpo/margin": 127.47908020019531, "fcm_dpo/q_t": 0.42304670810699463, "grad_norm": 28.638195037841797, "learning_rate": 4.136269950853473e-07, "logits/chosen": -0.36658668518066406, "logits/rejected": -0.3595519959926605, "logps/chosen": -279.5358581542969, "logps/ref_chosen": -54.168121337890625, "logps/ref_rejected": -94.78036499023438, "logps/rejected": -447.6271667480469, "loss": 1.1454, "margin_dpo/margin_mean": 127.47908020019531, "margin_dpo/margin_std": 202.83706665039062, "step": 237 }, { "KL/chosen_KL_mean": -207.52232360839844, "KL/mean": -273.25469970703125, "KL/rejected_KL_mean": -338.9870910644531, "KL/std": 176.7599639892578, "epoch": 0.34948604992657856, "fcm_dpo/beta": 0.002619014820083976, "fcm_dpo/delta": 0.005497526377439499, "fcm_dpo/margin": 131.4647674560547, "fcm_dpo/q_t": 0.4229828417301178, "grad_norm": 27.573131561279297, "learning_rate": 4.126545750510605e-07, "logits/chosen": -0.3411535620689392, "logits/rejected": -0.35550758242607117, "logps/chosen": -261.4954528808594, "logps/ref_chosen": -53.973121643066406, "logps/ref_rejected": -89.41795349121094, "logps/rejected": -428.405029296875, "loss": 1.1376, "margin_dpo/margin_mean": 131.4647674560547, "margin_dpo/margin_std": 203.32174682617188, "step": 238 }, { "KL/chosen_KL_mean": -223.70838928222656, "KL/mean": -298.6877746582031, "KL/rejected_KL_mean": -373.6671447753906, "KL/std": 167.540771484375, "epoch": 0.3509544787077827, "fcm_dpo/beta": 0.0026005804538726807, "fcm_dpo/delta": -0.0424700528383255, "fcm_dpo/margin": 149.958740234375, "fcm_dpo/q_t": 0.4087134003639221, "grad_norm": 24.893705368041992, "learning_rate": 4.116778689174514e-07, "logits/chosen": -0.3514358401298523, "logits/rejected": -0.3399575352668762, "logps/chosen": -281.80621337890625, "logps/ref_chosen": -58.09782409667969, "logps/ref_rejected": -93.59294128417969, "logps/rejected": -467.26007080078125, "loss": 1.0948, "margin_dpo/margin_mean": 149.95875549316406, "margin_dpo/margin_std": 190.29000854492188, "step": 239 }, { "KL/chosen_KL_mean": -231.37310791015625, "KL/mean": -291.67303466796875, "KL/rejected_KL_mean": -351.9729919433594, "KL/std": 181.56192016601562, "epoch": 0.3524229074889868, "fcm_dpo/beta": 0.0026145458687096834, "fcm_dpo/delta": 0.0357041209936142, "fcm_dpo/margin": 120.59989929199219, "fcm_dpo/q_t": 0.4276321530342102, "grad_norm": 30.90320587158203, "learning_rate": 4.106969024216348e-07, "logits/chosen": -0.38337743282318115, "logits/rejected": -0.3629865050315857, "logps/chosen": -291.98760986328125, "logps/ref_chosen": -60.6144905090332, "logps/ref_rejected": -74.1185302734375, "logps/rejected": -426.0915222167969, "loss": 1.1746, "margin_dpo/margin_mean": 120.59988403320312, "margin_dpo/margin_std": 218.30465698242188, "step": 240 }, { "KL/chosen_KL_mean": -191.98458862304688, "KL/mean": -287.6497497558594, "KL/rejected_KL_mean": -383.31494140625, "KL/std": 194.20921325683594, "epoch": 0.35389133627019087, "fcm_dpo/beta": 0.002574965823441744, "fcm_dpo/delta": -0.14985774457454681, "fcm_dpo/margin": 191.33033752441406, "fcm_dpo/q_t": 0.3884301781654358, "grad_norm": 25.951866149902344, "learning_rate": 4.097117014129903e-07, "logits/chosen": -0.4079846739768982, "logits/rejected": -0.38517332077026367, "logps/chosen": -258.0756530761719, "logps/ref_chosen": -66.091064453125, "logps/ref_rejected": -88.06088256835938, "logps/rejected": -471.37579345703125, "loss": 1.0299, "margin_dpo/margin_mean": 191.33033752441406, "margin_dpo/margin_std": 221.56744384765625, "step": 241 }, { "KL/chosen_KL_mean": -215.06796264648438, "KL/mean": -284.0657958984375, "KL/rejected_KL_mean": -353.0635986328125, "KL/std": 165.72854614257812, "epoch": 0.355359765051395, "fcm_dpo/beta": 0.0025596697814762592, "fcm_dpo/delta": -0.003671940416097641, "fcm_dpo/margin": 137.99563598632812, "fcm_dpo/q_t": 0.41891148686408997, "grad_norm": 24.505521774291992, "learning_rate": 4.087222918524807e-07, "logits/chosen": -0.3450043797492981, "logits/rejected": -0.3202177882194519, "logps/chosen": -282.931884765625, "logps/ref_chosen": -67.86392974853516, "logps/ref_rejected": -83.36033630371094, "logps/rejected": -436.4239501953125, "loss": 1.1243, "margin_dpo/margin_mean": 137.99563598632812, "margin_dpo/margin_std": 196.5327606201172, "step": 242 }, { "KL/chosen_KL_mean": -214.51419067382812, "KL/mean": -297.41632080078125, "KL/rejected_KL_mean": -380.3184814453125, "KL/std": 182.67172241210938, "epoch": 0.3568281938325991, "fcm_dpo/beta": 0.002519626170396805, "fcm_dpo/delta": -0.07080057263374329, "fcm_dpo/margin": 165.80429077148438, "fcm_dpo/q_t": 0.40310847759246826, "grad_norm": 24.171142578125, "learning_rate": 4.07728699811968e-07, "logits/chosen": -0.382672518491745, "logits/rejected": -0.35550010204315186, "logps/chosen": -277.59844970703125, "logps/ref_chosen": -63.0842399597168, "logps/ref_rejected": -76.33563232421875, "logps/rejected": -456.65411376953125, "loss": 1.0728, "margin_dpo/margin_mean": 165.80429077148438, "margin_dpo/margin_std": 204.3806610107422, "step": 243 }, { "KL/chosen_KL_mean": -183.78140258789062, "KL/mean": -269.9775695800781, "KL/rejected_KL_mean": -356.1737060546875, "KL/std": 174.60687255859375, "epoch": 0.35829662261380324, "fcm_dpo/beta": 0.002478396287187934, "fcm_dpo/delta": -0.08090801537036896, "fcm_dpo/margin": 172.39230346679688, "fcm_dpo/q_t": 0.3996942937374115, "grad_norm": 25.310270309448242, "learning_rate": 4.067309514735267e-07, "logits/chosen": -0.4503272771835327, "logits/rejected": -0.44473206996917725, "logps/chosen": -244.92210388183594, "logps/ref_chosen": -61.140689849853516, "logps/ref_rejected": -94.89193725585938, "logps/rejected": -451.0656433105469, "loss": 1.0502, "margin_dpo/margin_mean": 172.39230346679688, "margin_dpo/margin_std": 177.4683837890625, "step": 244 }, { "KL/chosen_KL_mean": -209.92994689941406, "KL/mean": -275.97613525390625, "KL/rejected_KL_mean": -342.02227783203125, "KL/std": 168.3325653076172, "epoch": 0.35976505139500736, "fcm_dpo/beta": 0.0024427759926766157, "fcm_dpo/delta": -0.060425255447626114, "fcm_dpo/margin": 132.0923309326172, "fcm_dpo/q_t": 0.42357954382896423, "grad_norm": 22.16962242126465, "learning_rate": 4.057290731287531e-07, "logits/chosen": -0.4270048141479492, "logits/rejected": -0.39348340034484863, "logps/chosen": -277.1922302246094, "logps/ref_chosen": -67.26228332519531, "logps/ref_rejected": -87.64010620117188, "logps/rejected": -429.6623840332031, "loss": 1.1442, "margin_dpo/margin_mean": 132.0923309326172, "margin_dpo/margin_std": 188.43490600585938, "step": 245 }, { "KL/chosen_KL_mean": -200.66488647460938, "KL/mean": -268.2365417480469, "KL/rejected_KL_mean": -335.80816650390625, "KL/std": 186.197265625, "epoch": 0.36123348017621143, "fcm_dpo/beta": 0.0024440872948616743, "fcm_dpo/delta": 0.02029471844434738, "fcm_dpo/margin": 135.14329528808594, "fcm_dpo/q_t": 0.4241391718387604, "grad_norm": 25.40897560119629, "learning_rate": 4.047230911780736e-07, "logits/chosen": -0.4735276699066162, "logits/rejected": -0.43419474363327026, "logps/chosen": -267.3618469238281, "logps/ref_chosen": -66.69696807861328, "logps/ref_rejected": -84.34634399414062, "logps/rejected": -420.154541015625, "loss": 1.1388, "margin_dpo/margin_mean": 135.14329528808594, "margin_dpo/margin_std": 200.08724975585938, "step": 246 }, { "KL/chosen_KL_mean": -234.4573516845703, "KL/mean": -329.88720703125, "KL/rejected_KL_mean": -425.31707763671875, "KL/std": 194.06973266601562, "epoch": 0.36270190895741555, "fcm_dpo/beta": 0.0024084593169391155, "fcm_dpo/delta": -0.11511167883872986, "fcm_dpo/margin": 190.85971069335938, "fcm_dpo/q_t": 0.3926267623901367, "grad_norm": 26.054208755493164, "learning_rate": 4.0371303213004814e-07, "logits/chosen": -0.4317302405834198, "logits/rejected": -0.4298396706581116, "logps/chosen": -291.06268310546875, "logps/ref_chosen": -56.6053466796875, "logps/ref_rejected": -106.29326629638672, "logps/rejected": -531.6103515625, "loss": 1.0387, "margin_dpo/margin_mean": 190.85972595214844, "margin_dpo/margin_std": 208.8062744140625, "step": 247 }, { "KL/chosen_KL_mean": -201.06793212890625, "KL/mean": -285.0498046875, "KL/rejected_KL_mean": -369.03167724609375, "KL/std": 158.5778350830078, "epoch": 0.3641703377386197, "fcm_dpo/beta": 0.0023695870768278837, "fcm_dpo/delta": -0.05050865188241005, "fcm_dpo/margin": 167.96372985839844, "fcm_dpo/q_t": 0.4044187664985657, "grad_norm": 25.75300407409668, "learning_rate": 4.0269892260067197e-07, "logits/chosen": -0.407694935798645, "logits/rejected": -0.42660677433013916, "logps/chosen": -245.1111602783203, "logps/ref_chosen": -44.043216705322266, "logps/ref_rejected": -91.85687255859375, "logps/rejected": -460.8885498046875, "loss": 1.0603, "margin_dpo/margin_mean": 167.96371459960938, "margin_dpo/margin_std": 153.2666015625, "step": 248 }, { "KL/chosen_KL_mean": -262.7314147949219, "KL/mean": -308.6989440917969, "KL/rejected_KL_mean": -354.66644287109375, "KL/std": 177.82664489746094, "epoch": 0.3656387665198238, "fcm_dpo/beta": 0.002401418052613735, "fcm_dpo/delta": 0.05994679778814316, "fcm_dpo/margin": 91.93504333496094, "fcm_dpo/q_t": 0.44910717010498047, "grad_norm": 33.85601806640625, "learning_rate": 4.0168078931267426e-07, "logits/chosen": -0.43958780169487, "logits/rejected": -0.4103413224220276, "logps/chosen": -325.17376708984375, "logps/ref_chosen": -62.442352294921875, "logps/ref_rejected": -80.46806335449219, "logps/rejected": -435.134521484375, "loss": 1.2408, "margin_dpo/margin_mean": 91.9350357055664, "margin_dpo/margin_std": 213.35684204101562, "step": 249 }, { "KL/chosen_KL_mean": -231.9749755859375, "KL/mean": -314.76470947265625, "KL/rejected_KL_mean": -397.55438232421875, "KL/std": 174.78016662597656, "epoch": 0.3671071953010279, "fcm_dpo/beta": 0.0023939479142427444, "fcm_dpo/delta": -0.04844534397125244, "fcm_dpo/margin": 165.57940673828125, "fcm_dpo/q_t": 0.40629374980926514, "grad_norm": 41.83970642089844, "learning_rate": 4.006586590948141e-07, "logits/chosen": -0.43187060952186584, "logits/rejected": -0.3728846311569214, "logps/chosen": -297.6116943359375, "logps/ref_chosen": -65.63668823242188, "logps/ref_rejected": -73.87184143066406, "logps/rejected": -471.42620849609375, "loss": 1.0762, "margin_dpo/margin_mean": 165.57940673828125, "margin_dpo/margin_std": 183.59869384765625, "step": 250 }, { "KL/chosen_KL_mean": -240.02685546875, "KL/mean": -298.3031005859375, "KL/rejected_KL_mean": -356.5793762207031, "KL/std": 184.30174255371094, "epoch": 0.368575624082232, "fcm_dpo/beta": 0.002399963792413473, "fcm_dpo/delta": 0.07236499339342117, "fcm_dpo/margin": 116.55254364013672, "fcm_dpo/q_t": 0.4352172613143921, "grad_norm": 23.73138427734375, "learning_rate": 3.9963255888117325e-07, "logits/chosen": -0.3889552354812622, "logits/rejected": -0.35478660464286804, "logps/chosen": -297.2095642089844, "logps/ref_chosen": -57.182716369628906, "logps/ref_rejected": -77.66343688964844, "logps/rejected": -434.2427978515625, "loss": 1.1848, "margin_dpo/margin_mean": 116.55254364013672, "margin_dpo/margin_std": 200.21017456054688, "step": 251 }, { "KL/chosen_KL_mean": -222.79547119140625, "KL/mean": -299.47967529296875, "KL/rejected_KL_mean": -376.16387939453125, "KL/std": 155.37213134765625, "epoch": 0.3700440528634361, "fcm_dpo/beta": 0.0024020476266741753, "fcm_dpo/delta": -0.01919008605182171, "fcm_dpo/margin": 153.368408203125, "fcm_dpo/q_t": 0.4108833074569702, "grad_norm": 22.33046531677246, "learning_rate": 3.9860251571044666e-07, "logits/chosen": -0.48540571331977844, "logits/rejected": -0.44611865282058716, "logps/chosen": -294.48114013671875, "logps/ref_chosen": -71.68563842773438, "logps/ref_rejected": -84.75799560546875, "logps/rejected": -460.921875, "loss": 1.086, "margin_dpo/margin_mean": 153.368408203125, "margin_dpo/margin_std": 156.39324951171875, "step": 252 }, { "KL/chosen_KL_mean": -193.2376708984375, "KL/mean": -271.1014404296875, "KL/rejected_KL_mean": -348.9652099609375, "KL/std": 167.41372680664062, "epoch": 0.37151248164464024, "fcm_dpo/beta": 0.002407216466963291, "fcm_dpo/delta": -0.02628351002931595, "fcm_dpo/margin": 155.72752380371094, "fcm_dpo/q_t": 0.4112589955329895, "grad_norm": 20.50514793395996, "learning_rate": 3.9756855672522986e-07, "logits/chosen": -0.44385430216789246, "logits/rejected": -0.43464112281799316, "logps/chosen": -262.3716125488281, "logps/ref_chosen": -69.1339340209961, "logps/ref_rejected": -98.70252990722656, "logps/rejected": -447.667724609375, "loss": 1.0973, "margin_dpo/margin_mean": 155.72752380371094, "margin_dpo/margin_std": 186.75709533691406, "step": 253 }, { "KL/chosen_KL_mean": -200.95816040039062, "KL/mean": -266.77008056640625, "KL/rejected_KL_mean": -332.58203125, "KL/std": 181.447998046875, "epoch": 0.37298091042584436, "fcm_dpo/beta": 0.002402001526206732, "fcm_dpo/delta": 0.03495839610695839, "fcm_dpo/margin": 131.62384033203125, "fcm_dpo/q_t": 0.42862290143966675, "grad_norm": 22.96527862548828, "learning_rate": 3.965307091713037e-07, "logits/chosen": -0.41199079155921936, "logits/rejected": -0.39535099267959595, "logps/chosen": -255.1131591796875, "logps/ref_chosen": -54.154998779296875, "logps/ref_rejected": -90.30764770507812, "logps/rejected": -422.8896484375, "loss": 1.1643, "margin_dpo/margin_mean": 131.62384033203125, "margin_dpo/margin_std": 228.5364532470703, "step": 254 }, { "KL/chosen_KL_mean": -213.82733154296875, "KL/mean": -285.01959228515625, "KL/rejected_KL_mean": -356.2118225097656, "KL/std": 163.05636596679688, "epoch": 0.3744493392070485, "fcm_dpo/beta": 0.0024028548505157232, "fcm_dpo/delta": 0.007957128807902336, "fcm_dpo/margin": 142.38449096679688, "fcm_dpo/q_t": 0.4198462963104248, "grad_norm": 21.276325225830078, "learning_rate": 3.954890003969163e-07, "logits/chosen": -0.3709820806980133, "logits/rejected": -0.3580097556114197, "logps/chosen": -270.968994140625, "logps/ref_chosen": -57.14167022705078, "logps/ref_rejected": -90.2085952758789, "logps/rejected": -446.42041015625, "loss": 1.1336, "margin_dpo/margin_mean": 142.38449096679688, "margin_dpo/margin_std": 206.68133544921875, "step": 255 }, { "KL/chosen_KL_mean": -192.61761474609375, "KL/mean": -271.03961181640625, "KL/rejected_KL_mean": -349.4615478515625, "KL/std": 179.253662109375, "epoch": 0.37591776798825255, "fcm_dpo/beta": 0.0024023070000112057, "fcm_dpo/delta": -0.027830585837364197, "fcm_dpo/margin": 156.8439483642578, "fcm_dpo/q_t": 0.41207343339920044, "grad_norm": 23.000877380371094, "learning_rate": 3.944434578520628e-07, "logits/chosen": -0.3232497274875641, "logits/rejected": -0.3319256007671356, "logps/chosen": -247.78109741210938, "logps/ref_chosen": -55.163490295410156, "logps/ref_rejected": -92.56291961669922, "logps/rejected": -442.02447509765625, "loss": 1.1021, "margin_dpo/margin_mean": 156.84396362304688, "margin_dpo/margin_std": 205.78041076660156, "step": 256 }, { "KL/chosen_KL_mean": -190.80270385742188, "KL/mean": -275.1641845703125, "KL/rejected_KL_mean": -359.5256652832031, "KL/std": 190.19740295410156, "epoch": 0.37738619676945667, "fcm_dpo/beta": 0.0023642941378057003, "fcm_dpo/delta": -0.052151791751384735, "fcm_dpo/margin": 168.72293090820312, "fcm_dpo/q_t": 0.4079574942588806, "grad_norm": 21.00106430053711, "learning_rate": 3.933941090877615e-07, "logits/chosen": -0.3763759434223175, "logits/rejected": -0.36591070890426636, "logps/chosen": -240.22640991210938, "logps/ref_chosen": -49.42369842529297, "logps/ref_rejected": -79.53791809082031, "logps/rejected": -439.0635986328125, "loss": 1.0929, "margin_dpo/margin_mean": 168.72293090820312, "margin_dpo/margin_std": 215.62423706054688, "step": 257 }, { "KL/chosen_KL_mean": -259.0428161621094, "KL/mean": -340.54241943359375, "KL/rejected_KL_mean": -422.0419921875, "KL/std": 201.83018493652344, "epoch": 0.3788546255506608, "fcm_dpo/beta": 0.002365153282880783, "fcm_dpo/delta": -0.036950256675481796, "fcm_dpo/margin": 162.99920654296875, "fcm_dpo/q_t": 0.4098193347454071, "grad_norm": 20.96170425415039, "learning_rate": 3.923409817553284e-07, "logits/chosen": -0.3399478793144226, "logits/rejected": -0.3372340202331543, "logps/chosen": -318.42694091796875, "logps/ref_chosen": -59.384124755859375, "logps/ref_rejected": -95.99010467529297, "logps/rejected": -518.0321044921875, "loss": 1.1166, "margin_dpo/margin_mean": 162.99920654296875, "margin_dpo/margin_std": 244.92196655273438, "step": 258 }, { "KL/chosen_KL_mean": -242.94644165039062, "KL/mean": -309.83868408203125, "KL/rejected_KL_mean": -376.73089599609375, "KL/std": 193.08038330078125, "epoch": 0.3803230543318649, "fcm_dpo/beta": 0.00236131576821208, "fcm_dpo/delta": 0.03512916713953018, "fcm_dpo/margin": 133.784423828125, "fcm_dpo/q_t": 0.4273309111595154, "grad_norm": 20.399890899658203, "learning_rate": 3.9128410360564793e-07, "logits/chosen": -0.4055570065975189, "logits/rejected": -0.40263453125953674, "logps/chosen": -295.7747802734375, "logps/ref_chosen": -52.828346252441406, "logps/ref_rejected": -89.191650390625, "logps/rejected": -465.92254638671875, "loss": 1.1583, "margin_dpo/margin_mean": 133.784423828125, "margin_dpo/margin_std": 217.29360961914062, "step": 259 }, { "KL/chosen_KL_mean": -235.74888610839844, "KL/mean": -328.3739929199219, "KL/rejected_KL_mean": -420.99908447265625, "KL/std": 199.5541534423828, "epoch": 0.38179148311306904, "fcm_dpo/beta": 0.0023506833240389824, "fcm_dpo/delta": -0.08932534605264664, "fcm_dpo/margin": 185.25018310546875, "fcm_dpo/q_t": 0.40073201060295105, "grad_norm": 22.08340072631836, "learning_rate": 3.9022350248844246e-07, "logits/chosen": -0.3329894542694092, "logits/rejected": -0.34919166564941406, "logps/chosen": -283.16656494140625, "logps/ref_chosen": -47.41767501831055, "logps/ref_rejected": -95.08978271484375, "logps/rejected": -516.0888671875, "loss": 1.0557, "margin_dpo/margin_mean": 185.25018310546875, "margin_dpo/margin_std": 216.7088623046875, "step": 260 }, { "KL/chosen_KL_mean": -244.0491943359375, "KL/mean": -328.637451171875, "KL/rejected_KL_mean": -413.2257080078125, "KL/std": 211.37762451171875, "epoch": 0.3832599118942731, "fcm_dpo/beta": 0.002309254603460431, "fcm_dpo/delta": -0.04260961338877678, "fcm_dpo/margin": 169.17654418945312, "fcm_dpo/q_t": 0.41040754318237305, "grad_norm": 18.893648147583008, "learning_rate": 3.891592063515376e-07, "logits/chosen": -0.30380675196647644, "logits/rejected": -0.30368590354919434, "logps/chosen": -297.08056640625, "logps/ref_chosen": -53.03137969970703, "logps/ref_rejected": -88.51494598388672, "logps/rejected": -501.74066162109375, "loss": 1.1033, "margin_dpo/margin_mean": 169.17652893066406, "margin_dpo/margin_std": 237.78912353515625, "step": 261 }, { "KL/chosen_KL_mean": -268.54522705078125, "KL/mean": -337.58929443359375, "KL/rejected_KL_mean": -406.63336181640625, "KL/std": 185.41744995117188, "epoch": 0.38472834067547723, "fcm_dpo/beta": 0.002329050563275814, "fcm_dpo/delta": 0.028969408944249153, "fcm_dpo/margin": 138.08815002441406, "fcm_dpo/q_t": 0.4243643879890442, "grad_norm": 25.439929962158203, "learning_rate": 3.880912432401264e-07, "logits/chosen": -0.344732403755188, "logits/rejected": -0.3178238272666931, "logps/chosen": -328.16534423828125, "logps/ref_chosen": -59.620140075683594, "logps/ref_rejected": -86.41853332519531, "logps/rejected": -493.0518798828125, "loss": 1.1353, "margin_dpo/margin_mean": 138.08815002441406, "margin_dpo/margin_std": 185.788330078125, "step": 262 }, { "KL/chosen_KL_mean": -235.66673278808594, "KL/mean": -331.74688720703125, "KL/rejected_KL_mean": -427.82708740234375, "KL/std": 215.85923767089844, "epoch": 0.38619676945668135, "fcm_dpo/beta": 0.0022833282127976418, "fcm_dpo/delta": -0.09349527209997177, "fcm_dpo/margin": 192.16030883789062, "fcm_dpo/q_t": 0.3982081413269043, "grad_norm": 23.847797393798828, "learning_rate": 3.870196412960302e-07, "logits/chosen": -0.36012065410614014, "logits/rejected": -0.3382565975189209, "logps/chosen": -295.0876770019531, "logps/ref_chosen": -59.42094421386719, "logps/ref_rejected": -96.85720825195312, "logps/rejected": -524.6842651367188, "loss": 1.061, "margin_dpo/margin_mean": 192.16033935546875, "margin_dpo/margin_std": 232.20291137695312, "step": 263 }, { "KL/chosen_KL_mean": -241.67784118652344, "KL/mean": -322.71044921875, "KL/rejected_KL_mean": -403.7430419921875, "KL/std": 196.90768432617188, "epoch": 0.3876651982378855, "fcm_dpo/beta": 0.002260984852910042, "fcm_dpo/delta": -0.017950933426618576, "fcm_dpo/margin": 162.06521606445312, "fcm_dpo/q_t": 0.4153786301612854, "grad_norm": 21.233423233032227, "learning_rate": 3.8594442875695665e-07, "logits/chosen": -0.4053404927253723, "logits/rejected": -0.3976580500602722, "logps/chosen": -304.3999328613281, "logps/ref_chosen": -62.722084045410156, "logps/ref_rejected": -93.85620880126953, "logps/rejected": -497.5992431640625, "loss": 1.1174, "margin_dpo/margin_mean": 162.06521606445312, "margin_dpo/margin_std": 225.04757690429688, "step": 264 }, { "KL/chosen_KL_mean": -266.4638366699219, "KL/mean": -342.8912353515625, "KL/rejected_KL_mean": -419.3185729980469, "KL/std": 219.68490600585938, "epoch": 0.3891336270190896, "fcm_dpo/beta": 0.002274434082210064, "fcm_dpo/delta": 0.0024239865597337484, "fcm_dpo/margin": 152.85475158691406, "fcm_dpo/q_t": 0.4211368262767792, "grad_norm": 23.371530532836914, "learning_rate": 3.848656339557562e-07, "logits/chosen": -0.3490335941314697, "logits/rejected": -0.3347071707248688, "logps/chosen": -328.435302734375, "logps/ref_chosen": -61.971466064453125, "logps/ref_rejected": -88.02059936523438, "logps/rejected": -507.33917236328125, "loss": 1.1501, "margin_dpo/margin_mean": 152.85475158691406, "margin_dpo/margin_std": 261.6484069824219, "step": 265 }, { "KL/chosen_KL_mean": -269.20404052734375, "KL/mean": -327.8632507324219, "KL/rejected_KL_mean": -386.5224609375, "KL/std": 181.0565948486328, "epoch": 0.39060205580029367, "fcm_dpo/beta": 0.0023039982188493013, "fcm_dpo/delta": 0.08182872086763382, "fcm_dpo/margin": 117.31837463378906, "fcm_dpo/q_t": 0.43739640712738037, "grad_norm": 47.123313903808594, "learning_rate": 3.8378328531967507e-07, "logits/chosen": -0.3748672306537628, "logits/rejected": -0.340556800365448, "logps/chosen": -336.3037109375, "logps/ref_chosen": -67.09967041015625, "logps/ref_rejected": -67.97122192382812, "logps/rejected": -454.4936828613281, "loss": 1.1911, "margin_dpo/margin_mean": 117.31837463378906, "margin_dpo/margin_std": 212.25018310546875, "step": 266 }, { "KL/chosen_KL_mean": -236.6424560546875, "KL/mean": -310.0815124511719, "KL/rejected_KL_mean": -383.52056884765625, "KL/std": 203.6986083984375, "epoch": 0.3920704845814978, "fcm_dpo/beta": 0.0023088420275598764, "fcm_dpo/delta": 0.011129170656204224, "fcm_dpo/margin": 146.87808227539062, "fcm_dpo/q_t": 0.4222862124443054, "grad_norm": 26.301164627075195, "learning_rate": 3.8269741136960646e-07, "logits/chosen": -0.3820560872554779, "logits/rejected": -0.3540261387825012, "logps/chosen": -305.61322021484375, "logps/ref_chosen": -68.97075653076172, "logps/ref_rejected": -90.16844940185547, "logps/rejected": -473.6889953613281, "loss": 1.1456, "margin_dpo/margin_mean": 146.87808227539062, "margin_dpo/margin_std": 237.48892211914062, "step": 267 }, { "KL/chosen_KL_mean": -249.36642456054688, "KL/mean": -328.30133056640625, "KL/rejected_KL_mean": -407.23626708984375, "KL/std": 204.91433715820312, "epoch": 0.3935389133627019, "fcm_dpo/beta": 0.002304880414158106, "fcm_dpo/delta": -0.014616616070270538, "fcm_dpo/margin": 157.86981201171875, "fcm_dpo/q_t": 0.41689735651016235, "grad_norm": 24.127378463745117, "learning_rate": 3.8160804071933894e-07, "logits/chosen": -0.35487100481987, "logits/rejected": -0.35821008682250977, "logps/chosen": -305.2667236328125, "logps/ref_chosen": -55.90031051635742, "logps/ref_rejected": -101.64763641357422, "logps/rejected": -508.8839111328125, "loss": 1.1305, "margin_dpo/margin_mean": 157.86981201171875, "margin_dpo/margin_std": 247.58233642578125, "step": 268 }, { "KL/chosen_KL_mean": -278.9156494140625, "KL/mean": -371.38323974609375, "KL/rejected_KL_mean": -463.85076904296875, "KL/std": 207.69430541992188, "epoch": 0.39500734214390604, "fcm_dpo/beta": 0.0022870278917253017, "fcm_dpo/delta": -0.0761084258556366, "fcm_dpo/margin": 184.93511962890625, "fcm_dpo/q_t": 0.40323197841644287, "grad_norm": 24.954824447631836, "learning_rate": 3.8051520207480204e-07, "logits/chosen": -0.3960835933685303, "logits/rejected": -0.37948912382125854, "logps/chosen": -348.9552307128906, "logps/ref_chosen": -70.03955841064453, "logps/ref_rejected": -107.34937286376953, "logps/rejected": -571.2001953125, "loss": 1.095, "margin_dpo/margin_mean": 184.93511962890625, "margin_dpo/margin_std": 267.55242919921875, "step": 269 }, { "KL/chosen_KL_mean": -247.30697631835938, "KL/mean": -317.04638671875, "KL/rejected_KL_mean": -386.7857971191406, "KL/std": 184.7318115234375, "epoch": 0.3964757709251101, "fcm_dpo/beta": 0.002292545512318611, "fcm_dpo/delta": 0.030957801267504692, "fcm_dpo/margin": 139.4788360595703, "fcm_dpo/q_t": 0.42589646577835083, "grad_norm": 21.303037643432617, "learning_rate": 3.794189242333106e-07, "logits/chosen": -0.4222424626350403, "logits/rejected": -0.4178420305252075, "logps/chosen": -316.8404541015625, "logps/ref_chosen": -69.53347778320312, "logps/ref_rejected": -109.92864990234375, "logps/rejected": -496.7144470214844, "loss": 1.1639, "margin_dpo/margin_mean": 139.4788360595703, "margin_dpo/margin_std": 239.0890350341797, "step": 270 }, { "KL/chosen_KL_mean": -228.56326293945312, "KL/mean": -315.3665466308594, "KL/rejected_KL_mean": -402.1697998046875, "KL/std": 187.98452758789062, "epoch": 0.39794419970631423, "fcm_dpo/beta": 0.0022674724459648132, "fcm_dpo/delta": -0.04571424052119255, "fcm_dpo/margin": 173.60653686523438, "fcm_dpo/q_t": 0.4088677167892456, "grad_norm": 21.11397361755371, "learning_rate": 3.7831923608280514e-07, "logits/chosen": -0.3680250346660614, "logits/rejected": -0.3530626893043518, "logps/chosen": -285.32781982421875, "logps/ref_chosen": -56.76456832885742, "logps/ref_rejected": -92.51383972167969, "logps/rejected": -494.68365478515625, "loss": 1.0878, "margin_dpo/margin_mean": 173.60653686523438, "margin_dpo/margin_std": 217.1529541015625, "step": 271 }, { "KL/chosen_KL_mean": -215.89785766601562, "KL/mean": -326.992431640625, "KL/rejected_KL_mean": -438.0870361328125, "KL/std": 207.18685913085938, "epoch": 0.39941262848751835, "fcm_dpo/beta": 0.0022256127558648586, "fcm_dpo/delta": -0.1517450511455536, "fcm_dpo/margin": 222.18914794921875, "fcm_dpo/q_t": 0.38386717438697815, "grad_norm": 29.405025482177734, "learning_rate": 3.772161666010912e-07, "logits/chosen": -0.35582485795021057, "logits/rejected": -0.36779093742370605, "logps/chosen": -265.39501953125, "logps/ref_chosen": -49.497154235839844, "logps/ref_rejected": -105.54279327392578, "logps/rejected": -543.6298217773438, "loss": 1.007, "margin_dpo/margin_mean": 222.18914794921875, "margin_dpo/margin_std": 215.62319946289062, "step": 272 }, { "KL/chosen_KL_mean": -264.82843017578125, "KL/mean": -371.188720703125, "KL/rejected_KL_mean": -477.54901123046875, "KL/std": 212.29110717773438, "epoch": 0.4008810572687225, "fcm_dpo/beta": 0.0021587321534752846, "fcm_dpo/delta": -0.11479845643043518, "fcm_dpo/margin": 212.72059631347656, "fcm_dpo/q_t": 0.3933224678039551, "grad_norm": 31.42890739440918, "learning_rate": 3.761097448550755e-07, "logits/chosen": -0.3145291805267334, "logits/rejected": -0.2950317859649658, "logps/chosen": -327.8038330078125, "logps/ref_chosen": -62.97539520263672, "logps/ref_rejected": -92.49858093261719, "logps/rejected": -570.047607421875, "loss": 1.0435, "margin_dpo/margin_mean": 212.72059631347656, "margin_dpo/margin_std": 245.05528259277344, "step": 273 }, { "KL/chosen_KL_mean": -320.0135498046875, "KL/mean": -398.0048828125, "KL/rejected_KL_mean": -475.9962158203125, "KL/std": 196.2461700439453, "epoch": 0.4023494860499266, "fcm_dpo/beta": 0.0021619275212287903, "fcm_dpo/delta": 0.012957748025655746, "fcm_dpo/margin": 155.98263549804688, "fcm_dpo/q_t": 0.42174428701400757, "grad_norm": 33.96529769897461, "learning_rate": 3.75e-07, "logits/chosen": -0.2863592803478241, "logits/rejected": -0.26870518922805786, "logps/chosen": -375.6812744140625, "logps/ref_chosen": -55.66770935058594, "logps/ref_rejected": -77.33308410644531, "logps/rejected": -553.3292846679688, "loss": 1.1361, "margin_dpo/margin_mean": 155.98263549804688, "margin_dpo/margin_std": 228.75283813476562, "step": 274 }, { "KL/chosen_KL_mean": -249.88211059570312, "KL/mean": -336.5286865234375, "KL/rejected_KL_mean": -423.1752014160156, "KL/std": 197.85572814941406, "epoch": 0.40381791483113066, "fcm_dpo/beta": 0.0021535661071538925, "fcm_dpo/delta": -0.024154571816325188, "fcm_dpo/margin": 173.29312133789062, "fcm_dpo/q_t": 0.4124983847141266, "grad_norm": 26.547245025634766, "learning_rate": 3.738869612786737e-07, "logits/chosen": -0.3534200191497803, "logits/rejected": -0.3581305146217346, "logps/chosen": -298.476806640625, "logps/ref_chosen": -48.594703674316406, "logps/ref_rejected": -93.30369567871094, "logps/rejected": -516.4788818359375, "loss": 1.1025, "margin_dpo/margin_mean": 173.29312133789062, "margin_dpo/margin_std": 222.11753845214844, "step": 275 }, { "KL/chosen_KL_mean": -279.96337890625, "KL/mean": -363.65765380859375, "KL/rejected_KL_mean": -447.3519287109375, "KL/std": 210.06130981445312, "epoch": 0.4052863436123348, "fcm_dpo/beta": 0.0021390733309090137, "fcm_dpo/delta": -0.008444213308393955, "fcm_dpo/margin": 167.38851928710938, "fcm_dpo/q_t": 0.4185035824775696, "grad_norm": 25.184022903442383, "learning_rate": 3.7277065802070204e-07, "logits/chosen": -0.34166771173477173, "logits/rejected": -0.323291152715683, "logps/chosen": -336.540771484375, "logps/ref_chosen": -56.57740783691406, "logps/ref_rejected": -70.36566925048828, "logps/rejected": -517.7175903320312, "loss": 1.131, "margin_dpo/margin_mean": 167.38851928710938, "margin_dpo/margin_std": 257.09930419921875, "step": 276 }, { "KL/chosen_KL_mean": -302.4512939453125, "KL/mean": -391.5897521972656, "KL/rejected_KL_mean": -480.72821044921875, "KL/std": 209.8077850341797, "epoch": 0.4067547723935389, "fcm_dpo/beta": 0.0021361508406698704, "fcm_dpo/delta": -0.032061617821455, "fcm_dpo/margin": 178.2769012451172, "fcm_dpo/q_t": 0.41165226697921753, "grad_norm": 24.468589782714844, "learning_rate": 3.71651119641714e-07, "logits/chosen": -0.3302989602088928, "logits/rejected": -0.3095225393772125, "logps/chosen": -358.7228698730469, "logps/ref_chosen": -56.27156066894531, "logps/ref_rejected": -92.88127136230469, "logps/rejected": -573.6094970703125, "loss": 1.1107, "margin_dpo/margin_mean": 178.2769012451172, "margin_dpo/margin_std": 252.54180908203125, "step": 277 }, { "KL/chosen_KL_mean": -257.0399169921875, "KL/mean": -361.0743103027344, "KL/rejected_KL_mean": -465.1087341308594, "KL/std": 211.69451904296875, "epoch": 0.40822320117474303, "fcm_dpo/beta": 0.002095486968755722, "fcm_dpo/delta": -0.09022445231676102, "fcm_dpo/margin": 208.06884765625, "fcm_dpo/q_t": 0.3987731635570526, "grad_norm": 20.533023834228516, "learning_rate": 3.705283756425872e-07, "logits/chosen": -0.3557165265083313, "logits/rejected": -0.36121219396591187, "logps/chosen": -309.9818420410156, "logps/ref_chosen": -52.94194030761719, "logps/ref_rejected": -91.25357818603516, "logps/rejected": -556.3623046875, "loss": 1.056, "margin_dpo/margin_mean": 208.06884765625, "margin_dpo/margin_std": 240.28025817871094, "step": 278 }, { "KL/chosen_KL_mean": -297.38983154296875, "KL/mean": -400.4170227050781, "KL/rejected_KL_mean": -503.44427490234375, "KL/std": 230.00518798828125, "epoch": 0.40969162995594716, "fcm_dpo/beta": 0.0020560654811561108, "fcm_dpo/delta": -0.07781803607940674, "fcm_dpo/margin": 206.0543975830078, "fcm_dpo/q_t": 0.40380701422691345, "grad_norm": 25.10165786743164, "learning_rate": 3.6940245560867e-07, "logits/chosen": -0.2838994860649109, "logits/rejected": -0.2843170762062073, "logps/chosen": -346.0311279296875, "logps/ref_chosen": -48.641319274902344, "logps/ref_rejected": -87.8514404296875, "logps/rejected": -591.2957153320312, "loss": 1.0884, "margin_dpo/margin_mean": 206.05441284179688, "margin_dpo/margin_std": 287.540771484375, "step": 279 }, { "KL/chosen_KL_mean": -289.91473388671875, "KL/mean": -394.880859375, "KL/rejected_KL_mean": -499.84698486328125, "KL/std": 210.3863525390625, "epoch": 0.4111600587371512, "fcm_dpo/beta": 0.0020319321192801, "fcm_dpo/delta": -0.08006896078586578, "fcm_dpo/margin": 209.9322052001953, "fcm_dpo/q_t": 0.39913302659988403, "grad_norm": 28.02513885498047, "learning_rate": 3.6827338920900253e-07, "logits/chosen": -0.3529083728790283, "logits/rejected": -0.35713738203048706, "logps/chosen": -348.71185302734375, "logps/ref_chosen": -58.797122955322266, "logps/ref_rejected": -98.61885070800781, "logps/rejected": -598.4658203125, "loss": 1.0566, "margin_dpo/margin_mean": 209.9322052001953, "margin_dpo/margin_std": 231.01846313476562, "step": 280 }, { "KL/chosen_KL_mean": -261.54742431640625, "KL/mean": -353.0968933105469, "KL/rejected_KL_mean": -444.6463623046875, "KL/std": 201.8822021484375, "epoch": 0.41262848751835535, "fcm_dpo/beta": 0.0020250456873327494, "fcm_dpo/delta": -0.02170453406870365, "fcm_dpo/margin": 183.09893798828125, "fcm_dpo/q_t": 0.41141772270202637, "grad_norm": 22.01675796508789, "learning_rate": 3.6714120619553435e-07, "logits/chosen": -0.38511866331100464, "logits/rejected": -0.3614617586135864, "logps/chosen": -317.03594970703125, "logps/ref_chosen": -55.488521575927734, "logps/ref_rejected": -80.88258361816406, "logps/rejected": -525.5289306640625, "loss": 1.0996, "margin_dpo/margin_mean": 183.0989227294922, "margin_dpo/margin_std": 222.86624145507812, "step": 281 }, { "KL/chosen_KL_mean": -272.9205322265625, "KL/mean": -343.1865234375, "KL/rejected_KL_mean": -413.4525146484375, "KL/std": 208.0769500732422, "epoch": 0.41409691629955947, "fcm_dpo/beta": 0.0020167033653706312, "fcm_dpo/delta": 7.871291745686904e-05, "fcm_dpo/margin": 140.53196716308594, "fcm_dpo/q_t": 0.43511486053466797, "grad_norm": 22.090290069580078, "learning_rate": 3.660059364023408e-07, "logits/chosen": -0.4389272928237915, "logits/rejected": -0.41942620277404785, "logps/chosen": -345.99066162109375, "logps/ref_chosen": -73.07014465332031, "logps/ref_rejected": -95.35098266601562, "logps/rejected": -508.80352783203125, "loss": 1.1751, "margin_dpo/margin_mean": 140.53196716308594, "margin_dpo/margin_std": 235.73179626464844, "step": 282 }, { "KL/chosen_KL_mean": -265.983154296875, "KL/mean": -372.7215576171875, "KL/rejected_KL_mean": -479.46002197265625, "KL/std": 232.66961669921875, "epoch": 0.4155653450807636, "fcm_dpo/beta": 0.0020004159305244684, "fcm_dpo/delta": -0.08047128468751907, "fcm_dpo/margin": 213.47683715820312, "fcm_dpo/q_t": 0.3984811007976532, "grad_norm": 32.03657531738281, "learning_rate": 3.6486760974483685e-07, "logits/chosen": -0.39184266328811646, "logits/rejected": -0.3945959210395813, "logps/chosen": -327.881591796875, "logps/ref_chosen": -61.89844512939453, "logps/ref_rejected": -96.98655700683594, "logps/rejected": -576.446533203125, "loss": 1.0528, "margin_dpo/margin_mean": 213.47683715820312, "margin_dpo/margin_std": 221.32455444335938, "step": 283 }, { "KL/chosen_KL_mean": -246.0831298828125, "KL/mean": -346.190185546875, "KL/rejected_KL_mean": -446.2972106933594, "KL/std": 212.89483642578125, "epoch": 0.4170337738619677, "fcm_dpo/beta": 0.0019676508381962776, "fcm_dpo/delta": -0.04589087888598442, "fcm_dpo/margin": 200.21408081054688, "fcm_dpo/q_t": 0.40764474868774414, "grad_norm": 24.56162452697754, "learning_rate": 3.6372625621898863e-07, "logits/chosen": -0.452556312084198, "logits/rejected": -0.43920326232910156, "logps/chosen": -304.5186767578125, "logps/ref_chosen": -58.4355354309082, "logps/ref_rejected": -93.46926879882812, "logps/rejected": -539.7664794921875, "loss": 1.0754, "margin_dpo/margin_mean": 200.21408081054688, "margin_dpo/margin_std": 222.98330688476562, "step": 284 }, { "KL/chosen_KL_mean": -289.398193359375, "KL/mean": -374.96954345703125, "KL/rejected_KL_mean": -460.54095458984375, "KL/std": 184.91241455078125, "epoch": 0.4185022026431718, "fcm_dpo/beta": 0.0019783233292400837, "fcm_dpo/delta": 0.011391036212444305, "fcm_dpo/margin": 171.1427764892578, "fcm_dpo/q_t": 0.41872942447662354, "grad_norm": 19.697229385375977, "learning_rate": 3.625819059005228e-07, "logits/chosen": -0.41778579354286194, "logits/rejected": -0.4037598967552185, "logps/chosen": -355.63037109375, "logps/ref_chosen": -66.23219299316406, "logps/ref_rejected": -99.1268310546875, "logps/rejected": -559.6677856445312, "loss": 1.1174, "margin_dpo/margin_mean": 171.14279174804688, "margin_dpo/margin_std": 203.82138061523438, "step": 285 }, { "KL/chosen_KL_mean": -287.46539306640625, "KL/mean": -386.77532958984375, "KL/rejected_KL_mean": -486.08526611328125, "KL/std": 219.8089141845703, "epoch": 0.4199706314243759, "fcm_dpo/beta": 0.001961304107680917, "fcm_dpo/delta": -0.04113783687353134, "fcm_dpo/margin": 198.61988830566406, "fcm_dpo/q_t": 0.4081750810146332, "grad_norm": 24.665103912353516, "learning_rate": 3.614345889441346e-07, "logits/chosen": -0.4218488931655884, "logits/rejected": -0.40640929341316223, "logps/chosen": -360.4164123535156, "logps/ref_chosen": -72.95100402832031, "logps/ref_rejected": -88.58845520019531, "logps/rejected": -574.6737060546875, "loss": 1.0874, "margin_dpo/margin_mean": 198.619873046875, "margin_dpo/margin_std": 240.56381225585938, "step": 286 }, { "KL/chosen_KL_mean": -281.83392333984375, "KL/mean": -363.5769958496094, "KL/rejected_KL_mean": -445.320068359375, "KL/std": 199.60064697265625, "epoch": 0.42143906020558003, "fcm_dpo/beta": 0.001964614959433675, "fcm_dpo/delta": 0.029743198305368423, "fcm_dpo/margin": 163.4861297607422, "fcm_dpo/q_t": 0.42457687854766846, "grad_norm": 21.264787673950195, "learning_rate": 3.6028433558269275e-07, "logits/chosen": -0.3639473021030426, "logits/rejected": -0.33845168352127075, "logps/chosen": -343.37506103515625, "logps/ref_chosen": -61.54115295410156, "logps/ref_rejected": -77.69607543945312, "logps/rejected": -523.01611328125, "loss": 1.1329, "margin_dpo/margin_mean": 163.4861297607422, "margin_dpo/margin_std": 215.17059326171875, "step": 287 }, { "KL/chosen_KL_mean": -272.1412048339844, "KL/mean": -371.6726989746094, "KL/rejected_KL_mean": -471.20416259765625, "KL/std": 197.98980712890625, "epoch": 0.42290748898678415, "fcm_dpo/beta": 0.0019424078054726124, "fcm_dpo/delta": -0.03923293575644493, "fcm_dpo/margin": 199.06297302246094, "fcm_dpo/q_t": 0.40794262290000916, "grad_norm": 23.609954833984375, "learning_rate": 3.5913117612644327e-07, "logits/chosen": -0.38256457448005676, "logits/rejected": -0.3666524887084961, "logps/chosen": -328.80242919921875, "logps/ref_chosen": -56.661224365234375, "logps/ref_rejected": -87.33570098876953, "logps/rejected": -558.5399169921875, "loss": 1.0781, "margin_dpo/margin_mean": 199.06297302246094, "margin_dpo/margin_std": 205.2132568359375, "step": 288 }, { "KL/chosen_KL_mean": -277.923828125, "KL/mean": -390.819091796875, "KL/rejected_KL_mean": -503.71429443359375, "KL/std": 227.17816162109375, "epoch": 0.4243759177679883, "fcm_dpo/beta": 0.001931286184117198, "fcm_dpo/delta": -0.08991145342588425, "fcm_dpo/margin": 225.79046630859375, "fcm_dpo/q_t": 0.3978983163833618, "grad_norm": 25.743732452392578, "learning_rate": 3.5797514096221024e-07, "logits/chosen": -0.3490581512451172, "logits/rejected": -0.35004904866218567, "logps/chosen": -323.15423583984375, "logps/ref_chosen": -45.23039245605469, "logps/ref_rejected": -87.64266967773438, "logps/rejected": -591.35693359375, "loss": 1.05, "margin_dpo/margin_mean": 225.7904815673828, "margin_dpo/margin_std": 243.15834045410156, "step": 289 }, { "KL/chosen_KL_mean": -311.5221252441406, "KL/mean": -425.03460693359375, "KL/rejected_KL_mean": -538.5470581054688, "KL/std": 240.62857055664062, "epoch": 0.42584434654919234, "fcm_dpo/beta": 0.0018900145078077912, "fcm_dpo/delta": -0.08279038965702057, "fcm_dpo/margin": 227.02496337890625, "fcm_dpo/q_t": 0.4012698531150818, "grad_norm": 22.0297794342041, "learning_rate": 3.568162605525952e-07, "logits/chosen": -0.3336731791496277, "logits/rejected": -0.3561541438102722, "logps/chosen": -366.9936218261719, "logps/ref_chosen": -55.47149658203125, "logps/ref_rejected": -116.70857238769531, "logps/rejected": -655.255615234375, "loss": 1.0686, "margin_dpo/margin_mean": 227.02493286132812, "margin_dpo/margin_std": 284.237060546875, "step": 290 }, { "KL/chosen_KL_mean": -282.7470703125, "KL/mean": -385.64276123046875, "KL/rejected_KL_mean": -488.5384216308594, "KL/std": 215.71090698242188, "epoch": 0.42731277533039647, "fcm_dpo/beta": 0.0018817877862602472, "fcm_dpo/delta": -0.03895223140716553, "fcm_dpo/margin": 205.79135131835938, "fcm_dpo/q_t": 0.4090266227722168, "grad_norm": 25.234416961669922, "learning_rate": 3.5565456543517485e-07, "logits/chosen": -0.33878666162490845, "logits/rejected": -0.3268297016620636, "logps/chosen": -346.0074462890625, "logps/ref_chosen": -63.26036834716797, "logps/ref_rejected": -89.29708862304688, "logps/rejected": -577.8355102539062, "loss": 1.0938, "margin_dpo/margin_mean": 205.79135131835938, "margin_dpo/margin_std": 257.6557312011719, "step": 291 }, { "KL/chosen_KL_mean": -299.9615173339844, "KL/mean": -412.6459045410156, "KL/rejected_KL_mean": -525.330322265625, "KL/std": 256.40252685546875, "epoch": 0.4287812041116006, "fcm_dpo/beta": 0.0018554746638983488, "fcm_dpo/delta": -0.07109744101762772, "fcm_dpo/margin": 225.3687744140625, "fcm_dpo/q_t": 0.40268850326538086, "grad_norm": 20.208969116210938, "learning_rate": 