{ "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": 83.5241928100586, "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": 72.2212142944336, "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": -0.001789093017578125, "KL/mean": 0.018673986196517944, "KL/rejected_KL_mean": 0.03913688659667969, "KL/std": 0.2510424256324768, "epoch": 0.004405286343612335, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.04093104600906372, "fcm_dpo/q_t": 0.5010229349136353, "grad_norm": 71.08148193359375, "learning_rate": 1.4492753623188406e-08, "logits/chosen": -0.48160529136657715, "logits/rejected": -0.4421182870864868, "logps/chosen": -60.98338317871094, "logps/ref_chosen": -60.981597900390625, "logps/ref_rejected": -68.67259216308594, "logps/rejected": -68.63346099853516, "loss": 1.3908, "margin_dpo/margin_mean": -0.04093062877655029, "margin_dpo/margin_std": 0.3883278965950012, "step": 3 }, { "KL/chosen_KL_mean": -0.035938262939453125, "KL/mean": -0.004994198679924011, "KL/rejected_KL_mean": 0.02594757080078125, "KL/std": 0.25746771693229675, "epoch": 0.005873715124816446, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.06188991665840149, "fcm_dpo/q_t": 0.5015468001365662, "grad_norm": 72.06177520751953, "learning_rate": 2.1739130434782606e-08, "logits/chosen": -0.4687862992286682, "logits/rejected": -0.4411006569862366, "logps/chosen": -56.80364990234375, "logps/ref_chosen": -56.7677116394043, "logps/ref_rejected": -86.64710998535156, "logps/rejected": -86.62116241455078, "loss": 1.3928, "margin_dpo/margin_mean": -0.06189015507698059, "margin_dpo/margin_std": 0.36542826890945435, "step": 4 }, { "KL/chosen_KL_mean": 0.041553497314453125, "KL/mean": 0.02368500828742981, "KL/rejected_KL_mean": 0.00582122802734375, "KL/std": 0.25065213441848755, "epoch": 0.007342143906020558, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.03573501110076904, "fcm_dpo/q_t": 0.4991065561771393, "grad_norm": 89.9738998413086, "learning_rate": 2.898550724637681e-08, "logits/chosen": -0.514451265335083, "logits/rejected": -0.4707820415496826, "logps/chosen": -53.81782150268555, "logps/ref_chosen": -53.859375, "logps/ref_rejected": -84.14918518066406, "logps/rejected": -84.14336395263672, "loss": 1.383, "margin_dpo/margin_mean": 0.035734742879867554, "margin_dpo/margin_std": 0.3362279534339905, "step": 5 }, { "KL/chosen_KL_mean": -0.03502845764160156, "KL/mean": -0.045771896839141846, "KL/rejected_KL_mean": -0.0565185546875, "KL/std": 0.302729070186615, "epoch": 0.00881057268722467, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.021490395069122314, "fcm_dpo/q_t": 0.4994634687900543, "grad_norm": 91.83248138427734, "learning_rate": 3.6231884057971014e-08, "logits/chosen": -0.4969807267189026, "logits/rejected": -0.4538990259170532, "logps/chosen": -63.042510986328125, "logps/ref_chosen": -63.007484436035156, "logps/ref_rejected": -92.64534759521484, "logps/rejected": -92.70186614990234, "loss": 1.3846, "margin_dpo/margin_mean": 0.02148991823196411, "margin_dpo/margin_std": 0.44182121753692627, "step": 6 }, { "KL/chosen_KL_mean": 0.03751373291015625, "KL/mean": 0.04191398620605469, "KL/rejected_KL_mean": 0.04631805419921875, "KL/std": 0.2821422517299652, "epoch": 0.010279001468428781, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.008798539638519287, "fcm_dpo/q_t": 0.5002202987670898, "grad_norm": 82.35990142822266, "learning_rate": 4.347826086956521e-08, "logits/chosen": -0.5038993954658508, "logits/rejected": -0.47004765272140503, "logps/chosen": -57.7373046875, "logps/ref_chosen": -57.774818420410156, "logps/ref_rejected": -103.92059326171875, "logps/rejected": -103.87427520751953, "loss": 1.3876, "margin_dpo/margin_mean": -0.00879824161529541, "margin_dpo/margin_std": 0.41485100984573364, "step": 7 }, { "KL/chosen_KL_mean": 0.012628555297851562, "KL/mean": -0.0013420134782791138, "KL/rejected_KL_mean": -0.01531219482421875, "KL/std": 0.3049571216106415, "epoch": 0.011747430249632892, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.027936726808547974, "fcm_dpo/q_t": 0.49930089712142944, "grad_norm": 78.54287719726562, "learning_rate": 5.0724637681159424e-08, "logits/chosen": -0.49774545431137085, "logits/rejected": -0.4704884886741638, "logps/chosen": -58.703407287597656, "logps/ref_chosen": -58.716033935546875, "logps/ref_rejected": -79.3114242553711, "logps/rejected": -79.32673645019531, "loss": 1.3839, "margin_dpo/margin_mean": 0.027936339378356934, "margin_dpo/margin_std": 0.38197433948516846, "step": 8 }, { "KL/chosen_KL_mean": 0.00943756103515625, "KL/mean": 0.014243200421333313, "KL/rejected_KL_mean": 0.019046783447265625, "KL/std": 0.28038617968559265, "epoch": 0.013215859030837005, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.009610921144485474, "fcm_dpo/q_t": 0.5002400875091553, "grad_norm": 85.46974182128906, "learning_rate": 5.797101449275362e-08, "logits/chosen": -0.5015987753868103, "logits/rejected": -0.4552796483039856, "logps/chosen": -69.85740661621094, "logps/ref_chosen": -69.8668441772461, "logps/ref_rejected": -99.6026611328125, "logps/rejected": -99.5836181640625, "loss": 1.3876, "margin_dpo/margin_mean": -0.009611576795578003, "margin_dpo/margin_std": 0.3819504380226135, "step": 9 }, { "KL/chosen_KL_mean": 0.03025054931640625, "KL/mean": 0.027018234133720398, "KL/rejected_KL_mean": 0.0237884521484375, "KL/std": 0.2487342655658722, "epoch": 0.014684287812041116, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.006460636854171753, "fcm_dpo/q_t": 0.4998389482498169, "grad_norm": 70.36607360839844, "learning_rate": 6.521739130434782e-08, "logits/chosen": -0.4691097140312195, "logits/rejected": -0.4247715175151825, "logps/chosen": -48.32743453979492, "logps/ref_chosen": -48.35768508911133, "logps/ref_rejected": -80.37206268310547, "logps/rejected": -80.34827423095703, "loss": 1.386, "margin_dpo/margin_mean": 0.006460994482040405, "margin_dpo/margin_std": 0.37741619348526, "step": 10 }, { "KL/chosen_KL_mean": -0.013975143432617188, "KL/mean": -0.009909212589263916, "KL/rejected_KL_mean": -0.005840301513671875, "KL/std": 0.25705647468566895, "epoch": 0.016152716593245228, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.008137047290802002, "fcm_dpo/q_t": 0.5002034902572632, "grad_norm": 68.1004867553711, "learning_rate": 7.246376811594203e-08, "logits/chosen": -0.4795861840248108, "logits/rejected": -0.4555514454841614, "logps/chosen": -53.03083038330078, "logps/ref_chosen": -53.01685333251953, "logps/ref_rejected": -87.78038024902344, "logps/rejected": -87.78621673583984, "loss": 1.3875, "margin_dpo/margin_mean": -0.008136779069900513, "margin_dpo/margin_std": 0.36586588621139526, "step": 11 }, { "KL/chosen_KL_mean": -0.026020050048828125, "KL/mean": -0.01642027497291565, "KL/rejected_KL_mean": -0.00681304931640625, "KL/std": 0.29991090297698975, "epoch": 0.01762114537444934, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.019200026988983154, "fcm_dpo/q_t": 0.5004792213439941, "grad_norm": 90.451904296875, "learning_rate": 7.971014492753623e-08, "logits/chosen": -0.523429811000824, "logits/rejected": -0.4867948889732361, "logps/chosen": -61.831459045410156, "logps/ref_chosen": -61.80543518066406, "logps/ref_rejected": -104.8582763671875, "logps/rejected": -104.86508178710938, "loss": 1.3886, "margin_dpo/margin_mean": -0.019200772047042847, "margin_dpo/margin_std": 0.40876132249832153, "step": 12 }, { "KL/chosen_KL_mean": 0.00020599365234375, "KL/mean": 0.01627969741821289, "KL/rejected_KL_mean": 0.03235626220703125, "KL/std": 0.30039823055267334, "epoch": 0.01908957415565345, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.03214287757873535, "fcm_dpo/q_t": 0.5008050203323364, "grad_norm": 79.43994903564453, "learning_rate": 8.695652173913042e-08, "logits/chosen": -0.46810948848724365, "logits/rejected": -0.43907326459884644, "logps/chosen": -64.26014709472656, "logps/ref_chosen": -64.2603530883789, "logps/ref_rejected": -87.20307922363281, "logps/rejected": -87.17073059082031, "loss": 1.39, "margin_dpo/margin_mean": -0.03214350342750549, "margin_dpo/margin_std": 0.41335833072662354, "step": 13 }, { "KL/chosen_KL_mean": -0.05523681640625, "KL/mean": -0.02267158031463623, "KL/rejected_KL_mean": 0.0098876953125, "KL/std": 0.2773570418357849, "epoch": 0.020558002936857563, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.06512236595153809, "fcm_dpo/q_t": 0.5016275644302368, "grad_norm": 85.54232025146484, "learning_rate": 9.420289855072464e-08, "logits/chosen": -0.461495578289032, "logits/rejected": -0.4215831756591797, "logps/chosen": -58.16544723510742, "logps/ref_chosen": -58.11021041870117, "logps/ref_rejected": -104.04708099365234, "logps/rejected": -104.03719329833984, "loss": 1.3932, "margin_dpo/margin_mean": -0.06512203812599182, "margin_dpo/margin_std": 0.40178561210632324, "step": 14 }, { "KL/chosen_KL_mean": 0.008823394775390625, "KL/mean": -0.01242247223854065, "KL/rejected_KL_mean": -0.03366851806640625, "KL/std": 0.24548175930976868, "epoch": 0.022026431718061675, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.04249519109725952, "fcm_dpo/q_t": 0.4989377558231354, "grad_norm": 63.90875244140625, "learning_rate": 1.0144927536231885e-07, "logits/chosen": -0.4897874891757965, "logits/rejected": -0.47086426615715027, "logps/chosen": -56.95808410644531, "logps/ref_chosen": -56.96691131591797, "logps/ref_rejected": -80.80863952636719, "logps/rejected": -80.84231567382812, "loss": 1.3825, "margin_dpo/margin_mean": 0.04249492287635803, "margin_dpo/margin_std": 0.40503597259521484, "step": 15 }, { "KL/chosen_KL_mean": -0.04467201232910156, "KL/mean": -0.026810422539711, "KL/rejected_KL_mean": -0.008953094482421875, "KL/std": 0.2630922496318817, "epoch": 0.023494860499265784, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.0357193648815155, "fcm_dpo/q_t": 0.500891923904419, "grad_norm": 84.26193237304688, "learning_rate": 1.0869565217391303e-07, "logits/chosen": -0.5425084233283997, "logits/rejected": -0.5031050443649292, "logps/chosen": -61.78456497192383, "logps/ref_chosen": -61.739891052246094, "logps/ref_rejected": -84.36947631835938, "logps/rejected": -84.37842559814453, "loss": 1.3902, "margin_dpo/margin_mean": -0.03571975231170654, "margin_dpo/margin_std": 0.38462013006210327, "step": 16 }, { "KL/chosen_KL_mean": 0.0346221923828125, "KL/mean": -0.005598008632659912, "KL/rejected_KL_mean": -0.045818328857421875, "KL/std": 0.26897895336151123, "epoch": 0.024963289280469897, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.08044159412384033, "fcm_dpo/q_t": 0.4979892373085022, "grad_norm": 78.46760559082031, "learning_rate": 1.1594202898550725e-07, "logits/chosen": -0.5154045820236206, "logits/rejected": -0.480136513710022, "logps/chosen": -67.67571258544922, "logps/ref_chosen": -67.71033477783203, "logps/ref_rejected": -85.37865447998047, "logps/rejected": -85.42446899414062, "loss": 1.3786, "margin_dpo/margin_mean": 0.08044150471687317, "margin_dpo/margin_std": 0.3672877550125122, "step": 17 }, { "KL/chosen_KL_mean": -0.008848190307617188, "KL/mean": -0.022856131196022034, "KL/rejected_KL_mean": -0.036865234375, "KL/std": 0.2776832580566406, "epoch": 0.02643171806167401, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.028015047311782837, "fcm_dpo/q_t": 0.4992990791797638, "grad_norm": 82.00312042236328, "learning_rate": 1.2318840579710146e-07, "logits/chosen": -0.516010582447052, "logits/rejected": -0.4631628394126892, "logps/chosen": -47.74833679199219, "logps/ref_chosen": -47.7394905090332, "logps/ref_rejected": -75.4722900390625, "logps/rejected": -75.5091552734375, "loss": 1.3838, "margin_dpo/margin_mean": 0.028014779090881348, "margin_dpo/margin_std": 0.3619588017463684, "step": 18 }, { "KL/chosen_KL_mean": 0.031284332275390625, "KL/mean": -0.0030418336391448975, "KL/rejected_KL_mean": -0.03736114501953125, "KL/std": 0.23613137006759644, "epoch": 0.027900146842878122, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.06864070892333984, "fcm_dpo/q_t": 0.49828577041625977, "grad_norm": 73.23355865478516, "learning_rate": 1.3043478260869563e-07, "logits/chosen": -0.48611390590667725, "logits/rejected": -0.435970276594162, "logps/chosen": -70.174072265625, "logps/ref_chosen": -70.20536041259766, "logps/ref_rejected": -89.7575912475586, "logps/rejected": -89.79495239257812, "loss": 1.3798, "margin_dpo/margin_mean": 0.06864088773727417, "margin_dpo/margin_std": 0.35450613498687744, "step": 19 }, { "KL/chosen_KL_mean": -0.012430191040039062, "KL/mean": -0.04596352577209473, "KL/rejected_KL_mean": -0.07949447631835938, "KL/std": 0.2700064778327942, "epoch": 0.02936857562408223, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.06706851720809937, "fcm_dpo/q_t": 0.49832427501678467, "grad_norm": 73.91919708251953, "learning_rate": 1.3768115942028986e-07, "logits/chosen": -0.5568511486053467, "logits/rejected": -0.5009536743164062, "logps/chosen": -50.815673828125, "logps/ref_chosen": -50.80324172973633, "logps/ref_rejected": -78.82334899902344, "logps/rejected": -78.90284729003906, "loss": 1.3799, "margin_dpo/margin_mean": 0.06706875562667847, "margin_dpo/margin_std": 0.3464469313621521, "step": 20 }, { "KL/chosen_KL_mean": -0.012804031372070312, "KL/mean": -0.05751395970582962, "KL/rejected_KL_mean": -0.10222244262695312, "KL/std": 0.2957477867603302, "epoch": 0.030837004405286344, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.08941756188869476, "fcm_dpo/q_t": 0.49776533246040344, "grad_norm": 77.51765441894531, "learning_rate": 1.4492753623188405e-07, "logits/chosen": -0.5174500346183777, "logits/rejected": -0.4953506886959076, "logps/chosen": -50.07582092285156, "logps/ref_chosen": -50.063018798828125, "logps/ref_rejected": -77.86878967285156, "logps/rejected": -77.97101593017578, "loss": 1.3777, "margin_dpo/margin_mean": 0.08941812813282013, "margin_dpo/margin_std": 0.3819068670272827, "step": 21 }, { "KL/chosen_KL_mean": 0.02753448486328125, "KL/mean": -0.06719014048576355, "KL/rejected_KL_mean": -0.16191482543945312, "KL/std": 0.3151050806045532, "epoch": 0.032305433186490456, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.18945235013961792, "fcm_dpo/q_t": 0.49526724219322205, "grad_norm": 84.16546630859375, "learning_rate": 1.5217391304347825e-07, "logits/chosen": -0.4989904761314392, "logits/rejected": -0.45593979954719543, "logps/chosen": -59.03010177612305, "logps/ref_chosen": -59.05763626098633, "logps/ref_rejected": -97.50466918945312, "logps/rejected": -97.66658782958984, "loss": 1.3679, "margin_dpo/margin_mean": 0.1894513964653015, "margin_dpo/margin_std": 0.44450777769088745, "step": 22 }, { "KL/chosen_KL_mean": 0.08102035522460938, "KL/mean": -0.04034559428691864, "KL/rejected_KL_mean": -0.161712646484375, "KL/std": 0.3124745488166809, "epoch": 0.033773861967694566, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.24273207783699036, "fcm_dpo/q_t": 0.49393430352211, "grad_norm": 80.9083251953125, "learning_rate": 1.5942028985507245e-07, "logits/chosen": -0.5055763721466064, "logits/rejected": -0.48390600085258484, "logps/chosen": -59.996673583984375, "logps/ref_chosen": -60.07769775390625, "logps/ref_rejected": -81.13955688476562, "logps/rejected": -81.30126953125, "loss": 1.3626, "margin_dpo/margin_mean": 0.24273264408111572, "margin_dpo/margin_std": 0.39859479665756226, "step": 23 }, { "KL/chosen_KL_mean": 0.04727935791015625, "KL/mean": -0.07922174781560898, "KL/rejected_KL_mean": -0.20572662353515625, "KL/std": 0.33435285091400146, "epoch": 0.03524229074889868, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.25300681591033936, "fcm_dpo/q_t": 0.4936782717704773, "grad_norm": 82.37586212158203, "learning_rate": 1.6666666666666665e-07, "logits/chosen": -0.5237699747085571, "logits/rejected": -0.5069272518157959, "logps/chosen": -44.24375534057617, "logps/ref_chosen": -44.29103469848633, "logps/ref_rejected": -99.12521362304688, "logps/rejected": -99.33094024658203, "loss": 1.3617, "margin_dpo/margin_mean": 0.2530064284801483, "margin_dpo/margin_std": 0.4634859561920166, "step": 24 }, { "KL/chosen_KL_mean": 0.016876220703125, "KL/mean": -0.10173620283603668, "KL/rejected_KL_mean": -0.22034454345703125, "KL/std": 0.35340842604637146, "epoch": 0.03671071953010279, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.23721817135810852, "fcm_dpo/q_t": 0.4940733313560486, "grad_norm": 74.06179809570312, "learning_rate": 1.7391304347826085e-07, "logits/chosen": -0.5053163170814514, "logits/rejected": -0.4756869077682495, "logps/chosen": -52.52017593383789, "logps/ref_chosen": -52.537052154541016, "logps/ref_rejected": -89.34219360351562, "logps/rejected": -89.56253051757812, "loss": 1.3632, "margin_dpo/margin_mean": 0.2372182309627533, "margin_dpo/margin_std": 0.4329448938369751, "step": 25 }, { "KL/chosen_KL_mean": 0.08753395080566406, "KL/mean": -0.11949370801448822, "KL/rejected_KL_mean": -0.3265228271484375, "KL/std": 0.44881671667099, "epoch": 0.0381791483113069, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.4140579402446747, "fcm_dpo/q_t": 0.4896630048751831, "grad_norm": 87.02017974853516, "learning_rate": 1.8115942028985507e-07, "logits/chosen": -0.5420371294021606, "logits/rejected": -0.5105117559432983, "logps/chosen": -53.83527374267578, "logps/ref_chosen": -53.92280578613281, "logps/ref_rejected": -103.35971069335938, "logps/rejected": -103.68623352050781, "loss": 1.3462, "margin_dpo/margin_mean": 0.4140591323375702, "margin_dpo/margin_std": 0.555227518081665, "step": 26 }, { "KL/chosen_KL_mean": 0.07442474365234375, "KL/mean": -0.15376684069633484, "KL/rejected_KL_mean": -0.3819580078125, "KL/std": 0.41409561038017273, "epoch": 0.039647577092511016, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.45638173818588257, "fcm_dpo/q_t": 0.4886014759540558, "grad_norm": 94.21731567382812, "learning_rate": 1.8840579710144927e-07, "logits/chosen": -0.5366708040237427, "logits/rejected": -0.500505805015564, "logps/chosen": -42.82410430908203, "logps/ref_chosen": -42.898529052734375, "logps/ref_rejected": -98.72419738769531, "logps/rejected": -99.10615539550781, "loss": 1.3419, "margin_dpo/margin_mean": 0.4563814401626587, "margin_dpo/margin_std": 0.5116335153579712, "step": 27 }, { "KL/chosen_KL_mean": 0.017200469970703125, "KL/mean": -0.14496609568595886, "KL/rejected_KL_mean": -0.3071327209472656, "KL/std": 0.4189748764038086, "epoch": 0.041116005873715125, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.3243354558944702, "fcm_dpo/q_t": 0.4918993413448334, "grad_norm": 75.2711410522461, "learning_rate": 1.9565217391304347e-07, "logits/chosen": -0.5030197501182556, "logits/rejected": -0.4474295377731323, "logps/chosen": -60.53929901123047, "logps/ref_chosen": -60.55650329589844, "logps/ref_rejected": -91.40111541748047, "logps/rejected": -91.708251953125, "loss": 1.3549, "margin_dpo/margin_mean": 0.32433581352233887, "margin_dpo/margin_std": 0.5495618581771851, "step": 28 }, { "KL/chosen_KL_mean": 0.10011100769042969, "KL/mean": -0.1491682380437851, "KL/rejected_KL_mean": -0.3984489440917969, "KL/std": 0.4559759795665741, "epoch": 0.042584434654919234, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.4985598027706146, "fcm_dpo/q_t": 0.4875490665435791, "grad_norm": 90.70254516601562, "learning_rate": 2.028985507246377e-07, "logits/chosen": -0.5579368472099304, "logits/rejected": -0.5119800567626953, "logps/chosen": -57.707672119140625, "logps/ref_chosen": -57.80778503417969, "logps/ref_rejected": -97.39434814453125, "logps/rejected": -97.79280090332031, "loss": 1.3378, "margin_dpo/margin_mean": 0.4985605478286743, "margin_dpo/margin_std": 0.5526378154754639, "step": 29 }, { "KL/chosen_KL_mean": 0.13415145874023438, "KL/mean": -0.2075020968914032, "KL/rejected_KL_mean": -0.5491600036621094, "KL/std": 0.5850201845169067, "epoch": 0.04405286343612335, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.6833061575889587, "fcm_dpo/q_t": 0.4829457402229309, "grad_norm": 87.75554656982422, "learning_rate": 2.1014492753623187e-07, "logits/chosen": -0.4961678981781006, "logits/rejected": -0.46552735567092896, "logps/chosen": -52.44322204589844, "logps/ref_chosen": -52.577369689941406, "logps/ref_rejected": -98.48920440673828, "logps/rejected": -99.03836059570312, "loss": 1.3202, "margin_dpo/margin_mean": 0.683306097984314, "margin_dpo/margin_std": 0.6441049575805664, "step": 30 }, { "KL/chosen_KL_mean": 0.11783027648925781, "KL/mean": -0.1323196440935135, "KL/rejected_KL_mean": -0.3824729919433594, "KL/std": 0.5714821219444275, "epoch": 0.04552129221732746, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.5003038644790649, "fcm_dpo/q_t": 0.4875211715698242, "grad_norm": 67.77152252197266, "learning_rate": 2.1739130434782607e-07, "logits/chosen": -0.4899521470069885, "logits/rejected": -0.44324806332588196, "logps/chosen": -63.689090728759766, "logps/ref_chosen": -63.806922912597656, "logps/ref_rejected": -72.89400482177734, "logps/rejected": -73.27647399902344, "loss": 1.3382, "margin_dpo/margin_mean": 0.5003036260604858, "margin_dpo/margin_std": 0.7216464877128601, "step": 31 }, { "KL/chosen_KL_mean": 0.15192794799804688, "KL/mean": -0.22155286371707916, "KL/rejected_KL_mean": -0.59503173828125, "KL/std": 0.7016799449920654, "epoch": 0.04698972099853157, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.7469567060470581, "fcm_dpo/q_t": 0.4813820719718933, "grad_norm": 82.0534896850586, "learning_rate": 2.2463768115942027e-07, "logits/chosen": -0.5268591046333313, "logits/rejected": -0.48644328117370605, "logps/chosen": -62.58759689331055, "logps/ref_chosen": -62.739524841308594, "logps/ref_rejected": -89.3175048828125, "logps/rejected": -89.91253662109375, "loss": 1.3148, "margin_dpo/margin_mean": 0.7469568252563477, "margin_dpo/margin_std": 0.8592814803123474, "step": 32 }, { "KL/chosen_KL_mean": 0.10763168334960938, "KL/mean": -0.20983782410621643, "KL/rejected_KL_mean": -0.5273056030273438, "KL/std": 0.5811012983322144, "epoch": 0.048458149779735685, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.6349318623542786, "fcm_dpo/q_t": 0.4841501712799072, "grad_norm": 72.05321502685547, "learning_rate": 2.318840579710145e-07, "logits/chosen": -0.4794453978538513, "logits/rejected": -0.45305970311164856, "logps/chosen": -53.153343200683594, "logps/ref_chosen": -53.26097106933594, "logps/ref_rejected": -87.8851318359375, "logps/rejected": -88.41244506835938, "loss": 1.3249, "margin_dpo/margin_mean": 0.6349321603775024, "margin_dpo/margin_std": 0.6593722701072693, "step": 33 }, { "KL/chosen_KL_mean": 0.07497215270996094, "KL/mean": -0.299121618270874, "KL/rejected_KL_mean": -0.6732177734375, "KL/std": 0.7037482857704163, "epoch": 0.049926578560939794, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.748187780380249, "fcm_dpo/q_t": 0.4813481569290161, "grad_norm": 77.933837890625, "learning_rate": 2.391304347826087e-07, "logits/chosen": -0.5027347803115845, "logits/rejected": -0.48531198501586914, "logps/chosen": -50.74235534667969, "logps/ref_chosen": -50.81732940673828, "logps/ref_rejected": -101.92184448242188, "logps/rejected": -102.59506225585938, "loss": 1.3148, "margin_dpo/margin_mean": 0.7481868267059326, "margin_dpo/margin_std": 0.8796716928482056, "step": 34 }, { "KL/chosen_KL_mean": 0.14807891845703125, "KL/mean": -0.4777490496635437, "KL/rejected_KL_mean": -1.10357666015625, "KL/std": 0.9910285472869873, "epoch": 0.0513950073421439, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.2516549825668335, "fcm_dpo/q_t": 0.4688939154148102, "grad_norm": 82.58599090576172, "learning_rate": 2.463768115942029e-07, "logits/chosen": -0.531462550163269, "logits/rejected": -0.4944959580898285, "logps/chosen": -50.87641143798828, "logps/ref_chosen": -51.02449035644531, "logps/ref_rejected": -106.82443237304688, "logps/rejected": -107.92800903320312, "loss": 1.268, "margin_dpo/margin_mean": 1.2516546249389648, "margin_dpo/margin_std": 1.0724127292633057, "step": 35 }, { "KL/chosen_KL_mean": 0.06724357604980469, "KL/mean": -0.5083455443382263, "KL/rejected_KL_mean": -1.0839347839355469, "KL/std": 1.0932717323303223, "epoch": 0.05286343612334802, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.1511791944503784, "fcm_dpo/q_t": 0.47139599919319153, "grad_norm": 72.46247100830078, "learning_rate": 2.536231884057971e-07, "logits/chosen": -0.5766887664794922, "logits/rejected": -0.5414172410964966, "logps/chosen": -51.92424774169922, "logps/ref_chosen": -51.991493225097656, "logps/ref_rejected": -86.0406265258789, "logps/rejected": -87.12455749511719, "loss": 1.2787, "margin_dpo/margin_mean": 1.1511785984039307, "margin_dpo/margin_std": 1.2563061714172363, "step": 36 }, { "KL/chosen_KL_mean": 0.0029850006103515625, "KL/mean": -0.5103100538253784, "KL/rejected_KL_mean": -1.0236091613769531, "KL/std": 1.0585589408874512, "epoch": 0.05433186490455213, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.0265923738479614, "fcm_dpo/q_t": 0.474545419216156, "grad_norm": 61.76712417602539, "learning_rate": 2.6086956521739126e-07, "logits/chosen": -0.5153898000717163, "logits/rejected": -0.4721587300300598, "logps/chosen": -62.80412292480469, "logps/ref_chosen": -62.807106018066406, "logps/ref_rejected": -77.89507293701172, "logps/rejected": -78.91868591308594, "loss": 1.2909, "margin_dpo/margin_mean": 1.0265920162200928, "margin_dpo/margin_std": 1.3249099254608154, "step": 37 }, { "KL/chosen_KL_mean": 0.15967178344726562, "KL/mean": -0.5315895080566406, "KL/rejected_KL_mean": -1.2228507995605469, "KL/std": 1.3578490018844604, "epoch": 0.055800293685756244, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.382524847984314, "fcm_dpo/q_t": 0.4658650755882263, "grad_norm": 69.12739562988281, "learning_rate": 2.681159420289855e-07, "logits/chosen": -0.5189441442489624, "logits/rejected": -0.4861838221549988, "logps/chosen": -48.23085021972656, "logps/ref_chosen": -48.39051818847656, "logps/ref_rejected": -97.91244506835938, "logps/rejected": -99.13529205322266, "loss": 1.2596, "margin_dpo/margin_mean": 1.3825253248214722, "margin_dpo/margin_std": 1.6585068702697754, "step": 38 }, { "KL/chosen_KL_mean": 0.07357215881347656, "KL/mean": -0.7325087785720825, "KL/rejected_KL_mean": -1.5385856628417969, "KL/std": 1.2871696949005127, "epoch": 0.05726872246696035, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.612156629562378, "fcm_dpo/q_t": 0.4599946141242981, "grad_norm": 73.47403717041016, "learning_rate": 2.753623188405797e-07, "logits/chosen": -0.5452001094818115, "logits/rejected": -0.5045498013496399, "logps/chosen": -50.676902770996094, "logps/ref_chosen": -50.75047302246094, "logps/ref_rejected": -78.56951141357422, "logps/rejected": -80.10809326171875, "loss": 1.2365, "margin_dpo/margin_mean": 1.612156867980957, "margin_dpo/margin_std": 1.3819551467895508, "step": 39 }, { "KL/chosen_KL_mean": 0.1919574737548828, "KL/mean": -0.5750719308853149, "KL/rejected_KL_mean": -1.34210205078125, "KL/std": 1.3759987354278564, "epoch": 0.05873715124816446, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.5340617895126343, "fcm_dpo/q_t": 0.4621471166610718, "grad_norm": 60.486148834228516, "learning_rate": 2.8260869565217386e-07, "logits/chosen": -0.5081291794776917, "logits/rejected": -0.477234423160553, "logps/chosen": -57.79310989379883, "logps/ref_chosen": -57.985069274902344, "logps/ref_rejected": -74.3000717163086, "logps/rejected": -75.64217376708984, "loss": 1.2461, "margin_dpo/margin_mean": 1.5340622663497925, "margin_dpo/margin_std": 1.6614691019058228, "step": 40 }, { "KL/chosen_KL_mean": 0.04014015197753906, "KL/mean": -0.9176149368286133, "KL/rejected_KL_mean": -1.8753662109375, "KL/std": 1.8172500133514404, "epoch": 0.06020558002936858, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.9155058860778809, "fcm_dpo/q_t": 0.4528723359107971, "grad_norm": 68.08113098144531, "learning_rate": 2.898550724637681e-07, "logits/chosen": -0.510447084903717, "logits/rejected": -0.4722723066806793, "logps/chosen": -62.655677795410156, "logps/ref_chosen": -62.69581604003906, "logps/ref_rejected": -97.02352905273438, "logps/rejected": -98.89889526367188, "loss": 1.2138, "margin_dpo/margin_mean": 1.9155060052871704, "margin_dpo/margin_std": 1.9704290628433228, "step": 41 }, { "KL/chosen_KL_mean": 0.2155475616455078, "KL/mean": -1.0993424654006958, "KL/rejected_KL_mean": -2.4142303466796875, "KL/std": 2.225177764892578, "epoch": 0.06167400881057269, "fcm_dpo/beta": 0.108363576233387, "fcm_dpo/delta": 0.529624342918396, "fcm_dpo/margin": 2.629781723022461, "fcm_dpo/q_t": 0.43431544303894043, "grad_norm": 83.5115737915039, "learning_rate": 2.971014492753623e-07, "logits/chosen": -0.5298647880554199, "logits/rejected": -0.48321884870529175, "logps/chosen": -58.750877380371094, "logps/ref_chosen": -58.966426849365234, "logps/ref_rejected": -109.90837097167969, "logps/rejected": -112.32260131835938, "loss": 1.1397, "margin_dpo/margin_mean": 2.629781723022461, "margin_dpo/margin_std": 2.5406675338745117, "step": 42 }, { "KL/chosen_KL_mean": 0.53778076171875, "KL/mean": -0.7240562438964844, "KL/rejected_KL_mean": -1.98590087890625, "KL/std": 1.8650085926055908, "epoch": 0.0631424375917768, "fcm_dpo/beta": 0.11999893188476562, "fcm_dpo/delta": 0.5129410028457642, "fcm_dpo/margin": 2.5236780643463135, "fcm_dpo/q_t": 0.42989984154701233, "grad_norm": 82.07591247558594, "learning_rate": 3.043478260869565e-07, "logits/chosen": -0.5556699633598328, "logits/rejected": -0.5314440727233887, "logps/chosen": -53.61821746826172, "logps/ref_chosen": -54.15599822998047, "logps/ref_rejected": -96.48019409179688, "logps/rejected": -98.46609497070312, "loss": 1.1214, "margin_dpo/margin_mean": 2.5236783027648926, "margin_dpo/margin_std": 2.0528616905212402, "step": 43 }, { "KL/chosen_KL_mean": 0.22237205505371094, "KL/mean": -1.1844167709350586, "KL/rejected_KL_mean": -2.591205596923828, "KL/std": 2.265498638153076, "epoch": 0.06461086637298091, "fcm_dpo/beta": 0.1317014992237091, "fcm_dpo/delta": 0.445910781621933, "fcm_dpo/margin": 2.8135740756988525, "fcm_dpo/q_t": 0.4144276976585388, "grad_norm": 95.687255859375, "learning_rate": 3.115942028985507e-07, "logits/chosen": -0.4916815161705017, "logits/rejected": -0.4712473750114441, "logps/chosen": -49.85612487792969, "logps/ref_chosen": -50.07849884033203, "logps/ref_rejected": -108.78376007080078, "logps/rejected": -111.37496948242188, "loss": 1.0708, "margin_dpo/margin_mean": 2.8135738372802734, "margin_dpo/margin_std": 2.256591796875, "step": 44 }, { "KL/chosen_KL_mean": 0.11939048767089844, "KL/mean": -1.027750015258789, "KL/rejected_KL_mean": -2.1748924255371094, "KL/std": 2.073944091796875, "epoch": 0.06607929515418502, "fcm_dpo/beta": 0.1451357901096344, "fcm_dpo/delta": 0.4816334843635559, "fcm_dpo/margin": 2.2942895889282227, "fcm_dpo/q_t": 0.42459607124328613, "grad_norm": 80.60940551757812, "learning_rate": 3.188405797101449e-07, "logits/chosen": -0.4741925597190857, "logits/rejected": -0.4613730311393738, "logps/chosen": -48.29553985595703, "logps/ref_chosen": -48.4149284362793, "logps/ref_rejected": -77.93643188476562, "logps/rejected": -80.11132049560547, "loss": 1.1144, "margin_dpo/margin_mean": 2.2942888736724854, "margin_dpo/margin_std": 2.5776329040527344, "step": 45 }, { "KL/chosen_KL_mean": 0.1797943115234375, "KL/mean": -1.272498369216919, "KL/rejected_KL_mean": -2.72479248046875, "KL/std": 2.6093244552612305, "epoch": 0.06754772393538913, "fcm_dpo/beta": 0.15693849325180054, "fcm_dpo/delta": 0.3592992424964905, "fcm_dpo/margin": 2.9045841693878174, "fcm_dpo/q_t": 0.3983314037322998, "grad_norm": 93.62377166748047, "learning_rate": 3.260869565217391e-07, "logits/chosen": -0.549965500831604, "logits/rejected": -0.5000295639038086, "logps/chosen": -55.81963348388672, "logps/ref_chosen": -55.999427795410156, "logps/ref_rejected": -95.652587890625, "logps/rejected": -98.37738037109375, "loss": 1.0358, "margin_dpo/margin_mean": 2.9045839309692383, "margin_dpo/margin_std": 3.1410748958587646, "step": 46 }, { "KL/chosen_KL_mean": 0.3869171142578125, "KL/mean": -1.106046199798584, "KL/rejected_KL_mean": -2.599010467529297, "KL/std": 2.578082799911499, "epoch": 0.06901615271659324, "fcm_dpo/beta": 0.16788116097450256, "fcm_dpo/delta": 0.3120081424713135, "fcm_dpo/margin": 2.985929012298584, "fcm_dpo/q_t": 0.3863651156425476, "grad_norm": 92.04369354248047, "learning_rate": 3.333333333333333e-07, "logits/chosen": -0.5671145915985107, "logits/rejected": -0.5139440298080444, "logps/chosen": -57.539161682128906, "logps/ref_chosen": -57.92607879638672, "logps/ref_rejected": -94.67920684814453, "logps/rejected": -97.27821350097656, "loss": 0.9928, "margin_dpo/margin_mean": 2.985929489135742, "margin_dpo/margin_std": 2.6357438564300537, "step": 47 }, { "KL/chosen_KL_mean": 0.05410575866699219, "KL/mean": -1.496936559677124, "KL/rejected_KL_mean": -3.0479812622070312, "KL/std": 2.6523523330688477, "epoch": 0.07048458149779736, "fcm_dpo/beta": 0.17588907480239868, "fcm_dpo/delta": 0.2697899341583252, "fcm_dpo/margin": 3.102083444595337, "fcm_dpo/q_t": 0.3784920573234558, "grad_norm": 104.22386932373047, "learning_rate": 3.4057971014492755e-07, "logits/chosen": -0.5789310932159424, "logits/rejected": -0.519860565662384, "logps/chosen": -57.133968353271484, "logps/ref_chosen": -57.188072204589844, "logps/ref_rejected": -88.0166015625, "logps/rejected": -91.06458282470703, "loss": 0.9819, "margin_dpo/margin_mean": 3.102083444595337, "margin_dpo/margin_std": 2.9072694778442383, "step": 48 }, { "KL/chosen_KL_mean": 0.3045520782470703, "KL/mean": -1.5718982219696045, "KL/rejected_KL_mean": -3.448345184326172, "KL/std": 3.377166748046875, "epoch": 0.07195301027900147, "fcm_dpo/beta": 0.1832749843597412, "fcm_dpo/delta": 0.11903807520866394, "fcm_dpo/margin": 3.752896547317505, "fcm_dpo/q_t": 0.3552005887031555, "grad_norm": 87.47305297851562, "learning_rate": 3.478260869565217e-07, "logits/chosen": -0.5585795640945435, "logits/rejected": -0.5012216567993164, "logps/chosen": -61.38071823120117, "logps/ref_chosen": -61.685272216796875, "logps/ref_rejected": -83.76747131347656, "logps/rejected": -87.2158203125, "loss": 0.9196, "margin_dpo/margin_mean": 3.752896308898926, "margin_dpo/margin_std": 3.921967029571533, "step": 49 }, { "KL/chosen_KL_mean": -0.16571998596191406, "KL/mean": -2.2112300395965576, "KL/rejected_KL_mean": -4.256736755371094, "KL/std": 3.468242645263672, "epoch": 0.07342143906020558, "fcm_dpo/beta": 0.18442076444625854, "fcm_dpo/delta": 0.04927580803632736, "fcm_dpo/margin": 4.091017246246338, "fcm_dpo/q_t": 0.34040141105651855, "grad_norm": 90.17442321777344, "learning_rate": 3.5507246376811595e-07, "logits/chosen": -0.5344624519348145, "logits/rejected": -0.49823105335235596, "logps/chosen": -58.88985824584961, "logps/ref_chosen": -58.72413635253906, "logps/ref_rejected": -96.35814666748047, "logps/rejected": -100.61488342285156, "loss": 0.8788, "margin_dpo/margin_mean": 4.091017246246338, "margin_dpo/margin_std": 3.9831931591033936, "step": 50 }, { "KL/chosen_KL_mean": -0.31114959716796875, "KL/mean": -2.3097453117370605, "KL/rejected_KL_mean": -4.308345794677734, "KL/std": 4.11016845703125, "epoch": 0.07488986784140969, "fcm_dpo/beta": 0.18537896871566772, "fcm_dpo/delta": 0.06288561224937439, "fcm_dpo/margin": 3.997190237045288, "fcm_dpo/q_t": 0.35509026050567627, "grad_norm": 76.0972900390625, "learning_rate": 3.6231884057971015e-07, "logits/chosen": -0.5231320858001709, "logits/rejected": -0.49016422033309937, "logps/chosen": -61.684818267822266, "logps/ref_chosen": -61.3736686706543, "logps/ref_rejected": -76.00199890136719, "logps/rejected": -80.31034851074219, "loss": 0.9433, "margin_dpo/margin_mean": 3.997189998626709, "margin_dpo/margin_std": 5.026128768920898, "step": 51 }, { "KL/chosen_KL_mean": 0.3505516052246094, "KL/mean": -2.698451519012451, "KL/rejected_KL_mean": -5.747459411621094, "KL/std": 4.724987030029297, "epoch": 0.0763582966226138, "fcm_dpo/beta": 0.17869603633880615, "fcm_dpo/delta": -0.32520201802253723, "fcm_dpo/margin": 6.09800910949707, "fcm_dpo/q_t": 0.27982762455940247, "grad_norm": 72.83785247802734, "learning_rate": 3.695652173913043e-07, "logits/chosen": -0.5447903871536255, "logits/rejected": -0.4885804057121277, "logps/chosen": -51.9868049621582, "logps/ref_chosen": -52.33735656738281, "logps/ref_rejected": -79.97391510009766, "logps/rejected": -85.72137451171875, "loss": 0.7172, "margin_dpo/margin_mean": 6.09800910949707, "margin_dpo/margin_std": 5.0404887199401855, "step": 52 }, { "KL/chosen_KL_mean": -0.15282440185546875, "KL/mean": -3.2396836280822754, "KL/rejected_KL_mean": -6.326541900634766, "KL/std": 5.3033671379089355, "epoch": 0.07782672540381791, "fcm_dpo/beta": 0.17080718278884888, "fcm_dpo/delta": -0.284501850605011, "fcm_dpo/margin": 6.173723220825195, "fcm_dpo/q_t": 0.3035929799079895, "grad_norm": 79.4716796875, "learning_rate": 3.7681159420289855e-07, "logits/chosen": -0.6213881969451904, "logits/rejected": -0.6000795364379883, "logps/chosen": -53.46747589111328, "logps/ref_chosen": -53.31465148925781, "logps/ref_rejected": -91.78359985351562, "logps/rejected": -98.11013793945312, "loss": 0.7851, "margin_dpo/margin_mean": 6.173722743988037, "margin_dpo/margin_std": 6.064870834350586, "step": 53 }, { "KL/chosen_KL_mean": -0.37441253662109375, "KL/mean": -3.029232978820801, "KL/rejected_KL_mean": -5.684051513671875, "KL/std": 5.031246662139893, "epoch": 0.07929515418502203, "fcm_dpo/beta": 0.16453491151332855, "fcm_dpo/delta": -0.08063030242919922, "fcm_dpo/margin": 5.309640407562256, "fcm_dpo/q_t": 0.321298748254776, "grad_norm": 69.15568542480469, "learning_rate": 3.8405797101449274e-07, "logits/chosen": -0.5816128253936768, "logits/rejected": -0.5275565385818481, "logps/chosen": -51.063072204589844, "logps/ref_chosen": -50.68865966796875, "logps/ref_rejected": -91.71539306640625, "logps/rejected": -97.39944458007812, "loss": 