3.5449008622169583e-07, "logits/chosen": -0.3124640882015228, "logits/rejected": -0.2998918890953064, "logps/chosen": -353.88006591796875, "logps/ref_chosen": -53.91852951049805, "logps/ref_rejected": -89.96138000488281, "logps/rejected": -615.2916870117188, "loss": 1.0843, "margin_dpo/margin_mean": 225.3687744140625, "margin_dpo/margin_std": 301.27008056640625, "step": 292 }, { "KL/chosen_KL_mean": -296.6073303222656, "KL/mean": -382.3493347167969, "KL/rejected_KL_mean": -468.09130859375, "KL/std": 245.44065856933594, "epoch": 0.4302496328928047, "fcm_dpo/beta": 0.0018549463711678982, "fcm_dpo/delta": 0.032890163362026215, "fcm_dpo/margin": 171.4840087890625, "fcm_dpo/q_t": 0.4270939826965332, "grad_norm": 38.751129150390625, "learning_rate": 3.5332285359726846e-07, "logits/chosen": -0.35250991582870483, "logits/rejected": -0.3470958471298218, "logps/chosen": -356.98333740234375, "logps/ref_chosen": -60.376033782958984, "logps/ref_rejected": -77.85244750976562, "logps/rejected": -545.9437255859375, "loss": 1.1539, "margin_dpo/margin_mean": 171.4840087890625, "margin_dpo/margin_std": 275.24749755859375, "step": 293 }, { "KL/chosen_KL_mean": -263.4585266113281, "KL/mean": -359.516845703125, "KL/rejected_KL_mean": -455.5751953125, "KL/std": 216.63278198242188, "epoch": 0.43171806167400884, "fcm_dpo/beta": 0.0018507996574044228, "fcm_dpo/delta": -0.005858566612005234, "fcm_dpo/margin": 192.11663818359375, "fcm_dpo/q_t": 0.4179803133010864, "grad_norm": 19.413515090942383, "learning_rate": 3.5215289831955786e-07, "logits/chosen": -0.34789374470710754, "logits/rejected": -0.3532540202140808, "logps/chosen": -311.54608154296875, "logps/ref_chosen": -48.0875358581543, "logps/ref_rejected": -81.89698791503906, "logps/rejected": -537.47216796875, "loss": 1.1213, "margin_dpo/margin_mean": 192.1166534423828, "margin_dpo/margin_std": 271.4549560546875, "step": 294 }, { "KL/chosen_KL_mean": -313.67718505859375, "KL/mean": -420.85565185546875, "KL/rejected_KL_mean": -528.0341186523438, "KL/std": 247.5594482421875, "epoch": 0.4331864904552129, "fcm_dpo/beta": 0.0018429165938869119, "fcm_dpo/delta": -0.046867527067661285, "fcm_dpo/margin": 214.35690307617188, "fcm_dpo/q_t": 0.40905821323394775, "grad_norm": 30.26310157775879, "learning_rate": 3.509802512179737e-07, "logits/chosen": -0.3324083089828491, "logits/rejected": -0.33607107400894165, "logps/chosen": -363.60186767578125, "logps/ref_chosen": -49.92467498779297, "logps/ref_rejected": -87.45632934570312, "logps/rejected": -615.4904174804688, "loss": 1.0985, "margin_dpo/margin_mean": 214.35690307617188, "margin_dpo/margin_std": 291.63604736328125, "step": 295 }, { "KL/chosen_KL_mean": -384.438232421875, "KL/mean": -456.2586669921875, "KL/rejected_KL_mean": -528.0791625976562, "KL/std": 237.65313720703125, "epoch": 0.434654919236417, "fcm_dpo/beta": 0.0018318177899345756, "fcm_dpo/delta": -0.012629139237105846, "fcm_dpo/margin": 143.6409454345703, "fcm_dpo/q_t": 0.43860769271850586, "grad_norm": 27.03304672241211, "learning_rate": 3.498049431928577e-07, "logits/chosen": -0.38435792922973633, "logits/rejected": -0.368790864944458, "logps/chosen": -449.929443359375, "logps/ref_chosen": -65.49124145507812, "logps/ref_rejected": -93.08908081054688, "logps/rejected": -621.168212890625, "loss": 1.2127, "margin_dpo/margin_mean": 143.6409454345703, "margin_dpo/margin_std": 295.96697998046875, "step": 296 }, { "KL/chosen_KL_mean": -326.37908935546875, "KL/mean": -411.85205078125, "KL/rejected_KL_mean": -497.32501220703125, "KL/std": 220.0688934326172, "epoch": 0.43612334801762115, "fcm_dpo/beta": 0.0018395561492070556, "fcm_dpo/delta": 0.03670303151011467, "fcm_dpo/margin": 170.94590759277344, "fcm_dpo/q_t": 0.425690233707428, "grad_norm": 36.579345703125, "learning_rate": 3.486270052146694e-07, "logits/chosen": -0.3833608627319336, "logits/rejected": -0.3880637586116791, "logps/chosen": -382.8560485839844, "logps/ref_chosen": -56.476951599121094, "logps/ref_rejected": -95.1385498046875, "logps/rejected": -592.4635620117188, "loss": 1.1364, "margin_dpo/margin_mean": 170.94590759277344, "margin_dpo/margin_std": 221.54315185546875, "step": 297 }, { "KL/chosen_KL_mean": -345.55328369140625, "KL/mean": -456.7684326171875, "KL/rejected_KL_mean": -567.9835815429688, "KL/std": 284.334228515625, "epoch": 0.43759177679882527, "fcm_dpo/beta": 0.0018349653109908104, "fcm_dpo/delta": -0.060655344277620316, "fcm_dpo/margin": 222.43023681640625, "fcm_dpo/q_t": 0.40840664505958557, "grad_norm": 25.856088638305664, "learning_rate": 3.474464683231698e-07, "logits/chosen": -0.3994889259338379, "logits/rejected": -0.41677361726760864, "logps/chosen": -412.87847900390625, "logps/ref_chosen": -67.32516479492188, "logps/ref_rejected": -116.66217041015625, "logps/rejected": -684.645751953125, "loss": 1.1025, "margin_dpo/margin_mean": 222.43026733398438, "margin_dpo/margin_std": 331.43804931640625, "step": 298 }, { "KL/chosen_KL_mean": -279.85418701171875, "KL/mean": -374.20880126953125, "KL/rejected_KL_mean": -468.5634460449219, "KL/std": 216.16893005371094, "epoch": 0.4390602055800294, "fcm_dpo/beta": 0.00183395785279572, "fcm_dpo/delta": 0.0034361593425273895, "fcm_dpo/margin": 188.70925903320312, "fcm_dpo/q_t": 0.41983699798583984, "grad_norm": 25.584741592407227, "learning_rate": 3.462633636266041e-07, "logits/chosen": -0.40785565972328186, "logits/rejected": -0.4132392108440399, "logps/chosen": -328.8162841796875, "logps/ref_chosen": -48.96209716796875, "logps/ref_rejected": -84.32823944091797, "logps/rejected": -552.8917236328125, "loss": 1.1295, "margin_dpo/margin_mean": 188.7092742919922, "margin_dpo/margin_std": 268.2356872558594, "step": 299 }, { "KL/chosen_KL_mean": -348.40313720703125, "KL/mean": -467.8739013671875, "KL/rejected_KL_mean": -587.3447265625, "KL/std": 258.7288818359375, "epoch": 0.44052863436123346, "fcm_dpo/beta": 0.0018063622992485762, "fcm_dpo/delta": -0.08521745353937149, "fcm_dpo/margin": 238.9415740966797, "fcm_dpo/q_t": 0.40012985467910767, "grad_norm": 30.562604904174805, "learning_rate": 3.4507772230088147e-07, "logits/chosen": -0.3769553303718567, "logits/rejected": -0.38268864154815674, "logps/chosen": -407.4768371582031, "logps/ref_chosen": -59.07371139526367, "logps/ref_rejected": -95.9664535522461, "logps/rejected": -683.3111572265625, "loss": 1.0755, "margin_dpo/margin_mean": 238.9415740966797, "margin_dpo/margin_std": 310.2802734375, "step": 300 }, { "KL/chosen_KL_mean": -289.6461486816406, "KL/mean": -401.08074951171875, "KL/rejected_KL_mean": -512.5152587890625, "KL/std": 234.33538818359375, "epoch": 0.4419970631424376, "fcm_dpo/beta": 0.0017770177219063044, "fcm_dpo/delta": -0.048126157373189926, "fcm_dpo/margin": 222.86912536621094, "fcm_dpo/q_t": 0.4084852337837219, "grad_norm": 24.22814178466797, "learning_rate": 3.4388957558875316e-07, "logits/chosen": -0.3939129412174225, "logits/rejected": -0.39491885900497437, "logps/chosen": -346.8955078125, "logps/ref_chosen": -57.249366760253906, "logps/ref_rejected": -92.35354614257812, "logps/rejected": -604.8688354492188, "loss": 1.0851, "margin_dpo/margin_mean": 222.86912536621094, "margin_dpo/margin_std": 277.33050537109375, "step": 301 }, { "KL/chosen_KL_mean": -257.5218200683594, "KL/mean": -349.50299072265625, "KL/rejected_KL_mean": -441.48419189453125, "KL/std": 211.3658447265625, "epoch": 0.4434654919236417, "fcm_dpo/beta": 0.001776274642907083, "fcm_dpo/delta": 0.02384105697274208, "fcm_dpo/margin": 183.96234130859375, "fcm_dpo/q_t": 0.4237297475337982, "grad_norm": 20.491802215576172, "learning_rate": 3.426989547989902e-07, "logits/chosen": -0.4090377688407898, "logits/rejected": -0.4143017828464508, "logps/chosen": -308.7198181152344, "logps/ref_chosen": -51.197994232177734, "logps/ref_rejected": -97.22636413574219, "logps/rejected": -538.7105712890625, "loss": 1.1344, "margin_dpo/margin_mean": 183.96234130859375, "margin_dpo/margin_std": 251.2886962890625, "step": 302 }, { "KL/chosen_KL_mean": -268.2716064453125, "KL/mean": -354.656494140625, "KL/rejected_KL_mean": -441.0413513183594, "KL/std": 220.9415740966797, "epoch": 0.44493392070484583, "fcm_dpo/beta": 0.0017914820928126574, "fcm_dpo/delta": 0.04175640642642975, "fcm_dpo/margin": 172.76974487304688, "fcm_dpo/q_t": 0.4276770353317261, "grad_norm": 21.24039649963379, "learning_rate": 3.4150589130555773e-07, "logits/chosen": -0.39839571714401245, "logits/rejected": -0.3856062889099121, "logps/chosen": -334.9855651855469, "logps/ref_chosen": -66.71394348144531, "logps/ref_rejected": -86.94542694091797, "logps/rejected": -527.98681640625, "loss": 1.1539, "margin_dpo/margin_mean": 172.76976013183594, "margin_dpo/margin_std": 260.4632568359375, "step": 303 }, { "KL/chosen_KL_mean": -234.28387451171875, "KL/mean": -328.23004150390625, "KL/rejected_KL_mean": -422.1761779785156, "KL/std": 197.538330078125, "epoch": 0.44640234948604995, "fcm_dpo/beta": 0.0018079555593430996, "fcm_dpo/delta": 0.010449519380927086, "fcm_dpo/margin": 187.8922882080078, "fcm_dpo/q_t": 0.41787901520729065, "grad_norm": 34.292579650878906, "learning_rate": 3.403104165467883e-07, "logits/chosen": -0.45170116424560547, "logits/rejected": -0.44620856642723083, "logps/chosen": -306.2345886230469, "logps/ref_chosen": -71.95069885253906, "logps/ref_rejected": -90.47203063964844, "logps/rejected": -512.648193359375, "loss": 1.0992, "margin_dpo/margin_mean": 187.89227294921875, "margin_dpo/margin_std": 169.62152099609375, "step": 304 }, { "KL/chosen_KL_mean": -262.1391906738281, "KL/mean": -348.18524169921875, "KL/rejected_KL_mean": -434.2312927246094, "KL/std": 229.55484008789062, "epoch": 0.447870778267254, "fcm_dpo/beta": 0.0018056455301120877, "fcm_dpo/delta": 0.040238022804260254, "fcm_dpo/margin": 172.0921173095703, "fcm_dpo/q_t": 0.4282124638557434, "grad_norm": 27.539531707763672, "learning_rate": 3.391125620245535e-07, "logits/chosen": -0.4220506548881531, "logits/rejected": -0.4064248204231262, "logps/chosen": -328.9344177246094, "logps/ref_chosen": -66.79523468017578, "logps/ref_rejected": -92.75459289550781, "logps/rejected": -526.9859008789062, "loss": 1.152, "margin_dpo/margin_mean": 172.09213256835938, "margin_dpo/margin_std": 254.4590301513672, "step": 305 }, { "KL/chosen_KL_mean": -257.07513427734375, "KL/mean": -341.1536560058594, "KL/rejected_KL_mean": -425.232177734375, "KL/std": 204.96353149414062, "epoch": 0.44933920704845814, "fcm_dpo/beta": 0.0018320318777114153, "fcm_dpo/delta": 0.04318378493189812, "fcm_dpo/margin": 168.15704345703125, "fcm_dpo/q_t": 0.4273602068424225, "grad_norm": 24.34828758239746, "learning_rate": 3.3791235930343417e-07, "logits/chosen": -0.40770423412323, "logits/rejected": -0.38708335161209106, "logps/chosen": -326.759033203125, "logps/ref_chosen": -69.68389892578125, "logps/ref_rejected": -85.15919494628906, "logps/rejected": -510.391357421875, "loss": 1.1388, "margin_dpo/margin_mean": 168.15704345703125, "margin_dpo/margin_std": 215.92709350585938, "step": 306 }, { "KL/chosen_KL_mean": -245.8076934814453, "KL/mean": -336.43817138671875, "KL/rejected_KL_mean": -427.0686340332031, "KL/std": 195.40953063964844, "epoch": 0.45080763582966227, "fcm_dpo/beta": 0.001836513401940465, "fcm_dpo/delta": 0.017692940309643745, "fcm_dpo/margin": 181.26095581054688, "fcm_dpo/q_t": 0.42179954051971436, "grad_norm": 24.07006072998047, "learning_rate": 3.367098400098881e-07, "logits/chosen": -0.3847889304161072, "logits/rejected": -0.3667571544647217, "logps/chosen": -315.9731140136719, "logps/ref_chosen": -70.16542053222656, "logps/ref_rejected": -86.97230529785156, "logps/rejected": -514.0409545898438, "loss": 1.1268, "margin_dpo/margin_mean": 181.26095581054688, "margin_dpo/margin_std": 240.51271057128906, "step": 307 }, { "KL/chosen_KL_mean": -257.4198303222656, "KL/mean": -357.9421691894531, "KL/rejected_KL_mean": -458.4644775390625, "KL/std": 220.29737854003906, "epoch": 0.4522760646108664, "fcm_dpo/beta": 0.0018358014058321714, "fcm_dpo/delta": -0.019804438576102257, "fcm_dpo/margin": 201.04469299316406, "fcm_dpo/q_t": 0.4117882549762726, "grad_norm": 31.097227096557617, "learning_rate": 3.355050358314172e-07, "logits/chosen": -0.376716673374176, "logits/rejected": -0.3686707019805908, "logps/chosen": -312.6648254394531, "logps/ref_chosen": -55.2449951171875, "logps/ref_rejected": -79.37226104736328, "logps/rejected": -537.8367919921875, "loss": 1.0855, "margin_dpo/margin_mean": 201.04469299316406, "margin_dpo/margin_std": 205.7581024169922, "step": 308 }, { "KL/chosen_KL_mean": -267.55694580078125, "KL/mean": -369.17938232421875, "KL/rejected_KL_mean": -470.80181884765625, "KL/std": 236.59249877929688, "epoch": 0.45374449339207046, "fcm_dpo/beta": 0.0018350511090829968, "fcm_dpo/delta": -0.024183016270399094, "fcm_dpo/margin": 203.24488830566406, "fcm_dpo/q_t": 0.4112602770328522, "grad_norm": 26.018009185791016, "learning_rate": 3.3429797851573183e-07, "logits/chosen": -0.37056082487106323, "logits/rejected": -0.36778467893600464, "logps/chosen": -316.5160217285156, "logps/ref_chosen": -48.959083557128906, "logps/ref_rejected": -82.34072875976562, "logps/rejected": -553.142578125, "loss": 1.1, "margin_dpo/margin_mean": 203.244873046875, "margin_dpo/margin_std": 249.24404907226562, "step": 309 }, { "KL/chosen_KL_mean": -313.85491943359375, "KL/mean": -406.3193359375, "KL/rejected_KL_mean": -498.78375244140625, "KL/std": 212.13653564453125, "epoch": 0.4552129221732746, "fcm_dpo/beta": 0.0018333324696868658, "fcm_dpo/delta": 0.011089498177170753, "fcm_dpo/margin": 184.9288787841797, "fcm_dpo/q_t": 0.41921311616897583, "grad_norm": 34.06706619262695, "learning_rate": 3.3308869986991487e-07, "logits/chosen": -0.35188862681388855, "logits/rejected": -0.3397538661956787, "logps/chosen": -376.5966796875, "logps/ref_chosen": -62.74177932739258, "logps/ref_rejected": -79.9302978515625, "logps/rejected": -578.7140502929688, "loss": 1.1142, "margin_dpo/margin_mean": 184.92886352539062, "margin_dpo/margin_std": 212.01107788085938, "step": 310 }, { "KL/chosen_KL_mean": -326.53228759765625, "KL/mean": -431.79541015625, "KL/rejected_KL_mean": -537.0585327148438, "KL/std": 269.36871337890625, "epoch": 0.4566813509544787, "fcm_dpo/beta": 0.0018190003465861082, "fcm_dpo/delta": -0.034259166568517685, "fcm_dpo/margin": 210.52621459960938, "fcm_dpo/q_t": 0.4119686484336853, "grad_norm": 24.23558807373047, "learning_rate": 3.3187723175958346e-07, "logits/chosen": -0.3476032316684723, "logits/rejected": -0.32333052158355713, "logps/chosen": -379.5602722167969, "logps/ref_chosen": -53.02798080444336, "logps/ref_rejected": -77.43820190429688, "logps/rejected": -614.4967041015625, "loss": 1.1021, "margin_dpo/margin_mean": 210.52622985839844, "margin_dpo/margin_std": 284.9056091308594, "step": 311 }, { "KL/chosen_KL_mean": -306.4524230957031, "KL/mean": -403.4335021972656, "KL/rejected_KL_mean": -500.41461181640625, "KL/std": 241.86181640625, "epoch": 0.4581497797356828, "fcm_dpo/beta": 0.0018220641650259495, "fcm_dpo/delta": -0.0038466081023216248, "fcm_dpo/margin": 193.96221923828125, "fcm_dpo/q_t": 0.4196726381778717, "grad_norm": 23.693248748779297, "learning_rate": 3.306636061080487e-07, "logits/chosen": -0.35380294919013977, "logits/rejected": -0.3449653089046478, "logps/chosen": -355.8446350097656, "logps/ref_chosen": -49.39221954345703, "logps/ref_rejected": -75.79280853271484, "logps/rejected": -576.2073974609375, "loss": 1.1271, "margin_dpo/margin_mean": 193.96221923828125, "margin_dpo/margin_std": 287.26824951171875, "step": 312 }, { "KL/chosen_KL_mean": -283.44970703125, "KL/mean": -383.09356689453125, "KL/rejected_KL_mean": -482.7374267578125, "KL/std": 249.87867736816406, "epoch": 0.45961820851688695, "fcm_dpo/beta": 0.001822406891733408, "fcm_dpo/delta": -0.014451660215854645, "fcm_dpo/margin": 199.28773498535156, "fcm_dpo/q_t": 0.4155638515949249, "grad_norm": 23.18164825439453, "learning_rate": 3.2944785489547537e-07, "logits/chosen": -0.4175724983215332, "logits/rejected": -0.41589704155921936, "logps/chosen": -333.60247802734375, "logps/ref_chosen": -50.152740478515625, "logps/ref_rejected": -86.40620422363281, "logps/rejected": -569.1436767578125, "loss": 1.1203, "margin_dpo/margin_mean": 199.2877197265625, "margin_dpo/margin_std": 278.8194580078125, "step": 313 }, { "KL/chosen_KL_mean": -298.1259765625, "KL/mean": -390.4653015136719, "KL/rejected_KL_mean": -482.80462646484375, "KL/std": 232.75111389160156, "epoch": 0.461086637298091, "fcm_dpo/beta": 0.0018075141124427319, "fcm_dpo/delta": 0.016570214182138443, "fcm_dpo/margin": 184.67868041992188, "fcm_dpo/q_t": 0.42391207814216614, "grad_norm": 20.25391960144043, "learning_rate": 3.2823001015803857e-07, "logits/chosen": -0.40884333848953247, "logits/rejected": -0.4110562801361084, "logps/chosen": -355.3635559082031, "logps/ref_chosen": -57.237579345703125, "logps/ref_rejected": -97.5965347290039, "logps/rejected": -580.4011840820312, "loss": 1.1494, "margin_dpo/margin_mean": 184.67868041992188, "margin_dpo/margin_std": 298.83831787109375, "step": 314 }, { "KL/chosen_KL_mean": -263.869873046875, "KL/mean": -349.53668212890625, "KL/rejected_KL_mean": -435.2034606933594, "KL/std": 211.77670288085938, "epoch": 0.46255506607929514, "fcm_dpo/beta": 0.001823345897719264, "fcm_dpo/delta": 0.03882795572280884, "fcm_dpo/margin": 171.33358764648438, "fcm_dpo/q_t": 0.42688193917274475, "grad_norm": 19.43107795715332, "learning_rate": 3.270101039870797e-07, "logits/chosen": -0.3321152329444885, "logits/rejected": -0.33484265208244324, "logps/chosen": -312.939453125, "logps/ref_chosen": -49.06958770751953, "logps/ref_rejected": -85.68087768554688, "logps/rejected": -520.8843383789062, "loss": 1.142, "margin_dpo/margin_mean": 171.33358764648438, "margin_dpo/margin_std": 235.25616455078125, "step": 315 }, { "KL/chosen_KL_mean": -255.12551879882812, "KL/mean": -380.3026123046875, "KL/rejected_KL_mean": -505.4797058105469, "KL/std": 246.68731689453125, "epoch": 0.46402349486049926, "fcm_dpo/beta": 0.0017926250584423542, "fcm_dpo/delta": -0.10438720881938934, "fcm_dpo/margin": 250.35418701171875, "fcm_dpo/q_t": 0.39585646986961365, "grad_norm": 27.900249481201172, "learning_rate": 3.2578816852826086e-07, "logits/chosen": -0.37772274017333984, "logits/rejected": -0.38135725259780884, "logps/chosen": -309.3862609863281, "logps/ref_chosen": -54.26074981689453, "logps/ref_rejected": -101.2814712524414, "logps/rejected": -606.7611694335938, "loss": 1.043, "margin_dpo/margin_mean": 250.3542022705078, "margin_dpo/margin_std": 271.3267517089844, "step": 316 }, { "KL/chosen_KL_mean": -264.39501953125, "KL/mean": -394.56658935546875, "KL/rejected_KL_mean": -524.7380981445312, "KL/std": 222.24476623535156, "epoch": 0.4654919236417034, "fcm_dpo/beta": 0.0017680514138191938, "fcm_dpo/delta": -0.11548471450805664, "fcm_dpo/margin": 260.3431091308594, "fcm_dpo/q_t": 0.39133790135383606, "grad_norm": 27.655803680419922, "learning_rate": 3.2456423598071783e-07, "logits/chosen": -0.40937548875808716, "logits/rejected": -0.4001927971839905, "logps/chosen": -320.4892578125, "logps/ref_chosen": -56.094207763671875, "logps/ref_rejected": -100.69905090332031, "logps/rejected": -625.4371337890625, "loss": 1.0271, "margin_dpo/margin_mean": 260.3431091308594, "margin_dpo/margin_std": 250.1036376953125, "step": 317 }, { "KL/chosen_KL_mean": -297.4501953125, "KL/mean": -396.00433349609375, "KL/rejected_KL_mean": -494.5584716796875, "KL/std": 234.63751220703125, "epoch": 0.4669603524229075, "fcm_dpo/beta": 0.0017554573714733124, "fcm_dpo/delta": 0.004025392234325409, "fcm_dpo/margin": 197.1082763671875, "fcm_dpo/q_t": 0.4192516803741455, "grad_norm": 23.755502700805664, "learning_rate": 3.233383385962115e-07, "logits/chosen": -0.47140195965766907, "logits/rejected": -0.44852572679519653, "logps/chosen": -362.09588623046875, "logps/ref_chosen": -64.64569854736328, "logps/ref_rejected": -82.76425170898438, "logps/rejected": -577.32275390625, "loss": 1.1185, "margin_dpo/margin_mean": 197.1082763671875, "margin_dpo/margin_std": 258.5344543457031, "step": 318 }, { "KL/chosen_KL_mean": -272.87457275390625, "KL/mean": -392.9670104980469, "KL/rejected_KL_mean": -513.0595092773438, "KL/std": 259.213134765625, "epoch": 0.4684287812041116, "fcm_dpo/beta": 0.0017336711753159761, "fcm_dpo/delta": -0.06930382549762726, "fcm_dpo/margin": 240.18490600585938, "fcm_dpo/q_t": 0.4022204279899597, "grad_norm": 23.22355842590332, "learning_rate": 3.2211050867827805e-07, "logits/chosen": -0.3927324414253235, "logits/rejected": -0.4057486057281494, "logps/chosen": -322.25830078125, "logps/ref_chosen": -49.383758544921875, "logps/ref_rejected": -113.90650939941406, "logps/rejected": -626.9660034179688, "loss": 1.0671, "margin_dpo/margin_mean": 240.18490600585938, "margin_dpo/margin_std": 278.5548095703125, "step": 319 }, { "KL/chosen_KL_mean": -305.7335205078125, "KL/mean": -437.4515380859375, "KL/rejected_KL_mean": -569.1694946289062, "KL/std": 278.39447021484375, "epoch": 0.4698972099853157, "fcm_dpo/beta": 0.0016967777628451586, "fcm_dpo/delta": -0.10202349722385406, "fcm_dpo/margin": 263.43597412109375, "fcm_dpo/q_t": 0.39611712098121643, "grad_norm": 24.74642562866211, "learning_rate": 3.208807785813777e-07, "logits/chosen": -0.4070661664009094, "logits/rejected": -0.41442254185676575, "logps/chosen": -365.2384033203125, "logps/ref_chosen": -59.50489044189453, "logps/ref_rejected": -97.66717529296875, "logps/rejected": -666.836669921875, "loss": 1.0469, "margin_dpo/margin_mean": 263.43597412109375, "margin_dpo/margin_std": 295.5460205078125, "step": 320 }, { "KL/chosen_KL_mean": -385.0860900878906, "KL/mean": -497.9861145019531, "KL/rejected_KL_mean": -610.8861694335938, "KL/std": 285.7553405761719, "epoch": 0.4713656387665198, "fcm_dpo/beta": 0.0016765834297984838, "fcm_dpo/delta": -0.03041578084230423, "fcm_dpo/margin": 225.800048828125, "fcm_dpo/q_t": 0.41292014718055725, "grad_norm": 27.752023696899414, "learning_rate": 3.1964918071004217e-07, "logits/chosen": -0.37062233686447144, "logits/rejected": -0.35946518182754517, "logps/chosen": -446.634765625, "logps/ref_chosen": -61.548683166503906, "logps/ref_rejected": -91.64103698730469, "logps/rejected": -702.5272216796875, "loss": 1.108, "margin_dpo/margin_mean": 225.80007934570312, "margin_dpo/margin_std": 308.14105224609375, "step": 321 }, { "KL/chosen_KL_mean": -346.7474365234375, "KL/mean": -466.8512268066406, "KL/rejected_KL_mean": -586.9550170898438, "KL/std": 259.3096923828125, "epoch": 0.47283406754772395, "fcm_dpo/beta": 0.001664046081714332, "fcm_dpo/delta": -0.0522020198404789, "fcm_dpo/margin": 240.2075958251953, "fcm_dpo/q_t": 0.4059278964996338, "grad_norm": 28.33850860595703, "learning_rate": 3.184157475180207e-07, "logits/chosen": -0.38744407892227173, "logits/rejected": -0.38698631525039673, "logps/chosen": -404.0374755859375, "logps/ref_chosen": -57.29003143310547, "logps/ref_rejected": -95.74992370605469, "logps/rejected": -682.7049560546875, "loss": 1.0774, "margin_dpo/margin_mean": 240.20761108398438, "margin_dpo/margin_std": 275.8708801269531, "step": 322 }, { "KL/chosen_KL_mean": -360.1468811035156, "KL/mean": -461.87078857421875, "KL/rejected_KL_mean": -563.5946655273438, "KL/std": 256.2729797363281, "epoch": 0.47430249632892807, "fcm_dpo/beta": 0.0016664534341543913, "fcm_dpo/delta": 0.011339662596583366, "fcm_dpo/margin": 203.44778442382812, "fcm_dpo/q_t": 0.4201313257217407, "grad_norm": 36.73489761352539, "learning_rate": 3.171805115074251e-07, "logits/chosen": -0.3776964545249939, "logits/rejected": -0.37534666061401367, "logps/chosen": -411.380859375, "logps/ref_chosen": -51.23395919799805, "logps/ref_rejected": -75.06192016601562, "logps/rejected": -638.6566162109375, "loss": 1.1293, "margin_dpo/margin_mean": 203.44778442382812, "margin_dpo/margin_std": 275.85626220703125, "step": 323 }, { "KL/chosen_KL_mean": -409.14288330078125, "KL/mean": -503.98602294921875, "KL/rejected_KL_mean": -598.8291625976562, "KL/std": 269.3505859375, "epoch": 0.47577092511013214, "fcm_dpo/beta": 0.0016613883199170232, "fcm_dpo/delta": -0.046242307871580124, "fcm_dpo/margin": 189.686279296875, "fcm_dpo/q_t": 0.42781496047973633, "grad_norm": 57.406150817871094, "learning_rate": 3.1594350522787295e-07, "logits/chosen": -0.40174317359924316, "logits/rejected": -0.385235071182251, "logps/chosen": -474.2780456542969, "logps/ref_chosen": -65.13516998291016, "logps/ref_rejected": -86.47750854492188, "logps/rejected": -685.306640625, "loss": 1.1672, "margin_dpo/margin_mean": 189.68629455566406, "margin_dpo/margin_std": 315.1155700683594, "step": 324 }, { "KL/chosen_KL_mean": -305.6237487792969, "KL/mean": -388.1728515625, "KL/rejected_KL_mean": -470.721923828125, "KL/std": 234.14627075195312, "epoch": 0.47723935389133626, "fcm_dpo/beta": 0.0016727466136217117, "fcm_dpo/delta": 0.07592117786407471, "fcm_dpo/margin": 165.09820556640625, "fcm_dpo/q_t": 0.43507638573646545, "grad_norm": 27.068408966064453, "learning_rate": 3.147047612756302e-07, "logits/chosen": -0.42593660950660706, "logits/rejected": -0.40495866537094116, "logps/chosen": -361.83935546875, "logps/ref_chosen": -56.215599060058594, "logps/ref_rejected": -70.08592987060547, "logps/rejected": -540.807861328125, "loss": 1.1704, "margin_dpo/margin_mean": 165.09820556640625, "margin_dpo/margin_std": 239.03897094726562, "step": 325 }, { "KL/chosen_KL_mean": -340.3418273925781, "KL/mean": -420.40203857421875, "KL/rejected_KL_mean": -500.4622497558594, "KL/std": 218.0757293701172, "epoch": 0.4787077826725404, "fcm_dpo/beta": 0.0016966955736279488, "fcm_dpo/delta": 0.0805881917476654, "fcm_dpo/margin": 160.12042236328125, "fcm_dpo/q_t": 0.43505293130874634, "grad_norm": 36.45564651489258, "learning_rate": 3.134643122927519e-07, "logits/chosen": -0.43507713079452515, "logits/rejected": -0.40912318229675293, "logps/chosen": -413.0667724609375, "logps/ref_chosen": -72.72496032714844, "logps/ref_rejected": -79.8467788696289, "logps/rejected": -580.3090209960938, "loss": 1.1649, "margin_dpo/margin_mean": 160.12042236328125, "margin_dpo/margin_std": 207.34747314453125, "step": 326 }, { "KL/chosen_KL_mean": -287.189208984375, "KL/mean": -408.7220153808594, "KL/rejected_KL_mean": -530.2548217773438, "KL/std": 226.6089324951172, "epoch": 0.4801762114537445, "fcm_dpo/beta": 0.001686271745711565, "fcm_dpo/delta": -0.06271137297153473, "fcm_dpo/margin": 243.0655517578125, "fcm_dpo/q_t": 0.40232932567596436, "grad_norm": 45.123172760009766, "learning_rate": 3.1222219096622264e-07, "logits/chosen": -0.4474729597568512, "logits/rejected": -0.4362829327583313, "logps/chosen": -356.32366943359375, "logps/ref_chosen": -69.13441467285156, "logps/ref_rejected": -111.93377685546875, "logps/rejected": -642.1885986328125, "loss": 1.0628, "margin_dpo/margin_mean": 243.0655517578125, "margin_dpo/margin_std": 256.05047607421875, "step": 327 }, { "KL/chosen_KL_mean": -290.7843933105469, "KL/mean": -401.93585205078125, "KL/rejected_KL_mean": -513.0873413085938, "KL/std": 251.8201446533203, "epoch": 0.48164464023494863, "fcm_dpo/beta": 0.0016804768238216639, "fcm_dpo/delta": -0.02448493242263794, "fcm_dpo/margin": 222.30294799804688, "fcm_dpo/q_t": 0.4131455719470978, "grad_norm": 26.31197166442871, "learning_rate": 3.1097843002709427e-07, "logits/chosen": -0.437290757894516, "logits/rejected": -0.4402696490287781, "logps/chosen": -350.4715881347656, "logps/ref_chosen": -59.68719482421875, "logps/ref_rejected": -90.85499572753906, "logps/rejected": -603.9423828125, "loss": 1.0994, "margin_dpo/margin_mean": 222.30294799804688, "margin_dpo/margin_std": 282.1927490234375, "step": 328 }, { "KL/chosen_KL_mean": -320.8580017089844, "KL/mean": -433.61395263671875, "KL/rejected_KL_mean": -546.369873046875, "KL/std": 260.9647216796875, "epoch": 0.4831130690161527, "fcm_dpo/beta": 0.0016629232559353113, "fcm_dpo/delta": -0.026486670598387718, "fcm_dpo/margin": 225.51190185546875, "fcm_dpo/q_t": 0.41196945309638977, "grad_norm": 24.17496109008789, "learning_rate": 3.0973306224962437e-07, "logits/chosen": -0.42437106370925903, "logits/rejected": -0.41386279463768005, "logps/chosen": -386.10418701171875, "logps/ref_chosen": -65.2461929321289, "logps/ref_rejected": -100.69770812988281, "logps/rejected": -647.067626953125, "loss": 1.1029, "margin_dpo/margin_mean": 225.51190185546875, "margin_dpo/margin_std": 284.83648681640625, "step": 329 }, { "KL/chosen_KL_mean": -273.79058837890625, "KL/mean": -391.7522888183594, "KL/rejected_KL_mean": -509.7140197753906, "KL/std": 246.3189697265625, "epoch": 0.4845814977973568, "fcm_dpo/beta": 0.001657246146351099, "fcm_dpo/delta": -0.04265393316745758, "fcm_dpo/margin": 235.92340087890625, "fcm_dpo/q_t": 0.4079107344150543, "grad_norm": 26.65233039855957, "learning_rate": 3.084861204504122e-07, "logits/chosen": -0.38325101137161255, "logits/rejected": -0.384086012840271, "logps/chosen": -320.7889404296875, "logps/ref_chosen": -46.998348236083984, "logps/ref_rejected": -86.87684631347656, "logps/rejected": -596.5908203125, "loss": 1.0803, "margin_dpo/margin_mean": 235.92337036132812, "margin_dpo/margin_std": 269.93707275390625, "step": 330 }, { "KL/chosen_KL_mean": -305.1080017089844, "KL/mean": -425.6927185058594, "KL/rejected_KL_mean": -546.2774658203125, "KL/std": 213.3463134765625, "epoch": 0.48604992657856094, "fcm_dpo/beta": 0.001642332412302494, "fcm_dpo/delta": -0.04797166585922241, "fcm_dpo/margin": 241.16941833496094, "fcm_dpo/q_t": 0.4048876166343689, "grad_norm": 22.422067642211914, "learning_rate": 3.072376374875335e-07, "logits/chosen": -0.4126804769039154, "logits/rejected": -0.40818193554878235, "logps/chosen": -355.63226318359375, "logps/ref_chosen": -50.52424621582031, "logps/ref_rejected": -89.01544189453125, "logps/rejected": -635.2928466796875, "loss": 1.0563, "margin_dpo/margin_mean": 241.16941833496094, "margin_dpo/margin_std": 207.7781982421875, "step": 331 }, { "KL/chosen_KL_mean": -311.51898193359375, "KL/mean": -397.1397705078125, "KL/rejected_KL_mean": -482.76055908203125, "KL/std": 219.59471130371094, "epoch": 0.48751835535976507, "fcm_dpo/beta": 0.0016566382255405188, "fcm_dpo/delta": 0.06814493983983994, "fcm_dpo/margin": 171.24156188964844, "fcm_dpo/q_t": 0.4337378740310669, "grad_norm": 18.958757400512695, "learning_rate": 3.059876462596758e-07, "logits/chosen": -0.42918598651885986, "logits/rejected": -0.40887415409088135, "logps/chosen": -360.69927978515625, "logps/ref_chosen": -49.18028259277344, "logps/ref_rejected": -76.48515319824219, "logps/rejected": -559.2457275390625, "loss": 1.1626, "margin_dpo/margin_mean": 171.24156188964844, "margin_dpo/margin_std": 248.4321746826172, "step": 332 }, { "KL/chosen_KL_mean": -342.23876953125, "KL/mean": -454.83258056640625, "KL/rejected_KL_mean": -567.4263916015625, "KL/std": 263.9354248046875, "epoch": 0.4889867841409692, "fcm_dpo/beta": 0.001645524287596345, "fcm_dpo/delta": -0.022140901535749435, "fcm_dpo/margin": 225.18756103515625, "fcm_dpo/q_t": 0.4143007695674896, "grad_norm": 23.2531795501709, "learning_rate": 3.0473617970527015e-07, "logits/chosen": -0.43771564960479736, "logits/rejected": -0.43037185072898865, "logps/chosen": -405.9945068359375, "logps/ref_chosen": -63.75574493408203, "logps/ref_rejected": -95.04411315917969, "logps/rejected": -662.470458984375, "loss": 1.1188, "margin_dpo/margin_mean": 225.1875762939453, "margin_dpo/margin_std": 320.94561767578125, "step": 333 }, { "KL/chosen_KL_mean": -339.3734130859375, "KL/mean": -440.17437744140625, "KL/rejected_KL_mean": -540.975341796875, "KL/std": 291.3700866699219, "epoch": 0.49045521292217326, "fcm_dpo/beta": 0.001655098283663392, "fcm_dpo/delta": 0.01687953621149063, "fcm_dpo/margin": 201.60191345214844, "fcm_dpo/q_t": 0.4232845902442932, "grad_norm": 21.963340759277344, "learning_rate": 3.034832708016243e-07, "logits/chosen": -0.4199391007423401, "logits/rejected": -0.41611427068710327, "logps/chosen": -406.3531494140625, "logps/ref_chosen": -66.97975158691406, "logps/ref_rejected": -95.31692504882812, "logps/rejected": -636.292236328125, "loss": 1.1508, "margin_dpo/margin_mean": 201.6019287109375, "margin_dpo/margin_std": 326.2679443359375, "step": 334 }, { "KL/chosen_KL_mean": -366.4427185058594, "KL/mean": -445.76446533203125, "KL/rejected_KL_mean": -525.086181640625, "KL/std": 274.3227844238281, "epoch": 0.4919236417033774, "fcm_dpo/beta": 0.0016588786384090781, "fcm_dpo/delta": 0.001063960837200284, "fcm_dpo/margin": 158.6435089111328, "fcm_dpo/q_t": 0.4391080439090729, "grad_norm": 30.97983741760254, "learning_rate": 3.022289525640531e-07, "logits/chosen": -0.4851636588573456, "logits/rejected": -0.4646086096763611, "logps/chosen": -428.9851989746094, "logps/ref_chosen": -62.54248046875, "logps/ref_rejected": -87.61770629882812, "logps/rejected": -612.7039184570312, "loss": 1.1978, "margin_dpo/margin_mean": 158.64349365234375, "margin_dpo/margin_std": 290.5368957519531, "step": 335 }, { "KL/chosen_KL_mean": -371.8771667480469, "KL/mean": -501.9913330078125, "KL/rejected_KL_mean": -632.10546875, "KL/std": 316.42431640625, "epoch": 0.4933920704845815, "fcm_dpo/beta": 0.0016450767870992422, "fcm_dpo/delta": -0.08156316727399826, "fcm_dpo/margin": 260.2283020019531, "fcm_dpo/q_t": 0.4031341075897217, "grad_norm": 30.599864959716797, "learning_rate": 3.009732580450086e-07, "logits/chosen": -0.4376946687698364, "logits/rejected": -0.4382895231246948, "logps/chosen": -426.4083251953125, "logps/ref_chosen": -54.53115463256836, "logps/ref_rejected": -104.40424346923828, "logps/rejected": -736.5097045898438, "loss": 1.0942, "margin_dpo/margin_mean": 260.2283020019531, "margin_dpo/margin_std": 375.32147216796875, "step": 336 }, { "KL/chosen_KL_mean": -335.18377685546875, "KL/mean": -462.601806640625, "KL/rejected_KL_mean": -590.0198364257812, "KL/std": 248.90377807617188, "epoch": 0.4948604992657856, "fcm_dpo/beta": 0.0016133894678205252, "fcm_dpo/delta": -0.06399152427911758, "fcm_dpo/margin": 254.83612060546875, "fcm_dpo/q_t": 0.40399160981178284, "grad_norm": 21.82675552368164, "learning_rate": 2.9971622033320914e-07, "logits/chosen": -0.46123170852661133, "logits/rejected": -0.44864845275878906, "logps/chosen": -400.31243896484375, "logps/ref_chosen": -65.12869262695312, "logps/ref_rejected": -101.72701263427734, "logps/rejected": -691.746826171875, "loss": 1.07, "margin_dpo/margin_mean": 254.83612060546875, "margin_dpo/margin_std": 295.6842041015625, "step": 337 }, { "KL/chosen_KL_mean": -298.3284606933594, "KL/mean": -423.231201171875, "KL/rejected_KL_mean": -548.1339111328125, "KL/std": 241.73153686523438, "epoch": 0.49632892804698975, "fcm_dpo/beta": 0.0015977486036717892, "fcm_dpo/delta": -0.05123209208250046, "fcm_dpo/margin": 249.80540466308594, "fcm_dpo/q_t": 0.4050595760345459, "grad_norm": 26.014801025390625, "learning_rate": 2.984578725527675e-07, "logits/chosen": -0.4533551335334778, "logits/rejected": -0.44745928049087524, "logps/chosen": -356.75115966796875, "logps/ref_chosen": -58.422706604003906, "logps/ref_rejected": -89.06854248046875, "logps/rejected": -637.202392578125, "loss": 1.0652, "margin_dpo/margin_mean": 249.805419921875, "margin_dpo/margin_std": 252.97518920898438, "step": 338 }, { "KL/chosen_KL_mean": -320.6482849121094, "KL/mean": -444.5311584472656, "KL/rejected_KL_mean": -568.4140625, "KL/std": 257.1028137207031, "epoch": 0.4977973568281938, "fcm_dpo/beta": 0.0015966382343322039, "fcm_dpo/delta": -0.04843657463788986, "fcm_dpo/margin": 247.76576232910156, "fcm_dpo/q_t": 0.4064652621746063, "grad_norm": 23.043846130371094, "learning_rate": 2.9719824786231796e-07, "logits/chosen": -0.5105829238891602, "logits/rejected": -0.49902158975601196, "logps/chosen": -380.64361572265625, "logps/ref_chosen": -59.99531555175781, "logps/ref_rejected": -103.9109115600586, "logps/rejected": -672.324951171875, "loss": 1.0773, "margin_dpo/margin_mean": 247.7657470703125, "margin_dpo/margin_std": 259.73455810546875, "step": 339 }, { "KL/chosen_KL_mean": -344.79083251953125, "KL/mean": -446.81793212890625, "KL/rejected_KL_mean": -548.844970703125, "KL/std": 248.848876953125, "epoch": 0.49926578560939794, "fcm_dpo/beta": 0.0015804399736225605, "fcm_dpo/delta": 0.02824411541223526, "fcm_dpo/margin": 204.05410766601562, "fcm_dpo/q_t": 0.4249947667121887, "grad_norm": 24.647438049316406, "learning_rate": 2.959373794541426e-07, "logits/chosen": -0.4062250852584839, "logits/rejected": -0.38350969552993774, "logps/chosen": -397.62109375, "logps/ref_chosen": -52.83022689819336, "logps/ref_rejected": -73.10723114013672, "logps/rejected": -621.9522094726562, "loss": 1.1452, "margin_dpo/margin_mean": 204.05410766601562, "margin_dpo/margin_std": 303.45513916015625, "step": 340 }, { "KL/chosen_KL_mean": -344.22686767578125, "KL/mean": -463.89288330078125, "KL/rejected_KL_mean": -583.558837890625, "KL/std": 270.5447082519531, "epoch": 0.5007342143906021, "fcm_dpo/beta": 0.001576824113726616, "fcm_dpo/delta": -0.028752204030752182, "fcm_dpo/margin": 239.3319549560547, "fcm_dpo/q_t": 0.41167351603507996, "grad_norm": 22.612895965576172, "learning_rate": 2.946753005532965e-07, "logits/chosen": -0.4313828945159912, "logits/rejected": -0.42949026823043823, "logps/chosen": -392.126708984375, "logps/ref_chosen": -47.899803161621094, "logps/ref_rejected": -101.80987548828125, "logps/rejected": -685.3687133789062, "loss": 1.0948, "margin_dpo/margin_mean": 239.33193969726562, "margin_dpo/margin_std": 287.29522705078125, "step": 341 }, { "KL/chosen_KL_mean": -340.2489318847656, "KL/mean": -441.2088317871094, "KL/rejected_KL_mean": -542.168701171875, "KL/std": 256.6587829589844, "epoch": 0.5022026431718062, "fcm_dpo/beta": 0.0015923329629004002, "fcm_dpo/delta": 0.029024792835116386, "fcm_dpo/margin": 201.91983032226562, "fcm_dpo/q_t": 0.4247163236141205, "grad_norm": 23.06468391418457, "learning_rate": 2.934120444167326e-07, "logits/chosen": -0.46542230248451233, "logits/rejected": -0.43691468238830566, "logps/chosen": -412.2455749511719, "logps/ref_chosen": -71.99664306640625, "logps/ref_rejected": -92.58959197998047, "logps/rejected": -634.75830078125, "loss": 1.1482, "margin_dpo/margin_mean": 201.91983032226562, "margin_dpo/margin_std": 308.49237060546875, "step": 342 }, { "KL/chosen_KL_mean": -311.03472900390625, "KL/mean": -440.37176513671875, "KL/rejected_KL_mean": -569.708740234375, "KL/std": 256.6684875488281, "epoch": 0.5036710719530103, "fcm_dpo/beta": 0.0015744275879114866, "fcm_dpo/delta": -0.05970253422856331, "fcm_dpo/margin": 258.6740417480469, "fcm_dpo/q_t": 0.40231946110725403, "grad_norm": 19.85938262939453, "learning_rate": 2.9214764433242476e-07, "logits/chosen": -0.4849863052368164, "logits/rejected": -0.4904642403125763, "logps/chosen": -365.44036865234375, "logps/ref_chosen": -54.405616760253906, "logps/ref_rejected": -111.04142761230469, "logps/rejected": -680.7501220703125, "loss": 1.0525, "margin_dpo/margin_mean": 258.67401123046875, "margin_dpo/margin_std": 228.75125122070312, "step": 343 }, { "KL/chosen_KL_mean": -311.48944091796875, "KL/mean": -426.0353698730469, "KL/rejected_KL_mean": -540.581298828125, "KL/std": 284.60888671875, "epoch": 0.5051395007342144, "fcm_dpo/beta": 0.0015560141764581203, "fcm_dpo/delta": -0.0887831598520279, "fcm_dpo/margin": 229.09188842773438, "fcm_dpo/q_t": 0.41803035140037537, "grad_norm": 29.873308181762695, "learning_rate": 2.9088213361849126e-07, "logits/chosen": -0.47299444675445557, "logits/rejected": -0.4747607707977295, "logps/chosen": -365.4541015625, "logps/ref_chosen": -53.96466827392578, "logps/ref_rejected": -90.62336730957031, "logps/rejected": -631.2046508789062, "loss": 1.1164, "margin_dpo/margin_mean": 229.09188842773438, "margin_dpo/margin_std": 292.5248718261719, "step": 344 }, { "KL/chosen_KL_mean": -370.3597106933594, "KL/mean": -495.03125, "KL/rejected_KL_mean": -619.7027587890625, "KL/std": 266.9612731933594, "epoch": 0.5066079295154186, "fcm_dpo/beta": 0.0015359357930719852, "fcm_dpo/delta": -0.03427436202764511, "fcm_dpo/margin": 249.343017578125, "fcm_dpo/q_t": 0.4097304940223694, "grad_norm": 23.3608455657959, "learning_rate": 2.896155456223163e-07, "logits/chosen": -0.4813809394836426, "logits/rejected": -0.4792475700378418, "logps/chosen": -432.04541015625, "logps/ref_chosen": -61.685699462890625, "logps/ref_rejected": -99.49041748046875, "logps/rejected": -719.193115234375, "loss": 1.0872, "margin_dpo/margin_mean": 249.34304809570312, "margin_dpo/margin_std": 292.41497802734375, "step": 345 }, { "KL/chosen_KL_mean": -363.1270751953125, "KL/mean": -480.3489685058594, "KL/rejected_KL_mean": -597.5708618164062, "KL/std": 265.65924072265625, "epoch": 0.5080763582966226, "fcm_dpo/beta": 0.0015292272437363863, "fcm_dpo/delta": -0.008838503621518612, "fcm_dpo/margin": 234.44375610351562, "fcm_dpo/q_t": 0.4146372973918915, "grad_norm": 20.73213005065918, "learning_rate": 2.883479137196714e-07, "logits/chosen": -0.44797492027282715, "logits/rejected": -0.4381539821624756, "logps/chosen": -418.38336181640625, "logps/ref_chosen": -55.256263732910156, "logps/ref_rejected": -77.41532135009766, "logps/rejected": -674.9862060546875, "loss": 1.1038, "margin_dpo/margin_mean": 234.4437713623047, "margin_dpo/margin_std": 281.268310546875, "step": 346 }, { "KL/chosen_KL_mean": -360.153564453125, "KL/mean": -475.7332763671875, "KL/rejected_KL_mean": -591.31298828125, "KL/std": 275.2700500488281, "epoch": 0.5095447870778267, "fcm_dpo/beta": 0.0015283192042261362, "fcm_dpo/delta": -0.0034055609721690416, "fcm_dpo/margin": 231.15940856933594, "fcm_dpo/q_t": 0.417599081993103, "grad_norm": 22.630468368530273, "learning_rate": 2.8707927131383614e-07, "logits/chosen": -0.4328559637069702, "logits/rejected": -0.42671138048171997, "logps/chosen": -417.71978759765625, "logps/ref_chosen": -57.56623840332031, "logps/ref_rejected": -92.35509490966797, "logps/rejected": -683.6680908203125, "loss": 1.1179, "margin_dpo/margin_mean": 231.15939331054688, "margin_dpo/margin_std": 312.8233642578125, "step": 347 }, { "KL/chosen_KL_mean": -321.64593505859375, "KL/mean": -419.4115905761719, "KL/rejected_KL_mean": -517.1773071289062, "KL/std": 238.86471557617188, "epoch": 0.5110132158590308, "fcm_dpo/beta": 0.0015206331154331565, "fcm_dpo/delta": -0.02389657124876976, "fcm_dpo/margin": 195.53135681152344, "fcm_dpo/q_t": 0.4300573468208313, "grad_norm": 22.362464904785156, "learning_rate": 2.858096518347179e-07, "logits/chosen": -0.4757729768753052, "logits/rejected": -0.47732916474342346, "logps/chosen": -377.963623046875, "logps/ref_chosen": -56.31770324707031, "logps/ref_rejected": -89.13836669921875, "logps/rejected": -606.315673828125, "loss": 1.1552, "margin_dpo/margin_mean": 195.53135681152344, "margin_dpo/margin_std": 277.104248046875, "step": 348 }, { "KL/chosen_KL_mean": -298.3360595703125, "KL/mean": -417.1045227050781, "KL/rejected_KL_mean": -535.8729858398438, "KL/std": 269.330322265625, "epoch": 0.5124816446402349, "fcm_dpo/beta": 0.0015224895905703306, "fcm_dpo/delta": -0.01223532110452652, "fcm_dpo/margin": 237.53692626953125, "fcm_dpo/q_t": 0.4166039228439331, "grad_norm": 18.555866241455078, "learning_rate": 2.845390887379706e-07, "logits/chosen": -0.4357994794845581, "logits/rejected": -0.43820104002952576, "logps/chosen": -356.361572265625, "logps/ref_chosen": -58.025516510009766, "logps/ref_rejected": -97.50515747070312, "logps/rejected": -633.378173828125, "loss": 1.1159, "margin_dpo/margin_mean": 237.5369415283203, "margin_dpo/margin_std": 327.4973449707031, "step": 349 }, { "KL/chosen_KL_mean": -326.14739990234375, "KL/mean": -436.9446716308594, "KL/rejected_KL_mean": -547.741943359375, "KL/std": 256.19500732421875, "epoch": 0.5139500734214391, "fcm_dpo/beta": 0.001514824340119958, "fcm_dpo/delta": 0.014613550156354904, "fcm_dpo/margin": 221.5946044921875, "fcm_dpo/q_t": 0.4211843013763428, "grad_norm": 25.532001495361328, "learning_rate": 2.8326761550411346e-07, "logits/chosen": -0.4588230848312378, "logits/rejected": -0.4639089107513428, "logps/chosen": -390.4779052734375, "logps/ref_chosen": -64.33049011230469, "logps/ref_rejected": -89.87164306640625, "logps/rejected": -637.6136474609375, "loss": 1.1337, "margin_dpo/margin_mean": 221.5946044921875, "margin_dpo/margin_std": 314.7109680175781, "step": 350 }, { "KL/chosen_KL_mean": -311.5458679199219, "KL/mean": -448.833984375, "KL/rejected_KL_mean": -586.1221313476562, "KL/std": 308.70892333984375, "epoch": 0.5154185022026432, "fcm_dpo/beta": 0.0015022414736449718, "fcm_dpo/delta": -0.0655267983675003, "fcm_dpo/margin": 274.5762634277344, "fcm_dpo/q_t": 0.4057248532772064, "grad_norm": 26.511186599731445, "learning_rate": 2.819952656376487e-07, "logits/chosen": -0.47476595640182495, "logits/rejected": -0.4755515456199646, "logps/chosen": -372.218017578125, "logps/ref_chosen": -60.6721305847168, "logps/ref_rejected": -101.5654296875, "logps/rejected": -687.6875610351562, "loss": 1.0834, "margin_dpo/margin_mean": 274.57623291015625, "margin_dpo/margin_std": 358.6259765625, "step": 351 }, { "KL/chosen_KL_mean": -366.9032287597656, "KL/mean": -447.29217529296875, "KL/rejected_KL_mean": -527.68115234375, "KL/std": 277.0421142578125, "epoch": 0.5168869309838473, "fcm_dpo/beta": 0.001509011024609208, "fcm_dpo/delta": 0.024786878377199173, "fcm_dpo/margin": 160.77786254882812, "fcm_dpo/q_t": 0.44290652871131897, "grad_norm": 30.402721405029297, "learning_rate": 2.8072207266617854e-07, "logits/chosen": -0.4430214464664459, "logits/rejected": -0.41058680415153503, "logps/chosen": -437.8466796875, "logps/ref_chosen": -70.9434585571289, "logps/ref_rejected": -76.6419677734375, "logps/rejected": -604.3231201171875, "loss": 1.2099, "margin_dpo/margin_mean": 160.77786254882812, "margin_dpo/margin_std": 302.5135498046875, "step": 352 }, { "KL/chosen_KL_mean": -346.7418212890625, "KL/mean": -460.20111083984375, "KL/rejected_KL_mean": -573.660400390625, "KL/std": 280.99822998046875, "epoch": 0.5183553597650514, "fcm_dpo/beta": 0.0015030490467324853, "fcm_dpo/delta": 0.008724931627511978, "fcm_dpo/margin": 226.9186248779297, "fcm_dpo/q_t": 0.42118746042251587, "grad_norm": 23.444734573364258, "learning_rate": 2.794480701395219e-07, "logits/chosen": -0.4665010869503021, "logits/rejected": -0.4548315405845642, "logps/chosen": -405.1371765136719, "logps/ref_chosen": -58.39533996582031, "logps/ref_rejected": -80.33553314208984, "logps/rejected": -653.9959716796875, "loss": 1.1327, "margin_dpo/margin_mean": 226.9186248779297, "margin_dpo/margin_std": 323.4075622558594, "step": 353 }, { "KL/chosen_KL_mean": -296.152099609375, "KL/mean": -417.95367431640625, "KL/rejected_KL_mean": -539.7552490234375, "KL/std": 264.796630859375, "epoch": 0.5198237885462555, "fcm_dpo/beta": 0.0015054075047373772, "fcm_dpo/delta": -0.01743352971971035, "fcm_dpo/margin": 243.60313415527344, "fcm_dpo/q_t": 0.4130624532699585, "grad_norm": 27.117420196533203, "learning_rate": 2.781732916288303e-07, "logits/chosen": -0.44472265243530273, "logits/rejected": -0.4349956214427948, "logps/chosen": -355.9551086425781, "logps/ref_chosen": -59.80299377441406, "logps/ref_rejected": -88.75750732421875, "logps/rejected": -628.5127563476562, "loss": 1.0915, "margin_dpo/margin_mean": 243.60311889648438, "margin_dpo/margin_std": 267.82965087890625, "step": 354 }, { "KL/chosen_KL_mean": -299.48431396484375, "KL/mean": -417.59002685546875, "KL/rejected_KL_mean": -535.6956787109375, "KL/std": 261.95001220703125, "epoch": 0.5212922173274597, "fcm_dpo/beta": 0.0015026122564449906, "fcm_dpo/delta": -0.005186091177165508, "fcm_dpo/margin": 236.21141052246094, "fcm_dpo/q_t": 0.4154089689254761, "grad_norm": 33.38981628417969, "learning_rate": 2.7689777072570284e-07, "logits/chosen": -0.5034153461456299, "logits/rejected": -0.4952772855758667, "logps/chosen": -353.61279296875, "logps/ref_chosen": -54.12849807739258, "logps/ref_rejected": -82.40606689453125, "logps/rejected": -618.101806640625, "loss": 1.1006, "margin_dpo/margin_mean": 236.21142578125, "margin_dpo/margin_std": 263.2491760253906, "step": 355 }, { "KL/chosen_KL_mean": -364.2958984375, "KL/mean": -427.6046142578125, "KL/rejected_KL_mean": -490.913330078125, "KL/std": 271.66748046875, "epoch": 0.5227606461086637, "fcm_dpo/beta": 0.0015103459591045976, "fcm_dpo/delta": 0.03656543791294098, "fcm_dpo/margin": 126.61747741699219, "fcm_dpo/q_t": 0.45585763454437256, "grad_norm": 26.874828338623047, "learning_rate": 2.7562154104130176e-07, "logits/chosen": -0.4800550043582916, "logits/rejected": -0.4602137804031372, "logps/chosen": -428.9696960449219, "logps/ref_chosen": -64.6738052368164, "logps/ref_rejected": -75.89926147460938, "logps/rejected": -566.8126220703125, "loss": 1.2663, "margin_dpo/margin_mean": 126.61747741699219, "margin_dpo/margin_std": 333.15924072265625, "step": 356 }, { "KL/chosen_KL_mean": -329.130859375, "KL/mean": -432.2911376953125, "KL/rejected_KL_mean": -535.451416015625, "KL/std": 263.6339111328125, "epoch": 0.5242290748898678, "fcm_dpo/beta": 0.0015188403194770217, "fcm_dpo/delta": 0.03768056631088257, "fcm_dpo/margin": 206.32057189941406, "fcm_dpo/q_t": 0.42600899934768677, "grad_norm": 33.71342468261719, "learning_rate": 2.7434463620546594e-07, "logits/chosen": -0.4496878385543823, "logits/rejected": -0.4382253885269165, "logps/chosen": -381.8566589355469, "logps/ref_chosen": -52.725799560546875, "logps/ref_rejected": -86.84115600585938, "logps/rejected": -622.2926025390625, "loss": 1.1443, "margin_dpo/margin_mean": 206.32058715820312, "margin_dpo/margin_std": 286.84808349609375, "step": 357 }, { "KL/chosen_KL_mean": -299.54290771484375, "KL/mean": -400.99774169921875, "KL/rejected_KL_mean": -502.4526062011719, "KL/std": 268.91558837890625, "epoch": 0.5256975036710719, "fcm_dpo/beta": 0.0015357965603470802, "fcm_dpo/delta": 0.03961861506104469, "fcm_dpo/margin": 202.90966796875, "fcm_dpo/q_t": 0.42794185876846313, "grad_norm": 21.35138702392578, "learning_rate": 2.730670898658255e-07, "logits/chosen": -0.5003876686096191, "logits/rejected": -0.484429270029068, "logps/chosen": -362.74835205078125, "logps/ref_chosen": -63.20543670654297, "logps/ref_rejected": -88.373291015625, "logps/rejected": -590.825927734375, "loss": 1.1493, "margin_dpo/margin_mean": 202.90966796875, "margin_dpo/margin_std": 300.5551452636719, "step": 358 }, { "KL/chosen_KL_mean": -351.01727294921875, "KL/mean": -474.3118591308594, "KL/rejected_KL_mean": -597.6065673828125, "KL/std": 270.4597473144531, "epoch": 0.527165932452276, "fcm_dpo/beta": 0.0015308901201933622, "fcm_dpo/delta": -0.028654001653194427, "fcm_dpo/margin": 246.58924865722656, "fcm_dpo/q_t": 0.41267070174217224, "grad_norm": 27.014921188354492, "learning_rate": 2.717889356869146e-07, "logits/chosen": -0.39821434020996094, "logits/rejected": -0.38216477632522583, "logps/chosen": -407.387451171875, "logps/ref_chosen": -56.370216369628906, "logps/ref_rejected": -82.17375183105469, "logps/rejected": -679.7802734375, "loss": 1.1054, "margin_dpo/margin_mean": 246.58926391601562, "margin_dpo/margin_std": 330.0404052734375, "step": 359 }, { "KL/chosen_KL_mean": -346.491943359375, "KL/mean": -435.66802978515625, "KL/rejected_KL_mean": -524.8441162109375, "KL/std": 230.1074676513672, "epoch": 0.5286343612334802, "fcm_dpo/beta": 0.0015494239050894976, "fcm_dpo/delta": 0.07566958665847778, "fcm_dpo/margin": 178.35223388671875, "fcm_dpo/q_t": 0.4340076446533203, "grad_norm": 44.442474365234375, "learning_rate": 2.7051020734928443e-07, "logits/chosen": -0.4111713767051697, "logits/rejected": -0.39715811610221863, "logps/chosen": -397.95233154296875, "logps/ref_chosen": -51.460384368896484, "logps/ref_rejected": -69.83892059326172, "logps/rejected": -594.6830444335938, "loss": 1.1574, "margin_dpo/margin_mean": 178.3522186279297, "margin_dpo/margin_std": 219.6265869140625, "step": 360 }, { "KL/chosen_KL_mean": -368.93402099609375, "KL/mean": -457.94183349609375, "KL/rejected_KL_mean": -546.9495849609375, "KL/std": 270.2475891113281, "epoch": 0.5301027900146843, "fcm_dpo/beta": 0.0015741336392238736, "fcm_dpo/delta": 0.07154025882482529, "fcm_dpo/margin": 178.01559448242188, "fcm_dpo/q_t": 0.43446093797683716, "grad_norm": 32.93223190307617, "learning_rate": 2.6923093854861593e-07, "logits/chosen": -0.4147510528564453, "logits/rejected": -0.4130870997905731, "logps/chosen": -422.80352783203125, "logps/ref_chosen": -53.86951446533203, "logps/ref_rejected": -90.7692642211914, "logps/rejected": -637.7188720703125, "loss": 1.176, "margin_dpo/margin_mean": 178.0155792236328, "margin_dpo/margin_std": 290.41571044921875, "step": 361 }, { "KL/chosen_KL_mean": -313.84429931640625, "KL/mean": -461.6890869140625, "KL/rejected_KL_mean": -609.533935546875, "KL/std": 282.345947265625, "epoch": 0.5315712187958884, "fcm_dpo/beta": 0.001544571714475751, "fcm_dpo/delta": -0.11250950396060944, "fcm_dpo/margin": 295.68963623046875, "fcm_dpo/q_t": 0.39361345767974854, "grad_norm": 26.073640823364258, "learning_rate": 2.679511629948319e-07, "logits/chosen": -0.44423967599868774, "logits/rejected": -0.4537912905216217, "logps/chosen": -372.48333740234375, "logps/ref_chosen": -58.639060974121094, "logps/ref_rejected": -105.58195495605469, "logps/rejected": -715.1158447265625, "loss": 1.0313, "margin_dpo/margin_mean": 295.68963623046875, "margin_dpo/margin_std": 301.7510986328125, "step": 362 }, { "KL/chosen_KL_mean": -284.67132568359375, "KL/mean": -432.5, "KL/rejected_KL_mean": -580.3286743164062, "KL/std": 269.4873962402344, "epoch": 0.5330396475770925, "fcm_dpo/beta": 0.0015216870233416557, "fcm_dpo/delta": -0.10447092354297638, "fcm_dpo/margin": 295.65728759765625, "fcm_dpo/q_t": 0.3945517838001251, "grad_norm": 30.12809944152832, "learning_rate": 2.6667091441120816e-07, "logits/chosen": -0.3808155059814453, "logits/rejected": -0.37547147274017334, "logps/chosen": -329.229736328125, "logps/ref_chosen": -44.558380126953125, "logps/ref_rejected": -74.69496154785156, "logps/rejected": -655.0236206054688, "loss": 1.0335, "margin_dpo/margin_mean": 295.65728759765625, "margin_dpo/margin_std": 295.49774169921875, "step": 363 }, { "KL/chosen_KL_mean": -308.2042236328125, "KL/mean": -420.90313720703125, "KL/rejected_KL_mean": -533.6021728515625, "KL/std": 261.20819091796875, "epoch": 0.5345080763582967, "fcm_dpo/beta": 0.0015164412325248122, "fcm_dpo/delta": 0.008228322491049767, "fcm_dpo/margin": 225.39791870117188, "fcm_dpo/q_t": 0.41953587532043457, "grad_norm": 19.665786743164062, "learning_rate": 2.6539022653348575e-07, "logits/chosen": -0.423043817281723, "logits/rejected": -0.4345157742500305, "logps/chosen": -357.09881591796875, "logps/ref_chosen": -48.894622802734375, "logps/ref_rejected": -91.395751953125, "logps/rejected": -624.9979248046875, "loss": 1.1234, "margin_dpo/margin_mean": 225.39791870117188, "margin_dpo/margin_std": 298.9825439453125, "step": 364 }, { "KL/chosen_KL_mean": -306.31304931640625, "KL/mean": -421.04754638671875, "KL/rejected_KL_mean": -535.781982421875, "KL/std": 272.7528076171875, "epoch": 0.5359765051395007, "fcm_dpo/beta": 0.001512871589511633, "fcm_dpo/delta": 0.002944107400253415, "fcm_dpo/margin": 229.468994140625, "fcm_dpo/q_t": 0.41882503032684326, "grad_norm": 22.943742752075195, "learning_rate": 2.641091331089811e-07, "logits/chosen": -0.4167822003364563, "logits/rejected": -0.4276372194290161, "logps/chosen": -357.8057861328125, "logps/ref_chosen": -51.49274444580078, "logps/ref_rejected": -92.70166778564453, "logps/rejected": -628.483642578125, "loss": 1.1113, "margin_dpo/margin_mean": 229.46900939941406, "margin_dpo/margin_std": 278.7212219238281, "step": 365 }, { "KL/chosen_KL_mean": -279.43597412109375, "KL/mean": -397.7274169921875, "KL/rejected_KL_mean": -516.0189208984375, "KL/std": 259.2245178222656, "epoch": 0.5374449339207048, "fcm_dpo/beta": 0.001505196327343583, "fcm_dpo/delta": -0.006696049124002457, "fcm_dpo/margin": 236.58290100097656, "fcm_dpo/q_t": 0.41680413484573364, "grad_norm": 20.372705459594727, "learning_rate": 2.6282766789569736e-07, "logits/chosen": -0.41744059324264526, "logits/rejected": -0.43434953689575195, "logps/chosen": -324.15655517578125, "logps/ref_chosen": -44.7205696105957, "logps/ref_rejected": -83.31040954589844, "logps/rejected": -599.3292846679688, "loss": 1.1141, "margin_dpo/margin_mean": 236.5828857421875, "margin_dpo/margin_std": 308.424072265625, "step": 366 }, { "KL/chosen_KL_mean": -296.7589111328125, "KL/mean": -394.7176818847656, "KL/rejected_KL_mean": -492.67645263671875, "KL/std": 231.99639892578125, "epoch": 0.5389133627019089, "fcm_dpo/beta": 0.0015253680758178234, "fcm_dpo/delta": 0.05257764458656311, "fcm_dpo/margin": 195.91758728027344, "fcm_dpo/q_t": 0.4285646378993988, "grad_norm": 20.739452362060547, "learning_rate": 2.615458646614349e-07, "logits/chosen": -0.4246237277984619, "logits/rejected": -0.40804946422576904, "logps/chosen": -355.164306640625, "logps/ref_chosen": -58.405418395996094, "logps/ref_rejected": -76.75132751464844, "logps/rejected": -569.4277954101562, "loss": 1.1444, "margin_dpo/margin_mean": 195.91757202148438, "margin_dpo/margin_std": 247.79733276367188, "step": 367 }, { "KL/chosen_KL_mean": -280.85968017578125, "KL/mean": -437.2155456542969, "KL/rejected_KL_mean": -593.5714111328125, "KL/std": 266.61279296875, "epoch": 0.540381791483113, "fcm_dpo/beta": 0.0014983892906457186, "fcm_dpo/delta": -0.12442989647388458, "fcm_dpo/margin": 312.7117919921875, "fcm_dpo/q_t": 0.38702040910720825, "grad_norm": 45.461280822753906, "learning_rate": 2.6026375718290083e-07, "logits/chosen": -0.4298338294029236, "logits/rejected": -0.43975430727005005, "logps/chosen": -325.31219482421875, "logps/ref_chosen": -44.452518463134766, "logps/ref_rejected": -98.55526733398438, "logps/rejected": -692.126708984375, "loss": 1.0009, "margin_dpo/margin_mean": 312.7117919921875, "margin_dpo/margin_std": 233.67335510253906, "step": 368 }, { "KL/chosen_KL_mean": -379.5709228515625, "KL/mean": -460.81072998046875, "KL/rejected_KL_mean": -542.0505981445312, "KL/std": 278.4993896484375, "epoch": 0.5418502202643172, "fcm_dpo/beta": 0.001493208110332489, "fcm_dpo/delta": 0.02635270729660988, "fcm_dpo/margin": 162.47970581054688, "fcm_dpo/q_t": 0.44229739904403687, "grad_norm": 22.148353576660156, "learning_rate": 2.589813792448196e-07, "logits/chosen": -0.4620640277862549, "logits/rejected": -0.4419480264186859, "logps/chosen": -450.9524230957031, "logps/ref_chosen": -71.38150024414062, "logps/ref_rejected": -91.29582214355469, "logps/rejected": -633.346435546875, "loss": 1.2114, "margin_dpo/margin_mean": 162.47970581054688, "margin_dpo/margin_std": 314.4382019042969, "step": 369 }, { "KL/chosen_KL_mean": -407.7558288574219, "KL/mean": -481.97650146484375, "KL/rejected_KL_mean": -556.1971435546875, "KL/std": 289.5532531738281, "epoch": 0.5433186490455213, "fcm_dpo/beta": 0.0015031839720904827, "fcm_dpo/delta": 0.04018635302782059, "fcm_dpo/margin": 148.44134521484375, "fcm_dpo/q_t": 0.4480086863040924, "grad_norm": 38.19746398925781, "learning_rate": 2.5769876463904263e-07, "logits/chosen": -0.476939857006073, "logits/rejected": -0.4694327414035797, "logps/chosen": -479.36334228515625, "logps/ref_chosen": -71.60749816894531, "logps/ref_rejected": -97.25978088378906, "logps/rejected": -653.4569702148438, "loss": 1.2328, "margin_dpo/margin_mean": 148.44134521484375, "margin_dpo/margin_std": 322.265869140625, "step": 370 }, { "KL/chosen_KL_mean": -384.25799560546875, "KL/mean": -499.07647705078125, "KL/rejected_KL_mean": -613.8948974609375, "KL/std": 296.7556457519531, "epoch": 0.5447870778267254, "fcm_dpo/beta": 0.0015145591460168362, "fcm_dpo/delta": 0.0020634010434150696, "fcm_dpo/margin": 229.63694763183594, "fcm_dpo/q_t": 0.41980546712875366, "grad_norm": 30.20676612854004, "learning_rate": 2.5641594716365744e-07, "logits/chosen": -0.4951871335506439, "logits/rejected": -0.4812811017036438, "logps/chosen": -453.6724853515625, "logps/ref_chosen": -69.41448974609375, "logps/ref_rejected": -99.17217254638672, "logps/rejected": -713.067138671875, "loss": 1.1315, "margin_dpo/margin_mean": 229.63694763183594, "margin_dpo/margin_std": 338.1903991699219, "step": 371 }, { "KL/chosen_KL_mean": -376.35491943359375, "KL/mean": -509.65484619140625, "KL/rejected_KL_mean": -642.9547119140625, "KL/std": 326.4896545410156, "epoch": 0.5462555066079295, "fcm_dpo/beta": 0.0014974682126194239, "fcm_dpo/delta": -0.05130906403064728, "fcm_dpo/margin": 266.59979248046875, "fcm_dpo/q_t": 0.4094997048377991, "grad_norm": 24.317825317382812, "learning_rate": 2.551329606220976e-07, "logits/chosen": -0.46680545806884766, "logits/rejected": -0.44958722591400146, "logps/chosen": -438.17291259765625, "logps/ref_chosen": -61.8179931640625, "logps/ref_rejected": -78.53948974609375, "logps/rejected": -721.4942626953125, "loss": 1.0968, "margin_dpo/margin_mean": 266.59979248046875, "margin_dpo/margin_std": 374.668212890625, "step": 372 }, { "KL/chosen_KL_mean": -410.6204833984375, "KL/mean": -533.13623046875, "KL/rejected_KL_mean": -655.6519775390625, "KL/std": 314.3919677734375, "epoch": 0.5477239353891337, "fcm_dpo/beta": 0.0014972374774515629, "fcm_dpo/delta": -0.017848707735538483, "fcm_dpo/margin": 245.0315399169922, "fcm_dpo/q_t": 0.4135189950466156, "grad_norm": 25.24411392211914, "learning_rate": 2.538498388222517e-07, "logits/chosen": -0.4661676287651062, "logits/rejected": -0.44771280884742737, "logps/chosen": -474.8376159667969, "logps/ref_chosen": -64.21713256835938, "logps/ref_rejected": -85.95960998535156, "logps/rejected": -741.6116333007812, "loss": 1.1063, "margin_dpo/margin_mean": 245.03155517578125, "margin_dpo/margin_std": 305.9546813964844, "step": 373 }, { "KL/chosen_KL_mean": -357.4342346191406, "KL/mean": -472.685302734375, "KL/rejected_KL_mean": -587.936279296875, "KL/std": 340.931396484375, "epoch": 0.5491923641703378, "fcm_dpo/beta": 0.001483630738221109, "fcm_dpo/delta": 0.007859412580728531, "fcm_dpo/margin": 230.50205993652344, "fcm_dpo/q_t": 0.4234431982040405, "grad_norm": 25.0133113861084, "learning_rate": 2.525666155755725e-07, "logits/chosen": -0.5473544597625732, "logits/rejected": -0.5300641059875488, "logps/chosen": -428.08441162109375, "logps/ref_chosen": -70.65018463134766, "logps/ref_rejected": -93.64016723632812, "logps/rejected": -681.5764770507812, "loss": 1.149, "margin_dpo/margin_mean": 230.50204467773438, "margin_dpo/margin_std": 383.9598388671875, "step": 374 }, { "KL/chosen_KL_mean": -355.9977722167969, "KL/mean": -470.6821594238281, "KL/rejected_KL_mean": -585.3665771484375, "KL/std": 272.306396484375, "epoch": 0.5506607929515418, "fcm_dpo/beta": 0.001486303168348968, "fcm_dpo/delta": 0.008981514722108841, "fcm_dpo/margin": 229.36874389648438, "fcm_dpo/q_t": 0.42030322551727295, "grad_norm": 28.873268127441406, "learning_rate": 2.512833246961859e-07, "logits/chosen": -0.5043726563453674, "logits/rejected": -0.502221941947937, "logps/chosen": -416.0780029296875, "logps/ref_chosen": -60.080223083496094, "logps/ref_rejected": -88.93830871582031, "logps/rejected": -674.3048095703125, "loss": 1.1302, "margin_dpo/margin_mean": 229.36875915527344, "margin_dpo/margin_std": 321.7081298828125, "step": 375 }, { "KL/chosen_KL_mean": -350.9241943359375, "KL/mean": -486.59490966796875, "KL/rejected_KL_mean": -622.265625, "KL/std": 287.79742431640625, "epoch": 0.5521292217327459, "fcm_dpo/beta": 0.0014795248862355947, "fcm_dpo/delta": -0.053773216903209686, "fcm_dpo/margin": 271.3414306640625, "fcm_dpo/q_t": 0.40609490871429443, "grad_norm": 24.37376594543457, "learning_rate": 2.5e-07, "logits/chosen": -0.4954999089241028, "logits/rejected": -0.4836694002151489, "logps/chosen": -413.58453369140625, "logps/ref_chosen": -62.660308837890625, "logps/ref_rejected": -105.52660369873047, "logps/rejected": -727.792236328125, "loss": 1.0827, "margin_dpo/margin_mean": 271.3414306640625, "margin_dpo/margin_std": 334.4213562011719, "step": 376 }, { "KL/chosen_KL_mean": -335.0908203125, "KL/mean": -465.4486389160156, "KL/rejected_KL_mean": -595.8064575195312, "KL/std": 300.1963195800781, "epoch": 0.55359765051395, "fcm_dpo/beta": 0.0014730466064065695, "fcm_dpo/delta": -0.03540883585810661, "fcm_dpo/margin": 260.71563720703125, "fcm_dpo/q_t": 0.40986379981040955, "grad_norm": 32.48672866821289, "learning_rate": 2.487166753038141e-07, "logits/chosen": -0.46442294120788574, "logits/rejected": -0.4637567102909088, "logps/chosen": -389.569580078125, "logps/ref_chosen": -54.478736877441406, "logps/ref_rejected": -98.70335388183594, "logps/rejected": -694.509765625, "loss": 1.0863, "margin_dpo/margin_mean": 260.71563720703125, "margin_dpo/margin_std": 304.24932861328125, "step": 377 }, { "KL/chosen_KL_mean": -324.253173828125, "KL/mean": -457.3460693359375, "KL/rejected_KL_mean": -590.43896484375, "KL/std": 271.17132568359375, "epoch": 0.5550660792951542, "fcm_dpo/beta": 0.0014589335769414902, "fcm_dpo/delta": -0.03989920765161514, "fcm_dpo/margin": 266.185791015625, "fcm_dpo/q_t": 0.4074614346027374, "grad_norm": 30.09598731994629, "learning_rate": 2.4743338442442754e-07, "logits/chosen": -0.44874757528305054, "logits/rejected": -0.4621288776397705, "logps/chosen": -369.273681640625, "logps/ref_chosen": -45.02053451538086, "logps/ref_rejected": -88.0469741821289, "logps/rejected": -678.4859619140625, "loss": 1.0774, "margin_dpo/margin_mean": 266.185791015625, "margin_dpo/margin_std": 289.00152587890625, "step": 378 }, { "KL/chosen_KL_mean": -363.91680908203125, "KL/mean": -496.1492919921875, "KL/rejected_KL_mean": -628.3817138671875, "KL/std": 281.95513916015625, "epoch": 0.5565345080763583, "fcm_dpo/beta": 0.001443482586182654, "fcm_dpo/delta": -0.03341230750083923, "fcm_dpo/margin": 264.4649658203125, "fcm_dpo/q_t": 0.41118955612182617, "grad_norm": 27.302921295166016, "learning_rate": 2.461501611777483e-07, "logits/chosen": -0.4446529150009155, "logits/rejected": -0.462520956993103, "logps/chosen": -417.09893798828125, "logps/ref_chosen": -53.182098388671875, "logps/ref_rejected": -114.3001708984375, "logps/rejected": -742.681884765625, "loss": 1.0973, "margin_dpo/margin_mean": 264.4649658203125, "margin_dpo/margin_std": 336.5331115722656, "step": 379 }, { "KL/chosen_KL_mean": -347.66009521484375, "KL/mean": -501.8836669921875, "KL/rejected_KL_mean": -656.1072998046875, "KL/std": 324.0223083496094, "epoch": 0.5580029368575624, "fcm_dpo/beta": 0.0014270804822444916, "fcm_dpo/delta": -0.0942816212773323, "fcm_dpo/margin": 308.44720458984375, "fcm_dpo/q_t": 0.3965052366256714, "grad_norm": 23.3795166015625, "learning_rate": 2.4486703937790243e-07, "logits/chosen": -0.4075812101364136, "logits/rejected": -0.4323120415210724, "logps/chosen": -399.01312255859375, "logps/ref_chosen": -51.3530387878418, "logps/ref_rejected": -104.19169616699219, "logps/rejected": -760.2989501953125, "loss": 1.0519, "margin_dpo/margin_mean": 308.44720458984375, "margin_dpo/margin_std": 349.7032470703125, "step": 380 }, { "KL/chosen_KL_mean": -357.15435791015625, "KL/mean": -460.5580749511719, "KL/rejected_KL_mean": -563.9618530273438, "KL/std": 264.73126220703125, "epoch": 0.5594713656387665, "fcm_dpo/beta": 0.0014290527906268835, "fcm_dpo/delta": 0.056157857179641724, "fcm_dpo/margin": 206.8074951171875, "fcm_dpo/q_t": 0.43153679370880127, "grad_norm": 26.060014724731445, "learning_rate": 2.435840528363426e-07, "logits/chosen": -0.4189026951789856, "logits/rejected": -0.3993133306503296, "logps/chosen": -414.9573974609375, "logps/ref_chosen": -57.80306625366211, "logps/ref_rejected": -79.21940612792969, "logps/rejected": -643.1812744140625, "loss": 1.1781, "margin_dpo/margin_mean": 206.8074951171875, "margin_dpo/margin_std": 366.61651611328125, "step": 381 }, { "KL/chosen_KL_mean": -346.9622497558594, "KL/mean": -476.4005126953125, "KL/rejected_KL_mean": -605.8387451171875, "KL/std": 250.68878173828125, "epoch": 0.5609397944199707, "fcm_dpo/beta": 0.0014304309152066708, "fcm_dpo/delta": -0.02109403908252716, "fcm_dpo/margin": 258.87652587890625, "fcm_dpo/q_t": 0.41199958324432373, "grad_norm": 26.346120834350586, "learning_rate": 2.4230123536095745e-07, "logits/chosen": -0.4333261251449585, "logits/rejected": -0.43255484104156494, "logps/chosen": -412.9825439453125, "logps/ref_chosen": -66.02030181884766, "logps/ref_rejected": -110.71016693115234, "logps/rejected": -716.5488891601562, "loss": 1.0869, "margin_dpo/margin_mean": 258.87652587890625, "margin_dpo/margin_std": 273.289306640625, "step": 382 }, { "KL/chosen_KL_mean": -361.38958740234375, "KL/mean": -491.83441162109375, "KL/rejected_KL_mean": -622.2792358398438, "KL/std": 291.9897766113281, "epoch": 0.5624082232011748, "fcm_dpo/beta": 0.0014227345818653703, "fcm_dpo/delta": -0.022058716043829918, "fcm_dpo/margin": 260.88958740234375, "fcm_dpo/q_t": 0.4132228195667267, "grad_norm": 29.732643127441406, "learning_rate": 2.4101862075518037e-07, "logits/chosen": -0.4310753345489502, "logits/rejected": -0.4417402744293213, "logps/chosen": -411.78106689453125, "logps/ref_chosen": -50.39148712158203, "logps/ref_rejected": -93.71589660644531, "logps/rejected": -715.9951171875, "loss": 1.1175, "margin_dpo/margin_mean": 260.8896179199219, "margin_dpo/margin_std": 378.602783203125, "step": 383 }, { "KL/chosen_KL_mean": -364.3046569824219, "KL/mean": -464.1553649902344, "KL/rejected_KL_mean": -564.006103515625, "KL/std": 258.0005187988281, "epoch": 0.5638766519823789, "fcm_dpo/beta": 0.0014413523022085428, "fcm_dpo/delta": 0.06357108056545258, "fcm_dpo/margin": 199.70144653320312, "fcm_dpo/q_t": 0.43101757764816284, "grad_norm": 23.503877639770508, "learning_rate": 2.397362428170992e-07, "logits/chosen": -0.46157699823379517, "logits/rejected": -0.452648401260376, "logps/chosen": -416.35076904296875, "logps/ref_chosen": -52.046104431152344, "logps/ref_rejected": -85.76089477539062, "logps/rejected": -649.7669677734375, "loss": 1.1487, "margin_dpo/margin_mean": 199.70143127441406, "margin_dpo/margin_std": 238.79812622070312, "step": 384 }, { "KL/chosen_KL_mean": -332.13482666015625, "KL/mean": -453.1705627441406, "KL/rejected_KL_mean": -574.2062377929688, "KL/std": 223.9749755859375, "epoch": 0.5653450807635829, "fcm_dpo/beta": 0.0014425960835069418, "fcm_dpo/delta": 0.000802924856543541, "fcm_dpo/margin": 242.0714569091797, "fcm_dpo/q_t": 0.4163161516189575, "grad_norm": 27.88534164428711, "learning_rate": 2.3845413533856514e-07, "logits/chosen": -0.5080987215042114, "logits/rejected": -0.48287850618362427, "logps/chosen": -397.6869812011719, "logps/ref_chosen": -65.55215454101562, "logps/ref_rejected": -77.82792663574219, "logps/rejected": -652.0341796875, "loss": 1.0991, "margin_dpo/margin_mean": 242.0714569091797, "margin_dpo/margin_std": 249.71163940429688, "step": 385 }, { "KL/chosen_KL_mean": -357.517578125, "KL/mean": -492.87701416015625, "KL/rejected_KL_mean": -628.2364501953125, "KL/std": 289.87762451171875, "epoch": 0.566813509544787, "fcm_dpo/beta": 0.001433402532711625, "fcm_dpo/delta": -0.03957575559616089, "fcm_dpo/margin": 270.7188720703125, "fcm_dpo/q_t": 0.4095271825790405, "grad_norm": 19.920806884765625, "learning_rate": 2.3717233210430254e-07, "logits/chosen": -0.4485896825790405, "logits/rejected": -0.44217997789382935, "logps/chosen": -415.7394104003906, "logps/ref_chosen": -58.22185516357422, "logps/ref_rejected": -92.32742309570312, "logps/rejected": -720.5638427734375, "loss": 1.0902, "margin_dpo/margin_mean": 270.7189025878906, "margin_dpo/margin_std": 338.1651306152344, "step": 386 }, { "KL/chosen_KL_mean": -396.12030029296875, "KL/mean": -507.7943115234375, "KL/rejected_KL_mean": -619.46826171875, "KL/std": 272.486328125, "epoch": 0.5682819383259912, "fcm_dpo/beta": 0.0014316510641947389, "fcm_dpo/delta": 0.031094016507267952, "fcm_dpo/margin": 223.3479461669922, "fcm_dpo/q_t": 0.42408478260040283, "grad_norm": 29.61167335510254, "learning_rate": 2.3589086689101889e-07, "logits/chosen": -0.5005359649658203, "logits/rejected": -0.4781208038330078, "logps/chosen": -462.5397644042969, "logps/ref_chosen": -66.41944885253906, "logps/ref_rejected": -92.16915893554688, "logps/rejected": -711.637451171875, "loss": 1.1323, "margin_dpo/margin_mean": 223.3479461669922, "margin_dpo/margin_std": 280.696044921875, "step": 387 }, { "KL/chosen_KL_mean": -375.528076171875, "KL/mean": -526.5117797851562, "KL/rejected_KL_mean": -677.4954833984375, "KL/std": 317.00433349609375, "epoch": 0.5697503671071953, "fcm_dpo/beta": 0.0014184208121150732, "fcm_dpo/delta": -0.08210138231515884, "fcm_dpo/margin": 301.9674072265625, "fcm_dpo/q_t": 0.40150347352027893, "grad_norm": 24.022911071777344, "learning_rate": 2.3460977346651428e-07, "logits/chosen": -0.41366052627563477, "logits/rejected": -0.4241475462913513, "logps/chosen": -425.65753173828125, "logps/ref_chosen": -50.129459381103516, "logps/ref_rejected": -104.43305969238281, "logps/rejected": -781.9285888671875, "loss": 1.0641, "margin_dpo/margin_mean": 301.9674072265625, "margin_dpo/margin_std": 364.0025634765625, "step": 388 }, { "KL/chosen_KL_mean": -425.42889404296875, "KL/mean": -547.1348876953125, "KL/rejected_KL_mean": -668.8409423828125, "KL/std": 310.88848876953125, "epoch": 0.5712187958883994, "fcm_dpo/beta": 0.0014143642038106918, "fcm_dpo/delta": 0.005878929980099201, "fcm_dpo/margin": 243.41204833984375, "fcm_dpo/q_t": 0.41971537470817566, "grad_norm": 36.894657135009766, "learning_rate": 2.3332908558879177e-07, "logits/chosen": -0.49010196328163147, "logits/rejected": -0.4835142195224762, "logps/chosen": -483.33544921875, "logps/ref_chosen": -57.906593322753906, "logps/ref_rejected": -77.91454315185547, "logps/rejected": -746.7554931640625, "loss": 1.1298, "margin_dpo/margin_mean": 243.41204833984375, "margin_dpo/margin_std": 350.9767761230469, "step": 389 }, { "KL/chosen_KL_mean": -416.3863220214844, "KL/mean": -545.6452026367188, "KL/rejected_KL_mean": -674.904052734375, "KL/std": 316.62554931640625, "epoch": 0.5726872246696035, "fcm_dpo/beta": 0.0014098630053922534, "fcm_dpo/delta": -0.01524612121284008, "fcm_dpo/margin": 258.5177307128906, "fcm_dpo/q_t": 0.4184497892856598, "grad_norm": 23.087417602539062, "learning_rate": 2.320488370051681e-07, "logits/chosen": -0.44321155548095703, "logits/rejected": -0.4361841678619385, "logps/chosen": -465.61224365234375, "logps/ref_chosen": -49.22591781616211, "logps/ref_rejected": -85.5281982421875, "logps/rejected": -760.4322509765625, "loss": 1.1294, "margin_dpo/margin_mean": 258.5177307128906, "margin_dpo/margin_std": 404.07928466796875, "step": 390 }, { "KL/chosen_KL_mean": -422.77203369140625, "KL/mean": -493.1278076171875, "KL/rejected_KL_mean": -563.4835205078125, "KL/std": 304.7773742675781, "epoch": 0.5741556534508077, "fcm_dpo/beta": 0.0014129449846222997, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 140.71145629882812, "fcm_dpo/q_t": 0.4527771472930908, "grad_norm": 47.28102111816406, "learning_rate": 2.3076906145138405e-07, "logits/chosen": -0.4732820391654968, "logits/rejected": -0.464574933052063, "logps/chosen": -487.1016845703125, "logps/ref_chosen": -64.32965087890625, "logps/ref_rejected": -86.73820495605469, "logps/rejected": -650.2216796875, "loss": 1.2471, "margin_dpo/margin_mean": 140.71145629882812, "margin_dpo/margin_std": 314.5218200683594, "step": 391 }, { "KL/chosen_KL_mean": -348.7242431640625, "KL/mean": -502.31378173828125, "KL/rejected_KL_mean": -655.9034423828125, "KL/std": 310.9810791015625, "epoch": 0.5756240822320118, "fcm_dpo/beta": 0.0014000183437019587, "fcm_dpo/delta": -0.08359719067811966, "fcm_dpo/margin": 307.1791076660156, "fcm_dpo/q_t": 0.39894986152648926, "grad_norm": 29.815181732177734, "learning_rate": 2.294897926507156e-07, "logits/chosen": -0.4261566400527954, "logits/rejected": -0.4199681282043457, "logps/chosen": -402.22821044921875, "logps/ref_chosen": -53.50397872924805, "logps/ref_rejected": -102.34584045410156, "logps/rejected": -758.249267578125, "loss": 1.0465, "margin_dpo/margin_mean": 307.17913818359375, "margin_dpo/margin_std": 307.54779052734375, "step": 392 }, { "KL/chosen_KL_mean": -347.76776123046875, "KL/mean": -462.5784912109375, "KL/rejected_KL_mean": -577.38916015625, "KL/std": 302.32501220703125, "epoch": 0.5770925110132159, "fcm_dpo/beta": 0.0013939584605395794, "fcm_dpo/delta": 0.03090941533446312, "fcm_dpo/margin": 229.621337890625, "fcm_dpo/q_t": 0.4278019666671753, "grad_norm": 20.847705841064453, "learning_rate": 2.2821106431308543e-07, "logits/chosen": -0.4470987319946289, "logits/rejected": -0.44747045636177063, "logps/chosen": -394.24169921875, "logps/ref_chosen": -46.473915100097656, "logps/ref_rejected": -71.96885681152344, "logps/rejected": -649.3580322265625, "loss": 1.1562, "margin_dpo/margin_mean": 229.62136840820312, "margin_dpo/margin_std": 383.7559509277344, "step": 393 }, { "KL/chosen_KL_mean": -405.95355224609375, "KL/mean": -526.41748046875, "KL/rejected_KL_mean": -646.88134765625, "KL/std": 321.18109130859375, "epoch": 0.57856093979442, "fcm_dpo/beta": 0.0013992037856951356, "fcm_dpo/delta": 0.01332851778715849, "fcm_dpo/margin": 240.9277801513672, "fcm_dpo/q_t": 0.4218076467514038, "grad_norm": 23.301321029663086, "learning_rate": 2.2693291013417452e-07, "logits/chosen": -0.4254276156425476, "logits/rejected": -0.42678093910217285, "logps/chosen": -458.8651123046875, "logps/ref_chosen": -52.91154861450195, "logps/ref_rejected": -90.8226318359375, "logps/rejected": -737.7039794921875, "loss": 1.1278, "margin_dpo/margin_mean": 240.9277801513672, "margin_dpo/margin_std": 331.58807373046875, "step": 394 }, { "KL/chosen_KL_mean": -393.3065185546875, "KL/mean": -526.68896484375, "KL/rejected_KL_mean": -660.071533203125, "KL/std": 314.0694274902344, "epoch": 0.580029368575624, "fcm_dpo/beta": 0.0013938779011368752, "fcm_dpo/delta": -0.02284281887114048, "fcm_dpo/margin": 266.7650146484375, "fcm_dpo/q_t": 0.41418999433517456, "grad_norm": 23.73855972290039, "learning_rate": 2.2565536379453404e-07, "logits/chosen": -0.49259886145591736, "logits/rejected": -0.4919501841068268, "logps/chosen": -455.85260009765625, "logps/ref_chosen": -62.546112060546875, "logps/ref_rejected": -83.78262329101562, "logps/rejected": -743.8541259765625, "loss": 1.1106, "margin_dpo/margin_mean": 266.7650146484375, "margin_dpo/margin_std": 368.64495849609375, "step": 395 }, { "KL/chosen_KL_mean": -391.6252136230469, "KL/mean": -510.94085693359375, "KL/rejected_KL_mean": -630.256591796875, "KL/std": 305.98333740234375, "epoch": 0.5814977973568282, "fcm_dpo/beta": 0.0013969738502055407, "fcm_dpo/delta": 0.017192337661981583, "fcm_dpo/margin": 238.63131713867188, "fcm_dpo/q_t": 0.42099273204803467, "grad_norm": 21.45150375366211, "learning_rate": 2.2437845895869825e-07, "logits/chosen": -0.5134637355804443, "logits/rejected": -0.4980122447013855, "logps/chosen": -460.62115478515625, "logps/ref_chosen": -68.99594116210938, "logps/ref_rejected": -88.64665985107422, "logps/rejected": -718.9031982421875, "loss": 1.1162, "margin_dpo/margin_mean": 238.6313018798828, "margin_dpo/margin_std": 278.1322021484375, "step": 396 }, { "KL/chosen_KL_mean": -366.0447082519531, "KL/mean": -518.56298828125, "KL/rejected_KL_mean": -671.0814208984375, "KL/std": 297.60498046875, "epoch": 0.5829662261380323, "fcm_dpo/beta": 0.0013800031738355756, "fcm_dpo/delta": -0.07457705587148666, "fcm_dpo/margin": 305.0367126464844, "fcm_dpo/q_t": 0.4003956913948059, "grad_norm": 33.268245697021484, "learning_rate": 2.2310222927429716e-07, "logits/chosen": -0.45172950625419617, "logits/rejected": -0.45972180366516113, "logps/chosen": -427.3218688964844, "logps/ref_chosen": -61.27716827392578, "logps/ref_rejected": -103.11612701416016, "logps/rejected": -774.197509765625, "loss": 1.0549, "margin_dpo/margin_mean": 305.03668212890625, "margin_dpo/margin_std": 317.1044616699219, "step": 397 }, { "KL/chosen_KL_mean": -392.1479797363281, "KL/mean": -530.8260498046875, "KL/rejected_KL_mean": -669.5042114257812, "KL/std": 315.48931884765625, "epoch": 0.5844346549192364, "fcm_dpo/beta": 0.0013728467747569084, "fcm_dpo/delta": -0.03198913484811783, "fcm_dpo/margin": 277.356201171875, "fcm_dpo/q_t": 0.41205263137817383, "grad_norm": 21.440746307373047, "learning_rate": 2.2182670837116972e-07, "logits/chosen": -0.5343962907791138, "logits/rejected": -0.5324475765228271, "logps/chosen": -460.2995300292969, "logps/ref_chosen": -68.15155029296875, "logps/ref_rejected": -108.52360534667969, "logps/rejected": -778.02783203125, "loss": 1.0983, "margin_dpo/margin_mean": 277.3562316894531, "margin_dpo/margin_std": 362.9688720703125, "step": 398 }, { "KL/chosen_KL_mean": -341.6549377441406, "KL/mean": -460.3332214355469, "KL/rejected_KL_mean": -579.0115356445312, "KL/std": 272.2782287597656, "epoch": 0.5859030837004405, "fcm_dpo/beta": 0.0013724948512390256, "fcm_dpo/delta": 0.024983234703540802, "fcm_dpo/margin": 237.35659790039062, "fcm_dpo/q_t": 0.4236915111541748, "grad_norm": 24.46724510192871, "learning_rate": 2.2055192986047804e-07, "logits/chosen": -0.4898013174533844, "logits/rejected": -0.45694947242736816, "logps/chosen": -402.54473876953125, "logps/ref_chosen": -60.889801025390625, "logps/ref_rejected": -77.965576171875, "logps/rejected": -656.9771118164062, "loss": 1.1392, "margin_dpo/margin_mean": 237.35659790039062, "margin_dpo/margin_std": 339.81365966796875, "step": 399 }, { "KL/chosen_KL_mean": -335.7412109375, "KL/mean": -509.16314697265625, "KL/rejected_KL_mean": -682.585205078125, "KL/std": 286.89813232421875, "epoch": 0.5873715124816447, "fcm_dpo/beta": 0.001349159749224782, "fcm_dpo/delta": -0.12415604293346405, "fcm_dpo/margin": 346.8439636230469, "fcm_dpo/q_t": 0.3882994055747986, "grad_norm": 20.42949104309082, "learning_rate": 2.192779273338215e-07, "logits/chosen": -0.5137765407562256, "logits/rejected": -0.5164045691490173, "logps/chosen": -399.3847961425781, "logps/ref_chosen": -63.64359664916992, "logps/ref_rejected": -105.252685546875, "logps/rejected": -787.837890625, "loss": 1.0159, "margin_dpo/margin_mean": 346.8439636230469, "margin_dpo/margin_std": 318.0387268066406, "step": 400 }, { "KL/chosen_KL_mean": -375.37005615234375, "KL/mean": -471.6986083984375, "KL/rejected_KL_mean": -568.027099609375, "KL/std": 302.75244140625, "epoch": 0.5888399412628488, "fcm_dpo/beta": 0.0013488534605130553, "fcm_dpo/delta": 0.018642518669366837, "fcm_dpo/margin": 192.65704345703125, "fcm_dpo/q_t": 0.44042983651161194, "grad_norm": 21.7044677734375, "learning_rate": 2.1800473436235136e-07, "logits/chosen": -0.4904178977012634, "logits/rejected": -0.48238423466682434, "logps/chosen": -432.5331115722656, "logps/ref_chosen": -57.16303253173828, "logps/ref_rejected": -83.79249572753906, "logps/rejected": -651.819580078125, "loss": 1.2158, "margin_dpo/margin_mean": 192.65704345703125, "margin_dpo/margin_std": 408.41400146484375, "step": 401 }, { "KL/chosen_KL_mean": -283.151611328125, "KL/mean": -461.8321228027344, "KL/rejected_KL_mean": -640.5125732421875, "KL/std": 318.71820068359375, "epoch": 0.5903083700440529, "fcm_dpo/beta": 0.0013233155477792025, "fcm_dpo/delta": -0.12899458408355713, "fcm_dpo/margin": 357.3609924316406, "fcm_dpo/q_t": 0.387925922870636, "grad_norm": 26.780061721801758, "learning_rate": 2.1673238449588665e-07, "logits/chosen": -0.5302881002426147, "logits/rejected": -0.5248584151268005, "logps/chosen": -333.8919982910156, "logps/ref_chosen": -50.74037170410156, "logps/ref_rejected": -81.0460433959961, "logps/rejected": -721.5586547851562, "loss": 1.009, "margin_dpo/margin_mean": 357.3609924316406, "margin_dpo/margin_std": 318.2284851074219, "step": 402 }, { "KL/chosen_KL_mean": -328.6024169921875, "KL/mean": -467.64849853515625, "KL/rejected_KL_mean": -606.694580078125, "KL/std": 301.4685974121094, "epoch": 0.591776798825257, "fcm_dpo/beta": 