0.8212, "margin_dpo/margin_mean": 5.309639930725098, "margin_dpo/margin_std": 5.0380449295043945, "step": 54 }, { "KL/chosen_KL_mean": -0.8430213928222656, "KL/mean": -4.0328216552734375, "KL/rejected_KL_mean": -7.222618103027344, "KL/std": 6.049294471740723, "epoch": 0.08076358296622614, "fcm_dpo/beta": 0.1586906611919403, "fcm_dpo/delta": -0.23564836382865906, "fcm_dpo/margin": 6.379594802856445, "fcm_dpo/q_t": 0.31021779775619507, "grad_norm": 74.39564514160156, "learning_rate": 3.9130434782608694e-07, "logits/chosen": -0.6505610942840576, "logits/rejected": -0.5886675119400024, "logps/chosen": -63.458255767822266, "logps/ref_chosen": -62.615234375, "logps/ref_rejected": -88.99349975585938, "logps/rejected": -96.21612548828125, "loss": 0.8466, "margin_dpo/margin_mean": 6.379594802856445, "margin_dpo/margin_std": 7.276340484619141, "step": 55 }, { "KL/chosen_KL_mean": -0.6147956848144531, "KL/mean": -3.5950727462768555, "KL/rejected_KL_mean": -6.575355529785156, "KL/std": 5.803818702697754, "epoch": 0.08223201174743025, "fcm_dpo/beta": 0.1531292051076889, "fcm_dpo/delta": -0.12394850701093674, "fcm_dpo/margin": 5.960559844970703, "fcm_dpo/q_t": 0.32575082778930664, "grad_norm": 61.359893798828125, "learning_rate": 3.9855072463768114e-07, "logits/chosen": -0.6318885087966919, "logits/rejected": -0.5902144908905029, "logps/chosen": -58.547523498535156, "logps/ref_chosen": -57.9327278137207, "logps/ref_rejected": -94.1744384765625, "logps/rejected": -100.74979400634766, "loss": 0.8844, "margin_dpo/margin_mean": 5.9605607986450195, "margin_dpo/margin_std": 7.014354705810547, "step": 56 }, { "KL/chosen_KL_mean": -0.6196041107177734, "KL/mean": -3.9665966033935547, "KL/rejected_KL_mean": -7.313587188720703, "KL/std": 5.700008392333984, "epoch": 0.08370044052863436, "fcm_dpo/beta": 0.14832691848278046, "fcm_dpo/delta": -0.2137039601802826, "fcm_dpo/margin": 6.693978309631348, "fcm_dpo/q_t": 0.29930901527404785, "grad_norm": 65.74341583251953, "learning_rate": 4.057971014492754e-07, "logits/chosen": -0.5702251195907593, "logits/rejected": -0.5417561531066895, "logps/chosen": -71.1148910522461, "logps/ref_chosen": -70.49528503417969, "logps/ref_rejected": -95.56546020507812, "logps/rejected": -102.87904357910156, "loss": 0.7907, "margin_dpo/margin_mean": 6.693978309631348, "margin_dpo/margin_std": 6.242074966430664, "step": 57 }, { "KL/chosen_KL_mean": -0.8837299346923828, "KL/mean": -4.569659233093262, "KL/rejected_KL_mean": -8.255584716796875, "KL/std": 6.463259220123291, "epoch": 0.08516886930983847, "fcm_dpo/beta": 0.14102259278297424, "fcm_dpo/delta": -0.26656439900398254, "fcm_dpo/margin": 7.371853351593018, "fcm_dpo/q_t": 0.3033452033996582, "grad_norm": 75.09760284423828, "learning_rate": 4.1304347826086954e-07, "logits/chosen": -0.5813875794410706, "logits/rejected": -0.5039342641830444, "logps/chosen": -63.01667022705078, "logps/ref_chosen": -62.13294219970703, "logps/ref_rejected": -84.61729431152344, "logps/rejected": -92.87287902832031, "loss": 0.8067, "margin_dpo/margin_mean": 7.371853828430176, "margin_dpo/margin_std": 7.66532039642334, "step": 58 }, { "KL/chosen_KL_mean": -1.2969722747802734, "KL/mean": -5.302949905395508, "KL/rejected_KL_mean": -9.30892562866211, "KL/std": 7.01795768737793, "epoch": 0.08663729809104258, "fcm_dpo/beta": 0.13059750199317932, "fcm_dpo/delta": -0.2830684185028076, "fcm_dpo/margin": 8.011954307556152, "fcm_dpo/q_t": 0.3002108633518219, "grad_norm": 64.80830383300781, "learning_rate": 4.2028985507246374e-07, "logits/chosen": -0.6242841482162476, "logits/rejected": -0.5837790966033936, "logps/chosen": -53.22949981689453, "logps/ref_chosen": -51.932525634765625, "logps/ref_rejected": -88.88520050048828, "logps/rejected": -98.19412231445312, "loss": 0.7975, "margin_dpo/margin_mean": 8.011954307556152, "margin_dpo/margin_std": 8.137838363647461, "step": 59 }, { "KL/chosen_KL_mean": -2.28204345703125, "KL/mean": -5.468286514282227, "KL/rejected_KL_mean": -8.654525756835938, "KL/std": 6.25811767578125, "epoch": 0.0881057268722467, "fcm_dpo/beta": 0.12927772104740143, "fcm_dpo/delta": -0.026638053357601166, "fcm_dpo/margin": 6.372478485107422, "fcm_dpo/q_t": 0.3292919993400574, "grad_norm": 78.46163177490234, "learning_rate": 4.2753623188405794e-07, "logits/chosen": -0.6169058084487915, "logits/rejected": -0.5573669672012329, "logps/chosen": -63.2242317199707, "logps/ref_chosen": -60.94218826293945, "logps/ref_rejected": -85.39340209960938, "logps/rejected": -94.04792785644531, "loss": 0.9026, "margin_dpo/margin_mean": 6.3724775314331055, "margin_dpo/margin_std": 7.160068035125732, "step": 60 }, { "KL/chosen_KL_mean": -1.1401500701904297, "KL/mean": -5.068685531616211, "KL/rejected_KL_mean": -8.997222900390625, "KL/std": 8.527802467346191, "epoch": 0.08957415565345081, "fcm_dpo/beta": 0.1250603199005127, "fcm_dpo/delta": -0.20481759309768677, "fcm_dpo/margin": 7.857068061828613, "fcm_dpo/q_t": 0.326485812664032, "grad_norm": 63.45138931274414, "learning_rate": 4.3478260869565214e-07, "logits/chosen": -0.6488014459609985, "logits/rejected": -0.6157772541046143, "logps/chosen": -61.77367401123047, "logps/ref_chosen": -60.633522033691406, "logps/ref_rejected": -89.85249328613281, "logps/rejected": -98.84971618652344, "loss": 0.891, "margin_dpo/margin_mean": 7.857068061828613, "margin_dpo/margin_std": 10.577825546264648, "step": 61 }, { "KL/chosen_KL_mean": -1.325113296508789, "KL/mean": -4.2535576820373535, "KL/rejected_KL_mean": -7.181999206542969, "KL/std": 6.108260154724121, "epoch": 0.09104258443465492, "fcm_dpo/beta": 0.1265747994184494, "fcm_dpo/delta": 0.06192735210061073, "fcm_dpo/margin": 5.856889724731445, "fcm_dpo/q_t": 0.353201687335968, "grad_norm": 66.93692779541016, "learning_rate": 4.420289855072464e-07, "logits/chosen": -0.6124836206436157, "logits/rejected": -0.5789198875427246, "logps/chosen": -57.47588348388672, "logps/ref_chosen": -56.15077209472656, "logps/ref_rejected": -75.56619262695312, "logps/rejected": -82.74819946289062, "loss": 0.9536, "margin_dpo/margin_mean": 5.856890678405762, "margin_dpo/margin_std": 7.419264316558838, "step": 62 }, { "KL/chosen_KL_mean": -2.2093944549560547, "KL/mean": -6.111651420593262, "KL/rejected_KL_mean": -10.013912200927734, "KL/std": 7.733986854553223, "epoch": 0.09251101321585903, "fcm_dpo/beta": 0.12191849946975708, "fcm_dpo/delta": -0.17046590149402618, "fcm_dpo/margin": 7.8045125007629395, "fcm_dpo/q_t": 0.30967673659324646, "grad_norm": 63.06834411621094, "learning_rate": 4.4927536231884053e-07, "logits/chosen": -0.5985692739486694, "logits/rejected": -0.5528501272201538, "logps/chosen": -75.3567886352539, "logps/ref_chosen": -73.14739227294922, "logps/ref_rejected": -97.61006164550781, "logps/rejected": -107.62397766113281, "loss": 0.8395, "margin_dpo/margin_mean": 7.804512023925781, "margin_dpo/margin_std": 8.159123420715332, "step": 63 }, { "KL/chosen_KL_mean": -0.6886730194091797, "KL/mean": -5.2321319580078125, "KL/rejected_KL_mean": -9.77560043334961, "KL/std": 8.251523971557617, "epoch": 0.09397944199706314, "fcm_dpo/beta": 0.11495494097471237, "fcm_dpo/delta": -0.28372639417648315, "fcm_dpo/margin": 9.086919784545898, "fcm_dpo/q_t": 0.3031163811683655, "grad_norm": 56.18279266357422, "learning_rate": 4.5652173913043473e-07, "logits/chosen": -0.563762366771698, "logits/rejected": -0.5307266712188721, "logps/chosen": -54.68727111816406, "logps/ref_chosen": -53.998600006103516, "logps/ref_rejected": -93.53019714355469, "logps/rejected": -103.30580139160156, "loss": 0.8288, "margin_dpo/margin_mean": 9.086919784545898, "margin_dpo/margin_std": 10.110519409179688, "step": 64 }, { "KL/chosen_KL_mean": -2.598611831665039, "KL/mean": -7.124964237213135, "KL/rejected_KL_mean": -11.651321411132812, "KL/std": 9.134474754333496, "epoch": 0.09544787077826726, "fcm_dpo/beta": 0.11054727435112, "fcm_dpo/delta": -0.22596801817417145, "fcm_dpo/margin": 9.052715301513672, "fcm_dpo/q_t": 0.3050973117351532, "grad_norm": 57.133758544921875, "learning_rate": 4.63768115942029e-07, "logits/chosen": -0.6590486764907837, "logits/rejected": -0.6450433731079102, "logps/chosen": -67.43461608886719, "logps/ref_chosen": -64.83599853515625, "logps/ref_rejected": -109.94645690917969, "logps/rejected": -121.5977783203125, "loss": 0.8482, "margin_dpo/margin_mean": 9.052714347839355, "margin_dpo/margin_std": 9.923721313476562, "step": 65 }, { "KL/chosen_KL_mean": -2.3348751068115234, "KL/mean": -6.542342185974121, "KL/rejected_KL_mean": -10.749805450439453, "KL/std": 8.355534553527832, "epoch": 0.09691629955947137, "fcm_dpo/beta": 0.10701608657836914, "fcm_dpo/delta": -0.11439692974090576, "fcm_dpo/margin": 8.41492748260498, "fcm_dpo/q_t": 0.3284803032875061, "grad_norm": 53.33970260620117, "learning_rate": 4.7101449275362313e-07, "logits/chosen": -0.6295303702354431, "logits/rejected": -0.5955126285552979, "logps/chosen": -53.77840042114258, "logps/ref_chosen": -51.44352722167969, "logps/ref_rejected": -75.63629913330078, "logps/rejected": -86.3861083984375, "loss": 0.8783, "margin_dpo/margin_mean": 8.414926528930664, "margin_dpo/margin_std": 10.007061004638672, "step": 66 }, { "KL/chosen_KL_mean": -1.9881420135498047, "KL/mean": -6.224081039428711, "KL/rejected_KL_mean": -10.46002197265625, "KL/std": 8.115812301635742, "epoch": 0.09838472834067548, "fcm_dpo/beta": 0.1063048467040062, "fcm_dpo/delta": -0.11047748476266861, "fcm_dpo/margin": 8.47187614440918, "fcm_dpo/q_t": 0.3267403841018677, "grad_norm": 56.267860412597656, "learning_rate": 4.782608695652174e-07, "logits/chosen": -0.5961214303970337, "logits/rejected": -0.5539319515228271, "logps/chosen": -61.328948974609375, "logps/ref_chosen": -59.34080505371094, "logps/ref_rejected": -72.78728485107422, "logps/rejected": -83.24730682373047, "loss": 0.8694, "margin_dpo/margin_mean": 8.47187614440918, "margin_dpo/margin_std": 9.791044235229492, "step": 67 }, { "KL/chosen_KL_mean": -2.0880203247070312, "KL/mean": -6.178552627563477, "KL/rejected_KL_mean": -10.269084930419922, "KL/std": 7.334725856781006, "epoch": 0.09985315712187959, "fcm_dpo/beta": 0.10470952093601227, "fcm_dpo/delta": -0.062055252492427826, "fcm_dpo/margin": 8.181063652038574, "fcm_dpo/q_t": 0.32555100321769714, "grad_norm": 51.56489181518555, "learning_rate": 4.855072463768116e-07, "logits/chosen": -0.6386753916740417, "logits/rejected": -0.5816385746002197, "logps/chosen": -67.29385375976562, "logps/ref_chosen": -65.2058334350586, "logps/ref_rejected": -77.20724487304688, "logps/rejected": -87.47633361816406, "loss": 0.844, "margin_dpo/margin_mean": 8.181063652038574, "margin_dpo/margin_std": 8.15043830871582, "step": 68 }, { "KL/chosen_KL_mean": -2.4980125427246094, "KL/mean": -7.213038444519043, "KL/rejected_KL_mean": -11.92806625366211, "KL/std": 8.368502616882324, "epoch": 0.1013215859030837, "fcm_dpo/beta": 0.10073809325695038, "fcm_dpo/delta": -0.16765105724334717, "fcm_dpo/margin": 9.430049896240234, "fcm_dpo/q_t": 0.3103695511817932, "grad_norm": 52.50539779663086, "learning_rate": 4.927536231884058e-07, "logits/chosen": -0.5846946239471436, "logits/rejected": -0.5604379177093506, "logps/chosen": -62.31725311279297, "logps/ref_chosen": -59.81924057006836, "logps/ref_rejected": -103.38886260986328, "logps/rejected": -115.31692504882812, "loss": 0.7993, "margin_dpo/margin_mean": 9.430049896240234, "margin_dpo/margin_std": 9.154158592224121, "step": 69 }, { "KL/chosen_KL_mean": -3.4951820373535156, "KL/mean": -8.612916946411133, "KL/rejected_KL_mean": -13.73065185546875, "KL/std": 9.930105209350586, "epoch": 0.1027900146842878, "fcm_dpo/beta": 0.09773986041545868, "fcm_dpo/delta": -0.22211110591888428, "fcm_dpo/margin": 10.235469818115234, "fcm_dpo/q_t": 0.3081403374671936, "grad_norm": 54.948020935058594, "learning_rate": 5e-07, "logits/chosen": -0.6405221819877625, "logits/rejected": -0.6063104867935181, "logps/chosen": -65.42582702636719, "logps/ref_chosen": -61.930641174316406, "logps/ref_rejected": -91.06078338623047, "logps/rejected": -104.79143524169922, "loss": 0.811, "margin_dpo/margin_mean": 10.23547077178955, "margin_dpo/margin_std": 10.585088729858398, "step": 70 }, { "KL/chosen_KL_mean": -2.8669185638427734, "KL/mean": -8.555381774902344, "KL/rejected_KL_mean": -14.243846893310547, "KL/std": 9.787435531616211, "epoch": 0.10425844346549193, "fcm_dpo/beta": 0.09209594875574112, "fcm_dpo/delta": -0.2774484157562256, "fcm_dpo/margin": 11.37692642211914, "fcm_dpo/q_t": 0.29503485560417175, "grad_norm": 47.97392272949219, "learning_rate": 4.999967061337492e-07, "logits/chosen": -0.6438924074172974, "logits/rejected": -0.6015328168869019, "logps/chosen": -64.61725616455078, "logps/ref_chosen": -61.750335693359375, "logps/ref_rejected": -97.33662414550781, "logps/rejected": -111.5804672241211, "loss": 0.7634, "margin_dpo/margin_mean": 11.37692642211914, "margin_dpo/margin_std": 10.718829154968262, "step": 71 }, { "KL/chosen_KL_mean": -2.9292354583740234, "KL/mean": -8.912168502807617, "KL/rejected_KL_mean": -14.895095825195312, "KL/std": 10.194162368774414, "epoch": 0.10572687224669604, "fcm_dpo/beta": 0.08702956140041351, "fcm_dpo/delta": -0.27092522382736206, "fcm_dpo/margin": 11.965860366821289, "fcm_dpo/q_t": 0.2947954833507538, "grad_norm": 52.500999450683594, "learning_rate": 4.999868246217933e-07, "logits/chosen": -0.6367168426513672, "logits/rejected": -0.6010754704475403, "logps/chosen": -68.98265075683594, "logps/ref_chosen": -66.05341339111328, "logps/ref_rejected": -95.2869873046875, "logps/rejected": -110.18208312988281, "loss": 0.7728, "margin_dpo/margin_mean": 11.965860366821289, "margin_dpo/margin_std": 11.45088005065918, "step": 72 }, { "KL/chosen_KL_mean": -4.294904708862305, "KL/mean": -10.067806243896484, "KL/rejected_KL_mean": -15.840709686279297, "KL/std": 12.866714477539062, "epoch": 0.10719530102790015, "fcm_dpo/beta": 0.08335524797439575, "fcm_dpo/delta": -0.1816135048866272, "fcm_dpo/margin": 11.545804977416992, "fcm_dpo/q_t": 0.330466091632843, "grad_norm": 52.436309814453125, "learning_rate": 4.999703557245192e-07, "logits/chosen": -0.6677882671356201, "logits/rejected": -0.6248770952224731, "logps/chosen": -70.55117797851562, "logps/ref_chosen": -66.25627136230469, "logps/ref_rejected": -90.45613098144531, "logps/rejected": -106.29684448242188, "loss": 0.9326, "margin_dpo/margin_mean": 11.545804977416992, "margin_dpo/margin_std": 15.709431648254395, "step": 73 }, { "KL/chosen_KL_mean": -4.670244216918945, "KL/mean": -10.651287078857422, "KL/rejected_KL_mean": -16.63233184814453, "KL/std": 13.789833068847656, "epoch": 0.10866372980910426, "fcm_dpo/beta": 0.08051745593547821, "fcm_dpo/delta": -0.18177367746829987, "fcm_dpo/margin": 11.96208381652832, "fcm_dpo/q_t": 0.32404544949531555, "grad_norm": 55.107704162597656, "learning_rate": 4.999472998758977e-07, "logits/chosen": -0.631071925163269, "logits/rejected": -0.6169995069503784, "logps/chosen": -58.09512710571289, "logps/ref_chosen": -53.42488098144531, "logps/ref_rejected": -95.94693756103516, "logps/rejected": -112.57926940917969, "loss": 0.9081, "margin_dpo/margin_mean": 11.96208381652832, "margin_dpo/margin_std": 16.845170974731445, "step": 74 }, { "KL/chosen_KL_mean": -4.217048645019531, "KL/mean": -12.636398315429688, "KL/rejected_KL_mean": -21.055740356445312, "KL/std": 14.347799301147461, "epoch": 0.11013215859030837, "fcm_dpo/beta": 0.0748315155506134, "fcm_dpo/delta": -0.5235542058944702, "fcm_dpo/margin": 16.838699340820312, "fcm_dpo/q_t": 0.2691539525985718, "grad_norm": 42.937564849853516, "learning_rate": 4.999176576834721e-07, "logits/chosen": -0.6606135368347168, "logits/rejected": -0.6492525339126587, "logps/chosen": -56.078712463378906, "logps/ref_chosen": -51.861663818359375, "logps/ref_rejected": -111.25398254394531, "logps/rejected": -132.30972290039062, "loss": 0.709, "margin_dpo/margin_mean": 16.838699340820312, "margin_dpo/margin_std": 15.839242935180664, "step": 75 }, { "KL/chosen_KL_mean": -5.221561431884766, "KL/mean": -10.881424903869629, "KL/rejected_KL_mean": -16.541290283203125, "KL/std": 11.600919723510742, "epoch": 0.11160058737151249, "fcm_dpo/beta": 0.07199326157569885, "fcm_dpo/delta": -0.016267387196421623, "fcm_dpo/margin": 11.319733619689941, "fcm_dpo/q_t": 0.3303011655807495, "grad_norm": 45.18712615966797, "learning_rate": 4.998814299283415e-07, "logits/chosen": -0.6939696073532104, "logits/rejected": -0.6507217884063721, "logps/chosen": -58.48760223388672, "logps/ref_chosen": -53.26603698730469, "logps/ref_rejected": -78.21662902832031, "logps/rejected": -94.75791931152344, "loss": 0.8827, "margin_dpo/margin_mean": 11.319733619689941, "margin_dpo/margin_std": 12.374711990356445, "step": 76 }, { "KL/chosen_KL_mean": -4.481769561767578, "KL/mean": -13.34714126586914, "KL/rejected_KL_mean": -22.21251678466797, "KL/std": 15.9805908203125, "epoch": 0.1130690161527166, "fcm_dpo/beta": 0.06670539081096649, "fcm_dpo/delta": -0.4409904479980469, "fcm_dpo/margin": 17.730743408203125, "fcm_dpo/q_t": 0.27410849928855896, "grad_norm": 48.95978927612305, "learning_rate": 4.998386175651409e-07, "logits/chosen": -0.6710631847381592, "logits/rejected": -0.6282619833946228, "logps/chosen": -62.57844924926758, "logps/ref_chosen": -58.0966796875, "logps/ref_rejected": -93.77361297607422, "logps/rejected": -115.98612976074219, "loss": 0.749, "margin_dpo/margin_mean": 17.730743408203125, "margin_dpo/margin_std": 16.987913131713867, "step": 77 }, { "KL/chosen_KL_mean": -4.559408187866211, "KL/mean": -11.559930801391602, "KL/rejected_KL_mean": -18.56045150756836, "KL/std": 12.783736228942871, "epoch": 0.1145374449339207, "fcm_dpo/beta": 0.06502692401409149, "fcm_dpo/delta": -0.12146870791912079, "fcm_dpo/margin": 14.001038551330566, "fcm_dpo/q_t": 0.31776103377342224, "grad_norm": 43.656578063964844, "learning_rate": 4.997892217220159e-07, "logits/chosen": -0.6281476020812988, "logits/rejected": -0.598828911781311, "logps/chosen": -60.173194885253906, "logps/ref_chosen": -55.61378479003906, "logps/ref_rejected": -84.93436431884766, "logps/rejected": -103.49481201171875, "loss": 0.8351, "margin_dpo/margin_mean": 14.001038551330566, "margin_dpo/margin_std": 14.177305221557617, "step": 78 }, { "KL/chosen_KL_mean": -4.683469772338867, "KL/mean": -12.436279296875, "KL/rejected_KL_mean": -20.1890869140625, "KL/std": 14.876789093017578, "epoch": 0.11600587371512482, "fcm_dpo/beta": 0.06247137486934662, "fcm_dpo/delta": -0.18688446283340454, "fcm_dpo/margin": 15.505611419677734, "fcm_dpo/q_t": 0.3226276934146881, "grad_norm": 37.53826141357422, "learning_rate": 4.997332437005931e-07, "logits/chosen": -0.6610733270645142, "logits/rejected": -0.6301044225692749, "logps/chosen": -60.13395690917969, "logps/ref_chosen": -55.45048522949219, "logps/ref_rejected": -87.64756774902344, "logps/rejected": -107.83665466308594, "loss": 0.8644, "margin_dpo/margin_mean": 15.505611419677734, "margin_dpo/margin_std": 18.106834411621094, "step": 79 }, { "KL/chosen_KL_mean": -6.907768249511719, "KL/mean": -13.883707046508789, "KL/rejected_KL_mean": -20.85964584350586, "KL/std": 15.121198654174805, "epoch": 0.11747430249632893, "fcm_dpo/beta": 0.060649387538433075, "fcm_dpo/delta": -0.05435388535261154, "fcm_dpo/margin": 13.95186996459961, "fcm_dpo/q_t": 0.34249839186668396, "grad_norm": 40.209007263183594, "learning_rate": 4.996706849759452e-07, "logits/chosen": -0.7091320753097534, "logits/rejected": -0.6612462401390076, "logps/chosen": -65.42706298828125, "logps/ref_chosen": -58.519290924072266, "logps/ref_rejected": -87.54750061035156, "logps/rejected": -108.40715026855469, "loss": 0.9236, "margin_dpo/margin_mean": 13.951870918273926, "margin_dpo/margin_std": 17.254470825195312, "step": 80 }, { "KL/chosen_KL_mean": -5.891490936279297, "KL/mean": -15.477046966552734, "KL/rejected_KL_mean": -25.062599182128906, "KL/std": 18.107501983642578, "epoch": 0.11894273127753303, "fcm_dpo/beta": 0.05760374665260315, "fcm_dpo/delta": -0.3476817011833191, "fcm_dpo/margin": 19.171104431152344, "fcm_dpo/q_t": 0.2975410223007202, "grad_norm": 43.146236419677734, "learning_rate": 4.996015471965529e-07, "logits/chosen": -0.7024596333503723, "logits/rejected": -0.6684285402297974, "logps/chosen": -72.34036254882812, "logps/ref_chosen": -66.44886779785156, "logps/ref_rejected": -129.66270446777344, "logps/rejected": -154.72531127929688, "loss": 0.8002, "margin_dpo/margin_mean": 19.171104431152344, "margin_dpo/margin_std": 21.296463012695312, "step": 81 }, { "KL/chosen_KL_mean": -8.008272171020508, "KL/mean": -15.765302658081055, "KL/rejected_KL_mean": -23.522335052490234, "KL/std": 16.664772033691406, "epoch": 0.12041116005873716, "fcm_dpo/beta": 0.05717352032661438, "fcm_dpo/delta": -0.09888464212417603, "fcm_dpo/margin": 15.514065742492676, "fcm_dpo/q_t": 0.3363991379737854, "grad_norm": 52.070457458496094, "learning_rate": 4.995258321842611e-07, "logits/chosen": -0.637446403503418, "logits/rejected": -0.6206815838813782, "logps/chosen": -60.240657806396484, "logps/ref_chosen": -52.232383728027344, "logps/ref_rejected": -90.74325561523438, "logps/rejected": -114.26558685302734, "loss": 0.9861, "margin_dpo/margin_mean": 15.514065742492676, "margin_dpo/margin_std": 22.472978591918945, "step": 82 }, { "KL/chosen_KL_mean": -8.266363143920898, "KL/mean": -17.189800262451172, "KL/rejected_KL_mean": -26.113239288330078, "KL/std": 17.502941131591797, "epoch": 0.12187958883994127, "fcm_dpo/beta": 0.05427241325378418, "fcm_dpo/delta": -0.18765102326869965, "fcm_dpo/margin": 17.846874237060547, "fcm_dpo/q_t": 0.3166254460811615, "grad_norm": 39.94942092895508, "learning_rate": 4.994435419342304e-07, "logits/chosen": -0.6970843076705933, "logits/rejected": -0.6541832685470581, "logps/chosen": -64.09375, "logps/ref_chosen": -55.82738494873047, "logps/ref_rejected": -103.71589660644531, "logps/rejected": -129.82913208007812, "loss": 0.8555, "margin_dpo/margin_mean": 17.846874237060547, "margin_dpo/margin_std": 20.124860763549805, "step": 83 }, { "KL/chosen_KL_mean": -6.939172744750977, "KL/mean": -15.054475784301758, "KL/rejected_KL_mean": -23.16977310180664, "KL/std": 15.09716796875, "epoch": 0.12334801762114538, "fcm_dpo/beta": 0.05286204814910889, "fcm_dpo/delta": -0.06635652482509613, "fcm_dpo/margin": 16.23060417175293, "fcm_dpo/q_t": 0.3202790319919586, "grad_norm": 38.52730941772461, "learning_rate": 4.993546786148857e-07, "logits/chosen": -0.6437501311302185, "logits/rejected": -0.604649007320404, "logps/chosen": -74.11534118652344, "logps/ref_chosen": -67.1761703491211, "logps/ref_rejected": -87.29859924316406, "logps/rejected": -110.46837615966797, "loss": 0.8436, "margin_dpo/margin_mean": 16.23060417175293, "margin_dpo/margin_std": 15.421857833862305, "step": 84 }, { "KL/chosen_KL_mean": -7.799615859985352, "KL/mean": -15.955020904541016, "KL/rejected_KL_mean": -24.110427856445312, "KL/std": 15.60763931274414, "epoch": 0.12481644640234948, "fcm_dpo/beta": 0.053284820169210434, "fcm_dpo/delta": -0.07798822969198227, "fcm_dpo/margin": 16.310810089111328, "fcm_dpo/q_t": 0.33197450637817383, "grad_norm": 40.09149169921875, "learning_rate": 4.992592445678582e-07, "logits/chosen": -0.6409064531326294, "logits/rejected": -0.6097261905670166, "logps/chosen": -66.20623779296875, "logps/ref_chosen": -58.4066162109375, "logps/ref_rejected": -78.63880157470703, "logps/rejected": -102.74922943115234, "loss": 0.8927, "margin_dpo/margin_mean": 16.31081199645996, "margin_dpo/margin_std": 18.370267868041992, "step": 85 }, { "KL/chosen_KL_mean": -10.027334213256836, "KL/mean": -18.718555450439453, "KL/rejected_KL_mean": -27.409774780273438, "KL/std": 19.970577239990234, "epoch": 0.1262848751835536, "fcm_dpo/beta": 0.0518500953912735, "fcm_dpo/delta": -0.11137088388204575, "fcm_dpo/margin": 17.3824462890625, "fcm_dpo/q_t": 0.34738343954086304, "grad_norm": 48.6260871887207, "learning_rate": 4.991572423079235e-07, "logits/chosen": -0.7073214650154114, "logits/rejected": -0.6920993328094482, "logps/chosen": -66.164794921875, "logps/ref_chosen": -56.13746643066406, "logps/ref_rejected": -88.12165069580078, "logps/rejected": -115.53142547607422, "loss": 0.9935, "margin_dpo/margin_mean": 17.3824462890625, "margin_dpo/margin_std": 26.812076568603516, "step": 86 }, { "KL/chosen_KL_mean": -9.976259231567383, "KL/mean": -19.749847412109375, "KL/rejected_KL_mean": -29.5234375, "KL/std": 20.053821563720703, "epoch": 0.1277533039647577, "fcm_dpo/beta": 0.049444474279880524, "fcm_dpo/delta": -0.1872720569372177, "fcm_dpo/margin": 19.547176361083984, "fcm_dpo/q_t": 0.3172088861465454, "grad_norm": 38.684879302978516, "learning_rate": 4.990486745229364e-07, "logits/chosen": -0.7268928289413452, "logits/rejected": -0.6944937705993652, "logps/chosen": -65.61235046386719, "logps/ref_chosen": -55.63609313964844, "logps/ref_rejected": -95.46757507324219, "logps/rejected": -124.99101257324219, "loss": 0.8925, "margin_dpo/margin_mean": 19.547176361083984, "margin_dpo/margin_std": 23.78310775756836, "step": 87 }, { "KL/chosen_KL_mean": -12.57611083984375, "KL/mean": -20.00674057006836, "KL/rejected_KL_mean": -27.437366485595703, "KL/std": 20.11369514465332, "epoch": 0.12922173274596183, "fcm_dpo/beta": 0.049487993121147156, "fcm_dpo/delta": 0.06895022094249725, "fcm_dpo/margin": 14.86125373840332, "fcm_dpo/q_t": 0.35909318923950195, "grad_norm": 45.062110900878906, "learning_rate": 4.989335440737586e-07, "logits/chosen": -0.6849937438964844, "logits/rejected": -0.6724286079406738, "logps/chosen": -86.24726104736328, "logps/ref_chosen": -73.67115020751953, "logps/ref_rejected": -106.70849609375, "logps/rejected": -134.1458740234375, "loss": 1.0091, "margin_dpo/margin_mean": 14.861252784729004, "margin_dpo/margin_std": 20.61050796508789, "step": 88 }, { "KL/chosen_KL_mean": -7.28764533996582, "KL/mean": -15.397772789001465, "KL/rejected_KL_mean": -23.507896423339844, "KL/std": 17.05303382873535, "epoch": 0.13069016152716592, "fcm_dpo/beta": 0.050317272543907166, "fcm_dpo/delta": -0.018804259598255157, "fcm_dpo/margin": 16.22024917602539, "fcm_dpo/q_t": 0.33643651008605957, "grad_norm": 36.859100341796875, "learning_rate": 4.988118539941847e-07, "logits/chosen": -0.7560995817184448, "logits/rejected": -0.7220498323440552, "logps/chosen": -67.91255950927734, "logps/ref_chosen": -60.624916076660156, "logps/ref_rejected": -82.08354949951172, "logps/rejected": -105.59144592285156, "loss": 0.8918, "margin_dpo/margin_mean": 16.220251083374023, "margin_dpo/margin_std": 18.740238189697266, "step": 89 }, { "KL/chosen_KL_mean": -9.896280288696289, "KL/mean": -20.272663116455078, "KL/rejected_KL_mean": -30.649044036865234, "KL/std": 22.79068374633789, "epoch": 0.13215859030837004, "fcm_dpo/beta": 0.048810940235853195, "fcm_dpo/delta": -0.23755799233913422, "fcm_dpo/margin": 20.752761840820312, "fcm_dpo/q_t": 0.32720205187797546, "grad_norm": 40.190128326416016, "learning_rate": 4.986836074908615e-07, "logits/chosen": -0.6363822221755981, "logits/rejected": -0.6379286050796509, "logps/chosen": -63.18159103393555, "logps/ref_chosen": -53.285308837890625, "logps/ref_rejected": -111.54470825195312, "logps/rejected": -142.19375610351562, "loss": 0.9185, "margin_dpo/margin_mean": 20.752761840820312, "margin_dpo/margin_std": 27.751495361328125, "step": 90 }, { "KL/chosen_KL_mean": -9.575201034545898, "KL/mean": -19.17307472229004, "KL/rejected_KL_mean": -28.770950317382812, "KL/std": 19.327835083007812, "epoch": 0.13362701908957417, "fcm_dpo/beta": 0.047079045325517654, "fcm_dpo/delta": -0.11422688513994217, "fcm_dpo/margin": 19.195751190185547, "fcm_dpo/q_t": 0.3320731520652771, "grad_norm": 40.31279373168945, "learning_rate": 4.985488079432037e-07, "logits/chosen": -0.7156394720077515, "logits/rejected": -0.6806433796882629, "logps/chosen": -71.37815856933594, "logps/ref_chosen": -61.802955627441406, "logps/ref_rejected": -87.87395477294922, "logps/rejected": -116.64490509033203, "loss": 0.9117, "margin_dpo/margin_mean": 19.195751190185547, "margin_dpo/margin_std": 23.792537689208984, "step": 91 }, { "KL/chosen_KL_mean": -8.991003036499023, "KL/mean": -18.25010108947754, "KL/rejected_KL_mean": -27.509204864501953, "KL/std": 19.42957878112793, "epoch": 0.13509544787077826, "fcm_dpo/beta": 0.046396687626838684, "fcm_dpo/delta": -0.06554846465587616, "fcm_dpo/margin": 18.51820182800293, "fcm_dpo/q_t": 0.3381293714046478, "grad_norm": 36.30815887451172, "learning_rate": 4.984074589033043e-07, "logits/chosen": -0.7363908886909485, "logits/rejected": -0.7118117809295654, "logps/chosen": -60.631771087646484, "logps/ref_chosen": -51.640769958496094, "logps/ref_rejected": -77.88117980957031, "logps/rejected": -105.39038848876953, "loss": 0.9257, "margin_dpo/margin_mean": 18.51820182800293, "margin_dpo/margin_std": 23.363544464111328, "step": 92 }, { "KL/chosen_KL_mean": -9.855321884155273, "KL/mean": -19.377309799194336, "KL/rejected_KL_mean": -28.89929962158203, "KL/std": 18.807472229003906, "epoch": 0.13656387665198239, "fcm_dpo/beta": 0.0453658290207386, "fcm_dpo/delta": -0.07005032896995544, "fcm_dpo/margin": 19.043975830078125, "fcm_dpo/q_t": 0.3281649351119995, "grad_norm": 33.93098068237305, "learning_rate": 4.982595640958425e-07, "logits/chosen": -0.7472556233406067, "logits/rejected": -0.6917161345481873, "logps/chosen": -62.384559631347656, "logps/ref_chosen": -52.529239654541016, "logps/ref_rejected": -77.16075134277344, "logps/rejected": -106.06004333496094, "loss": 0.8607, "margin_dpo/margin_mean": 19.043975830078125, "margin_dpo/margin_std": 21.03558349609375, "step": 93 }, { "KL/chosen_KL_mean": -10.773126602172852, "KL/mean": -21.24502944946289, "KL/rejected_KL_mean": -31.716930389404297, "KL/std": 20.481773376464844, "epoch": 0.13803230543318648, "fcm_dpo/beta": 0.04396844655275345, "fcm_dpo/delta": -0.13568861782550812, "fcm_dpo/margin": 20.943805694580078, "fcm_dpo/q_t": 0.317252516746521, "grad_norm": 34.56655502319336, "learning_rate": 4.98105127417984e-07, "logits/chosen": -0.6687697768211365, "logits/rejected": -0.6449911594390869, "logps/chosen": -71.99574279785156, "logps/ref_chosen": -61.22261047363281, "logps/ref_rejected": -99.59902954101562, "logps/rejected": -131.3159637451172, "loss": 0.8318, "margin_dpo/margin_mean": 20.943805694580078, "margin_dpo/margin_std": 21.08070182800293, "step": 94 }, { "KL/chosen_KL_mean": -9.53125, "KL/mean": -18.02033233642578, "KL/rejected_KL_mean": -26.509416580200195, "KL/std": 18.12246322631836, "epoch": 0.1395007342143906, "fcm_dpo/beta": 0.044233426451683044, "fcm_dpo/delta": 0.05294659361243248, "fcm_dpo/margin": 16.978164672851562, "fcm_dpo/q_t": 0.3434492349624634, "grad_norm": 33.531463623046875, "learning_rate": 4.979441529392784e-07, "logits/chosen": -0.7102452516555786, "logits/rejected": -0.6761167049407959, "logps/chosen": -62.054893493652344, "logps/ref_chosen": -52.523643493652344, "logps/ref_rejected": -75.8803482055664, "logps/rejected": -102.3897705078125, "loss": 0.914, "margin_dpo/margin_mean": 16.978164672851562, "margin_dpo/margin_std": 18.86602783203125, "step": 95 }, { "KL/chosen_KL_mean": -8.65886116027832, "KL/mean": -20.042604446411133, "KL/rejected_KL_mean": -31.426353454589844, "KL/std": 22.305217742919922, "epoch": 0.14096916299559473, "fcm_dpo/beta": 0.042825281620025635, "fcm_dpo/delta": -0.19741219282150269, "fcm_dpo/margin": 22.767492294311523, "fcm_dpo/q_t": 0.3105483055114746, "grad_norm": 33.64408493041992, "learning_rate": 4.977766449015534e-07, "logits/chosen": -0.7110648155212402, "logits/rejected": -0.6743212938308716, "logps/chosen": -70.81583404541016, "logps/ref_chosen": -62.15697479248047, "logps/ref_rejected": -96.59601593017578, "logps/rejected": -128.02236938476562, "loss": 0.814, "margin_dpo/margin_mean": 22.767492294311523, "margin_dpo/margin_std": 24.09496307373047, "step": 96 }, { "KL/chosen_KL_mean": -9.822381973266602, "KL/mean": -18.966472625732422, "KL/rejected_KL_mean": -28.110565185546875, "KL/std": 18.120513916015625, "epoch": 0.14243759177679882, "fcm_dpo/beta": 0.04335338622331619, "fcm_dpo/delta": 0.00421653687953949, "fcm_dpo/margin": 18.288185119628906, "fcm_dpo/q_t": 0.3337208330631256, "grad_norm": 36.38238525390625, "learning_rate": 4.976026077188012e-07, "logits/chosen": -0.6568082571029663, "logits/rejected": -0.6052370667457581, "logps/chosen": -64.46875, "logps/ref_chosen": -54.646366119384766, "logps/ref_rejected": -76.96475219726562, "logps/rejected": -105.0753173828125, "loss": 0.88, "margin_dpo/margin_mean": 18.288185119628906, "margin_dpo/margin_std": 17.953163146972656, "step": 97 }, { "KL/chosen_KL_mean": -11.879871368408203, "KL/mean": -22.035350799560547, "KL/rejected_KL_mean": -32.19082260131836, "KL/std": 20.222049713134766, "epoch": 0.14390602055800295, "fcm_dpo/beta": 0.04232582449913025, "fcm_dpo/delta": -0.06543941795825958, "fcm_dpo/margin": 20.310956954956055, "fcm_dpo/q_t": 0.3265250325202942, "grad_norm": 37.31782150268555, "learning_rate": 4.974220459770639e-07, "logits/chosen": -0.7118183374404907, "logits/rejected": -0.691297709941864, "logps/chosen": -77.13850402832031, "logps/ref_chosen": -65.25862884521484, "logps/ref_rejected": -96.5274887084961, "logps/rejected": -128.7183074951172, "loss": 0.903, "margin_dpo/margin_mean": 20.310956954956055, "margin_dpo/margin_std": 23.548891067504883, "step": 98 }, { "KL/chosen_KL_mean": -9.833019256591797, "KL/mean": -21.514665603637695, "KL/rejected_KL_mean": -33.19630432128906, "KL/std": 21.859981536865234, "epoch": 0.14537444933920704, "fcm_dpo/beta": 0.04106743261218071, "fcm_dpo/delta": -0.17689181864261627, "fcm_dpo/margin": 23.363290786743164, "fcm_dpo/q_t": 0.3184036612510681, "grad_norm": 31.978910446166992, "learning_rate": 4.972349644343108e-07, "logits/chosen": -0.6774739027023315, "logits/rejected": -0.6715878248214722, "logps/chosen": -55.47150421142578, "logps/ref_chosen": -45.638484954833984, "logps/ref_rejected": -86.43793487548828, "logps/rejected": -119.63423919677734, "loss": 0.8333, "margin_dpo/margin_mean": 23.363290786743164, "margin_dpo/margin_std": 26.378826141357422, "step": 99 }, { "KL/chosen_KL_mean": -10.707719802856445, "KL/mean": -18.45950698852539, "KL/rejected_KL_mean": -26.211288452148438, "KL/std": 17.909923553466797, "epoch": 0.14684287812041116, "fcm_dpo/beta": 0.041958265006542206, "fcm_dpo/delta": 0.15857800841331482, "fcm_dpo/margin": 15.503573417663574, "fcm_dpo/q_t": 0.36995577812194824, "grad_norm": 38.36394119262695, "learning_rate": 4.970413680203148e-07, "logits/chosen": -0.6655231714248657, "logits/rejected": -0.6192318201065063, "logps/chosen": -68.30169677734375, "logps/ref_chosen": -57.59397888183594, "logps/ref_rejected": -74.06021118164062, "logps/rejected": -100.27149963378906, "loss": 1.031, "margin_dpo/margin_mean": 15.503572463989258, "margin_dpo/margin_std": 22.743520736694336, "step": 100 }, { "KL/chosen_KL_mean": -12.964332580566406, "KL/mean": -21.973020553588867, "KL/rejected_KL_mean": -30.981712341308594, "KL/std": 20.87362289428711, "epoch": 0.14831130690161526, "fcm_dpo/beta": 0.04202631860971451, "fcm_dpo/delta": 0.045243848115205765, "fcm_dpo/margin": 18.01737403869629, "fcm_dpo/q_t": 0.35849660634994507, "grad_norm": 39.82667541503906, "learning_rate": 4.968412618365215e-07, "logits/chosen": -0.6872289180755615, "logits/rejected": -0.6496012210845947, "logps/chosen": -74.61318969726562, "logps/ref_chosen": -61.64885330200195, "logps/ref_rejected": -83.18968200683594, "logps/rejected": -114.17140197753906, "loss": 0.9806, "margin_dpo/margin_mean": 18.01737403869629, "margin_dpo/margin_std": 25.271141052246094, "step": 101 }, { "KL/chosen_KL_mean": -14.69502067565918, "KL/mean": -21.546857833862305, "KL/rejected_KL_mean": -28.398696899414062, "KL/std": 20.012174606323242, "epoch": 0.14977973568281938, "fcm_dpo/beta": 0.04423201084136963, "fcm_dpo/delta": 0.2019134908914566, "fcm_dpo/margin": 13.70367431640625, "fcm_dpo/q_t": 0.3829655945301056, "grad_norm": 49.598167419433594, "learning_rate": 4.966346511559149e-07, "logits/chosen": -0.7153527736663818, "logits/rejected": -0.6670923233032227, "logps/chosen": -78.77389526367188, "logps/ref_chosen": -64.0788803100586, "logps/ref_rejected": -68.18707275390625, "logps/rejected": -96.58576965332031, "loss": 1.0845, "margin_dpo/margin_mean": 13.703676223754883, "margin_dpo/margin_std": 22.846996307373047, "step": 102 }, { "KL/chosen_KL_mean": -10.015007019042969, "KL/mean": -22.716508865356445, "KL/rejected_KL_mean": -35.41801071166992, "KL/std": 21.94977569580078, "epoch": 0.1512481644640235, "fcm_dpo/beta": 0.042147666215896606, "fcm_dpo/delta": -0.3043977618217468, "fcm_dpo/margin": 25.402999877929688, "fcm_dpo/q_t": 0.29378920793533325, "grad_norm": 33.92921829223633, "learning_rate": 4.964215414228785e-07, "logits/chosen": -0.7019765377044678, "logits/rejected": -0.6657599210739136, "logps/chosen": -71.31428527832031, "logps/ref_chosen": -61.299278259277344, "logps/ref_rejected": -93.57270812988281, "logps/rejected": -128.99072265625, "loss": 0.7667, "margin_dpo/margin_mean": 25.402996063232422, "margin_dpo/margin_std": 24.849088668823242, "step": 103 }, { "KL/chosen_KL_mean": -11.889623641967773, "KL/mean": -23.839998245239258, "KL/rejected_KL_mean": -35.790382385253906, "KL/std": 24.284305572509766, "epoch": 0.1527165932452276, "fcm_dpo/beta": 0.040641963481903076, "fcm_dpo/delta": -0.18938633799552917, "fcm_dpo/margin": 23.9007568359375, "fcm_dpo/q_t": 0.3229272961616516, "grad_norm": 34.873538970947266, "learning_rate": 4.96201938253052e-07, "logits/chosen": -0.7309329509735107, "logits/rejected": -0.6916954517364502, "logps/chosen": -66.26239776611328, "logps/ref_chosen": -54.372772216796875, "logps/ref_rejected": -89.5647201538086, "logps/rejected": -125.3551025390625, "loss": 0.8711, "margin_dpo/margin_mean": 23.9007568359375, "margin_dpo/margin_std": 28.508708953857422, "step": 104 }, { "KL/chosen_KL_mean": -10.295255661010742, "KL/mean": -26.896554946899414, "KL/rejected_KL_mean": -43.49786376953125, "KL/std": 24.715675354003906, "epoch": 0.15418502202643172, "fcm_dpo/beta": 0.0372607484459877, "fcm_dpo/delta": -0.4987417459487915, "fcm_dpo/margin": 33.20261001586914, "fcm_dpo/q_t": 0.2553773522377014, "grad_norm": 30.176359176635742, "learning_rate": 4.959758474331832e-07, "logits/chosen": -0.7114518880844116, "logits/rejected": -0.6793392896652222, "logps/chosen": -64.9342041015625, "logps/ref_chosen": -54.638946533203125, "logps/ref_rejected": -97.97351837158203, "logps/rejected": -141.47137451171875, "loss": 0.6601, "margin_dpo/margin_mean": 33.20261001586914, "margin_dpo/margin_std": 26.298891067504883, "step": 105 }, { "KL/chosen_KL_mean": -12.043193817138672, "KL/mean": -22.559654235839844, "KL/rejected_KL_mean": -33.07612228393555, "KL/std": 20.727237701416016, "epoch": 0.15565345080763582, "fcm_dpo/beta": 0.0364394336938858, "fcm_dpo/delta": 0.03637177497148514, "fcm_dpo/margin": 21.03292465209961, "fcm_dpo/q_t": 0.3410166800022125, "grad_norm": 31.546979904174805, "learning_rate": 4.957432749209755e-07, "logits/chosen": -0.6615180969238281, "logits/rejected": -0.6151139140129089, "logps/chosen": -66.87609100341797, "logps/ref_chosen": -54.83289337158203, "logps/ref_rejected": -85.22461700439453, "logps/rejected": -118.30073547363281, "loss": 0.8977, "margin_dpo/margin_mean": 21.03292465209961, "margin_dpo/margin_std": 22.555233001708984, "step": 106 }, { "KL/chosen_KL_mean": -14.550010681152344, "KL/mean": -26.484832763671875, "KL/rejected_KL_mean": -38.419654846191406, "KL/std": 24.61465072631836, "epoch": 0.15712187958883994, "fcm_dpo/beta": 0.0360017865896225, "fcm_dpo/delta": -0.06571710109710693, "fcm_dpo/margin": 23.869640350341797, "fcm_dpo/q_t": 0.3263270854949951, "grad_norm": 30.792583465576172, "learning_rate": 4.955042268449307e-07, "logits/chosen": -0.7245312929153442, "logits/rejected": -0.6730589866638184, "logps/chosen": -84.25782775878906, "logps/ref_chosen": -69.70780944824219, "logps/ref_rejected": -94.73950958251953, "logps/rejected": -133.15916442871094, "loss": 0.8737, "margin_dpo/margin_mean": 23.869640350341797, "margin_dpo/margin_std": 25.510066986083984, "step": 107 }, { "KL/chosen_KL_mean": -13.597505569458008, "KL/mean": -26.603561401367188, "KL/rejected_KL_mean": -39.60961151123047, "KL/std": 26.066404342651367, "epoch": 0.15859030837004406, "fcm_dpo/beta": 0.03532955050468445, "fcm_dpo/delta": -0.1311628818511963, "fcm_dpo/margin": 26.012113571166992, "fcm_dpo/q_t": 0.33340299129486084, "grad_norm": 34.07270050048828, "learning_rate": 4.952587095041881e-07, "logits/chosen": -0.6948635578155518, "logits/rejected": -0.6569874882698059, "logps/chosen": -69.60739135742188, "logps/ref_chosen": -56.0098876953125, "logps/ref_rejected": -95.79601287841797, "logps/rejected": -135.40562438964844, "loss": 0.9154, "margin_dpo/margin_mean": 26.012113571166992, "margin_dpo/margin_std": 33.049766540527344, "step": 108 }, { "KL/chosen_KL_mean": -12.223468780517578, "KL/mean": -26.6085205078125, "KL/rejected_KL_mean": -40.99357604980469, "KL/std": 25.966686248779297, "epoch": 0.16005873715124816, "fcm_dpo/beta": 0.03398079425096512, "fcm_dpo/delta": -0.19790343940258026, "fcm_dpo/margin": 28.770111083984375, "fcm_dpo/q_t": 0.3054792881011963, "grad_norm": 30.806764602661133, "learning_rate": 4.95006729368358e-07, "logits/chosen": -0.6443037986755371, "logits/rejected": -0.618833065032959, "logps/chosen": -75.10895538330078, "logps/ref_chosen": -62.88549041748047, "logps/ref_rejected": -98.68573760986328, "logps/rejected": -139.6793212890625, "loss": 0.8036, "margin_dpo/margin_mean": 28.770111083984375, "margin_dpo/margin_std": 27.74138641357422, "step": 109 }, { "KL/chosen_KL_mean": -11.736026763916016, "KL/mean": -24.941783905029297, "KL/rejected_KL_mean": -38.14754104614258, "KL/std": 25.426937103271484, "epoch": 0.16152716593245228, "fcm_dpo/beta": 0.0328492596745491, "fcm_dpo/delta": -0.08019863814115524, "fcm_dpo/margin": 26.411510467529297, "fcm_dpo/q_t": 0.32906657457351685, "grad_norm": 29.275253295898438, "learning_rate": 4.947482930773511e-07, "logits/chosen": -0.6451650857925415, "logits/rejected": -0.5977135896682739, "logps/chosen": -70.48971557617188, "logps/ref_chosen": -58.753684997558594, "logps/ref_rejected": -79.75001525878906, "logps/rejected": -117.89755249023438, "loss": 0.893, "margin_dpo/margin_mean": 26.411510467529297, "margin_dpo/margin_std": 29.890933990478516, "step": 110 }, { "KL/chosen_KL_mean": -14.619787216186523, "KL/mean": -28.84766387939453, "KL/rejected_KL_mean": -43.075538635253906, "KL/std": 28.174575805664062, "epoch": 0.16299559471365638, "fcm_dpo/beta": 0.032323651015758514, "fcm_dpo/delta": -0.13409429788589478, "fcm_dpo/margin": 28.45575523376465, "fcm_dpo/q_t": 0.32236313819885254, "grad_norm": 30.668977737426758, "learning_rate": 4.944834074412042e-07, "logits/chosen": -0.7059623003005981, "logits/rejected": -0.6807644367218018, "logps/chosen": -83.243896484375, "logps/ref_chosen": -68.62410736083984, "logps/ref_rejected": -98.42886352539062, "logps/rejected": -141.50439453125, "loss": 0.8786, "margin_dpo/margin_mean": 28.45575523376465, "margin_dpo/margin_std": 32.44450378417969, "step": 111 }, { "KL/chosen_KL_mean": -13.895004272460938, "KL/mean": -23.563627243041992, "KL/rejected_KL_mean": -33.23225402832031, "KL/std": 21.184219360351562, "epoch": 0.1644640234948605, "fcm_dpo/beta": 0.032971762120723724, "fcm_dpo/delta": 0.1737350970506668, "fcm_dpo/margin": 19.33725357055664, "fcm_dpo/q_t": 0.3689352869987488, "grad_norm": 31.53493309020996, "learning_rate": 4.942120794399002e-07, "logits/chosen": -0.6969665884971619, "logits/rejected": -0.6505686640739441, "logps/chosen": -64.14464569091797, "logps/ref_chosen": -50.24964141845703, "logps/ref_rejected": -64.77442932128906, "logps/rejected": -98.00668334960938, "loss": 0.9915, "margin_dpo/margin_mean": 19.33725357055664, "margin_dpo/margin_std": 24.64777183532715, "step": 112 }, { "KL/chosen_KL_mean": -15.778499603271484, "KL/mean": -26.65062713623047, "KL/rejected_KL_mean": -37.522762298583984, "KL/std": 20.39154815673828, "epoch": 0.16593245227606462, "fcm_dpo/beta": 0.03382117301225662, "fcm_dpo/delta": 0.06972439587116241, "fcm_dpo/margin": 21.744258880615234, "fcm_dpo/q_t": 0.34526845812797546, "grad_norm": 30.338754653930664, "learning_rate": 4.939343162231841e-07, "logits/chosen": -0.6443773508071899, "logits/rejected": -0.5929930210113525, "logps/chosen": -82.491455078125, "logps/ref_chosen": -66.71295166015625, "logps/ref_rejected": -77.96870422363281, "logps/rejected": -115.49147033691406, "loss": 0.9041, "margin_dpo/margin_mean": 21.744258880615234, "margin_dpo/margin_std": 22.552532196044922, "step": 113 }, { "KL/chosen_KL_mean": -14.374372482299805, "KL/mean": -30.3890438079834, "KL/rejected_KL_mean": -46.403717041015625, "KL/std": 31.068870544433594, "epoch": 0.16740088105726872, "fcm_dpo/beta": 0.03228652477264404, "fcm_dpo/delta": -0.2666097581386566, "fcm_dpo/margin": 32.02934265136719, "fcm_dpo/q_t": 0.30826088786125183, "grad_norm": 28.729209899902344, "learning_rate": 4.936501251103751e-07, "logits/chosen": -0.6775549650192261, "logits/rejected": -0.6369887590408325, "logps/chosen": -72.15945434570312, "logps/ref_chosen": -57.78507995605469, "logps/ref_rejected": -87.10966491699219, "logps/rejected": -133.5133819580078, "loss": 0.8171, "margin_dpo/margin_mean": 32.02934265136719, "margin_dpo/margin_std": 36.730525970458984, "step": 114 }, { "KL/chosen_KL_mean": -19.756134033203125, "KL/mean": -31.28628921508789, "KL/rejected_KL_mean": -42.816444396972656, "KL/std": 28.551877975463867, "epoch": 0.16886930983847284, "fcm_dpo/beta": 0.03259563446044922, "fcm_dpo/delta": 0.0514962375164032, "fcm_dpo/margin": 23.0603084564209, "fcm_dpo/q_t": 0.3605424761772156, "grad_norm": 44.56121826171875, "learning_rate": 4.933595135901732e-07, "logits/chosen": -0.7127076387405396, "logits/rejected": -0.6788430213928223, "logps/chosen": -85.33877563476562, "logps/ref_chosen": -65.5826416015625, "logps/ref_rejected": -98.56552124023438, "logps/rejected": -141.3819580078125, "loss": 1.0188, "margin_dpo/margin_mean": 23.0603084564209, "margin_dpo/margin_std": 35.82677459716797, "step": 115 }, { "KL/chosen_KL_mean": -14.837322235107422, "KL/mean": -26.82990837097168, "KL/rejected_KL_mean": -38.82250213623047, "KL/std": 24.2340145111084, "epoch": 0.17033773861967694, "fcm_dpo/beta": 0.03275762125849724, "fcm_dpo/delta": 0.014920920133590698, "fcm_dpo/margin": 23.98516845703125, "fcm_dpo/q_t": 0.33930882811546326, "grad_norm": 31.535541534423828, "learning_rate": 4.930624893204624e-07, "logits/chosen": -0.6988054513931274, "logits/rejected": -0.6746259331703186, "logps/chosen": -66.23763275146484, "logps/ref_chosen": -51.40031433105469, "logps/ref_rejected": -80.5218505859375, "logps/rejected": -119.34435272216797, "loss": 0.8796, "margin_dpo/margin_mean": 23.985170364379883, "margin_dpo/margin_std": 25.571834564208984, "step": 116 }, { "KL/chosen_KL_mean": -17.652626037597656, "KL/mean": -28.647598266601562, "KL/rejected_KL_mean": -39.642574310302734, "KL/std": 25.565113067626953, "epoch": 0.17180616740088106, "fcm_dpo/beta": 0.03300020843744278, "fcm_dpo/delta": 0.080161452293396, "fcm_dpo/margin": 21.989948272705078, "fcm_dpo/q_t": 0.3538960814476013, "grad_norm": 39.028663635253906, "learning_rate": 4.927590601281083e-07, "logits/chosen": -0.6682844161987305, "logits/rejected": -0.628246545791626, "logps/chosen": -86.95103454589844, "logps/ref_chosen": -69.29840850830078, "logps/ref_rejected": -66.583984375, "logps/rejected": -106.2265625, "loss": 0.9556, "margin_dpo/margin_mean": 21.989948272705078, "margin_dpo/margin_std": 28.185462951660156, "step": 117 }, { "KL/chosen_KL_mean": -15.85186767578125, "KL/mean": -27.46802520751953, "KL/rejected_KL_mean": -39.08417892456055, "KL/std": 23.371967315673828, "epoch": 0.17327459618208516, "fcm_dpo/beta": 0.03330120071768761, "fcm_dpo/delta": 0.028526771813631058, "fcm_dpo/margin": 23.232316970825195, "fcm_dpo/q_t": 0.3429677188396454, "grad_norm": 48.863182067871094, "learning_rate": 4.924492340087524e-07, "logits/chosen": -0.6977423429489136, "logits/rejected": -0.6685633659362793, "logps/chosen": -71.49284362792969, "logps/ref_chosen": -55.6409797668457, "logps/ref_rejected": -75.66905975341797, "logps/rejected": -114.75323486328125, "loss": 0.8972, "margin_dpo/margin_mean": 23.232318878173828, "margin_dpo/margin_std": 25.548391342163086, "step": 118 }, { "KL/chosen_KL_mean": -19.729721069335938, "KL/mean": -31.524303436279297, "KL/rejected_KL_mean": -43.31888961791992, "KL/std": 25.345006942749023, "epoch": 0.17474302496328928, "fcm_dpo/beta": 0.03315918147563934, "fcm_dpo/delta": 0.018118254840373993, "fcm_dpo/margin": 23.58917236328125, "fcm_dpo/q_t": 0.3480648994445801, "grad_norm": 34.203346252441406, "learning_rate": 4.92133019126601e-07, "logits/chosen": -0.6777770519256592, "logits/rejected": -0.6563413739204407, "logps/chosen": -93.23991394042969, "logps/ref_chosen": -73.51019287109375, "logps/ref_rejected": -102.977294921875, "logps/rejected": -146.29617309570312, "loss": 0.9387, "margin_dpo/margin_mean": 23.58917236328125, "margin_dpo/margin_std": 29.872432708740234, "step": 119 }, { "KL/chosen_KL_mean": -21.17425537109375, "KL/mean": -36.770111083984375, "KL/rejected_KL_mean": -52.365962982177734, "KL/std": 31.41073226928711, "epoch": 0.1762114537444934, "fcm_dpo/beta": 0.03230596333742142, "fcm_dpo/delta": -0.23119711875915527, "fcm_dpo/margin": 31.19170379638672, "fcm_dpo/q_t": 0.30633899569511414, "grad_norm": 32.835208892822266, "learning_rate": 4.918104238142103e-07, "logits/chosen": -0.6844779253005981, "logits/rejected": -0.6430919766426086, "logps/chosen": -97.95509338378906, "logps/ref_chosen": -76.78083801269531, "logps/ref_rejected": -108.02374267578125, "logps/rejected": -160.38970947265625, "loss": 0.8181, "margin_dpo/margin_mean": 31.19170379638672, "margin_dpo/margin_std": 32.8134765625, "step": 120 }, { "KL/chosen_KL_mean": -20.149927139282227, "KL/mean": -37.27117156982422, "KL/rejected_KL_mean": -54.39240646362305, "KL/std": 31.102367401123047, "epoch": 0.1776798825256975, "fcm_dpo/beta": 0.03047577664256096, "fcm_dpo/delta": -0.27470406889915466, "fcm_dpo/margin": 34.24247360229492, "fcm_dpo/q_t": 0.30704832077026367, "grad_norm": 32.69995880126953, "learning_rate": 4.91481456572267e-07, "logits/chosen": -0.6667851209640503, "logits/rejected": -0.6555989384651184, "logps/chosen": -81.9398193359375, "logps/ref_chosen": -61.789894104003906, "logps/ref_rejected": -109.99456787109375, "logps/rejected": -164.38697814941406, "loss": 0.8259, "margin_dpo/margin_mean": 34.24247360229492, "margin_dpo/margin_std": 37.63690948486328, "step": 121 }, { "KL/chosen_KL_mean": -18.689184188842773, "KL/mean": -39.89842224121094, "KL/rejected_KL_mean": -61.107669830322266, "KL/std": 36.409278869628906, "epoch": 0.17914831130690162, "fcm_dpo/beta": 0.02825929969549179, "fcm_dpo/delta": -0.4536348581314087, "fcm_dpo/margin": 42.418479919433594, "fcm_dpo/q_t": 0.26600882411003113, "grad_norm": 31.267995834350586, "learning_rate": 4.911461260693638e-07, "logits/chosen": -0.6460617780685425, "logits/rejected": -0.6521382927894592, "logps/chosen": -65.59139251708984, "logps/ref_chosen": -46.9022102355957, "logps/ref_rejected": -106.71418762207031, "logps/rejected": -167.82186889648438, "loss": 0.6902, "margin_dpo/margin_mean": 42.41847610473633, "margin_dpo/margin_std": 35.58713150024414, "step": 122 }, { "KL/chosen_KL_mean": -19.802492141723633, "KL/mean": -33.98395919799805, "KL/rejected_KL_mean": -48.16543197631836, "KL/std": 30.98874282836914, "epoch": 0.18061674008810572, "fcm_dpo/beta": 0.027563009411096573, "fcm_dpo/delta": 0.01940111070871353, "fcm_dpo/margin": 28.362932205200195, "fcm_dpo/q_t": 0.35159093141555786, "grad_norm": 32.10316467285156, "learning_rate": 4.908044411417711e-07, "logits/chosen": -0.6464723348617554, "logits/rejected": -0.6184071898460388, "logps/chosen": -81.14112854003906, "logps/ref_chosen": -61.33863830566406, "logps/ref_rejected": -87.775390625, "logps/rejected": -135.94082641601562, "loss": 0.984, "margin_dpo/margin_mean": 28.362932205200195, "margin_dpo/margin_std": 40.174957275390625, "step": 123 }, { "KL/chosen_KL_mean": -21.405563354492188, "KL/mean": -40.972625732421875, "KL/rejected_KL_mean": -60.53968811035156, "KL/std": 40.26324462890625, "epoch": 0.18208516886930984, "fcm_dpo/beta": 0.0265456922352314, "fcm_dpo/delta": -0.26791417598724365, "fcm_dpo/margin": 39.134124755859375, "fcm_dpo/q_t": 0.3195335566997528, "grad_norm": 32.24935531616211, "learning_rate": 4.904564107932048e-07, "logits/chosen": -0.664286732673645, "logits/rejected": -0.6608208417892456, "logps/chosen": -92.85389709472656, "logps/ref_chosen": -71.44833374023438, "logps/ref_rejected": -117.58056640625, "logps/rejected": -178.12025451660156, "loss": 0.8779, "margin_dpo/margin_mean": 39.134124755859375, "margin_dpo/margin_std": 50.061702728271484, "step": 124 }, { "KL/chosen_KL_mean": -17.320648193359375, "KL/mean": -35.29172134399414, "KL/rejected_KL_mean": -53.2628059387207, "KL/std": 32.852684020996094, "epoch": 0.18355359765051396, "fcm_dpo/beta": 0.02581089735031128, "fcm_dpo/delta": -0.14052706956863403, "fcm_dpo/margin": 35.94215393066406, "fcm_dpo/q_t": 0.31883522868156433, "grad_norm": 27.765161514282227, "learning_rate": 4.90102044194588e-07, "logits/chosen": -0.6267136335372925, "logits/rejected": -0.615674614906311, "logps/chosen": -67.45758819580078, "logps/ref_chosen": -50.136940002441406, "logps/ref_rejected": -83.98861694335938, "logps/rejected": -137.2514190673828, "loss": 0.8366, "margin_dpo/margin_mean": 35.94215393066406, "margin_dpo/margin_std": 37.57044982910156, "step": 125 }, { "KL/chosen_KL_mean": -19.33326530456543, "KL/mean": -36.09067153930664, "KL/rejected_KL_mean": -52.84807586669922, "KL/std": 30.532608032226562, "epoch": 0.18502202643171806, "fcm_dpo/beta": 0.02531757578253746, "fcm_dpo/delta": -0.05310361459851265, "fcm_dpo/margin": 33.51481628417969, "fcm_dpo/q_t": 0.330096036195755, "grad_norm": 30.099266052246094, "learning_rate": 4.897413506838102e-07, "logits/chosen": -0.6573269367218018, "logits/rejected": -0.6356841921806335, "logps/chosen": -75.00033569335938, "logps/ref_chosen": -55.66706848144531, "logps/ref_rejected": -98.1297607421875, "logps/rejected": -150.97784423828125, "loss": 0.8715, "margin_dpo/margin_mean": 33.514808654785156, "margin_dpo/margin_std": 36.43012237548828, "step": 126 }, { "KL/chosen_KL_mean": -17.67395782470703, "KL/mean": -31.366748809814453, "KL/rejected_KL_mean": -45.05954360961914, "KL/std": 26.2576904296875, "epoch": 0.18649045521292218, "fcm_dpo/beta": 0.025727272033691406, "fcm_dpo/delta": 0.1022951602935791, "fcm_dpo/margin": 27.38558006286621, "fcm_dpo/q_t": 0.35243093967437744, "grad_norm": 29.7404727935791, "learning_rate": 4.89374339765481e-07, "logits/chosen": -0.6145305633544922, "logits/rejected": -0.5852265357971191, "logps/chosen": -74.2286376953125, "logps/ref_chosen": -56.55467987060547, "logps/ref_rejected": -76.7957763671875, "logps/rejected": -121.8553237915039, "loss": 0.9369, "margin_dpo/margin_mean": 27.38558006286621, "margin_dpo/margin_std": 30.984207153320312, "step": 127 }, { "KL/chosen_KL_mean": -19.56759262084961, "KL/mean": -33.658756256103516, "KL/rejected_KL_mean": -47.749916076660156, "KL/std": 32.29600524902344, "epoch": 0.18795888399412627, "fcm_dpo/beta": 0.026202715933322906, "fcm_dpo/delta": 0.06536929309368134, "fcm_dpo/margin": 28.182331085205078, "fcm_dpo/q_t": 0.355803906917572, "grad_norm": 37.05172348022461, "learning_rate": 4.890010211106795e-07, "logits/chosen": -0.6373168230056763, "logits/rejected": -0.598369836807251, "logps/chosen": -77.68855285644531, "logps/ref_chosen": -58.12095642089844, "logps/ref_rejected": -76.43896484375, "logps/rejected": -124.18888092041016, "loss": 0.9771, "margin_dpo/margin_mean": 28.182331085205078, "margin_dpo/margin_std": 38.03752136230469, "step": 128 }, { "KL/chosen_KL_mean": -25.679073333740234, "KL/mean": -40.25211715698242, "KL/rejected_KL_mean": -54.825164794921875, "KL/std": 33.20759201049805, "epoch": 0.1894273127753304, "fcm_dpo/beta": 0.02634388394653797, "fcm_dpo/delta": 0.03461331129074097, "fcm_dpo/margin": 29.14608383178711, "fcm_dpo/q_t": 0.3556867241859436, "grad_norm": 34.6231575012207, "learning_rate": 4.88621404556699e-07, "logits/chosen": -0.6257126331329346, "logits/rejected": -0.6025162935256958, "logps/chosen": -92.59544372558594, "logps/ref_chosen": -66.91637420654297, "logps/ref_rejected": -96.6422119140625, "logps/rejected": -151.46737670898438, "loss": 0.9911, "margin_dpo/margin_mean": 29.14608383178711, "margin_dpo/margin_std": 41.93766784667969, "step": 129 }, { "KL/chosen_KL_mean": -19.17430305480957, "KL/mean": -40.231685638427734, "KL/rejected_KL_mean": -61.28907012939453, "KL/std": 37.78228759765625, "epoch": 0.19089574155653452, "fcm_dpo/beta": 0.02557522803544998, "fcm_dpo/delta": -0.3111729323863983, "fcm_dpo/margin": 42.11476135253906, "fcm_dpo/q_t": 0.2972610890865326, "grad_norm": 27.576147079467773, "learning_rate": 4.882355001067891e-07, "logits/chosen": -0.6214011907577515, "logits/rejected": -0.6146156191825867, "logps/chosen": -63.841156005859375, "logps/ref_chosen": -44.66685104370117, "logps/ref_rejected": -82.78165435791016, "logps/rejected": -144.0707244873047, "loss": 0.8184, "margin_dpo/margin_mean": 42.11476135253906, "margin_dpo/margin_std": 42.95219802856445, "step": 130 }, { "KL/chosen_KL_mean": -15.329935073852539, "KL/mean": -35.1269416809082, "KL/rejected_KL_mean": -54.923946380615234, "KL/std": 34.3599967956543, "epoch": 0.19236417033773862, "fcm_dpo/beta": 0.02407342568039894, "fcm_dpo/delta": -0.169725239276886, "fcm_dpo/margin": 39.59401321411133, "fcm_dpo/q_t": 0.3058611750602722, "grad_norm": 28.302913665771484, "learning_rate": 4.878433179298909e-07, "logits/chosen": -0.6235455274581909, "logits/rejected": -0.6231019496917725, "logps/chosen": -60.25452423095703, "logps/ref_chosen": -44.924591064453125, "logps/ref_rejected": -88.44401550292969, "logps/rejected": -143.36795043945312, "loss": 0.8077, "margin_dpo/margin_mean": 39.59400939941406, "margin_dpo/margin_std": 37.83281326293945, "step": 131 }, { "KL/chosen_KL_mean": -21.70701026916504, "KL/mean": -39.37970733642578, "KL/rejected_KL_mean": -57.05239486694336, "KL/std": 35.41930389404297, "epoch": 0.19383259911894274, "fcm_dpo/beta": 0.023686932399868965, "fcm_dpo/delta": -0.04097011685371399, "fcm_dpo/margin": 35.34538269042969, "fcm_dpo/q_t": 0.33768439292907715, "grad_norm": 29.296648025512695, "learning_rate": 4.874448683603694e-07, "logits/chosen": -0.6430600881576538, "logits/rejected": -0.6259501576423645, "logps/chosen": -80.70809936523438, "logps/ref_chosen": -59.00108337402344, "logps/ref_rejected": -87.89215087890625, "logps/rejected": -144.94454956054688, "loss": 0.9004, "margin_dpo/margin_mean": 35.34538269042969, "margin_dpo/margin_std": 43.01177978515625, "step": 132 }, { "KL/chosen_KL_mean": -25.513275146484375, "KL/mean": -41.42220687866211, "KL/rejected_KL_mean": -57.33114242553711, "KL/std": 34.27680206298828, "epoch": 0.19530102790014683, "fcm_dpo/beta": 0.023830700665712357, "fcm_dpo/delta": 0.04520042985677719, "fcm_dpo/margin": 31.81786346435547, "fcm_dpo/q_t": 0.35364824533462524, "grad_norm": 29.429222106933594, "learning_rate": 4.870401618977415e-07, "logits/chosen": -0.6595304012298584, "logits/rejected": -0.6431485414505005, "logps/chosen": -92.1177749633789, "logps/ref_chosen": -66.60449981689453, "logps/ref_rejected": -96.33355712890625, "logps/rejected": -153.66470336914062, "loss": 0.9495, "margin_dpo/margin_mean": 31.81786346435547, "margin_dpo/margin_std": 41.558082580566406, "step": 133 }, { "KL/chosen_KL_mean": -19.793434143066406, "KL/mean": -37.411041259765625, "KL/rejected_KL_mean": -55.02864456176758, "KL/std": 33.86731719970703, "epoch": 0.19676945668135096, "fcm_dpo/beta": 0.023871179670095444, "fcm_dpo/delta": -0.045072250068187714, "fcm_dpo/margin": 35.2352180480957, "fcm_dpo/q_t": 0.3298512101173401, "grad_norm": 27.502216339111328, "learning_rate": 4.866292092063986e-07, "logits/chosen": -0.6473065614700317, "logits/rejected": -0.6194508075714111, "logps/chosen": -71.86268615722656, "logps/ref_chosen": -52.06925582885742, "logps/ref_rejected": -87.6545181274414, "logps/rejected": -142.68316650390625, "loss": 0.8507, "margin_dpo/margin_mean": 35.23521423339844, "margin_dpo/margin_std": 36.012550354003906, "step": 134 }, { "KL/chosen_KL_mean": -24.061479568481445, "KL/mean": -45.499755859375, "KL/rejected_KL_mean": -66.93803405761719, "KL/std": 40.95748519897461, "epoch": 0.19823788546255505, "fcm_dpo/beta": 0.023015718907117844, "fcm_dpo/delta": -0.20716674625873566, "fcm_dpo/margin": 42.87656021118164, "fcm_dpo/q_t": 0.3147345781326294, "grad_norm": 29.010801315307617, "learning_rate": 4.862120211153265e-07, "logits/chosen": -0.6196516752243042, "logits/rejected": -0.6404733657836914, "logps/chosen": -74.41534423828125, "logps/ref_chosen": -50.353858947753906, "logps/ref_rejected": -115.97975158691406, "logps/rejected": -182.91778564453125, "loss": 0.8439, "margin_dpo/margin_mean": 42.876556396484375, "margin_dpo/margin_std": 47.684814453125, "step": 135 }, { "KL/chosen_KL_mean": -27.37831687927246, "KL/mean": -42.851158142089844, "KL/rejected_KL_mean": -58.32400131225586, "KL/std": 36.29102325439453, "epoch": 0.19970631424375918, "fcm_dpo/beta": 0.022867830470204353, "fcm_dpo/delta": 0.09842132776975632, "fcm_dpo/margin": 30.945680618286133, "fcm_dpo/q_t": 0.36907869577407837, "grad_norm": 30.862062454223633, "learning_rate": 4.857886086178193e-07, "logits/chosen": -0.6548190116882324, "logits/rejected": -0.6352522373199463, "logps/chosen": -92.4508285522461, "logps/ref_chosen": -65.072509765625, "logps/ref_rejected": -96.32122802734375, "logps/rejected": -154.64523315429688, "loss": 0.9884, "margin_dpo/margin_mean": 30.945680618286133, "margin_dpo/margin_std": 44.39238357543945, "step": 136 }, { "KL/chosen_KL_mean": -28.5081844329834, "KL/mean": -51.93169403076172, "KL/rejected_KL_mean": -75.35520935058594, "KL/std": 50.29423141479492, "epoch": 0.2011747430249633, "fcm_dpo/beta": 0.022213459014892578, "fcm_dpo/delta": -0.26994192600250244, "fcm_dpo/margin": 46.847015380859375, "fcm_dpo/q_t": 0.3159807026386261, "grad_norm": 29.10369110107422, "learning_rate": 4.853589828711902e-07, "logits/chosen": -0.5867665410041809, "logits/rejected": -0.6017188429832458, "logps/chosen": -77.26730346679688, "logps/ref_chosen": -48.759117126464844, "logps/ref_rejected": -113.86376953125, "logps/rejected": -189.21897888183594, "loss": 0.8694, "margin_dpo/margin_mean": 46.847015380859375, "margin_dpo/margin_std": 57.939029693603516, "step": 137 }, { "KL/chosen_KL_mean": -29.796592712402344, "KL/mean": -47.36405563354492, "KL/rejected_KL_mean": -64.9315185546875, "KL/std": 38.10211944580078, "epoch": 0.2026431718061674, "fcm_dpo/beta": 0.022151703014969826, "fcm_dpo/delta": 0.023063896223902702, "fcm_dpo/margin": 35.134925842285156, "fcm_dpo/q_t": 0.340358704328537, "grad_norm": 30.460988998413086, "learning_rate": 4.849231551964771e-07, "logits/chosen": -0.5743345022201538, "logits/rejected": -0.5551059246063232, "logps/chosen": -90.31623840332031, "logps/ref_chosen": -60.519649505615234, "logps/ref_rejected": -93.19694519042969, "logps/rejected": -158.1284637451172, "loss": 0.9103, "margin_dpo/margin_mean": 35.134925842285156, "margin_dpo/margin_std": 39.568973541259766, "step": 138 }, { "KL/chosen_KL_mean": -23.029796600341797, "KL/mean": -45.01148986816406, "KL/rejected_KL_mean": -66.9931869506836, "KL/std": 37.149497985839844, "epoch": 0.20411160058737152, "fcm_dpo/beta": 0.021606380119919777, "fcm_dpo/delta": -0.16534382104873657, "fcm_dpo/margin": 43.96338653564453, "fcm_dpo/q_t": 0.31225913763046265, "grad_norm": 27.354082107543945, "learning_rate": 4.844811370781446e-07, "logits/chosen": -0.6141539812088013, "logits/rejected": -0.5956014394760132, "logps/chosen": -69.92118072509766, "logps/ref_chosen": -46.89138412475586, "logps/ref_rejected": -79.72798156738281, "logps/rejected": -146.72116088867188, "loss": 0.8173, "margin_dpo/margin_mean": 43.96338653564453, "margin_dpo/margin_std": 44.648704528808594, "step": 139 }, { "KL/chosen_KL_mean": -27.68996238708496, "KL/mean": -48.107666015625, "KL/rejected_KL_mean": -68.52537536621094, "KL/std": 42.47513961791992, "epoch": 0.2055800293685756, "fcm_dpo/beta": 0.02111215889453888, "fcm_dpo/delta": -0.06818899512290955, "fcm_dpo/margin": 40.83540344238281, "fcm_dpo/q_t": 0.33521249890327454, "grad_norm": 28.752126693725586, "learning_rate": 4.840329401637809e-07, "logits/chosen": -0.614875316619873, "logits/rejected": -0.5954533815383911, "logps/chosen": -86.6646728515625, "logps/ref_chosen": -58.97471618652344, "logps/ref_rejected": -83.28410339355469, "logps/rejected": -151.80947875976562, "loss": 0.9081, "margin_dpo/margin_mean": 40.83540344238281, "margin_dpo/margin_std": 49.618003845214844, "step": 140 }, { "KL/chosen_KL_mean": -27.771591186523438, "KL/mean": -47.71379470825195, "KL/rejected_KL_mean": -67.65599822998047, "KL/std": 41.21592712402344, "epoch": 0.20704845814977973, "fcm_dpo/beta": 0.0209644865244627, "fcm_dpo/delta": -0.03944108262658119, "fcm_dpo/margin": 39.88440704345703, "fcm_dpo/q_t": 0.33007389307022095, "grad_norm": 30.820947647094727, "learning_rate": 4.83578576263792e-07, "logits/chosen": -0.6197242736816406, "logits/rejected": -0.5986499190330505, "logps/chosen": -102.84725952148438, "logps/ref_chosen": -75.07566833496094, "logps/ref_rejected": -98.1922607421875, "logps/rejected": -165.8482666015625, "loss": 0.9013, "margin_dpo/margin_mean": 39.88440704345703, "margin_dpo/margin_std": 45.809532165527344, "step": 141 }, { "KL/chosen_KL_mean": -32.35279846191406, "KL/mean": -55.03770446777344, "KL/rejected_KL_mean": -77.72261047363281, "KL/std": 48.86778259277344, "epoch": 0.20851688693098386, "fcm_dpo/beta": 0.0206030011177063, "fcm_dpo/delta": -0.14863386750221252, "fcm_dpo/margin": 45.36981964111328, "fcm_dpo/q_t": 0.32887327671051025, "grad_norm": 30.7178955078125, "learning_rate": 4.83118057351089e-07, "logits/chosen": -0.5786978006362915, "logits/rejected": -0.5709136128425598, "logps/chosen": -90.38072204589844, "logps/ref_chosen": -58.027931213378906, "logps/ref_rejected": -94.58222961425781, "logps/rejected": -172.30484008789062, "loss": 0.9218, "margin_dpo/margin_mean": 45.36981964111328, "margin_dpo/margin_std": 57.22596740722656, "step": 142 }, { "KL/chosen_KL_mean": -35.13017654418945, "KL/mean": -49.68499755859375, "KL/rejected_KL_mean": -64.23982238769531, "KL/std": 42.61761474609375, "epoch": 0.20998531571218795, "fcm_dpo/beta": 0.0208355113863945, "fcm_dpo/delta": 0.20649181306362152, "fcm_dpo/margin": 29.10964584350586, "fcm_dpo/q_t": 0.38302189111709595, "grad_norm": 36.91963195800781, "learning_rate": 4.826513955607734e-07, "logits/chosen": -0.56812983751297, "logits/rejected": -0.543799102306366, "logps/chosen": -92.72662353515625, "logps/ref_chosen": -57.59645080566406, "logps/ref_rejected": -78.99957275390625, "logps/rejected": -143.23939514160156, "loss": 1.0707, "margin_dpo/margin_mean": 29.10964584350586, "margin_dpo/margin_std": 47.77528762817383, "step": 143 }, { "KL/chosen_KL_mean": -28.555076599121094, "KL/mean": -47.98714065551758, "KL/rejected_KL_mean": -67.41920471191406, "KL/std": 36.25341033935547, "epoch": 0.21145374449339208, "fcm_dpo/beta": 0.021065259352326393, "fcm_dpo/delta": -0.02040477842092514, "fcm_dpo/margin": 38.8641242980957, "fcm_dpo/q_t": 0.33234626054763794, "grad_norm": 28.399396896362305, "learning_rate": 4.821786031898176e-07, "logits/chosen": -0.5759509801864624, "logits/rejected": -0.5413784980773926, "logps/chosen": -88.46144104003906, "logps/ref_chosen": -59.90636444091797, "logps/ref_rejected": -82.00025939941406, "logps/rejected": -149.41946411132812, "loss": 0.8764, "margin_dpo/margin_mean": 38.8641242980957, "margin_dpo/margin_std": 41.08622741699219, "step": 144 }, { "KL/chosen_KL_mean": -31.771686553955078, "KL/mean": -52.151512145996094, "KL/rejected_KL_mean": -72.53134155273438, "KL/std": 41.140968322753906, "epoch": 0.21292217327459617, "fcm_dpo/beta": 0.020855430513620377, "fcm_dpo/delta": -0.05491182580590248, "fcm_dpo/margin": 40.75965881347656, "fcm_dpo/q_t": 0.3329458236694336, "grad_norm": 33.59305953979492, "learning_rate": 4.816996926967401e-07, "logits/chosen": -0.5763427019119263, "logits/rejected": -0.544802725315094, "logps/chosen": -88.37235260009766, "logps/ref_chosen": -56.60066604614258, "logps/ref_rejected": -77.86631774902344, "logps/rejected": -150.3976593017578, "loss": 0.8897, "margin_dpo/margin_mean": 40.75965881347656, "margin_dpo/margin_std": 46.583221435546875, "step": 145 }, { "KL/chosen_KL_mean": -44.07976150512695, "KL/mean": -60.510459899902344, "KL/rejected_KL_mean": -76.94116973876953, "KL/std": 42.426719665527344, "epoch": 0.2143906020558003, "fcm_dpo/beta": 0.021105386316776276, "fcm_dpo/delta": 0.1144891306757927, "fcm_dpo/margin": 32.861412048339844, "fcm_dpo/q_t": 0.3631611764431, "grad_norm": 44.34247970581055, "learning_rate": 4.812146767012779e-07, "logits/chosen": -0.5526940822601318, "logits/rejected": -0.5075163841247559, "logps/chosen": -110.08021545410156, "logps/ref_chosen": -66.00045013427734, "logps/ref_rejected": -81.70278930664062, "logps/rejected": -158.64395141601562, "loss": 1.0376, "margin_dpo/margin_mean": 32.861412048339844, "margin_dpo/margin_std": 49.71653747558594, "step": 146 }, { "KL/chosen_KL_mean": -33.446563720703125, "KL/mean": -55.87425231933594, "KL/rejected_KL_mean": -78.30194091796875, "KL/std": 45.601898193359375, "epoch": 0.21585903083700442, "fcm_dpo/beta": 0.020772868767380714, "fcm_dpo/delta": -0.14638394117355347, "fcm_dpo/margin": 44.855384826660156, "fcm_dpo/q_t": 0.3264108896255493, "grad_norm": 38.08588409423828, "learning_rate": 4.807235679840536e-07, "logits/chosen": -0.5531934499740601, "logits/rejected": -0.5197881460189819, "logps/chosen": -86.85205078125, "logps/ref_chosen": -53.405487060546875, "logps/ref_rejected": -71.39060974121094, "logps/rejected": -149.6925506591797, "loss": 0.8932, "margin_dpo/margin_mean": 44.85538101196289, "margin_dpo/margin_std": 54.91014862060547, "step": 147 }, { "KL/chosen_KL_mean": -32.03932571411133, "KL/mean": -49.7523307800293, "KL/rejected_KL_mean": -67.46534729003906, "KL/std": 41.54022216796875, "epoch": 0.2173274596182085, "fcm_dpo/beta": 0.021175362169742584, "fcm_dpo/delta": 0.048354372382164, "fcm_dpo/margin": 35.426021575927734, "fcm_dpo/q_t": 0.355111300945282, "grad_norm": 30.767854690551758, "learning_rate": 4.802263794862384e-07, "logits/chosen": -0.5951809287071228, "logits/rejected": -0.5793430805206299, "logps/chosen": -96.97640991210938, "logps/ref_chosen": -64.93708038330078, "logps/ref_rejected": -103.09384155273438, "logps/rejected": -170.55918884277344, "loss": 0.9545, "margin_dpo/margin_mean": 35.426021575927734, "margin_dpo/margin_std": 43.80073547363281, "step": 148 }, { "KL/chosen_KL_mean": -31.97378158569336, "KL/mean": -54.779823303222656, "KL/rejected_KL_mean": -77.58586120605469, "KL/std": 41.12773132324219, "epoch": 0.21879588839941264, "fcm_dpo/beta": 0.02021482028067112, "fcm_dpo/delta": -0.13983891904354095, "fcm_dpo/margin": 45.612083435058594, "fcm_dpo/q_t": 0.3158126473426819, "grad_norm": 31.028966903686523, "learning_rate": 4.797231243092118e-07, "logits/chosen": -0.6176996231079102, "logits/rejected": -0.596400797367096, "logps/chosen": -90.44754028320312, "logps/ref_chosen": -58.47376251220703, "logps/ref_rejected": -99.31474304199219, "logps/rejected": -176.90060424804688, "loss": 0.8414, "margin_dpo/margin_mean": 45.612083435058594, "margin_dpo/margin_std": 47.153900146484375, "step": 149 }, { "KL/chosen_KL_mean": -29.563812255859375, "KL/mean": -51.22605895996094, "KL/rejected_KL_mean": -72.8883056640625, "KL/std": 48.249725341796875, "epoch": 0.22026431718061673, "fcm_dpo/beta": 0.019835665822029114, "fcm_dpo/delta": -0.07024933397769928, "fcm_dpo/margin": 43.324493408203125, "fcm_dpo/q_t": 0.3418238162994385, "grad_norm": 30.609365463256836, "learning_rate": 4.792138157142157e-07, "logits/chosen": -0.5815879106521606, "logits/rejected": -0.5770955681800842, "logps/chosen": -75.26962280273438, "logps/ref_chosen": -45.705810546875, "logps/ref_rejected": -83.34759521484375, "logps/rejected": -156.23590087890625, "loss": 0.9152, "margin_dpo/margin_mean": 43.324493408203125, "margin_dpo/margin_std": 56.31333923339844, "step": 150 }, { "KL/chosen_KL_mean": -33.465301513671875, "KL/mean": -56.45378112792969, "KL/rejected_KL_mean": -79.4422607421875, "KL/std": 45.18247985839844, "epoch": 0.22173274596182085, "fcm_dpo/beta": 0.019716549664735794, "fcm_dpo/delta": -0.1170579120516777, "fcm_dpo/margin": 45.976966857910156, "fcm_dpo/q_t": 0.31761401891708374, "grad_norm": 29.541719436645508, "learning_rate": 4.786984671220053e-07, "logits/chosen": -0.6623918414115906, "logits/rejected": -0.6282558441162109, "logps/chosen": -104.0361328125, "logps/ref_chosen": -70.57083129882812, "logps/ref_rejected": -100.46382141113281, "logps/rejected": -179.90606689453125, "loss": 0.8425, "margin_dpo/margin_mean": 45.976966857910156, "margin_dpo/margin_std": 47.510108947753906, "step": 151 }, { "KL/chosen_KL_mean": -28.917739868164062, "KL/mean": -54.12328338623047, "KL/rejected_KL_mean": -79.32882690429688, "KL/std": 45.731834411621094, "epoch": 0.22320117474302498, "fcm_dpo/beta": 0.019173946231603622, "fcm_dpo/delta": -0.18405033648014069, "fcm_dpo/margin": 50.41108703613281, "fcm_dpo/q_t": 0.31479883193969727, "grad_norm": 27.932998657226562, "learning_rate": 4.78177092112495e-07, "logits/chosen": -0.6130504608154297, "logits/rejected": -0.6028048992156982, "logps/chosen": -89.08212280273438, "logps/ref_chosen": -60.16438674926758, "logps/ref_rejected": -106.14045715332031, "logps/rejected": -185.4692840576172, "loss": 0.8384, "margin_dpo/margin_mean": 50.41108703613281, "margin_dpo/margin_std": 53.3874397277832, "step": 152 }, { "KL/chosen_KL_mean": -28.78639030456543, "KL/mean": -50.361366271972656, "KL/rejected_KL_mean": -71.93634033203125, "KL/std": 47.18241500854492, "epoch": 0.22466960352422907, "fcm_dpo/beta": 0.018890701234340668, "fcm_dpo/delta": -0.016644855961203575, "fcm_dpo/margin": 43.14995193481445, "fcm_dpo/q_t": 0.3440234959125519, "grad_norm": 25.5745792388916, "learning_rate": 4.776497044244016e-07, "logits/chosen": -0.616934061050415, "logits/rejected": -0.6062427163124084, "logps/chosen": -85.10166931152344, "logps/ref_chosen": -56.315277099609375, "logps/ref_rejected": -85.65583801269531, "logps/rejected": -157.59217834472656, "loss": 0.9313, "margin_dpo/margin_mean": 43.14995193481445, "margin_dpo/margin_std": 55.69709014892578, "step": 153 }, { "KL/chosen_KL_mean": -33.958927154541016, "KL/mean": -55.02075958251953, "KL/rejected_KL_mean": -76.08258056640625, "KL/std": 47.15727615356445, "epoch": 0.2261380323054332, "fcm_dpo/beta": 0.018977440893650055, "fcm_dpo/delta": -0.0009660199284553528, "fcm_dpo/margin": 42.1236572265625, "fcm_dpo/q_t": 0.3485991954803467, "grad_norm": 31.434709548950195, "learning_rate": 4.771163179548808e-07, "logits/chosen": -0.6276504993438721, "logits/rejected": -0.6189817190170288, "logps/chosen": -96.70149230957031, "logps/ref_chosen": -62.74256896972656, "logps/ref_rejected": -104.24420166015625, "logps/rejected": -180.3267822265625, "loss": 0.9699, "margin_dpo/margin_mean": 42.123653411865234, "margin_dpo/margin_std": 56.15239715576172, "step": 154 }, { "KL/chosen_KL_mean": -28.95609474182129, "KL/mean": -51.265785217285156, "KL/rejected_KL_mean": -73.57547760009766, "KL/std": 45.23421859741211, "epoch": 0.2276064610866373, "fcm_dpo/beta": 0.01880238763988018, "fcm_dpo/delta": -0.04300792142748833, "fcm_dpo/margin": 44.61937713623047, "fcm_dpo/q_t": 0.33316513895988464, "grad_norm": 27.293209075927734, "learning_rate": 4.7657694675916247e-07, "logits/chosen": -0.6241185665130615, "logits/rejected": -0.5972343683242798, "logps/chosen": -89.6092758178711, "logps/ref_chosen": -60.65318298339844, "logps/ref_rejected": -77.49220275878906, "logps/rejected": -151.06768798828125, "loss": 0.896, "margin_dpo/margin_mean": 44.61937713623047, "margin_dpo/margin_std": 52.05685043334961, "step": 155 }, { "KL/chosen_KL_mean": -41.33610534667969, "KL/mean": -55.682823181152344, "KL/rejected_KL_mean": -70.02952575683594, "KL/std": 44.41982650756836, "epoch": 0.2290748898678414, "fcm_dpo/beta": 0.019381992518901825, "fcm_dpo/delta": 0.25817906856536865, "fcm_dpo/margin": 28.693424224853516, "fcm_dpo/q_t": 0.39033377170562744, "grad_norm": 43.62647247314453, "learning_rate": 4.7603160505017893e-07, "logits/chosen": -0.561880350112915, "logits/rejected": -0.5312899947166443, "logps/chosen": -110.82798767089844, "logps/ref_chosen": -69.49188232421875, "logps/ref_rejected": -77.16929626464844, "logps/rejected": -147.19882202148438, "loss": 1.176, "margin_dpo/margin_mean": 28.693424224853516, "margin_dpo/margin_std": 59.12420654296875, "step": 156 }, { "KL/chosen_KL_mean": -38.10362243652344, "KL/mean": -64.24569702148438, "KL/rejected_KL_mean": -90.38777923583984, "KL/std": 48.962730407714844, "epoch": 0.2305433186490455, "fcm_dpo/beta": 0.018893180415034294, "fcm_dpo/delta": -0.21114027500152588, "fcm_dpo/margin": 52.28416061401367, "fcm_dpo/q_t": 0.30563968420028687, "grad_norm": 30.84547233581543, "learning_rate": 4.7548030719819154e-07, "logits/chosen": -0.608643651008606, "logits/rejected": -0.6031553745269775, "logps/chosen": -99.47206115722656, "logps/ref_chosen": -61.368438720703125, "logps/ref_rejected": -107.64636993408203, "logps/rejected": -198.03414916992188, "loss": 0.8289, "margin_dpo/margin_mean": 52.28416061401367, "margin_dpo/margin_std": 53.84667205810547, "step": 157 }, { "KL/chosen_KL_mean": -34.95962142944336, "KL/mean": -60.97590255737305, "KL/rejected_KL_mean": -86.9921875, "KL/std": 58.75188446044922, "epoch": 0.23201174743024963, "fcm_dpo/beta": 0.018339555710554123, "fcm_dpo/delta": -0.1707034558057785, "fcm_dpo/margin": 52.03257751464844, "fcm_dpo/q_t": 0.3332700729370117, "grad_norm": 25.7038631439209, "learning_rate": 4.7492306773041136e-07, "logits/chosen": -0.5895273685455322, "logits/rejected": -0.5938005447387695, "logps/chosen": -92.57254028320312, "logps/ref_chosen": -57.612918853759766, "logps/ref_rejected": -113.6946792602539, "logps/rejected": -200.68685913085938, "loss": 0.9016, "margin_dpo/margin_mean": 52.03257751464844, "margin_dpo/margin_std": 68.59899139404297, "step": 158 }, { "KL/chosen_KL_mean": -39.52997970581055, "KL/mean": -60.189544677734375, "KL/rejected_KL_mean": -80.84910583496094, "KL/std": 51.77936553955078, "epoch": 0.23348017621145375, "fcm_dpo/beta": 0.018413972109556198, "fcm_dpo/delta": 0.0413820706307888, "fcm_dpo/margin": 41.31913375854492, "fcm_dpo/q_t": 0.35460609197616577, "grad_norm": 34.12101745605469, "learning_rate": 4.743599013306165e-07, "logits/chosen": -0.5919934511184692, "logits/rejected": -0.5467137694358826, "logps/chosen": -121.09032440185547, "logps/ref_chosen": -81.56034851074219, "logps/ref_rejected": -88.89871215820312, "logps/rejected": -169.74781799316406, "loss": 0.9757, "margin_dpo/margin_mean": 41.31913757324219, "margin_dpo/margin_std": 57.08066940307617, "step": 159 }, { "KL/chosen_KL_mean": -39.19923400878906, "KL/mean": -63.417083740234375, "KL/rejected_KL_mean": -87.63493347167969, "KL/std": 52.864295959472656, "epoch": 0.23494860499265785, "fcm_dpo/beta": 0.01803383231163025, "fcm_dpo/delta": -0.081442691385746, "fcm_dpo/margin": 48.43570327758789, "fcm_dpo/q_t": 0.34352871775627136, "grad_norm": 30.687528610229492, "learning_rate": 4.737908228387656e-07, "logits/chosen": -0.5659890174865723, "logits/rejected": -0.5416566729545593, "logps/chosen": -104.93011474609375, "logps/ref_chosen": -65.73088073730469, "logps/ref_rejected": -97.21781921386719, "logps/rejected": -184.85275268554688, "loss": 0.9425, "margin_dpo/margin_mean": 48.43570327758789, "margin_dpo/margin_std": 65.908935546875, "step": 160 }, { "KL/chosen_KL_mean": -35.278465270996094, "KL/mean": -57.27626037597656, "KL/rejected_KL_mean": -79.2740478515625, "KL/std": 46.100494384765625, "epoch": 0.23641703377386197, "fcm_dpo/beta": 0.0180988647043705, "fcm_dpo/delta": 0.0039350297302007675, "fcm_dpo/margin": 43.995582580566406, "fcm_dpo/q_t": 0.34731143712997437, "grad_norm": 30.22264862060547, "learning_rate": 4.7321584725060594e-07, "logits/chosen": -0.5966364741325378, "logits/rejected": -0.5794234275817871, "logps/chosen": -87.71493530273438, "logps/ref_chosen": -52.43647003173828, "logps/ref_rejected": -83.43095397949219, "logps/rejected": -162.7050018310547, "loss": 0.9439, "margin_dpo/margin_mean": 43.995582580566406, "margin_dpo/margin_std": 56.44627380371094, "step": 161 }, { "KL/chosen_KL_mean": -33.17871856689453, "KL/mean": -56.08073425292969, "KL/rejected_KL_mean": -78.98275756835938, "KL/std": 51.76576232910156, "epoch": 0.23788546255506607, "fcm_dpo/beta": 0.017755115404725075, "fcm_dpo/delta": -0.01901736855506897, "fcm_dpo/margin": 45.804046630859375, "fcm_dpo/q_t": 0.3460782766342163, "grad_norm": 25.410987854003906, "learning_rate": 4.7263498971727905e-07, "logits/chosen": -0.6189196109771729, "logits/rejected": -0.5900764465332031, "logps/chosen": -95.789306640625, "logps/ref_chosen": -62.6105842590332, "logps/ref_rejected": -89.39057922363281, "logps/rejected": -168.37335205078125, "loss": 0.9361, "margin_dpo/margin_mean": 45.804046630859375, "margin_dpo/margin_std": 59.16350173950195, "step": 162 }, { "KL/chosen_KL_mean": -34.73664474487305, "KL/mean": -56.72509002685547, "KL/rejected_KL_mean": -78.71353149414062, "KL/std": 45.937137603759766, "epoch": 0.2393538913362702, "fcm_dpo/beta": 0.0180361308157444, "fcm_dpo/delta": 0.007369140163064003, "fcm_dpo/margin": 43.976890563964844, "fcm_dpo/q_t": 0.34393471479415894, "grad_norm": 26.971351623535156, "learning_rate": 4.720482655449212e-07, "logits/chosen": -0.5807976126670837, "logits/rejected": -0.5515669584274292, "logps/chosen": -89.75827026367188, "logps/ref_chosen": -55.021629333496094, "logps/ref_rejected": -75.418212890625, "logps/rejected": -154.13174438476562, "loss": 0.9156, "margin_dpo/margin_mean": 43.976890563964844, "margin_dpo/margin_std": 52.59187316894531, "step": 163 }, { "KL/chosen_KL_mean": -32.905269622802734, "KL/mean": -58.444881439208984, "KL/rejected_KL_mean": -83.98448181152344, "KL/std": 46.80143737792969, "epoch": 0.24082232011747431, "fcm_dpo/beta": 0.017528720200061798, "fcm_dpo/delta": -0.10878677666187286, "fcm_dpo/margin": 51.079219818115234, "fcm_dpo/q_t": 0.3195659816265106, "grad_norm": 27.89398765563965, "learning_rate": 4.714556901942599e-07, "logits/chosen": -0.5946141481399536, "logits/rejected": -0.5662475228309631, "logps/chosen": -88.54593658447266, "logps/ref_chosen": -55.64066696166992, "logps/ref_rejected": -79.66463470458984, "logps/rejected": -163.64910888671875, "loss": 0.8391, "margin_dpo/margin_mean": 51.07921600341797, "margin_dpo/margin_std": 52.178611755371094, "step": 164 }, { "KL/chosen_KL_mean": -36.89280700683594, "KL/mean": -54.11333084106445, "KL/rejected_KL_mean": -71.3338623046875, "KL/std": 39.27880859375, "epoch": 0.2422907488986784, "fcm_dpo/beta": 0.018116144463419914, "fcm_dpo/delta": 0.18791693449020386, "fcm_dpo/margin": 34.4410514831543, "fcm_dpo/q_t": 0.37251168489456177, "grad_norm": 35.580360412597656, "learning_rate": 4.708572792802069e-07, "logits/chosen": -0.6063022613525391, "logits/rejected": -0.5636542439460754, "logps/chosen": -98.2034912109375, "logps/ref_chosen": -61.310691833496094, "logps/ref_rejected": -73.67060852050781, "logps/rejected": -145.0044708251953, "loss": 1.0123, "margin_dpo/margin_mean": 34.4410514831543, "margin_dpo/margin_std": 47.09235382080078, "step": 165 }, { "KL/chosen_KL_mean": -34.758277893066406, "KL/mean": -64.99706268310547, "KL/rejected_KL_mean": -95.23585510253906, "KL/std": 59.283321380615234, "epoch": 0.24375917767988253, "fcm_dpo/beta": 0.017409585416316986, "fcm_dpo/delta": -0.2863610088825226, "fcm_dpo/margin": 60.477569580078125, "fcm_dpo/q_t": 0.3133409023284912, "grad_norm": 25.431591033935547, "learning_rate": 4.702530485714461e-07, "logits/chosen": -0.5450478792190552, "logits/rejected": -0.5523861050605774, "logps/chosen": -85.74188232421875, "logps/ref_chosen": -50.98360061645508, "logps/ref_rejected": -98.09512329101562, "logps/rejected": -193.3309783935547, "loss": 0.8277, "margin_dpo/margin_mean": 60.477569580078125, "margin_dpo/margin_std": 71.7076187133789, "step": 166 }, { "KL/chosen_KL_mean": -35.36009979248047, "KL/mean": -67.31937408447266, "KL/rejected_KL_mean": -99.27864837646484, "KL/std": 54.32081604003906, "epoch": 0.24522760646108663, "fcm_dpo/beta": 0.01653553918004036, "fcm_dpo/delta": -0.28801417350769043, "fcm_dpo/margin": 63.918548583984375, "fcm_dpo/q_t": 0.29091453552246094, "grad_norm": 25.4301815032959, "learning_rate": 4.6964301399001877e-07, "logits/chosen": -0.49997496604919434, "logits/rejected": -0.4990348219871521, "logps/chosen": -85.78419494628906, "logps/ref_chosen": -50.424095153808594, "logps/ref_rejected": -96.03042602539062, "logps/rejected": -195.30908203125, "loss": 0.7685, "margin_dpo/margin_mean": 63.918548583984375, "margin_dpo/margin_std": 59.32880783081055, "step": 167 }, { "KL/chosen_KL_mean": -39.87940979003906, "KL/mean": -64.94029235839844, "KL/rejected_KL_mean": -90.00116729736328, "KL/std": 51.81836700439453, "epoch": 0.24669603524229075, "fcm_dpo/beta": 0.0162336602807045, "fcm_dpo/delta": -0.015112994238734245, "fcm_dpo/margin": 50.12175750732422, "fcm_dpo/q_t": 0.33712950348854065, "grad_norm": 28.887313842773438, "learning_rate": 4.690271916109034e-07, "logits/chosen": -0.5341615676879883, "logits/rejected": -0.5141159296035767, "logps/chosen": -89.34223937988281, "logps/ref_chosen": -49.462825775146484, "logps/ref_rejected": -75.30855560302734, "logps/rejected": -165.30972290039062, "loss": 0.8878, "margin_dpo/margin_mean": 50.12175750732422, "margin_dpo/margin_std": 55.842750549316406, "step": 168 }, { "KL/chosen_KL_mean": -38.55489730834961, "KL/mean": -60.85240936279297, "KL/rejected_KL_mean": -83.14991760253906, "KL/std": 52.6884765625, "epoch": 0.24816446402349487, "fcm_dpo/beta": 0.016293514519929886, "fcm_dpo/delta": 0.07776357233524323, "fcm_dpo/margin": 44.59502029418945, "fcm_dpo/q_t": 0.3672984838485718, "grad_norm": 31.865880966186523, "learning_rate": 4.6840559766159235e-07, "logits/chosen": -0.5834333896636963, "logits/rejected": -0.5621183514595032, "logps/chosen": -98.35833740234375, "logps/ref_chosen": -59.803443908691406, "logps/ref_rejected": -83.34574890136719, "logps/rejected": -166.49566650390625, "loss": 1.0159, "margin_dpo/margin_mean": 44.59502410888672, "margin_dpo/margin_std": 69.36247253417969, "step": 169 }, { "KL/chosen_KL_mean": -35.542457580566406, "KL/mean": -63.328102111816406, "KL/rejected_KL_mean": -91.11375427246094, "KL/std": 50.676841735839844, "epoch": 0.24963289280469897, "fcm_dpo/beta": 0.016118060797452927, "fcm_dpo/delta": -0.10781846940517426, "fcm_dpo/margin": 55.5712890625, "fcm_dpo/q_t": 0.3201714754104614, "grad_norm": 27.23699188232422, "learning_rate": 4.6777824852166437e-07, "logits/chosen": -0.504392147064209, "logits/rejected": -0.48632174730300903, "logps/chosen": -85.01423645019531, "logps/ref_chosen": -49.471771240234375, "logps/ref_rejected": -75.91734313964844, "logps/rejected": -167.03109741210938, "loss": 0.8564, "margin_dpo/margin_mean": 55.571292877197266, "margin_dpo/margin_std": 57.82054138183594, "step": 170 }, { "KL/chosen_KL_mean": -52.33512878417969, "KL/mean": -75.137939453125, "KL/rejected_KL_mean": -97.94075775146484, "KL/std": 56.281578063964844, "epoch": 0.2511013215859031, "fcm_dpo/beta": 0.01631205342710018, "fcm_dpo/delta": 0.060608040541410446, "fcm_dpo/margin": 45.605628967285156, "fcm_dpo/q_t": 0.3616752624511719, "grad_norm": 37.65127182006836, "learning_rate": 4.6714516072235273e-07, "logits/chosen": -0.5643775463104248, "logits/rejected": -0.5289890170097351, "logps/chosen": -136.83444213867188, "logps/ref_chosen": -84.49931335449219, "logps/ref_rejected": -109.38209533691406, "logps/rejected": -207.32284545898438, "loss": 1.0027, "margin_dpo/margin_mean": 45.605628967285156, "margin_dpo/margin_std": 68.96087646484375, "step": 171 }, { "KL/chosen_KL_mean": -51.69752502441406, "KL/mean": -75.31649780273438, "KL/rejected_KL_mean": -98.93547058105469, "KL/std": 57.76668930053711, "epoch": 0.2525697503671072, "fcm_dpo/beta": 0.016372080892324448, "fcm_dpo/delta": 0.028607267886400223, "fcm_dpo/margin": 47.237937927246094, "fcm_dpo/q_t": 0.3496333956718445, "grad_norm": 35.594093322753906, "learning_rate": 4.6650635094610966e-07, "logits/chosen": -0.5606852769851685, "logits/rejected": -0.5259276032447815, "logps/chosen": -120.3514404296875, "logps/ref_chosen": -68.65391540527344, "logps/ref_rejected": -85.43667602539062, "logps/rejected": -184.37213134765625, "loss": 0.9619, "margin_dpo/margin_mean": 47.237937927246094, "margin_dpo/margin_std": 63.03688049316406, "step": 172 }, { "KL/chosen_KL_mean": -45.648590087890625, "KL/mean": -69.3135757446289, "KL/rejected_KL_mean": -92.97855377197266, "KL/std": 52.03070831298828, "epoch": 0.2540381791483113, "fcm_dpo/beta": 0.016575288027524948, "fcm_dpo/delta": 0.016358211636543274, "fcm_dpo/margin": 47.3299560546875, "fcm_dpo/q_t": 0.3418443202972412, "grad_norm": 34.87374496459961, "learning_rate": 4.6586183602616687e-07, "logits/chosen": -0.5738804340362549, "logits/rejected": -0.5336560606956482, "logps/chosen": -108.69947052001953, "logps/ref_chosen": -63.050880432128906, "logps/ref_rejected": -78.68392181396484, "logps/rejected": -171.6624755859375, "loss": 0.9043, "margin_dpo/margin_mean": 47.329952239990234, "margin_dpo/margin_std": 54.01713943481445, "step": 173 }, { "KL/chosen_KL_mean": -46.467247009277344, "KL/mean": -72.35551452636719, "KL/rejected_KL_mean": -98.24378967285156, "KL/std": 57.77963638305664, "epoch": 0.2555066079295154, "fcm_dpo/beta": 0.01650991104543209, "fcm_dpo/delta": -0.06092230975627899, "fcm_dpo/margin": 51.776527404785156, "fcm_dpo/q_t": 0.33887800574302673, "grad_norm": 39.11122131347656, "learning_rate": 4.652116329460919e-07, "logits/chosen": -0.49907761812210083, "logits/rejected": -0.5138338804244995, "logps/chosen": -99.83021545410156, "logps/ref_chosen": -53.36296844482422, "logps/ref_rejected": -101.91120910644531, "logps/rejected": -200.15499877929688, "loss": 0.918, "margin_dpo/margin_mean": 51.776527404785156, "margin_dpo/margin_std": 64.13847351074219, "step": 174 }, { "KL/chosen_KL_mean": -44.84705352783203, "KL/mean": -81.27606201171875, "KL/rejected_KL_mean": -117.70506286621094, "KL/std": 65.36940002441406, "epoch": 0.25697503671071953, "fcm_dpo/beta": 0.01541909109801054, "fcm_dpo/delta": -0.36564406752586365, "fcm_dpo/margin": 72.85799407958984, "fcm_dpo/q_t": 0.28254061937332153, "grad_norm": 37.300682067871094, "learning_rate": 4.645557588393406e-07, "logits/chosen": -0.4535369873046875, "logits/rejected": -0.4410516023635864, "logps/chosen": -90.26481628417969, "logps/ref_chosen": -45.417762756347656, "logps/ref_rejected": -89.50579833984375, "logps/rejected": -207.2108612060547, "loss": 0.7328, "margin_dpo/margin_mean": 72.85800170898438, "margin_dpo/margin_std": 66.75201416015625, "step": 175 }, { "KL/chosen_KL_mean": -46.950782775878906, "KL/mean": -76.24058532714844, "KL/rejected_KL_mean": -105.53038024902344, "KL/std": 59.06817626953125, "epoch": 0.25844346549192365, "fcm_dpo/beta": 0.015022006817162037, "fcm_dpo/delta": -0.08763821423053741, "fcm_dpo/margin": 58.57960510253906, "fcm_dpo/q_t": 0.3318549394607544, "grad_norm": 40.32584762573242, "learning_rate": 4.638942309888058e-07, "logits/chosen": -0.42935019731521606, "logits/rejected": -0.44252365827560425, "logps/chosen": -97.40362548828125, "logps/ref_chosen": -50.452842712402344, "logps/ref_rejected": -95.5589599609375, "logps/rejected": -201.08935546875, "loss": 0.8749, "margin_dpo/margin_mean": 58.57960510253906, "margin_dpo/margin_std": 67.5830307006836, "step": 176 }, { "KL/chosen_KL_mean": -52.787109375, "KL/mean": -83.38577270507812, "KL/rejected_KL_mean": -113.98443603515625, "KL/std": 65.90585327148438, "epoch": 0.2599118942731278, "fcm_dpo/beta": 0.014719298109412193, "fcm_dpo/delta": -0.11064187437295914, "fcm_dpo/margin": 61.197330474853516, "fcm_dpo/q_t": 0.32784807682037354, "grad_norm": 42.42573928833008, "learning_rate": 4.6322706682636137e-07, "logits/chosen": -0.5027165412902832, "logits/rejected": -0.48637324571609497, "logps/chosen": -114.00357055664062, "logps/ref_chosen": -61.216468811035156, "logps/ref_rejected": -95.89378356933594, "logps/rejected": -209.8782196044922, "loss": 0.8677, "margin_dpo/margin_mean": 61.19733428955078, "margin_dpo/margin_std": 69.97938537597656, "step": 177 }, { "KL/chosen_KL_mean": -60.033023834228516, "KL/mean": -97.84568786621094, "KL/rejected_KL_mean": -135.65835571289062, "KL/std": 80.01490020751953, "epoch": 0.26138032305433184, "fcm_dpo/beta": 0.013894779607653618, "fcm_dpo/delta": -0.282906174659729, "fcm_dpo/margin": 75.62533569335938, "fcm_dpo/q_t": 0.30688485503196716, "grad_norm": 39.59940719604492, "learning_rate": 4.6255428393240354e-07, "logits/chosen": -0.4129112660884857, "logits/rejected": -0.4165341258049011, "logps/chosen": -118.29781341552734, "logps/ref_chosen": -58.26478958129883, "logps/ref_rejected": -105.3653335571289, "logps/rejected": -241.02369689941406, "loss": 0.8254, "margin_dpo/margin_mean": 75.62533569335938, "margin_dpo/margin_std": 86.36256408691406, "step": 178 }, { "KL/chosen_KL_mean": -57.656192779541016, "KL/mean": -87.85360717773438, "KL/rejected_KL_mean": -118.05103302001953, "KL/std": 66.25624084472656, "epoch": 0.26284875183553597, "fcm_dpo/beta": 0.01362140104174614, "fcm_dpo/delta": -0.02604658529162407, "fcm_dpo/margin": 60.394832611083984, "fcm_dpo/q_t": 0.3425036668777466, "grad_norm": 57.43539047241211, "learning_rate": 4.6187590003538724e-07, "logits/chosen": -0.4739261269569397, "logits/rejected": -0.47013968229293823, "logps/chosen": -118.71451568603516, "logps/ref_chosen": -61.05832290649414, "logps/ref_rejected": -90.52782440185547, "logps/rejected": -208.578857421875, "loss": 0.9654, "margin_dpo/margin_mean": 60.394832611083984, "margin_dpo/margin_std": 80.08963775634766, "step": 179 }, { "KL/chosen_KL_mean": -48.66742706298828, "KL/mean": -82.73980712890625, "KL/rejected_KL_mean": -116.81217193603516, "KL/std": 60.23838424682617, "epoch": 0.2643171806167401, "fcm_dpo/beta": 0.013405999168753624, "fcm_dpo/delta": -0.12565259635448456, "fcm_dpo/margin": 68.14473724365234, "fcm_dpo/q_t": 0.3189559280872345, "grad_norm": 52.571556091308594, "learning_rate": 4.611919330113591e-07, "logits/chosen": -0.47179630398750305, "logits/rejected": -0.45903003215789795, "logps/chosen": -103.01014709472656, "logps/ref_chosen": -54.34272003173828, "logps/ref_rejected": -98.21183776855469, "logps/rejected": -215.02401733398438, "loss": 0.8536, "margin_dpo/margin_mean": 68.14473724365234, "margin_dpo/margin_std": 72.47191619873047, "step": 180 }, { "KL/chosen_KL_mean": -39.98610305786133, "KL/mean": -62.656532287597656, "KL/rejected_KL_mean": -85.32696533203125, "KL/std": 53.778289794921875, "epoch": 0.2657856093979442, "fcm_dpo/beta": 0.013776125386357307, "fcm_dpo/delta": 0.18641813099384308, "fcm_dpo/margin": 45.34085464477539, "fcm_dpo/q_t": 0.37443986535072327, "grad_norm": 26.93648338317871, "learning_rate": 4.605024008834863e-07, "logits/chosen": -0.5331726670265198, "logits/rejected": -0.5003119707107544, "logps/chosen": -94.98655700683594, "logps/ref_chosen": -55.000457763671875, "logps/ref_rejected": -61.656166076660156, "logps/rejected": -146.98312377929688, "loss": 1.0225, "margin_dpo/margin_mean": 45.34085464477539, "margin_dpo/margin_std": 65.32242584228516, "step": 181 }, { "KL/chosen_KL_mean": -38.98612594604492, "KL/mean": -74.91706848144531, "KL/rejected_KL_mean": -110.84800720214844, "KL/std": 66.33081817626953, "epoch": 0.26725403817914833, "fcm_dpo/beta": 0.013408947736024857, "fcm_dpo/delta": -0.18331019580364227, "fcm_dpo/margin": 71.86188507080078, "fcm_dpo/q_t": 0.3110619783401489, "grad_norm": 27.608600616455078, "learning_rate": 4.598073218215817e-07, "logits/chosen": -0.4613451063632965, "logits/rejected": -0.4604541063308716, "logps/chosen": -80.09397888183594, "logps/ref_chosen": -41.107852935791016, "logps/ref_rejected": -89.5215835571289, "logps/rejected": -200.36959838867188, "loss": 0.8519, "margin_dpo/margin_mean": 71.86187744140625, "margin_dpo/margin_std": 77.41173553466797, "step": 182 }, { "KL/chosen_KL_mean": -55.72860336303711, "KL/mean": -78.53468322753906, "KL/rejected_KL_mean": -101.34075927734375, "KL/std": 50.84642791748047, "epoch": 0.2687224669603524, "fcm_dpo/beta": 0.013560278341174126, "fcm_dpo/delta": 0.19175204634666443, "fcm_dpo/margin": 45.612159729003906, "fcm_dpo/q_t": 0.3727971911430359, "grad_norm": 30.786500930786133, "learning_rate": 4.5910671414162484e-07, "logits/chosen": -0.5146129131317139, "logits/rejected": -0.4842214584350586, "logps/chosen": -113.25316619873047, "logps/ref_chosen": -57.52456283569336, "logps/ref_rejected": -75.97572326660156, "logps/rejected": -177.3164825439453, "loss": 1.0039, "margin_dpo/margin_mean": 45.612159729003906, "margin_dpo/margin_std": 55.62657928466797, "step": 183 }, { "KL/chosen_KL_mean": -42.92633056640625, "KL/mean": -64.14622497558594, "KL/rejected_KL_mean": -85.3661117553711, "KL/std": 48.59588623046875, "epoch": 0.2701908957415565, "fcm_dpo/beta": 0.014429578557610512, "fcm_dpo/delta": 0.19666652381420135, "fcm_dpo/margin": 42.43977737426758, "fcm_dpo/q_t": 0.37740087509155273, "grad_norm": 24.578208923339844, "learning_rate": 4.5840059630527985e-07, "logits/chosen": -0.5529508590698242, "logits/rejected": -0.5374755263328552, "logps/chosen": -101.47128295898438, "logps/ref_chosen": -58.544952392578125, "logps/ref_rejected": -76.63406372070312, "logps/rejected": -162.00018310546875, "loss": 1.0119, "margin_dpo/margin_mean": 42.43977737426758, "margin_dpo/margin_std": 58.118682861328125, "step": 184 }, { "KL/chosen_KL_mean": -47.32001495361328, "KL/mean": -64.94092559814453, "KL/rejected_KL_mean": -82.56183624267578, "KL/std": 51.73418045043945, "epoch": 0.27165932452276065, "fcm_dpo/beta": 0.01505757961422205, "fcm_dpo/delta": 0.2848103940486908, "fcm_dpo/margin": 35.24181365966797, "fcm_dpo/q_t": 0.3963721990585327, "grad_norm": 37.11123275756836, "learning_rate": 4.5768898691940836e-07, "logits/chosen": -0.549095094203949, "logits/rejected": -0.5053994655609131, "logps/chosen": -109.34586334228516, "logps/ref_chosen": -62.025848388671875, "logps/ref_rejected": -73.7625961303711, "logps/rejected": -156.32443237304688, "loss": 1.0906, "margin_dpo/margin_mean": 35.24181365966797, "margin_dpo/margin_std": 59.88246154785156, "step": 185 }, { "KL/chosen_KL_mean": -44.749820709228516, "KL/mean": -75.26815795898438, "KL/rejected_KL_mean": -105.7864990234375, "KL/std": 57.48197937011719, "epoch": 0.27312775330396477, "fcm_dpo/beta": 0.014994775876402855, "fcm_dpo/delta": -0.12706023454666138, "fcm_dpo/margin": 61.03667449951172, "fcm_dpo/q_t": 0.3158210217952728, "grad_norm": 40.3712043762207, "learning_rate": 4.5697190473557947e-07, "logits/chosen": -0.5640110373497009, "logits/rejected": -0.5185353755950928, "logps/chosen": -114.10328674316406, "logps/ref_chosen": -69.35346984863281, "logps/ref_rejected": -88.07244873046875, "logps/rejected": -193.85894775390625, "loss": 0.8233, "margin_dpo/margin_mean": 61.03667068481445, "margin_dpo/margin_std": 60.81195831298828, "step": 186 }, { "KL/chosen_KL_mean": -43.97365951538086, "KL/mean": -68.62959289550781, "KL/rejected_KL_mean": -93.28553009033203, "KL/std": 53.63896179199219, "epoch": 0.2745961820851689, "fcm_dpo/beta": 0.015185382217168808, "fcm_dpo/delta": 0.05268273502588272, "fcm_dpo/margin": 49.31187438964844, "fcm_dpo/q_t": 0.35000455379486084, "grad_norm": 30.36644172668457, "learning_rate": 4.5624936864957555e-07, "logits/chosen": -0.5303990840911865, "logits/rejected": -0.5252684354782104, "logps/chosen": -96.73011779785156, "logps/ref_chosen": -52.7564582824707, "logps/ref_rejected": -81.96910095214844, "logps/rejected": -175.254638671875, "loss": 0.9437, "margin_dpo/margin_mean": 49.3118782043457, "margin_dpo/margin_std": 60.085113525390625, "step": 187 }, { "KL/chosen_KL_mean": -39.8115234375, "KL/mean": -70.60044860839844, "KL/rejected_KL_mean": -101.38937377929688, "KL/std": 61.41886520385742, "epoch": 0.27606461086637296, "fcm_dpo/beta": 0.01480124332010746, "fcm_dpo/delta": -0.12262607365846634, "fcm_dpo/margin": 61.577857971191406, "fcm_dpo/q_t": 0.3225354552268982, "grad_norm": 36.456275939941406, "learning_rate": 4.5552139770089454e-07, "logits/chosen": -0.5304605960845947, "logits/rejected": -0.525503396987915, "logps/chosen": -89.22700500488281, "logps/ref_chosen": -49.415489196777344, "logps/ref_rejected": -89.54043579101562, "logps/rejected": -190.9298095703125, "loss": 0.8553, "margin_dpo/margin_mean": 61.577857971191406, "margin_dpo/margin_std": 67.12760925292969, "step": 188 }, { "KL/chosen_KL_mean": -46.008487701416016, "KL/mean": -73.29472351074219, "KL/rejected_KL_mean": -100.58096313476562, "KL/std": 59.5657844543457, "epoch": 0.2775330396475771, "fcm_dpo/beta": 0.014734582975506783, "fcm_dpo/delta": -0.004572561010718346, "fcm_dpo/margin": 54.572471618652344, "fcm_dpo/q_t": 0.3462943434715271, "grad_norm": 32.416141510009766, "learning_rate": 4.5478801107224794e-07, "logits/chosen": -0.5101655721664429, "logits/rejected": -0.47071373462677, "logps/chosen": -98.4074478149414, "logps/ref_chosen": -52.39896011352539, "logps/ref_rejected": -72.16735076904297, "logps/rejected": -172.74832153320312, "loss": 0.9359, "margin_dpo/margin_mean": 54.572471618652344, "margin_dpo/margin_std": 71.166015625, "step": 189 }, { "KL/chosen_KL_mean": -49.97734069824219, "KL/mean": -79.54035949707031, "KL/rejected_KL_mean": -109.10337829589844, "KL/std": 62.99628829956055, "epoch": 0.2790014684287812, "fcm_dpo/beta": 0.014730885624885559, "fcm_dpo/delta": -0.08081679046154022, "fcm_dpo/margin": 59.12603759765625, "fcm_dpo/q_t": 0.336655855178833, "grad_norm": 32.33114242553711, "learning_rate": 4.5404922808905543e-07, "logits/chosen": -0.5516223907470703, "logits/rejected": -0.525041937828064, "logps/chosen": -114.660400390625, "logps/ref_chosen": -64.68305969238281, "logps/ref_rejected": -102.55052185058594, "logps/rejected": -211.65390014648438, "loss": 0.9259, "margin_dpo/margin_mean": 59.12604522705078, "margin_dpo/margin_std": 73.30282592773438, "step": 190 }, { "KL/chosen_KL_mean": -43.832725524902344, "KL/mean": -86.67939758300781, "KL/rejected_KL_mean": -129.52606201171875, "KL/std": 77.8781967163086, "epoch": 0.28046989720998533, "fcm_dpo/beta": 0.013516737148165703, "fcm_dpo/delta": -0.40941235423088074, "fcm_dpo/margin": 85.6933364868164, "fcm_dpo/q_t": 0.2823325991630554, "grad_norm": 29.902969360351562, "learning_rate": 4.5330506821893565e-07, "logits/chosen": -0.5110219717025757, "logits/rejected": -0.49608632922172546, "logps/chosen": -112.49159240722656, "logps/ref_chosen": -68.65887451171875, "logps/ref_rejected": -110.1396713256836, "logps/rejected": -239.66574096679688, "loss": 0.7371, "margin_dpo/margin_mean": 85.6933364868164, "margin_dpo/margin_std": 83.78256225585938, "step": 191 }, { "KL/chosen_KL_mean": -60.607242584228516, "KL/mean": -91.75163269042969, "KL/rejected_KL_mean": -122.89604187011719, "KL/std": 67.90768432617188, "epoch": 0.28193832599118945, "fcm_dpo/beta": 0.01330653764307499, "fcm_dpo/delta": -0.031592316925525665, "fcm_dpo/margin": 62.288795471191406, "fcm_dpo/q_t": 0.3421240448951721, "grad_norm": 32.041038513183594, "learning_rate": 4.5255555107119336e-07, "logits/chosen": -0.5030392408370972, "logits/rejected": -0.4946763217449188, "logps/chosen": -130.3341522216797, "logps/ref_chosen": -69.72691345214844, "logps/ref_rejected": -103.32135009765625, "logps/rejected": -226.21737670898438, "loss": 0.9197, "margin_dpo/margin_mean": 62.28879928588867, "margin_dpo/margin_std": 77.69465637207031, "step": 192 }, { "KL/chosen_KL_mean": -63.92502212524414, "KL/mean": -82.90367889404297, "KL/rejected_KL_mean": -101.88233947753906, "KL/std": 57.34717559814453, "epoch": 0.2834067547723935, "fcm_dpo/beta": 0.013927249237895012, "fcm_dpo/delta": 0.2825108766555786, "fcm_dpo/margin": 37.95730972290039, "fcm_dpo/q_t": 0.3886550962924957, "grad_norm": 43.13148498535156, "learning_rate": 4.5180069639630236e-07, "logits/chosen": -0.5010656118392944, "logits/rejected": -0.4694574475288391, "logps/chosen": -124.11552429199219, "logps/ref_chosen": -60.19049835205078, "logps/ref_rejected": -76.40755462646484, "logps/rejected": -178.28988647460938, "loss": 1.1345, "margin_dpo/margin_mean": 37.95730972290039, "margin_dpo/margin_std": 70.45993041992188, "step": 193 }, { "KL/chosen_KL_mean": -39.9882698059082, "KL/mean": -70.1005859375, "KL/rejected_KL_mean": -100.21292114257812, "KL/std": 56.194984436035156, "epoch": 0.28487518355359764, "fcm_dpo/beta": 0.013896675780415535, "fcm_dpo/delta": -0.040464356541633606, "fcm_dpo/margin": 60.22463607788086, "fcm_dpo/q_t": 0.3275321424007416, "grad_norm": 26.552078247070312, "learning_rate": 4.510405240853854e-07, "logits/chosen": -0.4238455891609192, "logits/rejected": -0.4025278389453888, "logps/chosen": -77.82864379882812, "logps/ref_chosen": -37.84037399291992, "logps/ref_rejected": -60.684783935546875, "logps/rejected": -160.897705078125, "loss": 0.8503, "margin_dpo/margin_mean": 60.224639892578125, "margin_dpo/margin_std": 58.84598922729492, "step": 194 }, { "KL/chosen_KL_mean": -66.05638122558594, "KL/mean": -98.05033874511719, "KL/rejected_KL_mean": -130.04428100585938, "KL/std": 64.62409210205078, "epoch": 0.28634361233480177, "fcm_dpo/beta": 0.013699124567210674, "fcm_dpo/delta": -0.0841585025191307, "fcm_dpo/margin": 63.98790740966797, "fcm_dpo/q_t": 0.32316848635673523, "grad_norm": 37.01463317871094, "learning_rate": 4.5027505416968985e-07, "logits/chosen": -0.4037872552871704, "logits/rejected": -0.4184109568595886, "logps/chosen": -120.94795227050781, "logps/ref_chosen": -54.891571044921875, "logps/ref_rejected": -96.77095794677734, "logps/rejected": -226.81524658203125, "loss": 0.8456, "margin_dpo/margin_mean": 63.98790740966797, "margin_dpo/margin_std": 66.32554626464844, "step": 195 }, { "KL/chosen_KL_mean": -52.29330825805664, "KL/mean": -89.10833740234375, "KL/rejected_KL_mean": -125.9233627319336, "KL/std": 72.76502990722656, "epoch": 0.2878120411160059, "fcm_dpo/beta": 0.013190150260925293, "fcm_dpo/delta": -0.19196948409080505, "fcm_dpo/margin": 73.63005828857422, "fcm_dpo/q_t": 0.3175709843635559, "grad_norm": 33.40938949584961, "learning_rate": 4.495043068200599e-07, "logits/chosen": -0.44675952196121216, "logits/rejected": -0.4224822521209717, "logps/chosen": -105.5385513305664, "logps/ref_chosen": -53.245243072509766, "logps/ref_rejected": -76.05294799804688, "logps/rejected": -201.976318359375, "loss": 0.8571, "margin_dpo/margin_mean": 73.63005065917969, "margin_dpo/margin_std": 82.35310363769531, "step": 196 }, { "KL/chosen_KL_mean": -53.17420959472656, "KL/mean": -82.42511749267578, "KL/rejected_KL_mean": -111.676025390625, "KL/std": 61.11585235595703, "epoch": 0.28928046989721, "fcm_dpo/beta": 0.013252440840005875, "fcm_dpo/delta": 0.02636050246655941, "fcm_dpo/margin": 58.50181198120117, "fcm_dpo/q_t": 0.34405672550201416, "grad_norm": 43.14238739013672, "learning_rate": 4.4872830234640493e-07, "logits/chosen": -0.3930720090866089, "logits/rejected": -0.38080155849456787, "logps/chosen": -113.59454345703125, "logps/ref_chosen": -60.42033386230469, "logps/ref_rejected": -77.20890808105469, "logps/rejected": -188.8849334716797, "loss": 0.9062, "margin_dpo/margin_mean": 58.50181198120117, "margin_dpo/margin_std": 66.8076400756836, "step": 197 }, { "KL/chosen_KL_mean": -57.16075134277344, "KL/mean": -91.8175277709961, "KL/rejected_KL_mean": -126.47431945800781, "KL/std": 73.49171447753906, "epoch": 0.2907488986784141, "fcm_dpo/beta": 0.013048326596617699, "fcm_dpo/delta": -0.1146821528673172, "fcm_dpo/margin": 69.31356048583984, "fcm_dpo/q_t": 0.3285406827926636, "grad_norm": 33.66745376586914, "learning_rate": 4.479470611971645e-07, "logits/chosen": -0.4547293186187744, "logits/rejected": -0.4567200243473053, "logps/chosen": -112.19693756103516, "logps/ref_chosen": -55.03618621826172, "logps/ref_rejected": -97.24325561523438, "logps/rejected": -223.7175750732422, "loss": 0.8665, "margin_dpo/margin_mean": 69.31356048583984, "margin_dpo/margin_std": 82.79634094238281, "step": 198 }, { "KL/chosen_KL_mean": -57.0550537109375, "KL/mean": -89.62132263183594, "KL/rejected_KL_mean": -122.18759155273438, "KL/std": 66.67041015625, "epoch": 0.2922173274596182, "fcm_dpo/beta": 0.012743506580591202, "fcm_dpo/delta": -0.03456920385360718, "fcm_dpo/margin": 65.13253021240234, "fcm_dpo/q_t": 0.33725377917289734, "grad_norm": 36.01947021484375, "learning_rate": 4.471606039587695e-07, "logits/chosen": -0.4379258155822754, "logits/rejected": -0.4226016104221344, "logps/chosen": -113.88388061523438, "logps/ref_chosen": -56.828826904296875, "logps/ref_rejected": -84.64820861816406, "logps/rejected": -206.83580017089844, "loss": 0.9154, "margin_dpo/margin_mean": 65.13253021240234, "margin_dpo/margin_std": 77.29293060302734, "step": 199 }, { "KL/chosen_KL_mean": -54.75391387939453, "KL/mean": -90.25743103027344, "KL/rejected_KL_mean": -125.76095581054688, "KL/std": 74.021484375, "epoch": 0.2936857562408223, "fcm_dpo/beta": 0.012570657767355442, "fcm_dpo/delta": -0.10279311239719391, "fcm_dpo/margin": 71.00704956054688, "fcm_dpo/q_t": 0.33011093735694885, "grad_norm": 34.09638977050781, "learning_rate": 4.4636895135509966e-07, "logits/chosen": -0.398794949054718, "logits/rejected": -0.3711370825767517, "logps/chosen": -107.82096862792969, "logps/ref_chosen": -53.06706237792969, "logps/ref_rejected": -80.60843658447266, "logps/rejected": -206.369384765625, "loss": 0.8994, "margin_dpo/margin_mean": 71.00704956054688, "margin_dpo/margin_std": 86.69010925292969, "step": 200 }, { "KL/chosen_KL_mean": -54.42976379394531, "KL/mean": -89.63890075683594, "KL/rejected_KL_mean": -124.8480224609375, "KL/std": 72.89269256591797, "epoch": 0.29515418502202645, "fcm_dpo/beta": 0.012411890551447868, "fcm_dpo/delta": -0.08116737008094788, "fcm_dpo/margin": 70.41826629638672, "fcm_dpo/q_t": 0.32952481508255005, "grad_norm": 35.047454833984375, "learning_rate": 4.455721242469372e-07, "logits/chosen": -0.4590557813644409, "logits/rejected": -0.4471399188041687, "logps/chosen": -129.8319854736328, "logps/ref_chosen": -75.4022216796875, "logps/ref_rejected": -114.80821990966797, "logps/rejected": -239.65625, "loss": 0.8873, "margin_dpo/margin_mean": 70.41826629638672, "margin_dpo/margin_std": 82.54559326171875, "step": 201 }, { "KL/chosen_KL_mean": -58.886390686035156, "KL/mean": -83.84607696533203, "KL/rejected_KL_mean": -108.8057632446289, "KL/std": 64.16035461425781, "epoch": 0.2966226138032305, "fcm_dpo/beta": 0.012727733701467514, "fcm_dpo/delta": 0.17515869438648224, "fcm_dpo/margin": 49.91937255859375, "fcm_dpo/q_t": 0.3769991397857666, "grad_norm": 33.56517791748047, "learning_rate": 4.4477014363141755e-07, "logits/chosen": -0.43319663405418396, "logits/rejected": -0.435769647359848, "logps/chosen": -108.98770904541016, "logps/ref_chosen": -50.101318359375, "logps/ref_rejected": -86.98503112792969, "logps/rejected": -195.79080200195312, "loss": 1.053, "margin_dpo/margin_mean": 49.919376373291016, "margin_dpo/margin_std": 78.88993835449219, "step": 202 }, { "KL/chosen_KL_mean": -55.39906311035156, "KL/mean": -85.28140258789062, "KL/rejected_KL_mean": -115.16374206542969, "KL/std": 64.33355712890625, "epoch": 0.29809104258443464, "fcm_dpo/beta": 0.012859424576163292, "fcm_dpo/delta": 0.03409314155578613, "fcm_dpo/margin": 59.764678955078125, "fcm_dpo/q_t": 0.34479662775993347, "grad_norm": 36.9451904296875, "learning_rate": 4.439630306414758e-07, "logits/chosen": -0.4340917766094208, "logits/rejected": -0.4139591157436371, "logps/chosen": -116.00875854492188, "logps/ref_chosen": -60.60969543457031, "logps/ref_rejected": -85.89596557617188, "logps/rejected": -201.05970764160156, "loss": 0.9075, "margin_dpo/margin_mean": 59.764678955078125, "margin_dpo/margin_std": 68.13716125488281, "step": 203 }, { "KL/chosen_KL_mean": -66.02056121826172, "KL/mean": -94.26062774658203, "KL/rejected_KL_mean": -122.50070190429688, "KL/std": 72.59966278076172, "epoch": 0.29955947136563876, "fcm_dpo/beta": 0.013025782071053982, "fcm_dpo/delta": 0.06942086666822433, "fcm_dpo/margin": 56.480140686035156, "fcm_dpo/q_t": 0.3637349605560303, "grad_norm": 41.804168701171875, "learning_rate": 4.431508065452897e-07, "logits/chosen": -0.4650980234146118, "logits/rejected": -0.42012155055999756, "logps/chosen": -146.18551635742188, "logps/ref_chosen": -80.16496276855469, "logps/ref_rejected": -87.69590759277344, "logps/rejected": -210.1966094970703, "loss": 0.9857, "margin_dpo/margin_mean": 56.480140686035156, "margin_dpo/margin_std": 81.36882019042969, "step": 204 }, { "KL/chosen_KL_mean": -67.76717376708984, "KL/mean": -103.71720886230469, "KL/rejected_KL_mean": -139.667236328125, "KL/std": 76.89108276367188, "epoch": 0.3010279001468429, "fcm_dpo/beta": 0.012646548449993134, "fcm_dpo/delta": -0.1265978217124939, "fcm_dpo/margin": 