0.001311097526922822, "fcm_dpo/delta": -0.015165827237069607, "fcm_dpo/margin": 278.0921325683594, "fcm_dpo/q_t": 0.4141251742839813, "grad_norm": 29.468088150024414, "learning_rate": 2.154609112620295e-07, "logits/chosen": -0.525422990322113, "logits/rejected": -0.5270965099334717, "logps/chosen": -375.749755859375, "logps/ref_chosen": -47.14731216430664, "logps/ref_rejected": -77.2666015625, "logps/rejected": -683.961181640625, "loss": 1.0966, "margin_dpo/margin_mean": 278.0921325683594, "margin_dpo/margin_std": 317.0299987792969, "step": 403 }, { "KL/chosen_KL_mean": -375.371337890625, "KL/mean": -508.93017578125, "KL/rejected_KL_mean": -642.4889526367188, "KL/std": 296.78961181640625, "epoch": 0.593245227606461, "fcm_dpo/beta": 0.0013087866827845573, "fcm_dpo/delta": 0.00037629250437021255, "fcm_dpo/margin": 267.11761474609375, "fcm_dpo/q_t": 0.41891252994537354, "grad_norm": 28.834339141845703, "learning_rate": 2.1419034816528218e-07, "logits/chosen": -0.5048468708992004, "logits/rejected": -0.4995875358581543, "logps/chosen": -423.24664306640625, "logps/ref_chosen": -47.875274658203125, "logps/ref_rejected": -77.15499877929688, "logps/rejected": -719.6439208984375, "loss": 1.1319, "margin_dpo/margin_mean": 267.11761474609375, "margin_dpo/margin_std": 398.306884765625, "step": 404 }, { "KL/chosen_KL_mean": -412.86749267578125, "KL/mean": -529.7939453125, "KL/rejected_KL_mean": -646.7203979492188, "KL/std": 337.2768249511719, "epoch": 0.5947136563876652, "fcm_dpo/beta": 0.001304717967286706, "fcm_dpo/delta": -0.04521414637565613, "fcm_dpo/margin": 233.85287475585938, "fcm_dpo/q_t": 0.4310234785079956, "grad_norm": 30.699373245239258, "learning_rate": 2.129207286861638e-07, "logits/chosen": -0.5107825398445129, "logits/rejected": -0.5024291276931763, "logps/chosen": -478.0303955078125, "logps/ref_chosen": -65.16290283203125, "logps/ref_rejected": -87.18678283691406, "logps/rejected": -733.9072265625, "loss": 1.1835, "margin_dpo/margin_mean": 233.85287475585938, "margin_dpo/margin_std": 432.97540283203125, "step": 405 }, { "KL/chosen_KL_mean": -389.0184020996094, "KL/mean": -537.796142578125, "KL/rejected_KL_mean": -686.5738525390625, "KL/std": 333.91748046875, "epoch": 0.5961820851688693, "fcm_dpo/beta": 0.0012974301353096962, "fcm_dpo/delta": -0.03775372728705406, "fcm_dpo/margin": 297.55548095703125, "fcm_dpo/q_t": 0.4103890061378479, "grad_norm": 29.44150733947754, "learning_rate": 2.1165208628032861e-07, "logits/chosen": -0.5451512932777405, "logits/rejected": -0.5550295114517212, "logps/chosen": -438.75921630859375, "logps/ref_chosen": -49.740814208984375, "logps/ref_rejected": -92.07862854003906, "logps/rejected": -778.6525268554688, "loss": 1.0925, "margin_dpo/margin_mean": 297.55548095703125, "margin_dpo/margin_std": 369.20068359375, "step": 406 }, { "KL/chosen_KL_mean": -414.57012939453125, "KL/mean": -509.9847106933594, "KL/rejected_KL_mean": -605.3992919921875, "KL/std": 269.21441650390625, "epoch": 0.5976505139500734, "fcm_dpo/beta": 0.001289546606130898, "fcm_dpo/delta": 0.003860359312966466, "fcm_dpo/margin": 190.8292236328125, "fcm_dpo/q_t": 0.4413040280342102, "grad_norm": 29.11326789855957, "learning_rate": 2.1038445437768375e-07, "logits/chosen": -0.5535376071929932, "logits/rejected": -0.523551344871521, "logps/chosen": -470.90081787109375, "logps/ref_chosen": -56.33069610595703, "logps/ref_rejected": -77.51209259033203, "logps/rejected": -682.911376953125, "loss": 1.2067, "margin_dpo/margin_mean": 190.82920837402344, "margin_dpo/margin_std": 349.5740966796875, "step": 407 }, { "KL/chosen_KL_mean": -411.55230712890625, "KL/mean": -515.71240234375, "KL/rejected_KL_mean": -619.8724365234375, "KL/std": 258.9462585449219, "epoch": 0.5991189427312775, "fcm_dpo/beta": 0.0013068197295069695, "fcm_dpo/delta": 0.07983996719121933, "fcm_dpo/margin": 208.32012939453125, "fcm_dpo/q_t": 0.43545645475387573, "grad_norm": 23.367450714111328, "learning_rate": 2.0911786638150872e-07, "logits/chosen": -0.5809944272041321, "logits/rejected": -0.5597822666168213, "logps/chosen": -481.34161376953125, "logps/ref_chosen": -69.789306640625, "logps/ref_rejected": -90.09693908691406, "logps/rejected": -709.9693603515625, "loss": 1.1688, "margin_dpo/margin_mean": 208.32012939453125, "margin_dpo/margin_std": 296.0139465332031, "step": 408 }, { "KL/chosen_KL_mean": -399.9471435546875, "KL/mean": -502.8631286621094, "KL/rejected_KL_mean": -605.7791137695312, "KL/std": 295.7550048828125, "epoch": 0.6005873715124816, "fcm_dpo/beta": 0.0013286788016557693, "fcm_dpo/delta": 0.07845626771450043, "fcm_dpo/margin": 205.8319549560547, "fcm_dpo/q_t": 0.4356931746006012, "grad_norm": 44.062828063964844, "learning_rate": 2.0785235566757517e-07, "logits/chosen": -0.5095489025115967, "logits/rejected": -0.4916442632675171, "logps/chosen": -467.26458740234375, "logps/ref_chosen": -67.31744384765625, "logps/ref_rejected": -84.904296875, "logps/rejected": -690.6834106445312, "loss": 1.1764, "margin_dpo/margin_mean": 205.8319549560547, "margin_dpo/margin_std": 322.6492919921875, "step": 409 }, { "KL/chosen_KL_mean": -363.33453369140625, "KL/mean": -491.4117431640625, "KL/rejected_KL_mean": -619.4889526367188, "KL/std": 283.58184814453125, "epoch": 0.6020558002936858, "fcm_dpo/beta": 0.00133224087767303, "fcm_dpo/delta": 0.009043154306709766, "fcm_dpo/margin": 256.1543884277344, "fcm_dpo/q_t": 0.4186854362487793, "grad_norm": 23.003204345703125, "learning_rate": 2.065879555832674e-07, "logits/chosen": -0.5263328552246094, "logits/rejected": -0.5296115875244141, "logps/chosen": -414.7998962402344, "logps/ref_chosen": -51.465354919433594, "logps/ref_rejected": -83.198974609375, "logps/rejected": -702.6879272460938, "loss": 1.1177, "margin_dpo/margin_mean": 256.1543884277344, "margin_dpo/margin_std": 320.51104736328125, "step": 410 }, { "KL/chosen_KL_mean": -383.42071533203125, "KL/mean": -508.7550354003906, "KL/rejected_KL_mean": -634.08935546875, "KL/std": 312.1067810058594, "epoch": 0.6035242290748899, "fcm_dpo/beta": 0.0013171627651900053, "fcm_dpo/delta": -0.06332255154848099, "fcm_dpo/margin": 250.6686553955078, "fcm_dpo/q_t": 0.42369675636291504, "grad_norm": 32.008201599121094, "learning_rate": 2.0532469944670343e-07, "logits/chosen": -0.5449320077896118, "logits/rejected": -0.5536935925483704, "logps/chosen": -435.72796630859375, "logps/ref_chosen": -52.30727005004883, "logps/ref_rejected": -80.69495391845703, "logps/rejected": -714.7843017578125, "loss": 1.142, "margin_dpo/margin_mean": 250.6686553955078, "margin_dpo/margin_std": 367.11444091796875, "step": 411 }, { "KL/chosen_KL_mean": -392.28662109375, "KL/mean": -521.355712890625, "KL/rejected_KL_mean": -650.4248046875, "KL/std": 303.547119140625, "epoch": 0.604992657856094, "fcm_dpo/beta": 0.00132085126824677, "fcm_dpo/delta": 0.009303496219217777, "fcm_dpo/margin": 258.1380920410156, "fcm_dpo/q_t": 0.42002564668655396, "grad_norm": 34.010955810546875, "learning_rate": 2.0406262054585738e-07, "logits/chosen": -0.6012539863586426, "logits/rejected": -0.6298267841339111, "logps/chosen": -445.4307861328125, "logps/ref_chosen": -53.144126892089844, "logps/ref_rejected": -100.0608139038086, "logps/rejected": -750.485595703125, "loss": 1.1274, "margin_dpo/margin_mean": 258.1380920410156, "margin_dpo/margin_std": 356.703857421875, "step": 412 }, { "KL/chosen_KL_mean": -416.2436218261719, "KL/mean": -545.9385986328125, "KL/rejected_KL_mean": -675.6336059570312, "KL/std": 308.3495178222656, "epoch": 0.6064610866372981, "fcm_dpo/beta": 0.001323135569691658, "fcm_dpo/delta": 0.006968270055949688, "fcm_dpo/margin": 259.3899841308594, "fcm_dpo/q_t": 0.41812664270401, "grad_norm": 36.40060806274414, "learning_rate": 2.0280175213768205e-07, "logits/chosen": -0.4915761649608612, "logits/rejected": -0.49571770429611206, "logps/chosen": -477.8255615234375, "logps/ref_chosen": -61.58196258544922, "logps/ref_rejected": -99.47340393066406, "logps/rejected": -775.1070556640625, "loss": 1.12, "margin_dpo/margin_mean": 259.3900146484375, "margin_dpo/margin_std": 336.85467529296875, "step": 413 }, { "KL/chosen_KL_mean": -375.42376708984375, "KL/mean": -514.7646484375, "KL/rejected_KL_mean": -654.1056518554688, "KL/std": 286.4696960449219, "epoch": 0.6079295154185022, "fcm_dpo/beta": 0.0013254533987492323, "fcm_dpo/delta": -0.02075226604938507, "fcm_dpo/margin": 278.681884765625, "fcm_dpo/q_t": 0.41263848543167114, "grad_norm": 27.233083724975586, "learning_rate": 2.0154212744723247e-07, "logits/chosen": -0.4962560534477234, "logits/rejected": -0.4908558428287506, "logps/chosen": -422.05523681640625, "logps/ref_chosen": -46.63148498535156, "logps/ref_rejected": -87.64653015136719, "logps/rejected": -741.752197265625, "loss": 1.0978, "margin_dpo/margin_mean": 278.681884765625, "margin_dpo/margin_std": 322.50115966796875, "step": 414 }, { "KL/chosen_KL_mean": -410.77630615234375, "KL/mean": -507.6982421875, "KL/rejected_KL_mean": -604.6201782226562, "KL/std": 278.3897399902344, "epoch": 0.6093979441997063, "fcm_dpo/beta": 0.0013193815248087049, "fcm_dpo/delta": 0.012895876541733742, "fcm_dpo/margin": 193.84385681152344, "fcm_dpo/q_t": 0.43898284435272217, "grad_norm": 24.22504997253418, "learning_rate": 2.002837796667909e-07, "logits/chosen": -0.5504616498947144, "logits/rejected": -0.5457872152328491, "logps/chosen": -489.39459228515625, "logps/ref_chosen": -78.6182861328125, "logps/ref_rejected": -100.47752380371094, "logps/rejected": -705.09765625, "loss": 1.1892, "margin_dpo/margin_mean": 193.84385681152344, "margin_dpo/margin_std": 313.9031982421875, "step": 415 }, { "KL/chosen_KL_mean": -355.70068359375, "KL/mean": -515.9644165039062, "KL/rejected_KL_mean": -676.2281494140625, "KL/std": 308.6180419921875, "epoch": 0.6108663729809104, "fcm_dpo/beta": 0.0013083845842629671, "fcm_dpo/delta": -0.07235361635684967, "fcm_dpo/margin": 320.5274658203125, "fcm_dpo/q_t": 0.4001033306121826, "grad_norm": 31.607568740844727, "learning_rate": 1.990267419549914e-07, "logits/chosen": -0.5469499230384827, "logits/rejected": -0.5526891946792603, "logps/chosen": -413.97979736328125, "logps/ref_chosen": -58.27912521362305, "logps/ref_rejected": -90.56871795654297, "logps/rejected": -766.796875, "loss": 1.0473, "margin_dpo/margin_mean": 320.5274658203125, "margin_dpo/margin_std": 299.2843322753906, "step": 416 }, { "KL/chosen_KL_mean": -364.55999755859375, "KL/mean": -502.7241516113281, "KL/rejected_KL_mean": -640.8883056640625, "KL/std": 286.73431396484375, "epoch": 0.6123348017621145, "fcm_dpo/beta": 0.0013004662469029427, "fcm_dpo/delta": -0.009734570048749447, "fcm_dpo/margin": 276.32830810546875, "fcm_dpo/q_t": 0.4140380620956421, "grad_norm": 24.474712371826172, "learning_rate": 1.9777104743594686e-07, "logits/chosen": -0.5460329651832581, "logits/rejected": -0.5258671045303345, "logps/chosen": -414.75872802734375, "logps/ref_chosen": -50.1987190246582, "logps/ref_rejected": -68.15184020996094, "logps/rejected": -709.0401611328125, "loss": 1.0922, "margin_dpo/margin_mean": 276.32830810546875, "margin_dpo/margin_std": 285.49676513671875, "step": 417 }, { "KL/chosen_KL_mean": -389.93212890625, "KL/mean": -532.0841674804688, "KL/rejected_KL_mean": -674.2362060546875, "KL/std": 330.60418701171875, "epoch": 0.6138032305433186, "fcm_dpo/beta": 0.001303232740610838, "fcm_dpo/delta": -0.02225048840045929, "fcm_dpo/margin": 284.3041076660156, "fcm_dpo/q_t": 0.41474616527557373, "grad_norm": 26.265411376953125, "learning_rate": 1.965167291983757e-07, "logits/chosen": -0.6291038990020752, "logits/rejected": -0.6125690937042236, "logps/chosen": -471.91058349609375, "logps/ref_chosen": -81.97846984863281, "logps/ref_rejected": -104.69148254394531, "logps/rejected": -778.9276733398438, "loss": 1.1142, "margin_dpo/margin_mean": 284.3041076660156, "margin_dpo/margin_std": 386.27587890625, "step": 418 }, { "KL/chosen_KL_mean": -362.5212097167969, "KL/mean": -517.69580078125, "KL/rejected_KL_mean": -672.8703002929688, "KL/std": 289.50018310546875, "epoch": 0.6152716593245228, "fcm_dpo/beta": 0.001282334327697754, "fcm_dpo/delta": -0.049968693405389786, "fcm_dpo/margin": 310.3490905761719, "fcm_dpo/q_t": 0.40543508529663086, "grad_norm": 22.777507781982422, "learning_rate": 1.9526382029472988e-07, "logits/chosen": -0.5401506423950195, "logits/rejected": -0.5378715395927429, "logps/chosen": -415.4698486328125, "logps/ref_chosen": -52.948646545410156, "logps/ref_rejected": -91.58309936523438, "logps/rejected": -764.453369140625, "loss": 1.0724, "margin_dpo/margin_mean": 310.34912109375, "margin_dpo/margin_std": 338.61309814453125, "step": 419 }, { "KL/chosen_KL_mean": -469.8270263671875, "KL/mean": -561.0233154296875, "KL/rejected_KL_mean": -652.219482421875, "KL/std": 309.34521484375, "epoch": 0.6167400881057269, "fcm_dpo/beta": 0.0013024189975112677, "fcm_dpo/delta": 0.11509222537279129, "fcm_dpo/margin": 182.3925018310547, "fcm_dpo/q_t": 0.4456251859664917, "grad_norm": 52.509613037109375, "learning_rate": 1.9401235374032425e-07, "logits/chosen": -0.5761592388153076, "logits/rejected": -0.542568564414978, "logps/chosen": -547.596923828125, "logps/ref_chosen": -77.7699203491211, "logps/ref_rejected": -69.31985473632812, "logps/rejected": -721.5394287109375, "loss": 1.2333, "margin_dpo/margin_mean": 182.3925018310547, "margin_dpo/margin_std": 416.1588134765625, "step": 420 }, { "KL/chosen_KL_mean": -383.7919921875, "KL/mean": -492.5055236816406, "KL/rejected_KL_mean": -601.218994140625, "KL/std": 306.576416015625, "epoch": 0.618208516886931, "fcm_dpo/beta": 0.001326603116467595, "fcm_dpo/delta": 0.06283024698495865, "fcm_dpo/margin": 217.42698669433594, "fcm_dpo/q_t": 0.43084651231765747, "grad_norm": 25.688739776611328, "learning_rate": 1.9276236251246653e-07, "logits/chosen": -0.5750212669372559, "logits/rejected": -0.5611597299575806, "logps/chosen": -437.557861328125, "logps/ref_chosen": -53.765865325927734, "logps/ref_rejected": -89.28144836425781, "logps/rejected": -690.5004272460938, "loss": 1.1646, "margin_dpo/margin_mean": 217.42698669433594, "margin_dpo/margin_std": 318.6001892089844, "step": 421 }, { "KL/chosen_KL_mean": -440.12738037109375, "KL/mean": -564.8936767578125, "KL/rejected_KL_mean": -689.659912109375, "KL/std": 308.4464111328125, "epoch": 0.6196769456681351, "fcm_dpo/beta": 0.0013281775172799826, "fcm_dpo/delta": 0.019213195890188217, "fcm_dpo/margin": 249.53256225585938, "fcm_dpo/q_t": 0.4216609597206116, "grad_norm": 33.2342414855957, "learning_rate": 1.9151387954958792e-07, "logits/chosen": -0.5859231352806091, "logits/rejected": -0.5860297083854675, "logps/chosen": -508.7611389160156, "logps/ref_chosen": -68.6337661743164, "logps/ref_rejected": -87.86351013183594, "logps/rejected": -777.5234375, "loss": 1.1347, "margin_dpo/margin_mean": 249.53257751464844, "margin_dpo/margin_std": 353.23614501953125, "step": 422 }, { "KL/chosen_KL_mean": -419.96142578125, "KL/mean": -560.3416748046875, "KL/rejected_KL_mean": -700.721923828125, "KL/std": 307.92205810546875, "epoch": 0.6211453744493393, "fcm_dpo/beta": 0.001325579360127449, "fcm_dpo/delta": -0.023041730746626854, "fcm_dpo/margin": 280.76055908203125, "fcm_dpo/q_t": 0.412263423204422, "grad_norm": 29.936748504638672, "learning_rate": 1.902669377503756e-07, "logits/chosen": -0.5792564153671265, "logits/rejected": -0.5877513885498047, "logps/chosen": -474.95172119140625, "logps/ref_chosen": -54.99030303955078, "logps/ref_rejected": -86.30654907226562, "logps/rejected": -787.028564453125, "loss": 1.0937, "margin_dpo/margin_mean": 280.7605895996094, "margin_dpo/margin_std": 330.3087158203125, "step": 423 }, { "KL/chosen_KL_mean": -394.10980224609375, "KL/mean": -522.6301879882812, "KL/rejected_KL_mean": -651.1505126953125, "KL/std": 303.40478515625, "epoch": 0.6226138032305433, "fcm_dpo/beta": 0.0013219558168202639, "fcm_dpo/delta": 0.010341478511691093, "fcm_dpo/margin": 257.0407409667969, "fcm_dpo/q_t": 0.42126062512397766, "grad_norm": 36.76081085205078, "learning_rate": 1.890215699729057e-07, "logits/chosen": -0.5852949619293213, "logits/rejected": -0.5621461868286133, "logps/chosen": -450.1217041015625, "logps/ref_chosen": -56.01192092895508, "logps/ref_rejected": -66.47896575927734, "logps/rejected": -717.6295166015625, "loss": 1.1318, "margin_dpo/margin_mean": 257.040771484375, "margin_dpo/margin_std": 368.2303161621094, "step": 424 }, { "KL/chosen_KL_mean": -452.0614013671875, "KL/mean": -553.6272583007812, "KL/rejected_KL_mean": -655.193115234375, "KL/std": 295.1512145996094, "epoch": 0.6240822320117474, "fcm_dpo/beta": 0.0013455725274980068, "fcm_dpo/delta": 0.07860462367534637, "fcm_dpo/margin": 203.1317138671875, "fcm_dpo/q_t": 0.4350726306438446, "grad_norm": 33.804561614990234, "learning_rate": 1.8777780903377732e-07, "logits/chosen": -0.5849053859710693, "logits/rejected": -0.5874682664871216, "logps/chosen": -498.9303894042969, "logps/ref_chosen": -46.86899948120117, "logps/ref_rejected": -95.92545318603516, "logps/rejected": -751.1185302734375, "loss": 1.1895, "margin_dpo/margin_mean": 203.1317138671875, "margin_dpo/margin_std": 363.55322265625, "step": 425 }, { "KL/chosen_KL_mean": -423.40118408203125, "KL/mean": -549.4375, "KL/rejected_KL_mean": -675.473876953125, "KL/std": 297.34173583984375, "epoch": 0.6255506607929515, "fcm_dpo/beta": 0.0013556771446019411, "fcm_dpo/delta": 0.008176274597644806, "fcm_dpo/margin": 252.0726776123047, "fcm_dpo/q_t": 0.41926294565200806, "grad_norm": 27.19607162475586, "learning_rate": 1.8653568770724803e-07, "logits/chosen": -0.6059207916259766, "logits/rejected": -0.5767127871513367, "logps/chosen": -499.9847412109375, "logps/ref_chosen": -76.58354187011719, "logps/ref_rejected": -81.26658630371094, "logps/rejected": -756.740478515625, "loss": 1.1239, "margin_dpo/margin_mean": 252.0726776123047, "margin_dpo/margin_std": 330.4275207519531, "step": 426 }, { "KL/chosen_KL_mean": -393.76983642578125, "KL/mean": -488.9925537109375, "KL/rejected_KL_mean": -584.21533203125, "KL/std": 265.6219482421875, "epoch": 0.6270190895741556, "fcm_dpo/beta": 0.001356072025373578, "fcm_dpo/delta": 0.02023770846426487, "fcm_dpo/margin": 190.4454803466797, "fcm_dpo/q_t": 0.4393557906150818, "grad_norm": 28.08988380432129, "learning_rate": 1.8529523872436977e-07, "logits/chosen": -0.6063523292541504, "logits/rejected": -0.5898270010948181, "logps/chosen": -458.62371826171875, "logps/ref_chosen": -64.8538818359375, "logps/ref_rejected": -78.5660171508789, "logps/rejected": -662.7813110351562, "loss": 1.1898, "margin_dpo/margin_mean": 190.44546508789062, "margin_dpo/margin_std": 315.81085205078125, "step": 427 }, { "KL/chosen_KL_mean": -493.20880126953125, "KL/mean": -632.4622802734375, "KL/rejected_KL_mean": -771.7158203125, "KL/std": 357.8115234375, "epoch": 0.6284875183553598, "fcm_dpo/beta": 0.001346116536296904, "fcm_dpo/delta": -0.02620844356715679, "fcm_dpo/margin": 278.5070495605469, "fcm_dpo/q_t": 0.413205087184906, "grad_norm": 37.65504455566406, "learning_rate": 1.8405649477212697e-07, "logits/chosen": -0.5920642614364624, "logits/rejected": -0.5966956615447998, "logps/chosen": -555.845458984375, "logps/ref_chosen": -62.63666534423828, "logps/ref_rejected": -103.28181457519531, "logps/rejected": -874.9976196289062, "loss": 1.1263, "margin_dpo/margin_mean": 278.5070495605469, "margin_dpo/margin_std": 432.5724792480469, "step": 428 }, { "KL/chosen_KL_mean": -506.41094970703125, "KL/mean": -613.6200561523438, "KL/rejected_KL_mean": -720.8291625976562, "KL/std": 329.47845458984375, "epoch": 0.6299559471365639, "fcm_dpo/beta": 0.0013451819540932775, "fcm_dpo/delta": -0.028253547847270966, "fcm_dpo/margin": 214.41818237304688, "fcm_dpo/q_t": 0.43296605348587036, "grad_norm": 36.45220947265625, "learning_rate": 1.828194884925749e-07, "logits/chosen": -0.6370459794998169, "logits/rejected": -0.6188497543334961, "logps/chosen": -587.6449584960938, "logps/ref_chosen": -81.23401641845703, "logps/ref_rejected": -91.79493713378906, "logps/rejected": -812.6240844726562, "loss": 1.1891, "margin_dpo/margin_mean": 214.41818237304688, "margin_dpo/margin_std": 390.85205078125, "step": 429 }, { "KL/chosen_KL_mean": -406.6867370605469, "KL/mean": -513.8056640625, "KL/rejected_KL_mean": -620.9246215820312, "KL/std": 296.41180419921875, "epoch": 0.631424375917768, "fcm_dpo/beta": 0.0013558154460042715, "fcm_dpo/delta": 0.06116756424307823, "fcm_dpo/margin": 214.23789978027344, "fcm_dpo/q_t": 0.43244469165802, "grad_norm": 34.47238540649414, "learning_rate": 1.8158425248197928e-07, "logits/chosen": -0.5731499791145325, "logits/rejected": -0.5749542713165283, "logps/chosen": -467.6070556640625, "logps/ref_chosen": -60.920326232910156, "logps/ref_rejected": -104.42280578613281, "logps/rejected": -725.347412109375, "loss": 1.1677, "margin_dpo/margin_mean": 214.23788452148438, "margin_dpo/margin_std": 341.7822265625, "step": 430 }, { "KL/chosen_KL_mean": -371.7575988769531, "KL/mean": -523.7132568359375, "KL/rejected_KL_mean": -675.6690063476562, "KL/std": 307.078857421875, "epoch": 0.6328928046989721, "fcm_dpo/beta": 0.0013448914978653193, "fcm_dpo/delta": -0.06126604229211807, "fcm_dpo/margin": 303.9114074707031, "fcm_dpo/q_t": 0.40351831912994385, "grad_norm": 35.25053405761719, "learning_rate": 1.8035081928995788e-07, "logits/chosen": -0.5623865723609924, "logits/rejected": -0.5650370717048645, "logps/chosen": -429.1063232421875, "logps/ref_chosen": -57.34874725341797, "logps/ref_rejected": -92.84022521972656, "logps/rejected": -768.50927734375, "loss": 1.0697, "margin_dpo/margin_mean": 303.91143798828125, "margin_dpo/margin_std": 338.7742919921875, "step": 431 }, { "KL/chosen_KL_mean": -368.0189514160156, "KL/mean": -524.6700439453125, "KL/rejected_KL_mean": -681.3211669921875, "KL/std": 317.7171936035156, "epoch": 0.6343612334801763, "fcm_dpo/beta": 0.001335380831733346, "fcm_dpo/delta": -0.0716981440782547, "fcm_dpo/margin": 313.3021545410156, "fcm_dpo/q_t": 0.4016353487968445, "grad_norm": 37.07876205444336, "learning_rate": 1.791192214186223e-07, "logits/chosen": -0.5415102243423462, "logits/rejected": -0.5325489044189453, "logps/chosen": -439.09375, "logps/ref_chosen": -71.07479095458984, "logps/ref_rejected": -98.57952880859375, "logps/rejected": -779.900634765625, "loss": 1.0569, "margin_dpo/margin_mean": 313.3021545410156, "margin_dpo/margin_std": 315.588134765625, "step": 432 }, { "KL/chosen_KL_mean": -469.5299377441406, "KL/mean": -571.39794921875, "KL/rejected_KL_mean": -673.2659912109375, "KL/std": 318.712890625, "epoch": 0.6358296622613803, "fcm_dpo/beta": 0.0013404275523498654, "fcm_dpo/delta": 0.07884688675403595, "fcm_dpo/margin": 203.736083984375, "fcm_dpo/q_t": 0.4341526925563812, "grad_norm": 34.616695404052734, "learning_rate": 1.7788949132172193e-07, "logits/chosen": -0.630656361579895, "logits/rejected": -0.6229075193405151, "logps/chosen": -527.8031005859375, "logps/ref_chosen": -58.273193359375, "logps/ref_rejected": -95.95089721679688, "logps/rejected": -769.2169189453125, "loss": 1.1937, "margin_dpo/margin_mean": 203.73606872558594, "margin_dpo/margin_std": 376.9918212890625, "step": 433 }, { "KL/chosen_KL_mean": -403.37701416015625, "KL/mean": -527.6131591796875, "KL/rejected_KL_mean": -651.8493041992188, "KL/std": 324.9095153808594, "epoch": 0.6372980910425844, "fcm_dpo/beta": 0.0013430386316031218, "fcm_dpo/delta": 0.01675334945321083, "fcm_dpo/margin": 248.47225952148438, "fcm_dpo/q_t": 0.42638862133026123, "grad_norm": 28.57368278503418, "learning_rate": 1.7666166140378853e-07, "logits/chosen": -0.6396076083183289, "logits/rejected": -0.6434687376022339, "logps/chosen": -465.3507080078125, "logps/ref_chosen": -61.97370147705078, "logps/ref_rejected": -78.49861145019531, "logps/rejected": -730.347900390625, "loss": 1.147, "margin_dpo/margin_mean": 248.4722900390625, "margin_dpo/margin_std": 408.96661376953125, "step": 434 }, { "KL/chosen_KL_mean": -392.12298583984375, "KL/mean": -531.1943359375, "KL/rejected_KL_mean": -670.2657470703125, "KL/std": 317.26971435546875, "epoch": 0.6387665198237885, "fcm_dpo/beta": 0.0013403587508946657, "fcm_dpo/delta": -0.023820772767066956, "fcm_dpo/margin": 278.14276123046875, "fcm_dpo/q_t": 0.41278964281082153, "grad_norm": 31.200889587402344, "learning_rate": 1.7543576401928218e-07, "logits/chosen": -0.7370610237121582, "logits/rejected": -0.731311559677124, "logps/chosen": -443.6250305175781, "logps/ref_chosen": -51.502052307128906, "logps/ref_rejected": -87.56689453125, "logps/rejected": -757.8326416015625, "loss": 1.1053, "margin_dpo/margin_mean": 278.1427307128906, "margin_dpo/margin_std": 360.19976806640625, "step": 435 }, { "KL/chosen_KL_mean": -413.7061767578125, "KL/mean": -527.6519775390625, "KL/rejected_KL_mean": -641.597900390625, "KL/std": 283.1363525390625, "epoch": 0.6402349486049926, "fcm_dpo/beta": 0.001346162287518382, "fcm_dpo/delta": 0.04453161358833313, "fcm_dpo/margin": 227.89169311523438, "fcm_dpo/q_t": 0.4281315207481384, "grad_norm": 32.2522087097168, "learning_rate": 1.742118314717391e-07, "logits/chosen": -0.6348098516464233, "logits/rejected": -0.6036412715911865, "logps/chosen": -485.10986328125, "logps/ref_chosen": -71.40371704101562, "logps/ref_rejected": -82.72775268554688, "logps/rejected": -724.3256225585938, "loss": 1.1522, "margin_dpo/margin_mean": 227.89169311523438, "margin_dpo/margin_std": 335.2723083496094, "step": 436 }, { "KL/chosen_KL_mean": -422.2236328125, "KL/mean": -543.794189453125, "KL/rejected_KL_mean": -665.36474609375, "KL/std": 282.27435302734375, "epoch": 0.6417033773861968, "fcm_dpo/beta": 0.0013560389634221792, "fcm_dpo/delta": 0.020979713648557663, "fcm_dpo/margin": 243.1411590576172, "fcm_dpo/q_t": 0.4221380054950714, "grad_norm": 22.420637130737305, "learning_rate": 1.7298989601292036e-07, "logits/chosen": -0.6972805857658386, "logits/rejected": -0.67779940366745, "logps/chosen": -486.9678955078125, "logps/ref_chosen": -64.7442626953125, "logps/ref_rejected": -82.04356384277344, "logps/rejected": -747.4083251953125, "loss": 1.1318, "margin_dpo/margin_mean": 243.14114379882812, "margin_dpo/margin_std": 330.1854248046875, "step": 437 }, { "KL/chosen_KL_mean": -457.51593017578125, "KL/mean": -587.1756591796875, "KL/rejected_KL_mean": -716.8353881835938, "KL/std": 319.5601806640625, "epoch": 0.6431718061674009, "fcm_dpo/beta": 0.0013543383684009314, "fcm_dpo/delta": -0.0014728754758834839, "fcm_dpo/margin": 259.3194274902344, "fcm_dpo/q_t": 0.41706225275993347, "grad_norm": 29.479515075683594, "learning_rate": 1.7176998984196144e-07, "logits/chosen": -0.6999799013137817, "logits/rejected": -0.6740041971206665, "logps/chosen": -516.5346069335938, "logps/ref_chosen": -59.0186653137207, "logps/ref_rejected": -83.07682800292969, "logps/rejected": -799.9122314453125, "loss": 1.1196, "margin_dpo/margin_mean": 259.3194274902344, "margin_dpo/margin_std": 347.9755859375, "step": 438 }, { "KL/chosen_KL_mean": -469.77227783203125, "KL/mean": -582.2373046875, "KL/rejected_KL_mean": -694.7022705078125, "KL/std": 336.91986083984375, "epoch": 0.644640234948605, "fcm_dpo/beta": 0.0013382842298597097, "fcm_dpo/delta": -0.07557918131351471, "fcm_dpo/margin": 224.93003845214844, "fcm_dpo/q_t": 0.43068742752075195, "grad_norm": 33.094661712646484, "learning_rate": 1.7055214510452458e-07, "logits/chosen": -0.6910462379455566, "logits/rejected": -0.6919522285461426, "logps/chosen": -523.5563354492188, "logps/ref_chosen": -53.78407669067383, "logps/ref_rejected": -83.98545837402344, "logps/rejected": -778.687744140625, "loss": 1.1801, "margin_dpo/margin_mean": 224.9300537109375, "margin_dpo/margin_std": 390.89849853515625, "step": 439 }, { "KL/chosen_KL_mean": -492.5137939453125, "KL/mean": -635.44140625, "KL/rejected_KL_mean": -778.3690185546875, "KL/std": 411.0914611816406, "epoch": 0.6461086637298091, "fcm_dpo/beta": 0.001335415756329894, "fcm_dpo/delta": -0.03305999934673309, "fcm_dpo/margin": 285.8551330566406, "fcm_dpo/q_t": 0.4155203104019165, "grad_norm": 49.92783737182617, "learning_rate": 1.6933639389195134e-07, "logits/chosen": -0.7562768459320068, "logits/rejected": -0.7523295879364014, "logps/chosen": -571.08056640625, "logps/ref_chosen": -78.56671905517578, "logps/ref_rejected": -96.49775695800781, "logps/rejected": -874.86669921875, "loss": 1.1137, "margin_dpo/margin_mean": 285.8551025390625, "margin_dpo/margin_std": 428.9686584472656, "step": 440 }, { "KL/chosen_KL_mean": -570.6612548828125, "KL/mean": -697.5283813476562, "KL/rejected_KL_mean": -824.3955688476562, "KL/std": 409.962890625, "epoch": 0.6475770925110133, "fcm_dpo/beta": 0.0013370553497225046, "fcm_dpo/delta": 0.01073683425784111, "fcm_dpo/margin": 253.73431396484375, "fcm_dpo/q_t": 0.4260130524635315, "grad_norm": 45.495601654052734, "learning_rate": 1.681227682404166e-07, "logits/chosen": -0.7495392560958862, "logits/rejected": -0.7344924807548523, "logps/chosen": -631.4856567382812, "logps/ref_chosen": -60.824440002441406, "logps/ref_rejected": -96.47080993652344, "logps/rejected": -920.8663940429688, "loss": 1.1806, "margin_dpo/margin_mean": 253.73431396484375, "margin_dpo/margin_std": 494.9446105957031, "step": 441 }, { "KL/chosen_KL_mean": -478.4798583984375, "KL/mean": -630.342041015625, "KL/rejected_KL_mean": -782.2042236328125, "KL/std": 391.163330078125, "epoch": 0.6490455212922174, "fcm_dpo/beta": 0.001329011982306838, "fcm_dpo/delta": -0.056281156837940216, "fcm_dpo/margin": 303.7243347167969, "fcm_dpo/q_t": 0.4097488522529602, "grad_norm": 31.81182861328125, "learning_rate": 1.669113001300851e-07, "logits/chosen": -0.718780517578125, "logits/rejected": -0.7090387344360352, "logps/chosen": -525.4910888671875, "logps/ref_chosen": -47.01121520996094, "logps/ref_rejected": -76.53926086425781, "logps/rejected": -858.7434692382812, "loss": 1.1065, "margin_dpo/margin_mean": 303.7243347167969, "margin_dpo/margin_std": 445.0675964355469, "step": 442 }, { "KL/chosen_KL_mean": -530.4336547851562, "KL/mean": -621.8048706054688, "KL/rejected_KL_mean": -713.1761474609375, "KL/std": 391.05328369140625, "epoch": 0.6505139500734214, "fcm_dpo/beta": 0.0013202856061980128, "fcm_dpo/delta": 0.0109033128246665, "fcm_dpo/margin": 182.74246215820312, "fcm_dpo/q_t": 0.4463666081428528, "grad_norm": 38.57601547241211, "learning_rate": 1.6570202148426815e-07, "logits/chosen": -0.724666953086853, "logits/rejected": -0.701187014579773, "logps/chosen": -601.7066650390625, "logps/ref_chosen": -71.27301788330078, "logps/ref_rejected": -86.679931640625, "logps/rejected": -799.8560791015625, "loss": 1.2554, "margin_dpo/margin_mean": 182.74249267578125, "margin_dpo/margin_std": 479.74188232421875, "step": 443 }, { "KL/chosen_KL_mean": -528.6512451171875, "KL/mean": -687.4662475585938, "KL/rejected_KL_mean": -846.2813110351562, "KL/std": 420.274169921875, "epoch": 0.6519823788546255, "fcm_dpo/beta": 0.0013028013054281473, "fcm_dpo/delta": -0.06701706349849701, "fcm_dpo/margin": 317.630126953125, "fcm_dpo/q_t": 0.40774786472320557, "grad_norm": 30.262189865112305, "learning_rate": 1.6449496416858282e-07, "logits/chosen": -0.6718389391899109, "logits/rejected": -0.6785389184951782, "logps/chosen": -585.8648681640625, "logps/ref_chosen": -57.213706970214844, "logps/ref_rejected": -97.25489807128906, "logps/rejected": -943.5361938476562, "loss": 1.1052, "margin_dpo/margin_mean": 317.630126953125, "margin_dpo/margin_std": 487.72283935546875, "step": 444 }, { "KL/chosen_KL_mean": -456.7267761230469, "KL/mean": -602.949951171875, "KL/rejected_KL_mean": -749.173095703125, "KL/std": 321.0992431640625, "epoch": 0.6534508076358296, "fcm_dpo/beta": 0.0012994688004255295, "fcm_dpo/delta": -0.031238578259944916, "fcm_dpo/margin": 292.4462890625, "fcm_dpo/q_t": 0.41294771432876587, "grad_norm": 30.900972366333008, "learning_rate": 1.6329015999011182e-07, "logits/chosen": -0.6935943365097046, "logits/rejected": -0.6810604333877563, "logps/chosen": -524.026611328125, "logps/ref_chosen": -67.29979705810547, "logps/ref_rejected": -92.68267059326172, "logps/rejected": -841.855712890625, "loss": 1.1085, "margin_dpo/margin_mean": 292.4463195800781, "margin_dpo/margin_std": 402.79986572265625, "step": 445 }, { "KL/chosen_KL_mean": -402.2188720703125, "KL/mean": -560.820556640625, "KL/rejected_KL_mean": -719.4222412109375, "KL/std": 337.19036865234375, "epoch": 0.6549192364170338, "fcm_dpo/beta": 0.0012904771137982607, "fcm_dpo/delta": -0.062246501445770264, "fcm_dpo/margin": 317.2033386230469, "fcm_dpo/q_t": 0.4033682346343994, "grad_norm": 34.864933013916016, "learning_rate": 1.6208764069656578e-07, "logits/chosen": -0.6820211410522461, "logits/rejected": -0.6916550397872925, "logps/chosen": -461.3173522949219, "logps/ref_chosen": -59.098487854003906, "logps/ref_rejected": -101.26419067382812, "logps/rejected": -820.6864013671875, "loss": 1.0721, "margin_dpo/margin_mean": 317.2033386230469, "margin_dpo/margin_std": 355.0349426269531, "step": 446 }, { "KL/chosen_KL_mean": -425.578857421875, "KL/mean": -594.2220458984375, "KL/rejected_KL_mean": -762.8651733398438, "KL/std": 410.32159423828125, "epoch": 0.6563876651982379, "fcm_dpo/beta": 0.0012605185620486736, "fcm_dpo/delta": -0.07881560921669006, "fcm_dpo/margin": 337.28631591796875, "fcm_dpo/q_t": 0.4043826460838318, "grad_norm": 26.480072021484375, "learning_rate": 1.608874379754465e-07, "logits/chosen": -0.7632917761802673, "logits/rejected": -0.7825505137443542, "logps/chosen": -481.6541748046875, "logps/ref_chosen": -56.07533264160156, "logps/ref_rejected": -98.69475555419922, "logps/rejected": -861.5599365234375, "loss": 1.079, "margin_dpo/margin_mean": 337.28631591796875, "margin_dpo/margin_std": 460.5667419433594, "step": 447 }, { "KL/chosen_KL_mean": -467.5102844238281, "KL/mean": -623.7568359375, "KL/rejected_KL_mean": -780.00341796875, "KL/std": 348.59857177734375, "epoch": 0.657856093979442, "fcm_dpo/beta": 0.001254360075108707, "fcm_dpo/delta": -0.04375584051012993, "fcm_dpo/margin": 312.4931335449219, "fcm_dpo/q_t": 0.40883946418762207, "grad_norm": 32.5056266784668, "learning_rate": 1.5968958345321177e-07, "logits/chosen": -0.6593608856201172, "logits/rejected": -0.665467381477356, "logps/chosen": -527.51416015625, "logps/ref_chosen": -60.00384521484375, "logps/ref_rejected": -102.26465606689453, "logps/rejected": -882.26806640625, "loss": 1.0861, "margin_dpo/margin_mean": 312.4931335449219, "margin_dpo/margin_std": 379.61871337890625, "step": 448 }, { "KL/chosen_KL_mean": -483.6077575683594, "KL/mean": -639.1844482421875, "KL/rejected_KL_mean": -794.76123046875, "KL/std": 414.4624938964844, "epoch": 0.6593245227606461, "fcm_dpo/beta": 0.001239268109202385, "fcm_dpo/delta": -0.03710303083062172, "fcm_dpo/margin": 311.15350341796875, "fcm_dpo/q_t": 0.41368383169174194, "grad_norm": 29.851238250732422, "learning_rate": 1.584941086944423e-07, "logits/chosen": -0.7158821821212769, "logits/rejected": -0.7080788016319275, "logps/chosen": -551.1343383789062, "logps/ref_chosen": -67.52661895751953, "logps/ref_rejected": -88.59690856933594, "logps/rejected": -883.358154296875, "loss": 1.1186, "margin_dpo/margin_mean": 311.15350341796875, "margin_dpo/margin_std": 484.8404846191406, "step": 449 }, { "KL/chosen_KL_mean": -386.138671875, "KL/mean": -558.7864990234375, "KL/rejected_KL_mean": -731.434326171875, "KL/std": 357.3294982910156, "epoch": 0.6607929515418502, "fcm_dpo/beta": 0.001226372318342328, "fcm_dpo/delta": -0.07663866132497787, "fcm_dpo/margin": 345.29571533203125, "fcm_dpo/q_t": 0.3992771804332733, "grad_norm": 48.51375198364258, "learning_rate": 1.573010452010098e-07, "logits/chosen": -0.7433021068572998, "logits/rejected": -0.7565724849700928, "logps/chosen": -443.24676513671875, "logps/ref_chosen": -57.10811996459961, "logps/ref_rejected": -102.75494384765625, "logps/rejected": -834.1893310546875, "loss": 1.0505, "margin_dpo/margin_mean": 345.29571533203125, "margin_dpo/margin_std": 349.22174072265625, "step": 450 }, { "KL/chosen_KL_mean": -553.4360961914062, "KL/mean": -673.3494873046875, "KL/rejected_KL_mean": -793.2628784179688, "KL/std": 449.323486328125, "epoch": 0.6622613803230544, "fcm_dpo/beta": 0.0012323290575295687, "fcm_dpo/delta": 0.05590929836034775, "fcm_dpo/margin": 239.82675170898438, "fcm_dpo/q_t": 0.4289303421974182, "grad_norm": 36.48598861694336, "learning_rate": 1.5611042441124687e-07, "logits/chosen": -0.7902913689613342, "logits/rejected": -0.7614456415176392, "logps/chosen": -611.9049072265625, "logps/ref_chosen": -58.46883010864258, "logps/ref_rejected": -72.92941284179688, "logps/rejected": -866.1922607421875, "loss": 1.2143, "margin_dpo/margin_mean": 239.8267364501953, "margin_dpo/margin_std": 536.7329711914062, "step": 451 }, { "KL/chosen_KL_mean": -394.72314453125, "KL/mean": -541.702392578125, "KL/rejected_KL_mean": -688.681640625, "KL/std": 334.7508239746094, "epoch": 0.6637298091042585, "fcm_dpo/beta": 0.0012294357875362039, "fcm_dpo/delta": -0.01188319455832243, "fcm_dpo/margin": 293.95849609375, "fcm_dpo/q_t": 0.4138975441455841, "grad_norm": 26.67286491394043, "learning_rate": 1.549222776991186e-07, "logits/chosen": -0.6999271512031555, "logits/rejected": -0.7178478240966797, "logps/chosen": -445.11370849609375, "logps/ref_chosen": -50.39055252075195, "logps/ref_rejected": -97.77142333984375, "logps/rejected": -786.4530639648438, "loss": 1.0979, "margin_dpo/margin_mean": 293.95849609375, "margin_dpo/margin_std": 333.3941955566406, "step": 452 }, { "KL/chosen_KL_mean": -457.27532958984375, "KL/mean": -610.49267578125, "KL/rejected_KL_mean": -763.7099609375, "KL/std": 360.0783386230469, "epoch": 0.6651982378854625, "fcm_dpo/beta": 0.0012230616994202137, "fcm_dpo/delta": -0.025895066559314728, "fcm_dpo/margin": 306.43463134765625, "fcm_dpo/q_t": 0.41480809450149536, "grad_norm": 35.29240036010742, "learning_rate": 1.5373663637339584e-07, "logits/chosen": -0.7407764792442322, "logits/rejected": -0.7278668880462646, "logps/chosen": -514.9901733398438, "logps/ref_chosen": -57.71485137939453, "logps/ref_rejected": -82.20741271972656, "logps/rejected": -845.9173583984375, "loss": 1.1113, "margin_dpo/margin_mean": 306.43463134765625, "margin_dpo/margin_std": 436.13421630859375, "step": 453 }, { "KL/chosen_KL_mean": -560.2719116210938, "KL/mean": -723.659912109375, "KL/rejected_KL_mean": -887.0479736328125, "KL/std": 408.50616455078125, "epoch": 0.6666666666666666, "fcm_dpo/beta": 0.0012127361260354519, "fcm_dpo/delta": -0.04838482663035393, "fcm_dpo/margin": 326.7760009765625, "fcm_dpo/q_t": 0.4091429114341736, "grad_norm": 25.13304328918457, "learning_rate": 1.5255353167683017e-07, "logits/chosen": -0.7615550756454468, "logits/rejected": -0.7526893019676208, "logps/chosen": -621.217529296875, "logps/ref_chosen": -60.945648193359375, "logps/ref_rejected": -84.95079040527344, "logps/rejected": -971.9987182617188, "loss": 1.0997, "margin_dpo/margin_mean": 326.7760314941406, "margin_dpo/margin_std": 461.345947265625, "step": 454 }, { "KL/chosen_KL_mean": -504.2913818359375, "KL/mean": -699.232421875, "KL/rejected_KL_mean": -894.17333984375, "KL/std": 421.8465576171875, "epoch": 0.6681350954478708, "fcm_dpo/beta": 0.001196699682623148, "fcm_dpo/delta": -0.12219969183206558, "fcm_dpo/margin": 389.8819580078125, "fcm_dpo/q_t": 0.39322376251220703, "grad_norm": 33.46556854248047, "learning_rate": 1.5137299478533064e-07, "logits/chosen": -0.7414263486862183, "logits/rejected": -0.7670072317123413, "logps/chosen": -549.1781005859375, "logps/ref_chosen": -44.88671112060547, "logps/ref_rejected": -115.30147552490234, "logps/rejected": -1009.4747924804688, "loss": 1.0514, "margin_dpo/margin_mean": 389.8819580078125, "margin_dpo/margin_std": 477.7015380859375, "step": 455 }, { "KL/chosen_KL_mean": -507.195556640625, "KL/mean": -697.402587890625, "KL/rejected_KL_mean": -887.6096801757812, "KL/std": 429.1546630859375, "epoch": 0.6696035242290749, "fcm_dpo/beta": 0.0011668999213725328, "fcm_dpo/delta": -0.09819206595420837, "fcm_dpo/margin": 380.4140625, "fcm_dpo/q_t": 0.39691370725631714, "grad_norm": 38.52052688598633, "learning_rate": 1.5019505680714232e-07, "logits/chosen": -0.7470618486404419, "logits/rejected": -0.7721505165100098, "logps/chosen": -564.2323608398438, "logps/ref_chosen": -57.036781311035156, "logps/ref_rejected": -105.21784210205078, "logps/rejected": -992.8275146484375, "loss": 1.0411, "margin_dpo/margin_mean": 380.4140625, "margin_dpo/margin_std": 404.1457214355469, "step": 456 }, { "KL/chosen_KL_mean": -478.6456298828125, "KL/mean": -666.3101806640625, "KL/rejected_KL_mean": -853.9747314453125, "KL/std": 396.79046630859375, "epoch": 0.671071953010279, "fcm_dpo/beta": 0.00114365853369236, "fcm_dpo/delta": -0.0830128863453865, "fcm_dpo/margin": 375.3291015625, "fcm_dpo/q_t": 0.3985850214958191, "grad_norm": 33.69874572753906, "learning_rate": 1.4901974878202627e-07, "logits/chosen": -0.7634145617485046, "logits/rejected": -0.7695783376693726, "logps/chosen": -532.88818359375, "logps/ref_chosen": -54.24253845214844, "logps/ref_rejected": -85.10956573486328, "logps/rejected": -939.0842895507812, "loss": 1.0498, "margin_dpo/margin_mean": 375.3291015625, "margin_dpo/margin_std": 390.97637939453125, "step": 457 }, { "KL/chosen_KL_mean": -554.390869140625, "KL/mean": -707.5039672851562, "KL/rejected_KL_mean": -860.6171264648438, "KL/std": 387.835693359375, "epoch": 0.6725403817914831, "fcm_dpo/beta": 0.0011345113161951303, "fcm_dpo/delta": 0.001986939460039139, "fcm_dpo/margin": 306.226318359375, "fcm_dpo/q_t": 0.42130082845687866, "grad_norm": 35.67972183227539, "learning_rate": 1.4784710168044212e-07, "logits/chosen": -0.8082045316696167, "logits/rejected": -0.8004375100135803, "logps/chosen": -609.7996826171875, "logps/ref_chosen": -55.40888214111328, "logps/ref_rejected": -97.68325805664062, "logps/rejected": -958.3004150390625, "loss": 1.1318, "margin_dpo/margin_mean": 306.226318359375, "margin_dpo/margin_std": 