71.9000473022461, "fcm_dpo/q_t": 0.32160481810569763, "grad_norm": 31.984642028808594, "learning_rate": 4.4233349274571974e-07, "logits/chosen": -0.40963441133499146, "logits/rejected": -0.3742339015007019, "logps/chosen": -127.15190887451172, "logps/ref_chosen": -59.384735107421875, "logps/ref_rejected": -85.12505340576172, "logps/rejected": -224.7922821044922, "loss": 0.898, "margin_dpo/margin_mean": 71.9000473022461, "margin_dpo/margin_std": 81.00496673583984, "step": 205 }, { "KL/chosen_KL_mean": -65.09518432617188, "KL/mean": -104.39747619628906, "KL/rejected_KL_mean": -143.6997833251953, "KL/std": 76.27721405029297, "epoch": 0.302496328928047, "fcm_dpo/beta": 0.01231662929058075, "fcm_dpo/delta": -0.18836334347724915, "fcm_dpo/margin": 78.60458374023438, "fcm_dpo/q_t": 0.3099524676799774, "grad_norm": 42.828025817871094, "learning_rate": 4.415111107797445e-07, "logits/chosen": -0.39149826765060425, "logits/rejected": -0.3944673538208008, "logps/chosen": -112.0596923828125, "logps/ref_chosen": -46.964500427246094, "logps/ref_rejected": -98.9534912109375, "logps/rejected": -242.6532745361328, "loss": 0.8418, "margin_dpo/margin_mean": 78.60458374023438, "margin_dpo/margin_std": 83.66462707519531, "step": 206 }, { "KL/chosen_KL_mean": -51.632415771484375, "KL/mean": -96.7965087890625, "KL/rejected_KL_mean": -141.96058654785156, "KL/std": 84.35016632080078, "epoch": 0.3039647577092511, "fcm_dpo/beta": 0.011775311082601547, "fcm_dpo/delta": -0.2946253716945648, "fcm_dpo/margin": 90.32818603515625, "fcm_dpo/q_t": 0.30258244276046753, "grad_norm": 29.97208595275879, "learning_rate": 4.4068368231789365e-07, "logits/chosen": -0.4526169002056122, "logits/rejected": -0.42235511541366577, "logps/chosen": -107.68867492675781, "logps/ref_chosen": -56.05625915527344, "logps/ref_rejected": -84.44779968261719, "logps/rejected": -226.40838623046875, "loss": 0.7953, "margin_dpo/margin_mean": 90.32817840576172, "margin_dpo/margin_std": 96.11925506591797, "step": 207 }, { "KL/chosen_KL_mean": -75.16487121582031, "KL/mean": -111.59002685546875, "KL/rejected_KL_mean": -148.01516723632812, "KL/std": 76.63873291015625, "epoch": 0.3054331864904552, "fcm_dpo/beta": 0.011480957269668579, "fcm_dpo/delta": -0.03990346938371658, "fcm_dpo/margin": 72.85030364990234, "fcm_dpo/q_t": 0.33449286222457886, "grad_norm": 32.494258880615234, "learning_rate": 4.398512291636768e-07, "logits/chosen": -0.4866720736026764, "logits/rejected": -0.466644287109375, "logps/chosen": -142.23248291015625, "logps/ref_chosen": -67.06761169433594, "logps/ref_rejected": -94.28689575195312, "logps/rejected": -242.30206298828125, "loss": 0.9049, "margin_dpo/margin_mean": 72.85029602050781, "margin_dpo/margin_std": 86.24763488769531, "step": 208 }, { "KL/chosen_KL_mean": -63.27362060546875, "KL/mean": -94.08541107177734, "KL/rejected_KL_mean": -124.89720916748047, "KL/std": 64.80196380615234, "epoch": 0.3069016152716593, "fcm_dpo/beta": 0.011625498533248901, "fcm_dpo/delta": 0.09011469781398773, "fcm_dpo/margin": 61.62358093261719, "fcm_dpo/q_t": 0.35670265555381775, "grad_norm": 35.18954086303711, "learning_rate": 4.3901377325300857e-07, "logits/chosen": -0.4050047993659973, "logits/rejected": -0.390215665102005, "logps/chosen": -119.45531463623047, "logps/ref_chosen": -56.18169403076172, "logps/ref_rejected": -80.94152069091797, "logps/rejected": -205.83872985839844, "loss": 0.9732, "margin_dpo/margin_mean": 61.62357711791992, "margin_dpo/margin_std": 79.23416137695312, "step": 209 }, { "KL/chosen_KL_mean": -55.82368469238281, "KL/mean": -89.87747955322266, "KL/rejected_KL_mean": -123.9312744140625, "KL/std": 65.2939682006836, "epoch": 0.30837004405286345, "fcm_dpo/beta": 0.011724742129445076, "fcm_dpo/delta": 0.0014328230172395706, "fcm_dpo/margin": 68.10758972167969, "fcm_dpo/q_t": 0.3393843472003937, "grad_norm": 25.020299911499023, "learning_rate": 4.381713366536311e-07, "logits/chosen": -0.4426725208759308, "logits/rejected": -0.42541852593421936, "logps/chosen": -102.19551086425781, "logps/ref_chosen": -46.371822357177734, "logps/ref_rejected": -76.68162536621094, "logps/rejected": -200.61288452148438, "loss": 0.8903, "margin_dpo/margin_mean": 68.10758972167969, "margin_dpo/margin_std": 74.40350341796875, "step": 210 }, { "KL/chosen_KL_mean": -80.90629577636719, "KL/mean": -109.43853759765625, "KL/rejected_KL_mean": -137.9707794189453, "KL/std": 67.03673553466797, "epoch": 0.30983847283406757, "fcm_dpo/beta": 0.011907186359167099, "fcm_dpo/delta": 0.1294347047805786, "fcm_dpo/margin": 57.064483642578125, "fcm_dpo/q_t": 0.36575454473495483, "grad_norm": 36.52604675292969, "learning_rate": 4.373239415645323e-07, "logits/chosen": -0.48629775643348694, "logits/rejected": -0.43736910820007324, "logps/chosen": -159.83865356445312, "logps/ref_chosen": -78.93235778808594, "logps/ref_rejected": -86.82098388671875, "logps/rejected": -224.79176330566406, "loss": 0.9892, "margin_dpo/margin_mean": 57.064483642578125, "margin_dpo/margin_std": 76.28172302246094, "step": 211 }, { "KL/chosen_KL_mean": -65.99049377441406, "KL/mean": -107.37837982177734, "KL/rejected_KL_mean": -148.76626586914062, "KL/std": 81.93545532226562, "epoch": 0.31130690161527164, "fcm_dpo/beta": 0.011535340920090675, "fcm_dpo/delta": -0.17738127708435059, "fcm_dpo/margin": 82.77577209472656, "fcm_dpo/q_t": 0.31630581617355347, "grad_norm": 26.731220245361328, "learning_rate": 4.3647161031536086e-07, "logits/chosen": -0.49336063861846924, "logits/rejected": -0.48159685730934143, "logps/chosen": -124.18750762939453, "logps/ref_chosen": -58.19701385498047, "logps/ref_rejected": -103.05785369873047, "logps/rejected": -251.82412719726562, "loss": 0.8567, "margin_dpo/margin_mean": 82.77577209472656, "margin_dpo/margin_std": 87.4062728881836, "step": 212 }, { "KL/chosen_KL_mean": -59.183773040771484, "KL/mean": -100.06889343261719, "KL/rejected_KL_mean": -140.95401000976562, "KL/std": 71.30158996582031, "epoch": 0.31277533039647576, "fcm_dpo/beta": 0.011260983534157276, "fcm_dpo/delta": -0.13558340072631836, "fcm_dpo/margin": 81.77023315429688, "fcm_dpo/q_t": 0.316591739654541, "grad_norm": 29.752309799194336, "learning_rate": 4.3561436536583774e-07, "logits/chosen": -0.4815593957901001, "logits/rejected": -0.443857342004776, "logps/chosen": -126.69648742675781, "logps/ref_chosen": -67.51271057128906, "logps/ref_rejected": -93.91471862792969, "logps/rejected": -234.8687286376953, "loss": 0.843, "margin_dpo/margin_mean": 81.77023315429688, "margin_dpo/margin_std": 84.07875061035156, "step": 213 }, { "KL/chosen_KL_mean": -57.820125579833984, "KL/mean": -92.26338195800781, "KL/rejected_KL_mean": -126.70663452148438, "KL/std": 73.20215606689453, "epoch": 0.3142437591776799, "fcm_dpo/beta": 0.011330001056194305, "fcm_dpo/delta": 0.02105110138654709, "fcm_dpo/margin": 68.88651275634766, "fcm_dpo/q_t": 0.3464926779270172, "grad_norm": 35.64681625366211, "learning_rate": 4.3475222930516473e-07, "logits/chosen": -0.43472471833229065, "logits/rejected": -0.4402396082878113, "logps/chosen": -99.42501831054688, "logps/ref_chosen": -41.604888916015625, "logps/ref_rejected": -77.51741027832031, "logps/rejected": -204.22406005859375, "loss": 0.9323, "margin_dpo/margin_mean": 68.88650512695312, "margin_dpo/margin_std": 85.51469421386719, "step": 214 }, { "KL/chosen_KL_mean": -63.354827880859375, "KL/mean": -100.20165252685547, "KL/rejected_KL_mean": -137.04847717285156, "KL/std": 71.63882446289062, "epoch": 0.315712187958884, "fcm_dpo/beta": 0.011217828840017319, "fcm_dpo/delta": -0.02969864010810852, "fcm_dpo/margin": 73.69364929199219, "fcm_dpo/q_t": 0.33209890127182007, "grad_norm": 32.47001647949219, "learning_rate": 4.3388522485142885e-07, "logits/chosen": -0.4393133521080017, "logits/rejected": -0.42265331745147705, "logps/chosen": -116.63409423828125, "logps/ref_chosen": -53.279266357421875, "logps/ref_rejected": -89.96464538574219, "logps/rejected": -227.01312255859375, "loss": 0.8721, "margin_dpo/margin_mean": 73.69364929199219, "margin_dpo/margin_std": 77.87715911865234, "step": 215 }, { "KL/chosen_KL_mean": -65.12640380859375, "KL/mean": -104.15386199951172, "KL/rejected_KL_mean": -143.18133544921875, "KL/std": 75.26974487304688, "epoch": 0.31718061674008813, "fcm_dpo/beta": 0.011194197461009026, "fcm_dpo/delta": -0.08120834827423096, "fcm_dpo/margin": 78.054931640625, "fcm_dpo/q_t": 0.3296348750591278, "grad_norm": 33.180416107177734, "learning_rate": 4.330133748510036e-07, "logits/chosen": -0.46606436371803284, "logits/rejected": -0.44526204466819763, "logps/chosen": -114.01419067382812, "logps/ref_chosen": -48.887794494628906, "logps/ref_rejected": -77.19892883300781, "logps/rejected": -220.3802490234375, "loss": 0.8957, "margin_dpo/margin_mean": 78.054931640625, "margin_dpo/margin_std": 89.83686828613281, "step": 216 }, { "KL/chosen_KL_mean": -61.40077209472656, "KL/mean": -104.93330383300781, "KL/rejected_KL_mean": -148.4658203125, "KL/std": 74.56744384765625, "epoch": 0.3186490455212922, "fcm_dpo/beta": 0.010798503644764423, "fcm_dpo/delta": -0.15515443682670593, "fcm_dpo/margin": 87.06504821777344, "fcm_dpo/q_t": 0.3086259365081787, "grad_norm": 29.548845291137695, "learning_rate": 4.3213670227794757e-07, "logits/chosen": -0.461231529712677, "logits/rejected": -0.4476476311683655, "logps/chosen": -111.24607849121094, "logps/ref_chosen": -49.845306396484375, "logps/ref_rejected": -100.07832336425781, "logps/rejected": -248.54415893554688, "loss": 0.7976, "margin_dpo/margin_mean": 87.06504821777344, "margin_dpo/margin_std": 80.49182891845703, "step": 217 }, { "KL/chosen_KL_mean": -65.54306030273438, "KL/mean": -100.61727142333984, "KL/rejected_KL_mean": -135.6914825439453, "KL/std": 72.58134460449219, "epoch": 0.3201174743024963, "fcm_dpo/beta": 0.010796687565743923, "fcm_dpo/delta": 0.046147942543029785, "fcm_dpo/margin": 70.14842224121094, "fcm_dpo/q_t": 0.3473072052001953, "grad_norm": 28.300230026245117, "learning_rate": 4.3125523023339815e-07, "logits/chosen": -0.4414158761501312, "logits/rejected": -0.42348459362983704, "logps/chosen": -124.11973571777344, "logps/ref_chosen": -58.576683044433594, "logps/ref_rejected": -87.84639739990234, "logps/rejected": -223.53787231445312, "loss": 0.9229, "margin_dpo/margin_mean": 70.14842224121094, "margin_dpo/margin_std": 81.61410522460938, "step": 218 }, { "KL/chosen_KL_mean": -69.83723449707031, "KL/mean": -99.55513000488281, "KL/rejected_KL_mean": -129.27301025390625, "KL/std": 70.54728698730469, "epoch": 0.32158590308370044, "fcm_dpo/beta": 0.011129364371299744, "fcm_dpo/delta": 0.14702974259853363, "fcm_dpo/margin": 59.4357795715332, "fcm_dpo/q_t": 0.3667854070663452, "grad_norm": 33.41205596923828, "learning_rate": 4.303689819449636e-07, "logits/chosen": -0.470586359500885, "logits/rejected": -0.45085620880126953, "logps/chosen": -130.9210968017578, "logps/ref_chosen": -61.083858489990234, "logps/ref_rejected": -85.83042907714844, "logps/rejected": -215.1034393310547, "loss": 0.9974, "margin_dpo/margin_mean": 59.43577575683594, "margin_dpo/margin_std": 81.409912109375, "step": 219 }, { "KL/chosen_KL_mean": -76.43949890136719, "KL/mean": -103.16355895996094, "KL/rejected_KL_mean": -129.88760375976562, "KL/std": 62.09259796142578, "epoch": 0.32305433186490456, "fcm_dpo/beta": 0.011432506144046783, "fcm_dpo/delta": 0.20173068344593048, "fcm_dpo/margin": 53.44810485839844, "fcm_dpo/q_t": 0.3675389885902405, "grad_norm": 29.649065017700195, "learning_rate": 4.2947798076611047e-07, "logits/chosen": -0.46881765127182007, "logits/rejected": -0.4399240016937256, "logps/chosen": -146.47079467773438, "logps/ref_chosen": -70.03128051757812, "logps/ref_rejected": -87.68551635742188, "logps/rejected": -217.5731201171875, "loss": 0.9669, "margin_dpo/margin_mean": 53.4481086730957, "margin_dpo/margin_std": 57.649662017822266, "step": 220 }, { "KL/chosen_KL_mean": -77.00071716308594, "KL/mean": -127.14787292480469, "KL/rejected_KL_mean": -177.29502868652344, "KL/std": 87.18041229248047, "epoch": 0.3245227606461087, "fcm_dpo/beta": 0.011109141632914543, "fcm_dpo/delta": -0.3521580696105957, "fcm_dpo/margin": 100.2943115234375, "fcm_dpo/q_t": 0.2772984504699707, "grad_norm": 28.87704086303711, "learning_rate": 4.285822501755485e-07, "logits/chosen": -0.4105208218097687, "logits/rejected": -0.416123628616333, "logps/chosen": -129.15542602539062, "logps/ref_chosen": -52.15470886230469, "logps/ref_rejected": -106.46768188476562, "logps/rejected": -283.7626953125, "loss": 0.7239, "margin_dpo/margin_mean": 100.2943115234375, "margin_dpo/margin_std": 82.78035736083984, "step": 221 }, { "KL/chosen_KL_mean": -80.87545013427734, "KL/mean": -119.49418640136719, "KL/rejected_KL_mean": -158.1129150390625, "KL/std": 77.89344024658203, "epoch": 0.32599118942731276, "fcm_dpo/beta": 0.010816777125000954, "fcm_dpo/delta": -0.0389985591173172, "fcm_dpo/margin": 77.23746490478516, "fcm_dpo/q_t": 0.33166736364364624, "grad_norm": 46.94334411621094, "learning_rate": 4.276818137766118e-07, "logits/chosen": -0.4480839669704437, "logits/rejected": -0.4343772232532501, "logps/chosen": -141.8465576171875, "logps/ref_chosen": -60.971099853515625, "logps/ref_rejected": -100.00115203857422, "logps/rejected": -258.11407470703125, "loss": 0.8845, "margin_dpo/margin_mean": 77.23746490478516, "margin_dpo/margin_std": 84.92472839355469, "step": 222 }, { "KL/chosen_KL_mean": -88.08724212646484, "KL/mean": -121.95101928710938, "KL/rejected_KL_mean": -155.81478881835938, "KL/std": 75.61756896972656, "epoch": 0.3274596182085169, "fcm_dpo/beta": 0.010851439088582993, "fcm_dpo/delta": 0.07023735344409943, "fcm_dpo/margin": 67.72756958007812, "fcm_dpo/q_t": 0.3549031615257263, "grad_norm": 38.04423522949219, "learning_rate": 4.2677669529663686e-07, "logits/chosen": -0.382046639919281, "logits/rejected": -0.36027002334594727, "logps/chosen": -140.72781372070312, "logps/ref_chosen": -52.64057540893555, "logps/ref_rejected": -82.82502746582031, "logps/rejected": -238.63983154296875, "loss": 1.002, "margin_dpo/margin_mean": 67.72756958007812, "margin_dpo/margin_std": 95.45079040527344, "step": 223 }, { "KL/chosen_KL_mean": -76.32827758789062, "KL/mean": -117.27737426757812, "KL/rejected_KL_mean": -158.2264862060547, "KL/std": 86.974365234375, "epoch": 0.328928046989721, "fcm_dpo/beta": 0.010628387331962585, "fcm_dpo/delta": -0.0815625712275505, "fcm_dpo/margin": 81.89820861816406, "fcm_dpo/q_t": 0.3337979316711426, "grad_norm": 36.384342193603516, "learning_rate": 4.2586691858633747e-07, "logits/chosen": -0.3687441349029541, "logits/rejected": -0.3405214548110962, "logps/chosen": -124.9236831665039, "logps/ref_chosen": -48.59541320800781, "logps/ref_rejected": -77.11648559570312, "logps/rejected": -235.3429718017578, "loss": 0.9033, "margin_dpo/margin_mean": 81.89820861816406, "margin_dpo/margin_std": 97.37171936035156, "step": 224 }, { "KL/chosen_KL_mean": -89.83382415771484, "KL/mean": -139.490966796875, "KL/rejected_KL_mean": -189.14810180664062, "KL/std": 89.47193908691406, "epoch": 0.3303964757709251, "fcm_dpo/beta": 0.010304208844900131, "fcm_dpo/delta": -0.2496224194765091, "fcm_dpo/margin": 99.31427001953125, "fcm_dpo/q_t": 0.3028189539909363, "grad_norm": 30.707063674926758, "learning_rate": 4.249525076191759e-07, "logits/chosen": -0.4086596965789795, "logits/rejected": -0.3914273977279663, "logps/chosen": -147.83428955078125, "logps/ref_chosen": -58.000465393066406, "logps/ref_rejected": -99.90291595458984, "logps/rejected": -289.051025390625, "loss": 0.8152, "margin_dpo/margin_mean": 99.31427001953125, "margin_dpo/margin_std": 103.35362243652344, "step": 225 }, { "KL/chosen_KL_mean": -76.8957290649414, "KL/mean": -115.76048278808594, "KL/rejected_KL_mean": -154.625244140625, "KL/std": 85.1742935180664, "epoch": 0.33186490455212925, "fcm_dpo/beta": 0.010075679048895836, "fcm_dpo/delta": 0.014316320419311523, "fcm_dpo/margin": 77.7295150756836, "fcm_dpo/q_t": 0.34836602210998535, "grad_norm": 35.826454162597656, "learning_rate": 4.2403348649073167e-07, "logits/chosen": -0.3905969262123108, "logits/rejected": -0.3390669822692871, "logps/chosen": -135.79452514648438, "logps/ref_chosen": -58.898799896240234, "logps/ref_rejected": -78.68775939941406, "logps/rejected": -233.31300354003906, "loss": 0.9375, "margin_dpo/margin_mean": 77.7295150756836, "margin_dpo/margin_std": 94.68783569335938, "step": 226 }, { "KL/chosen_KL_mean": -88.30928802490234, "KL/mean": -137.89309692382812, "KL/rejected_KL_mean": -187.4768829345703, "KL/std": 100.16815948486328, "epoch": 0.3333333333333333, "fcm_dpo/beta": 0.009845739230513573, "fcm_dpo/delta": -0.19868168234825134, "fcm_dpo/margin": 99.1676025390625, "fcm_dpo/q_t": 0.3060121238231659, "grad_norm": 42.023475646972656, "learning_rate": 4.2310987941806615e-07, "logits/chosen": -0.41084548830986023, "logits/rejected": -0.3941611647605896, "logps/chosen": -147.3814697265625, "logps/ref_chosen": -59.072181701660156, "logps/ref_rejected": -99.41236877441406, "logps/rejected": -286.8892517089844, "loss": 0.8302, "margin_dpo/margin_mean": 99.1676025390625, "margin_dpo/margin_std": 103.44757080078125, "step": 227 }, { "KL/chosen_KL_mean": -86.19977569580078, "KL/mean": -121.89463806152344, "KL/rejected_KL_mean": -157.5894775390625, "KL/std": 78.76786804199219, "epoch": 0.33480176211453744, "fcm_dpo/beta": 0.009987346827983856, "fcm_dpo/delta": 0.09334658831357956, "fcm_dpo/margin": 71.38971710205078, "fcm_dpo/q_t": 0.35477495193481445, "grad_norm": 29.949947357177734, "learning_rate": 4.2218171073908463e-07, "logits/chosen": -0.39673376083374023, "logits/rejected": -0.37368690967559814, "logps/chosen": -152.091064453125, "logps/ref_chosen": -65.89128875732422, "logps/ref_rejected": -91.04875183105469, "logps/rejected": -248.6382293701172, "loss": 0.961, "margin_dpo/margin_mean": 71.38971710205078, "margin_dpo/margin_std": 89.57862091064453, "step": 228 }, { "KL/chosen_KL_mean": -86.62499237060547, "KL/mean": -127.21802520751953, "KL/rejected_KL_mean": -167.81106567382812, "KL/std": 92.57134246826172, "epoch": 0.33627019089574156, "fcm_dpo/beta": 0.010008785873651505, "fcm_dpo/delta": -0.013723690994083881, "fcm_dpo/margin": 81.18606567382812, "fcm_dpo/q_t": 0.33619070053100586, "grad_norm": 38.73611068725586, "learning_rate": 4.212490049118951e-07, "logits/chosen": -0.49296534061431885, "logits/rejected": -0.46168243885040283, "logps/chosen": -157.33135986328125, "logps/ref_chosen": -70.70637512207031, "logps/ref_rejected": -84.52741241455078, "logps/rejected": -252.33847045898438, "loss": 0.9191, "margin_dpo/margin_mean": 81.18606567382812, "margin_dpo/margin_std": 97.13127899169922, "step": 229 }, { "KL/chosen_KL_mean": -73.65254211425781, "KL/mean": -126.47062683105469, "KL/rejected_KL_mean": -179.28871154785156, "KL/std": 91.7388916015625, "epoch": 0.3377386196769457, "fcm_dpo/beta": 0.009593733586370945, "fcm_dpo/delta": -0.23911604285240173, "fcm_dpo/margin": 105.63615417480469, "fcm_dpo/q_t": 0.29790061712265015, "grad_norm": 28.22924041748047, "learning_rate": 4.203117865141635e-07, "logits/chosen": -0.3197930157184601, "logits/rejected": -0.3206769824028015, "logps/chosen": -112.93455505371094, "logps/ref_chosen": -39.282005310058594, "logps/ref_rejected": -85.62191009521484, "logps/rejected": -264.9106140136719, "loss": 0.793, "margin_dpo/margin_mean": 105.63615417480469, "margin_dpo/margin_std": 102.19107055664062, "step": 230 }, { "KL/chosen_KL_mean": -75.81376647949219, "KL/mean": -114.97604370117188, "KL/rejected_KL_mean": -154.13833618164062, "KL/std": 78.87530517578125, "epoch": 0.3392070484581498, "fcm_dpo/beta": 0.009577736258506775, "fcm_dpo/delta": 0.05389944836497307, "fcm_dpo/margin": 78.32456970214844, "fcm_dpo/q_t": 0.34839433431625366, "grad_norm": 44.196414947509766, "learning_rate": 4.1937008024246625e-07, "logits/chosen": -0.41479164361953735, "logits/rejected": -0.38350310921669006, "logps/chosen": -139.0902099609375, "logps/ref_chosen": -63.27644348144531, "logps/ref_rejected": -74.1239013671875, "logps/rejected": -228.26223754882812, "loss": 0.9068, "margin_dpo/margin_mean": 78.32456970214844, "margin_dpo/margin_std": 87.84222412109375, "step": 231 }, { "KL/chosen_KL_mean": -92.11940002441406, "KL/mean": -123.58616638183594, "KL/rejected_KL_mean": -155.05294799804688, "KL/std": 88.39076232910156, "epoch": 0.3406754772393539, "fcm_dpo/beta": 0.009844740852713585, "fcm_dpo/delta": 0.19277611374855042, "fcm_dpo/margin": 62.93353271484375, "fcm_dpo/q_t": 0.3812215328216553, "grad_norm": 31.937786102294922, "learning_rate": 4.1842391091163933e-07, "logits/chosen": -0.4062514305114746, "logits/rejected": -0.37542420625686646, "logps/chosen": -162.8681640625, "logps/ref_chosen": -70.74876403808594, "logps/ref_rejected": -83.97706604003906, "logps/rejected": -239.03001403808594, "loss": 1.0336, "margin_dpo/margin_mean": 62.93353271484375, "margin_dpo/margin_std": 95.02970886230469, "step": 232 }, { "KL/chosen_KL_mean": -80.99494171142578, "KL/mean": -128.9691162109375, "KL/rejected_KL_mean": -176.94326782226562, "KL/std": 97.7330322265625, "epoch": 0.342143906020558, "fcm_dpo/beta": 0.009784163907170296, "fcm_dpo/delta": -0.15304714441299438, "fcm_dpo/margin": 95.94833374023438, "fcm_dpo/q_t": 0.33123689889907837, "grad_norm": 36.1855354309082, "learning_rate": 4.174733034541245e-07, "logits/chosen": -0.4026934802532196, "logits/rejected": -0.39757007360458374, "logps/chosen": -135.87786865234375, "logps/ref_chosen": -54.8829345703125, "logps/ref_rejected": -107.4800796508789, "logps/rejected": -284.42333984375, "loss": 0.9175, "margin_dpo/margin_mean": 95.94833374023438, "margin_dpo/margin_std": 123.36692810058594, "step": 233 }, { "KL/chosen_KL_mean": -76.04293823242188, "KL/mean": -124.41976928710938, "KL/rejected_KL_mean": -172.7965850830078, "KL/std": 83.86376953125, "epoch": 0.3436123348017621, "fcm_dpo/beta": 0.009408731944859028, "fcm_dpo/delta": -0.12575244903564453, "fcm_dpo/margin": 96.75364685058594, "fcm_dpo/q_t": 0.3136642873287201, "grad_norm": 33.3326530456543, "learning_rate": 4.165182829193126e-07, "logits/chosen": -0.36011114716529846, "logits/rejected": -0.3862539529800415, "logps/chosen": -120.13746643066406, "logps/ref_chosen": -44.094520568847656, "logps/ref_rejected": -100.00663757324219, "logps/rejected": -272.80322265625, "loss": 0.8233, "margin_dpo/margin_mean": 96.75364685058594, "margin_dpo/margin_std": 91.30296325683594, "step": 234 }, { "KL/chosen_KL_mean": -94.54927062988281, "KL/mean": -127.44790649414062, "KL/rejected_KL_mean": -160.34654235839844, "KL/std": 81.41526794433594, "epoch": 0.34508076358296624, "fcm_dpo/beta": 0.009737150743603706, "fcm_dpo/delta": 0.16920268535614014, "fcm_dpo/margin": 65.79728698730469, "fcm_dpo/q_t": 0.3712586462497711, "grad_norm": 35.418949127197266, "learning_rate": 4.1555887447288255e-07, "logits/chosen": -0.4128161072731018, "logits/rejected": -0.38685333728790283, "logps/chosen": -156.7871856689453, "logps/ref_chosen": -62.237911224365234, "logps/ref_rejected": -90.39506530761719, "logps/rejected": -250.74160766601562, "loss": 1.0157, "margin_dpo/margin_mean": 65.79728698730469, "margin_dpo/margin_std": 92.96914672851562, "step": 235 }, { "KL/chosen_KL_mean": -71.0826187133789, "KL/mean": -118.72409057617188, "KL/rejected_KL_mean": -166.3655548095703, "KL/std": 86.84884643554688, "epoch": 0.3465491923641703, "fcm_dpo/beta": 0.009598923847079277, "fcm_dpo/delta": -0.12620067596435547, "fcm_dpo/margin": 95.28294372558594, "fcm_dpo/q_t": 0.3114420175552368, "grad_norm": 42.24114990234375, "learning_rate": 4.1459510339613946e-07, "logits/chosen": -0.34095823764801025, "logits/rejected": -0.34158775210380554, "logps/chosen": -120.42398071289062, "logps/ref_chosen": -49.34136199951172, "logps/ref_rejected": -103.51162719726562, "logps/rejected": -269.877197265625, "loss": 0.8009, "margin_dpo/margin_mean": 95.2829360961914, "margin_dpo/margin_std": 85.3992691040039, "step": 236 }, { "KL/chosen_KL_mean": -89.83419799804688, "KL/mean": -129.98583984375, "KL/rejected_KL_mean": -170.13746643066406, "KL/std": 84.88386535644531, "epoch": 0.34801762114537443, "fcm_dpo/beta": 0.009580841287970543, "fcm_dpo/delta": 0.0331917479634285, "fcm_dpo/margin": 80.30326843261719, "fcm_dpo/q_t": 0.3448735475540161, "grad_norm": 36.753971099853516, "learning_rate": 4.136269950853473e-07, "logits/chosen": -0.3638402223587036, "logits/rejected": -0.3566875457763672, "logps/chosen": -144.0023193359375, "logps/ref_chosen": -54.168121337890625, "logps/ref_rejected": -94.78036499023438, "logps/rejected": -264.9178466796875, "loss": 0.9228, "margin_dpo/margin_mean": 80.30326843261719, "margin_dpo/margin_std": 94.76294708251953, "step": 237 }, { "KL/chosen_KL_mean": -87.62262725830078, "KL/mean": -125.47653198242188, "KL/rejected_KL_mean": -163.3304443359375, "KL/std": 84.19970703125, "epoch": 0.34948604992657856, "fcm_dpo/beta": 0.009638959541916847, "fcm_dpo/delta": 0.0752006322145462, "fcm_dpo/margin": 75.70780944824219, "fcm_dpo/q_t": 0.35555845499038696, "grad_norm": 36.07196044921875, "learning_rate": 4.126545750510605e-07, "logits/chosen": -0.33357322216033936, "logits/rejected": -0.3504067659378052, "logps/chosen": -141.5957489013672, "logps/ref_chosen": -53.973121643066406, "logps/ref_rejected": -89.41795349121094, "logps/rejected": -252.74838256835938, "loss": 0.94, "margin_dpo/margin_mean": 75.70780944824219, "margin_dpo/margin_std": 92.58589172363281, "step": 238 }, { "KL/chosen_KL_mean": -94.18663024902344, "KL/mean": -137.20260620117188, "KL/rejected_KL_mean": -180.21859741210938, "KL/std": 83.78192138671875, "epoch": 0.3509544787077827, "fcm_dpo/beta": 0.009586915373802185, "fcm_dpo/delta": -0.03049781173467636, "fcm_dpo/margin": 86.03196716308594, "fcm_dpo/q_t": 0.3317285478115082, "grad_norm": 34.0923957824707, "learning_rate": 4.116778689174514e-07, "logits/chosen": -0.3576841354370117, "logits/rejected": -0.33470335602760315, "logps/chosen": -152.28445434570312, "logps/ref_chosen": -58.09782409667969, "logps/ref_rejected": -93.59294128417969, "logps/rejected": -273.8115234375, "loss": 0.8762, "margin_dpo/margin_mean": 86.03196716308594, "margin_dpo/margin_std": 86.97271728515625, "step": 239 }, { "KL/chosen_KL_mean": -98.12992858886719, "KL/mean": -137.9344482421875, "KL/rejected_KL_mean": -177.7389678955078, "KL/std": 91.66488647460938, "epoch": 0.3524229074889868, "fcm_dpo/beta": 0.009678095579147339, "fcm_dpo/delta": 0.031485963612794876, "fcm_dpo/margin": 79.60904693603516, "fcm_dpo/q_t": 0.3433542251586914, "grad_norm": 35.59693908691406, "learning_rate": 4.106969024216348e-07, "logits/chosen": -0.34670311212539673, "logits/rejected": -0.32258307933807373, "logps/chosen": -158.74441528320312, "logps/ref_chosen": -60.6144905090332, "logps/ref_rejected": -74.1185302734375, "logps/rejected": -251.8574981689453, "loss": 0.9304, "margin_dpo/margin_mean": 79.60903930664062, "margin_dpo/margin_std": 93.16104125976562, "step": 240 }, { "KL/chosen_KL_mean": -81.70965576171875, "KL/mean": -138.0463409423828, "KL/rejected_KL_mean": -194.38299560546875, "KL/std": 101.58145141601562, "epoch": 0.35389133627019087, "fcm_dpo/beta": 0.009372754022479057, "fcm_dpo/delta": -0.28571805357933044, "fcm_dpo/margin": 112.67334747314453, "fcm_dpo/q_t": 0.29759377241134644, "grad_norm": 34.78485870361328, "learning_rate": 4.097117014129903e-07, "logits/chosen": -0.3599684238433838, "logits/rejected": -0.3252626061439514, "logps/chosen": -147.80072021484375, "logps/ref_chosen": -66.091064453125, "logps/ref_rejected": -88.06088256835938, "logps/rejected": -282.44390869140625, "loss": 0.7835, "margin_dpo/margin_mean": 112.67334747314453, "margin_dpo/margin_std": 113.32879638671875, "step": 241 }, { "KL/chosen_KL_mean": -99.06041717529297, "KL/mean": -137.75704956054688, "KL/rejected_KL_mean": -176.4536590576172, "KL/std": 85.93316650390625, "epoch": 0.355359765051395, "fcm_dpo/beta": 0.009386187419295311, "fcm_dpo/delta": 0.07753384858369827, "fcm_dpo/margin": 77.39323425292969, "fcm_dpo/q_t": 0.3551555573940277, "grad_norm": 39.032955169677734, "learning_rate": 4.087222918524807e-07, "logits/chosen": -0.29589563608169556, "logits/rejected": -0.26947832107543945, "logps/chosen": -166.92434692382812, "logps/ref_chosen": -67.86392974853516, "logps/ref_rejected": -83.36033630371094, "logps/rejected": -259.8139953613281, "loss": 0.9389, "margin_dpo/margin_mean": 77.39323425292969, "margin_dpo/margin_std": 94.50575256347656, "step": 242 }, { "KL/chosen_KL_mean": -101.89714050292969, "KL/mean": -153.20147705078125, "KL/rejected_KL_mean": -204.5058135986328, "KL/std": 97.73469543457031, "epoch": 0.3568281938325991, "fcm_dpo/beta": 0.009123127907514572, "fcm_dpo/delta": -0.15070104598999023, "fcm_dpo/margin": 102.60867309570312, "fcm_dpo/q_t": 0.3097137212753296, "grad_norm": 33.98747634887695, "learning_rate": 4.07728699811968e-07, "logits/chosen": -0.34125879406929016, "logits/rejected": -0.2980634570121765, "logps/chosen": -164.98138427734375, "logps/ref_chosen": -63.0842399597168, "logps/ref_rejected": -76.33563232421875, "logps/rejected": -280.8414306640625, "loss": 0.8068, "margin_dpo/margin_mean": 102.60867309570312, "margin_dpo/margin_std": 97.82844543457031, "step": 243 }, { "KL/chosen_KL_mean": -88.12919616699219, "KL/mean": -137.82073974609375, "KL/rejected_KL_mean": -187.51231384277344, "KL/std": 93.26637268066406, "epoch": 0.35829662261380324, "fcm_dpo/beta": 0.008928896859288216, "fcm_dpo/delta": -0.09616594016551971, "fcm_dpo/margin": 99.38310241699219, "fcm_dpo/q_t": 0.31590431928634644, "grad_norm": 32.82498550415039, "learning_rate": 4.067309514735267e-07, "logits/chosen": -0.4070456027984619, "logits/rejected": -0.3957531154155731, "logps/chosen": -149.26988220214844, "logps/ref_chosen": -61.140689849853516, "logps/ref_rejected": -94.89193725585938, "logps/rejected": -282.40423583984375, "loss": 0.8176, "margin_dpo/margin_mean": 99.38311004638672, "margin_dpo/margin_std": 90.44486999511719, "step": 244 }, { "KL/chosen_KL_mean": -102.30755615234375, "KL/mean": -147.13352966308594, "KL/rejected_KL_mean": -191.95950317382812, "KL/std": 100.39794921875, "epoch": 0.35976505139500736, "fcm_dpo/beta": 0.00877309963107109, "fcm_dpo/delta": 0.009379252791404724, "fcm_dpo/margin": 89.65194702148438, "fcm_dpo/q_t": 0.34968093037605286, "grad_norm": 49.55756759643555, "learning_rate": 4.057290731287531e-07, "logits/chosen": -0.40635842084884644, "logits/rejected": -0.3686367869377136, "logps/chosen": -169.56985473632812, "logps/ref_chosen": -67.26228332519531, "logps/ref_rejected": -87.64010620117188, "logps/rejected": -279.599609375, "loss": 0.9526, "margin_dpo/margin_mean": 89.65194702148438, "margin_dpo/margin_std": 113.21443176269531, "step": 245 }, { "KL/chosen_KL_mean": -94.72513580322266, "KL/mean": -136.26837158203125, "KL/rejected_KL_mean": -177.8115997314453, "KL/std": 99.95783996582031, "epoch": 0.36123348017621143, "fcm_dpo/beta": 0.008978039026260376, "fcm_dpo/delta": 0.05843641608953476, "fcm_dpo/margin": 83.08646392822266, "fcm_dpo/q_t": 0.35539257526397705, "grad_norm": 33.92315673828125, "learning_rate": 4.047230911780736e-07, "logits/chosen": -0.45548391342163086, "logits/rejected": -0.41600751876831055, "logps/chosen": -161.42210388183594, "logps/ref_chosen": -66.69696807861328, "logps/ref_rejected": -84.34634399414062, "logps/rejected": -262.157958984375, "loss": 0.9558, "margin_dpo/margin_mean": 83.08645629882812, "margin_dpo/margin_std": 109.47613525390625, "step": 246 }, { "KL/chosen_KL_mean": -109.76311492919922, "KL/mean": -170.4364013671875, "KL/rejected_KL_mean": -231.10968017578125, "KL/std": 111.96722412109375, "epoch": 0.36270190895741555, "fcm_dpo/beta": 0.008609181270003319, "fcm_dpo/delta": -0.27586841583251953, "fcm_dpo/margin": 121.34656524658203, "fcm_dpo/q_t": 0.2987045645713806, "grad_norm": 29.220176696777344, "learning_rate": 4.0371303213004814e-07, "logits/chosen": -0.38043373823165894, "logits/rejected": -0.37580886483192444, "logps/chosen": -166.36846923828125, "logps/ref_chosen": -56.6053466796875, "logps/ref_rejected": -106.29326629638672, "logps/rejected": -337.4029541015625, "loss": 0.8024, "margin_dpo/margin_mean": 121.34656524658203, "margin_dpo/margin_std": 122.87495422363281, "step": 247 }, { "KL/chosen_KL_mean": -87.84927368164062, "KL/mean": -134.8021697998047, "KL/rejected_KL_mean": -181.75506591796875, "KL/std": 83.55899047851562, "epoch": 0.3641703377386197, "fcm_dpo/beta": 0.008513124659657478, "fcm_dpo/delta": 0.0004185140132904053, "fcm_dpo/margin": 93.90579223632812, "fcm_dpo/q_t": 0.3283466696739197, "grad_norm": 48.29165267944336, "learning_rate": 4.0269892260067197e-07, "logits/chosen": -0.37384480237960815, "logits/rejected": -0.3868759274482727, "logps/chosen": -131.89248657226562, "logps/ref_chosen": -44.043216705322266, "logps/ref_rejected": -91.85687255859375, "logps/rejected": -273.6119384765625, "loss": 0.8374, "margin_dpo/margin_mean": 93.90579223632812, "margin_dpo/margin_std": 80.23796081542969, "step": 248 }, { "KL/chosen_KL_mean": -112.0308609008789, "KL/mean": -145.31228637695312, "KL/rejected_KL_mean": -178.59371948242188, "KL/std": 91.14482116699219, "epoch": 0.3656387665198238, "fcm_dpo/beta": 0.008806029334664345, "fcm_dpo/delta": 0.22767800092697144, "fcm_dpo/margin": 66.5628662109375, "fcm_dpo/q_t": 0.38379842042922974, "grad_norm": 38.947914123535156, "learning_rate": 4.0168078931267426e-07, "logits/chosen": -0.4205751419067383, "logits/rejected": -0.394936203956604, "logps/chosen": -174.47320556640625, "logps/ref_chosen": -62.442352294921875, "logps/ref_rejected": -80.46806335449219, "logps/rejected": -259.061767578125, "loss": 1.0696, "margin_dpo/margin_mean": 66.5628662109375, "margin_dpo/margin_std": 105.92036437988281, "step": 249 }, { "KL/chosen_KL_mean": -95.54054260253906, "KL/mean": -145.7799072265625, "KL/rejected_KL_mean": -196.019287109375, "KL/std": 85.64958190917969, "epoch": 0.3671071953010279, "fcm_dpo/beta": 0.008858118206262589, "fcm_dpo/delta": -0.09895782917737961, "fcm_dpo/margin": 100.47872924804688, "fcm_dpo/q_t": 0.3090527057647705, "grad_norm": 46.5678596496582, "learning_rate": 4.006586590948141e-07, "logits/chosen": -0.3768117129802704, "logits/rejected": -0.31270933151245117, "logps/chosen": -161.17723083496094, "logps/ref_chosen": -65.63668823242188, "logps/ref_rejected": -73.87184143066406, "logps/rejected": -269.89111328125, "loss": 0.7945, "margin_dpo/margin_mean": 100.4787368774414, "margin_dpo/margin_std": 79.92070007324219, "step": 250 }, { "KL/chosen_KL_mean": -116.66812896728516, "KL/mean": -158.00106811523438, "KL/rejected_KL_mean": -199.3340301513672, "KL/std": 99.83149719238281, "epoch": 0.368575624082232, "fcm_dpo/beta": 0.008802777156233788, "fcm_dpo/delta": 0.07752127945423126, "fcm_dpo/margin": 82.6658935546875, "fcm_dpo/q_t": 0.35495853424072266, "grad_norm": 48.53387451171875, "learning_rate": 3.9963255888117325e-07, "logits/chosen": -0.3401832580566406, "logits/rejected": -0.3037635087966919, "logps/chosen": -173.85084533691406, "logps/ref_chosen": -57.182716369628906, "logps/ref_rejected": -77.66343688964844, "logps/rejected": -276.9974670410156, "loss": 0.968, "margin_dpo/margin_mean": 82.6658935546875, "margin_dpo/margin_std": 105.33753967285156, "step": 251 }, { "KL/chosen_KL_mean": -115.15412902832031, "KL/mean": -166.94317626953125, "KL/rejected_KL_mean": -218.7322540283203, "KL/std": 89.63955688476562, "epoch": 0.3700440528634361, "fcm_dpo/beta": 0.008697323501110077, "fcm_dpo/delta": -0.11240165680646896, "fcm_dpo/margin": 103.57811737060547, "fcm_dpo/q_t": 0.30444052815437317, "grad_norm": 34.029197692871094, "learning_rate": 3.9860251571044666e-07, "logits/chosen": -0.39935481548309326, "logits/rejected": -0.35331428050994873, "logps/chosen": -186.8397674560547, "logps/ref_chosen": -71.68563842773438, "logps/ref_rejected": -84.75799560546875, "logps/rejected": -303.490234375, "loss": 0.7994, "margin_dpo/margin_mean": 103.57811737060547, "margin_dpo/margin_std": 86.66522216796875, "step": 252 }, { "KL/chosen_KL_mean": -109.57987976074219, "KL/mean": -156.1011962890625, "KL/rejected_KL_mean": -202.62249755859375, "KL/std": 100.52079010009766, "epoch": 0.37151248164464024, "fcm_dpo/beta": 0.008769432082772255, "fcm_dpo/delta": -0.019483350217342377, "fcm_dpo/margin": 93.04261779785156, "fcm_dpo/q_t": 0.3355570137500763, "grad_norm": 38.78324508666992, "learning_rate": 3.9756855672522986e-07, "logits/chosen": -0.4148038625717163, "logits/rejected": -0.4067913293838501, "logps/chosen": -178.71380615234375, "logps/ref_chosen": -69.1339340209961, "logps/ref_rejected": -98.70252990722656, "logps/rejected": -301.32501220703125, "loss": 0.9266, "margin_dpo/margin_mean": 93.04261779785156, "margin_dpo/margin_std": 110.63420867919922, "step": 253 }, { "KL/chosen_KL_mean": -120.14738464355469, "KL/mean": -163.5091552734375, "KL/rejected_KL_mean": -206.8709716796875, "KL/std": 104.5792007446289, "epoch": 0.37298091042584436, "fcm_dpo/beta": 0.00873111467808485, "fcm_dpo/delta": 0.046288929879665375, "fcm_dpo/margin": 86.72355651855469, "fcm_dpo/q_t": 0.36139702796936035, "grad_norm": 52.724220275878906, "learning_rate": 3.965307091713037e-07, "logits/chosen": -0.33431658148765564, "logits/rejected": -0.30974212288856506, "logps/chosen": -174.30238342285156, "logps/ref_chosen": -54.154998779296875, "logps/ref_rejected": -90.30764770507812, "logps/rejected": -297.1785888671875, "loss": 0.9987, "margin_dpo/margin_mean": 86.72355651855469, "margin_dpo/margin_std": 128.11468505859375, "step": 254 }, { "KL/chosen_KL_mean": -123.59725952148438, "KL/mean": -169.68804931640625, "KL/rejected_KL_mean": -215.7788543701172, "KL/std": 99.052001953125, "epoch": 0.3744493392070485, "fcm_dpo/beta": 0.008680582977831364, "fcm_dpo/delta": -0.0014177486300468445, "fcm_dpo/margin": 92.18161010742188, "fcm_dpo/q_t": 0.34611931443214417, "grad_norm": 48.63689041137695, "learning_rate": 3.954890003969163e-07, "logits/chosen": -0.30374979972839355, "logits/rejected": -0.2946004271507263, "logps/chosen": -180.73892211914062, "logps/ref_chosen": -57.14167022705078, "logps/ref_rejected": -90.2085952758789, "logps/rejected": -305.9874572753906, "loss": 0.9727, "margin_dpo/margin_mean": 92.18161010742188, "margin_dpo/margin_std": 123.77351379394531, "step": 255 }, { "KL/chosen_KL_mean": -105.18153381347656, "KL/mean": -155.66441345214844, "KL/rejected_KL_mean": -206.14730834960938, "KL/std": 106.35296630859375, "epoch": 0.37591776798825255, "fcm_dpo/beta": 0.008626077324151993, "fcm_dpo/delta": -0.0779896229505539, "fcm_dpo/margin": 100.96575927734375, "fcm_dpo/q_t": 0.3262209892272949, "grad_norm": 29.191234588623047, "learning_rate": 3.944434578520628e-07, "logits/chosen": -0.30876320600509644, "logits/rejected": -0.30598020553588867, "logps/chosen": -160.34503173828125, "logps/ref_chosen": -55.163490295410156, "logps/ref_rejected": -92.56291961669922, "logps/rejected": -298.710205078125, "loss": 0.8687, "margin_dpo/margin_mean": 100.96575927734375, "margin_dpo/margin_std": 109.50328063964844, "step": 256 }, { "KL/chosen_KL_mean": -97.52660369873047, "KL/mean": -149.33360290527344, "KL/rejected_KL_mean": -201.14059448242188, "KL/std": 103.94354248046875, "epoch": 0.37738619676945667, "fcm_dpo/beta": 0.00838499329984188, "fcm_dpo/delta": -0.08100643008947372, "fcm_dpo/margin": 103.61398315429688, "fcm_dpo/q_t": 0.3288852870464325, "grad_norm": 28.96324348449707, "learning_rate": 3.933941090877615e-07, "logits/chosen": -0.3473857343196869, "logits/rejected": -0.33152860403060913, "logps/chosen": -146.95030212402344, "logps/ref_chosen": -49.42369842529297, "logps/ref_rejected": -79.53791809082031, "logps/rejected": -280.67852783203125, "loss": 0.8754, "margin_dpo/margin_mean": 103.61398315429688, "margin_dpo/margin_std": 109.51292419433594, "step": 257 }, { "KL/chosen_KL_mean": -112.60980224609375, "KL/mean": -157.80865478515625, "KL/rejected_KL_mean": -203.00750732421875, "KL/std": 97.22439575195312, "epoch": 0.3788546255506608, "fcm_dpo/beta": 0.008527854457497597, "fcm_dpo/delta": 0.03139530122280121, "fcm_dpo/margin": 90.397705078125, "fcm_dpo/q_t": 0.34141361713409424, "grad_norm": 37.659915924072266, "learning_rate": 3.923409817553284e-07, "logits/chosen": -0.32153844833374023, "logits/rejected": -0.31202760338783264, "logps/chosen": -171.99392700195312, "logps/ref_chosen": -59.384124755859375, "logps/ref_rejected": -95.99010467529297, "logps/rejected": -298.99761962890625, "loss": 0.9525, "margin_dpo/margin_mean": 90.39769744873047, "margin_dpo/margin_std": 114.62164306640625, "step": 258 }, { "KL/chosen_KL_mean": -101.266845703125, "KL/mean": -141.80453491210938, "KL/rejected_KL_mean": -182.3422088623047, "KL/std": 91.08666229248047, "epoch": 0.3803230543318649, "fcm_dpo/beta": 0.008614392951130867, "fcm_dpo/delta": 0.10913976281881332, "fcm_dpo/margin": 81.07535552978516, "fcm_dpo/q_t": 0.3593909740447998, "grad_norm": 40.97324752807617, "learning_rate": 3.9128410360564793e-07, "logits/chosen": -0.39695829153060913, "logits/rejected": -0.3891292214393616, "logps/chosen": -154.09518432617188, "logps/ref_chosen": -52.828346252441406, "logps/ref_rejected": -89.191650390625, "logps/rejected": -271.53387451171875, "loss": 0.9726, "margin_dpo/margin_mean": 81.07535552978516, "margin_dpo/margin_std": 102.6622314453125, "step": 259 }, { "KL/chosen_KL_mean": -96.43135833740234, "KL/mean": -142.50563049316406, "KL/rejected_KL_mean": -188.57989501953125, "KL/std": 89.96287536621094, "epoch": 0.38179148311306904, "fcm_dpo/beta": 0.008739551529288292, "fcm_dpo/delta": -0.006093751639127731, "fcm_dpo/margin": 92.14852905273438, "fcm_dpo/q_t": 0.33704131841659546, "grad_norm": 45.9984016418457, "learning_rate": 3.9022350248844246e-07, "logits/chosen": -0.3932284116744995, "logits/rejected": -0.4087497591972351, "logps/chosen": -143.84902954101562, "logps/ref_chosen": -47.41767501831055, "logps/ref_rejected": -95.08978271484375, "logps/rejected": -283.669677734375, "loss": 0.8789, "margin_dpo/margin_mean": 92.14852905273438, "margin_dpo/margin_std": 98.96159362792969, "step": 260 }, { "KL/chosen_KL_mean": -85.78534698486328, "KL/mean": -136.07266235351562, "KL/rejected_KL_mean": -186.35995483398438, "KL/std": 94.51484680175781, "epoch": 0.3832599118942731, "fcm_dpo/beta": 0.008594926446676254, "fcm_dpo/delta": -0.0707588791847229, "fcm_dpo/margin": 100.57462310791016, "fcm_dpo/q_t": 0.32291507720947266, "grad_norm": 32.12379837036133, "learning_rate": 3.891592063515376e-07, "logits/chosen": -0.34815192222595215, "logits/rejected": -0.3409876227378845, "logps/chosen": -138.8167266845703, "logps/ref_chosen": -53.03137969970703, "logps/ref_rejected": -88.51494598388672, "logps/rejected": -274.8749084472656, "loss": 0.8429, "margin_dpo/margin_mean": 100.57461547851562, "margin_dpo/margin_std": 100.58946990966797, "step": 261 }, { "KL/chosen_KL_mean": -99.97845458984375, "KL/mean": -139.77328491210938, "KL/rejected_KL_mean": -179.568115234375, "KL/std": 84.39540100097656, "epoch": 0.38472834067547723, "fcm_dpo/beta": 0.008759372867643833, "fcm_dpo/delta": 0.10967819392681122, "fcm_dpo/margin": 79.58967590332031, "fcm_dpo/q_t": 0.3497657775878906, "grad_norm": 31.033384323120117, "learning_rate": 3.880912432401264e-07, "logits/chosen": -0.3228227496147156, "logits/rejected": -0.2938918173313141, "logps/chosen": -159.59860229492188, "logps/ref_chosen": -59.620140075683594, "logps/ref_rejected": -86.41853332519531, "logps/rejected": -265.98663330078125, "loss": 0.9248, "margin_dpo/margin_mean": 79.58967590332031, "margin_dpo/margin_std": 84.16622924804688, "step": 262 }, { "KL/chosen_KL_mean": -91.52005767822266, "KL/mean": -147.3134002685547, "KL/rejected_KL_mean": -203.10671997070312, "KL/std": 100.20832061767578, "epoch": 0.38619676945668135, "fcm_dpo/beta": 0.008518587797880173, "fcm_dpo/delta": -0.1675512194633484, "fcm_dpo/margin": 111.58668518066406, "fcm_dpo/q_t": 0.3041931092739105, "grad_norm": 32.451473236083984, "learning_rate": 3.870196412960302e-07, "logits/chosen": -0.3325992822647095, "logits/rejected": -0.29685109853744507, "logps/chosen": -150.94100952148438, "logps/ref_chosen": -59.42094421386719, "logps/ref_rejected": -96.85720825195312, "logps/rejected": -299.96392822265625, "loss": 0.7942, "margin_dpo/margin_mean": 111.58668518066406, "margin_dpo/margin_std": 99.71820068359375, "step": 263 }, { "KL/chosen_KL_mean": -96.28630828857422, "KL/mean": -141.96990966796875, "KL/rejected_KL_mean": -187.65350341796875, "KL/std": 88.48318481445312, "epoch": 0.3876651982378855, "fcm_dpo/beta": 0.00846080295741558, "fcm_dpo/delta": 0.028465010225772858, "fcm_dpo/margin": 91.36720275878906, "fcm_dpo/q_t": 0.33899131417274475, "grad_norm": 36.09725570678711, "learning_rate": 3.8594442875695665e-07, "logits/chosen": -0.33131206035614014, "logits/rejected": -0.3153602182865143, "logps/chosen": -159.00839233398438, "logps/ref_chosen": -62.722084045410156, "logps/ref_rejected": -93.85620880126953, "logps/rejected": -281.50970458984375, "loss": 0.8846, "margin_dpo/margin_mean": 91.36720275878906, "margin_dpo/margin_std": 92.86323547363281, "step": 264 }, { "KL/chosen_KL_mean": -110.69151306152344, "KL/mean": -156.87734985351562, "KL/rejected_KL_mean": -203.06320190429688, "KL/std": 102.73265075683594, "epoch": 0.3891336270190896, "fcm_dpo/beta": 0.008556404151022434, "fcm_dpo/delta": 0.010393224656581879, "fcm_dpo/margin": 92.3717041015625, "fcm_dpo/q_t": 0.3452988862991333, "grad_norm": 44.408111572265625, "learning_rate": 3.848656339557562e-07, "logits/chosen": -0.2442556917667389, "logits/rejected": -0.22311729192733765, "logps/chosen": -172.66299438476562, "logps/ref_chosen": -61.971466064453125, "logps/ref_rejected": -88.02059936523438, "logps/rejected": -291.08380126953125, "loss": 0.9235, "margin_dpo/margin_mean": 92.3717041015625, "margin_dpo/margin_std": 112.91886138916016, "step": 265 }, { "KL/chosen_KL_mean": -107.76649475097656, "KL/mean": -154.02108764648438, "KL/rejected_KL_mean": -200.27566528320312, "KL/std": 99.5924072265625, "epoch": 0.39060205580029367, "fcm_dpo/beta": 0.008531241677701473, "fcm_dpo/delta": 0.01146605983376503, "fcm_dpo/margin": 92.50917053222656, "fcm_dpo/q_t": 0.33881711959838867, "grad_norm": 54.370849609375, "learning_rate": 3.8378328531967507e-07, "logits/chosen": -0.3046592175960541, "logits/rejected": -0.24882930517196655, "logps/chosen": -174.8661651611328, "logps/ref_chosen": -67.09967041015625, "logps/ref_rejected": -67.97122192382812, "logps/rejected": -268.24688720703125, "loss": 0.8963, "margin_dpo/margin_mean": 92.50917053222656, "margin_dpo/margin_std": 100.46192932128906, "step": 266 }, { "KL/chosen_KL_mean": -111.7256088256836, "KL/mean": -161.23422241210938, "KL/rejected_KL_mean": -210.74285888671875, "KL/std": 98.82826232910156, "epoch": 0.3920704845814978, "fcm_dpo/beta": 0.008480764925479889, "fcm_dpo/delta": -0.04383730888366699, "fcm_dpo/margin": 99.01722717285156, "fcm_dpo/q_t": 0.3242519497871399, "grad_norm": 41.02730941772461, "learning_rate": 3.8269741136960646e-07, "logits/chosen": -0.26957884430885315, "logits/rejected": -0.2300885021686554, "logps/chosen": -180.6963653564453, "logps/ref_chosen": -68.97075653076172, "logps/ref_rejected": -90.16844940185547, "logps/rejected": -300.91131591796875, "loss": 0.8434, "margin_dpo/margin_mean": 99.01722717285156, "margin_dpo/margin_std": 95.28477478027344, "step": 267 }, { "KL/chosen_KL_mean": -126.69245910644531, "KL/mean": -174.75869750976562, "KL/rejected_KL_mean": -222.824951171875, "KL/std": 94.38973236083984, "epoch": 0.3935389133627019, "fcm_dpo/beta": 0.008443417027592659, "fcm_dpo/delta": -0.013090118765830994, "fcm_dpo/margin": 96.13249206542969, "fcm_dpo/q_t": 0.3327893316745758, "grad_norm": 47.08164978027344, "learning_rate": 3.8160804071933894e-07, "logits/chosen": -0.20129185914993286, "logits/rejected": -0.20935595035552979, "logps/chosen": -182.5927734375, "logps/ref_chosen": -55.90031051635742, "logps/ref_rejected": -101.64763641357422, "logps/rejected": -324.47259521484375, "loss": 0.8821, "margin_dpo/margin_mean": 96.13249206542969, "margin_dpo/margin_std": 103.66873931884766, "step": 268 }, { "KL/chosen_KL_mean": -144.3126220703125, "KL/mean": -207.30047607421875, "KL/rejected_KL_mean": -270.288330078125, "KL/std": 112.39837646484375, "epoch": 0.39500734214390604, "fcm_dpo/beta": 0.008144214749336243, "fcm_dpo/delta": -0.25245746970176697, "fcm_dpo/margin": 125.9757080078125, "fcm_dpo/q_t": 0.30247214436531067, "grad_norm": 60.25404357910156, "learning_rate": 3.8051520207480204e-07, "logits/chosen": -0.2495516538619995, "logits/rejected": -0.2241310477256775, "logps/chosen": -214.35218811035156, "logps/ref_chosen": -70.03955841064453, "logps/ref_rejected": -107.34937286376953, "logps/rejected": -377.6376953125, "loss": 0.8291, "margin_dpo/margin_mean": 125.9757080078125, "margin_dpo/margin_std": 133.42056274414062, "step": 269 }, { "KL/chosen_KL_mean": -136.29586791992188, "KL/mean": -183.19334411621094, "KL/rejected_KL_mean": -230.0908203125, "KL/std": 105.35012817382812, "epoch": 0.3964757709251101, "fcm_dpo/beta": 0.008130217902362347, "fcm_dpo/delta": 0.04011562466621399, "fcm_dpo/margin": 93.7949447631836, "fcm_dpo/q_t": 0.35046982765197754, "grad_norm": 50.62470245361328, "learning_rate": 3.794189242333106e-07, "logits/chosen": -0.29408591985702515, "logits/rejected": -0.2865193486213684, "logps/chosen": -205.829345703125, "logps/ref_chosen": -69.53347778320312, "logps/ref_rejected": -109.92864990234375, "logps/rejected": -340.01947021484375, "loss": 0.9617, "margin_dpo/margin_mean": 93.79493713378906, "margin_dpo/margin_std": 123.04978942871094, "step": 270 }, { "KL/chosen_KL_mean": -136.893310546875, "KL/mean": -195.67169189453125, "KL/rejected_KL_mean": -254.4500274658203, "KL/std": 104.8451156616211, "epoch": 0.39794419970631423, "fcm_dpo/beta": 0.007922438904643059, "fcm_dpo/delta": -0.1454475224018097, "fcm_dpo/margin": 117.55671691894531, "fcm_dpo/q_t": 0.3143436908721924, "grad_norm": 33.90323257446289, "learning_rate": 3.7831923608280514e-07, "logits/chosen": -0.20189276337623596, "logits/rejected": -0.17152735590934753, "logps/chosen": -193.65789794921875, "logps/ref_chosen": -56.76456832885742, "logps/ref_rejected": -92.51383972167969, "logps/rejected": -346.9638671875, "loss": 0.8179, "margin_dpo/margin_mean": 117.55671691894531, "margin_dpo/margin_std": 115.97308349609375, "step": 271 }, { "KL/chosen_KL_mean": -128.50242614746094, "KL/mean": -197.3144073486328, "KL/rejected_KL_mean": -266.1263732910156, "KL/std": 122.61846160888672, "epoch": 0.39941262848751835, "fcm_dpo/beta": 0.00761681143194437, "fcm_dpo/delta": -0.27643194794654846, "fcm_dpo/margin": 137.62393188476562, "fcm_dpo/q_t": 0.2913571000099182, "grad_norm": 40.26355743408203, "learning_rate": 3.772161666010912e-07, "logits/chosen": -0.16119909286499023, "logits/rejected": -0.17272311449050903, "logps/chosen": -177.9995880126953, "logps/ref_chosen": -49.497154235839844, "logps/ref_rejected": -105.54279327392578, "logps/rejected": -371.669189453125, "loss": 0.7877, "margin_dpo/margin_mean": 137.62393188476562, "margin_dpo/margin_std": 130.5201416015625, "step": 272 }, { "KL/chosen_KL_mean": -142.26593017578125, "KL/mean": -211.93959045410156, "KL/rejected_KL_mean": -281.61322021484375, "KL/std": 124.14240264892578, "epoch": 0.4008810572687225, "fcm_dpo/beta": 0.007157725282013416, "fcm_dpo/delta": -0.22283779084682465, "fcm_dpo/margin": 139.3472900390625, "fcm_dpo/q_t": 0.3012547492980957, "grad_norm": 32.3958854675293, "learning_rate": 3.761097448550755e-07, "logits/chosen": -0.19352608919143677, "logits/rejected": -0.16837653517723083, "logps/chosen": -205.2413330078125, "logps/ref_chosen": -62.97539520263672, "logps/ref_rejected": -92.49858093261719, "logps/rejected": -374.11181640625, "loss": 0.7903, "margin_dpo/margin_mean": 139.34730529785156, "margin_dpo/margin_std": 130.78785705566406, "step": 273 }, { "KL/chosen_KL_mean": -173.3973388671875, "KL/mean": -229.00474548339844, "KL/rejected_KL_mean": -284.6121520996094, "KL/std": 112.57540893554688, "epoch": 0.4023494860499266, "fcm_dpo/beta": 0.007162545807659626, "fcm_dpo/delta": 0.0034975409507751465, "fcm_dpo/margin": 111.2148208618164, "fcm_dpo/q_t": 0.33956801891326904, "grad_norm": 45.80356979370117, "learning_rate": 3.75e-07, "logits/chosen": -0.11668691784143448, "logits/rejected": -0.08984885364770889, "logps/chosen": -229.06503295898438, "logps/ref_chosen": -55.66770935058594, "logps/ref_rejected": -77.33308410644531, "logps/rejected": -361.94525146484375, "loss": 0.9127, "margin_dpo/margin_mean": 111.21481323242188, "margin_dpo/margin_std": 127.79832458496094, "step": 274 }, { "KL/chosen_KL_mean": -135.02244567871094, "KL/mean": -191.32406616210938, "KL/rejected_KL_mean": -247.62567138671875, "KL/std": 117.45345306396484, "epoch": 0.40381791483113066, "fcm_dpo/beta": 0.007176370359957218, "fcm_dpo/delta": -0.009557224810123444, "fcm_dpo/margin": 112.60321044921875, "fcm_dpo/q_t": 0.339751273393631, "grad_norm": 38.28922653198242, "learning_rate": 3.738869612786737e-07, "logits/chosen": -0.20961615443229675, "logits/rejected": -0.21241699159145355, "logps/chosen": -183.61715698242188, "logps/ref_chosen": -48.594703674316406, "logps/ref_rejected": -93.30369567871094, "logps/rejected": -340.92938232421875, "loss": 0.9185, "margin_dpo/margin_mean": 112.60321044921875, "margin_dpo/margin_std": 132.3553466796875, "step": 275 }, { "KL/chosen_KL_mean": -146.90843200683594, "KL/mean": -208.81777954101562, "KL/rejected_KL_mean": -270.72711181640625, "KL/std": 123.26631164550781, "epoch": 0.4052863436123348, "fcm_dpo/beta": 0.007015553303062916, "fcm_dpo/delta": -0.07598960399627686, "fcm_dpo/margin": 123.81869506835938, "fcm_dpo/q_t": 0.3311005234718323, "grad_norm": 38.56481170654297, "learning_rate": 3.7277065802070204e-07, "logits/chosen": -0.19943909347057343, "logits/rejected": -0.16966615617275238, "logps/chosen": -203.48583984375, "logps/ref_chosen": -56.57740783691406, "logps/ref_rejected": -70.36566925048828, "logps/rejected": -341.0927734375, "loss": 0.8881, "margin_dpo/margin_mean": 123.81867980957031, "margin_dpo/margin_std": 144.08128356933594, "step": 276 }, { "KL/chosen_KL_mean": -158.66197204589844, "KL/mean": -219.00457763671875, "KL/rejected_KL_mean": -279.34716796875, "KL/std": 118.71607208251953, "epoch": 0.4067547723935389, "fcm_dpo/beta": 0.007000372279435396, "fcm_dpo/delta": -0.04916992783546448, "fcm_dpo/margin": 120.68519592285156, "fcm_dpo/q_t": 0.3313535451889038, "grad_norm": 34.503780364990234, "learning_rate": 3.71651119641714e-07, "logits/chosen": -0.1559976041316986, "logits/rejected": -0.1391219049692154, "logps/chosen": -214.93353271484375, "logps/ref_chosen": -56.27156066894531, "logps/ref_rejected": -92.88127136230469, "logps/rejected": -372.22845458984375, "loss": 0.8744, "margin_dpo/margin_mean": 120.68519592285156, "margin_dpo/margin_std": 130.1613311767578, "step": 277 }, { "KL/chosen_KL_mean": -151.7635955810547, "KL/mean": -215.63674926757812, "KL/rejected_KL_mean": -279.5098876953125, "KL/std": 126.30606079101562, "epoch": 0.40822320117474303, "fcm_dpo/beta": 0.006845751777291298, "fcm_dpo/delta": -0.08213748037815094, "fcm_dpo/margin": 127.74630737304688, "fcm_dpo/q_t": 0.3263602554798126, "grad_norm": 38.64687728881836, "learning_rate": 3.705283756425872e-07, "logits/chosen": -0.17192326486110687, "logits/rejected": -0.17586654424667358, "logps/chosen": -204.70553588867188, "logps/ref_chosen": -52.94194030761719, "logps/ref_rejected": -91.25357818603516, "logps/rejected": -370.7634582519531, "loss": 0.8594, "margin_dpo/margin_mean": 127.74630737304688, "margin_dpo/margin_std": 135.8467559814453, "step": 278 }, { "KL/chosen_KL_mean": -168.48736572265625, "KL/mean": -237.06497192382812, "KL/rejected_KL_mean": -305.642578125, "KL/std": 130.68185424804688, "epoch": 0.40969162995594716, "fcm_dpo/beta": 0.006637511774897575, "fcm_dpo/delta": -0.12550733983516693, "fcm_dpo/margin": 137.15524291992188, "fcm_dpo/q_t": 0.3223535418510437, "grad_norm": 44.90956497192383, "learning_rate": 3.6940245560867e-07, "logits/chosen": -0.06602071225643158, "logits/rejected": -0.05742315948009491, "logps/chosen": -217.12867736816406, "logps/ref_chosen": -48.641319274902344, "logps/ref_rejected": -87.8514404296875, "logps/rejected": -393.4940185546875, "loss": 0.8748, "margin_dpo/margin_mean": 137.15524291992188, "margin_dpo/margin_std": 151.80455017089844, "step": 279 }, { "KL/chosen_KL_mean": -147.5689697265625, "KL/mean": -216.75357055664062, "KL/rejected_KL_mean": -285.93817138671875, "KL/std": 113.64622497558594, "epoch": 0.4111600587371512, "fcm_dpo/beta": 0.006537692155689001, "fcm_dpo/delta": -0.1155073344707489, "fcm_dpo/margin": 138.36923217773438, "fcm_dpo/q_t": 0.3063077926635742, "grad_norm": 55.07219314575195, "learning_rate": 3.6827338920900253e-07, "logits/chosen": -0.15549999475479126, "logits/rejected": -0.15061771869659424, "logps/chosen": -206.3660888671875, "logps/ref_chosen": -58.797122955322266, "logps/ref_rejected": -98.61885070800781, "logps/rejected": -384.5570373535156, "loss": 0.7877, "margin_dpo/margin_mean": 138.3692169189453, "margin_dpo/margin_std": 111.95368957519531, "step": 280 }, { "KL/chosen_KL_mean": -141.1031494140625, "KL/mean": -203.22886657714844, "KL/rejected_KL_mean": -265.35455322265625, "KL/std": 114.74964904785156, "epoch": 0.41262848751835535, "fcm_dpo/beta": 0.006536037661135197, "fcm_dpo/delta": -0.013637367635965347, "fcm_dpo/margin": 124.25141143798828, "fcm_dpo/q_t": 0.32296425104141235, "grad_norm": 49.556819915771484, "learning_rate": 3.6714120619553435e-07, "logits/chosen": -0.2126597762107849, "logits/rejected": -0.17675194144248962, "logps/chosen": -196.5916748046875, "logps/ref_chosen": -55.488521575927734, "logps/ref_rejected": -80.88258361816406, "logps/rejected": -346.2371520996094, "loss": 0.8527, "margin_dpo/margin_mean": 124.25141143798828, "margin_dpo/margin_std": 114.7934799194336, "step": 281 }, { "KL/chosen_KL_mean": -156.95260620117188, "KL/mean": -207.12942504882812, "KL/rejected_KL_mean": -257.3062438964844, "KL/std": 108.96510314941406, "epoch": 0.41409691629955947, "fcm_dpo/beta": 0.006672097835689783, "fcm_dpo/delta": 0.13827921450138092, "fcm_dpo/margin": 100.3536376953125, "fcm_dpo/q_t": 0.36175787448883057, "grad_norm": 41.35542297363281, "learning_rate": 3.660059364023408e-07, "logits/chosen": -0.2611793577671051, "logits/rejected": -0.22383667528629303, "logps/chosen": -230.02276611328125, "logps/ref_chosen": -73.07014465332031, "logps/ref_rejected": -95.35098266601562, "logps/rejected": -352.6572265625, "loss": 0.9472, "margin_dpo/margin_mean": 100.3536376953125, "margin_dpo/margin_std": 119.709716796875, "step": 282 }, { "KL/chosen_KL_mean": -165.55203247070312, "KL/mean": -235.3925018310547, "KL/rejected_KL_mean": -305.2330017089844, "KL/std": 139.40737915039062, "epoch": 0.4155653450807636, "fcm_dpo/beta": 0.0065710763446986675, "fcm_dpo/delta": -0.1295718550682068, "fcm_dpo/margin": 139.68096923828125, "fcm_dpo/q_t": 0.30924761295318604, "grad_norm": 42.45329666137695, "learning_rate": 3.6486760974483685e-07, "logits/chosen": -0.184697687625885, "logits/rejected": -0.17743512988090515, "logps/chosen": -227.45046997070312, "logps/ref_chosen": -61.89844512939453, "logps/ref_rejected": -96.98655700683594, "logps/rejected": -402.21954345703125, "loss": 0.8136, "margin_dpo/margin_mean": 139.68096923828125, "margin_dpo/margin_std": 129.1880340576172, "step": 283 }, { "KL/chosen_KL_mean": -171.84423828125, "KL/mean": -239.68804931640625, "KL/rejected_KL_mean": -307.5318603515625, "KL/std": 133.614990234375, "epoch": 0.4170337738619677, "fcm_dpo/beta": 0.006425424944609404, "fcm_dpo/delta": -0.07886916399002075, "fcm_dpo/margin": 135.68765258789062, "fcm_dpo/q_t": 0.3206017017364502, "grad_norm": 45.275691986083984, "learning_rate": 3.6372625621898863e-07, "logits/chosen": -0.16477805376052856, "logits/rejected": -0.14584243297576904, "logps/chosen": -230.27975463867188, "logps/ref_chosen": -58.4355354309082, "logps/ref_rejected": -93.46926879882812, "logps/rejected": -401.0011291503906, "loss": 0.8227, "margin_dpo/margin_mean": 135.68765258789062, "margin_dpo/margin_std": 128.515380859375, "step": 284 }, { "KL/chosen_KL_mean": -185.8503875732422, "KL/mean": -248.33926391601562, "KL/rejected_KL_mean": -310.8281555175781, "KL/std": 114.6676025390625, "epoch": 0.4185022026431718, "fcm_dpo/beta": 0.00645300280302763, "fcm_dpo/delta": -0.008379101753234863, "fcm_dpo/margin": 124.9777603149414, "fcm_dpo/q_t": 0.3266201615333557, "grad_norm": 48.88724899291992, "learning_rate": 3.625819059005228e-07, "logits/chosen": -0.20939572155475616, "logits/rejected": -0.18859480321407318, "logps/chosen": -252.08258056640625, "logps/ref_chosen": -66.23219299316406, "logps/ref_rejected": -99.1268310546875, "logps/rejected": -409.9549865722656, "loss": 0.85, "margin_dpo/margin_mean": 124.9777603149414, "margin_dpo/margin_std": 111.27662658691406, "step": 285 }, { "KL/chosen_KL_mean": -186.20968627929688, "KL/mean": -254.66603088378906, "KL/rejected_KL_mean": -323.12237548828125, "KL/std": 133.13705444335938, "epoch": 0.4199706314243759, "fcm_dpo/beta": 0.0063394177705049515, "fcm_dpo/delta": -0.07433885335922241, "fcm_dpo/margin": 136.91268920898438, "fcm_dpo/q_t": 0.32065117359161377, "grad_norm": 33.741024017333984, "learning_rate": 3.614345889441346e-07, "logits/chosen": -0.19529663026332855, "logits/rejected": -0.17008443176746368, "logps/chosen": -259.16070556640625, "logps/ref_chosen": -72.95100402832031, "logps/ref_rejected": -88.58845520019531, "logps/rejected": -411.7108154296875, "loss": 0.8456, "margin_dpo/margin_mean": 136.91268920898438, "margin_dpo/margin_std": 134.44070434570312, "step": 286 }, { "KL/chosen_KL_mean": -194.35592651367188, "KL/mean": -251.70867919921875, "KL/rejected_KL_mean": -309.0614013671875, "KL/std": 127.58320617675781, "epoch": 0.42143906020558003, "fcm_dpo/beta": 0.006387564353644848, "fcm_dpo/delta": 0.07230030000209808, "fcm_dpo/margin": 114.70552825927734, "fcm_dpo/q_t": 0.3514924943447113, "grad_norm": 35.43113708496094, "learning_rate": 3.6028433558269275e-07, "logits/chosen": -0.09007483720779419, "logits/rejected": -0.06799876689910889, "logps/chosen": -255.89706420898438, "logps/ref_chosen": -61.54115295410156, "logps/ref_rejected": -77.69607543945312, "logps/rejected": -386.75750732421875, "loss": 0.9391, "margin_dpo/margin_mean": 114.70552825927734, "margin_dpo/margin_std": 139.09039306640625, "step": 287 }, { "KL/chosen_KL_mean": -190.921142578125, "KL/mean": -257.931396484375, "KL/rejected_KL_mean": -324.941650390625, "KL/std": 118.37663269042969, "epoch": 0.42290748898678415, "fcm_dpo/beta": 0.0062710437923669815, "fcm_dpo/delta": -0.04802323132753372, "fcm_dpo/margin": 134.02052307128906, "fcm_dpo/q_t": 0.3206808567047119, "grad_norm": 33.899898529052734, "learning_rate": 3.5913117612644327e-07, "logits/chosen": -0.14238840341567993, "logits/rejected": -0.12358824163675308, "logps/chosen": -247.58236694335938, "logps/ref_chosen": -56.661224365234375, "logps/ref_rejected": -87.33570098876953, "logps/rejected": -412.27734375, "loss": 0.8339, "margin_dpo/margin_mean": 134.02052307128906, "margin_dpo/margin_std": 115.77912139892578, "step": 288 }, { "KL/chosen_KL_mean": -187.45516967773438, "KL/mean": -263.76019287109375, "KL/rejected_KL_mean": -340.065185546875, "KL/std": 135.79379272460938, "epoch": 0.4243759177679883, "fcm_dpo/beta": 0.006219586357474327, "fcm_dpo/delta": -0.1645835041999817, "fcm_dpo/margin": 152.6099853515625, "fcm_dpo/q_t": 0.31164634227752686, "grad_norm": 37.334537506103516, "learning_rate": 3.5797514096221024e-07, "logits/chosen": -0.028572116047143936, "logits/rejected": -0.022187216207385063, "logps/chosen": -232.68557739257812, "logps/ref_chosen": -45.23039245605469, "logps/ref_rejected": -87.64266967773438, "logps/rejected": -427.70782470703125, "loss": 0.8196, "margin_dpo/margin_mean": 152.6099853515625, "margin_dpo/margin_std": 149.71493530273438, "step": 289 }, { "KL/chosen_KL_mean": -209.78424072265625, "KL/mean": -285.5294189453125, "KL/rejected_KL_mean": -361.274658203125, "KL/std": 147.62713623046875, "epoch": 0.42584434654919234, "fcm_dpo/beta": 0.006003453396260738, "fcm_dpo/delta": -0.12091632187366486, "fcm_dpo/margin": 151.49041748046875, "fcm_dpo/q_t": 0.32481837272644043, "grad_norm": 39.83347702026367, "learning_rate": 3.568162605525952e-07, "logits/chosen": -0.036675482988357544, "logits/rejected": -0.05918949097394943, "logps/chosen": -265.2557373046875, "logps/ref_chosen": -55.47149658203125, "logps/ref_rejected": -116.70857238769531, "logps/rejected": -477.98321533203125, "loss": 0.8687, "margin_dpo/margin_mean": 151.4904022216797, "margin_dpo/margin_std": 172.48365783691406, "step": 290 }, { "KL/chosen_KL_mean": -181.04437255859375, "KL/mean": -251.22125244140625, "KL/rejected_KL_mean": -321.39813232421875, "KL/std": 126.02388000488281, "epoch": 0.42731277533039647, "fcm_dpo/beta": 0.005997511558234692, "fcm_dpo/delta": -0.04703492671251297, "fcm_dpo/margin": 140.35374450683594, "fcm_dpo/q_t": 0.32573235034942627, "grad_norm": 40.698150634765625, "learning_rate": 3.5565456543517485e-07, "logits/chosen": -0.08782187104225159, "logits/rejected": -0.07490293681621552, "logps/chosen": -244.30474853515625, "logps/ref_chosen": -63.26036834716797, "logps/ref_rejected": -89.29708862304688, "logps/rejected": -410.6952209472656, "loss": 0.8633, "margin_dpo/margin_mean": 140.35374450683594, "margin_dpo/margin_std": 138.77685546875, "step": 291 }, { "KL/chosen_KL_mean": -203.10101318359375, "KL/mean": -273.4295654296875, "KL/rejected_KL_mean": -343.75811767578125, "KL/std": 140.24241638183594, "epoch": 0.4287812041116006, "fcm_dpo/beta": 0.005930889863520861, "fcm_dpo/delta": -0.03794461488723755, "fcm_dpo/margin": 140.6571044921875, "fcm_dpo/q_t": 0.32827457785606384, "grad_norm": 52.17742156982422, "learning_rate": 3.5449008622169583e-07, "logits/chosen": -0.009431811049580574, "logits/rejected": 0.00803426280617714, "logps/chosen": -257.01953125, "logps/ref_chosen": -53.91852951049805, "logps/ref_rejected": -89.96138000488281, "logps/rejected": -433.7195129394531, "loss": 0.9046, "margin_dpo/margin_mean": 140.6571044921875, "margin_dpo/margin_std": 160.97979736328125, "step": 292 }, { "KL/chosen_KL_mean": -186.6754913330078, "KL/mean": -246.67132568359375, "KL/rejected_KL_mean": -306.66717529296875, "KL/std": 132.06301879882812, "epoch": 0.4302496328928047, "fcm_dpo/beta": 0.005984225310385227, "fcm_dpo/delta": 0.08749930560588837, "fcm_dpo/margin": 119.99168395996094, "fcm_dpo/q_t": 0.3568994402885437, "grad_norm": 45.794559478759766, "learning_rate": 3.5332285359726846e-07, "logits/chosen": -0.056711144745349884, "logits/rejected": -0.039980124682188034, "logps/chosen": -247.051513671875, "logps/ref_chosen": -60.376033782958984, "logps/ref_rejected": -77.85244750976562, "logps/rejected": -384.5196228027344, "loss": 0.9677, "margin_dpo/margin_mean": 119.99169158935547, "margin_dpo/margin_std": 153.38482666015625, "step": 293 }, { "KL/chosen_KL_mean": -164.24774169921875, "KL/mean": -227.4092559814453, "KL/rejected_KL_mean": -290.5707702636719, "KL/std": 132.69287109375, "epoch": 0.43171806167400884, "fcm_dpo/beta": 0.0059990473091602325, "fcm_dpo/delta": 0.04543343558907509, "fcm_dpo/margin": 126.32302856445312, "fcm_dpo/q_t": 0.35002630949020386, "grad_norm": 38.24577713012695, "learning_rate": 3.5215289831955786e-07, "logits/chosen": -0.07339977473020554, "logits/rejected": -0.07351949065923691, "logps/chosen": -212.3352813720703, "logps/ref_chosen": -48.0875358581543, "logps/ref_rejected": -81.89698791503906, "logps/rejected": -372.4677734375, "loss": 0.9284, "margin_dpo/margin_mean": 126.32302856445312, "margin_dpo/margin_std": 151.93936157226562, "step": 294 }, { "KL/chosen_KL_mean": -187.2119598388672, "KL/mean": -249.87669372558594, "KL/rejected_KL_mean": -312.54144287109375, "KL/std": 127.01498413085938, "epoch": 0.4331864904552129, "fcm_dpo/beta": 0.0060641467571258545, "fcm_dpo/delta": 0.04325701668858528, "fcm_dpo/margin": 125.32946014404297, "fcm_dpo/q_t": 0.3435119688510895, "grad_norm": 37.575748443603516, "learning_rate": 3.509802512179737e-07, "logits/chosen": -0.06643851101398468, "logits/rejected": -0.06460548937320709, "logps/chosen": -237.13662719726562, "logps/ref_chosen": -49.92467498779297, "logps/ref_rejected": -87.45632934570312, "logps/rejected": -399.99774169921875, "loss": 0.9108, "margin_dpo/margin_mean": 125.3294677734375, "margin_dpo/margin_std": 138.71270751953125, "step": 295 }, { "KL/chosen_KL_mean": -210.7562255859375, "KL/mean": -261.470947265625, "KL/rejected_KL_mean": -312.18572998046875, "KL/std": 124.9061508178711, "epoch": 0.434654919236417, "fcm_dpo/beta": 0.0062162261456251144, "fcm_dpo/delta": 0.1811649054288864, "fcm_dpo/margin": 101.42951965332031, "fcm_dpo/q_t": 0.37338194251060486, "grad_norm": 66.42903137207031, "learning_rate": 3.498049431928577e-07, "logits/chosen": -0.0994357317686081, "logits/rejected": -0.07531915605068207, "logps/chosen": -276.2474365234375, "logps/ref_chosen": -65.49124145507812, "logps/ref_rejected": -93.08908081054688, "logps/rejected": -405.27484130859375, "loss": 1.0423, "margin_dpo/margin_mean": 101.42951965332031, "margin_dpo/margin_std": 150.9395294189453, "step": 296 }, { "KL/chosen_KL_mean": -186.30535888671875, "KL/mean": -245.25868225097656, "KL/rejected_KL_mean": -304.2120056152344, "KL/std": 122.74223327636719, "epoch": 0.43612334801762115, "fcm_dpo/beta": 0.006364773027598858, "fcm_dpo/delta": 0.05358999967575073, "fcm_dpo/margin": 117.90662384033203, "fcm_dpo/q_t": 0.34062981605529785, "grad_norm": 36.62229919433594, "learning_rate": 3.486270052146694e-07, "logits/chosen": -0.14602035284042358, "logits/rejected": -0.14422789216041565, "logps/chosen": -242.78231811523438, "logps/ref_chosen": -56.476951599121094, "logps/ref_rejected": -95.1385498046875, "logps/rejected": -399.3505554199219, "loss": 0.8876, "margin_dpo/margin_mean": 117.90663146972656, "margin_dpo/margin_std": 117.52006530761719, "step": 297 }, { "KL/chosen_KL_mean": -183.15670776367188, "KL/mean": -253.70401000976562, "KL/rejected_KL_mean": -324.25128173828125, "KL/std": 148.4191436767578, "epoch": 0.43759177679882527, "fcm_dpo/beta": 0.006309167016297579, "fcm_dpo/delta": -0.09888706356287003, "fcm_dpo/margin": 141.0946044921875, "fcm_dpo/q_t": 0.3272933065891266, "grad_norm": 44.55033493041992, "learning_rate": 3.474464683231698e-07, "logits/chosen": -0.1122806966304779, "logits/rejected": -0.13088667392730713, "logps/chosen": -250.48187255859375, "logps/ref_chosen": -67.32516479492188, "logps/ref_rejected": -116.66217041015625, "logps/rejected": -440.9134521484375, "loss": 0.8618, "margin_dpo/margin_mean": 141.0946044921875, "margin_dpo/margin_std": 156.77532958984375, "step": 298 }, { "KL/chosen_KL_mean": -160.49649047851562, "KL/mean": -227.01962280273438, "KL/rejected_KL_mean": -293.54278564453125, "KL/std": 127.14181518554688, "epoch": 0.4390602055800294, "fcm_dpo/beta": 0.0062877098098397255, "fcm_dpo/delta": -0.0417584553360939, "fcm_dpo/margin": 133.04627990722656, "fcm_dpo/q_t": 0.33548539876937866, "grad_norm": 44.61274337768555, "learning_rate": 3.462633636266041e-07, "logits/chosen": -0.17389070987701416, "logits/rejected": -0.16740745306015015, "logps/chosen": -209.45858764648438, "logps/ref_chosen": -48.96209716796875, "logps/ref_rejected": -84.32823944091797, "logps/rejected": -377.87103271484375, "loss": 0.9029, "margin_dpo/margin_mean": 133.04627990722656, "margin_dpo/margin_std": 153.0653076171875, "step": 299 }, { "KL/chosen_KL_mean": -201.64463806152344, "KL/mean": -279.50714111328125, "KL/rejected_KL_mean": -357.3695983886719, "KL/std": 150.42608642578125, "epoch": 0.44052863436123346, "fcm_dpo/beta": 0.00608530780300498, "fcm_dpo/delta": -0.16277390718460083, "fcm_dpo/margin": 155.72494506835938, "fcm_dpo/q_t": 0.3149811625480652, "grad_norm": 62.3836784362793, "learning_rate": 3.4507772230088147e-07, "logits/chosen": -0.12772464752197266, "logits/rejected": -0.11891786009073257, "logps/chosen": -260.7183532714844, "logps/ref_chosen": -59.07371139526367, "logps/ref_rejected": -95.9664535522461, "logps/rejected": -453.3360595703125, "loss": 0.8776, "margin_dpo/margin_mean": 155.72494506835938, "margin_dpo/margin_std": 176.29461669921875, "step": 300 }, { "KL/chosen_KL_mean": -177.3877716064453, "KL/mean": -250.5698699951172, "KL/rejected_KL_mean": -323.751953125, "KL/std": 135.59161376953125, "epoch": 0.4419970631424376, "fcm_dpo/beta": 0.005912788212299347, "fcm_dpo/delta": -0.0726335421204567, "fcm_dpo/margin": 146.36419677734375, "fcm_dpo/q_t": 0.3244306743144989, "grad_norm": 47.50827407836914, "learning_rate": 3.4388957558875316e-07, "logits/chosen": -0.10594043135643005, "logits/rejected": -0.09627512097358704, "logps/chosen": -234.63714599609375, "logps/ref_chosen": -57.249366760253906, "logps/ref_rejected": -92.35354614257812, "logps/rejected": -416.10552978515625, "loss": 0.8447, "margin_dpo/margin_mean": 146.36419677734375, "margin_dpo/margin_std": 145.9784393310547, "step": 301 }, { "KL/chosen_KL_mean": -150.9648895263672, "KL/mean": -216.1134033203125, "KL/rejected_KL_mean": -281.26190185546875, "KL/std": 123.28852844238281, "epoch": 0.4434654919236417, "fcm_dpo/beta": 0.005917476490139961, "fcm_dpo/delta": 0.03093942627310753, "fcm_dpo/margin": 130.2970428466797, "fcm_dpo/q_t": 0.3413656949996948, "grad_norm": 45.89029312133789, "learning_rate": 3.426989547989902e-07, "logits/chosen": -0.13110940158367157, "logits/rejected": -0.12976717948913574, "logps/chosen": -202.16287231445312, "logps/ref_chosen": -51.197994232177734, "logps/ref_rejected": -97.22636413574219, "logps/rejected": -378.48828125, "loss": 0.9054, "margin_dpo/margin_mean": 130.2970428466797, "margin_dpo/margin_std": 143.6829071044922, "step": 302 }, { "KL/chosen_KL_mean": -161.89434814453125, "KL/mean": -225.97076416015625, "KL/rejected_KL_mean": -290.04718017578125, "KL/std": 136.81471252441406, "epoch": 0.44493392070484583, "fcm_dpo/beta": 0.005963338539004326, "fcm_dpo/delta": 0.03840842843055725, "fcm_dpo/margin": 128.15284729003906, "fcm_dpo/q_t": 0.34974241256713867, "grad_norm": 39.980674743652344, "learning_rate": 3.4150589130555773e-07, "logits/chosen": -0.12430644035339355, "logits/rejected": -0.09778697788715363, "logps/chosen": -228.60829162597656, "logps/ref_chosen": -66.71394348144531, "logps/ref_rejected": -86.94542694091797, "logps/rejected": -376.99261474609375, "loss": 0.9415, "margin_dpo/margin_mean": 128.15284729003906, "margin_dpo/margin_std": 157.72259521484375, "step": 303 }, { "KL/chosen_KL_mean": -149.53477478027344, "KL/mean": -216.70220947265625, "KL/rejected_KL_mean": -283.86962890625, "KL/std": 127.11853790283203, "epoch": 0.44640234948604995, "fcm_dpo/beta": 0.006036032922565937, "fcm_dpo/delta": -0.01273266226053238, "fcm_dpo/margin": 134.3348388671875, "fcm_dpo/q_t": 0.3229079246520996, "grad_norm": 56.761173248291016, "learning_rate": 3.403104165467883e-07, "logits/chosen": -0.13418614864349365, "logits/rejected": -0.11346621066331863, "logps/chosen": -221.4854736328125, "logps/ref_chosen": -71.95069885253906, "logps/ref_rejected": -90.47203063964844, "logps/rejected": -374.3416442871094, "loss": 0.829, "margin_dpo/margin_mean": 134.3348388671875, "margin_dpo/margin_std": 107.37701416015625, "step": 304 }, { "KL/chosen_KL_mean": -170.97412109375, "KL/mean": -234.58453369140625, "KL/rejected_KL_mean": -298.1949462890625, "KL/std": 147.8951416015625, "epoch": 0.447870778267254, "fcm_dpo/beta": 