444.1019287109375, "step": 458 }, { "KL/chosen_KL_mean": -591.978515625, "KL/mean": -779.3623657226562, "KL/rejected_KL_mean": -966.7462158203125, "KL/std": 468.7334289550781, "epoch": 0.6740088105726872, "fcm_dpo/beta": 0.0011250160168856382, "fcm_dpo/delta": -0.07521242648363113, "fcm_dpo/margin": 374.7676696777344, "fcm_dpo/q_t": 0.4042002856731415, "grad_norm": 38.14002990722656, "learning_rate": 1.466771464027316e-07, "logits/chosen": -0.7572959661483765, "logits/rejected": -0.7787726521492004, "logps/chosen": -638.5360107421875, "logps/ref_chosen": -46.55748748779297, "logps/ref_rejected": -86.16854095458984, "logps/rejected": -1052.914794921875, "loss": 1.0933, "margin_dpo/margin_mean": 374.7676696777344, "margin_dpo/margin_std": 536.9511108398438, "step": 459 }, { "KL/chosen_KL_mean": -629.0224609375, "KL/mean": -831.0809936523438, "KL/rejected_KL_mean": -1033.1395263671875, "KL/std": 450.43658447265625, "epoch": 0.6754772393538914, "fcm_dpo/beta": 0.0011082333512604237, "fcm_dpo/delta": -0.10236979275941849, "fcm_dpo/margin": 404.1170349121094, "fcm_dpo/q_t": 0.39786800742149353, "grad_norm": 28.9811954498291, "learning_rate": 1.4550991377830423e-07, "logits/chosen": -0.8412591814994812, "logits/rejected": -0.8749679923057556, "logps/chosen": -680.6573486328125, "logps/ref_chosen": -51.63489532470703, "logps/ref_rejected": -104.11935424804688, "logps/rejected": -1137.2587890625, "loss": 1.055, "margin_dpo/margin_mean": 404.1170349121094, "margin_dpo/margin_std": 494.65667724609375, "step": 460 }, { "KL/chosen_KL_mean": -658.8389892578125, "KL/mean": -815.6547241210938, "KL/rejected_KL_mean": -972.470458984375, "KL/std": 444.0308837890625, "epoch": 0.6769456681350955, "fcm_dpo/beta": 0.0011054275091737509, "fcm_dpo/delta": 0.0032142307609319687, "fcm_dpo/margin": 313.631591796875, "fcm_dpo/q_t": 0.42101699113845825, "grad_norm": 28.833742141723633, "learning_rate": 1.4434543456482518e-07, "logits/chosen": -0.8508638739585876, "logits/rejected": -0.8650791049003601, "logps/chosen": -714.0208740234375, "logps/ref_chosen": -55.18195724487305, "logps/ref_rejected": -86.47689819335938, "logps/rejected": -1058.9473876953125, "loss": 1.1359, "margin_dpo/margin_mean": 313.631591796875, "margin_dpo/margin_std": 482.70458984375, "step": 461 }, { "KL/chosen_KL_mean": -635.347412109375, "KL/mean": -766.1812744140625, "KL/rejected_KL_mean": -897.0151977539062, "KL/std": 428.46905517578125, "epoch": 0.6784140969162996, "fcm_dpo/beta": 0.0011110971681773663, "fcm_dpo/delta": 0.06107814237475395, "fcm_dpo/margin": 261.667724609375, "fcm_dpo/q_t": 0.43434569239616394, "grad_norm": 37.025150299072266, "learning_rate": 1.4318373944740484e-07, "logits/chosen": -0.8966616988182068, "logits/rejected": -0.8862354755401611, "logps/chosen": -705.2755126953125, "logps/ref_chosen": -69.92803192138672, "logps/ref_rejected": -78.84111022949219, "logps/rejected": -975.8563232421875, "loss": 1.1858, "margin_dpo/margin_mean": 261.667724609375, "margin_dpo/margin_std": 488.41827392578125, "step": 462 }, { "KL/chosen_KL_mean": -617.084716796875, "KL/mean": -794.1759033203125, "KL/rejected_KL_mean": -971.2671508789062, "KL/std": 443.2431640625, "epoch": 0.6798825256975036, "fcm_dpo/beta": 0.0011113372165709734, "fcm_dpo/delta": -0.045437999069690704, "fcm_dpo/margin": 354.18243408203125, "fcm_dpo/q_t": 0.4097776710987091, "grad_norm": 36.853878021240234, "learning_rate": 1.4202485903778976e-07, "logits/chosen": -0.8798764944076538, "logits/rejected": -0.893752932548523, "logps/chosen": -672.3590698242188, "logps/ref_chosen": -55.27437210083008, "logps/ref_rejected": -89.02497863769531, "logps/rejected": -1060.2921142578125, "loss": 1.101, "margin_dpo/margin_mean": 354.18243408203125, "margin_dpo/margin_std": 491.94415283203125, "step": 463 }, { "KL/chosen_KL_mean": -604.09423828125, "KL/mean": -862.4444580078125, "KL/rejected_KL_mean": -1120.794677734375, "KL/std": 517.349853515625, "epoch": 0.6813509544787077, "fcm_dpo/beta": 0.0010644260328263044, "fcm_dpo/delta": -0.2137536257505417, "fcm_dpo/margin": 516.7005615234375, "fcm_dpo/q_t": 0.37335121631622314, "grad_norm": 34.50212478637695, "learning_rate": 1.4086882387355658e-07, "logits/chosen": -0.8490620851516724, "logits/rejected": -0.9159963726997375, "logps/chosen": -655.0065307617188, "logps/ref_chosen": -50.91230010986328, "logps/ref_rejected": -102.4893798828125, "logps/rejected": -1223.2841796875, "loss": 0.9802, "margin_dpo/margin_mean": 516.7005615234375, "margin_dpo/margin_std": 523.5531005859375, "step": 464 }, { "KL/chosen_KL_mean": -608.3139038085938, "KL/mean": -818.8243408203125, "KL/rejected_KL_mean": -1029.334716796875, "KL/std": 525.769775390625, "epoch": 0.6828193832599119, "fcm_dpo/beta": 0.0010484305676072836, "fcm_dpo/delta": -0.09556633979082108, "fcm_dpo/margin": 421.0207824707031, "fcm_dpo/q_t": 0.39575302600860596, "grad_norm": 34.12406539916992, "learning_rate": 1.3971566441730714e-07, "logits/chosen": -0.8353086709976196, "logits/rejected": -0.8615255951881409, "logps/chosen": -668.4307861328125, "logps/ref_chosen": -60.116851806640625, "logps/ref_rejected": -113.94602966308594, "logps/rejected": -1143.28076171875, "loss": 1.0727, "margin_dpo/margin_mean": 421.020751953125, "margin_dpo/margin_std": 551.701416015625, "step": 465 }, { "KL/chosen_KL_mean": -665.450439453125, "KL/mean": -846.2392578125, "KL/rejected_KL_mean": -1027.0279541015625, "KL/std": 486.2896728515625, "epoch": 0.684287812041116, "fcm_dpo/beta": 0.0010309141362085938, "fcm_dpo/delta": -0.02415800467133522, "fcm_dpo/margin": 361.57757568359375, "fcm_dpo/q_t": 0.41319042444229126, "grad_norm": 30.475038528442383, "learning_rate": 1.3856541105586545e-07, "logits/chosen": -0.8484971523284912, "logits/rejected": -0.850833535194397, "logps/chosen": -718.371337890625, "logps/ref_chosen": -52.920921325683594, "logps/ref_rejected": -90.3154296875, "logps/rejected": -1117.343505859375, "loss": 1.1194, "margin_dpo/margin_mean": 361.57757568359375, "margin_dpo/margin_std": 526.1229248046875, "step": 466 }, { "KL/chosen_KL_mean": -806.384033203125, "KL/mean": -1001.665283203125, "KL/rejected_KL_mean": -1196.946533203125, "KL/std": 618.6810913085938, "epoch": 0.6857562408223201, "fcm_dpo/beta": 0.0010187213774770498, "fcm_dpo/delta": -0.05139738321304321, "fcm_dpo/margin": 390.5625, "fcm_dpo/q_t": 0.412168025970459, "grad_norm": 39.5161018371582, "learning_rate": 1.3741809409947729e-07, "logits/chosen": -0.9231687784194946, "logits/rejected": -0.9177477359771729, "logps/chosen": -885.099853515625, "logps/ref_chosen": -78.7158203125, "logps/ref_rejected": -102.86019897460938, "logps/rejected": -1299.806640625, "loss": 1.153, "margin_dpo/margin_mean": 390.5625, "margin_dpo/margin_std": 716.0555419921875, "step": 467 }, { "KL/chosen_KL_mean": -630.0722045898438, "KL/mean": -863.9874877929688, "KL/rejected_KL_mean": -1097.9027099609375, "KL/std": 546.4240112304688, "epoch": 0.6872246696035242, "fcm_dpo/beta": 0.001004476216621697, "fcm_dpo/delta": -0.12587127089500427, "fcm_dpo/margin": 467.8305358886719, "fcm_dpo/q_t": 0.39484673738479614, "grad_norm": 56.975738525390625, "learning_rate": 1.362737437810114e-07, "logits/chosen": -0.8668639659881592, "logits/rejected": -0.8785967826843262, "logps/chosen": -700.007568359375, "logps/ref_chosen": -69.93536376953125, "logps/ref_rejected": -101.02880859375, "logps/rejected": -1198.9315185546875, "loss": 1.052, "margin_dpo/margin_mean": 467.83050537109375, "margin_dpo/margin_std": 610.6904296875, "step": 468 }, { "KL/chosen_KL_mean": -658.3848876953125, "KL/mean": -892.792724609375, "KL/rejected_KL_mean": -1127.20068359375, "KL/std": 478.67694091796875, "epoch": 0.6886930983847284, "fcm_dpo/beta": 0.0009764886926859617, "fcm_dpo/delta": -0.11406978219747543, "fcm_dpo/margin": 468.81585693359375, "fcm_dpo/q_t": 0.3934182822704315, "grad_norm": 53.98129653930664, "learning_rate": 1.351323902551631e-07, "logits/chosen": -0.8874689340591431, "logits/rejected": -0.9010331630706787, "logps/chosen": -726.509521484375, "logps/ref_chosen": -68.12469482421875, "logps/ref_rejected": -104.78640747070312, "logps/rejected": -1231.987060546875, "loss": 1.0452, "margin_dpo/margin_mean": 468.81585693359375, "margin_dpo/margin_std": 534.7791748046875, "step": 469 }, { "KL/chosen_KL_mean": -561.26513671875, "KL/mean": -760.3869018554688, "KL/rejected_KL_mean": -959.5086669921875, "KL/std": 498.2298278808594, "epoch": 0.6901615271659325, "fcm_dpo/beta": 0.0009724711999297142, "fcm_dpo/delta": -0.03880758211016655, "fcm_dpo/margin": 398.2434387207031, "fcm_dpo/q_t": 0.41007864475250244, "grad_norm": 30.534019470214844, "learning_rate": 1.339940635976592e-07, "logits/chosen": -0.8203660249710083, "logits/rejected": -0.8314099311828613, "logps/chosen": -605.05712890625, "logps/ref_chosen": -43.791927337646484, "logps/ref_rejected": -82.70285034179688, "logps/rejected": -1042.21142578125, "loss": 1.1043, "margin_dpo/margin_mean": 398.2434387207031, "margin_dpo/margin_std": 553.9703979492188, "step": 470 }, { "KL/chosen_KL_mean": -689.858642578125, "KL/mean": -864.8088989257812, "KL/rejected_KL_mean": -1039.7591552734375, "KL/std": 511.3665771484375, "epoch": 0.6916299559471366, "fcm_dpo/beta": 0.0009652703301981091, "fcm_dpo/delta": 0.01225661113858223, "fcm_dpo/margin": 349.9005432128906, "fcm_dpo/q_t": 0.42378807067871094, "grad_norm": 45.94660568237305, "learning_rate": 1.3285879380446563e-07, "logits/chosen": -0.903639554977417, "logits/rejected": -0.9123305082321167, "logps/chosen": -753.1981201171875, "logps/ref_chosen": -63.33952331542969, "logps/ref_rejected": -83.61048126220703, "logps/rejected": -1123.36962890625, "loss": 1.1473, "margin_dpo/margin_mean": 349.90057373046875, "margin_dpo/margin_std": 566.9776000976562, "step": 471 }, { "KL/chosen_KL_mean": -701.3533935546875, "KL/mean": -908.082763671875, "KL/rejected_KL_mean": -1114.812255859375, "KL/std": 622.212158203125, "epoch": 0.6930983847283406, "fcm_dpo/beta": 0.0009603024809621274, "fcm_dpo/delta": -0.04937269911170006, "fcm_dpo/margin": 413.4587707519531, "fcm_dpo/q_t": 0.4133405089378357, "grad_norm": 30.506114959716797, "learning_rate": 1.317266107909975e-07, "logits/chosen": -0.9473692178726196, "logits/rejected": -0.9324079751968384, "logps/chosen": -785.01953125, "logps/ref_chosen": -83.66610717773438, "logps/ref_rejected": -117.20919799804688, "logps/rejected": -1232.021484375, "loss": 1.1316, "margin_dpo/margin_mean": 413.458740234375, "margin_dpo/margin_std": 694.000732421875, "step": 472 }, { "KL/chosen_KL_mean": -834.3516845703125, "KL/mean": -932.49462890625, "KL/rejected_KL_mean": -1030.637451171875, "KL/std": 639.5214233398438, "epoch": 0.6945668135095447, "fcm_dpo/beta": 0.0009610364213585854, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 196.28573608398438, "fcm_dpo/q_t": 0.4586009979248047, "grad_norm": 95.01997375488281, "learning_rate": 1.3059754439133002e-07, "logits/chosen": -0.9266204833984375, "logits/rejected": -0.9034590721130371, "logps/chosen": -897.8486938476562, "logps/ref_chosen": -63.49696731567383, "logps/ref_rejected": -81.14657592773438, "logps/rejected": -1111.7840576171875, "loss": 1.3466, "margin_dpo/margin_mean": 196.28573608398438, "margin_dpo/margin_std": 829.6607666015625, "step": 473 }, { "KL/chosen_KL_mean": -665.4111328125, "KL/mean": -841.2649536132812, "KL/rejected_KL_mean": -1017.1187744140625, "KL/std": 553.4940795898438, "epoch": 0.6960352422907489, "fcm_dpo/beta": 0.0009541058680042624, "fcm_dpo/delta": -0.07264073938131332, "fcm_dpo/margin": 351.7076416015625, "fcm_dpo/q_t": 0.4222460687160492, "grad_norm": 34.34967041015625, "learning_rate": 1.2947162435741277e-07, "logits/chosen": -0.8338485956192017, "logits/rejected": -0.8360385894775391, "logps/chosen": -718.0230712890625, "logps/ref_chosen": -52.6119384765625, "logps/ref_rejected": -90.08041381835938, "logps/rejected": -1107.19921875, "loss": 1.1678, "margin_dpo/margin_mean": 351.7076416015625, "margin_dpo/margin_std": 621.35986328125, "step": 474 }, { "KL/chosen_KL_mean": -480.55462646484375, "KL/mean": -702.83447265625, "KL/rejected_KL_mean": -925.1143798828125, "KL/std": 437.4728088378906, "epoch": 0.697503671071953, "fcm_dpo/beta": 0.0009357599192298949, "fcm_dpo/delta": -0.06906232237815857, "fcm_dpo/margin": 444.5597229003906, "fcm_dpo/q_t": 0.40210431814193726, "grad_norm": 49.64398956298828, "learning_rate": 1.2834888035828596e-07, "logits/chosen": -0.8846492767333984, "logits/rejected": -0.911199688911438, "logps/chosen": -523.0498046875, "logps/ref_chosen": -42.49519348144531, "logps/ref_rejected": -90.06294250488281, "logps/rejected": -1015.17724609375, "loss": 1.0605, "margin_dpo/margin_mean": 444.5597229003906, "margin_dpo/margin_std": 480.71630859375, "step": 475 }, { "KL/chosen_KL_mean": -625.072509765625, "KL/mean": -801.80712890625, "KL/rejected_KL_mean": -978.5418090820312, "KL/std": 506.44488525390625, "epoch": 0.6989720998531571, "fcm_dpo/beta": 0.0009366124868392944, "fcm_dpo/delta": 0.019588805735111237, "fcm_dpo/margin": 353.46929931640625, "fcm_dpo/q_t": 0.4231525659561157, "grad_norm": 63.98448181152344, "learning_rate": 1.2722934197929802e-07, "logits/chosen": -0.8908638954162598, "logits/rejected": -0.9078314304351807, "logps/chosen": -668.0218505859375, "logps/ref_chosen": -42.94938278198242, "logps/ref_rejected": -73.71023559570312, "logps/rejected": -1052.2520751953125, "loss": 1.132, "margin_dpo/margin_mean": 353.46929931640625, "margin_dpo/margin_std": 492.0439758300781, "step": 476 }, { "KL/chosen_KL_mean": -637.2841186523438, "KL/mean": -817.2775268554688, "KL/rejected_KL_mean": -997.2708740234375, "KL/std": 515.0693359375, "epoch": 0.7004405286343612, "fcm_dpo/beta": 0.0009418263216502964, "fcm_dpo/delta": 0.011177122592926025, "fcm_dpo/margin": 359.98681640625, "fcm_dpo/q_t": 0.42166703939437866, "grad_norm": 36.05009460449219, "learning_rate": 1.2611303872132631e-07, "logits/chosen": -0.8980028629302979, "logits/rejected": -0.8670873641967773, "logps/chosen": -708.0567626953125, "logps/ref_chosen": -70.77261352539062, "logps/ref_rejected": -76.13737487792969, "logps/rejected": -1073.408203125, "loss": 1.1558, "margin_dpo/margin_mean": 359.9867858886719, "margin_dpo/margin_std": 619.1052856445312, "step": 477 }, { "KL/chosen_KL_mean": -511.648193359375, "KL/mean": -720.3424072265625, "KL/rejected_KL_mean": -929.03662109375, "KL/std": 461.0159912109375, "epoch": 0.7019089574155654, "fcm_dpo/beta": 0.0009357367525808513, "fcm_dpo/delta": -0.04220425337553024, "fcm_dpo/margin": 417.388427734375, "fcm_dpo/q_t": 0.4093322157859802, "grad_norm": 60.29072189331055, "learning_rate": 1.2500000000000005e-07, "logits/chosen": -0.7977697849273682, "logits/rejected": -0.8219680190086365, "logps/chosen": -553.0887451171875, "logps/ref_chosen": -41.440513610839844, "logps/ref_rejected": -85.36196899414062, "logps/rejected": -1014.3985595703125, "loss": 1.0931, "margin_dpo/margin_mean": 417.388427734375, "margin_dpo/margin_std": 540.1747436523438, "step": 478 }, { "KL/chosen_KL_mean": -685.029541015625, "KL/mean": -879.80517578125, "KL/rejected_KL_mean": -1074.580810546875, "KL/std": 583.0810546875, "epoch": 0.7033773861967695, "fcm_dpo/beta": 0.000935424177441746, "fcm_dpo/delta": -0.015560656785964966, "fcm_dpo/margin": 389.55133056640625, "fcm_dpo/q_t": 0.4196696877479553, "grad_norm": 31.228153228759766, "learning_rate": 1.2389025514492456e-07, "logits/chosen": -0.8409562706947327, "logits/rejected": -0.8747595548629761, "logps/chosen": -738.9374389648438, "logps/ref_chosen": -53.907920837402344, "logps/ref_rejected": -95.1163330078125, "logps/rejected": -1169.697265625, "loss": 1.1435, "margin_dpo/margin_mean": 389.55126953125, "margin_dpo/margin_std": 670.361328125, "step": 479 }, { "KL/chosen_KL_mean": -805.6644897460938, "KL/mean": -962.97802734375, "KL/rejected_KL_mean": -1120.29150390625, "KL/std": 531.8762817382812, "epoch": 0.7048458149779736, "fcm_dpo/beta": 0.0009248163551092148, "fcm_dpo/delta": -0.046877436339855194, "fcm_dpo/margin": 314.626953125, "fcm_dpo/q_t": 0.43354925513267517, "grad_norm": 39.37574768066406, "learning_rate": 1.227838333989088e-07, "logits/chosen": -0.8963419198989868, "logits/rejected": -0.898643970489502, "logps/chosen": -864.34716796875, "logps/ref_chosen": -58.682701110839844, "logps/ref_rejected": -82.93248748779297, "logps/rejected": -1203.223876953125, "loss": 1.1908, "margin_dpo/margin_mean": 314.626953125, "margin_dpo/margin_std": 573.584228515625, "step": 480 }, { "KL/chosen_KL_mean": -680.197509765625, "KL/mean": -925.4300537109375, "KL/rejected_KL_mean": -1170.6624755859375, "KL/std": 552.0270385742188, "epoch": 0.7063142437591777, "fcm_dpo/beta": 0.0009076240239664912, "fcm_dpo/delta": -0.09952923655509949, "fcm_dpo/margin": 490.46502685546875, "fcm_dpo/q_t": 0.397752046585083, "grad_norm": 46.874427795410156, "learning_rate": 1.2168076391719489e-07, "logits/chosen": -0.9476274251937866, "logits/rejected": -0.9800101518630981, "logps/chosen": -735.1617431640625, "logps/ref_chosen": -54.964271545410156, "logps/ref_rejected": -92.42044067382812, "logps/rejected": -1263.0830078125, "loss": 1.0567, "margin_dpo/margin_mean": 490.46502685546875, "margin_dpo/margin_std": 596.1909790039062, "step": 481 }, { "KL/chosen_KL_mean": -753.5008544921875, "KL/mean": -868.485595703125, "KL/rejected_KL_mean": -983.4703369140625, "KL/std": 577.3681640625, "epoch": 0.7077826725403817, "fcm_dpo/beta": 0.0009085052879527211, "fcm_dpo/delta": 0.06787097454071045, "fcm_dpo/margin": 229.9693603515625, "fcm_dpo/q_t": 0.4502810835838318, "grad_norm": 55.56328201293945, "learning_rate": 1.2058107576668938e-07, "logits/chosen": -0.8725358843803406, "logits/rejected": -0.8671541213989258, "logps/chosen": -821.0543212890625, "logps/ref_chosen": -67.553466796875, "logps/ref_rejected": -87.58953857421875, "logps/rejected": -1071.059814453125, "loss": 1.2717, "margin_dpo/margin_mean": 229.9693603515625, "margin_dpo/margin_std": 657.5565185546875, "step": 482 }, { "KL/chosen_KL_mean": -682.5269165039062, "KL/mean": -933.0989990234375, "KL/rejected_KL_mean": -1183.6710205078125, "KL/std": 570.397216796875, "epoch": 0.7092511013215859, "fcm_dpo/beta": 0.0008986013708636165, "fcm_dpo/delta": -0.10536154359579086, "fcm_dpo/margin": 501.1441345214844, "fcm_dpo/q_t": 0.3984888792037964, "grad_norm": 29.073741912841797, "learning_rate": 1.194847979251979e-07, "logits/chosen": -0.9419503808021545, "logits/rejected": -0.9568164348602295, "logps/chosen": -745.856689453125, "logps/ref_chosen": -63.32981872558594, "logps/ref_rejected": -95.78697204589844, "logps/rejected": -1279.4580078125, "loss": 1.0585, "margin_dpo/margin_mean": 501.14410400390625, "margin_dpo/margin_std": 632.0308227539062, "step": 483 }, { "KL/chosen_KL_mean": -616.1393432617188, "KL/mean": -855.525634765625, "KL/rejected_KL_mean": -1094.911865234375, "KL/std": 583.80126953125, "epoch": 0.71071953010279, "fcm_dpo/beta": 0.0008886742289178073, "fcm_dpo/delta": -0.07883388549089432, "fcm_dpo/margin": 478.7725830078125, "fcm_dpo/q_t": 0.40235435962677, "grad_norm": 42.01279067993164, "learning_rate": 1.1839195928066101e-07, "logits/chosen": -0.9948852062225342, "logits/rejected": -1.0258457660675049, "logps/chosen": -675.2774658203125, "logps/ref_chosen": -59.13812255859375, "logps/ref_rejected": -84.37144470214844, "logps/rejected": -1179.2833251953125, "loss": 1.0683, "margin_dpo/margin_mean": 478.7725830078125, "margin_dpo/margin_std": 582.26611328125, "step": 484 }, { "KL/chosen_KL_mean": -638.5044555664062, "KL/mean": -869.9237060546875, "KL/rejected_KL_mean": -1101.343017578125, "KL/std": 587.0706176757812, "epoch": 0.7121879588839941, "fcm_dpo/beta": 0.0008760282071307302, "fcm_dpo/delta": -0.05777723342180252, "fcm_dpo/margin": 462.83856201171875, "fcm_dpo/q_t": 0.40835797786712646, "grad_norm": 32.41249084472656, "learning_rate": 1.1730258863039347e-07, "logits/chosen": -0.9031185507774353, "logits/rejected": -0.9280800223350525, "logps/chosen": -697.35400390625, "logps/ref_chosen": -58.849571228027344, "logps/ref_rejected": -103.36408233642578, "logps/rejected": -1204.70703125, "loss": 1.0999, "margin_dpo/margin_mean": 462.83856201171875, "margin_dpo/margin_std": 665.9918823242188, "step": 485 }, { "KL/chosen_KL_mean": -735.9765625, "KL/mean": -987.707763671875, "KL/rejected_KL_mean": -1239.43896484375, "KL/std": 638.144287109375, "epoch": 0.7136563876651982, "fcm_dpo/beta": 0.0008581732981838286, "fcm_dpo/delta": -0.0861148089170456, "fcm_dpo/margin": 503.46240234375, "fcm_dpo/q_t": 0.40256860852241516, "grad_norm": 36.858909606933594, "learning_rate": 1.1621671468032493e-07, "logits/chosen": -1.0510118007659912, "logits/rejected": -1.0759706497192383, "logps/chosen": -791.2362060546875, "logps/ref_chosen": -55.25966262817383, "logps/ref_rejected": -92.13936614990234, "logps/rejected": -1331.578369140625, "loss": 1.1, "margin_dpo/margin_mean": 503.4623718261719, "margin_dpo/margin_std": 770.2353515625, "step": 486 }, { "KL/chosen_KL_mean": -800.8622436523438, "KL/mean": -982.6766357421875, "KL/rejected_KL_mean": -1164.4910888671875, "KL/std": 589.8531494140625, "epoch": 0.7151248164464024, "fcm_dpo/beta": 0.0008646865608170629, "fcm_dpo/delta": 0.03625814616680145, "fcm_dpo/margin": 363.62884521484375, "fcm_dpo/q_t": 0.42635592818260193, "grad_norm": 31.64628791809082, "learning_rate": 1.1513436604424378e-07, "logits/chosen": -1.0411834716796875, "logits/rejected": -1.0501723289489746, "logps/chosen": -853.925537109375, "logps/ref_chosen": -53.06330871582031, "logps/ref_rejected": -92.41883087158203, "logps/rejected": -1256.909912109375, "loss": 1.1617, "margin_dpo/margin_mean": 363.62884521484375, "margin_dpo/margin_std": 604.9837646484375, "step": 487 }, { "KL/chosen_KL_mean": -663.748046875, "KL/mean": -853.4315185546875, "KL/rejected_KL_mean": -1043.114990234375, "KL/std": 536.4071044921875, "epoch": 0.7165932452276065, "fcm_dpo/beta": 0.0008684194181114435, "fcm_dpo/delta": 0.020952310413122177, "fcm_dpo/margin": 379.3669738769531, "fcm_dpo/q_t": 0.423104465007782, "grad_norm": 31.610628128051758, "learning_rate": 1.1405557124304335e-07, "logits/chosen": -0.9978197813034058, "logits/rejected": -1.009468913078308, "logps/chosen": -715.9761962890625, "logps/ref_chosen": -52.22815704345703, "logps/ref_rejected": -84.00656127929688, "logps/rejected": -1127.12158203125, "loss": 1.1324, "margin_dpo/margin_mean": 379.36700439453125, "margin_dpo/margin_std": 518.4456787109375, "step": 488 }, { "KL/chosen_KL_mean": -616.62109375, "KL/mean": -808.0659790039062, "KL/rejected_KL_mean": -999.5108642578125, "KL/std": 516.0118408203125, "epoch": 0.7180616740088106, "fcm_dpo/beta": 0.0008714280556887388, "fcm_dpo/delta": 0.016632717102766037, "fcm_dpo/margin": 382.88970947265625, "fcm_dpo/q_t": 0.4236697554588318, "grad_norm": 24.073930740356445, "learning_rate": 1.1298035870396985e-07, "logits/chosen": -0.9905928373336792, "logits/rejected": -0.991652250289917, "logps/chosen": -672.6107177734375, "logps/ref_chosen": -55.989627838134766, "logps/ref_rejected": -79.39812469482422, "logps/rejected": -1078.908935546875, "loss": 1.1387, "margin_dpo/margin_mean": 382.8896484375, "margin_dpo/margin_std": 572.0997314453125, "step": 489 }, { "KL/chosen_KL_mean": -702.3668823242188, "KL/mean": -895.4848022460938, "KL/rejected_KL_mean": -1088.602783203125, "KL/std": 633.2512817382812, "epoch": 0.7195301027900147, "fcm_dpo/beta": 0.0008734596194699407, "fcm_dpo/delta": 0.01288134977221489, "fcm_dpo/margin": 386.2358703613281, "fcm_dpo/q_t": 0.4240713119506836, "grad_norm": 39.06137466430664, "learning_rate": 1.1190875675987355e-07, "logits/chosen": -1.0082063674926758, "logits/rejected": -1.0535714626312256, "logps/chosen": -754.7332763671875, "logps/ref_chosen": -52.36639404296875, "logps/ref_rejected": -110.4090576171875, "logps/rejected": -1199.0118408203125, "loss": 1.1739, "margin_dpo/margin_mean": 386.2358703613281, "margin_dpo/margin_std": 737.7401123046875, "step": 490 }, { "KL/chosen_KL_mean": -649.2156982421875, "KL/mean": -784.3421630859375, "KL/rejected_KL_mean": -919.4685668945312, "KL/std": 507.2773132324219, "epoch": 0.7209985315712188, "fcm_dpo/beta": 0.0008869500597938895, "fcm_dpo/delta": 0.11293124407529831, "fcm_dpo/margin": 270.2528991699219, "fcm_dpo/q_t": 0.44444432854652405, "grad_norm": 29.092905044555664, "learning_rate": 1.1084079364846241e-07, "logits/chosen": -0.9913074970245361, "logits/rejected": -0.9897642135620117, "logps/chosen": -709.3319702148438, "logps/ref_chosen": -60.11626434326172, "logps/ref_rejected": -73.27278900146484, "logps/rejected": -992.7413330078125, "loss": 1.2104, "margin_dpo/margin_mean": 270.2528991699219, "margin_dpo/margin_std": 513.86279296875, "step": 491 }, { "KL/chosen_KL_mean": -699.1614990234375, "KL/mean": -827.0299072265625, "KL/rejected_KL_mean": -954.8983154296875, "KL/std": 548.1110229492188, "epoch": 0.7224669603524229, "fcm_dpo/beta": 0.0009072008542716503, "fcm_dpo/delta": 0.12086275219917297, "fcm_dpo/margin": 255.7367401123047, "fcm_dpo/q_t": 0.44698652625083923, "grad_norm": 44.75581359863281, "learning_rate": 1.097764975115576e-07, "logits/chosen": -1.0495141744613647, "logits/rejected": -1.0257760286331177, "logps/chosen": -753.1557006835938, "logps/ref_chosen": -53.994178771972656, "logps/ref_rejected": -72.65962219238281, "logps/rejected": -1027.557861328125, "loss": 1.2431, "margin_dpo/margin_mean": 255.73672485351562, "margin_dpo/margin_std": 618.5252685546875, "step": 492 }, { "KL/chosen_KL_mean": -726.62841796875, "KL/mean": -883.1748046875, "KL/rejected_KL_mean": -1039.721435546875, "KL/std": 605.7178955078125, "epoch": 0.723935389133627, "fcm_dpo/beta": 0.0009097631555050611, "fcm_dpo/delta": -0.019333893433213234, "fcm_dpo/margin": 313.09295654296875, "fcm_dpo/q_t": 0.4324401617050171, "grad_norm": 32.448036193847656, "learning_rate": 1.0871589639435203e-07, "logits/chosen": -1.0681705474853516, "logits/rejected": -1.0370638370513916, "logps/chosen": -802.1256103515625, "logps/ref_chosen": -75.49723815917969, "logps/ref_rejected": -87.32301330566406, "logps/rejected": -1127.04443359375, "loss": 1.187, "margin_dpo/margin_mean": 313.09295654296875, "margin_dpo/margin_std": 570.6629638671875, "step": 493 }, { "KL/chosen_KL_mean": -579.3143310546875, "KL/mean": -823.8994140625, "KL/rejected_KL_mean": -1068.484619140625, "KL/std": 546.38134765625, "epoch": 0.7254038179148311, "fcm_dpo/beta": 0.0008967167232185602, "fcm_dpo/delta": -0.09278169274330139, "fcm_dpo/margin": 489.1702575683594, "fcm_dpo/q_t": 0.3974645435810089, "grad_norm": 38.209651947021484, "learning_rate": 1.0765901824467166e-07, "logits/chosen": -0.9156688451766968, "logits/rejected": -0.9544728994369507, "logps/chosen": -620.673583984375, "logps/ref_chosen": -41.35926818847656, "logps/ref_rejected": -86.09136962890625, "logps/rejected": -1154.575927734375, "loss": 1.0522, "margin_dpo/margin_mean": 489.1702880859375, "margin_dpo/margin_std": 552.1724853515625, "step": 494 }, { "KL/chosen_KL_mean": -666.56298828125, "KL/mean": -872.1007080078125, "KL/rejected_KL_mean": -1077.638427734375, "KL/std": 544.0244140625, "epoch": 0.7268722466960352, "fcm_dpo/beta": 0.0008921578410081565, "fcm_dpo/delta": -0.017406316474080086, "fcm_dpo/margin": 411.0754699707031, "fcm_dpo/q_t": 0.4172486662864685, "grad_norm": 37.34906768798828, "learning_rate": 1.0660589091223854e-07, "logits/chosen": -1.006931185722351, "logits/rejected": -1.0146992206573486, "logps/chosen": -730.0980224609375, "logps/ref_chosen": -63.53507995605469, "logps/ref_rejected": -91.42443084716797, "logps/rejected": -1169.06298828125, "loss": 1.1308, "margin_dpo/margin_mean": 411.07550048828125, "margin_dpo/margin_std": 657.4510498046875, "step": 495 }, { "KL/chosen_KL_mean": -814.9395751953125, "KL/mean": -919.1443481445312, "KL/rejected_KL_mean": -1023.34912109375, "KL/std": 457.9229736328125, "epoch": 0.7283406754772394, "fcm_dpo/beta": 0.0008899334352463484, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 208.4095458984375, "fcm_dpo/q_t": 0.45584726333618164, "grad_norm": 70.3835678100586, "learning_rate": 1.0555654214793722e-07, "logits/chosen": -0.9732378721237183, "logits/rejected": -0.9386388063430786, "logps/chosen": -887.531494140625, "logps/ref_chosen": -72.5919189453125, "logps/ref_rejected": -84.32933807373047, "logps/rejected": -1107.678466796875, "loss": 1.2632, "margin_dpo/margin_mean": 208.4095458984375, "margin_dpo/margin_std": 522.48583984375, "step": 496 }, { "KL/chosen_KL_mean": -753.662353515625, "KL/mean": -864.582763671875, "KL/rejected_KL_mean": -975.5030517578125, "KL/std": 551.1240844726562, "epoch": 0.7298091042584435, "fcm_dpo/beta": 0.0008933351491577923, "fcm_dpo/delta": 0.01907578855752945, "fcm_dpo/margin": 221.8406524658203, "fcm_dpo/q_t": 0.453231543302536, "grad_norm": 33.05242919921875, "learning_rate": 1.0451099960308374e-07, "logits/chosen": -0.9515343904495239, "logits/rejected": -0.9357314109802246, "logps/chosen": -812.25634765625, "logps/ref_chosen": -58.59397506713867, "logps/ref_rejected": -76.28836822509766, "logps/rejected": -1051.7913818359375, "loss": 1.2493, "margin_dpo/margin_mean": 221.84066772460938, "margin_dpo/margin_std": 499.05657958984375, "step": 497 }, { "KL/chosen_KL_mean": -736.4669189453125, "KL/mean": -919.92578125, "KL/rejected_KL_mean": -1103.384521484375, "KL/std": 573.03173828125, "epoch": 0.7312775330396476, "fcm_dpo/beta": 0.0008958862163126469, "fcm_dpo/delta": 0.02201257273554802, "fcm_dpo/margin": 366.917724609375, "fcm_dpo/q_t": 0.4242166578769684, "grad_norm": 28.690711975097656, "learning_rate": 1.0346929082869641e-07, "logits/chosen": -0.9741387367248535, "logits/rejected": -0.9623622894287109, "logps/chosen": -807.672607421875, "logps/ref_chosen": -71.20565795898438, "logps/ref_rejected": -83.95803833007812, "logps/rejected": -1187.3426513671875, "loss": 1.1655, "margin_dpo/margin_mean": 366.917724609375, "margin_dpo/margin_std": 657.1267700195312, "step": 498 }, { "KL/chosen_KL_mean": -616.2876586914062, "KL/mean": -847.274169921875, "KL/rejected_KL_mean": -1078.2607421875, "KL/std": 576.01513671875, "epoch": 0.7327459618208517, "fcm_dpo/beta": 0.0008876968640834093, "fcm_dpo/delta": -0.06278754025697708, "fcm_dpo/margin": 461.9729309082031, "fcm_dpo/q_t": 0.4059438109397888, "grad_norm": 41.73751449584961, "learning_rate": 1.0243144327477013e-07, "logits/chosen": -0.9784067869186401, "logits/rejected": -1.0155872106552124, "logps/chosen": -667.5428466796875, "logps/ref_chosen": -51.25519561767578, "logps/ref_rejected": -101.07870483398438, "logps/rejected": -1179.33935546875, "loss": 1.0909, "margin_dpo/margin_mean": 461.972900390625, "margin_dpo/margin_std": 637.4395751953125, "step": 499 }, { "KL/chosen_KL_mean": -716.6522216796875, "KL/mean": -908.4356689453125, "KL/rejected_KL_mean": -1100.218994140625, "KL/std": 504.1438293457031, "epoch": 0.7342143906020558, "fcm_dpo/beta": 0.0008866304415278137, "fcm_dpo/delta": 0.010259722359478474, "fcm_dpo/margin": 383.56671142578125, "fcm_dpo/q_t": 0.42117249965667725, "grad_norm": 27.376264572143555, "learning_rate": 1.0139748428955333e-07, "logits/chosen": -0.9571424722671509, "logits/rejected": -0.9902607202529907, "logps/chosen": -773.6796875, "logps/ref_chosen": -57.027442932128906, "logps/ref_rejected": -93.93421173095703, "logps/rejected": -1194.1531982421875, "loss": 1.1487, "margin_dpo/margin_mean": 383.56671142578125, "margin_dpo/margin_std": 624.0594482421875, "step": 500 }, { "KL/chosen_KL_mean": -627.3432006835938, "KL/mean": -835.041259765625, "KL/rejected_KL_mean": -1042.7393798828125, "KL/std": 518.466796875, "epoch": 0.73568281938326, "fcm_dpo/beta": 0.0008866213611327112, "fcm_dpo/delta": -0.019011791795492172, "fcm_dpo/margin": 415.3961486816406, "fcm_dpo/q_t": 0.415674090385437, "grad_norm": 28.41216278076172, "learning_rate": 1.0036744111882672e-07, "logits/chosen": -0.8981431722640991, "logits/rejected": -0.88893723487854, "logps/chosen": -681.7027587890625, "logps/ref_chosen": -54.359527587890625, "logps/ref_rejected": -80.15670013427734, "logps/rejected": -1122.89599609375, "loss": 1.1279, "margin_dpo/margin_mean": 415.3961486816406, "margin_dpo/margin_std": 643.92919921875, "step": 501 }, { "KL/chosen_KL_mean": -574.186279296875, "KL/mean": -764.1837158203125, "KL/rejected_KL_mean": -954.1812133789062, "KL/std": 467.6396484375, "epoch": 0.737151248164464, "fcm_dpo/beta": 0.0008872783510014415, "fcm_dpo/delta": 0.013246024027466774, "fcm_dpo/margin": 379.9948425292969, "fcm_dpo/q_t": 0.4216458797454834, "grad_norm": 25.857147216796875, "learning_rate": 9.934134090518592e-08, "logits/chosen": -0.854095995426178, "logits/rejected": -0.8424793481826782, "logps/chosen": -641.786865234375, "logps/ref_chosen": -67.60050964355469, "logps/ref_rejected": -82.94876098632812, "logps/rejected": -1037.1300048828125, "loss": 1.1233, "margin_dpo/margin_mean": 379.9948425292969, "margin_dpo/margin_std": 498.8470458984375, "step": 502 }, { "KL/chosen_KL_mean": -585.6944580078125, "KL/mean": -766.4293212890625, "KL/rejected_KL_mean": -947.1641845703125, "KL/std": 479.78033447265625, "epoch": 0.7386196769456681, "fcm_dpo/beta": 0.0008887014118954539, "fcm_dpo/delta": 0.029671432450413704, "fcm_dpo/margin": 361.46978759765625, "fcm_dpo/q_t": 0.42668020725250244, "grad_norm": 25.81717872619629, "learning_rate": 9.831921068732571e-08, "logits/chosen": -0.8250192403793335, "logits/rejected": -0.8123034834861755, "logps/chosen": -640.7728271484375, "logps/ref_chosen": -55.078407287597656, "logps/ref_rejected": -82.50544738769531, "logps/rejected": -1029.669677734375, "loss": 1.1462, "margin_dpo/margin_mean": 361.46978759765625, "margin_dpo/margin_std": 550.151611328125, "step": 503 }, { "KL/chosen_KL_mean": -624.8487548828125, "KL/mean": -847.6356201171875, "KL/rejected_KL_mean": -1070.4224853515625, "KL/std": 563.498779296875, "epoch": 0.7400881057268722, "fcm_dpo/beta": 0.0008875064086169004, "fcm_dpo/delta": -0.047302231192588806, "fcm_dpo/margin": 445.57373046875, "fcm_dpo/q_t": 0.409305602312088, "grad_norm": 32.672576904296875, "learning_rate": 9.730107739932805e-08, "logits/chosen": -0.8826575875282288, "logits/rejected": -0.9081119298934937, "logps/chosen": -684.8145141601562, "logps/ref_chosen": -59.96575164794922, "logps/ref_rejected": -103.76212310791016, "logps/rejected": -1174.1845703125, "loss": 1.1008, "margin_dpo/margin_mean": 445.57373046875, "margin_dpo/margin_std": 609.791015625, "step": 504 }, { "KL/chosen_KL_mean": -699.5025024414062, "KL/mean": -810.5694580078125, "KL/rejected_KL_mean": -921.6364135742188, "KL/std": 508.30511474609375, "epoch": 0.7415565345080763, "fcm_dpo/beta": 0.0008896557847037911, "fcm_dpo/delta": 0.06633193790912628, "fcm_dpo/margin": 222.13388061523438, "fcm_dpo/q_t": 0.45340317487716675, "grad_norm": 50.9668083190918, "learning_rate": 9.628696786995188e-08, "logits/chosen": -0.9198693037033081, "logits/rejected": -0.8950393199920654, "logps/chosen": -775.6574096679688, "logps/ref_chosen": -76.1549072265625, "logps/ref_rejected": -88.58537292480469, "logps/rejected": -1010.2218017578125, "loss": 1.2443, "margin_dpo/margin_mean": 222.13388061523438, "margin_dpo/margin_std": 481.71832275390625, "step": 505 }, { "KL/chosen_KL_mean": -559.4972534179688, "KL/mean": -758.4755859375, "KL/rejected_KL_mean": -957.4539794921875, "KL/std": 527.7215576171875, "epoch": 0.7430249632892805, "fcm_dpo/beta": 0.0008918484672904015, "fcm_dpo/delta": -0.00534551776945591, "fcm_dpo/margin": 397.9566650390625, "fcm_dpo/q_t": 0.4175879955291748, "grad_norm": 34.8582878112793, "learning_rate": 9.527690882192635e-08, "logits/chosen": -0.8557130098342896, "logits/rejected": -0.8738311529159546, "logps/chosen": -608.457763671875, "logps/ref_chosen": -48.96050262451172, "logps/ref_rejected": -78.41505432128906, "logps/rejected": -1035.8690185546875, "loss": 1.1269, "margin_dpo/margin_mean": 397.9566650390625, "margin_dpo/margin_std": 581.9273681640625, "step": 506 }, { "KL/chosen_KL_mean": -641.53173828125, "KL/mean": -815.253662109375, "KL/rejected_KL_mean": -988.9755859375, "KL/std": 610.8743896484375, "epoch": 0.7444933920704846, "fcm_dpo/beta": 0.0008992149960249662, "fcm_dpo/delta": 0.03880162909626961, "fcm_dpo/margin": 347.44384765625, "fcm_dpo/q_t": 0.4313592314720154, "grad_norm": 27.87244987487793, "learning_rate": 9.427092687124691e-08, "logits/chosen": -0.8712353706359863, "logits/rejected": -0.8783326148986816, "logps/chosen": -708.333251953125, "logps/ref_chosen": -66.80149841308594, "logps/ref_rejected": -95.37289428710938, "logps/rejected": -1084.3485107421875, "loss": 1.181, "margin_dpo/margin_mean": 347.44384765625, "margin_dpo/margin_std": 672.9422607421875, "step": 507 }, { "KL/chosen_KL_mean": -666.12744140625, "KL/mean": -816.7818603515625, "KL/rejected_KL_mean": -967.436279296875, "KL/std": 569.1419677734375, "epoch": 0.7459618208516887, "fcm_dpo/beta": 0.0009134809952229261, "fcm_dpo/delta": 0.0766909047961235, "fcm_dpo/margin": 301.30889892578125, "fcm_dpo/q_t": 0.43825656175613403, "grad_norm": 29.6184139251709, "learning_rate": 9.326904852647344e-08, "logits/chosen": -0.8354727029800415, "logits/rejected": -0.8378840684890747, "logps/chosen": -737.430908203125, "logps/ref_chosen": -71.303466796875, "logps/ref_rejected": -95.6275405883789, "logps/rejected": -1063.0638427734375, "loss": 1.2145, "margin_dpo/margin_mean": 301.308837890625, "margin_dpo/margin_std": 656.0669555664062, "step": 508 }, { "KL/chosen_KL_mean": -506.78466796875, "KL/mean": -663.2677001953125, "KL/rejected_KL_mean": -819.750732421875, "KL/std": 414.9226989746094, "epoch": 0.7474302496328928, "fcm_dpo/beta": 0.0009152308921329677, "fcm_dpo/delta": -0.002988607855513692, "fcm_dpo/margin": 312.9660949707031, "fcm_dpo/q_t": 0.43361881375312805, "grad_norm": 33.5871467590332, "learning_rate": 9.227130018803195e-08, "logits/chosen": -0.765209972858429, "logits/rejected": -0.762215256690979, "logps/chosen": -570.6036376953125, "logps/ref_chosen": -63.81895065307617, "logps/ref_rejected": -83.25643920898438, "logps/rejected": -903.0072021484375, "loss": 1.1739, "margin_dpo/margin_mean": 312.966064453125, "margin_dpo/margin_std": 512.6011962890625, "step": 509 }, { "KL/chosen_KL_mean": -638.8026123046875, "KL/mean": -840.986328125, "KL/rejected_KL_mean": -1043.169921875, "KL/std": 484.48175048828125, "epoch": 0.748898678414097, "fcm_dpo/beta": 0.0009141661575995386, "fcm_dpo/delta": -0.02046576887369156, "fcm_dpo/margin": 404.3673095703125, "fcm_dpo/q_t": 0.4124307632446289, "grad_norm": 39.50413513183594, "learning_rate": 9.127770814751932e-08, "logits/chosen": -0.783023476600647, "logits/rejected": -0.8079743385314941, "logps/chosen": -690.6810913085938, "logps/ref_chosen": -51.878448486328125, "logps/ref_rejected": -102.7651596069336, "logps/rejected": -1145.93505859375, "loss": 1.095, "margin_dpo/margin_mean": 404.3673095703125, "margin_dpo/margin_std": 470.9833984375, "step": 510 }, { "KL/chosen_KL_mean": -606.0228271484375, "KL/mean": -764.6362915039062, "KL/rejected_KL_mean": -923.249755859375, "KL/std": 534.836669921875, "epoch": 0.750367107195301, "fcm_dpo/beta": 0.0009196955943480134, "fcm_dpo/delta": 0.05997824668884277, "fcm_dpo/margin": 317.2269287109375, "fcm_dpo/q_t": 0.433167427778244, "grad_norm": 34.168601989746094, "learning_rate": 9.028829858700973e-08, "logits/chosen": -0.8136961460113525, "logits/rejected": -0.8219088315963745, "logps/chosen": -666.260986328125, "logps/ref_chosen": -60.23811721801758, "logps/ref_rejected": -92.85676574707031, "logps/rejected": -1016.1065063476562, "loss": 1.1989, "margin_dpo/margin_mean": 317.2269287109375, "margin_dpo/margin_std": 654.623046875, "step": 511 }, { "KL/chosen_KL_mean": -481.8533935546875, "KL/mean": -713.4747314453125, "KL/rejected_KL_mean": -945.095947265625, "KL/std": 487.7607116699219, "epoch": 0.7518355359765051, "fcm_dpo/beta": 0.0009103405755013227, "fcm_dpo/delta": -0.07506463676691055, "fcm_dpo/margin": 463.2425537109375, "fcm_dpo/q_t": 0.4011862874031067, "grad_norm": 54.72198486328125, "learning_rate": 8.930309757836516e-08, "logits/chosen": -0.812955915927887, "logits/rejected": -0.8373688459396362, "logps/chosen": -536.7589111328125, "logps/ref_chosen": -54.905494689941406, "logps/ref_rejected": -81.87586975097656, "logps/rejected": -1026.9718017578125, "loss": 1.0568, "margin_dpo/margin_mean": 463.2425537109375, "margin_dpo/margin_std": 502.77996826171875, "step": 512 }, { "KL/chosen_KL_mean": -666.6439819335938, "KL/mean": -828.3573608398438, "KL/rejected_KL_mean": -990.07080078125, "KL/std": 492.90380859375, "epoch": 0.7533039647577092, "fcm_dpo/beta": 0.0009010280482470989, "fcm_dpo/delta": -0.041153233498334885, "fcm_dpo/margin": 323.4268493652344, "fcm_dpo/q_t": 0.43294602632522583, "grad_norm": 58.28951644897461, "learning_rate": 8.832213108254863e-08, "logits/chosen": -0.8997926712036133, "logits/rejected": -0.8882004022598267, "logps/chosen": -731.5604248046875, "logps/ref_chosen": -64.91644287109375, "logps/ref_rejected": -76.06245422363281, "logps/rejected": -1066.13330078125, "loss": 1.1844, "margin_dpo/margin_mean": 323.4268493652344, "margin_dpo/margin_std": 577.8350830078125, "step": 513 }, { "KL/chosen_KL_mean": -695.1303100585938, "KL/mean": -851.9722900390625, "KL/rejected_KL_mean": -1008.8141479492188, "KL/std": 538.831298828125, "epoch": 0.7547723935389133, "fcm_dpo/beta": 0.0009115079883486032, "fcm_dpo/delta": 0.06578437238931656, "fcm_dpo/margin": 313.683837890625, "fcm_dpo/q_t": 0.4355640113353729, "grad_norm": 39.681461334228516, "learning_rate": 8.734542494893954e-08, "logits/chosen": -0.8688774108886719, "logits/rejected": -0.8621048927307129, "logps/chosen": -769.35986328125, "logps/ref_chosen": -74.22957611083984, "logps/ref_rejected": -78.945556640625, "logps/rejected": -1087.759765625, "loss": 