0.005968155339360237, "fcm_dpo/delta": 0.04241461306810379, "fcm_dpo/margin": 127.22084045410156, "fcm_dpo/q_t": 0.34988775849342346, "grad_norm": 41.87591552734375, "learning_rate": 3.391125620245535e-07, "logits/chosen": -0.14380377531051636, "logits/rejected": -0.11721345782279968, "logps/chosen": -237.76934814453125, "logps/ref_chosen": -66.79523468017578, "logps/ref_rejected": -92.75459289550781, "logps/rejected": -390.94952392578125, "loss": 0.9292, "margin_dpo/margin_mean": 127.22084045410156, "margin_dpo/margin_std": 151.81414794921875, "step": 305 }, { "KL/chosen_KL_mean": -176.7681427001953, "KL/mean": -238.14883422851562, "KL/rejected_KL_mean": -299.529541015625, "KL/std": 127.993408203125, "epoch": 0.44933920704845814, "fcm_dpo/beta": 0.006088206544518471, "fcm_dpo/delta": 0.05688028782606125, "fcm_dpo/margin": 122.76139831542969, "fcm_dpo/q_t": 0.3437831401824951, "grad_norm": 42.77898025512695, "learning_rate": 3.3791235930343417e-07, "logits/chosen": -0.08690765500068665, "logits/rejected": -0.056158192455768585, "logps/chosen": -246.45204162597656, "logps/ref_chosen": -69.68389892578125, "logps/ref_rejected": -85.15919494628906, "logps/rejected": -384.688720703125, "loss": 0.9036, "margin_dpo/margin_mean": 122.76140594482422, "margin_dpo/margin_std": 131.52952575683594, "step": 306 }, { "KL/chosen_KL_mean": -173.75515747070312, "KL/mean": -236.67062377929688, "KL/rejected_KL_mean": -299.5860595703125, "KL/std": 121.59797668457031, "epoch": 0.45080763582966227, "fcm_dpo/beta": 0.0061085280030965805, "fcm_dpo/delta": 0.03357131406664848, "fcm_dpo/margin": 125.83091735839844, "fcm_dpo/q_t": 0.3417189121246338, "grad_norm": 44.65004348754883, "learning_rate": 3.367098400098881e-07, "logits/chosen": -0.06562909483909607, "logits/rejected": -0.04333928972482681, "logps/chosen": -243.92059326171875, "logps/ref_chosen": -70.16542053222656, "logps/ref_rejected": -86.97230529785156, "logps/rejected": -386.5583801269531, "loss": 0.9153, "margin_dpo/margin_mean": 125.83091735839844, "margin_dpo/margin_std": 141.49557495117188, "step": 307 }, { "KL/chosen_KL_mean": -178.3945770263672, "KL/mean": -248.44857788085938, "KL/rejected_KL_mean": -318.5025634765625, "KL/std": 136.55337524414062, "epoch": 0.4522760646108664, "fcm_dpo/beta": 0.00610099732875824, "fcm_dpo/delta": -0.059890978038311005, "fcm_dpo/margin": 140.10801696777344, "fcm_dpo/q_t": 0.3226816654205322, "grad_norm": 44.11930847167969, "learning_rate": 3.355050358314172e-07, "logits/chosen": -0.03536057472229004, "logits/rejected": -0.012509562075138092, "logps/chosen": -233.6395721435547, "logps/ref_chosen": -55.2449951171875, "logps/ref_rejected": -79.37226104736328, "logps/rejected": -397.8748474121094, "loss": 0.8489, "margin_dpo/margin_mean": 140.10801696777344, "margin_dpo/margin_std": 136.1858367919922, "step": 308 }, { "KL/chosen_KL_mean": -182.00643920898438, "KL/mean": -252.7855224609375, "KL/rejected_KL_mean": -323.56463623046875, "KL/std": 137.23944091796875, "epoch": 0.45374449339207046, "fcm_dpo/beta": 0.006070786155760288, "fcm_dpo/delta": -0.06589590013027191, "fcm_dpo/margin": 141.55816650390625, "fcm_dpo/q_t": 0.3239052891731262, "grad_norm": 40.19969177246094, "learning_rate": 3.3429797851573183e-07, "logits/chosen": -0.010355365462601185, "logits/rejected": 0.010411388240754604, "logps/chosen": -230.9655303955078, "logps/ref_chosen": -48.959083557128906, "logps/ref_rejected": -82.34072875976562, "logps/rejected": -405.90533447265625, "loss": 0.8536, "margin_dpo/margin_mean": 141.5581512451172, "margin_dpo/margin_std": 140.39210510253906, "step": 309 }, { "KL/chosen_KL_mean": -226.34396362304688, "KL/mean": -292.3310546875, "KL/rejected_KL_mean": -358.31805419921875, "KL/std": 120.46112060546875, "epoch": 0.4552129221732746, "fcm_dpo/beta": 0.006043789908289909, "fcm_dpo/delta": 0.001462891697883606, "fcm_dpo/margin": 131.97410583496094, "fcm_dpo/q_t": 0.3287044167518616, "grad_norm": 69.3414535522461, "learning_rate": 3.3308869986991487e-07, "logits/chosen": 0.0008979048579931259, "logits/rejected": 0.024873077869415283, "logps/chosen": -289.08575439453125, "logps/ref_chosen": -62.74177932739258, "logps/ref_rejected": -79.9302978515625, "logps/rejected": -438.2483825683594, "loss": 0.8421, "margin_dpo/margin_mean": 131.97410583496094, "margin_dpo/margin_std": 113.02054595947266, "step": 310 }, { "KL/chosen_KL_mean": -230.95571899414062, "KL/mean": -309.91021728515625, "KL/rejected_KL_mean": -388.86474609375, "KL/std": 167.3114013671875, "epoch": 0.4566813509544787, "fcm_dpo/beta": 0.005860151723027229, "fcm_dpo/delta": -0.1387987583875656, "fcm_dpo/margin": 157.90902709960938, "fcm_dpo/q_t": 0.3198055028915405, "grad_norm": 36.636531829833984, "learning_rate": 3.3187723175958346e-07, "logits/chosen": -0.007311457768082619, "logits/rejected": 0.02352697029709816, "logps/chosen": -283.98370361328125, "logps/ref_chosen": -53.02798080444336, "logps/ref_rejected": -77.43820190429688, "logps/rejected": -466.3029479980469, "loss": 0.85, "margin_dpo/margin_mean": 157.90902709960938, "margin_dpo/margin_std": 170.51254272460938, "step": 311 }, { "KL/chosen_KL_mean": -220.95730590820312, "KL/mean": -290.677001953125, "KL/rejected_KL_mean": -360.3967590332031, "KL/std": 133.24356079101562, "epoch": 0.4581497797356828, "fcm_dpo/beta": 0.005845385603606701, "fcm_dpo/delta": -0.01669427752494812, "fcm_dpo/margin": 139.439453125, "fcm_dpo/q_t": 0.33251118659973145, "grad_norm": 41.69878005981445, "learning_rate": 3.306636061080487e-07, "logits/chosen": 0.03852830454707146, "logits/rejected": 0.05525093898177147, "logps/chosen": -270.3495178222656, "logps/ref_chosen": -49.39221954345703, "logps/ref_rejected": -75.79280853271484, "logps/rejected": -436.1895751953125, "loss": 0.8841, "margin_dpo/margin_mean": 139.439453125, "margin_dpo/margin_std": 151.3997039794922, "step": 312 }, { "KL/chosen_KL_mean": -219.05848693847656, "KL/mean": -286.2684020996094, "KL/rejected_KL_mean": -353.47833251953125, "KL/std": 144.6976318359375, "epoch": 0.45961820851688695, "fcm_dpo/beta": 0.005889068823307753, "fcm_dpo/delta": 0.006327494978904724, "fcm_dpo/margin": 134.4198455810547, "fcm_dpo/q_t": 0.3411150276660919, "grad_norm": 43.50709533691406, "learning_rate": 3.2944785489547537e-07, "logits/chosen": 0.026514366269111633, "logits/rejected": 0.03467664122581482, "logps/chosen": -269.21124267578125, "logps/ref_chosen": -50.152740478515625, "logps/ref_rejected": -86.40620422363281, "logps/rejected": -439.884521484375, "loss": 0.932, "margin_dpo/margin_mean": 134.41983032226562, "margin_dpo/margin_std": 159.38235473632812, "step": 313 }, { "KL/chosen_KL_mean": -235.50747680664062, "KL/mean": -299.62200927734375, "KL/rejected_KL_mean": -363.73651123046875, "KL/std": 141.76861572265625, "epoch": 0.461086637298091, "fcm_dpo/beta": 0.005829677917063236, "fcm_dpo/delta": 0.05599457398056984, "fcm_dpo/margin": 128.22901916503906, "fcm_dpo/q_t": 0.35217374563217163, "grad_norm": 33.33452224731445, "learning_rate": 3.2823001015803857e-07, "logits/chosen": -0.06579890102148056, "logits/rejected": -0.06030385196208954, "logps/chosen": -292.74505615234375, "logps/ref_chosen": -57.237579345703125, "logps/ref_rejected": -97.5965347290039, "logps/rejected": -461.3330383300781, "loss": 0.9653, "margin_dpo/margin_mean": 128.22901916503906, "margin_dpo/margin_std": 166.1632080078125, "step": 314 }, { "KL/chosen_KL_mean": -230.164306640625, "KL/mean": -291.27130126953125, "KL/rejected_KL_mean": -352.3782958984375, "KL/std": 119.44525146484375, "epoch": 0.46255506607929514, "fcm_dpo/beta": 0.005951759405434132, "fcm_dpo/delta": 0.07836839556694031, "fcm_dpo/margin": 122.21400451660156, "fcm_dpo/q_t": 0.3487578332424164, "grad_norm": 51.986087799072266, "learning_rate": 3.270101039870797e-07, "logits/chosen": 0.05683774873614311, "logits/rejected": 0.0601893775165081, "logps/chosen": -279.23388671875, "logps/ref_chosen": -49.06958770751953, "logps/ref_rejected": -85.68087768554688, "logps/rejected": -438.0592041015625, "loss": 0.9161, "margin_dpo/margin_mean": 122.21400451660156, "margin_dpo/margin_std": 134.08251953125, "step": 315 }, { "KL/chosen_KL_mean": -211.80429077148438, "KL/mean": -290.564208984375, "KL/rejected_KL_mean": -369.32415771484375, "KL/std": 143.4363250732422, "epoch": 0.46402349486049926, "fcm_dpo/beta": 0.005822393111884594, "fcm_dpo/delta": -0.13124880194664001, "fcm_dpo/margin": 157.5198974609375, "fcm_dpo/q_t": 0.31461644172668457, "grad_norm": 42.63252258300781, "learning_rate": 3.2578816852826086e-07, "logits/chosen": -0.016533054411411285, "logits/rejected": -0.012745167128741741, "logps/chosen": -266.0650329589844, "logps/ref_chosen": -54.26074981689453, "logps/ref_rejected": -101.2814712524414, "logps/rejected": -470.60565185546875, "loss": 0.8139, "margin_dpo/margin_mean": 157.5198974609375, "margin_dpo/margin_std": 150.1714324951172, "step": 316 }, { "KL/chosen_KL_mean": -211.88148498535156, "KL/mean": -298.3778991699219, "KL/rejected_KL_mean": -384.8742980957031, "KL/std": 135.2079620361328, "epoch": 0.4654919236417034, "fcm_dpo/beta": 0.005675586871802807, "fcm_dpo/delta": -0.20120689272880554, "fcm_dpo/margin": 172.99285888671875, "fcm_dpo/q_t": 0.29535406827926636, "grad_norm": 43.179039001464844, "learning_rate": 3.2456423598071783e-07, "logits/chosen": -0.02477482333779335, "logits/rejected": -0.006942405365407467, "logps/chosen": -267.9757080078125, "logps/ref_chosen": -56.094207763671875, "logps/ref_rejected": -100.69905090332031, "logps/rejected": -485.5733642578125, "loss": 0.7599, "margin_dpo/margin_mean": 172.9928436279297, "margin_dpo/margin_std": 140.1666259765625, "step": 317 }, { "KL/chosen_KL_mean": -212.18310546875, "KL/mean": -289.941162109375, "KL/rejected_KL_mean": -367.69927978515625, "KL/std": 141.094970703125, "epoch": 0.4669603524229075, "fcm_dpo/beta": 0.005568217486143112, "fcm_dpo/delta": -0.07241801917552948, "fcm_dpo/margin": 155.5161590576172, "fcm_dpo/q_t": 0.31813758611679077, "grad_norm": 38.82643508911133, "learning_rate": 3.233383385962115e-07, "logits/chosen": -0.08865199983119965, "logits/rejected": -0.057291001081466675, "logps/chosen": -276.82879638671875, "logps/ref_chosen": -64.64569854736328, "logps/ref_rejected": -82.76425170898438, "logps/rejected": -450.4635009765625, "loss": 0.8259, "margin_dpo/margin_mean": 155.51617431640625, "margin_dpo/margin_std": 140.63192749023438, "step": 318 }, { "KL/chosen_KL_mean": -218.0725860595703, "KL/mean": -295.23101806640625, "KL/rejected_KL_mean": -372.389404296875, "KL/std": 148.6660919189453, "epoch": 0.4684287812041116, "fcm_dpo/beta": 0.005480584688484669, "fcm_dpo/delta": -0.04994754120707512, "fcm_dpo/margin": 154.31683349609375, "fcm_dpo/q_t": 0.32287895679473877, "grad_norm": 31.670095443725586, "learning_rate": 3.2211050867827805e-07, "logits/chosen": -0.042262256145477295, "logits/rejected": -0.05314238369464874, "logps/chosen": -267.45635986328125, "logps/ref_chosen": -49.383758544921875, "logps/ref_rejected": -113.90650939941406, "logps/rejected": -486.2959289550781, "loss": 0.8427, "margin_dpo/margin_mean": 154.31683349609375, "margin_dpo/margin_std": 146.704833984375, "step": 319 }, { "KL/chosen_KL_mean": -237.90200805664062, "KL/mean": -325.74908447265625, "KL/rejected_KL_mean": -413.5962219238281, "KL/std": 172.5784149169922, "epoch": 0.4698972099853157, "fcm_dpo/beta": 0.005322999320924282, "fcm_dpo/delta": -0.15001486241817474, "fcm_dpo/margin": 175.69418334960938, "fcm_dpo/q_t": 0.3112744092941284, "grad_norm": 39.102046966552734, "learning_rate": 3.208807785813777e-07, "logits/chosen": -0.01164646353572607, "logits/rejected": -0.012258168309926987, "logps/chosen": -297.4068908691406, "logps/ref_chosen": -59.50489044189453, "logps/ref_rejected": -97.66717529296875, "logps/rejected": -511.26336669921875, "loss": 0.8104, "margin_dpo/margin_mean": 175.69418334960938, "margin_dpo/margin_std": 170.92335510253906, "step": 320 }, { "KL/chosen_KL_mean": -300.34954833984375, "KL/mean": -372.80084228515625, "KL/rejected_KL_mean": -445.2521667480469, "KL/std": 164.09548950195312, "epoch": 0.4713656387665198, "fcm_dpo/beta": 0.005263539031147957, "fcm_dpo/delta": 0.03748725354671478, "fcm_dpo/margin": 144.90261840820312, "fcm_dpo/q_t": 0.3461604416370392, "grad_norm": 42.18912124633789, "learning_rate": 3.1964918071004217e-07, "logits/chosen": 0.18075895309448242, "logits/rejected": 0.189798966050148, "logps/chosen": -361.898193359375, "logps/ref_chosen": -61.548683166503906, "logps/ref_rejected": -91.64103698730469, "logps/rejected": -536.8931884765625, "loss": 0.9492, "margin_dpo/margin_mean": 144.90261840820312, "margin_dpo/margin_std": 174.24356079101562, "step": 321 }, { "KL/chosen_KL_mean": -268.3616638183594, "KL/mean": -350.6944580078125, "KL/rejected_KL_mean": -433.0272216796875, "KL/std": 155.32608032226562, "epoch": 0.47283406754772395, "fcm_dpo/beta": 0.005247790366411209, "fcm_dpo/delta": -0.0717114806175232, "fcm_dpo/margin": 164.66555786132812, "fcm_dpo/q_t": 0.32040244340896606, "grad_norm": 47.565155029296875, "learning_rate": 3.184157475180207e-07, "logits/chosen": 0.15934163331985474, "logits/rejected": 0.16657748818397522, "logps/chosen": -325.6517028808594, "logps/ref_chosen": -57.29003143310547, "logps/ref_rejected": -95.74992370605469, "logps/rejected": -528.7771606445312, "loss": 0.8352, "margin_dpo/margin_mean": 164.66555786132812, "margin_dpo/margin_std": 153.64788818359375, "step": 322 }, { "KL/chosen_KL_mean": -287.80499267578125, "KL/mean": -361.19696044921875, "KL/rejected_KL_mean": -434.5888671875, "KL/std": 157.22926330566406, "epoch": 0.47430249632892807, "fcm_dpo/beta": 0.005266541615128517, "fcm_dpo/delta": 0.029003996402025223, "fcm_dpo/margin": 146.7838897705078, "fcm_dpo/q_t": 0.33969342708587646, "grad_norm": 94.16871643066406, "learning_rate": 3.171805115074251e-07, "logits/chosen": 0.12828153371810913, "logits/rejected": 0.13148348033428192, "logps/chosen": -339.0389404296875, "logps/ref_chosen": -51.23395919799805, "logps/ref_rejected": -75.06192016601562, "logps/rejected": -509.6507873535156, "loss": 0.9039, "margin_dpo/margin_mean": 146.7838897705078, "margin_dpo/margin_std": 155.2939910888672, "step": 323 }, { "KL/chosen_KL_mean": -308.6185302734375, "KL/mean": -386.80194091796875, "KL/rejected_KL_mean": -464.9853820800781, "KL/std": 166.91270446777344, "epoch": 0.47577092511013214, "fcm_dpo/beta": 0.005319544579833746, "fcm_dpo/delta": -0.036364443600177765, "fcm_dpo/margin": 156.36685180664062, "fcm_dpo/q_t": 0.33428698778152466, "grad_norm": 102.16787719726562, "learning_rate": 3.1594350522787295e-07, "logits/chosen": 0.04309145733714104, "logits/rejected": 0.06093376502394676, "logps/chosen": -373.75372314453125, "logps/ref_chosen": -65.13516998291016, "logps/ref_rejected": -86.47750854492188, "logps/rejected": -551.462890625, "loss": 0.9097, "margin_dpo/margin_mean": 156.36685180664062, "margin_dpo/margin_std": 177.79598999023438, "step": 324 }, { "KL/chosen_KL_mean": -250.1693878173828, "KL/mean": -315.89727783203125, "KL/rejected_KL_mean": -381.625244140625, "KL/std": 165.94528198242188, "epoch": 0.47723935389133626, "fcm_dpo/beta": 0.005339835770428181, "fcm_dpo/delta": 0.10544876754283905, "fcm_dpo/margin": 131.45582580566406, "fcm_dpo/q_t": 0.35164737701416016, "grad_norm": 72.66465759277344, "learning_rate": 3.147047612756302e-07, "logits/chosen": -0.03497472405433655, "logits/rejected": -0.006054544821381569, "logps/chosen": -306.3849792480469, "logps/ref_chosen": -56.215599060058594, "logps/ref_rejected": -70.08592987060547, "logps/rejected": -451.7111511230469, "loss": 0.9295, "margin_dpo/margin_mean": 131.45584106445312, "margin_dpo/margin_std": 139.95791625976562, "step": 325 }, { "KL/chosen_KL_mean": -265.7798767089844, "KL/mean": -331.59429931640625, "KL/rejected_KL_mean": -397.40869140625, "KL/std": 143.18637084960938, "epoch": 0.4787077826725404, "fcm_dpo/beta": 0.005434621125459671, "fcm_dpo/delta": 0.09125562757253647, "fcm_dpo/margin": 131.62879943847656, "fcm_dpo/q_t": 0.34775280952453613, "grad_norm": 52.132789611816406, "learning_rate": 3.134643122927519e-07, "logits/chosen": -0.028336074203252792, "logits/rejected": 0.004011590965092182, "logps/chosen": -338.5048522949219, "logps/ref_chosen": -72.72496032714844, "logps/ref_rejected": -79.8467788696289, "logps/rejected": -477.2554931640625, "loss": 0.8987, "margin_dpo/margin_mean": 131.62879943847656, "margin_dpo/margin_std": 130.20550537109375, "step": 326 }, { "KL/chosen_KL_mean": -220.91165161132812, "KL/mean": -315.80706787109375, "KL/rejected_KL_mean": -410.7025146484375, "KL/std": 157.95884704589844, "epoch": 0.4801762114537445, "fcm_dpo/beta": 0.005260833539068699, "fcm_dpo/delta": -0.2224833071231842, "fcm_dpo/margin": 189.79087829589844, "fcm_dpo/q_t": 0.29437848925590515, "grad_norm": 56.85773468017578, "learning_rate": 3.1222219096622264e-07, "logits/chosen": -0.03550358861684799, "logits/rejected": -0.011621898040175438, "logps/chosen": -290.04608154296875, "logps/ref_chosen": -69.13441467285156, "logps/ref_rejected": -111.93377685546875, "logps/rejected": -522.6362915039062, "loss": 0.7735, "margin_dpo/margin_mean": 189.7908935546875, "margin_dpo/margin_std": 165.75531005859375, "step": 327 }, { "KL/chosen_KL_mean": -234.7518768310547, "KL/mean": -312.504638671875, "KL/rejected_KL_mean": -390.25738525390625, "KL/std": 155.8070068359375, "epoch": 0.48164464023494863, "fcm_dpo/beta": 0.005238644778728485, "fcm_dpo/delta": -0.01621408388018608, "fcm_dpo/margin": 155.5054931640625, "fcm_dpo/q_t": 0.3311880826950073, "grad_norm": 46.42213439941406, "learning_rate": 3.1097843002709427e-07, "logits/chosen": -0.05805660039186478, "logits/rejected": -0.05620625987648964, "logps/chosen": -294.4390869140625, "logps/ref_chosen": -59.68719482421875, "logps/ref_rejected": -90.85499572753906, "logps/rejected": -481.11236572265625, "loss": 0.8629, "margin_dpo/margin_mean": 155.5054931640625, "margin_dpo/margin_std": 155.49652099609375, "step": 328 }, { "KL/chosen_KL_mean": -249.91339111328125, "KL/mean": -333.9342956542969, "KL/rejected_KL_mean": -417.9552001953125, "KL/std": 158.51559448242188, "epoch": 0.4831130690161527, "fcm_dpo/beta": 0.005131312645971775, "fcm_dpo/delta": -0.0691857784986496, "fcm_dpo/margin": 168.04180908203125, "fcm_dpo/q_t": 0.32326847314834595, "grad_norm": 43.19540023803711, "learning_rate": 3.0973306224962437e-07, "logits/chosen": 0.032189346849918365, "logits/rejected": 0.0457240529358387, "logps/chosen": -315.1595764160156, "logps/ref_chosen": -65.2461929321289, "logps/ref_rejected": -100.69770812988281, "logps/rejected": -518.6528930664062, "loss": 0.8604, "margin_dpo/margin_mean": 168.04180908203125, "margin_dpo/margin_std": 168.69622802734375, "step": 329 }, { "KL/chosen_KL_mean": -229.13954162597656, "KL/mean": -311.7605285644531, "KL/rejected_KL_mean": -394.3815002441406, "KL/std": 157.78042602539062, "epoch": 0.4845814977973568, "fcm_dpo/beta": 0.005100243724882603, "fcm_dpo/delta": -0.04678843915462494, "fcm_dpo/margin": 165.24192810058594, "fcm_dpo/q_t": 0.3230311870574951, "grad_norm": 47.458736419677734, "learning_rate": 3.084861204504122e-07, "logits/chosen": 0.039801113307476044, "logits/rejected": 0.03778820112347603, "logps/chosen": -276.13787841796875, "logps/ref_chosen": -46.998348236083984, "logps/ref_rejected": -86.87684631347656, "logps/rejected": -481.25836181640625, "loss": 0.8448, "margin_dpo/margin_mean": 165.241943359375, "margin_dpo/margin_std": 157.53878784179688, "step": 330 }, { "KL/chosen_KL_mean": -226.46022033691406, "KL/mean": -313.51763916015625, "KL/rejected_KL_mean": -400.5750732421875, "KL/std": 132.8838348388672, "epoch": 0.48604992657856094, "fcm_dpo/beta": 0.005020303651690483, "fcm_dpo/delta": -0.08152840286493301, "fcm_dpo/margin": 174.1148681640625, "fcm_dpo/q_t": 0.3074917197227478, "grad_norm": 47.758262634277344, "learning_rate": 3.072376374875335e-07, "logits/chosen": 0.02427823096513748, "logits/rejected": 0.03714999929070473, "logps/chosen": -276.9844665527344, "logps/ref_chosen": -50.52424621582031, "logps/ref_rejected": -89.01544189453125, "logps/rejected": -489.59051513671875, "loss": 0.7647, "margin_dpo/margin_mean": 174.1148681640625, "margin_dpo/margin_std": 117.05330657958984, "step": 331 }, { "KL/chosen_KL_mean": -238.39886474609375, "KL/mean": -301.0345458984375, "KL/rejected_KL_mean": -363.67022705078125, "KL/std": 137.32064819335938, "epoch": 0.48751835535976507, "fcm_dpo/beta": 0.005146116018295288, "fcm_dpo/delta": 0.16562792658805847, "fcm_dpo/margin": 125.2713623046875, "fcm_dpo/q_t": 0.362199991941452, "grad_norm": 37.70057678222656, "learning_rate": 3.059876462596758e-07, "logits/chosen": 0.062703438103199, "logits/rejected": 0.07360726594924927, "logps/chosen": -287.57916259765625, "logps/ref_chosen": -49.18028259277344, "logps/ref_rejected": -76.48515319824219, "logps/rejected": -440.1553955078125, "loss": 0.949, "margin_dpo/margin_mean": 125.2713623046875, "margin_dpo/margin_std": 136.968017578125, "step": 332 }, { "KL/chosen_KL_mean": -251.75677490234375, "KL/mean": -335.0605773925781, "KL/rejected_KL_mean": -418.3643798828125, "KL/std": 162.35433959960938, "epoch": 0.4889867841409692, "fcm_dpo/beta": 0.005091669969260693, "fcm_dpo/delta": -0.05500401183962822, "fcm_dpo/margin": 166.6075897216797, "fcm_dpo/q_t": 0.3314821720123291, "grad_norm": 37.29023742675781, "learning_rate": 3.0473617970527015e-07, "logits/chosen": 0.00028313975781202316, "logits/rejected": 0.0058424267917871475, "logps/chosen": -315.51251220703125, "logps/ref_chosen": -63.75574493408203, "logps/ref_rejected": -95.04411315917969, "logps/rejected": -513.408447265625, "loss": 0.8865, "margin_dpo/margin_mean": 166.6075897216797, "margin_dpo/margin_std": 180.63882446289062, "step": 333 }, { "KL/chosen_KL_mean": -250.1717529296875, "KL/mean": -329.30615234375, "KL/rejected_KL_mean": -408.4405822753906, "KL/std": 176.424072265625, "epoch": 0.49045521292217326, "fcm_dpo/beta": 0.005118607543408871, "fcm_dpo/delta": -0.011011124588549137, "fcm_dpo/margin": 158.2688446044922, "fcm_dpo/q_t": 0.33261972665786743, "grad_norm": 39.07373809814453, "learning_rate": 3.034832708016243e-07, "logits/chosen": -0.023045614361763, "logits/rejected": -0.018983149901032448, "logps/chosen": -317.1514892578125, "logps/ref_chosen": -66.97975158691406, "logps/ref_rejected": -95.31692504882812, "logps/rejected": -503.75750732421875, "loss": 0.8932, "margin_dpo/margin_mean": 158.26882934570312, "margin_dpo/margin_std": 170.21026611328125, "step": 334 }, { "KL/chosen_KL_mean": -293.2091979980469, "KL/mean": -356.44830322265625, "KL/rejected_KL_mean": -419.68743896484375, "KL/std": 160.33944702148438, "epoch": 0.4919236417033774, "fcm_dpo/beta": 0.005251425784081221, "fcm_dpo/delta": 0.1440257877111435, "fcm_dpo/margin": 126.47823333740234, "fcm_dpo/q_t": 0.36713701486587524, "grad_norm": 71.73628234863281, "learning_rate": 3.022289525640531e-07, "logits/chosen": 0.0950508713722229, "logits/rejected": 0.11580017954111099, "logps/chosen": -355.7516784667969, "logps/ref_chosen": -62.54248046875, "logps/ref_rejected": -87.61770629882812, "logps/rejected": -507.30511474609375, "loss": 1.0083, "margin_dpo/margin_mean": 126.47822570800781, "margin_dpo/margin_std": 176.1946258544922, "step": 335 }, { "KL/chosen_KL_mean": -285.6962585449219, "KL/mean": -374.312744140625, "KL/rejected_KL_mean": -462.92919921875, "KL/std": 179.97970581054688, "epoch": 0.4933920704845815, "fcm_dpo/beta": 0.005166906863451004, "fcm_dpo/delta": -0.1273384988307953, "fcm_dpo/margin": 177.23297119140625, "fcm_dpo/q_t": 0.31727689504623413, "grad_norm": 44.789981842041016, "learning_rate": 3.009732580450086e-07, "logits/chosen": 0.06262758374214172, "logits/rejected": 0.06866885721683502, "logps/chosen": -340.2274169921875, "logps/ref_chosen": -54.53115463256836, "logps/ref_rejected": -104.40424346923828, "logps/rejected": -567.33349609375, "loss": 0.8731, "margin_dpo/margin_mean": 177.23297119140625, "margin_dpo/margin_std": 196.13185119628906, "step": 336 }, { "KL/chosen_KL_mean": -257.1330871582031, "KL/mean": -352.3122253417969, "KL/rejected_KL_mean": -447.4913635253906, "KL/std": 158.125244140625, "epoch": 0.4948604992657856, "fcm_dpo/beta": 0.004982330836355686, "fcm_dpo/delta": -0.165202796459198, "fcm_dpo/margin": 190.35829162597656, "fcm_dpo/q_t": 0.30310046672821045, "grad_norm": 44.3165283203125, "learning_rate": 2.9971622033320914e-07, "logits/chosen": 0.0833229273557663, "logits/rejected": 0.098611980676651, "logps/chosen": -322.26177978515625, "logps/ref_chosen": -65.12869262695312, "logps/ref_rejected": -101.72701263427734, "logps/rejected": -549.2183837890625, "loss": 0.7904, "margin_dpo/margin_mean": 190.35829162597656, "margin_dpo/margin_std": 169.05369567871094, "step": 337 }, { "KL/chosen_KL_mean": -253.24179077148438, "KL/mean": -342.19232177734375, "KL/rejected_KL_mean": -431.1428527832031, "KL/std": 160.1546630859375, "epoch": 0.49632892804698975, "fcm_dpo/beta": 0.004904153756797314, "fcm_dpo/delta": -0.07938512414693832, "fcm_dpo/margin": 177.90106201171875, "fcm_dpo/q_t": 0.3190838396549225, "grad_norm": 40.79475784301758, "learning_rate": 2.984578725527675e-07, "logits/chosen": 0.05606719106435776, "logits/rejected": 0.0772060826420784, "logps/chosen": -311.66448974609375, "logps/ref_chosen": -58.422706604003906, "logps/ref_rejected": -89.06854248046875, "logps/rejected": -520.2113647460938, "loss": 0.8318, "margin_dpo/margin_mean": 177.90106201171875, "margin_dpo/margin_std": 167.9755096435547, "step": 338 }, { "KL/chosen_KL_mean": -238.2997283935547, "KL/mean": -323.3148193359375, "KL/rejected_KL_mean": -408.3299560546875, "KL/std": 160.2054443359375, "epoch": 0.4977973568281938, "fcm_dpo/beta": 0.0049000889994204044, "fcm_dpo/delta": -0.03744722157716751, "fcm_dpo/margin": 170.03021240234375, "fcm_dpo/q_t": 0.32219401001930237, "grad_norm": 38.09584045410156, "learning_rate": 2.9719824786231796e-07, "logits/chosen": 0.05355466157197952, "logits/rejected": 0.06788984686136246, "logps/chosen": -298.2950439453125, "logps/ref_chosen": -59.99531555175781, "logps/ref_rejected": -103.9109115600586, "logps/rejected": -512.2408447265625, "loss": 0.8261, "margin_dpo/margin_mean": 170.03021240234375, "margin_dpo/margin_std": 143.59585571289062, "step": 339 }, { "KL/chosen_KL_mean": -240.2142333984375, "KL/mean": -317.741455078125, "KL/rejected_KL_mean": -395.26861572265625, "KL/std": 146.14395141601562, "epoch": 0.49926578560939794, "fcm_dpo/beta": 0.0048608845099806786, "fcm_dpo/delta": 0.04984866455197334, "fcm_dpo/margin": 155.05438232421875, "fcm_dpo/q_t": 0.3420332968235016, "grad_norm": 35.83970260620117, "learning_rate": 2.959373794541426e-07, "logits/chosen": 0.038565389811992645, "logits/rejected": 0.06811020523309708, "logps/chosen": -293.0444641113281, "logps/ref_chosen": -52.83022689819336, "logps/ref_rejected": -73.10723114013672, "logps/rejected": -468.3758544921875, "loss": 0.8975, "margin_dpo/margin_mean": 155.0543975830078, "margin_dpo/margin_std": 162.0891876220703, "step": 340 }, { "KL/chosen_KL_mean": -218.0098876953125, "KL/mean": -304.40362548828125, "KL/rejected_KL_mean": -390.79736328125, "KL/std": 153.83116149902344, "epoch": 0.5007342143906021, "fcm_dpo/beta": 0.004849275108426809, "fcm_dpo/delta": -0.04175635054707527, "fcm_dpo/margin": 172.7874755859375, "fcm_dpo/q_t": 0.32173144817352295, "grad_norm": 34.084495544433594, "learning_rate": 2.946753005532965e-07, "logits/chosen": -0.0034986878745257854, "logits/rejected": -0.00451930146664381, "logps/chosen": -265.90966796875, "logps/ref_chosen": -47.899803161621094, "logps/ref_rejected": -101.80987548828125, "logps/rejected": -492.60723876953125, "loss": 0.8354, "margin_dpo/margin_mean": 172.7874755859375, "margin_dpo/margin_std": 155.39715576171875, "step": 341 }, { "KL/chosen_KL_mean": -207.13504028320312, "KL/mean": -284.558349609375, "KL/rejected_KL_mean": -361.981689453125, "KL/std": 152.72943115234375, "epoch": 0.5022026431718062, "fcm_dpo/beta": 0.004914829973131418, "fcm_dpo/delta": 0.041156478226184845, "fcm_dpo/margin": 154.84661865234375, "fcm_dpo/q_t": 0.3427678644657135, "grad_norm": 34.09379959106445, "learning_rate": 2.934120444167326e-07, "logits/chosen": -0.09954602271318436, "logits/rejected": -0.07372716814279556, "logps/chosen": -279.1316833496094, "logps/ref_chosen": -71.99664306640625, "logps/ref_rejected": -92.58959197998047, "logps/rejected": -454.5712585449219, "loss": 0.908, "margin_dpo/margin_mean": 154.84661865234375, "margin_dpo/margin_std": 171.0548095703125, "step": 342 }, { "KL/chosen_KL_mean": -189.46636962890625, "KL/mean": -279.3402099609375, "KL/rejected_KL_mean": -369.21405029296875, "KL/std": 155.5145263671875, "epoch": 0.5036710719530103, "fcm_dpo/beta": 0.004830049350857735, "fcm_dpo/delta": -0.07497724145650864, "fcm_dpo/margin": 179.74769592285156, "fcm_dpo/q_t": 0.31141358613967896, "grad_norm": 55.63039016723633, "learning_rate": 2.9214764433242476e-07, "logits/chosen": -0.04048309847712517, "logits/rejected": -0.02954668179154396, "logps/chosen": -243.87197875976562, "logps/ref_chosen": -54.405616760253906, "logps/ref_rejected": -111.04142761230469, "logps/rejected": -480.2554931640625, "loss": 0.7885, "margin_dpo/margin_mean": 179.74771118164062, "margin_dpo/margin_std": 134.98263549804688, "step": 343 }, { "KL/chosen_KL_mean": -204.9948272705078, "KL/mean": -283.34881591796875, "KL/rejected_KL_mean": -361.70281982421875, "KL/std": 170.766357421875, "epoch": 0.5051395007342144, "fcm_dpo/beta": 0.004901527427136898, "fcm_dpo/delta": 0.03143307566642761, "fcm_dpo/margin": 156.7080078125, "fcm_dpo/q_t": 0.34064382314682007, "grad_norm": 54.35795974731445, "learning_rate": 2.9088213361849126e-07, "logits/chosen": -0.04297008737921715, "logits/rejected": -0.03432103246450424, "logps/chosen": -258.95947265625, "logps/ref_chosen": -53.96466827392578, "logps/ref_rejected": -90.62336730957031, "logps/rejected": -452.3262023925781, "loss": 0.8886, "margin_dpo/margin_mean": 156.7080078125, "margin_dpo/margin_std": 158.40911865234375, "step": 344 }, { "KL/chosen_KL_mean": -249.43026733398438, "KL/mean": -335.73760986328125, "KL/rejected_KL_mean": -422.04498291015625, "KL/std": 163.56846618652344, "epoch": 0.5066079295154186, "fcm_dpo/beta": 0.004851485602557659, "fcm_dpo/delta": -0.04130681976675987, "fcm_dpo/margin": 172.61471557617188, "fcm_dpo/q_t": 0.32575783133506775, "grad_norm": 45.86104965209961, "learning_rate": 2.896155456223163e-07, "logits/chosen": -0.04765704274177551, "logits/rejected": -0.031171437352895737, "logps/chosen": -311.115966796875, "logps/ref_chosen": -61.685699462890625, "logps/ref_rejected": -99.49041748046875, "logps/rejected": -521.535400390625, "loss": 0.8542, "margin_dpo/margin_mean": 172.61473083496094, "margin_dpo/margin_std": 169.39071655273438, "step": 345 }, { "KL/chosen_KL_mean": -275.68389892578125, "KL/mean": -357.99700927734375, "KL/rejected_KL_mean": -440.31011962890625, "KL/std": 157.96783447265625, "epoch": 0.5080763582966226, "fcm_dpo/beta": 0.004806653130799532, "fcm_dpo/delta": 0.009312797337770462, "fcm_dpo/margin": 164.626220703125, "fcm_dpo/q_t": 0.33165764808654785, "grad_norm": 36.5350341796875, "learning_rate": 2.883479137196714e-07, "logits/chosen": 0.11625039577484131, "logits/rejected": 0.13555414974689484, "logps/chosen": -330.9401550292969, "logps/ref_chosen": -55.256263732910156, "logps/ref_rejected": -77.41532135009766, "logps/rejected": -517.7254638671875, "loss": 0.877, "margin_dpo/margin_mean": 164.626220703125, "margin_dpo/margin_std": 163.69869995117188, "step": 346 }, { "KL/chosen_KL_mean": -289.11468505859375, "KL/mean": -371.96826171875, "KL/rejected_KL_mean": -454.8218078613281, "KL/std": 172.33523559570312, "epoch": 0.5095447870778267, "fcm_dpo/beta": 0.004811136983335018, "fcm_dpo/delta": 0.002851152792572975, "fcm_dpo/margin": 165.7071075439453, "fcm_dpo/q_t": 0.3361477851867676, "grad_norm": 40.908966064453125, "learning_rate": 2.8707927131383614e-07, "logits/chosen": 0.10305792093276978, "logits/rejected": 0.11407352983951569, "logps/chosen": -346.6809387207031, "logps/ref_chosen": -57.56623840332031, "logps/ref_rejected": -92.35509490966797, "logps/rejected": -547.1768798828125, "loss": 0.9029, "margin_dpo/margin_mean": 165.7071075439453, "margin_dpo/margin_std": 183.31068420410156, "step": 347 }, { "KL/chosen_KL_mean": -262.79388427734375, "KL/mean": -337.61395263671875, "KL/rejected_KL_mean": -412.4340515136719, "KL/std": 159.94393920898438, "epoch": 0.5110132158590308, "fcm_dpo/beta": 0.004831107333302498, "fcm_dpo/delta": 0.08148808032274246, "fcm_dpo/margin": 149.6401824951172, "fcm_dpo/q_t": 0.3513612449169159, "grad_norm": 45.900245666503906, "learning_rate": 2.858096518347179e-07, "logits/chosen": 0.0368988960981369, "logits/rejected": 0.042798325419425964, "logps/chosen": -319.111572265625, "logps/ref_chosen": -56.31770324707031, "logps/ref_rejected": -89.13836669921875, "logps/rejected": -501.5724182128906, "loss": 0.9404, "margin_dpo/margin_mean": 149.6401824951172, "margin_dpo/margin_std": 172.48953247070312, "step": 348 }, { "KL/chosen_KL_mean": -276.16656494140625, "KL/mean": -357.14825439453125, "KL/rejected_KL_mean": -438.1300048828125, "KL/std": 166.79330444335938, "epoch": 0.5124816446402349, "fcm_dpo/beta": 0.004948951303958893, "fcm_dpo/delta": -0.0035505443811416626, "fcm_dpo/margin": 161.96343994140625, "fcm_dpo/q_t": 0.34265533089637756, "grad_norm": 38.599334716796875, "learning_rate": 2.845390887379706e-07, "logits/chosen": 0.12962684035301208, "logits/rejected": 0.1368744671344757, "logps/chosen": -334.19207763671875, "logps/ref_chosen": -58.025516510009766, "logps/ref_rejected": -97.50515747070312, "logps/rejected": -535.6351318359375, "loss": 0.9288, "margin_dpo/margin_mean": 161.96343994140625, "margin_dpo/margin_std": 197.3702392578125, "step": 349 }, { "KL/chosen_KL_mean": -290.5421142578125, "KL/mean": -370.68487548828125, "KL/rejected_KL_mean": -450.82763671875, "KL/std": 159.0634765625, "epoch": 0.5139500734214391, "fcm_dpo/beta": 0.004874521866440773, "fcm_dpo/delta": 0.019010931253433228, "fcm_dpo/margin": 160.2855224609375, "fcm_dpo/q_t": 0.33771812915802, "grad_norm": 52.47856521606445, "learning_rate": 2.8326761550411346e-07, "logits/chosen": 0.047298818826675415, "logits/rejected": 0.0552336648106575, "logps/chosen": -354.87261962890625, "logps/ref_chosen": -64.33049011230469, "logps/ref_rejected": -89.87164306640625, "logps/rejected": -540.6992797851562, "loss": 0.9145, "margin_dpo/margin_mean": 160.2855224609375, "margin_dpo/margin_std": 178.19192504882812, "step": 350 }, { "KL/chosen_KL_mean": -277.630859375, "KL/mean": -372.445556640625, "KL/rejected_KL_mean": -467.26031494140625, "KL/std": 192.2869415283203, "epoch": 0.5154185022026432, "fcm_dpo/beta": 0.004822996445000172, "fcm_dpo/delta": -0.12629525363445282, "fcm_dpo/margin": 189.62950134277344, "fcm_dpo/q_t": 0.31934210658073425, "grad_norm": 39.9083251953125, "learning_rate": 2.819952656376487e-07, "logits/chosen": 0.024956397712230682, "logits/rejected": 0.0348396897315979, "logps/chosen": -338.302978515625, "logps/ref_chosen": -60.6721305847168, "logps/ref_rejected": -101.5654296875, "logps/rejected": -568.8257446289062, "loss": 0.8645, "margin_dpo/margin_mean": 189.62950134277344, "margin_dpo/margin_std": 208.29437255859375, "step": 351 }, { "KL/chosen_KL_mean": -329.90045166015625, "KL/mean": -390.7127685546875, "KL/rejected_KL_mean": -451.52508544921875, "KL/std": 176.01547241210938, "epoch": 0.5168869309838473, "fcm_dpo/beta": 0.004911536350846291, "fcm_dpo/delta": 0.21583901345729828, "fcm_dpo/margin": 121.62461853027344, "fcm_dpo/q_t": 0.3810230791568756, "grad_norm": 78.79871368408203, "learning_rate": 2.8072207266617854e-07, "logits/chosen": 0.135191410779953, "logits/rejected": 0.1668703854084015, "logps/chosen": -400.8439025878906, "logps/ref_chosen": -70.9434585571289, "logps/ref_rejected": -76.6419677734375, "logps/rejected": -528.1670532226562, "loss": 1.0591, "margin_dpo/margin_mean": 121.62461853027344, "margin_dpo/margin_std": 185.77769470214844, "step": 352 }, { "KL/chosen_KL_mean": -304.80230712890625, "KL/mean": -384.9066162109375, "KL/rejected_KL_mean": -465.01092529296875, "KL/std": 172.11077880859375, "epoch": 0.5183553597650514, "fcm_dpo/beta": 0.004962503910064697, "fcm_dpo/delta": 0.003439009189605713, "fcm_dpo/margin": 160.20860290527344, "fcm_dpo/q_t": 0.3434601128101349, "grad_norm": 39.81656265258789, "learning_rate": 2.794480701395219e-07, "logits/chosen": 0.17260809242725372, "logits/rejected": 0.19139912724494934, "logps/chosen": -363.1976623535156, "logps/ref_chosen": -58.39533996582031, "logps/ref_rejected": -80.33553314208984, "logps/rejected": -545.346435546875, "loss": 0.9295, "margin_dpo/margin_mean": 160.20860290527344, "margin_dpo/margin_std": 190.71505737304688, "step": 353 }, { "KL/chosen_KL_mean": -259.389404296875, "KL/mean": -348.7931823730469, "KL/rejected_KL_mean": -438.19696044921875, "KL/std": 158.2752685546875, "epoch": 0.5198237885462555, "fcm_dpo/beta": 0.004931221716105938, "fcm_dpo/delta": -0.09008358418941498, "fcm_dpo/margin": 178.80755615234375, "fcm_dpo/q_t": 0.31218570470809937, "grad_norm": 39.17101287841797, "learning_rate": 2.781732916288303e-07, "logits/chosen": 0.1611638367176056, "logits/rejected": 0.18379822373390198, "logps/chosen": -319.1923828125, "logps/ref_chosen": -59.80299377441406, "logps/ref_rejected": -88.75750732421875, "logps/rejected": -526.9544677734375, "loss": 0.7976, "margin_dpo/margin_mean": 178.80755615234375, "margin_dpo/margin_std": 147.66680908203125, "step": 354 }, { "KL/chosen_KL_mean": -251.05117797851562, "KL/mean": -341.536865234375, "KL/rejected_KL_mean": -432.0225524902344, "KL/std": 175.2047119140625, "epoch": 0.5212922173274597, "fcm_dpo/beta": 