1.1896, "margin_dpo/margin_mean": 313.683837890625, "margin_dpo/margin_std": 597.1491088867188, "step": 514 }, { "KL/chosen_KL_mean": -576.4310302734375, "KL/mean": -700.958251953125, "KL/rejected_KL_mean": -825.4854736328125, "KL/std": 449.62347412109375, "epoch": 0.7562408223201175, "fcm_dpo/beta": 0.0009299020748585463, "fcm_dpo/delta": 0.12117768824100494, "fcm_dpo/margin": 249.0543975830078, "fcm_dpo/q_t": 0.444853812456131, "grad_norm": 43.30710983276367, "learning_rate": 8.637300491465272e-08, "logits/chosen": -0.8548616170883179, "logits/rejected": -0.8692120313644409, "logps/chosen": -626.8326416015625, "logps/ref_chosen": -50.40156555175781, "logps/ref_rejected": -87.09774780273438, "logps/rejected": -912.5831909179688, "loss": 1.2272, "margin_dpo/margin_mean": 249.05441284179688, "margin_dpo/margin_std": 538.2990112304688, "step": 515 }, { "KL/chosen_KL_mean": -610.23974609375, "KL/mean": -794.4168090820312, "KL/rejected_KL_mean": -978.5938110351562, "KL/std": 516.6522216796875, "epoch": 0.7577092511013216, "fcm_dpo/beta": 0.000938748475164175, "fcm_dpo/delta": 0.0041639544069767, "fcm_dpo/margin": 368.35406494140625, "fcm_dpo/q_t": 0.41812512278556824, "grad_norm": 38.28722381591797, "learning_rate": 8.540489660386064e-08, "logits/chosen": -0.9362125396728516, "logits/rejected": -0.9661528468132019, "logps/chosen": -674.8893432617188, "logps/ref_chosen": -64.64956665039062, "logps/ref_rejected": -111.72237396240234, "logps/rejected": -1090.316162109375, "loss": 1.128, "margin_dpo/margin_mean": 368.3540344238281, "margin_dpo/margin_std": 521.5465698242188, "step": 516 }, { "KL/chosen_KL_mean": -608.512451171875, "KL/mean": -828.2478637695312, "KL/rejected_KL_mean": -1047.983154296875, "KL/std": 547.919677734375, "epoch": 0.7591776798825257, "fcm_dpo/beta": 0.0009262575185857713, "fcm_dpo/delta": -0.05964816361665726, "fcm_dpo/margin": 439.4707946777344, "fcm_dpo/q_t": 0.4080393314361572, "grad_norm": 26.922121047973633, "learning_rate": 8.444112552711752e-08, "logits/chosen": -0.8890372514724731, "logits/rejected": -0.8907492160797119, "logps/chosen": -669.426025390625, "logps/ref_chosen": -60.913551330566406, "logps/ref_rejected": -89.08308410644531, "logps/rejected": -1137.0662841796875, "loss": 1.0943, "margin_dpo/margin_mean": 439.4707946777344, "margin_dpo/margin_std": 619.3283081054688, "step": 517 }, { "KL/chosen_KL_mean": -538.5008544921875, "KL/mean": -726.7237548828125, "KL/rejected_KL_mean": -914.9466552734375, "KL/std": 435.29779052734375, "epoch": 0.7606461086637298, "fcm_dpo/beta": 0.0009226626716554165, "fcm_dpo/delta": 0.00259285606443882, "fcm_dpo/margin": 376.4457702636719, "fcm_dpo/q_t": 0.4179810881614685, "grad_norm": 46.04099655151367, "learning_rate": 8.348171708068747e-08, "logits/chosen": -0.8912504315376282, "logits/rejected": -0.9100457429885864, "logps/chosen": -595.956787109375, "logps/ref_chosen": -57.45589065551758, "logps/ref_rejected": -85.31269836425781, "logps/rejected": -1000.2593383789062, "loss": 1.1172, "margin_dpo/margin_mean": 376.4457702636719, "margin_dpo/margin_std": 483.531005859375, "step": 518 }, { "KL/chosen_KL_mean": -560.754150390625, "KL/mean": -688.62060546875, "KL/rejected_KL_mean": -816.4871215820312, "KL/std": 400.0428466796875, "epoch": 0.762114537444934, "fcm_dpo/beta": 0.0009278228390030563, "fcm_dpo/delta": 0.02487240917980671, "fcm_dpo/margin": 255.73304748535156, "fcm_dpo/q_t": 0.44359180331230164, "grad_norm": 25.208648681640625, "learning_rate": 8.25266965458755e-08, "logits/chosen": -0.8528093695640564, "logits/rejected": -0.8358422517776489, "logps/chosen": -634.8174438476562, "logps/ref_chosen": -74.06331634521484, "logps/ref_rejected": -104.44416809082031, "logps/rejected": -920.9312744140625, "loss": 1.2158, "margin_dpo/margin_mean": 255.7330322265625, "margin_dpo/margin_std": 497.2356872558594, "step": 519 }, { "KL/chosen_KL_mean": -596.6747436523438, "KL/mean": -771.2025146484375, "KL/rejected_KL_mean": -945.730224609375, "KL/std": 479.77325439453125, "epoch": 0.7635829662261381, "fcm_dpo/beta": 0.0009306604042649269, "fcm_dpo/delta": 0.025856276974081993, "fcm_dpo/margin": 349.0555114746094, "fcm_dpo/q_t": 0.42620331048965454, "grad_norm": 33.09168243408203, "learning_rate": 8.15760890883607e-08, "logits/chosen": -0.8668011426925659, "logits/rejected": -0.8748083114624023, "logps/chosen": -666.974609375, "logps/ref_chosen": -70.2998275756836, "logps/ref_rejected": -99.98133850097656, "logps/rejected": -1045.7115478515625, "loss": 1.151, "margin_dpo/margin_mean": 349.05548095703125, "margin_dpo/margin_std": 541.3212890625, "step": 520 }, { "KL/chosen_KL_mean": -511.3445739746094, "KL/mean": -700.4781494140625, "KL/rejected_KL_mean": -889.61181640625, "KL/std": 478.0359802246094, "epoch": 0.7650513950073421, "fcm_dpo/beta": 0.0009390891063958406, "fcm_dpo/delta": -0.0059568993747234344, "fcm_dpo/margin": 378.2672119140625, "fcm_dpo/q_t": 0.41709330677986145, "grad_norm": 33.98050308227539, "learning_rate": 8.062991975753378e-08, "logits/chosen": -0.8779432773590088, "logits/rejected": -0.8840290904045105, "logps/chosen": -569.4874877929688, "logps/ref_chosen": -58.14292526245117, "logps/ref_rejected": -83.28060913085938, "logps/rejected": -972.8924560546875, "loss": 1.1127, "margin_dpo/margin_mean": 378.2672119140625, "margin_dpo/margin_std": 479.72076416015625, "step": 521 }, { "KL/chosen_KL_mean": -599.886962890625, "KL/mean": -763.5291748046875, "KL/rejected_KL_mean": -927.1713256835938, "KL/std": 497.3731689453125, "epoch": 0.7665198237885462, "fcm_dpo/beta": 0.0009402185096405447, "fcm_dpo/delta": 0.043608807027339935, "fcm_dpo/margin": 327.2843017578125, "fcm_dpo/q_t": 0.42860668897628784, "grad_norm": 30.430320739746094, "learning_rate": 7.968821348583643e-08, "logits/chosen": -0.8880289793014526, "logits/rejected": -0.8938380479812622, "logps/chosen": -646.4346923828125, "logps/ref_chosen": -46.54766845703125, "logps/ref_rejected": -66.01388549804688, "logps/rejected": -993.1851806640625, "loss": 1.1638, "margin_dpo/margin_mean": 327.2843017578125, "margin_dpo/margin_std": 539.9135131835938, "step": 522 }, { "KL/chosen_KL_mean": -660.646240234375, "KL/mean": -847.5511474609375, "KL/rejected_KL_mean": -1034.4560546875, "KL/std": 619.6279296875, "epoch": 0.7679882525697503, "fcm_dpo/beta": 0.0009420792339369655, "fcm_dpo/delta": -0.002239507157355547, "fcm_dpo/margin": 373.80975341796875, "fcm_dpo/q_t": 0.4196031093597412, "grad_norm": 37.910400390625, "learning_rate": 7.875099508810484e-08, "logits/chosen": -0.9370777606964111, "logits/rejected": -0.9447219371795654, "logps/chosen": -722.4158935546875, "logps/ref_chosen": -61.76960372924805, "logps/ref_rejected": -83.76141357421875, "logps/rejected": -1118.217529296875, "loss": 1.1449, "margin_dpo/margin_mean": 373.80975341796875, "margin_dpo/margin_std": 621.3948364257812, "step": 523 }, { "KL/chosen_KL_mean": -649.7862548828125, "KL/mean": -849.8914794921875, "KL/rejected_KL_mean": -1049.99658203125, "KL/std": 567.035888671875, "epoch": 0.7694566813509545, "fcm_dpo/beta": 0.0009336816146969795, "fcm_dpo/delta": -0.02515629678964615, "fcm_dpo/margin": 400.2103271484375, "fcm_dpo/q_t": 0.4117085337638855, "grad_norm": 40.177085876464844, "learning_rate": 7.781828926091535e-08, "logits/chosen": -0.9887799024581909, "logits/rejected": -0.9843355417251587, "logps/chosen": -727.8583374023438, "logps/ref_chosen": -78.0720443725586, "logps/ref_rejected": -81.30198669433594, "logps/rejected": -1131.298583984375, "loss": 1.1197, "margin_dpo/margin_mean": 400.2103576660156, "margin_dpo/margin_std": 576.9923095703125, "step": 524 }, { "KL/chosen_KL_mean": -653.83642578125, "KL/mean": -902.3298950195312, "KL/rejected_KL_mean": -1150.8232421875, "KL/std": 586.9584350585938, "epoch": 0.7709251101321586, "fcm_dpo/beta": 0.0009176610037684441, "fcm_dpo/delta": -0.11192645132541656, "fcm_dpo/margin": 496.9869384765625, "fcm_dpo/q_t": 0.3969133496284485, "grad_norm": 28.977869033813477, "learning_rate": 7.689012058193384e-08, "logits/chosen": -0.9288259744644165, "logits/rejected": -0.9718555212020874, "logps/chosen": -704.664306640625, "logps/ref_chosen": -50.827857971191406, "logps/ref_rejected": -100.05294036865234, "logps/rejected": -1250.876220703125, "loss": 1.0598, "margin_dpo/margin_mean": 496.9869384765625, "margin_dpo/margin_std": 639.2304077148438, "step": 525 }, { "KL/chosen_KL_mean": -707.6033935546875, "KL/mean": -946.774169921875, "KL/rejected_KL_mean": -1185.945068359375, "KL/std": 587.451171875, "epoch": 0.7723935389133627, "fcm_dpo/beta": 0.0009070096421055496, "fcm_dpo/delta": -0.08758525550365448, "fcm_dpo/margin": 478.34161376953125, "fcm_dpo/q_t": 0.4001006484031677, "grad_norm": 27.308345794677734, "learning_rate": 7.596651350926836e-08, "logits/chosen": -0.9423408508300781, "logits/rejected": -0.943749189376831, "logps/chosen": -770.7706298828125, "logps/ref_chosen": -63.167236328125, "logps/ref_rejected": -86.30934143066406, "logps/rejected": -1272.25439453125, "loss": 1.0858, "margin_dpo/margin_mean": 478.34161376953125, "margin_dpo/margin_std": 668.8135986328125, "step": 526 }, { "KL/chosen_KL_mean": -757.494384765625, "KL/mean": -920.9222412109375, "KL/rejected_KL_mean": -1084.3499755859375, "KL/std": 633.6776123046875, "epoch": 0.7738619676945668, "fcm_dpo/beta": 0.0009061114978976548, "fcm_dpo/delta": 0.05550983548164368, "fcm_dpo/margin": 326.85565185546875, "fcm_dpo/q_t": 0.43042802810668945, "grad_norm": 39.842193603515625, "learning_rate": 7.504749238082414e-08, "logits/chosen": -1.1526522636413574, "logits/rejected": -1.125427007675171, "logps/chosen": -828.623046875, "logps/ref_chosen": -71.12867736816406, "logps/ref_rejected": -78.3425521850586, "logps/rejected": -1162.692626953125, "loss": 1.1639, "margin_dpo/margin_mean": 326.85565185546875, "margin_dpo/margin_std": 512.5347900390625, "step": 527 }, { "KL/chosen_KL_mean": -766.772705078125, "KL/mean": -964.36474609375, "KL/rejected_KL_mean": -1161.956787109375, "KL/std": 593.4571533203125, "epoch": 0.775330396475771, "fcm_dpo/beta": 0.0009081118041649461, "fcm_dpo/delta": -0.00923129916191101, "fcm_dpo/margin": 395.18408203125, "fcm_dpo/q_t": 0.42053911089897156, "grad_norm": 42.25053024291992, "learning_rate": 7.413308141366254e-08, "logits/chosen": -1.0484631061553955, "logits/rejected": -1.0378153324127197, "logps/chosen": -834.8621826171875, "logps/ref_chosen": -68.0894546508789, "logps/ref_rejected": -93.91006469726562, "logps/rejected": -1255.866943359375, "loss": 1.1581, "margin_dpo/margin_mean": 395.18408203125, "margin_dpo/margin_std": 718.8801879882812, "step": 528 }, { "KL/chosen_KL_mean": -914.0716552734375, "KL/mean": -1060.9349365234375, "KL/rejected_KL_mean": -1207.7982177734375, "KL/std": 576.7805786132812, "epoch": 0.7767988252569751, "fcm_dpo/beta": 0.0009093580883927643, "fcm_dpo/delta": 0.004190761595964432, "fcm_dpo/margin": 293.7265319824219, "fcm_dpo/q_t": 0.43876832723617554, "grad_norm": 51.59579086303711, "learning_rate": 7.322330470336313e-08, "logits/chosen": -1.104278326034546, "logits/rejected": -1.1218066215515137, "logps/chosen": -969.6466064453125, "logps/ref_chosen": -55.57495880126953, "logps/ref_rejected": -89.20909118652344, "logps/rejected": -1297.00732421875, "loss": 1.245, "margin_dpo/margin_mean": 293.7265319824219, "margin_dpo/margin_std": 738.239990234375, "step": 529 }, { "KL/chosen_KL_mean": -789.41162109375, "KL/mean": -1014.2277221679688, "KL/rejected_KL_mean": -1239.0438232421875, "KL/std": 690.6798095703125, "epoch": 0.7782672540381792, "fcm_dpo/beta": 0.000902860308997333, "fcm_dpo/delta": -0.058281224220991135, "fcm_dpo/margin": 449.6322021484375, "fcm_dpo/q_t": 0.41204962134361267, "grad_norm": 49.076507568359375, "learning_rate": 7.231818622338822e-08, "logits/chosen": -0.9682959318161011, "logits/rejected": -0.9685269594192505, "logps/chosen": -837.0130615234375, "logps/ref_chosen": -47.601417541503906, "logps/ref_rejected": -87.2845230102539, "logps/rejected": -1326.328369140625, "loss": 1.1613, "margin_dpo/margin_mean": 449.6322021484375, "margin_dpo/margin_std": 891.6737060546875, "step": 530 }, { "KL/chosen_KL_mean": -859.981201171875, "KL/mean": -1053.0469970703125, "KL/rejected_KL_mean": -1246.1129150390625, "KL/std": 698.8087158203125, "epoch": 0.7797356828193832, "fcm_dpo/beta": 0.000901082472410053, "fcm_dpo/delta": 0.0020340122282505035, "fcm_dpo/margin": 386.13177490234375, "fcm_dpo/q_t": 0.42149466276168823, "grad_norm": 49.02323913574219, "learning_rate": 7.141774982445147e-08, "logits/chosen": -1.0975751876831055, "logits/rejected": -1.0849671363830566, "logps/chosen": -915.2272338867188, "logps/ref_chosen": -55.246063232421875, "logps/ref_rejected": -70.60598754882812, "logps/rejected": -1316.718994140625, "loss": 1.1536, "margin_dpo/margin_mean": 386.1317443847656, "margin_dpo/margin_std": 673.9454956054688, "step": 531 }, { "KL/chosen_KL_mean": -877.0020751953125, "KL/mean": -1081.432373046875, "KL/rejected_KL_mean": -1285.862548828125, "KL/std": 670.0946044921875, "epoch": 0.7812041116005873, "fcm_dpo/beta": 0.0008910692995414138, "fcm_dpo/delta": -0.01589033380150795, "fcm_dpo/margin": 408.8604431152344, "fcm_dpo/q_t": 0.4186711609363556, "grad_norm": 81.9896469116211, "learning_rate": 7.052201923388953e-08, "logits/chosen": -1.0430657863616943, "logits/rejected": -1.0236797332763672, "logps/chosen": -947.2880859375, "logps/ref_chosen": -70.28601837158203, "logps/ref_rejected": -86.5913314819336, "logps/rejected": -1372.453857421875, "loss": 1.1835, "margin_dpo/margin_mean": 408.8604736328125, "margin_dpo/margin_std": 825.3141479492188, "step": 532 }, { "KL/chosen_KL_mean": -810.05078125, "KL/mean": -956.464599609375, "KL/rejected_KL_mean": -1102.878173828125, "KL/std": 583.2452392578125, "epoch": 0.7826725403817915, "fcm_dpo/beta": 0.0008952566422522068, "fcm_dpo/delta": -0.012017594650387764, "fcm_dpo/margin": 292.827392578125, "fcm_dpo/q_t": 0.4408678412437439, "grad_norm": 35.946956634521484, "learning_rate": 6.963101805503646e-08, "logits/chosen": -1.065936803817749, "logits/rejected": -1.0475103855133057, "logps/chosen": -874.9058837890625, "logps/ref_chosen": -64.8551025390625, "logps/ref_rejected": -76.58805847167969, "logps/rejected": -1179.46630859375, "loss": 1.232, "margin_dpo/margin_mean": 292.827392578125, "margin_dpo/margin_std": 690.3145141601562, "step": 533 }, { "KL/chosen_KL_mean": -825.17578125, "KL/mean": -1033.3275146484375, "KL/rejected_KL_mean": -1241.479248046875, "KL/std": 642.9903564453125, "epoch": 0.7841409691629956, "fcm_dpo/beta": 0.0008851932943798602, "fcm_dpo/delta": -0.020383629947900772, "fcm_dpo/margin": 416.30340576171875, "fcm_dpo/q_t": 0.41717272996902466, "grad_norm": 40.804595947265625, "learning_rate": 6.874476976660184e-08, "logits/chosen": -1.0429396629333496, "logits/rejected": -1.0457737445831299, "logps/chosen": -885.295166015625, "logps/ref_chosen": -60.119388580322266, "logps/ref_rejected": -78.54347229003906, "logps/rejected": -1320.022705078125, "loss": 1.1406, "margin_dpo/margin_mean": 416.30340576171875, "margin_dpo/margin_std": 685.5845947265625, "step": 534 }, { "KL/chosen_KL_mean": -707.3360595703125, "KL/mean": -944.3709716796875, "KL/rejected_KL_mean": -1181.406005859375, "KL/std": 619.294189453125, "epoch": 0.7856093979441997, "fcm_dpo/beta": 0.000885907094925642, "fcm_dpo/delta": -0.07328834384679794, "fcm_dpo/margin": 474.0699462890625, "fcm_dpo/q_t": 0.40550702810287476, "grad_norm": 24.909685134887695, "learning_rate": 6.786329772205246e-08, "logits/chosen": -0.9549378752708435, "logits/rejected": -0.9660934209823608, "logps/chosen": -761.666259765625, "logps/ref_chosen": -54.330238342285156, "logps/ref_rejected": -96.30763244628906, "logps/rejected": -1277.713623046875, "loss": 1.0916, "margin_dpo/margin_mean": 474.0699462890625, "margin_dpo/margin_std": 659.700927734375, "step": 535 }, { "KL/chosen_KL_mean": -628.810791015625, "KL/mean": -908.8687744140625, "KL/rejected_KL_mean": -1188.9267578125, "KL/std": 691.511962890625, "epoch": 0.7870778267254038, "fcm_dpo/beta": 0.0008606475312262774, "fcm_dpo/delta": -0.13862332701683044, "fcm_dpo/margin": 560.1160278320312, "fcm_dpo/q_t": 0.39692431688308716, "grad_norm": 28.304698944091797, "learning_rate": 6.698662514899638e-08, "logits/chosen": -0.9188047051429749, "logits/rejected": -0.9584248065948486, "logps/chosen": -675.8912963867188, "logps/ref_chosen": -47.08053207397461, "logps/ref_rejected": -89.09783935546875, "logps/rejected": -1278.024658203125, "loss": 1.0688, "margin_dpo/margin_mean": 560.115966796875, "margin_dpo/margin_std": 843.6351318359375, "step": 536 }, { "KL/chosen_KL_mean": -652.699462890625, "KL/mean": -839.2890625, "KL/rejected_KL_mean": -1025.878662109375, "KL/std": 556.4027099609375, "epoch": 0.788546255506608, "fcm_dpo/beta": 0.0008501795819029212, "fcm_dpo/delta": -0.04009390249848366, "fcm_dpo/margin": 373.17919921875, "fcm_dpo/q_t": 0.42610064148902893, "grad_norm": 43.84196090698242, "learning_rate": 6.611477514857114e-08, "logits/chosen": -1.0005278587341309, "logits/rejected": -0.9875226020812988, "logps/chosen": -710.4468994140625, "logps/ref_chosen": -57.747467041015625, "logps/ref_rejected": -70.43838500976562, "logps/rejected": -1096.3170166015625, "loss": 1.1769, "margin_dpo/margin_mean": 373.17919921875, "margin_dpo/margin_std": 682.537841796875, "step": 537 }, { "KL/chosen_KL_mean": -809.7417602539062, "KL/mean": -1029.7225341796875, "KL/rejected_KL_mean": -1249.703369140625, "KL/std": 651.2928466796875, "epoch": 0.7900146842878121, "fcm_dpo/beta": 0.0008414477342739701, "fcm_dpo/delta": -0.02124924398958683, "fcm_dpo/margin": 439.9615478515625, "fcm_dpo/q_t": 0.4152023196220398, "grad_norm": 31.94780158996582, "learning_rate": 6.524777069483525e-08, "logits/chosen": -1.0291862487792969, "logits/rejected": -1.023388147354126, "logps/chosen": -876.15771484375, "logps/ref_chosen": -66.41594696044922, "logps/ref_rejected": -84.22808837890625, "logps/rejected": -1333.931396484375, "loss": 1.1207, "margin_dpo/margin_mean": 439.9615478515625, "margin_dpo/margin_std": 652.1756591796875, "step": 538 }, { "KL/chosen_KL_mean": -676.092529296875, "KL/mean": -877.2006225585938, "KL/rejected_KL_mean": -1078.3087158203125, "KL/std": 510.26898193359375, "epoch": 0.7914831130690162, "fcm_dpo/beta": 0.000846359645947814, "fcm_dpo/delta": 0.009787611663341522, "fcm_dpo/margin": 402.2161865234375, "fcm_dpo/q_t": 0.42187222838401794, "grad_norm": 27.017549514770508, "learning_rate": 6.438563463416221e-08, "logits/chosen": -1.0366239547729492, "logits/rejected": -1.0354104042053223, "logps/chosen": -734.5853881835938, "logps/ref_chosen": -58.492855072021484, "logps/ref_rejected": -91.85395050048828, "logps/rejected": -1170.16259765625, "loss": 1.133, "margin_dpo/margin_mean": 402.2161865234375, "margin_dpo/margin_std": 587.4945068359375, "step": 539 }, { "KL/chosen_KL_mean": -674.0075073242188, "KL/mean": -940.6689453125, "KL/rejected_KL_mean": -1207.330322265625, "KL/std": 633.3556518554688, "epoch": 0.7929515418502202, "fcm_dpo/beta": 0.0008355865720659494, "fcm_dpo/delta": -0.1000160276889801, "fcm_dpo/margin": 533.3228149414062, "fcm_dpo/q_t": 0.40245524048805237, "grad_norm": 33.840572357177734, "learning_rate": 6.352838968463919e-08, "logits/chosen": -0.9341703653335571, "logits/rejected": -0.9644461870193481, "logps/chosen": -737.489990234375, "logps/ref_chosen": -63.482513427734375, "logps/ref_rejected": -116.42999267578125, "logps/rejected": -1323.7603759765625, "loss": 1.0855, "margin_dpo/margin_mean": 533.3228149414062, "margin_dpo/margin_std": 773.7269287109375, "step": 540 }, { "KL/chosen_KL_mean": -790.6764526367188, "KL/mean": -917.277099609375, "KL/rejected_KL_mean": -1043.877685546875, "KL/std": 539.9120483398438, "epoch": 0.7944199706314243, "fcm_dpo/beta": 0.0008294563158415258, "fcm_dpo/delta": 0.008179640397429466, "fcm_dpo/margin": 253.2011260986328, "fcm_dpo/q_t": 0.4516167938709259, "grad_norm": 48.75702667236328, "learning_rate": 6.267605843546767e-08, "logits/chosen": -1.0939087867736816, "logits/rejected": -1.086462378501892, "logps/chosen": -868.956787109375, "logps/ref_chosen": -78.28036499023438, "logps/ref_rejected": -103.273681640625, "logps/rejected": -1147.1513671875, "loss": 1.2617, "margin_dpo/margin_mean": 253.2011260986328, "margin_dpo/margin_std": 658.490478515625, "step": 541 }, { "KL/chosen_KL_mean": -658.6339721679688, "KL/mean": -913.677978515625, "KL/rejected_KL_mean": -1168.721923828125, "KL/std": 608.7216186523438, "epoch": 0.7958883994126285, "fcm_dpo/beta": 0.0008145465981215239, "fcm_dpo/delta": -0.07008374482393265, "fcm_dpo/margin": 510.0878601074219, "fcm_dpo/q_t": 0.4053560495376587, "grad_norm": 48.96900939941406, "learning_rate": 6.182866334636888e-08, "logits/chosen": -1.0721611976623535, "logits/rejected": -1.1146684885025024, "logps/chosen": -716.1189575195312, "logps/ref_chosen": -57.48497009277344, "logps/ref_rejected": -96.47506713867188, "logps/rejected": -1265.197021484375, "loss": 1.0974, "margin_dpo/margin_mean": 510.08782958984375, "margin_dpo/margin_std": 724.8299560546875, "step": 542 }, { "KL/chosen_KL_mean": -741.8212280273438, "KL/mean": -929.9910888671875, "KL/rejected_KL_mean": -1118.1610107421875, "KL/std": 717.8450317382812, "epoch": 0.7973568281938326, "fcm_dpo/beta": 0.0008209783118218184, "fcm_dpo/delta": 0.0423150472342968, "fcm_dpo/margin": 376.3399353027344, "fcm_dpo/q_t": 0.4371965527534485, "grad_norm": 36.86077117919922, "learning_rate": 6.098622674699147e-08, "logits/chosen": -0.9811362028121948, "logits/rejected": -1.0160373449325562, "logps/chosen": -802.438720703125, "logps/ref_chosen": -60.61750793457031, "logps/ref_rejected": -105.59896850585938, "logps/rejected": -1223.760009765625, "loss": 1.2106, "margin_dpo/margin_mean": 376.33990478515625, "margin_dpo/margin_std": 861.6043701171875, "step": 543 }, { "KL/chosen_KL_mean": -747.0059204101562, "KL/mean": -959.01513671875, "KL/rejected_KL_mean": -1171.0242919921875, "KL/std": 579.2896728515625, "epoch": 0.7988252569750367, "fcm_dpo/beta": 0.0008243329357355833, "fcm_dpo/delta": 0.0004746345803141594, "fcm_dpo/margin": 424.0183410644531, "fcm_dpo/q_t": 0.41962099075317383, "grad_norm": 35.00315475463867, "learning_rate": 6.01487708363232e-08, "logits/chosen": -1.0107485055923462, "logits/rejected": -1.0369402170181274, "logps/chosen": -806.648193359375, "logps/ref_chosen": -59.642303466796875, "logps/ref_rejected": -100.95469665527344, "logps/rejected": -1271.97900390625, "loss": 1.1376, "margin_dpo/margin_mean": 424.01837158203125, "margin_dpo/margin_std": 667.390380859375, "step": 544 }, { "KL/chosen_KL_mean": -678.4371948242188, "KL/mean": -930.0258178710938, "KL/rejected_KL_mean": -1181.614501953125, "KL/std": 583.45068359375, "epoch": 0.8002936857562408, "fcm_dpo/beta": 0.0008170028449967504, "fcm_dpo/delta": -0.06372885406017303, "fcm_dpo/margin": 503.17724609375, "fcm_dpo/q_t": 0.4060240387916565, "grad_norm": 38.147010803222656, "learning_rate": 5.9316317682106294e-08, "logits/chosen": -0.9224939346313477, "logits/rejected": -0.9622963666915894, "logps/chosen": -746.0858154296875, "logps/ref_chosen": -67.64859771728516, "logps/ref_rejected": -95.90800476074219, "logps/rejected": -1277.5224609375, "loss": 1.0858, "margin_dpo/margin_mean": 503.17724609375, "margin_dpo/margin_std": 667.4920043945312, "step": 545 }, { "KL/chosen_KL_mean": -660.3955688476562, "KL/mean": -818.3583984375, "KL/rejected_KL_mean": -976.3211669921875, "KL/std": 512.590087890625, "epoch": 0.801762114537445, "fcm_dpo/beta": 0.0008259355090558529, "fcm_dpo/delta": 0.09142404794692993, "fcm_dpo/margin": 315.9256591796875, "fcm_dpo/q_t": 0.43779951333999634, "grad_norm": 28.25826644897461, "learning_rate": 5.848888922025552e-08, "logits/chosen": -0.990828275680542, "logits/rejected": -0.9813590049743652, "logps/chosen": -711.1397705078125, "logps/ref_chosen": -50.744232177734375, "logps/ref_rejected": -81.86622619628906, "logps/rejected": -1058.1875, "loss": 1.1939, "margin_dpo/margin_mean": 315.9256591796875, "margin_dpo/margin_std": 555.3873291015625, "step": 546 }, { "KL/chosen_KL_mean": -655.533935546875, "KL/mean": -874.7945556640625, "KL/rejected_KL_mean": -1094.05517578125, "KL/std": 580.2192993164062, "epoch": 0.8032305433186491, "fcm_dpo/beta": 0.0008292993879877031, "fcm_dpo/delta": -0.014224422164261341, "fcm_dpo/margin": 438.5212707519531, "fcm_dpo/q_t": 0.4169241786003113, "grad_norm": 42.50648498535156, "learning_rate": 5.7666507254280265e-08, "logits/chosen": -0.9036662578582764, "logits/rejected": -0.9190746545791626, "logps/chosen": -729.2216796875, "logps/ref_chosen": -73.6877212524414, "logps/ref_rejected": -90.76136779785156, "logps/rejected": -1184.816650390625, "loss": 1.1249, "margin_dpo/margin_mean": 438.5212707519531, "margin_dpo/margin_std": 658.936279296875, "step": 547 }, { "KL/chosen_KL_mean": -698.337890625, "KL/mean": -896.851806640625, "KL/rejected_KL_mean": -1095.3656005859375, "KL/std": 622.6420288085938, "epoch": 0.8046989720998532, "fcm_dpo/beta": 0.0008297248277813196, "fcm_dpo/delta": 0.02127872034907341, "fcm_dpo/margin": 397.02783203125, "fcm_dpo/q_t": 0.427689790725708, "grad_norm": 23.98332977294922, "learning_rate": 5.684919345471029e-08, "logits/chosen": -1.0400688648223877, "logits/rejected": -1.0478882789611816, "logps/chosen": -763.584228515625, "logps/ref_chosen": -65.24634552001953, "logps/ref_rejected": -94.11807250976562, "logps/rejected": -1189.483642578125, "loss": 1.1561, "margin_dpo/margin_mean": 397.0278625488281, "margin_dpo/margin_std": 693.2183837890625, "step": 548 }, { "KL/chosen_KL_mean": -738.9791870117188, "KL/mean": -889.6734008789062, "KL/rejected_KL_mean": -1040.36767578125, "KL/std": 521.7401733398438, "epoch": 0.8061674008810573, "fcm_dpo/beta": 0.0008309698896482587, "fcm_dpo/delta": 0.0033192974515259266, "fcm_dpo/margin": 301.388427734375, "fcm_dpo/q_t": 0.4423815608024597, "grad_norm": 49.45306396484375, "learning_rate": 5.603696935852426e-08, "logits/chosen": -1.0476170778274536, "logits/rejected": -1.0386322736740112, "logps/chosen": -788.1915283203125, "logps/ref_chosen": -49.21235656738281, "logps/ref_rejected": -73.91031646728516, "logps/rejected": -1114.2779541015625, "loss": 1.2187, "margin_dpo/margin_mean": 301.388427734375, "margin_dpo/margin_std": 629.03125, "step": 549 }, { "KL/chosen_KL_mean": -725.575927734375, "KL/mean": -906.271240234375, "KL/rejected_KL_mean": -1086.966552734375, "KL/std": 568.6636962890625, "epoch": 0.8076358296622613, "fcm_dpo/beta": 0.0008367022383026779, "fcm_dpo/delta": 0.04913487285375595, "fcm_dpo/margin": 361.39056396484375, "fcm_dpo/q_t": 0.42964881658554077, "grad_norm": 34.70016098022461, "learning_rate": 5.5229856368582376e-08, "logits/chosen": -0.9834055304527283, "logits/rejected": -1.0090551376342773, "logps/chosen": -782.3829345703125, "logps/ref_chosen": -56.80695343017578, "logps/ref_rejected": -95.12580871582031, "logps/rejected": -1182.09228515625, "loss": 1.164, "margin_dpo/margin_mean": 361.3905334472656, "margin_dpo/margin_std": 592.5504150390625, "step": 550 }, { "KL/chosen_KL_mean": -631.9072265625, "KL/mean": -930.6796875, "KL/rejected_KL_mean": -1229.4521484375, "KL/std": 619.4073486328125, "epoch": 0.8091042584434655, "fcm_dpo/beta": 0.0008201512973755598, "fcm_dpo/delta": -0.14753000438213348, "fcm_dpo/margin": 597.544921875, "fcm_dpo/q_t": 0.38655346632003784, "grad_norm": 45.9125862121582, "learning_rate": 5.4427875753062734e-08, "logits/chosen": -0.9045864343643188, "logits/rejected": -0.9689292907714844, "logps/chosen": -691.0135498046875, "logps/ref_chosen": -59.10633087158203, "logps/ref_rejected": -111.67280578613281, "logps/rejected": -1341.1248779296875, "loss": 1.0182, "margin_dpo/margin_mean": 597.544921875, "margin_dpo/margin_std": 639.778564453125, "step": 551 }, { "KL/chosen_KL_mean": -622.3704223632812, "KL/mean": -966.5551147460938, "KL/rejected_KL_mean": -1310.73974609375, "KL/std": 730.783203125, "epoch": 0.8105726872246696, "fcm_dpo/beta": 0.0007834373391233385, "fcm_dpo/delta": -0.20374326407909393, "fcm_dpo/margin": 688.369384765625, "fcm_dpo/q_t": 0.3815649747848511, "grad_norm": 56.078468322753906, "learning_rate": 5.363104864490034e-08, "logits/chosen": -1.027387022972107, "logits/rejected": -1.0806533098220825, "logps/chosen": -684.7250366210938, "logps/ref_chosen": -62.35459899902344, "logps/ref_rejected": -104.56210327148438, "logps/rejected": -1415.302001953125, "loss": 1.016, "margin_dpo/margin_mean": 688.369384765625, "margin_dpo/margin_std": 843.5326538085938, "step": 552 }, { "KL/chosen_KL_mean": -718.2479248046875, "KL/mean": -891.719970703125, "KL/rejected_KL_mean": -1065.19189453125, "KL/std": 602.6245727539062, "epoch": 0.8120411160058737, "fcm_dpo/beta": 0.0007897895993664861, "fcm_dpo/delta": 0.07819828391075134, "fcm_dpo/margin": 346.944091796875, "fcm_dpo/q_t": 0.4390897750854492, "grad_norm": 29.118030548095703, "learning_rate": 5.2839396041230415e-08, "logits/chosen": -0.9896628856658936, "logits/rejected": -0.9859640598297119, "logps/chosen": -786.5067138671875, "logps/ref_chosen": -68.25881958007812, "logps/ref_rejected": -98.0971450805664, "logps/rejected": -1163.2890625, "loss": 1.1954, "margin_dpo/margin_mean": 346.944091796875, "margin_dpo/margin_std": 672.8452758789062, "step": 553 }, { "KL/chosen_KL_mean": -761.680419921875, "KL/mean": -982.8804931640625, "KL/rejected_KL_mean": -1204.08056640625, "KL/std": 681.6918334960938, "epoch": 0.8135095447870778, "fcm_dpo/beta": 0.000800039735622704, "fcm_dpo/delta": -0.00511806458234787, "fcm_dpo/margin": 442.40008544921875, "fcm_dpo/q_t": 0.41944724321365356, "grad_norm": 130.93975830078125, "learning_rate": 5.205293880283551e-08, "logits/chosen": -0.9642994403839111, "logits/rejected": -0.9368813037872314, "logps/chosen": -829.6281127929688, "logps/ref_chosen": -67.94767761230469, "logps/ref_rejected": -89.78272247314453, "logps/rejected": -1293.86328125, "loss": 1.1836, "margin_dpo/margin_mean": 442.40008544921875, "margin_dpo/margin_std": 893.775634765625, "step": 554 }, { "KL/chosen_KL_mean": -751.6488037109375, "KL/mean": -1048.9979248046875, "KL/rejected_KL_mean": -1346.3470458984375, "KL/std": 696.0157470703125, "epoch": 0.8149779735682819, "fcm_dpo/beta": 0.0007814068230800331, "fcm_dpo/delta": -0.120139941573143, "fcm_dpo/margin": 594.6982421875, "fcm_dpo/q_t": 0.3991040587425232, "grad_norm": 35.379878997802734, "learning_rate": 5.127169765359515e-08, "logits/chosen": -1.0353598594665527, "logits/rejected": -1.1041009426116943, "logps/chosen": -804.979248046875, "logps/ref_chosen": -53.33049011230469, "logps/ref_rejected": -108.47937774658203, "logps/rejected": -1454.826416015625, "loss": 1.0837, "margin_dpo/margin_mean": 594.6982421875, "margin_dpo/margin_std": 901.941162109375, "step": 555 }, { "KL/chosen_KL_mean": -742.638916015625, "KL/mean": -909.952880859375, "KL/rejected_KL_mean": -1077.2667236328125, "KL/std": 527.5067138671875, "epoch": 0.8164464023494861, "fcm_dpo/beta": 0.0007855216972529888, "fcm_dpo/delta": 0.0895049199461937, "fcm_dpo/margin": 334.6279296875, "fcm_dpo/q_t": 0.43837928771972656, "grad_norm": 30.330717086791992, "learning_rate": 5.049569317994012e-08, "logits/chosen": -1.0571699142456055, "logits/rejected": -1.0535497665405273, "logps/chosen": -801.2833862304688, "logps/ref_chosen": -58.64447021484375, "logps/ref_rejected": -101.34040832519531, "logps/rejected": -1178.607177734375, "loss": 1.1816, "margin_dpo/margin_mean": 334.6279296875, "margin_dpo/margin_std": 526.4218139648438, "step": 556 }, { "KL/chosen_KL_mean": -823.6038818359375, "KL/mean": -1078.040283203125, "KL/rejected_KL_mean": -1332.476806640625, "KL/std": 743.6300048828125, "epoch": 0.8179148311306902, "fcm_dpo/beta": 0.000781252165324986, "fcm_dpo/delta": -0.049805186688899994, "fcm_dpo/margin": 508.87274169921875, "fcm_dpo/q_t": 0.4107533097267151, "grad_norm": 44.83199691772461, "learning_rate": 4.9724945830310144e-08, "logits/chosen": -1.0895347595214844, "logits/rejected": -1.1284986734390259, "logps/chosen": -891.444580078125, "logps/ref_chosen": -67.84066009521484, "logps/ref_rejected": -109.93965911865234, "logps/rejected": -1442.4163818359375, "loss": 1.1255, "margin_dpo/margin_mean": 508.8727722167969, "margin_dpo/margin_std": 828.09326171875, "step": 557 }, { "KL/chosen_KL_mean": -763.7918090820312, "KL/mean": -1106.19775390625, "KL/rejected_KL_mean": -1448.603759765625, "KL/std": 703.3863525390625, "epoch": 0.8193832599118943, "fcm_dpo/beta": 0.0007613534107804298, "fcm_dpo/delta": -0.18066798150539398, "fcm_dpo/margin": 684.81201171875, "fcm_dpo/q_t": 0.37840861082077026, "grad_norm": 27.58951187133789, "learning_rate": 4.8959475914614554e-08, "logits/chosen": -1.1353588104248047, "logits/rejected": -1.158677577972412, "logps/chosen": -826.1600341796875, "logps/ref_chosen": -62.36824035644531, "logps/ref_rejected": -102.16102600097656, "logps/rejected": -1550.7647705078125, "loss": 1.0166, "margin_dpo/margin_mean": 684.81201171875, "margin_dpo/margin_std": 793.9622802734375, "step": 558 }, { "KL/chosen_KL_mean": -867.53369140625, "KL/mean": -1147.8270263671875, "KL/rejected_KL_mean": -1428.120361328125, "KL/std": 725.3985595703125, "epoch": 0.8208516886930984, "fcm_dpo/beta": 0.0007472336874343455, "fcm_dpo/delta": -0.07184131443500519, "fcm_dpo/margin": 560.5865478515625, "fcm_dpo/q_t": 0.4061052203178406, "grad_norm": 31.313798904418945, "learning_rate": 4.8199303603697614e-08, "logits/chosen": -1.2728649377822876, "logits/rejected": -1.2823593616485596, "logps/chosen": -928.2860717773438, "logps/ref_chosen": -60.752323150634766, "logps/ref_rejected": -93.44229125976562, "logps/rejected": -1521.5626220703125, "loss": 1.0946, "margin_dpo/margin_mean": 560.5865478515625, "margin_dpo/margin_std": 811.6783447265625, "step": 559 }, { "KL/chosen_KL_mean": -794.177734375, "KL/mean": -987.329833984375, "KL/rejected_KL_mean": -1180.4820556640625, "KL/std": 631.4752197265625, "epoch": 0.8223201174743024, "fcm_dpo/beta": 0.000741536554414779, "fcm_dpo/delta": -0.008997570723295212, "fcm_dpo/margin": 386.3043212890625, "fcm_dpo/q_t": 0.4339349865913391, "grad_norm": 36.68315505981445, "learning_rate": 4.7444448928806615e-08, "logits/chosen": -0.9954744577407837, "logits/rejected": -0.9756410121917725, "logps/chosen": -852.2815551757812, "logps/ref_chosen": -58.10382080078125, "logps/ref_rejected": -79.99122619628906, "logps/rejected": -1260.4732666015625, "loss": 1.1833, "margin_dpo/margin_mean": 386.3043212890625, "margin_dpo/margin_std": 689.2628173828125, "step": 560 }, { "KL/chosen_KL_mean": -918.6915283203125, "KL/mean": -1091.2904052734375, "KL/rejected_KL_mean": -1263.8892822265625, "KL/std": 642.873291015625, "epoch": 0.8237885462555066, "fcm_dpo/beta": 0.0007538319332525134, "fcm_dpo/delta": 0.09177450835704803, "fcm_dpo/margin": 345.19769287109375, "fcm_dpo/q_t": 0.43946483731269836, "grad_norm": 40.61189651489258, "learning_rate": 4.669493178106432e-08, "logits/chosen": -1.1234174966812134, "logits/rejected": -1.147684097290039, "logps/chosen": -969.6044311523438, "logps/ref_chosen": -50.912879943847656, "logps/ref_rejected": -99.06856536865234, "logps/rejected": -1362.957763671875, "loss": 1.2324, "margin_dpo/margin_mean": 345.19769287109375, "margin_dpo/margin_std": 824.038330078125, "step": 561 }, { "KL/chosen_KL_mean": -841.947509765625, "KL/mean": -1096.912841796875, "KL/rejected_KL_mean": -1351.8782958984375, "KL/std": 714.989990234375, "epoch": 0.8252569750367107, "fcm_dpo/beta": 0.0007471415447071195, "fcm_dpo/delta": -0.03312261030077934, "fcm_dpo/margin": 509.93084716796875, "fcm_dpo/q_t": 0.4142550826072693, "grad_norm": 37.721954345703125, "learning_rate": 4.5950771910944596e-08, "logits/chosen": -1.1088937520980835, "logits/rejected": -1.1244970560073853, "logps/chosen": -901.411865234375, "logps/ref_chosen": -59.46440124511719, "logps/ref_rejected": -96.54266357421875, "logps/rejected": -1448.4208984375, "loss": 1.1201, "margin_dpo/margin_mean": 509.93084716796875, "margin_dpo/margin_std": 781.1636962890625, "step": 562 }, { "KL/chosen_KL_mean": -965.8650512695312, "KL/mean": -1130.153564453125, "KL/rejected_KL_mean": -1294.44189453125, "KL/std": 774.4320068359375, "epoch": 0.8267254038179148, "fcm_dpo/beta": 0.00074770423816517, "fcm_dpo/delta": -0.0345328263938427, "fcm_dpo/margin": 328.5767822265625, "fcm_dpo/q_t": 0.43725377321243286, "grad_norm": 60.61211013793945, "learning_rate": 4.521198892775202e-08, "logits/chosen": -1.101536512374878, "logits/rejected": -1.1053555011749268, "logps/chosen": -1026.4732666015625, "logps/ref_chosen": -60.60819625854492, "logps/ref_rejected": -94.56770324707031, "logps/rejected": -1389.009521484375, "loss": 1.2732, "margin_dpo/margin_mean": 328.5767822265625, "margin_dpo/margin_std": 918.204345703125, "step": 563 }, { "KL/chosen_KL_mean": -839.5714111328125, "KL/mean": -1072.73291015625, "KL/rejected_KL_mean": -1305.894287109375, "KL/std": 678.1080322265625, "epoch": 0.8281938325991189, "fcm_dpo/beta": 0.0007467868272215128, "fcm_dpo/delta": 0.0017370246350765228, "fcm_dpo/margin": 466.32293701171875, "fcm_dpo/q_t": 0.42049744725227356, "grad_norm": 39.77863693237305, "learning_rate": 4.447860229910544e-08, "logits/chosen": -1.1387296915054321, "logits/rejected": -1.1342732906341553, "logps/chosen": -913.8397827148438, "logps/ref_chosen": -74.26837921142578, "logps/ref_rejected": -93.23818969726562, "logps/rejected": -1399.132568359375, "loss": 1.1292, "margin_dpo/margin_mean": 466.32293701171875, "margin_dpo/margin_std": 674.2352294921875, "step": 564 }, { "KL/chosen_KL_mean": -903.0402221679688, "KL/mean": -1139.46142578125, "KL/rejected_KL_mean": -1375.882568359375, "KL/std": 793.594482421875, "epoch": 0.8296622613803231, "fcm_dpo/beta": 0.0007455106824636459, "fcm_dpo/delta": -0.002605642657727003, "fcm_dpo/margin": 472.84234619140625, "fcm_dpo/q_t": 0.424316942691803, "grad_norm": 58.753501892089844, "learning_rate": 4.375063135042445e-08, "logits/chosen": -1.083191156387329, "logits/rejected": -1.084733486175537, "logps/chosen": -972.0601806640625, "logps/ref_chosen": -69.0199203491211, "logps/ref_rejected": -85.7789306640625, "logps/rejected": -1461.6614990234375, "loss": 1.1859, "margin_dpo/margin_mean": 472.84234619140625, "margin_dpo/margin_std": 984.09130859375, "step": 565 }, { "KL/chosen_KL_mean": -851.157958984375, "KL/mean": -1113.0625, "KL/rejected_KL_mean": -1374.967041015625, "KL/std": 766.4264526367188, "epoch": 0.8311306901615272, "fcm_dpo/beta": 0.0007446268573403358, "fcm_dpo/delta": -0.04210010543465614, "fcm_dpo/margin": 523.8090209960938, "fcm_dpo/q_t": 0.41205883026123047, "grad_norm": 30.277292251586914, "learning_rate": 4.3028095264420525e-08, "logits/chosen": -1.0920714139938354, "logits/rejected": -1.1179513931274414, "logps/chosen": -917.7032470703125, "logps/ref_chosen": -66.5453109741211, "logps/ref_rejected": -103.86932373046875, "logps/rejected": -1478.8363037109375, "loss": 1.1391, "margin_dpo/margin_mean": 523.8089599609375, "margin_dpo/margin_std": 888.4176025390625, "step": 566 }, { "KL/chosen_KL_mean": -797.819091796875, "KL/mean": -1001.6353759765625, "KL/rejected_KL_mean": -1205.4515380859375, "KL/std": 539.5685424804688, "epoch": 0.8325991189427313, "fcm_dpo/beta": 0.0007430259138345718, "fcm_dpo/delta": 0.04861754924058914, "fcm_dpo/margin": 407.6325378417969, "fcm_dpo/q_t": 0.42830806970596313, "grad_norm": 25.824920654296875, "learning_rate": 4.231101308059165e-08, "logits/chosen": -1.213765263557434, "logits/rejected": -1.21918523311615, "logps/chosen": -850.6773681640625, "logps/ref_chosen": -52.85829544067383, "logps/ref_rejected": -85.37095642089844, "logps/rejected": -1290.822509765625, "loss": 1.1507, "margin_dpo/margin_mean": 407.632568359375, "margin_dpo/margin_std": 579.9044189453125, "step": 567 }, { "KL/chosen_KL_mean": -774.791748046875, "KL/mean": -1049.3804931640625, "KL/rejected_KL_mean": -1323.96923828125, "KL/std": 630.48486328125, "epoch": 0.8340675477239354, "fcm_dpo/beta": 0.0007391650578938425, "fcm_dpo/delta": -0.05833129584789276, "fcm_dpo/margin": 549.1774291992188, "fcm_dpo/q_t": 0.40518689155578613, "grad_norm": 27.453920364379883, "learning_rate": 4.1599403694720145e-08, "logits/chosen": -1.0664093494415283, "logits/rejected": -1.1100949048995972, "logps/chosen": -819.984130859375, "logps/ref_chosen": -45.1923828125, "logps/ref_rejected": -89.09236907958984, "logps/rejected": -1413.0615234375, "loss": 1.0805, "margin_dpo/margin_mean": 549.1774291992188, "margin_dpo/margin_std": 684.4533081054688, "step": 568 }, { "KL/chosen_KL_mean": -870.124267578125, "KL/mean": -1090.1845703125, "KL/rejected_KL_mean": -1310.244873046875, "KL/std": 802.5746459960938, "epoch": 0.8355359765051396, "fcm_dpo/beta": 0.0007340764277614653, "fcm_dpo/delta": -0.04408378154039383, "fcm_dpo/margin": 440.1204833984375, "fcm_dpo/q_t": 0.42503905296325684, "grad_norm": 42.93728256225586, "learning_rate": 4.089328585837512e-08, "logits/chosen": -1.109398365020752, "logits/rejected": -1.1187589168548584, "logps/chosen": -933.8448486328125, "logps/ref_chosen": -63.72056198120117, "logps/ref_rejected": -79.10325622558594, "logps/rejected": -1389.3480224609375, "loss": 1.1831, "margin_dpo/margin_mean": 440.1204833984375, "margin_dpo/margin_std": 843.69482421875, "step": 569 }, { "KL/chosen_KL_mean": -786.7838134765625, "KL/mean": -990.749755859375, "KL/rejected_KL_mean": -1194.715576171875, "KL/std": 626.913818359375, "epoch": 0.8370044052863436, "fcm_dpo/beta": 0.0007354548433795571, "fcm_dpo/delta": 0.05156288668513298, "fcm_dpo/margin": 407.93182373046875, "fcm_dpo/q_t": 0.43151092529296875, "grad_norm": 29.369874954223633, "learning_rate": 4.019267817841834e-08, "logits/chosen": -1.1785709857940674, "logits/rejected": -1.1703860759735107, "logps/chosen": -848.3983764648438, "logps/ref_chosen": -61.61454391479492, "logps/ref_rejected": -82.14186096191406, "logps/rejected": -1276.8575439453125, "loss": 1.1709, "margin_dpo/margin_mean": 407.9317932128906, "margin_dpo/margin_std": 689.3609619140625, "step": 570 }, { "KL/chosen_KL_mean": -815.0283813476562, "KL/mean": -1055.030029296875, "KL/rejected_KL_mean": -1295.03173828125, "KL/std": 642.4834594726562, "epoch": 0.8384728340675477, "fcm_dpo/beta": 0.0007371066603809595, "fcm_dpo/delta": -0.0039865849539637566, "fcm_dpo/margin": 480.00323486328125, "fcm_dpo/q_t": 0.41918835043907166, "grad_norm": 41.169490814208984, "learning_rate": 3.9497599116513705e-08, "logits/chosen": -1.0355522632598877, "logits/rejected": -1.0509165525436401, "logps/chosen": -868.0824584960938, "logps/ref_chosen": -53.05406188964844, "logps/ref_rejected": -91.33682250976562, "logps/rejected": -1386.3685302734375, "loss": 1.1453, "margin_dpo/margin_mean": 480.0032653808594, "margin_dpo/margin_std": 803.4883422851562, "step": 571 }, { "KL/chosen_KL_mean": -836.864013671875, "KL/mean": -1090.23388671875, "KL/rejected_KL_mean": -1343.603759765625, "KL/std": 737.3343505859375, "epoch": 0.8399412628487518, "fcm_dpo/beta": 0.0007338828290812671, "fcm_dpo/delta": -0.022855721414089203, "fcm_dpo/margin": 506.73992919921875, "fcm_dpo/q_t": 0.41840416193008423, "grad_norm": 32.44576644897461, "learning_rate": 3.880806698864086e-08, "logits/chosen": -1.0318219661712646, "logits/rejected": -1.06544828414917, "logps/chosen": -885.3232421875, "logps/ref_chosen": -48.45928955078125, "logps/ref_rejected": -83.55703735351562, "logps/rejected": -1427.160888671875, "loss": 1.1474, "margin_dpo/margin_mean": 506.73992919921875, "margin_dpo/margin_std": 905.3984985351562, "step": 572 }, { "KL/chosen_KL_mean": -788.7066040039062, "KL/mean": -1008.2884521484375, "KL/rejected_KL_mean": -1227.870361328125, "KL/std": 634.485107421875, "epoch": 0.8414096916299559, "fcm_dpo/beta": 0.0007377726724371314, "fcm_dpo/delta": 0.02684948407113552, "fcm_dpo/margin": 439.1636962890625, "fcm_dpo/q_t": 0.42565691471099854, "grad_norm": 25.155153274536133, "learning_rate": 3.812409996461275e-08, "logits/chosen": -1.1160049438476562, "logits/rejected": -1.1299877166748047, "logps/chosen": -840.3292236328125, "logps/ref_chosen": -51.62262725830078, "logps/ref_rejected": -85.32499694824219, "logps/rejected": -1313.1953125, "loss": 1.1386, "margin_dpo/margin_mean": 439.1636657714844, "margin_dpo/margin_std": 636.15185546875, "step": 573 }, { "KL/chosen_KL_mean": -715.6622314453125, "KL/mean": -959.97314453125, "KL/rejected_KL_mean": -1204.2841796875, "KL/std": 584.0929565429688, "epoch": 0.8428781204111601, "fcm_dpo/beta": 0.0007384730270132422, "fcm_dpo/delta": -0.011282573454082012, "fcm_dpo/margin": 488.62188720703125, "fcm_dpo/q_t": 0.416146457195282, "grad_norm": 30.116424560546875, "learning_rate": 3.74457160675965e-08, "logits/chosen": -1.0976953506469727, "logits/rejected": -1.1268978118896484, "logps/chosen": -766.7066650390625, "logps/ref_chosen": -51.04446029663086, "logps/ref_rejected": -92.80640411376953, "logps/rejected": -1297.090576171875, "loss": 1.1183, "margin_dpo/margin_mean": 488.62188720703125, "margin_dpo/margin_std": 683.5946044921875, "step": 574 }, { "KL/chosen_KL_mean": -779.8887939453125, "KL/mean": -997.6763916015625, "KL/rejected_KL_mean": -1215.464111328125, "KL/std": 598.38037109375, "epoch": 0.8443465491923642, "fcm_dpo/beta": 0.0007295336690731347, "fcm_dpo/delta": -0.049235548824071884, "fcm_dpo/margin": 435.5752258300781, "fcm_dpo/q_t": 0.4261569082736969, "grad_norm": 35.20523452758789, "learning_rate": 3.677293317363864e-08, "logits/chosen": -0.9492102861404419, "logits/rejected": -0.9579809904098511, "logps/chosen": -851.678955078125, "logps/ref_chosen": -71.7901382446289, "logps/ref_rejected": -95.38619995117188, "logps/rejected": -1310.850341796875, "loss": 1.1713, "margin_dpo/margin_mean": 435.57525634765625, "margin_dpo/margin_std": 770.2555541992188, "step": 575 }, { "KL/chosen_KL_mean": -756.765869140625, "KL/mean": -921.8978271484375, "KL/rejected_KL_mean": -1087.0296630859375, "KL/std": 545.06005859375, "epoch": 0.8458149779735683, "fcm_dpo/beta": 0.0007330116932280362, "fcm_dpo/delta": 0.023780547082424164, "fcm_dpo/margin": 330.26385498046875, "fcm_dpo/q_t": 0.4433874785900116, "grad_norm": 32.09224319458008, "learning_rate": 3.6105769011194224e-08, "logits/chosen": -1.070650339126587, "logits/rejected": -1.1040256023406982, "logps/chosen": -811.02880859375, "logps/ref_chosen": -54.262962341308594, "logps/ref_rejected": -100.75428009033203, "logps/rejected": -1187.783935546875, "loss": 1.211, "margin_dpo/margin_mean": 330.2638854980469, "margin_dpo/margin_std": 628.616943359375, "step": 576 }, { "KL/chosen_KL_mean": -669.3681640625, "KL/mean": -892.837890625, "KL/rejected_KL_mean": -1116.3076171875, "KL/std": 621.8170776367188, "epoch": 0.8472834067547724, "fcm_dpo/beta": 0.0007347659557126462, "fcm_dpo/delta": 0.022340623661875725, "fcm_dpo/margin": 446.939453125, "fcm_dpo/q_t": 0.424075722694397, "grad_norm": 25.360074996948242, "learning_rate": 3.5444241160659304e-08, "logits/chosen": -1.0318996906280518, "logits/rejected": -1.0215978622436523, "logps/chosen": -731.2778930664062, "logps/ref_chosen": -61.909706115722656, "logps/ref_rejected": -84.07069396972656, "logps/rejected": -1200.3782958984375, "loss": 1.1501, "margin_dpo/margin_mean": 446.93951416015625, "margin_dpo/margin_std": 677.6488037109375, "step": 577 }, { "KL/chosen_KL_mean": -641.9940185546875, "KL/mean": -880.4281005859375, "KL/rejected_KL_mean": -1118.8621826171875, "KL/std": 590.3482055664062, "epoch": 0.8487518355359766, "fcm_dpo/beta": 0.000733315828256309, "fcm_dpo/delta": -1.9393861293792725e-05, "fcm_dpo/margin": 476.8681335449219, "fcm_dpo/q_t": 0.41781163215637207, "grad_norm": 46.754920959472656, "learning_rate": 3.478836705390808e-08, "logits/chosen": -0.9348374605178833, "logits/rejected": -0.9725053310394287, "logps/chosen": -691.2577514648438, "logps/ref_chosen": -49.26368713378906, "logps/ref_rejected": -83.4362564086914, "logps/rejected": -1202.2984619140625, "loss": 1.1143, "margin_dpo/margin_mean": 476.8681335449219, "margin_dpo/margin_std": 598.9371337890625, "step": 578 }, { "KL/chosen_KL_mean": -752.2142333984375, "KL/mean": -893.55517578125, "KL/rejected_KL_mean": -1034.8961181640625, "KL/std": 593.69775390625, "epoch": 0.8502202643171806, "fcm_dpo/beta": 0.0007440147455781698, "fcm_dpo/delta": 0.052174538373947144, "fcm_dpo/margin": 282.68194580078125, "fcm_dpo/q_t": 0.4504472315311432, "grad_norm": 50.663330078125, "learning_rate": 3.41381639738331e-08, "logits/chosen": -0.9605880975723267, "logits/rejected": -0.964146077632904, "logps/chosen": -811.10009765625, "logps/ref_chosen": -58.88581848144531, "logps/ref_rejected": -94.78762817382812, "logps/rejected": -1129.683837890625, "loss": 1.2467, "margin_dpo/margin_mean": 282.68194580078125, "margin_dpo/margin_std": 665.921142578125, "step": 579 }, { "KL/chosen_KL_mean": -525.0825805664062, "KL/mean": -796.9617919921875, "KL/rejected_KL_mean": -1068.841064453125, "KL/std": 658.41455078125, "epoch": 0.8516886930983847, "fcm_dpo/beta": 0.0007361972820945084, "fcm_dpo/delta": -0.05270449444651604, "fcm_dpo/margin": 543.7584228515625, "fcm_dpo/q_t": 0.4098510146141052, "grad_norm": 40.2070426940918, "learning_rate": 3.349364905389032e-08, "logits/chosen": -0.8550982475280762, "logits/rejected": -0.8935449719429016, "logps/chosen": -573.7894287109375, "logps/ref_chosen": -48.70683670043945, "logps/ref_rejected": -81.7583999633789, "logps/rejected": -1150.5994873046875, "loss": 1.0934, "margin_dpo/margin_mean": 543.7584228515625, "margin_dpo/margin_std": 754.7635498046875, "step": 580 }, { "KL/chosen_KL_mean": -729.9277954101562, "KL/mean": -922.0654296875, "KL/rejected_KL_mean": -1114.202880859375, "KL/std": 613.97216796875, "epoch": 0.8531571218795888, "fcm_dpo/beta": 0.0007442636415362358, "fcm_dpo/delta": 0.06579051911830902, "fcm_dpo/margin": 384.27520751953125, "fcm_dpo/q_t": 0.434751033782959, "grad_norm": 25.00486946105957, "learning_rate": 3.285483927764726e-08, "logits/chosen": -1.0764214992523193, "logits/rejected": -1.0914995670318604, "logps/chosen": -792.150146484375, "logps/ref_chosen": -62.22235107421875, "logps/ref_rejected": -91.73568725585938, "logps/rejected": -1205.938720703125, "loss": 1.1788, "margin_dpo/margin_mean": 384.27520751953125, "margin_dpo/margin_std": 675.5390625, "step": 581 }, { "KL/chosen_KL_mean": -657.9926147460938, "KL/mean": -858.8969116210938, "KL/rejected_KL_mean": -1059.80126953125, "KL/std": 526.9097290039062, "epoch": 0.8546255506607929, "fcm_dpo/beta": 0.0007423992501571774, "fcm_dpo/delta": -0.04802374914288521, "fcm_dpo/margin": 401.8086242675781, "fcm_dpo/q_t": 0.4290406405925751, "grad_norm": 27.108118057250977, "learning_rate": 3.222175147833556e-08, "logits/chosen": -0.9977607727050781, "logits/rejected": -1.0202760696411133, "logps/chosen": -716.2213134765625, "logps/ref_chosen": -58.228660583496094, "logps/ref_rejected": -110.06959533691406, "logps/rejected": -1169.870849609375, "loss": 1.1596, "margin_dpo/margin_mean": 401.80865478515625, "margin_dpo/margin_std": 577.4631958007812, "step": 582 }, { "KL/chosen_KL_mean": -748.9545288085938, "KL/mean": -883.51416015625, "KL/rejected_KL_mean": -1018.0738525390625, "KL/std": 600.503662109375, "epoch": 0.856093979441997, "fcm_dpo/beta": 0.0007386098150163889, "fcm_dpo/delta": -0.0015174677828326821, "fcm_dpo/margin": 269.1192626953125, "fcm_dpo/q_t": 0.45400410890579224, "grad_norm": 39.618011474609375, "learning_rate": 3.159440233840763e-08, "logits/chosen": -0.953858494758606, "logits/rejected": -0.9514139890670776, "logps/chosen": -805.8173828125, "logps/ref_chosen": -56.86286163330078, "logps/ref_rejected": -88.4039306640625, "logps/rejected": -1106.477783203125, "loss": 1.2716, "margin_dpo/margin_mean": 269.1192626953125, "margin_dpo/margin_std": 726.809814453125, "step": 583 }, { "KL/chosen_KL_mean": -650.4002075195312, "KL/mean": -920.8975830078125, "KL/rejected_KL_mean": -1191.39501953125, "KL/std": 627.6343383789062, "epoch": 0.8575624082232012, "fcm_dpo/beta": 0.0007337032584473491, "fcm_dpo/delta": -0.0488472580909729, "fcm_dpo/margin": 540.9948120117188, "fcm_dpo/q_t": 0.40848594903945923, "grad_norm": 33.2059326171875, "learning_rate": 3.0972808389096635e-08, "logits/chosen": -0.9727611541748047, "logits/rejected": -0.9878000020980835, "logps/chosen": -707.3009033203125, "logps/ref_chosen": -56.90068054199219, "logps/ref_rejected": -97.63606262207031, "logps/rejected": -1289.0311279296875, "loss": 1.0841, "margin_dpo/margin_mean": 540.9948120117188, "margin_dpo/margin_std": 661.6787109375, "step": 584 }, { "KL/chosen_KL_mean": -755.4912109375, "KL/mean": -1001.9600219726562, "KL/rejected_KL_mean": -1248.4288330078125, "KL/std": 691.0889892578125, "epoch": 0.8590308370044053, "fcm_dpo/beta": 0.0007283834274858236, "fcm_dpo/delta": -0.009564774110913277, "fcm_dpo/margin": 492.93756103515625, "fcm_dpo/q_t": 0.4175519645214081, "grad_norm": 28.804832458496094, "learning_rate": 3.035698600998121e-08, "logits/chosen": -1.0524985790252686, "logits/rejected": -1.0806963443756104, "logps/chosen": -816.4652099609375, "logps/ref_chosen": -60.973968505859375, "logps/ref_rejected": -84.16952514648438, "logps/rejected": -1332.598388671875, "loss": 1.1445, "margin_dpo/margin_mean": 492.93756103515625, "margin_dpo/margin_std": 832.9927978515625, "step": 585 }, { "KL/chosen_KL_mean": -810.9488525390625, "KL/mean": -974.180419921875, "KL/rejected_KL_mean": -1137.412109375, "KL/std": 624.2877197265625, "epoch": 0.8604992657856094, "fcm_dpo/beta": 0.0007359388982877135, "fcm_dpo/delta": 0.04029836505651474, "fcm_dpo/margin": 326.46331787109375, "fcm_dpo/q_t": 0.4438827335834503, "grad_norm": 31.14897918701172, "learning_rate": 2.974695142855388e-08, "logits/chosen": -1.0293405055999756, "logits/rejected": -1.0565171241760254, "logps/chosen": -867.804443359375, "logps/ref_chosen": -56.85559844970703, "logps/ref_rejected": -91.80261993408203, "logps/rejected": -1229.2147216796875, "loss": 1.2212, "margin_dpo/margin_mean": 326.46331787109375, "margin_dpo/margin_std": 682.7779541015625, "step": 586 }, { "KL/chosen_KL_mean": -576.955810546875, "KL/mean": -799.3982543945312, "KL/rejected_KL_mean": -1021.8407592773438, "KL/std": 647.3914794921875, "epoch": 0.8619676945668135, "fcm_dpo/beta": 0.0007399095338769257, "fcm_dpo/delta": 0.02138981781899929, "fcm_dpo/margin": 444.885009765625, "fcm_dpo/q_t": 0.42297637462615967, "grad_norm": 35.817604064941406, "learning_rate": 2.9142720719793122e-08, "logits/chosen": -1.0655786991119385, "logits/rejected": -1.0954036712646484, "logps/chosen": -621.6473388671875, "logps/ref_chosen": -44.69159698486328, "logps/ref_rejected": -82.62385559082031, "logps/rejected": -1104.464599609375, "loss": 1.1398, "margin_dpo/margin_mean": 444.885009765625, "margin_dpo/margin_std": 656.8631591796875, "step": 587 }, { "KL/chosen_KL_mean": -763.5947265625, "KL/mean": -951.0715942382812, "KL/rejected_KL_mean": -1138.5484619140625, "KL/std": 554.4877319335938, "epoch": 0.8634361233480177, "fcm_dpo/beta": 0.0007451215060427785, "fcm_dpo/delta": 0.07268988341093063, "fcm_dpo/margin": 374.9537353515625, "fcm_dpo/q_t": 0.4341343939304352, "grad_norm": 28.395036697387695, "learning_rate": 2.8544309805740018e-08, "logits/chosen": -1.0229034423828125, "logits/rejected": -1.052525281906128, "logps/chosen": -813.8897094726562, "logps/ref_chosen": -50.29494857788086, "logps/ref_rejected": -107.36988067626953, "logps/rejected": -1245.9183349609375, "loss": 1.1757, "margin_dpo/margin_mean": 374.9537353515625, "margin_dpo/margin_std": 601.4207763671875, "step": 588 }, { "KL/chosen_KL_mean": -765.1441650390625, "KL/mean": -1015.72314453125, "KL/rejected_KL_mean": -1266.30224609375, "KL/std": 625.617431640625, "epoch": 0.8649045521292217, "fcm_dpo/beta": 0.000747905985917896, "fcm_dpo/delta": -0.025782965123653412, "fcm_dpo/margin": 501.15802001953125, "fcm_dpo/q_t": 0.4127042889595032, "grad_norm": 36.19021224975586, "learning_rate": 2.7951734455078786e-08, "logits/chosen": -1.0197863578796387, "logits/rejected": -1.0424385070800781, "logps/chosen": -825.0740356445312, "logps/ref_chosen": -59.929908752441406, "logps/ref_rejected": -111.65534973144531, "logps/rejected": -1377.95751953125, "loss": 1.1029, "margin_dpo/margin_mean": 501.1579895019531, "margin_dpo/margin_std": 657.3384399414062, "step": 589 }, { "KL/chosen_KL_mean": -667.9539794921875, "KL/mean": -920.4093627929688, "KL/rejected_KL_mean": -1172.86474609375, "KL/std": 630.0244140625, "epoch": 0.8663729809104258, "fcm_dpo/beta": 0.000742416363209486, "fcm_dpo/delta": -0.02587101422250271, "fcm_dpo/margin": 504.91082763671875, "fcm_dpo/q_t": 0.41454923152923584, "grad_norm": 20.466495513916016, "learning_rate": 2.736501028272095e-08, "logits/chosen": -1.0065345764160156, "logits/rejected": -1.039345145225525, "logps/chosen": -723.7637939453125, "logps/ref_chosen": -55.80979537963867, "logps/ref_rejected": -106.06282043457031, "logps/rejected": -1278.927734375, "loss": 1.109, "margin_dpo/margin_mean": 504.91082763671875, "margin_dpo/margin_std": 708.3488159179688, "step": 590 }, { "KL/chosen_KL_mean": -719.654296875, "KL/mean": -956.586181640625, "KL/rejected_KL_mean": -1193.51806640625, "KL/std": 603.654052734375, "epoch": 0.8678414096916299, "fcm_dpo/beta": 0.0007421453483402729, "fcm_dpo/delta": -0.0017371654976159334, "fcm_dpo/margin": 473.86370849609375, "fcm_dpo/q_t": 0.41778093576431274, "grad_norm": 29.069992065429688, "learning_rate": 2.678415274939408e-08, "logits/chosen": -1.1044466495513916, "logits/rejected": -1.1089308261871338, "logps/chosen": -775.8948974609375, "logps/ref_chosen": -56.24061965942383, "logps/ref_rejected": -83.78629302978516, "logps/rejected": -1277.3043212890625, "loss": 1.1284, "margin_dpo/margin_mean": 473.86370849609375, "margin_dpo/margin_std": 696.78857421875, "step": 591 }, { "KL/chosen_KL_mean": -779.455078125, "KL/mean": -986.9091186523438, "KL/rejected_KL_mean": -1194.3631591796875, "KL/std": 615.1773681640625, "epoch": 0.869309838472834, "fcm_dpo/beta": 0.0007462240173481405, "fcm_dpo/delta": 0.04169132187962532, "fcm_dpo/margin": 414.90814208984375, "fcm_dpo/q_t": 0.4289587736129761, "grad_norm": 39.19664764404297, "learning_rate": 2.6209177161234442e-08, "logits/chosen": -1.0807688236236572, "logits/rejected": -1.090897560119629, "logps/chosen": -827.3953247070312, "logps/ref_chosen": -47.94025421142578, "logps/ref_rejected": -75.73287963867188, "logps/rejected": -1270.0960693359375, "loss": 1.1963, "margin_dpo/margin_mean": 414.90814208984375, "margin_dpo/margin_std": 837.1720581054688, "step": 592 }, { "KL/chosen_KL_mean": -763.120849609375, "KL/mean": -918.5281982421875, "KL/rejected_KL_mean": -1073.9356689453125, "KL/std": 664.6796264648438, "epoch": 0.8707782672540382, "fcm_dpo/beta": 0.0007523976382799447, "fcm_dpo/delta": 0.028210198506712914, "fcm_dpo/margin": 310.814697265625, "fcm_dpo/q_t": 0.4463445246219635, "grad_norm": 45.363460540771484, "learning_rate": 2.564009866938349e-08, "logits/chosen": -0.9572064280509949, "logits/rejected": -0.9502384662628174, "logps/chosen": -811.8116455078125, "logps/ref_chosen": -48.690757751464844, "logps/ref_rejected": -60.90800094604492, "logps/rejected": -1134.8436279296875, "loss": 1.239, "margin_dpo/margin_mean": 310.814697265625, "margin_dpo/margin_std": 729.053955078125, "step": 593 }, { "KL/chosen_KL_mean": -731.2477416992188, "KL/mean": -935.2242431640625, "KL/rejected_KL_mean": -1139.2008056640625, "KL/std": 653.4246215820312, "epoch": 0.8722466960352423, "fcm_dpo/beta": 0.0007495468016713858, "fcm_dpo/delta": -0.03803475201129913, "fcm_dpo/margin": 407.953125, "fcm_dpo/q_t": 0.429071843624115, "grad_norm": 32.1541633605957, "learning_rate": 2.5076932269588708e-08, "logits/chosen": -1.05826735496521, "logits/rejected": -1.0492217540740967, "logps/chosen": -786.1826171875, "logps/ref_chosen": -54.93488693237305, "logps/ref_rejected": -86.09967803955078, "logps/rejected": -1225.300537109375, "loss": 1.1734, "margin_dpo/margin_mean": 407.953125, "margin_dpo/margin_std": 699.8004150390625, "step": 594 }, { "KL/chosen_KL_mean": -674.6976928710938, "KL/mean": -915.5961303710938, "KL/rejected_KL_mean": -1156.4945068359375, "KL/std": 617.6531982421875, "epoch": 0.8737151248164464, "fcm_dpo/beta": 0.0007444015936926007, "fcm_dpo/delta": -0.00904869381338358, "fcm_dpo/margin": 481.7968444824219, "fcm_dpo/q_t": 0.4190996289253235, "grad_norm": 51.79603958129883, "learning_rate": 2.451969280180849e-08, "logits/chosen": -1.0517390966415405, "logits/rejected": -1.0796115398406982, "logps/chosen": -724.1181030273438, "logps/ref_chosen": -49.4204216003418, "logps/ref_rejected": -80.62731170654297, "logps/rejected": -1237.121826171875, "loss": 1.1206, "margin_dpo/margin_mean": 481.7968444824219, "margin_dpo/margin_std": 694.1858520507812, "step": 595 }, { "KL/chosen_KL_mean": -815.898193359375, "KL/mean": -957.6881103515625, "KL/rejected_KL_mean": -1099.47802734375, "KL/std": 608.5431518554688, "epoch": 0.8751835535976505, "fcm_dpo/beta": 0.0007497658370994031, "fcm_dpo/delta": 0.05895322561264038, "fcm_dpo/margin": 283.5798645019531, "fcm_dpo/q_t": 0.4523196518421173, "grad_norm": 73.0402603149414, "learning_rate": 2.396839494982103e-08, "logits/chosen": -1.0392305850982666, "logits/rejected": -1.003946304321289, "logps/chosen": -875.6898803710938, "logps/ref_chosen": -59.791683197021484, "logps/ref_rejected": -80.09111785888672, "logps/rejected": -1179.569091796875, "loss": 1.2571, "margin_dpo/margin_mean": 283.5798645019531, "margin_dpo/margin_std": 738.8785400390625, "step": 596 }, { "KL/chosen_KL_mean": -754.0137939453125, "KL/mean": -1037.7945556640625, "KL/rejected_KL_mean": -1321.5755615234375, "KL/std": 707.1693725585938, "epoch": 0.8766519823788547, "fcm_dpo/beta": 0.000738929957151413, "fcm_dpo/delta": -0.0748884379863739, "fcm_dpo/margin": 567.561767578125, "fcm_dpo/q_t": 0.405043363571167, "grad_norm": 26.4957218170166, "learning_rate": 2.3423053240837514e-08, "logits/chosen": -0.9764111042022705, "logits/rejected": -1.030094861984253, "logps/chosen": -811.2745361328125, "logps/ref_chosen": -57.26078796386719, "logps/ref_rejected": -100.6937255859375, "logps/rejected": -1422.269287109375, "loss": 1.0919, "margin_dpo/margin_mean": 567.561767578125, "margin_dpo/margin_std": 766.98095703125, "step": 597 }, { "KL/chosen_KL_mean": -748.9894409179688, "KL/mean": -973.43359375, "KL/rejected_KL_mean": -1197.8779296875, "KL/std": 626.2220458984375, "epoch": 0.8781204111600588, "fcm_dpo/beta": 0.0007343441247940063, "fcm_dpo/delta": -0.058417148888111115, "fcm_dpo/margin": 448.888427734375, "fcm_dpo/q_t": 0.42271342873573303, "grad_norm": 28.324201583862305, "learning_rate": 2.2883682045119062e-08, "logits/chosen": -1.1111347675323486, "logits/rejected": -1.1283093690872192, "logps/chosen": -801.5079345703125, "logps/ref_chosen": -52.51850509643555, "logps/ref_rejected": -89.44385528564453, "logps/rejected": -1287.32177734375, "loss": 1.1504, "margin_dpo/margin_mean": 448.888427734375, "margin_dpo/margin_std": 702.4248657226562, "step": 598 }, { "KL/chosen_KL_mean": -776.1297607421875, "KL/mean": -960.819580078125, "KL/rejected_KL_mean": -1145.50927734375, "KL/std": 598.0007934570312, "epoch": 0.8795888399412628, "fcm_dpo/beta": 0.0007289372151717544, "fcm_dpo/delta": -0.03695056587457657, "fcm_dpo/margin": 369.37945556640625, "fcm_dpo/q_t": 0.43398356437683105, "grad_norm": 35.090877532958984, "learning_rate": 2.2350295575598367e-08, "logits/chosen": -1.0137587785720825, "logits/rejected": -1.0232412815093994, "logps/chosen": -825.9324951171875, "logps/ref_chosen": -49.802677154541016, "logps/ref_rejected": -82.978515625, "logps/rejected": -1228.48779296875, "loss": 1.1817, "margin_dpo/margin_mean": 369.3795166015625, "margin_dpo/margin_std": 569.4313354492188, "step": 599 }, { "KL/chosen_KL_mean": -844.0216064453125, "KL/mean": -1028.517578125, "KL/rejected_KL_mean": -1213.0135498046875, "KL/std": 647.4127197265625, "epoch": 0.8810572687224669, "fcm_dpo/beta": 0.0007366039790213108, "fcm_dpo/delta": 0.08046019077301025, "fcm_dpo/margin": 368.991943359375, "fcm_dpo/q_t": 0.4391680657863617, "grad_norm": 28.268569946289062, "learning_rate": 2.1822907887504932e-08, "logits/chosen": -1.1482144594192505, "logits/rejected": -1.1489068269729614, "logps/chosen": -910.4564208984375, "logps/ref_chosen": -66.43487548828125, "logps/ref_rejected": -85.45649719238281, "logps/rejected": -1298.469970703125, "loss": 1.2098, "margin_dpo/margin_mean": 368.991943359375, "margin_dpo/margin_std": 782.9544677734375, "step": 600 }, { "KL/chosen_KL_mean": -827.043701171875, "KL/mean": -1054.9534912109375, "KL/rejected_KL_mean": -1282.863037109375, "KL/std": 646.4471435546875, "epoch": 0.882525697503671, "fcm_dpo/beta": 0.0007413103012368083, "fcm_dpo/delta": 0.012507637962698936, "fcm_dpo/margin": 455.8194274902344, "fcm_dpo/q_t": 0.42060860991477966, "grad_norm": 33.074012756347656, "learning_rate": 2.1301532877994742e-08, "logits/chosen": -1.059302568435669, "logits/rejected": -1.0803344249725342, "logps/chosen": -886.1773681640625, "logps/ref_chosen": -59.13361358642578, "logps/ref_rejected": -94.69093322753906, "logps/rejected": -1377.5540771484375, "loss": 1.1316, "margin_dpo/margin_mean": 455.81939697265625, "margin_dpo/margin_std": 648.2555541992188, "step": 601 }, { "KL/chosen_KL_mean": -575.5924072265625, "KL/mean": -873.7586669921875, "KL/rejected_KL_mean": -1171.925048828125, "KL/std": 629.0206298828125, "epoch": 0.8839941262848752, "fcm_dpo/beta": 0.0007364454213529825, "fcm_dpo/delta": -0.09336760640144348, "fcm_dpo/margin": 596.3326416015625, "fcm_dpo/q_t": 0.3994067311286926, "grad_norm": 68.0593032836914, "learning_rate": 2.0786184285784298e-08, "logits/chosen": -1.0809543132781982, "logits/rejected": -1.1263089179992676, "logps/chosen": -624.1859130859375, "logps/ref_chosen": -48.59352111816406, "logps/ref_rejected": -87.6685562133789, "logps/rejected": -1259.593505859375, "loss": 1.0572, "margin_dpo/margin_mean": 596.3326416015625, "margin_dpo/margin_std": 703.1005859375, "step": 602 }, { "KL/chosen_KL_mean": -738.450927734375, "KL/mean": -995.5992431640625, "KL/rejected_KL_mean": -1252.74755859375, "KL/std": 682.3411865234375, "epoch": 0.8854625550660793, "fcm_dpo/beta": 0.0007258434779942036, "fcm_dpo/delta": -0.024222414940595627, "fcm_dpo/margin": 514.296630859375, "fcm_dpo/q_t": 0.41643890738487244, "grad_norm": 37.625732421875, "learning_rate": 2.0276875690788204e-08, "logits/chosen": -1.057875156402588, "logits/rejected": -1.049300193786621, "logps/chosen": -808.8655395507812, "logps/ref_chosen": -70.41461944580078, "logps/ref_rejected": -100.32559967041016, "logps/rejected": -1353.0731201171875, "loss": 1.1244, "margin_dpo/margin_mean": 514.2965698242188, "margin_dpo/margin_std": 806.4513549804688, "step": 603 }, { "KL/chosen_KL_mean": -721.03076171875, "KL/mean": -1012.0377197265625, "KL/rejected_KL_mean": -1303.0445556640625, "KL/std": 684.166015625, "epoch": 0.8869309838472834, "fcm_dpo/beta": 0.0007178003434091806, "fcm_dpo/delta": -0.0707206279039383, "fcm_dpo/margin": 582.013916015625, "fcm_dpo/q_t": 0.406802773475647, "grad_norm": 36.53494644165039, "learning_rate": 1.977362051376158e-08, "logits/chosen": -1.0587637424468994, "logits/rejected": -1.1084504127502441, "logps/chosen": -767.4888305664062, "logps/ref_chosen": -46.45808029174805, "logps/ref_rejected": -91.8544921875, "logps/rejected": -1394.8990478515625, "loss": 1.098, "margin_dpo/margin_mean": 582.0138549804688, "margin_dpo/margin_std": 858.50927734375, "step": 604 }, { "KL/chosen_KL_mean": -756.2188720703125, "KL/mean": -961.7646484375, "KL/rejected_KL_mean": -1167.310546875, "KL/std": 611.2559814453125, "epoch": 0.8883994126284875, "fcm_dpo/beta": 0.0007192428456619382, "fcm_dpo/delta": 0.055984023958444595, "fcm_dpo/margin": 411.09173583984375, "fcm_dpo/q_t": 0.43355774879455566, "grad_norm": 26.453216552734375, "learning_rate": 1.9276432015946446e-08, "logits/chosen": -1.0269505977630615, "logits/rejected": -1.0451889038085938, "logps/chosen": -822.4681396484375, "logps/ref_chosen": -66.24933624267578, "logps/ref_rejected": -102.30496978759766, "logps/rejected": -1269.615478515625, "loss": 1.1786, "margin_dpo/margin_mean": 411.09173583984375, "margin_dpo/margin_std": 747.6804809570312, "step": 605 }, { "KL/chosen_KL_mean": -741.6259765625, "KL/mean": -970.2193603515625, "KL/rejected_KL_mean": -1198.8126220703125, "KL/std": 623.1279296875, "epoch": 0.8898678414096917, "fcm_dpo/beta": 0.0007271652575582266, "fcm_dpo/delta": 0.017948858439922333, "fcm_dpo/margin": 457.1866760253906, "fcm_dpo/q_t": 0.4236467778682709, "grad_norm": 27.061796188354492, "learning_rate": 1.8785323298722093e-08, "logits/chosen": -1.0122807025909424, "logits/rejected": -1.0320955514907837, "logps/chosen": -796.445068359375, "logps/ref_chosen": -54.819122314453125, "logps/ref_rejected": -98.37146759033203, "logps/rejected": -1297.18408203125, "loss": 1.1365, "margin_dpo/margin_mean": 457.1866760253906, "margin_dpo/margin_std": 670.32568359375, "step": 606 }, { "KL/chosen_KL_mean": -794.986083984375, "KL/mean": -985.2572021484375, "KL/rejected_KL_mean": -1175.5283203125, "KL/std": 660.7117919921875, "epoch": 0.8913362701908958, "fcm_dpo/beta": 0.0007356663700193167, "fcm_dpo/delta": 0.0717119500041008, "fcm_dpo/margin": 380.5422668457031, "fcm_dpo/q_t": 0.43587759137153625, "grad_norm": 24.40001106262207, "learning_rate": 1.8300307303259904e-08, "logits/chosen": -1.0124876499176025, "logits/rejected": -1.0030550956726074, "logps/chosen": -853.070068359375, "logps/ref_chosen": -58.08403778076172, "logps/ref_rejected": -79.777099609375, "logps/rejected": -1255.305419921875, "loss": 1.191, "margin_dpo/margin_mean": 380.542236328125, "margin_dpo/margin_std": 708.0584106445312, "step": 607 }, { "KL/chosen_KL_mean": -712.60595703125, "KL/mean": -935.021484375, "KL/rejected_KL_mean": -1157.4371337890625, "KL/std": 576.3223876953125, "epoch": 0.8928046989720999, "fcm_dpo/beta": 0.0007404502248391509, "fcm_dpo/delta": 0.02115270122885704, "fcm_dpo/margin": 444.8311767578125, "fcm_dpo/q_t": 0.4227282702922821, "grad_norm": 26.049560546875, "learning_rate": 1.7821396810182437e-08, "logits/chosen": -1.084040880203247, "logits/rejected": -1.1021060943603516, "logps/chosen": -770.0567626953125, "logps/ref_chosen": -57.450836181640625, "logps/ref_rejected": -94.77339172363281, "logps/rejected": -1252.21044921875, "loss": 1.133, "margin_dpo/margin_mean": 444.8311767578125, "margin_dpo/margin_std": 611.6470947265625, "step": 608 }, { "KL/chosen_KL_mean": -711.9459228515625, "KL/mean": -1001.8514404296875, "KL/rejected_KL_mean": -1291.7567138671875, "KL/std": 775.0626220703125, "epoch": 0.8942731277533039, "fcm_dpo/beta": 0.0007308817002922297, "fcm_dpo/delta": -0.0770505964756012, "fcm_dpo/margin": 579.810791015625, "fcm_dpo/q_t": 0.408769428730011, "grad_norm": 27.387683868408203, "learning_rate": 1.7348604439226617e-08, "logits/chosen": -1.1206148862838745, "logits/rejected": -1.1497302055358887, "logps/chosen": -770.7513427734375, "logps/ref_chosen": -58.805355072021484, "logps/ref_rejected": -88.81600952148438, "logps/rejected": -1380.57275390625, "loss": 1.104, "margin_dpo/margin_mean": 579.810791015625, "margin_dpo/margin_std": 940.2999877929688, "step": 609 }, { "KL/chosen_KL_mean": -716.533447265625, "KL/mean": -895.5557861328125, "KL/rejected_KL_mean": -1074.578125, "KL/std": 590.9163818359375, "epoch": 0.895741556534508, "fcm_dpo/beta": 0.0007388823432847857, "fcm_dpo/delta": 0.08757883310317993, "fcm_dpo/margin": 358.044677734375, "fcm_dpo/q_t": 0.43797656893730164, "grad_norm": 45.32019805908203, "learning_rate": 1.6881942648911074e-08, "logits/chosen": -0.9885549545288086, "logits/rejected": -0.9654494524002075, "logps/chosen": -782.228515625, "logps/ref_chosen": -65.69503784179688, "logps/ref_rejected": -83.40538787841797, "logps/rejected": -1157.9835205078125, "loss": 1.2023, "margin_dpo/margin_mean": 358.044677734375, "margin_dpo/margin_std": 698.3663330078125, "step": 610 }, { "KL/chosen_KL_mean": -749.7844848632812, "KL/mean": -1059.1370849609375, "KL/rejected_KL_mean": -1368.4896240234375, "KL/std": 790.021728515625, "epoch": 0.8972099853157122, "fcm_dpo/beta": 0.0007288827328011394, "fcm_dpo/delta": -0.10581991076469421, "fcm_dpo/margin": 618.7051391601562, "fcm_dpo/q_t": 0.4007767140865326, "grad_norm": 31.054868698120117, "learning_rate": 1.6421423736208e-08, "logits/chosen": -1.0566998720169067, "logits/rejected": -1.1077455282211304, "logps/chosen": -802.3839111328125, "logps/ref_chosen": -52.59946823120117, "logps/ref_rejected": -86.33099365234375, "logps/rejected": -1454.820556640625, "loss": 1.0823, "margin_dpo/margin_mean": 618.7051391601562, "margin_dpo/margin_std": 904.1781616210938, "step": 611 }, { "KL/chosen_KL_mean": -812.09033203125, "KL/mean": -1034.2083740234375, "KL/rejected_KL_mean": -1256.326416015625, "KL/std": 622.982177734375, "epoch": 0.8986784140969163, "fcm_dpo/beta": 0.0007307927589863539, "fcm_dpo/delta": 0.02587536908686161, "fcm_dpo/margin": 444.236083984375, "fcm_dpo/q_t": 0.42509013414382935, "grad_norm": 33.46781539916992, "learning_rate": 1.5967059836219042e-08, "logits/chosen": -1.0426099300384521, "logits/rejected": -1.0452769994735718, "logps/chosen": -871.4140625, "logps/ref_chosen": -59.32372283935547, "logps/ref_rejected": -88.31239318847656, "logps/rejected": -1344.638916015625, "loss": 1.1473, "margin_dpo/margin_mean": 444.23602294921875, "margin_dpo/margin_std": 683.746337890625, "step": 612 }, { "KL/chosen_KL_mean": -720.933837890625, "KL/mean": -993.418212890625, "KL/rejected_KL_mean": -1265.902587890625, "KL/std": 689.9868774414062, "epoch": 0.9001468428781204, "fcm_dpo/beta": 0.0007219946710392833, "fcm_dpo/delta": -0.045596893876791, "fcm_dpo/margin": 544.9688110351562, "fcm_dpo/q_t": 0.40862947702407837, "grad_norm": 33.90407180786133, "learning_rate": 1.551886292185553e-08, "logits/chosen": -1.073573112487793, "logits/rejected": -1.12931227684021, "logps/chosen": -780.663818359375, "logps/ref_chosen": -59.72996520996094, "logps/ref_rejected": -105.10752868652344, "logps/rejected": -1371.01025390625, "loss": 1.0944, "margin_dpo/margin_mean": 544.9688110351562, "margin_dpo/margin_std": 720.446044921875, "step": 613 }, { "KL/chosen_KL_mean": -778.2044677734375, "KL/mean": -1059.3028564453125, "KL/rejected_KL_mean": -1340.4013671875, "KL/std": 686.830078125, "epoch": 0.9016152716593245, "fcm_dpo/beta": 0.0007176370127126575, "fcm_dpo/delta": -0.05566265806555748, "fcm_dpo/margin": 562.1969604492188, "fcm_dpo/q_t": 0.40819481015205383, "grad_norm": 38.53693771362305, "learning_rate": 1.507684480352292e-08, "logits/chosen": -1.0261731147766113, "logits/rejected": -1.1077706813812256, "logps/chosen": -831.1434326171875, "logps/ref_chosen": -52.93898010253906, "logps/ref_rejected": -104.67938232421875, "logps/rejected": -1445.080810546875, "loss": 1.1008, "margin_dpo/margin_mean": 562.1968994140625, "margin_dpo/margin_std": 811.7864379882812, "step": 614 }, { "KL/chosen_KL_mean": -728.3233642578125, "KL/mean": -960.8916015625, "KL/rejected_KL_mean": -1193.4598388671875, "KL/std": 708.7926025390625, "epoch": 0.9030837004405287, "fcm_dpo/beta": 0.0007183545967563987, "fcm_dpo/delta": 0.016210701316595078, "fcm_dpo/margin": 465.13641357421875, "fcm_dpo/q_t": 0.4231938123703003, "grad_norm": 28.220327377319336, "learning_rate": 1.4641017128809801e-08, "logits/chosen": -1.0408499240875244, "logits/rejected": -1.0693387985229492, "logps/chosen": -794.1406860351562, "logps/ref_chosen": -65.81727600097656, "logps/ref_rejected": -95.17749786376953, "logps/rejected": -1288.6373291015625, "loss": 1.15, "margin_dpo/margin_mean": 465.13641357421875, "margin_dpo/margin_std": 769.0079956054688, "step": 615 }, { "KL/chosen_KL_mean": -863.922119140625, "KL/mean": -1049.33056640625, "KL/rejected_KL_mean": -1234.73876953125, "KL/std": 591.33251953125, "epoch": 0.9045521292217328, "fcm_dpo/beta": 0.0007266043685376644, "fcm_dpo/delta": 0.08268511295318604, "fcm_dpo/margin": 370.8167419433594, "fcm_dpo/q_t": 0.4374980330467224, "grad_norm": 35.82476043701172, "learning_rate": 1.4211391382180637e-08, "logits/chosen": -1.0865511894226074, "logits/rejected": -1.0711218118667603, "logps/chosen": -929.0549926757812, "logps/ref_chosen": -65.13285827636719, "logps/ref_rejected": -74.70050048828125, "logps/rejected": -1309.439453125, "loss": 1.1912, "margin_dpo/margin_mean": 370.8167724609375, "margin_dpo/margin_std": 663.240234375, "step": 616 }, { "KL/chosen_KL_mean": -840.3502197265625, "KL/mean": -974.4063720703125, "KL/rejected_KL_mean": -1108.4625244140625, "KL/std": 581.7191772460938, "epoch": 0.9060205580029369, "fcm_dpo/beta": 0.0007288898341357708, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 268.1123046875, "fcm_dpo/q_t": 0.4547814726829529, "grad_norm": 50.21003723144531, "learning_rate": 1.378797888467345e-08, "logits/chosen": -1.0201683044433594, "logits/rejected": -0.9916302561759949, "logps/chosen": -903.3557739257812, "logps/ref_chosen": -63.005550384521484, "logps/ref_rejected": -64.234130859375, "logps/rejected": -1172.6966552734375, "loss": 1.2642, "margin_dpo/margin_mean": 268.1123046875, "margin_dpo/margin_std": 707.6526489257812, "step": 617 }, { "KL/chosen_KL_mean": -859.969482421875, "KL/mean": -1152.5023193359375, "KL/rejected_KL_mean": -1445.03515625, "KL/std": 750.560791015625, "epoch": 0.9074889867841409, "fcm_dpo/beta": 0.0007248672191053629, "fcm_dpo/delta": -0.07764283567667007, "fcm_dpo/margin": 585.065673828125, "fcm_dpo/q_t": 0.4056781530380249, "grad_norm": 38.26071548461914, "learning_rate": 1.3370790793601371e-08, "logits/chosen": -1.059650182723999, "logits/rejected": -1.103996753692627, "logps/chosen": -927.0708618164062, "logps/ref_chosen": -67.10134887695312, "logps/ref_rejected": -92.15340423583984, "logps/rejected": -1537.1884765625, "loss": 1.1236, "margin_dpo/margin_mean": 585.065673828125, "margin_dpo/margin_std": 982.0968017578125, "step": 618 }, { "KL/chosen_KL_mean": -856.5992431640625, "KL/mean": -1078.2330322265625, "KL/rejected_KL_mean": -1299.86669921875, "KL/std": 716.7399291992188, "epoch": 0.908957415565345, "fcm_dpo/beta": 0.0007212937925942242, "fcm_dpo/delta": 0.03125857189297676, "fcm_dpo/margin": 443.2675476074219, "fcm_dpo/q_t": 0.4321790337562561, "grad_norm": 46.859371185302734, "learning_rate": 1.2959838102258535e-08, "logits/chosen": -1.0055409669876099, "logits/rejected": -1.019460916519165, "logps/chosen": -912.5774536132812, "logps/ref_chosen": -55.978233337402344, "logps/ref_rejected": -93.1854019165039, "logps/rejected": -1393.0521240234375, "loss": 1.2044, "margin_dpo/margin_mean": 443.2675476074219, "margin_dpo/margin_std": 974.4349365234375, "step": 619 }, { "KL/chosen_KL_mean": -782.436279296875, "KL/mean": -988.9633178710938, "KL/rejected_KL_mean": -1195.490478515625, "KL/std": 634.9549560546875, "epoch": 0.9104258443465492, "fcm_dpo/beta": 0.0007255082600750029, "fcm_dpo/delta": 0.051843997091054916, "fcm_dpo/margin": 413.05419921875, "fcm_dpo/q_t": 0.4310418963432312, "grad_norm": 33.33245086669922, "learning_rate": 1.2555131639630567e-08, "logits/chosen": -1.0578522682189941, "logits/rejected": -1.0688400268554688, "logps/chosen": -842.2337646484375, "logps/ref_chosen": -59.79750061035156, "logps/ref_rejected": -78.41075134277344, "logps/rejected": -1273.9012451171875, "loss": 1.1738, "margin_dpo/margin_mean": 413.05419921875, "margin_dpo/margin_std": 718.75439453125, "step": 620 }, { "KL/chosen_KL_mean": -774.94873046875, "KL/mean": -1094.044677734375, "KL/rejected_KL_mean": -1413.1407470703125, "KL/std": 769.3406982421875, "epoch": 0.9118942731277533, "fcm_dpo/beta": 0.0007223983411677182, "fcm_dpo/delta": -0.11662234365940094, "fcm_dpo/margin": 638.1920166015625, "fcm_dpo/q_t": 0.3942115306854248, "grad_norm": 33.19838333129883, "learning_rate": 1.2156682070109086e-08, "logits/chosen": -1.1112885475158691, "logits/rejected": -1.1674792766571045, "logps/chosen": -828.8825073242188, "logps/ref_chosen": -53.93375778198242, "logps/ref_rejected": -88.36951446533203, "logps/rejected": -1501.51025390625, "loss": 1.0704, "margin_dpo/margin_mean": 638.1920166015625, "margin_dpo/margin_std": 859.8441772460938, "step": 621 }, { "KL/chosen_KL_mean": -750.149169921875, "KL/mean": -967.1200561523438, "KL/rejected_KL_mean": -1184.0909423828125, "KL/std": 593.5496826171875, "epoch": 0.9133627019089574, "fcm_dpo/beta": 0.0007141625974327326, "fcm_dpo/delta": 0.04117283225059509, "fcm_dpo/margin": 433.9416809082031, "fcm_dpo/q_t": 0.4296455979347229, "grad_norm": 34.48857116699219, "learning_rate": 1.1764499893210878e-08, "logits/chosen": -0.9647907018661499, "logits/rejected": -0.9519217014312744, "logps/chosen": -810.43505859375, "logps/ref_chosen": -60.28582000732422, "logps/ref_rejected": -85.51873779296875, "logps/rejected": -1269.609619140625, "loss": 1.1662, "margin_dpo/margin_mean": 433.9416809082031, "margin_dpo/margin_std": 747.2349853515625, "step": 622 }, { "KL/chosen_KL_mean": -820.90869140625, "KL/mean": -982.7512817382812, "KL/rejected_KL_mean": -1144.5938720703125, "KL/std": 598.2750244140625, "epoch": 0.9148311306901615, "fcm_dpo/beta": 0.0007311557419598103, "fcm_dpo/delta": 0.11611227691173553, "fcm_dpo/margin": 323.68511962890625, "fcm_dpo/q_t": 0.446946382522583, "grad_norm": 42.85483932495117, "learning_rate": 1.1378595443300998e-08, "logits/chosen": -1.119500756263733, "logits/rejected": -1.1187694072723389, "logps/chosen": -885.065673828125, "logps/ref_chosen": -64.1569595336914, "logps/ref_rejected": -85.08304595947266, "logps/rejected": -1229.6768798828125, "loss": 1.2308, "margin_dpo/margin_mean": 323.68511962890625, "margin_dpo/margin_std": 731.9136962890625, "step": 623 }, { "KL/chosen_KL_mean": -821.5892944335938, "KL/mean": -1092.430419921875, "KL/rejected_KL_mean": -1363.271484375, "KL/std": 638.8692016601562, "epoch": 0.9162995594713657, "fcm_dpo/beta": 0.0007304972968995571, "fcm_dpo/delta": -0.047577153891325, "fcm_dpo/margin": 541.6821899414062, "fcm_dpo/q_t": 0.4065777659416199, "grad_norm": 38.01239776611328, "learning_rate": 1.0998978889320582e-08, "logits/chosen": -1.1294810771942139, "logits/rejected": -1.137779951095581, "logps/chosen": -893.5079345703125, "logps/ref_chosen": -71.91862487792969, "logps/ref_rejected": -97.13203430175781, "logps/rejected": -1460.403564453125, "loss": 1.0894, "margin_dpo/margin_mean": 541.6821899414062, "margin_dpo/margin_std": 687.8084716796875, "step": 624 }, { "KL/chosen_KL_mean": -741.5074462890625, "KL/mean": -1046.6978759765625, "KL/rejected_KL_mean": -1351.88818359375, "KL/std": 684.56396484375, "epoch": 0.9177679882525698, "fcm_dpo/beta": 0.0007204815046861768, "fcm_dpo/delta": -0.09379212558269501, "fcm_dpo/margin": 610.380859375, "fcm_dpo/q_t": 0.39741039276123047, "grad_norm": 39.902103424072266, "learning_rate": 1.0625660234518913e-08, "logits/chosen": -1.004154920578003, "logits/rejected": -1.0372995138168335, "logps/chosen": -799.8494873046875, "logps/ref_chosen": -58.342071533203125, "logps/ref_rejected": -86.09038543701172, "logps/rejected": -1437.978515625, "loss": 1.0431, "margin_dpo/margin_mean": 610.3807983398438, "margin_dpo/margin_std": 639.5909423828125, "step": 625 }, { "KL/chosen_KL_mean": -940.815185546875, "KL/mean": -1116.69189453125, "KL/rejected_KL_mean": -1292.5687255859375, "KL/std": 766.899169921875, "epoch": 0.9192364170337739, "fcm_dpo/beta": 0.0007251551141962409, "fcm_dpo/delta": 0.09746433794498444, "fcm_dpo/margin": 351.75341796875, "fcm_dpo/q_t": 0.439464807510376, "grad_norm": 44.423484802246094, "learning_rate": 1.0258649316189721e-08, "logits/chosen": -0.9951568841934204, "logits/rejected": -0.9892057776451111, "logps/chosen": -1015.9278564453125, "logps/ref_chosen": -75.11260986328125, "logps/ref_rejected": -99.188720703125, "logps/rejected": -1391.7574462890625, "loss": 1.2212, "margin_dpo/margin_mean": 351.75341796875, "margin_dpo/margin_std": 743.8319702148438, "step": 626 }, { "KL/chosen_KL_mean": -655.391357421875, "KL/mean": -983.7683715820312, "KL/rejected_KL_mean": -1312.145263671875, "KL/std": 831.3623046875, "epoch": 0.920704845814978, "fcm_dpo/beta": 0.0007200604304671288, "fcm_dpo/delta": -0.12902742624282837, "fcm_dpo/margin": 656.75390625, "fcm_dpo/q_t": 0.40084049105644226, "grad_norm": 29.196271896362305, "learning_rate": 9.897955805412e-09, "logits/chosen": -0.9449664354324341, "logits/rejected": -1.0268559455871582, "logps/chosen": -703.134521484375, "logps/ref_chosen": -47.74314880371094, "logps/ref_rejected": -106.75448608398438, "logps/rejected": -1418.8997802734375, "loss": 1.0784, "margin_dpo/margin_mean": 656.7538452148438, "margin_dpo/margin_std": 977.4828491210938, "step": 627 }, { "KL/chosen_KL_mean": -853.753662109375, "KL/mean": -1100.697265625, "KL/rejected_KL_mean": -1347.640869140625, "KL/std": 691.8834228515625, "epoch": 0.922173274596182, "fcm_dpo/beta": 0.0007095912005752325, "fcm_dpo/delta": -0.0005148975178599358, "fcm_dpo/margin": 493.88702392578125, "fcm_dpo/q_t": 0.4203591048717499, "grad_norm": 30.30500602722168, "learning_rate": 9.543589206795238e-09, "logits/chosen": -1.106518268585205, "logits/rejected": -1.131546139717102, "logps/chosen": -913.9365844726562, "logps/ref_chosen": -60.182945251464844, "logps/ref_rejected": -101.55467224121094, "logps/rejected": -1449.1954345703125, "loss": 1.1425, "margin_dpo/margin_mean": 493.88702392578125, "margin_dpo/margin_std": 802.3726196289062, "step": 628 }, { "KL/chosen_KL_mean": -840.9974365234375, "KL/mean": -1075.9169921875, "KL/rejected_KL_mean": -1310.8365478515625, "KL/std": 669.786376953125, "epoch": 0.9236417033773862, "fcm_dpo/beta": 0.0007125267293304205, "fcm_dpo/delta": 0.015744555741548538, "fcm_dpo/margin": 469.839111328125, "fcm_dpo/q_t": 0.42130935192108154, "grad_norm": 34.061126708984375, "learning_rate": 9.19555885822887e-09, "logits/chosen": -1.0905547142028809, "logits/rejected": -1.110518455505371, "logps/chosen": -905.2109375, "logps/ref_chosen": -64.21354675292969, "logps/ref_rejected": -91.65367126464844, "logps/rejected": -1402.490234375, "loss": 1.1358, "margin_dpo/margin_mean": 469.839111328125, "margin_dpo/margin_std": 684.5360107421875, "step": 629 }, { "KL/chosen_KL_mean": -732.7456665039062, "KL/mean": -860.0938720703125, "KL/rejected_KL_mean": -987.442138671875, "KL/std": 659.4423828125, "epoch": 0.9251101321585903, "fcm_dpo/beta": 0.000715771981049329, "fcm_dpo/delta": 0.04146602377295494, "fcm_dpo/margin": 254.6964874267578, "fcm_dpo/q_t": 0.46173208951950073, "grad_norm": 44.93278503417969, "learning_rate": 8.85387393063622e-09, "logits/chosen": -1.049222469329834, "logits/rejected": -1.0251140594482422, "logps/chosen": -792.0366821289062, "logps/ref_chosen": -59.29100036621094, "logps/ref_rejected": -83.59829711914062, "logps/rejected": -1071.04052734375, "loss": 1.2901, "margin_dpo/margin_mean": 254.6964874267578, "margin_dpo/margin_std": 802.978271484375, "step": 630 }, { "KL/chosen_KL_mean": -897.2791748046875, "KL/mean": -1110.93359375, "KL/rejected_KL_mean": -1324.5877685546875, "KL/std": 662.3657836914062, "epoch": 0.9265785609397944, "fcm_dpo/beta": 0.0007242325227707624, "fcm_dpo/delta": 0.041726578027009964, "fcm_dpo/margin": 427.30865478515625, "fcm_dpo/q_t": 0.4278268814086914, "grad_norm": 31.293115615844727, "learning_rate": 8.518543427732949e-09, "logits/chosen": -1.1521632671356201, "logits/rejected": -1.1659929752349854, "logps/chosen": -956.7327880859375, "logps/ref_chosen": -59.45360565185547, "logps/ref_rejected": -80.95156860351562, "logps/rejected": -1405.539306640625, "loss": 1.174, "margin_dpo/margin_mean": 427.3086242675781, "margin_dpo/margin_std": 762.1250610351562, "step": 631 }, { "KL/chosen_KL_mean": -787.10400390625, "KL/mean": -995.85400390625, "KL/rejected_KL_mean": -1204.6038818359375, "KL/std": 610.8785400390625, "epoch": 0.9280469897209985, "fcm_dpo/beta": 0.0007272514048963785, "fcm_dpo/delta": 0.04772178828716278, "fcm_dpo/margin": 417.49993896484375, "fcm_dpo/q_t": 0.4296777546405792, "grad_norm": 37.82929229736328, "learning_rate": 8.189576185789637e-09, "logits/chosen": -1.1057119369506836, "logits/rejected": -1.1097162961959839, "logps/chosen": -848.45556640625, "logps/ref_chosen": -61.35155487060547, "logps/ref_rejected": -86.16017150878906, "logps/rejected": -1290.7640380859375, "loss": 1.1786, "margin_dpo/margin_mean": 417.49993896484375, "margin_dpo/margin_std": 755.839111328125, "step": 632 }, { "KL/chosen_KL_mean": -875.8489379882812, "KL/mean": -1027.587890625, "KL/rejected_KL_mean": -1179.326904296875, "KL/std": 598.5611572265625, "epoch": 0.9295154185022027, "fcm_dpo/beta": 0.0007348891813308001, "fcm_dpo/delta": 0.043210797011852264, "fcm_dpo/margin": 303.47796630859375, "fcm_dpo/q_t": 0.44790202379226685, "grad_norm": 37.137454986572266, "learning_rate": 7.866980873399015e-09, "logits/chosen": -1.1341545581817627, "logits/rejected": -1.1472094058990479, "logps/chosen": -933.1270751953125, "logps/ref_chosen": -57.27816390991211, "logps/ref_rejected": -91.58395385742188, "logps/rejected": -1270.910888671875, "loss": 1.2351, "margin_dpo/margin_mean": 303.4779968261719, "margin_dpo/margin_std": 675.3363037109375, "step": 633 }, { "KL/chosen_KL_mean": -960.07275390625, "KL/mean": -1097.663330078125, "KL/rejected_KL_mean": -1235.254150390625, "KL/std": 704.655517578125, "epoch": 0.9309838472834068, "fcm_dpo/beta": 0.0007380647002719343, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 275.1812744140625, "fcm_dpo/q_t": 0.4533958435058594, "grad_norm": 27.810382843017578, "learning_rate": 7.550765991247654e-09, "logits/chosen": -1.0175862312316895, "logits/rejected": -1.0139384269714355, "logps/chosen": -1026.6917724609375, "logps/ref_chosen": -66.61896514892578, "logps/ref_rejected": -107.12564849853516, "logps/rejected": -1342.379638671875, "loss": 1.2542, "margin_dpo/margin_mean": 275.1812744140625, "margin_dpo/margin_std": 683.193115234375, "step": 634 }, { "KL/chosen_KL_mean": -855.4681396484375, "KL/mean": -1050.290771484375, "KL/rejected_KL_mean": -1245.11328125, "KL/std": 766.2486572265625, "epoch": 0.9324522760646109, "fcm_dpo/beta": 0.0007452410645782948, "fcm_dpo/delta": 0.06136660277843475, "fcm_dpo/margin": 389.6451721191406, "fcm_dpo/q_t": 0.4358566403388977, "grad_norm": 46.93721008300781, "learning_rate": 7.240939871891699e-09, "logits/chosen": -1.081568717956543, "logits/rejected": -1.064734697341919, "logps/chosen": -929.4236450195312, "logps/ref_chosen": -73.95551300048828, "logps/ref_rejected": -82.50045776367188, "logps/rejected": -1327.61376953125, "loss": 1.1961, "margin_dpo/margin_mean": 389.64520263671875, "margin_dpo/margin_std": 794.4766235351562, "step": 635 }, { "KL/chosen_KL_mean": -784.0025634765625, "KL/mean": -1020.8980712890625, "KL/rejected_KL_mean": -1257.793701171875, "KL/std": 716.466064453125, "epoch": 0.933920704845815, "fcm_dpo/beta": 0.0007509600836783648, "fcm_dpo/delta": -0.00671178475022316, "fcm_dpo/margin": 473.79107666015625, "fcm_dpo/q_t": 0.42060697078704834, "grad_norm": 27.554264068603516, "learning_rate": 6.937510679537628e-09, "logits/chosen": -0.9968219995498657, "logits/rejected": -0.9979686141014099, "logps/chosen": -843.6314697265625, "logps/ref_chosen": -59.628910064697266, "logps/ref_rejected": -81.97883605957031, "logps/rejected": -1339.7724609375, "loss": 1.1335, "margin_dpo/margin_mean": 473.79107666015625, "margin_dpo/margin_std": 754.580810546875, "step": 636 }, { "KL/chosen_KL_mean": -779.475341796875, "KL/mean": -1054.064453125, "KL/rejected_KL_mean": -1328.653564453125, "KL/std": 737.46728515625, "epoch": 0.9353891336270191, "fcm_dpo/beta": 0.0007367645157501101, "fcm_dpo/delta": -0.05750354379415512, "fcm_dpo/margin": 549.1782836914062, "fcm_dpo/q_t": 0.4083985686302185, "grad_norm": 25.79489517211914, "learning_rate": 6.640486409826785e-09, "logits/chosen": -1.1218140125274658, "logits/rejected": -1.1801137924194336, "logps/chosen": -829.1279907226562, "logps/ref_chosen": -49.652687072753906, "logps/ref_rejected": -98.40513610839844, "logps/rejected": -1427.0587158203125, "loss": 1.0945, "margin_dpo/margin_mean": 549.1782836914062, "margin_dpo/margin_std": 762.412841796875, "step": 637 }, { "KL/chosen_KL_mean": -761.9863891601562, "KL/mean": -976.8961791992188, "KL/rejected_KL_mean": -1191.8060302734375, "KL/std": 677.3199462890625, "epoch": 0.9368575624082232, "fcm_dpo/beta": 0.0007319189608097076, "fcm_dpo/delta": -0.07822597026824951, "fcm_dpo/margin": 429.8195495605469, "fcm_dpo/q_t": 0.4225819408893585, "grad_norm": 34.85166931152344, "learning_rate": 6.349874889624962e-09, "logits/chosen": -0.981365978717804, "logits/rejected": -0.9689379334449768, "logps/chosen": -820.14306640625, "logps/ref_chosen": -58.156639099121094, "logps/ref_rejected": -79.3014907836914, "logps/rejected": -1271.107421875, "loss": 1.1841, "margin_dpo/margin_mean": 429.8195495605469, "margin_dpo/margin_std": 802.29248046875, "step": 638 }, { "KL/chosen_KL_mean": -1073.5750732421875, "KL/mean": -1151.7091064453125, "KL/rejected_KL_mean": -1229.84326171875, "KL/std": 669.1240844726562, "epoch": 0.9383259911894273, "fcm_dpo/beta": 0.0007261935388669372, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 156.26809692382812, "fcm_dpo/q_t": 0.47418278455734253, "grad_norm": 120.29761505126953, "learning_rate": 6.065683776815933e-09, "logits/chosen": -0.9551519155502319, "logits/rejected": -0.8928138017654419, "logps/chosen": -1145.8983154296875, "logps/ref_chosen": -72.32319641113281, "logps/ref_rejected": -74.2749252319336, "logps/rejected": -1304.1181640625, "loss": 1.3672, "margin_dpo/margin_mean": 156.26809692382812, "margin_dpo/margin_std": 847.606201171875, "step": 639 }, { "KL/chosen_KL_mean": -795.9225463867188, "KL/mean": -1108.8028564453125, "KL/rejected_KL_mean": -1421.683349609375, "KL/std": 764.7296752929688, "epoch": 0.9397944199706314, "fcm_dpo/beta": 0.0007176444632932544, "fcm_dpo/delta": -0.10362177342176437, "fcm_dpo/margin": 625.7607421875, "fcm_dpo/q_t": 0.40179121494293213, "grad_norm": 35.442649841308594, "learning_rate": 5.7879205600998296e-09, "logits/chosen": -0.9908477663993835, "logits/rejected": -1.0309135913848877, "logps/chosen": -852.056884765625, "logps/ref_chosen": -56.13436508178711, "logps/ref_rejected": -108.60014343261719, "logps/rejected": -1530.283447265625, "loss": 1.0819, "margin_dpo/margin_mean": 625.7606201171875, "margin_dpo/margin_std": 917.645751953125, "step": 640 }, { "KL/chosen_KL_mean": -917.562744140625, "KL/mean": -1132.0595703125, "KL/rejected_KL_mean": -1346.556396484375, "KL/std": 655.5465087890625, "epoch": 0.9412628487518355, "fcm_dpo/beta": 0.000714754918590188, "fcm_dpo/delta": 0.04476023092865944, "fcm_dpo/margin": 428.99371337890625, "fcm_dpo/q_t": 0.43139177560806274, "grad_norm": 29.970199584960938, "learning_rate": 5.516592558795746e-09, "logits/chosen": -1.057366132736206, "logits/rejected": -1.0727591514587402, "logps/chosen": -982.5596313476562, "logps/ref_chosen": -64.99689483642578, "logps/ref_rejected": -86.99232482910156, "logps/rejected": -1433.548828125, "loss": 1.1919, "margin_dpo/margin_mean": 428.99371337890625, "margin_dpo/margin_std": 868.6700439453125, "step": 641 }, { "KL/chosen_KL_mean": -880.2049560546875, "KL/mean": -1136.517333984375, "KL/rejected_KL_mean": -1392.829833984375, "KL/std": 870.1756591796875, "epoch": 0.9427312775330396, "fcm_dpo/beta": 0.0007182994158938527, "fcm_dpo/delta": -0.01912054233253002, "fcm_dpo/margin": 512.6249389648438, "fcm_dpo/q_t": 0.42391371726989746, "grad_norm": 37.84080505371094, "learning_rate": 5.251706922648868e-09, "logits/chosen": -0.9898715019226074, "logits/rejected": -1.0340471267700195, "logps/chosen": -945.8941650390625, "logps/ref_chosen": -65.68924713134766, "logps/ref_rejected": -110.24205017089844, "logps/rejected": -1503.0718994140625, "loss": 1.1715, "margin_dpo/margin_mean": 512.625, "margin_dpo/margin_std": 1061.2357177734375, "step": 642 }, { "KL/chosen_KL_mean": -796.40869140625, "KL/mean": -983.58837890625, "KL/rejected_KL_mean": -1170.7681884765625, "KL/std": 615.07666015625, "epoch": 0.9441997063142438, "fcm_dpo/beta": 0.0007112710736691952, "fcm_dpo/delta": -0.025832796469330788, "fcm_dpo/margin": 374.35943603515625, "fcm_dpo/q_t": 0.43726229667663574, "grad_norm": 34.36274719238281, "learning_rate": 4.993270631642038e-09, "logits/chosen": -1.1126079559326172, "logits/rejected": -1.1123908758163452, "logps/chosen": -848.3587036132812, "logps/ref_chosen": -51.94999694824219, "logps/ref_rejected": -87.46833801269531, "logps/rejected": -1258.236572265625, "loss": 1.1891, "margin_dpo/margin_mean": 374.3594665527344, "margin_dpo/margin_std": 623.792724609375, "step": 643 }, { "KL/chosen_KL_mean": -790.0933227539062, "KL/mean": -986.8052978515625, "KL/rejected_KL_mean": -1183.517333984375, "KL/std": 741.110107421875, "epoch": 0.9456681350954479, "fcm_dpo/beta": 0.0007180780521593988, "fcm_dpo/delta": 0.0694977343082428, "fcm_dpo/margin": 393.4239807128906, "fcm_dpo/q_t": 0.43679407238960266, "grad_norm": 47.92481231689453, "learning_rate": 4.741290495811873e-09, "logits/chosen": -1.0378338098526, "logits/rejected": -1.0484693050384521, "logps/chosen": -849.1109619140625, "logps/ref_chosen": -59.017662048339844, "logps/ref_rejected": -87.13668823242188, "logps/rejected": -1270.654052734375, "loss": 1.2172, "margin_dpo/margin_mean": 393.4239501953125, "margin_dpo/margin_std": 875.6864013671875, "step": 644 }, { "KL/chosen_KL_mean": -823.27685546875, "KL/mean": -912.3017578125, "KL/rejected_KL_mean": -1001.3265380859375, "KL/std": 572.473876953125, "epoch": 0.947136563876652, "fcm_dpo/beta": 0.0007212264463305473, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 178.04962158203125, "fcm_dpo/q_t": 0.47089576721191406, "grad_norm": 86.09916687011719, "learning_rate": 4.495773155069299e-09, "logits/chosen": -1.0229761600494385, "logits/rejected": -1.0112075805664062, "logps/chosen": -879.1529541015625, "logps/ref_chosen": -55.87602233886719, "logps/ref_rejected": -97.78080749511719, "logps/rejected": -1099.1072998046875, "loss": 1.3404, "margin_dpo/margin_mean": 178.04962158203125, "margin_dpo/margin_std": 762.8621826171875, "step": 645 }, { "KL/chosen_KL_mean": -781.0083618164062, "KL/mean": -941.683837890625, "KL/rejected_KL_mean": -1102.3594970703125, "KL/std": 565.9122924804688, "epoch": 0.9486049926578561, "fcm_dpo/beta": 0.0007259917329065502, "fcm_dpo/delta": 0.032927513122558594, "fcm_dpo/margin": 321.35107421875, "fcm_dpo/q_t": 0.4461080729961395, "grad_norm": 55.66950225830078, "learning_rate": 4.256725079024553e-09, "logits/chosen": -1.0636550188064575, "logits/rejected": -1.0476542711257935, "logps/chosen": -842.2841796875, "logps/ref_chosen": -61.275787353515625, "logps/ref_rejected": -77.50580596923828, "logps/rejected": -1179.865234375, "loss": 1.2275, "margin_dpo/margin_mean": 321.3511047363281, "margin_dpo/margin_std": 680.1424560546875, "step": 646 }, { "KL/chosen_KL_mean": -680.39453125, "KL/mean": -889.831787109375, "KL/rejected_KL_mean": -1099.26904296875, "KL/std": 623.091796875, "epoch": 0.9500734214390602, "fcm_dpo/beta": 0.0007311520748771727, "fcm_dpo/delta": 0.04510268568992615, "fcm_dpo/margin": 418.8743896484375, "fcm_dpo/q_t": 0.4280046820640564, "grad_norm": 27.29217529296875, "learning_rate": 4.024152566816791e-09, "logits/chosen": -0.9219083786010742, "logits/rejected": -0.95496666431427, "logps/chosen": -735.2469482421875, "logps/ref_chosen": -54.8524169921875, "logps/ref_rejected": -93.5194091796875, "logps/rejected": -1192.788330078125, "loss": 1.1548, "margin_dpo/margin_mean": 418.8743896484375, "margin_dpo/margin_std": 625.203857421875, "step": 647 }, { "KL/chosen_KL_mean": -736.6182861328125, "KL/mean": -1046.520263671875, "KL/rejected_KL_mean": -1356.42236328125, "KL/std": 752.701904296875, "epoch": 0.9515418502202643, "fcm_dpo/beta": 0.0007215134100988507, "fcm_dpo/delta": -0.10175478458404541, "fcm_dpo/margin": 619.80419921875, "fcm_dpo/q_t": 0.4019627571105957, "grad_norm": 25.925161361694336, "learning_rate": 3.798061746947995e-09, "logits/chosen": -1.08555269241333, "logits/rejected": -1.1510529518127441, "logps/chosen": -790.7897338867188, "logps/ref_chosen": -54.17146682739258, "logps/ref_rejected": -98.7127914428711, "logps/rejected": -1455.13525390625, "loss": 1.0815, "margin_dpo/margin_mean": 619.80419921875, "margin_dpo/margin_std": 915.1447143554688, "step": 648 }, { "KL/chosen_KL_mean": -784.3709106445312, "KL/mean": -931.5806884765625, "KL/rejected_KL_mean": -1078.79052734375, "KL/std": 599.5159301757812, "epoch": 0.9530102790014684, "fcm_dpo/beta": 0.0007213630015030503, "fcm_dpo/delta": 0.04922113195061684, "fcm_dpo/margin": 294.419677734375, "fcm_dpo/q_t": 0.4535311460494995, "grad_norm": 27.51808738708496, "learning_rate": 3.5784585771215235e-09, "logits/chosen": -1.1117441654205322, "logits/rejected": -1.1074135303497314, "logps/chosen": -846.8512573242188, "logps/ref_chosen": -62.480350494384766, "logps/ref_rejected": -80.07717895507812, "logps/rejected": -1158.86767578125, "loss": 1.2587, "margin_dpo/margin_mean": 294.4196472167969, "margin_dpo/margin_std": 746.974853515625, "step": 649 }, { "KL/chosen_KL_mean": -838.1929931640625, "KL/mean": -1064.818359375, "KL/rejected_KL_mean": -1291.44384765625, "KL/std": 749.1234130859375, "epoch": 0.9544787077826725, "fcm_dpo/beta": 0.0007260285783559084, "fcm_dpo/delta": 0.021617352962493896, "fcm_dpo/margin": 453.2508544921875, "fcm_dpo/q_t": 0.4258856177330017, "grad_norm": 36.681617736816406, "learning_rate": 3.3653488440851253e-09, "logits/chosen": -1.0278213024139404, "logits/rejected": -1.050445795059204, "logps/chosen": -894.2857666015625, "logps/ref_chosen": -56.09281921386719, "logps/ref_rejected": -98.26483917236328, "logps/rejected": -1389.708740234375, "loss": 1.175, "margin_dpo/margin_mean": 453.2508544921875, "margin_dpo/margin_std": 861.7562255859375, "step": 650 }, { "KL/chosen_KL_mean": -549.6793212890625, "KL/mean": -854.8587036132812, "KL/rejected_KL_mean": -1160.0380859375, "KL/std": 682.4044799804688, "epoch": 0.9559471365638766, "fcm_dpo/beta": 0.0007174712372943759, "fcm_dpo/delta": -0.09204652160406113, "fcm_dpo/margin": 610.3588256835938, "fcm_dpo/q_t": 0.39922603964805603, "grad_norm": 39.13344955444336, "learning_rate": 3.158738163478475e-09, "logits/chosen": -1.0335227251052856, "logits/rejected": -1.0952156782150269, "logps/chosen": -593.104736328125, "logps/ref_chosen": -43.42544937133789, "logps/ref_rejected": -99.95791625976562, "logps/rejected": -1259.99609375, "loss": 1.055, "margin_dpo/margin_mean": 610.3587646484375, "margin_dpo/margin_std": 714.5980224609375, "step": 651 }, { "KL/chosen_KL_mean": -708.1835327148438, "KL/mean": -933.1541137695312, "KL/rejected_KL_mean": -1158.1246337890625, "KL/std": 705.0360107421875, "epoch": 0.9574155653450808, "fcm_dpo/beta": 0.0007186533184722066, "fcm_dpo/delta": 0.0274675115942955, "fcm_dpo/margin": 449.94110107421875, "fcm_dpo/q_t": 0.42639029026031494, "grad_norm": 38.71635818481445, "learning_rate": 2.9586319796851555e-09, "logits/chosen": -1.0511879920959473, "logits/rejected": -1.0809803009033203, "logps/chosen": -770.7603759765625, "logps/ref_chosen": -62.57680892944336, "logps/ref_rejected": -111.76779174804688, "logps/rejected": -1269.8924560546875, "loss": 1.1698, "margin_dpo/margin_mean": 449.941162109375, "margin_dpo/margin_std": 828.396240234375, "step": 652 }, { "KL/chosen_KL_mean": -871.048583984375, "KL/mean": -1087.985107421875, "KL/rejected_KL_mean": -1304.921630859375, "KL/std": 736.3079833984375, "epoch": 0.9588839941262849, "fcm_dpo/beta": 0.0007236155215650797, "fcm_dpo/delta": 0.03713195025920868, "fcm_dpo/margin": 433.8730773925781, "fcm_dpo/q_t": 0.4294109344482422, "grad_norm": 38.670223236083984, "learning_rate": 2.7650355656892166e-09, "logits/chosen": -1.1091747283935547, "logits/rejected": -1.138980746269226, "logps/chosen": -932.16162109375, "logps/ref_chosen": -61.11295700073242, "logps/ref_rejected": -103.24960327148438, "logps/rejected": -1408.17138671875, "loss": 1.1733, "margin_dpo/margin_mean": 433.8730773925781, "margin_dpo/margin_std": 794.478271484375, "step": 653 }, { "KL/chosen_KL_mean": -779.94384765625, "KL/mean": -982.3485107421875, "KL/rejected_KL_mean": -1184.753173828125, "KL/std": 596.6171875, "epoch": 0.960352422907489, "fcm_dpo/beta": 0.0007300974684767425, "fcm_dpo/delta": 0.05611284077167511, "fcm_dpo/margin": 404.8092041015625, "fcm_dpo/q_t": 0.4334927797317505, "grad_norm": 37.371498107910156, "learning_rate": 2.577954022936174e-09, "logits/chosen": -1.0734694004058838, "logits/rejected": -1.092237949371338, "logps/chosen": -841.6719970703125, "logps/ref_chosen": -61.7281379699707, "logps/ref_rejected": -98.7738037109375, "logps/rejected": -1283.5269775390625, "loss": 1.1781, "margin_dpo/margin_mean": 404.8092346191406, "margin_dpo/margin_std": 726.0567626953125, "step": 654 }, { "KL/chosen_KL_mean": -724.4075317382812, "KL/mean": -935.876220703125, "KL/rejected_KL_mean": -1147.344970703125, "KL/std": 602.5899658203125, "epoch": 0.9618208516886931, "fcm_dpo/beta": 0.0007367508951574564, "fcm_dpo/delta": 0.03962551802396774, "fcm_dpo/margin": 422.9373779296875, "fcm_dpo/q_t": 0.4289171099662781, "grad_norm": 30.014909744262695, "learning_rate": 2.397392281198729e-09, "logits/chosen": -1.055437684059143, "logits/rejected": -1.1015443801879883, "logps/chosen": -773.984375, "logps/ref_chosen": -49.576812744140625, "logps/ref_rejected": -98.29183197021484, "logps/rejected": -1245.63671875, "loss": 1.166, "margin_dpo/margin_mean": 422.9373779296875, "margin_dpo/margin_std": 725.2291259765625, "step": 655 }, { "KL/chosen_KL_mean": -779.2088623046875, "KL/mean": -1150.2603759765625, "KL/rejected_KL_mean": -1521.31201171875, "KL/std": 814.0103759765625, "epoch": 0.9632892804698973, "fcm_dpo/beta": 0.0007195958169177175, "fcm_dpo/delta": -0.1939731240272522, "fcm_dpo/margin": 742.10302734375, "fcm_dpo/q_t": 0.3803486227989197, "grad_norm": 68.15776824951172, "learning_rate": 2.223355098446622e-09, "logits/chosen": -0.9589822888374329, "logits/rejected": -1.0362895727157593, "logps/chosen": -831.75830078125, "logps/ref_chosen": -52.54943084716797, "logps/ref_rejected": -113.67464447021484, "logps/rejected": -1634.986572265625, "loss": 1.0025, "margin_dpo/margin_mean": 742.10302734375, "margin_dpo/margin_std": 839.2947998046875, "step": 656 }, { "KL/chosen_KL_mean": -735.045654296875, "KL/mean": -1036.506591796875, "KL/rejected_KL_mean": -1337.9677734375, "KL/std": 803.0293579101562, "epoch": 0.9647577092511013, "fcm_dpo/beta": 0.00070092489477247, "fcm_dpo/delta": -0.07597053050994873, "fcm_dpo/margin": 602.922119140625, "fcm_dpo/q_t": 0.4036826491355896, "grad_norm": 41.26325225830078, "learning_rate": 2.055847060721566e-09, "logits/chosen": -1.0972576141357422, "logits/rejected": -1.155198335647583, "logps/chosen": -781.7461547851562, "logps/ref_chosen": -46.700538635253906, "logps/ref_rejected": -97.91487121582031, "logps/rejected": -1435.882568359375, "loss": 1.0879, "margin_dpo/margin_mean": 602.922119140625, "margin_dpo/margin_std": 850.9063720703125, "step": 657 }, { "KL/chosen_KL_mean": -768.9866943359375, "KL/mean": -999.5380249023438, "KL/rejected_KL_mean": -1230.08935546875, "KL/std": 592.4909057617188, "epoch": 0.9662261380323054, "fcm_dpo/beta": 0.0006991151021793485, "fcm_dpo/delta": 0.02827742137014866, "fcm_dpo/margin": 461.10272216796875, "fcm_dpo/q_t": 0.42443007230758667, "grad_norm": 35.22886276245117, "learning_rate": 1.8948725820160662e-09, "logits/chosen": -1.0550436973571777, "logits/rejected": -1.0875539779663086, "logps/chosen": -829.9449462890625, "logps/ref_chosen": -60.95820999145508, "logps/ref_rejected": -95.93949127197266, "logps/rejected": -1326.02880859375, "loss": 1.1446, "margin_dpo/margin_mean": 461.10272216796875, "margin_dpo/margin_std": 672.0286865234375, "step": 658 }, { "KL/chosen_KL_mean": -714.39892578125, "KL/mean": -941.6336059570312, "KL/rejected_KL_mean": -1168.868408203125, "KL/std": 622.14306640625, "epoch": 0.9676945668135095, "fcm_dpo/beta": 0.0007049725973047316, "fcm_dpo/delta": 0.0305656585842371, "fcm_dpo/margin": 454.4694519042969, "fcm_dpo/q_t": 0.4260675311088562, "grad_norm": 35.09855270385742, "learning_rate": 1.7404359041573723e-09, "logits/chosen": -0.985196053981781, "logits/rejected": -0.9553881883621216, "logps/chosen": -791.1419067382812, "logps/ref_chosen": -76.74298095703125, "logps/ref_rejected": -87.4709701538086, "logps/rejected": -1256.33935546875, "loss": 1.1443, "margin_dpo/margin_mean": 454.469482421875, "margin_dpo/margin_std": 673.2275390625, "step": 659 }, { "KL/chosen_KL_mean": -742.4683837890625, "KL/mean": -1035.2835693359375, "KL/rejected_KL_mean": -1328.0987548828125, "KL/std": 711.6279296875, "epoch": 0.9691629955947136, "fcm_dpo/beta": 0.0007018651231192052, "fcm_dpo/delta": -0.0636008009314537, "fcm_dpo/margin": 585.63037109375, "fcm_dpo/q_t": 0.4052135944366455, "grad_norm": 43.428226470947266, "learning_rate": 1.592541096695571e-09, "logits/chosen": -1.0797405242919922, "logits/rejected": -1.1055940389633179, "logps/chosen": -801.5162963867188, "logps/ref_chosen": -59.04788589477539, "logps/ref_rejected": -75.96005249023438, "logps/rejected": -1404.058837890625, "loss": 1.0847, "margin_dpo/margin_mean": 585.63037109375, "margin_dpo/margin_std": 770.469970703125, "step": 660 }, { "KL/chosen_KL_mean": -663.244140625, "KL/mean": -936.0105590820312, "KL/rejected_KL_mean": -1208.77685546875, "KL/std": 813.6663818359375, "epoch": 0.9706314243759178, "fcm_dpo/beta": 0.0006977942539379001, "fcm_dpo/delta": -0.03203122317790985, "fcm_dpo/margin": 545.5327758789062, "fcm_dpo/q_t": 0.4161611795425415, "grad_norm": 36.02547836303711, "learning_rate": 1.4511920567963908e-09, "logits/chosen": -1.0755589008331299, "logits/rejected": -1.100312352180481, "logps/chosen": -713.9181518554688, "logps/ref_chosen": -50.673973083496094, "logps/ref_rejected": -86.00569152832031, "logps/rejected": -1294.7825927734375, "loss": 1.112, "margin_dpo/margin_mean": 545.5327758789062, "margin_dpo/margin_std": 835.273681640625, "step": 661 }, { "KL/chosen_KL_mean": -777.9444580078125, "KL/mean": -982.4157104492188, "KL/rejected_KL_mean": -1186.886962890625, "KL/std": 655.0043334960938, "epoch": 0.9720998531571219, "fcm_dpo/beta": 0.0007006356026977301, "fcm_dpo/delta": 0.06537885963916779, "fcm_dpo/margin": 408.9425048828125, "fcm_dpo/q_t": 0.4345260560512543, "grad_norm": 26.648513793945312, "learning_rate": 1.3163925091384532e-09, "logits/chosen": -1.0024073123931885, "logits/rejected": -1.0070304870605469, "logps/chosen": -847.20556640625, "logps/ref_chosen": -69.26106262207031, "logps/ref_rejected": -89.05593872070312, "logps/rejected": -1275.94287109375, "loss": 1.1959, "margin_dpo/margin_mean": 408.9425048828125, "margin_dpo/margin_std": 810.3031005859375, "step": 662 }, { "KL/chosen_KL_mean": -734.29833984375, "KL/mean": -990.4869995117188, "KL/rejected_KL_mean": -1246.6756591796875, "KL/std": 746.890625, "epoch": 0.973568281938326, "fcm_dpo/beta": 0.000701136770658195, "fcm_dpo/delta": -0.009660286828875542, "fcm_dpo/margin": 512.3773193359375, "fcm_dpo/q_t": 0.42002010345458984, "grad_norm": 25.236740112304688, "learning_rate": 1.1881460058152382e-09, "logits/chosen": -1.055614709854126, "logits/rejected": -1.0850775241851807, "logps/chosen": -799.17724609375, "logps/ref_chosen": -64.87890625, "logps/ref_rejected": -113.92536926269531, "logps/rejected": -1360.60107421875, "loss": 1.1439, "margin_dpo/margin_mean": 512.3772583007812, "margin_dpo/margin_std": 886.00927734375, "step": 663 }, { "KL/chosen_KL_mean": -807.125732421875, "KL/mean": -1088.34619140625, "KL/rejected_KL_mean": -1369.56689453125, "KL/std": 771.4961547851562, "epoch": 0.9750367107195301, "fcm_dpo/beta": 0.0006953716510906816, "fcm_dpo/delta": -0.04308585077524185, "fcm_dpo/margin": 562.4410400390625, "fcm_dpo/q_t": 0.41181105375289917, "grad_norm": 27.388290405273438, "learning_rate": 1.066455926241383e-09, "logits/chosen": -1.0791889429092407, "logits/rejected": -1.1189603805541992, "logps/chosen": -868.0142211914062, "logps/ref_chosen": -60.88847351074219, "logps/ref_rejected": -105.521728515625, "logps/rejected": -1475.088623046875, "loss": 1.1093, "margin_dpo/margin_mean": 562.4410400390625, "margin_dpo/margin_std": 830.4803466796875, "step": 664 }, { "KL/chosen_KL_mean": -685.2548828125, "KL/mean": -923.9151000976562, "KL/rejected_KL_mean": -1162.5753173828125, "KL/std": 609.5552368164062, "epoch": 0.9765051395007343, "fcm_dpo/beta": 0.0006976306904107332, "fcm_dpo/delta": 0.017594607546925545, "fcm_dpo/margin": 477.3204345703125, "fcm_dpo/q_t": 0.42177683115005493, "grad_norm": 50.124088287353516, "learning_rate": 9.513254770636137e-10, "logits/chosen": -1.1326605081558228, "logits/rejected": -1.160711646080017, "logps/chosen": -745.8190307617188, "logps/ref_chosen": -60.56413269042969, "logps/ref_rejected": -84.80882263183594, "logps/rejected": -1247.384033203125, "loss": 1.1244, "margin_dpo/margin_mean": 477.3204345703125, "margin_dpo/margin_std": 607.7765502929688, "step": 665 }, { "KL/chosen_KL_mean": -722.7728271484375, "KL/mean": -959.1403198242188, "KL/rejected_KL_mean": -1195.5078125, "KL/std": 615.6036376953125, "epoch": 0.9779735682819384, "fcm_dpo/beta": 0.0006999197648838162, "fcm_dpo/delta": 0.019781440496444702, "fcm_dpo/margin": 472.7351379394531, "fcm_dpo/q_t": 0.4240819215774536, "grad_norm": 30.830541610717773, "learning_rate": 8.427576920763956e-10, "logits/chosen": -0.9478716254234314, "logits/rejected": -0.9615623950958252, "logps/chosen": -787.1927490234375, "logps/ref_chosen": -64.41996002197266, "logps/ref_rejected": -95.8916244506836, "logps/rejected": -1291.3995361328125, "loss": 1.1399, "margin_dpo/margin_mean": 472.73516845703125, "margin_dpo/margin_std": 689.4872436523438, "step": 666 }, { "KL/chosen_KL_mean": -859.1533203125, "KL/mean": -1136.722900390625, "KL/rejected_KL_mean": -1414.29248046875, "KL/std": 699.344482421875, "epoch": 0.9794419970631424, "fcm_dpo/beta": 0.0006951624527573586, "fcm_dpo/delta": -0.037639595568180084, "fcm_dpo/margin": 555.1390380859375, "fcm_dpo/q_t": 0.4112476706504822, "grad_norm": 36.62522506713867, "learning_rate": 7.407554321417764e-10, "logits/chosen": -0.9919800758361816, "logits/rejected": -0.995948076248169, "logps/chosen": -928.4303588867188, "logps/ref_chosen": -69.27702331542969, "logps/ref_rejected": -87.83549499511719, "logps/rejected": -1502.1279296875, "loss": 1.1057, "margin_dpo/margin_mean": 555.1390380859375, "margin_dpo/margin_std": 778.8447875976562, "step": 667 }, { "KL/chosen_KL_mean": -890.34130859375, "KL/mean": -1082.375732421875, "KL/rejected_KL_mean": -1274.4102783203125, "KL/std": 727.134765625, "epoch": 0.9809104258443465, "fcm_dpo/beta": 0.0006961192702874541, "fcm_dpo/delta": 0.003592910710722208, "fcm_dpo/margin": 384.0690002441406, "fcm_dpo/q_t": 0.4404904246330261, "grad_norm": 48.40626525878906, "learning_rate": 6.453213851142225e-10, "logits/chosen": -1.0494754314422607, "logits/rejected": -1.0595271587371826, "logps/chosen": -962.9453125, "logps/ref_chosen": -72.60400390625, "logps/ref_rejected": -103.73905944824219, "logps/rejected": -1378.1494140625, "loss": 1.2243, "margin_dpo/margin_mean": 384.0690002441406, "margin_dpo/margin_std": 874.9485473632812, "step": 668 }, { "KL/chosen_KL_mean": -679.6414184570312, "KL/mean": -960.998291015625, "KL/rejected_KL_mean": -1242.3551025390625, "KL/std": 686.39013671875, "epoch": 0.9823788546255506, "fcm_dpo/beta": 0.000693183159455657, "fcm_dpo/delta": -0.04167580232024193, "fcm_dpo/margin": 562.7136840820312, "fcm_dpo/q_t": 0.4098045825958252, "grad_norm": 26.838964462280273, "learning_rate": 5.564580657695939e-10, "logits/chosen": -1.047389268875122, "logits/rejected": -1.0622926950454712, "logps/chosen": -725.7578125, "logps/ref_chosen": -46.116416931152344, "logps/ref_rejected": -77.92434692382812, "logps/rejected": -1320.279541015625, "loss": 1.101, "margin_dpo/margin_mean": 562.7137451171875, "margin_dpo/margin_std": 774.2056884765625, "step": 669 }, { "KL/chosen_KL_mean": -649.9046630859375, "KL/mean": -927.82958984375, "KL/rejected_KL_mean": -1205.754638671875, "KL/std": 644.1104736328125, "epoch": 0.9838472834067548, "fcm_dpo/beta": 0.0006895088008604944, "fcm_dpo/delta": -0.03472786396741867, "fcm_dpo/margin": 555.849853515625, "fcm_dpo/q_t": 0.41115692257881165, "grad_norm": 25.038528442382812, "learning_rate": 4.741678157389739e-10, "logits/chosen": -0.946910560131073, "logits/rejected": -0.9696442484855652, "logps/chosen": -712.2504272460938, "logps/ref_chosen": -62.34575271606445, "logps/ref_rejected": -96.9405517578125, "logps/rejected": -1302.695068359375, "loss": 1.1056, "margin_dpo/margin_mean": 555.849853515625, "margin_dpo/margin_std": 760.0970458984375, "step": 670 }, { "KL/chosen_KL_mean": -844.815673828125, "KL/mean": -1066.053955078125, "KL/rejected_KL_mean": -1287.292236328125, "KL/std": 671.5584716796875, "epoch": 0.9853157121879589, "fcm_dpo/beta": 0.0006929665105417371, "fcm_dpo/delta": 0.044294700026512146, "fcm_dpo/margin": 442.4765625, "fcm_dpo/q_t": 0.4280014634132385, "grad_norm": 30.275554656982422, "learning_rate": 3.9845280344705245e-10, "logits/chosen": -1.0639939308166504, "logits/rejected": -1.0980405807495117, "logps/chosen": -892.8157958984375, "logps/ref_chosen": -48.00010681152344, "logps/ref_rejected": -83.81932067871094, "logps/rejected": -1371.111572265625, "loss": 1.1821, "margin_dpo/margin_mean": 442.4765625, "margin_dpo/margin_std": 831.6336669921875, "step": 671 }, { "KL/chosen_KL_mean": -921.38525390625, "KL/mean": -1149.651123046875, "KL/rejected_KL_mean": -1377.9171142578125, "KL/std": 785.9630737304688, "epoch": 0.986784140969163, "fcm_dpo/beta": 0.0006937608122825623, "fcm_dpo/delta": 0.0343257300555706, "fcm_dpo/margin": 456.5320129394531, "fcm_dpo/q_t": 0.4287392497062683, "grad_norm": 48.789405822753906, "learning_rate": 3.293150240547549e-10, "logits/chosen": -1.1506716012954712, "logits/rejected": -1.1607093811035156, "logps/chosen": -979.968505859375, "logps/ref_chosen": -58.58328628540039, "logps/ref_rejected": -93.14015197753906, "logps/rejected": -1471.057373046875, "loss": 1.1798, "margin_dpo/margin_mean": 456.5320129394531, "margin_dpo/margin_std": 864.0927124023438, "step": 672 }, { "KL/chosen_KL_mean": -810.012939453125, "KL/mean": -1019.599365234375, "KL/rejected_KL_mean": -1229.185791015625, "KL/std": 655.4266357421875, "epoch": 0.9882525697503671, "fcm_dpo/beta": 0.0007023122161626816, "fcm_dpo/delta": 0.05732039362192154, "fcm_dpo/margin": 419.1728210449219, "fcm_dpo/q_t": 0.4328360855579376, "grad_norm": 23.517154693603516, "learning_rate": 2.6675629940689504e-10, "logits/chosen": -1.093796968460083, "logits/rejected": -1.101298451423645, "logps/chosen": -856.7362060546875, "logps/ref_chosen": -46.72320556640625, "logps/ref_rejected": -85.29623413085938, "logps/rejected": -1314.4820556640625, "loss": 1.1744, "margin_dpo/margin_mean": 419.1728515625, "margin_dpo/margin_std": 737.906982421875, "step": 673 }, { "KL/chosen_KL_mean": -668.804443359375, "KL/mean": -940.5889892578125, "KL/rejected_KL_mean": -1212.37353515625, "KL/std": 636.3111572265625, "epoch": 0.9897209985315712, "fcm_dpo/beta": 0.0007032062858343124, "fcm_dpo/delta": -0.033593274652957916, "fcm_dpo/margin": 543.5691528320312, "fcm_dpo/q_t": 0.4136511981487274, "grad_norm": 31.608436584472656, "learning_rate": 2.1077827798404725e-10, "logits/chosen": -0.9837385416030884, "logits/rejected": -1.0068650245666504, "logps/chosen": -714.25, "logps/ref_chosen": -45.445526123046875, "logps/ref_rejected": -70.04593658447266, "logps/rejected": -1282.41943359375, "loss": 1.1079, "margin_dpo/margin_mean": 543.5691528320312, "margin_dpo/margin_std": 769.9119873046875, "step": 674 }, { "KL/chosen_KL_mean": -757.3980712890625, "KL/mean": -1031.643798828125, "KL/rejected_KL_mean": -1305.8896484375, "KL/std": 698.0523681640625, "epoch": 0.9911894273127754, "fcm_dpo/beta": 0.0006924683111719787, "fcm_dpo/delta": -0.0319318063557148, "fcm_dpo/margin": 548.4915771484375, "fcm_dpo/q_t": 0.4143555462360382, "grad_norm": 27.629240036010742, "learning_rate": 1.6138243485910863e-10, "logits/chosen": -1.0433604717254639, "logits/rejected": -1.0594494342803955, "logps/chosen": -801.5743408203125, "logps/ref_chosen": -44.17628479003906, "logps/ref_rejected": -74.09197998046875, "logps/rejected": -1379.981689453125, "loss": 1.1111, "margin_dpo/margin_mean": 548.4916381835938, "margin_dpo/margin_std": 764.034912109375, "step": 675 }, { "KL/chosen_KL_mean": -832.3480834960938, "KL/mean": -1099.799560546875, "KL/rejected_KL_mean": -1367.2510986328125, "KL/std": 692.36865234375, "epoch": 0.9926578560939795, "fcm_dpo/beta": 0.0006945295608602464, "fcm_dpo/delta": -0.02236580289900303, "fcm_dpo/margin": 534.9030151367188, "fcm_dpo/q_t": 0.4144228398799896, "grad_norm": 33.126529693603516, "learning_rate": 1.1857007165852472e-10, "logits/chosen": -1.004886269569397, "logits/rejected": -1.0242345333099365, "logps/chosen": -903.74658203125, "logps/ref_chosen": -71.39852905273438, "logps/ref_rejected": -88.3587646484375, "logps/rejected": -1455.60986328125, "loss": 1.109, "margin_dpo/margin_mean": 534.9030151367188, "margin_dpo/margin_std": 720.2886352539062, "step": 676 }, { "KL/chosen_KL_mean": -826.7559814453125, "KL/mean": -1088.8828125, "KL/rejected_KL_mean": -1351.010009765625, "KL/std": 718.5059204101562, "epoch": 0.9941262848751835, "fcm_dpo/beta": 0.0006902614841237664, "fcm_dpo/delta": -0.012368248775601387, "fcm_dpo/margin": 524.25390625, "fcm_dpo/q_t": 0.4213380217552185, "grad_norm": 35.2880859375, "learning_rate": 8.23423165278725e-11, "logits/chosen": -1.128098487854004, "logits/rejected": -1.131150722503662, "logps/chosen": -883.2833862304688, "logps/ref_chosen": -56.527435302734375, "logps/ref_rejected": -78.22654724121094, "logps/rejected": -1429.2364501953125, "loss": 1.1378, "margin_dpo/margin_mean": 524.25390625, "margin_dpo/margin_std": 882.9517822265625, "step": 677 }, { "KL/chosen_KL_mean": -681.46044921875, "KL/mean": -999.7894287109375, "KL/rejected_KL_mean": -1318.118408203125, "KL/std": 790.3182983398438, "epoch": 0.9955947136563876, "fcm_dpo/beta": 0.0006847955519333482, "fcm_dpo/delta": -0.08993227779865265, "fcm_dpo/margin": 636.6580810546875, "fcm_dpo/q_t": 0.40213915705680847, "grad_norm": 32.831016540527344, "learning_rate": 5.270012410216185e-11, "logits/chosen": -1.0311408042907715, "logits/rejected": -1.0793263912200928, "logps/chosen": -727.5949096679688, "logps/ref_chosen": -46.13447570800781, "logps/ref_rejected": -80.60462951660156, "logps/rejected": -1398.72314453125, "loss": 1.0853, "margin_dpo/margin_mean": 636.6580200195312, "margin_dpo/margin_std": 912.1826171875, "step": 678 }, { "KL/chosen_KL_mean": -798.2120361328125, "KL/mean": -1003.3867797851562, "KL/rejected_KL_mean": -1208.5615234375, "KL/std": 617.0496826171875, "epoch": 0.9970631424375918, "fcm_dpo/beta": 0.000684911385178566, "fcm_dpo/delta": 0.07099346816539764, "fcm_dpo/margin": 410.34954833984375, "fcm_dpo/q_t": 0.43571871519088745, "grad_norm": 32.56108093261719, "learning_rate": 2.9644275480772416e-11, "logits/chosen": -1.0531508922576904, "logits/rejected": -1.048056960105896, "logps/chosen": -848.5069580078125, "logps/ref_chosen": -50.294921875, "logps/ref_rejected": -76.59813690185547, "logps/rejected": -1285.15966796875, "loss": 1.1832, "margin_dpo/margin_mean": 410.3495178222656, "margin_dpo/margin_std": 727.0447387695312, "step": 679 }, { "KL/chosen_KL_mean": -812.5574951171875, "KL/mean": -1099.232177734375, "KL/rejected_KL_mean": -1385.906982421875, "KL/std": 805.0675659179688, "epoch": 0.9985315712187959, "fcm_dpo/beta": 0.0006806158926337957, "fcm_dpo/delta": -0.04249640554189682, "fcm_dpo/margin": 573.349365234375, "fcm_dpo/q_t": 0.4099177122116089, "grad_norm": 40.93464660644531, "learning_rate": 1.31753782067201e-11, "logits/chosen": -1.0480668544769287, "logits/rejected": -1.0822138786315918, "logps/chosen": -889.4732055664062, "logps/ref_chosen": -76.91569519042969, "logps/ref_rejected": -112.384765625, "logps/rejected": -1498.291748046875, "loss": 1.1149, "margin_dpo/margin_mean": 573.349365234375, "margin_dpo/margin_std": 864.7239990234375, "step": 680 }, { "KL/chosen_KL_mean": -819.076416015625, "KL/mean": -1035.499755859375, "KL/rejected_KL_mean": -1251.92333984375, "KL/std": 668.314697265625, "epoch": 1.0, "fcm_dpo/beta": 0.0006794101791456342, "fcm_dpo/delta": -0.042327266186475754, "fcm_dpo/margin": 432.8468933105469, "fcm_dpo/q_t": 0.4326745867729187, "grad_norm": 40.8109130859375, "learning_rate": 3.2938662507808745e-12, "logits/chosen": -1.1136659383773804, "logits/rejected": -1.1350388526916504, "logps/chosen": -880.03369140625, "logps/ref_chosen": -60.957279205322266, "logps/ref_rejected": -88.55797576904297, "logps/rejected": -1340.481201171875, "loss": 1.1847, "margin_dpo/margin_mean": 432.84686279296875, "margin_dpo/margin_std": 766.4469604492188, "step": 681 }, { "epoch": 1.0, "step": 681, "total_flos": 0.0, "train_loss": 1.1324475014612363, "train_runtime": 1725.4575, "train_samples_per_second": 25.268, "train_steps_per_second": 0.395 } ], "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 }