0.004824481904506683, "fcm_dpo/delta": -0.08206754922866821, "fcm_dpo/margin": 180.97137451171875, "fcm_dpo/q_t": 0.31630951166152954, "grad_norm": 51.323665618896484, "learning_rate": 2.7689777072570284e-07, "logits/chosen": 0.11667881906032562, "logits/rejected": 0.14499685168266296, "logps/chosen": -305.1796875, "logps/ref_chosen": -54.12849807739258, "logps/ref_rejected": -82.40606689453125, "logps/rejected": -514.4285888671875, "loss": 0.8141, "margin_dpo/margin_mean": 180.97137451171875, "margin_dpo/margin_std": 157.3625030517578, "step": 355 }, { "KL/chosen_KL_mean": -290.5279541015625, "KL/mean": -346.39935302734375, "KL/rejected_KL_mean": -402.2707214355469, "KL/std": 172.05081176757812, "epoch": 0.5227606461086637, "fcm_dpo/beta": 0.005076043773442507, "fcm_dpo/delta": 0.24142667651176453, "fcm_dpo/margin": 111.74278259277344, "fcm_dpo/q_t": 0.39116936922073364, "grad_norm": 69.66039276123047, "learning_rate": 2.7562154104130176e-07, "logits/chosen": 0.11355781555175781, "logits/rejected": 0.12858667969703674, "logps/chosen": -355.2017517089844, "logps/ref_chosen": -64.6738052368164, "logps/ref_rejected": -75.89926147460938, "logps/rejected": -478.16998291015625, "loss": 1.1223, "margin_dpo/margin_mean": 111.74279022216797, "margin_dpo/margin_std": 199.85433959960938, "step": 356 }, { "KL/chosen_KL_mean": -250.232177734375, "KL/mean": -336.46234130859375, "KL/rejected_KL_mean": -422.6925048828125, "KL/std": 173.98196411132812, "epoch": 0.5242290748898678, "fcm_dpo/beta": 0.004994173534214497, "fcm_dpo/delta": -0.06866991519927979, "fcm_dpo/margin": 172.4603271484375, "fcm_dpo/q_t": 0.31907975673675537, "grad_norm": 39.19725799560547, "learning_rate": 2.7434463620546594e-07, "logits/chosen": 0.11515648663043976, "logits/rejected": 0.13530930876731873, "logps/chosen": -302.9579772949219, "logps/ref_chosen": -52.725799560546875, "logps/ref_rejected": -86.84115600585938, "logps/rejected": -509.5336608886719, "loss": 0.8274, "margin_dpo/margin_mean": 172.46034240722656, "margin_dpo/margin_std": 156.9672393798828, "step": 357 }, { "KL/chosen_KL_mean": -236.85372924804688, "KL/mean": -309.67779541015625, "KL/rejected_KL_mean": -382.50189208984375, "KL/std": 164.68772888183594, "epoch": 0.5256975036710719, "fcm_dpo/beta": 0.005067657679319382, "fcm_dpo/delta": 0.06674303859472275, "fcm_dpo/margin": 145.64813232421875, "fcm_dpo/q_t": 0.35047268867492676, "grad_norm": 39.29417419433594, "learning_rate": 2.730670898658255e-07, "logits/chosen": -0.0026182327419519424, "logits/rejected": 0.017924560233950615, "logps/chosen": -300.0591735839844, "logps/ref_chosen": -63.20543670654297, "logps/ref_rejected": -88.373291015625, "logps/rejected": -470.87518310546875, "loss": 0.9235, "margin_dpo/margin_mean": 145.64813232421875, "margin_dpo/margin_std": 169.32843017578125, "step": 358 }, { "KL/chosen_KL_mean": -261.9931640625, "KL/mean": -344.5310974121094, "KL/rejected_KL_mean": -427.0690612792969, "KL/std": 152.8817138671875, "epoch": 0.527165932452276, "fcm_dpo/beta": 0.005035270005464554, "fcm_dpo/delta": -0.03444623947143555, "fcm_dpo/margin": 165.0758819580078, "fcm_dpo/q_t": 0.33515065908432007, "grad_norm": 47.493412017822266, "learning_rate": 2.717889356869146e-07, "logits/chosen": 0.1439620852470398, "logits/rejected": 0.16959136724472046, "logps/chosen": -318.3633728027344, "logps/ref_chosen": -56.370216369628906, "logps/ref_rejected": -82.17375183105469, "logps/rejected": -509.2427978515625, "loss": 0.8822, "margin_dpo/margin_mean": 165.07586669921875, "margin_dpo/margin_std": 180.14077758789062, "step": 359 }, { "KL/chosen_KL_mean": -242.22293090820312, "KL/mean": -319.6527099609375, "KL/rejected_KL_mean": -397.08245849609375, "KL/std": 135.70486450195312, "epoch": 0.5286343612334802, "fcm_dpo/beta": 0.005076201166957617, "fcm_dpo/delta": 0.014695055782794952, "fcm_dpo/margin": 154.85951232910156, "fcm_dpo/q_t": 0.3311746418476105, "grad_norm": 39.480186462402344, "learning_rate": 2.7051020734928443e-07, "logits/chosen": 0.1422397643327713, "logits/rejected": 0.15580809116363525, "logps/chosen": -293.6833190917969, "logps/ref_chosen": -51.460384368896484, "logps/ref_rejected": -69.83892059326172, "logps/rejected": -466.92138671875, "loss": 0.85, "margin_dpo/margin_mean": 154.85951232910156, "margin_dpo/margin_std": 136.16104125976562, "step": 360 }, { "KL/chosen_KL_mean": -244.15972900390625, "KL/mean": -311.66357421875, "KL/rejected_KL_mean": -379.16741943359375, "KL/std": 147.64564514160156, "epoch": 0.5301027900146843, "fcm_dpo/beta": 0.005172476172447205, "fcm_dpo/delta": 0.10789966583251953, "fcm_dpo/margin": 135.00767517089844, "fcm_dpo/q_t": 0.35620319843292236, "grad_norm": 40.17778015136719, "learning_rate": 2.6923093854861593e-07, "logits/chosen": 0.06714671850204468, "logits/rejected": 0.0685434490442276, "logps/chosen": -298.02923583984375, "logps/ref_chosen": -53.86951446533203, "logps/ref_rejected": -90.7692642211914, "logps/rejected": -469.9366760253906, "loss": 0.9569, "margin_dpo/margin_mean": 135.0076904296875, "margin_dpo/margin_std": 163.23028564453125, "step": 361 }, { "KL/chosen_KL_mean": -209.6553955078125, "KL/mean": -306.86016845703125, "KL/rejected_KL_mean": -404.0649108886719, "KL/std": 155.64691162109375, "epoch": 0.5315712187958884, "fcm_dpo/beta": 0.0050034294836223125, "fcm_dpo/delta": -0.19257816672325134, "fcm_dpo/margin": 194.40948486328125, "fcm_dpo/q_t": 0.2972559332847595, "grad_norm": 34.24903869628906, "learning_rate": 2.679511629948319e-07, "logits/chosen": 0.06750906258821487, "logits/rejected": 0.06738758087158203, "logps/chosen": -268.2944641113281, "logps/ref_chosen": -58.639060974121094, "logps/ref_rejected": -105.58195495605469, "logps/rejected": -509.6468505859375, "loss": 0.7571, "margin_dpo/margin_mean": 194.40948486328125, "margin_dpo/margin_std": 160.616455078125, "step": 362 }, { "KL/chosen_KL_mean": -185.4617919921875, "KL/mean": -288.15216064453125, "KL/rejected_KL_mean": -390.8425598144531, "KL/std": 156.06210327148438, "epoch": 0.5330396475770925, "fcm_dpo/beta": 0.004827477503567934, "fcm_dpo/delta": -0.21232551336288452, "fcm_dpo/margin": 205.38076782226562, "fcm_dpo/q_t": 0.2914503514766693, "grad_norm": 42.787349700927734, "learning_rate": 2.6667091441120816e-07, "logits/chosen": 0.10203859210014343, "logits/rejected": 0.12859748303890228, "logps/chosen": -230.02017211914062, "logps/ref_chosen": -44.558380126953125, "logps/ref_rejected": -74.69496154785156, "logps/rejected": -465.53753662109375, "loss": 0.7421, "margin_dpo/margin_mean": 205.38076782226562, "margin_dpo/margin_std": 159.50564575195312, "step": 363 }, { "KL/chosen_KL_mean": -205.2596435546875, "KL/mean": -279.29486083984375, "KL/rejected_KL_mean": -353.3301086425781, "KL/std": 146.0043182373047, "epoch": 0.5345080763582967, "fcm_dpo/beta": 0.004854832775890827, "fcm_dpo/delta": 0.08639908581972122, "fcm_dpo/margin": 148.07046508789062, "fcm_dpo/q_t": 0.3479136824607849, "grad_norm": 30.45250129699707, "learning_rate": 2.6539022653348575e-07, "logits/chosen": -0.006259392946958542, "logits/rejected": -0.008601278066635132, "logps/chosen": -254.15426635742188, "logps/ref_chosen": -48.894622802734375, "logps/ref_rejected": -91.395751953125, "logps/rejected": -444.725830078125, "loss": 0.9176, "margin_dpo/margin_mean": 148.07046508789062, "margin_dpo/margin_std": 158.57998657226562, "step": 364 }, { "KL/chosen_KL_mean": -199.53936767578125, "KL/mean": -275.67108154296875, "KL/rejected_KL_mean": -351.80279541015625, "KL/std": 160.76339721679688, "epoch": 0.5359765051395007, "fcm_dpo/beta": 0.004905831068754196, "fcm_dpo/delta": 0.05698072165250778, "fcm_dpo/margin": 152.26344299316406, "fcm_dpo/q_t": 0.346111923456192, "grad_norm": 37.139583587646484, "learning_rate": 2.641091331089811e-07, "logits/chosen": 0.01942162960767746, "logits/rejected": 0.012668520212173462, "logps/chosen": -251.0321044921875, "logps/ref_chosen": -51.49274444580078, "logps/ref_rejected": -92.70166778564453, "logps/rejected": -444.50445556640625, "loss": 0.897, "margin_dpo/margin_mean": 152.263427734375, "margin_dpo/margin_std": 162.59872436523438, "step": 365 }, { "KL/chosen_KL_mean": -197.44915771484375, "KL/mean": -276.533203125, "KL/rejected_KL_mean": -355.6172790527344, "KL/std": 154.4979248046875, "epoch": 0.5374449339207048, "fcm_dpo/beta": 0.004907799884676933, "fcm_dpo/delta": 0.025566942989826202, "fcm_dpo/margin": 158.16812133789062, "fcm_dpo/q_t": 0.3392502963542938, "grad_norm": 35.77729034423828, "learning_rate": 2.6282766789569736e-07, "logits/chosen": 0.06100982427597046, "logits/rejected": 0.05832307040691376, "logps/chosen": -242.16970825195312, "logps/ref_chosen": -44.7205696105957, "logps/ref_rejected": -83.31040954589844, "logps/rejected": -438.92767333984375, "loss": 0.89, "margin_dpo/margin_mean": 158.16812133789062, "margin_dpo/margin_std": 166.03973388671875, "step": 366 }, { "KL/chosen_KL_mean": -210.3382568359375, "KL/mean": -274.84576416015625, "KL/rejected_KL_mean": -339.353271484375, "KL/std": 141.46987915039062, "epoch": 0.5389133627019089, "fcm_dpo/beta": 0.005055001471191645, "fcm_dpo/delta": 0.15791787207126617, "fcm_dpo/margin": 129.01502990722656, "fcm_dpo/q_t": 0.3598521947860718, "grad_norm": 33.63018798828125, "learning_rate": 2.615458646614349e-07, "logits/chosen": 0.01299678348004818, "logits/rejected": 0.038110073655843735, "logps/chosen": -268.7436828613281, "logps/ref_chosen": -58.405418395996094, "logps/ref_rejected": -76.75132751464844, "logps/rejected": -416.1046142578125, "loss": 0.9426, "margin_dpo/margin_mean": 129.01502990722656, "margin_dpo/margin_std": 137.5300750732422, "step": 367 }, { "KL/chosen_KL_mean": -205.1249237060547, "KL/mean": -294.59051513671875, "KL/rejected_KL_mean": -384.05609130859375, "KL/std": 149.47842407226562, "epoch": 0.540381791483113, "fcm_dpo/beta": 0.005028966814279556, "fcm_dpo/delta": -0.10961154103279114, "fcm_dpo/margin": 178.93115234375, "fcm_dpo/q_t": 0.3045145273208618, "grad_norm": 58.015506744384766, "learning_rate": 2.6026375718290083e-07, "logits/chosen": 0.03757400065660477, "logits/rejected": 0.0448252409696579, "logps/chosen": -249.5774383544922, "logps/ref_chosen": -44.452518463134766, "logps/ref_rejected": -98.55526733398438, "logps/rejected": -482.6113586425781, "loss": 0.765, "margin_dpo/margin_mean": 178.93115234375, "margin_dpo/margin_std": 128.32945251464844, "step": 368 }, { "KL/chosen_KL_mean": -243.73876953125, "KL/mean": -307.91265869140625, "KL/rejected_KL_mean": -372.08648681640625, "KL/std": 149.72775268554688, "epoch": 0.5418502202643172, "fcm_dpo/beta": 0.0051218681037425995, "fcm_dpo/delta": 0.15110597014427185, "fcm_dpo/margin": 128.3477020263672, "fcm_dpo/q_t": 0.3617558777332306, "grad_norm": 54.27680969238281, "learning_rate": 2.589813792448196e-07, "logits/chosen": 0.04510130733251572, "logits/rejected": 0.063881054520607, "logps/chosen": -315.12030029296875, "logps/ref_chosen": -71.38150024414062, "logps/ref_rejected": -91.29582214355469, "logps/rejected": -463.38232421875, "loss": 0.9757, "margin_dpo/margin_mean": 128.3477020263672, "margin_dpo/margin_std": 156.42678833007812, "step": 369 }, { "KL/chosen_KL_mean": -277.29473876953125, "KL/mean": -331.53424072265625, "KL/rejected_KL_mean": -385.7738037109375, "KL/std": 153.50787353515625, "epoch": 0.5433186490455213, "fcm_dpo/beta": 0.005329503677785397, "fcm_dpo/delta": 0.2346639633178711, "fcm_dpo/margin": 108.47908020019531, "fcm_dpo/q_t": 0.3781193196773529, "grad_norm": 43.87196731567383, "learning_rate": 2.5769876463904263e-07, "logits/chosen": 0.07844299077987671, "logits/rejected": 0.08098983019590378, "logps/chosen": -348.9022216796875, "logps/ref_chosen": -71.60749816894531, "logps/ref_rejected": -97.25978088378906, "logps/rejected": -483.0335693359375, "loss": 1.0206, "margin_dpo/margin_mean": 108.47908020019531, "margin_dpo/margin_std": 144.95086669921875, "step": 370 }, { "KL/chosen_KL_mean": -262.21136474609375, "KL/mean": -335.576171875, "KL/rejected_KL_mean": -408.94091796875, "KL/std": 151.41275024414062, "epoch": 0.5447870778267254, "fcm_dpo/beta": 0.005401424132287502, "fcm_dpo/delta": 0.007885420694947243, "fcm_dpo/margin": 146.72958374023438, "fcm_dpo/q_t": 0.33563292026519775, "grad_norm": 42.30845642089844, "learning_rate": 2.5641594716365744e-07, "logits/chosen": 0.06122782081365585, "logits/rejected": 0.07856868207454681, "logps/chosen": -331.6258544921875, "logps/ref_chosen": -69.41448974609375, "logps/ref_rejected": -99.17217254638672, "logps/rejected": -508.11309814453125, "loss": 0.8961, "margin_dpo/margin_mean": 146.72958374023438, "margin_dpo/margin_std": 158.2304229736328, "step": 371 }, { "KL/chosen_KL_mean": -268.47515869140625, "KL/mean": -363.11712646484375, "KL/rejected_KL_mean": -457.75909423828125, "KL/std": 186.968017578125, "epoch": 0.5462555066079295, "fcm_dpo/beta": 0.005196971818804741, "fcm_dpo/delta": -0.20561370253562927, "fcm_dpo/margin": 189.283935546875, "fcm_dpo/q_t": 0.307757705450058, "grad_norm": 38.48178482055664, "learning_rate": 2.551329606220976e-07, "logits/chosen": 0.14845509827136993, "logits/rejected": 0.18487697839736938, "logps/chosen": -330.29315185546875, "logps/ref_chosen": -61.8179931640625, "logps/ref_rejected": -78.53948974609375, "logps/rejected": -536.298583984375, "loss": 0.8196, "margin_dpo/margin_mean": 189.283935546875, "margin_dpo/margin_std": 192.58099365234375, "step": 372 }, { "KL/chosen_KL_mean": -282.3094177246094, "KL/mean": -373.96246337890625, "KL/rejected_KL_mean": -465.6155090332031, "KL/std": 172.3424530029297, "epoch": 0.5477239353891337, "fcm_dpo/beta": 0.005096299573779106, "fcm_dpo/delta": -0.1480431854724884, "fcm_dpo/margin": 183.30609130859375, "fcm_dpo/q_t": 0.30489110946655273, "grad_norm": 46.1933479309082, "learning_rate": 2.538498388222517e-07, "logits/chosen": 0.16775904595851898, "logits/rejected": 0.2070429027080536, "logps/chosen": -346.52655029296875, "logps/ref_chosen": -64.21713256835938, "logps/ref_rejected": -85.95960998535156, "logps/rejected": -551.5751342773438, "loss": 0.7973, "margin_dpo/margin_mean": 183.30609130859375, "margin_dpo/margin_std": 159.746826171875, "step": 373 }, { "KL/chosen_KL_mean": -259.88995361328125, "KL/mean": -342.8182373046875, "KL/rejected_KL_mean": -425.74652099609375, "KL/std": 191.01841735839844, "epoch": 0.5491923641703378, "fcm_dpo/beta": 0.00495956651866436, "fcm_dpo/delta": -0.02621932327747345, "fcm_dpo/margin": 165.8565673828125, "fcm_dpo/q_t": 0.33612874150276184, "grad_norm": 43.68080139160156, "learning_rate": 2.525666155755725e-07, "logits/chosen": 0.014575677923858166, "logits/rejected": 0.04540920630097389, "logps/chosen": -330.5401306152344, "logps/ref_chosen": -70.65018463134766, "logps/ref_rejected": -93.64016723632812, "logps/rejected": -519.38671875, "loss": 0.8808, "margin_dpo/margin_mean": 165.85658264160156, "margin_dpo/margin_std": 178.73219299316406, "step": 374 }, { "KL/chosen_KL_mean": -302.646484375, "KL/mean": -383.7525634765625, "KL/rejected_KL_mean": -464.8586730957031, "KL/std": 173.11106872558594, "epoch": 0.5506607929515418, "fcm_dpo/beta": 0.004924945533275604, "fcm_dpo/delta": -0.0026347339153289795, "fcm_dpo/margin": 162.21218872070312, "fcm_dpo/q_t": 0.33666491508483887, "grad_norm": 42.200557708740234, "learning_rate": 2.512833246961859e-07, "logits/chosen": 0.13259188830852509, "logits/rejected": 0.14011941850185394, "logps/chosen": -362.7267150878906, "logps/ref_chosen": -60.080223083496094, "logps/ref_rejected": -88.93830871582031, "logps/rejected": -553.7969970703125, "loss": 0.9159, "margin_dpo/margin_mean": 162.21218872070312, "margin_dpo/margin_std": 178.72955322265625, "step": 375 }, { "KL/chosen_KL_mean": -284.9696350097656, "KL/mean": -380.9700927734375, "KL/rejected_KL_mean": -476.97052001953125, "KL/std": 175.26528930664062, "epoch": 0.5521292217327459, "fcm_dpo/beta": 0.0048746224492788315, "fcm_dpo/delta": -0.15022152662277222, "fcm_dpo/margin": 192.00088500976562, "fcm_dpo/q_t": 0.31024953722953796, "grad_norm": 39.80452346801758, "learning_rate": 2.5e-07, "logits/chosen": 0.10005451738834381, "logits/rejected": 0.11893795430660248, "logps/chosen": -347.62994384765625, "logps/ref_chosen": -62.660308837890625, "logps/ref_rejected": -105.52660369873047, "logps/rejected": -582.4970703125, "loss": 0.8218, "margin_dpo/margin_mean": 192.00088500976562, "margin_dpo/margin_std": 187.91757202148438, "step": 376 }, { "KL/chosen_KL_mean": -312.7281188964844, "KL/mean": -402.2392883300781, "KL/rejected_KL_mean": -491.75048828125, "KL/std": 178.24826049804688, "epoch": 0.55359765051395, "fcm_dpo/beta": 0.004800399765372276, "fcm_dpo/delta": -0.06492528319358826, "fcm_dpo/margin": 179.02236938476562, "fcm_dpo/q_t": 0.32147014141082764, "grad_norm": 41.688804626464844, "learning_rate": 2.487166753038141e-07, "logits/chosen": 0.17259091138839722, "logits/rejected": 0.1837162971496582, "logps/chosen": -367.20684814453125, "logps/ref_chosen": -54.478736877441406, "logps/ref_rejected": -98.70335388183594, "logps/rejected": -590.453857421875, "loss": 0.8519, "margin_dpo/margin_mean": 179.02236938476562, "margin_dpo/margin_std": 176.6744384765625, "step": 377 }, { "KL/chosen_KL_mean": -329.71490478515625, "KL/mean": -428.11505126953125, "KL/rejected_KL_mean": -526.5152587890625, "KL/std": 173.01528930664062, "epoch": 0.5550660792951542, "fcm_dpo/beta": 0.0046610780991613865, "fcm_dpo/delta": -0.1307007521390915, "fcm_dpo/margin": 196.8003692626953, "fcm_dpo/q_t": 0.30623000860214233, "grad_norm": 49.080684661865234, "learning_rate": 2.4743338442442754e-07, "logits/chosen": 0.33253681659698486, "logits/rejected": 0.3434640169143677, "logps/chosen": -374.73541259765625, "logps/ref_chosen": -45.02053451538086, "logps/ref_rejected": -88.0469741821289, "logps/rejected": -614.562255859375, "loss": 0.7959, "margin_dpo/margin_mean": 196.8003692626953, "margin_dpo/margin_std": 168.24053955078125, "step": 378 }, { "KL/chosen_KL_mean": -347.2498779296875, "KL/mean": -441.2623291015625, "KL/rejected_KL_mean": -535.2748413085938, "KL/std": 171.8399658203125, "epoch": 0.5565345080763583, "fcm_dpo/beta": 0.004580418113619089, "fcm_dpo/delta": -0.0683475211262703, "fcm_dpo/margin": 188.02496337890625, "fcm_dpo/q_t": 0.3261038661003113, "grad_norm": 56.207244873046875, "learning_rate": 2.461501611777483e-07, "logits/chosen": 0.23632130026817322, "logits/rejected": 0.22753140330314636, "logps/chosen": -400.4319763183594, "logps/ref_chosen": -53.182098388671875, "logps/ref_rejected": -114.3001708984375, "logps/rejected": -649.574951171875, "loss": 0.8721, "margin_dpo/margin_mean": 188.0249481201172, "margin_dpo/margin_std": 201.0380859375, "step": 379 }, { "KL/chosen_KL_mean": -346.2129211425781, "KL/mean": -453.26580810546875, "KL/rejected_KL_mean": -560.3186645507812, "KL/std": 215.87939453125, "epoch": 0.5580029368575624, "fcm_dpo/beta": 0.004501840099692345, "fcm_dpo/delta": -0.18104371428489685, "fcm_dpo/margin": 214.10574340820312, "fcm_dpo/q_t": 0.3031790256500244, "grad_norm": 38.03917694091797, "learning_rate": 2.4486703937790243e-07, "logits/chosen": 0.21140098571777344, "logits/rejected": 0.19853919744491577, "logps/chosen": -397.5659484863281, "logps/ref_chosen": -51.3530387878418, "logps/ref_rejected": -104.19169616699219, "logps/rejected": -664.5103759765625, "loss": 0.8193, "margin_dpo/margin_mean": 214.10574340820312, "margin_dpo/margin_std": 205.07131958007812, "step": 380 }, { "KL/chosen_KL_mean": -359.42694091796875, "KL/mean": -447.6293029785156, "KL/rejected_KL_mean": -535.8316650390625, "KL/std": 172.031494140625, "epoch": 0.5594713656387665, "fcm_dpo/beta": 0.004441162571310997, "fcm_dpo/delta": 0.01796901971101761, "fcm_dpo/margin": 176.40469360351562, "fcm_dpo/q_t": 0.34440547227859497, "grad_norm": 51.2555046081543, "learning_rate": 2.435840528363426e-07, "logits/chosen": 0.264553427696228, "logits/rejected": 0.2853265404701233, "logps/chosen": -417.2300109863281, "logps/ref_chosen": -57.80306625366211, "logps/ref_rejected": -79.21940612792969, "logps/rejected": -615.051025390625, "loss": 0.9444, "margin_dpo/margin_mean": 176.40469360351562, "margin_dpo/margin_std": 220.600341796875, "step": 381 }, { "KL/chosen_KL_mean": -359.14471435546875, "KL/mean": -449.2369689941406, "KL/rejected_KL_mean": -539.3292236328125, "KL/std": 180.70993041992188, "epoch": 0.5609397944199707, "fcm_dpo/beta": 0.004444487858563662, "fcm_dpo/delta": -0.0009026522748172283, "fcm_dpo/margin": 180.18447875976562, "fcm_dpo/q_t": 0.3322370648384094, "grad_norm": 40.537044525146484, "learning_rate": 2.4230123536095745e-07, "logits/chosen": 0.19075685739517212, "logits/rejected": 0.19895179569721222, "logps/chosen": -425.1650390625, "logps/ref_chosen": -66.02030181884766, "logps/ref_rejected": -110.71016693115234, "logps/rejected": -650.0394287109375, "loss": 0.868, "margin_dpo/margin_mean": 180.18447875976562, "margin_dpo/margin_std": 177.90121459960938, "step": 382 }, { "KL/chosen_KL_mean": -351.47509765625, "KL/mean": -444.9332275390625, "KL/rejected_KL_mean": -538.391357421875, "KL/std": 187.61868286132812, "epoch": 0.5624082232011748, "fcm_dpo/beta": 0.004414629191160202, "fcm_dpo/delta": -0.027692466974258423, "fcm_dpo/margin": 186.916259765625, "fcm_dpo/q_t": 0.3323889970779419, "grad_norm": 50.61336898803711, "learning_rate": 2.4101862075518037e-07, "logits/chosen": 0.2367815375328064, "logits/rejected": 0.23397362232208252, "logps/chosen": -401.8665771484375, "logps/ref_chosen": -50.39148712158203, "logps/ref_rejected": -93.71589660644531, "logps/rejected": -632.1072998046875, "loss": 0.9123, "margin_dpo/margin_mean": 186.916259765625, "margin_dpo/margin_std": 220.93585205078125, "step": 383 }, { "KL/chosen_KL_mean": -350.86578369140625, "KL/mean": -428.3222351074219, "KL/rejected_KL_mean": -505.7787170410156, "KL/std": 191.98023986816406, "epoch": 0.5638766519823789, "fcm_dpo/beta": 0.004525776021182537, "fcm_dpo/delta": 0.10422958433628082, "fcm_dpo/margin": 154.9129638671875, "fcm_dpo/q_t": 0.3500513434410095, "grad_norm": 42.488250732421875, "learning_rate": 2.397362428170992e-07, "logits/chosen": 0.13534685969352722, "logits/rejected": 0.14616771042346954, "logps/chosen": -402.911865234375, "logps/ref_chosen": -52.046104431152344, "logps/ref_rejected": -85.76089477539062, "logps/rejected": -591.5396118164062, "loss": 0.9165, "margin_dpo/margin_mean": 154.91297912597656, "margin_dpo/margin_std": 157.85365295410156, "step": 384 }, { "KL/chosen_KL_mean": -349.88470458984375, "KL/mean": -442.55535888671875, "KL/rejected_KL_mean": -535.2260131835938, "KL/std": 163.54031372070312, "epoch": 0.5653450807635829, "fcm_dpo/beta": 0.004492057487368584, "fcm_dpo/delta": -0.035512764006853104, "fcm_dpo/margin": 185.34132385253906, "fcm_dpo/q_t": 0.32157087326049805, "grad_norm": 45.575958251953125, "learning_rate": 2.3845413533856514e-07, "logits/chosen": 0.24671843647956848, "logits/rejected": 0.2860070466995239, "logps/chosen": -415.4368591308594, "logps/ref_chosen": -65.55215454101562, "logps/ref_rejected": -77.82792663574219, "logps/rejected": -613.053955078125, "loss": 0.8289, "margin_dpo/margin_mean": 185.34132385253906, "margin_dpo/margin_std": 160.22723388671875, "step": 385 }, { "KL/chosen_KL_mean": -359.40350341796875, "KL/mean": -456.8614501953125, "KL/rejected_KL_mean": -554.3193359375, "KL/std": 196.9163360595703, "epoch": 0.566813509544787, "fcm_dpo/beta": 0.004448466002941132, "fcm_dpo/delta": -0.07342489063739777, "fcm_dpo/margin": 194.91583251953125, "fcm_dpo/q_t": 0.324515163898468, "grad_norm": 42.45697784423828, "learning_rate": 2.3717233210430254e-07, "logits/chosen": 0.16371941566467285, "logits/rejected": 0.1747831106185913, "logps/chosen": -417.6253662109375, "logps/ref_chosen": -58.22185516357422, "logps/ref_rejected": -92.32742309570312, "logps/rejected": -646.646728515625, "loss": 0.8613, "margin_dpo/margin_mean": 194.91583251953125, "margin_dpo/margin_std": 205.5290069580078, "step": 386 }, { "KL/chosen_KL_mean": -372.31109619140625, "KL/mean": -450.6243591308594, "KL/rejected_KL_mean": -528.9375610351562, "KL/std": 182.24484252929688, "epoch": 0.5682819383259912, "fcm_dpo/beta": 0.004475068300962448, "fcm_dpo/delta": 0.1066901832818985, "fcm_dpo/margin": 156.62646484375, "fcm_dpo/q_t": 0.35136184096336365, "grad_norm": 49.93186569213867, "learning_rate": 2.3589086689101889e-07, "logits/chosen": 0.1222163736820221, "logits/rejected": 0.1534290611743927, "logps/chosen": -438.73052978515625, "logps/ref_chosen": -66.41944885253906, "logps/ref_rejected": -92.16915893554688, "logps/rejected": -621.1067504882812, "loss": 0.9202, "margin_dpo/margin_mean": 156.62648010253906, "margin_dpo/margin_std": 165.41067504882812, "step": 387 }, { "KL/chosen_KL_mean": -345.98480224609375, "KL/mean": -450.4599914550781, "KL/rejected_KL_mean": -554.9351806640625, "KL/std": 201.11993408203125, "epoch": 0.5697503671071953, "fcm_dpo/beta": 0.00440019927918911, "fcm_dpo/delta": -0.13290224969387054, "fcm_dpo/margin": 208.95037841796875, "fcm_dpo/q_t": 0.31572598218917847, "grad_norm": 45.941741943359375, "learning_rate": 2.3460977346651428e-07, "logits/chosen": 0.2326255887746811, "logits/rejected": 0.22134818136692047, "logps/chosen": -396.1142578125, "logps/ref_chosen": -50.129459381103516, "logps/ref_rejected": -104.43305969238281, "logps/rejected": -659.3682250976562, "loss": 0.8373, "margin_dpo/margin_mean": 208.95037841796875, "margin_dpo/margin_std": 217.96234130859375, "step": 388 }, { "KL/chosen_KL_mean": -377.5467834472656, "KL/mean": -470.57232666015625, "KL/rejected_KL_mean": -563.5978393554688, "KL/std": 187.90357971191406, "epoch": 0.5712187958883994, "fcm_dpo/beta": 0.004369369708001614, "fcm_dpo/delta": -0.014375818893313408, "fcm_dpo/margin": 186.05105590820312, "fcm_dpo/q_t": 0.33482223749160767, "grad_norm": 44.91924285888672, "learning_rate": 2.3332908558879177e-07, "logits/chosen": 0.22911961376667023, "logits/rejected": 0.2511199712753296, "logps/chosen": -435.453369140625, "logps/ref_chosen": -57.906593322753906, "logps/ref_rejected": -77.91454315185547, "logps/rejected": -641.5123901367188, "loss": 0.8948, "margin_dpo/margin_mean": 186.05105590820312, "margin_dpo/margin_std": 203.96209716796875, "step": 389 }, { "KL/chosen_KL_mean": -358.6050720214844, "KL/mean": -444.9243469238281, "KL/rejected_KL_mean": -531.2435913085938, "KL/std": 184.04891967773438, "epoch": 0.5726872246696035, "fcm_dpo/beta": 0.00439393799751997, "fcm_dpo/delta": 0.044710010290145874, "fcm_dpo/margin": 172.63853454589844, "fcm_dpo/q_t": 0.3482525050640106, "grad_norm": 39.236610412597656, "learning_rate": 2.320488370051681e-07, "logits/chosen": 0.1646582931280136, "logits/rejected": 0.18338599801063538, "logps/chosen": -407.83099365234375, "logps/ref_chosen": -49.22591781616211, "logps/ref_rejected": -85.5281982421875, "logps/rejected": -616.7717895507812, "loss": 0.9706, "margin_dpo/margin_mean": 172.63853454589844, "margin_dpo/margin_std": 231.70809936523438, "step": 390 }, { "KL/chosen_KL_mean": -329.9169006347656, "KL/mean": -392.60296630859375, "KL/rejected_KL_mean": -455.28900146484375, "KL/std": 190.03704833984375, "epoch": 0.5741556534508077, "fcm_dpo/beta": 0.004583893343806267, "fcm_dpo/delta": 0.2388683259487152, "fcm_dpo/margin": 125.37210845947266, "fcm_dpo/q_t": 0.3850371241569519, "grad_norm": 55.410194396972656, "learning_rate": 2.3076906145138405e-07, "logits/chosen": 0.14501428604125977, "logits/rejected": 0.15393584966659546, "logps/chosen": -394.2465515136719, "logps/ref_chosen": -64.32965087890625, "logps/ref_rejected": -86.73820495605469, "logps/rejected": -542.0272216796875, "loss": 1.0517, "margin_dpo/margin_mean": 125.37210845947266, "margin_dpo/margin_std": 190.40362548828125, "step": 391 }, { "KL/chosen_KL_mean": -277.9818115234375, "KL/mean": -376.48553466796875, "KL/rejected_KL_mean": -474.9892883300781, "KL/std": 181.52938842773438, "epoch": 0.5756240822320118, "fcm_dpo/beta": 0.004586499184370041, "fcm_dpo/delta": -0.1139056608080864, "fcm_dpo/margin": 197.00746154785156, "fcm_dpo/q_t": 0.311518132686615, "grad_norm": 51.724281311035156, "learning_rate": 2.294897926507156e-07, "logits/chosen": 0.06240957975387573, "logits/rejected": 0.07568895816802979, "logps/chosen": -331.48577880859375, "logps/ref_chosen": -53.50397872924805, "logps/ref_rejected": -102.34584045410156, "logps/rejected": -577.3350830078125, "loss": 0.8014, "margin_dpo/margin_mean": 197.00747680664062, "margin_dpo/margin_std": 170.6578369140625, "step": 392 }, { "KL/chosen_KL_mean": -279.2279052734375, "KL/mean": -355.4307861328125, "KL/rejected_KL_mean": -431.6336669921875, "KL/std": 180.89584350585938, "epoch": 0.5770925110132159, "fcm_dpo/beta": 0.004597535356879234, "fcm_dpo/delta": 0.10691172629594803, "fcm_dpo/margin": 152.40574645996094, "fcm_dpo/q_t": 0.36370301246643066, "grad_norm": 33.592952728271484, "learning_rate": 2.2821106431308543e-07, "logits/chosen": 0.13287240266799927, "logits/rejected": 0.14408522844314575, "logps/chosen": -325.7018127441406, "logps/ref_chosen": -46.473915100097656, "logps/ref_rejected": -71.96885681152344, "logps/rejected": -503.6025390625, "loss": 0.9837, "margin_dpo/margin_mean": 152.40574645996094, "margin_dpo/margin_std": 210.06790161132812, "step": 393 }, { "KL/chosen_KL_mean": -281.6102294921875, "KL/mean": -368.0687255859375, "KL/rejected_KL_mean": -454.5271911621094, "KL/std": 181.987060546875, "epoch": 0.57856093979442, "fcm_dpo/beta": 0.004621568135917187, "fcm_dpo/delta": 0.0008383970707654953, "fcm_dpo/margin": 172.91696166992188, "fcm_dpo/q_t": 0.3378085494041443, "grad_norm": 48.079124450683594, "learning_rate": 2.2693291013417452e-07, "logits/chosen": 0.05104764923453331, "logits/rejected": 0.05909823626279831, "logps/chosen": -334.52178955078125, "logps/ref_chosen": -52.91154861450195, "logps/ref_rejected": -90.8226318359375, "logps/rejected": -545.349853515625, "loss": 0.8954, "margin_dpo/margin_mean": 172.91696166992188, "margin_dpo/margin_std": 190.8291778564453, "step": 394 }, { "KL/chosen_KL_mean": -271.431640625, "KL/mean": -362.2996520996094, "KL/rejected_KL_mean": -453.16766357421875, "KL/std": 188.79595947265625, "epoch": 0.580029368575624, "fcm_dpo/beta": 0.004584716632962227, "fcm_dpo/delta": -0.03685643523931503, "fcm_dpo/margin": 181.7360076904297, "fcm_dpo/q_t": 0.3331453800201416, "grad_norm": 40.45199966430664, "learning_rate": 2.2565536379453404e-07, "logits/chosen": 0.03748972713947296, "logits/rejected": 0.048588261008262634, "logps/chosen": -333.977783203125, "logps/ref_chosen": -62.546112060546875, "logps/ref_rejected": -83.78262329101562, "logps/rejected": -536.9503173828125, "loss": 0.8995, "margin_dpo/margin_mean": 181.73599243164062, "margin_dpo/margin_std": 206.99716186523438, "step": 395 }, { "KL/chosen_KL_mean": -254.0207977294922, "KL/mean": -335.33380126953125, "KL/rejected_KL_mean": -416.64678955078125, "KL/std": 170.42239379882812, "epoch": 0.5814977973568282, "fcm_dpo/beta": 0.004617318511009216, "fcm_dpo/delta": 0.05301779881119728, "fcm_dpo/margin": 162.62600708007812, "fcm_dpo/q_t": 0.3398272395133972, "grad_norm": 42.922828674316406, "learning_rate": 2.2437845895869825e-07, "logits/chosen": -0.03782034292817116, "logits/rejected": -0.010618546046316624, "logps/chosen": -323.0167236328125, "logps/ref_chosen": -68.99594116210938, "logps/ref_rejected": -88.64665985107422, "logps/rejected": -505.29345703125, "loss": 0.8781, "margin_dpo/margin_mean": 162.62600708007812, "margin_dpo/margin_std": 156.22705078125, "step": 396 }, { "KL/chosen_KL_mean": -230.32138061523438, "KL/mean": -323.5391845703125, "KL/rejected_KL_mean": -416.7569580078125, "KL/std": 165.25631713867188, "epoch": 0.5829662261380323, "fcm_dpo/beta": 0.004583822563290596, "fcm_dpo/delta": -0.060415178537368774, "fcm_dpo/margin": 186.43557739257812, "fcm_dpo/q_t": 0.31888556480407715, "grad_norm": 55.33768081665039, "learning_rate": 2.2310222927429716e-07, "logits/chosen": 0.06463508307933807, "logits/rejected": 0.08042390644550323, "logps/chosen": -291.59857177734375, "logps/ref_chosen": -61.27716827392578, "logps/ref_rejected": -103.11612701416016, "logps/rejected": -519.873046875, "loss": 0.8176, "margin_dpo/margin_mean": 186.43557739257812, "margin_dpo/margin_std": 161.7078857421875, "step": 397 }, { "KL/chosen_KL_mean": -252.4274139404297, "KL/mean": -340.54632568359375, "KL/rejected_KL_mean": -428.6652526855469, "KL/std": 179.56756591796875, "epoch": 0.5844346549192364, "fcm_dpo/beta": 0.004576831124722958, "fcm_dpo/delta": -0.0072813937440514565, "fcm_dpo/margin": 176.23779296875, "fcm_dpo/q_t": 0.33511707186698914, "grad_norm": 33.04292297363281, "learning_rate": 2.2182670837116972e-07, "logits/chosen": -0.08657974749803543, "logits/rejected": -0.07474084943532944, "logps/chosen": -320.5789794921875, "logps/ref_chosen": -68.15155029296875, "logps/ref_rejected": -108.52360534667969, "logps/rejected": -537.1888427734375, "loss": 0.884, "margin_dpo/margin_mean": 176.23780822753906, "margin_dpo/margin_std": 189.9404296875, "step": 398 }, { "KL/chosen_KL_mean": -215.05654907226562, "KL/mean": -306.60455322265625, "KL/rejected_KL_mean": -398.15252685546875, "KL/std": 165.11349487304688, "epoch": 0.5859030837004405, "fcm_dpo/beta": 0.004536399617791176, "fcm_dpo/delta": -0.03415355086326599, "fcm_dpo/margin": 183.09597778320312, "fcm_dpo/q_t": 0.3255794048309326, "grad_norm": 30.613500595092773, "learning_rate": 2.2055192986047804e-07, "logits/chosen": 0.02934134751558304, "logits/rejected": 0.0851244181394577, "logps/chosen": -275.94635009765625, "logps/ref_chosen": -60.889801025390625, "logps/ref_rejected": -77.965576171875, "logps/rejected": -476.11810302734375, "loss": 0.8583, "margin_dpo/margin_mean": 183.09597778320312, "margin_dpo/margin_std": 178.49737548828125, "step": 399 }, { "KL/chosen_KL_mean": -231.09521484375, "KL/mean": -350.0788879394531, "KL/rejected_KL_mean": -469.06256103515625, "KL/std": 186.72280883789062, "epoch": 0.5873715124816447, "fcm_dpo/beta": 0.004352675750851631, "fcm_dpo/delta": -0.26532918214797974, "fcm_dpo/margin": 237.9673309326172, "fcm_dpo/q_t": 0.2851880192756653, "grad_norm": 44.4686164855957, "learning_rate": 2.192779273338215e-07, "logits/chosen": 0.05125574767589569, "logits/rejected": 0.06881796568632126, "logps/chosen": -294.73883056640625, "logps/ref_chosen": -63.64359664916992, "logps/ref_rejected": -105.252685546875, "logps/rejected": -574.3152465820312, "loss": 0.7417, "margin_dpo/margin_mean": 237.9673614501953, "margin_dpo/margin_std": 195.74899291992188, "step": 400 }, { "KL/chosen_KL_mean": -268.1168212890625, "KL/mean": -336.1388854980469, "KL/rejected_KL_mean": -404.1609191894531, "KL/std": 176.03399658203125, "epoch": 0.5888399412628488, "fcm_dpo/beta": 0.004442146047949791, "fcm_dpo/delta": 0.2086574286222458, "fcm_dpo/margin": 136.04408264160156, "fcm_dpo/q_t": 0.3828345537185669, "grad_norm": 39.05641555786133, "learning_rate": 2.1800473436235136e-07, "logits/chosen": 0.07737800478935242, "logits/rejected": 0.08538491278886795, "logps/chosen": -325.27984619140625, "logps/ref_chosen": -57.16303253173828, "logps/ref_rejected": -83.79249572753906, "logps/rejected": -487.95343017578125, "loss": 1.0859, "margin_dpo/margin_mean": 136.0440673828125, "margin_dpo/margin_std": 228.36392211914062, "step": 401 }, { "KL/chosen_KL_mean": -213.17237854003906, "KL/mean": -337.1200256347656, "KL/rejected_KL_mean": -461.06768798828125, "KL/std": 202.32061767578125, "epoch": 0.5903083700440529, "fcm_dpo/beta": 0.004299672320485115, "fcm_dpo/delta": -0.29827845096588135, "fcm_dpo/margin": 247.89532470703125, "fcm_dpo/q_t": 0.2777937650680542, "grad_norm": 44.950050354003906, "learning_rate": 2.1673238449588665e-07, "logits/chosen": 0.08306563645601273, "logits/rejected": 0.10771772265434265, "logps/chosen": -263.9127502441406, "logps/ref_chosen": -50.74037170410156, "logps/ref_rejected": -81.0460433959961, "logps/rejected": -542.11376953125, "loss": 0.7104, "margin_dpo/margin_mean": 247.89532470703125, "margin_dpo/margin_std": 189.4843292236328, "step": 402 }, { "KL/chosen_KL_mean": -266.988037109375, "KL/mean": -361.72894287109375, "KL/rejected_KL_mean": -456.4698181152344, "KL/std": 186.5218505859375, "epoch": 0.591776798825257, "fcm_dpo/beta": 0.00422732625156641, "fcm_dpo/delta": -0.0012762416154146194, "fcm_dpo/margin": 189.48178100585938, "fcm_dpo/q_t": 0.33159884810447693, "grad_norm": 44.431949615478516, "learning_rate": 2.154609112620295e-07, "logits/chosen": 0.18813556432724, "logits/rejected": 0.19890309870243073, "logps/chosen": -314.1353454589844, "logps/ref_chosen": -47.14731216430664, "logps/ref_rejected": -77.2666015625, "logps/rejected": -533.7364501953125, "loss": 0.8712, "margin_dpo/margin_mean": 189.4817657470703, "margin_dpo/margin_std": 184.798583984375, "step": 403 }, { "KL/chosen_KL_mean": -298.219482421875, "KL/mean": -398.1234130859375, "KL/rejected_KL_mean": -498.02740478515625, "KL/std": 176.61068725585938, "epoch": 0.593245227606461, "fcm_dpo/beta": 0.004205549135804176, "fcm_dpo/delta": -0.044117238372564316, "fcm_dpo/margin": 199.80795288085938, "fcm_dpo/q_t": 0.3256443738937378, "grad_norm": 42.222572326660156, "learning_rate": 2.1419034816528218e-07, "logits/chosen": 0.2749345898628235, "logits/rejected": 0.29835087060928345, "logps/chosen": -346.0947265625, "logps/ref_chosen": -47.875274658203125, "logps/ref_rejected": -77.15499877929688, "logps/rejected": -575.182373046875, "loss": 0.8728, "margin_dpo/margin_mean": 199.80795288085938, "margin_dpo/margin_std": 206.64486694335938, "step": 404 }, { "KL/chosen_KL_mean": -318.4598083496094, "KL/mean": -398.934326171875, "KL/rejected_KL_mean": -479.40887451171875, "KL/std": 186.36871337890625, "epoch": 0.5947136563876652, "fcm_dpo/beta": 0.004319686442613602, "fcm_dpo/delta": 0.10862886905670166, "fcm_dpo/margin": 160.9490203857422, "fcm_dpo/q_t": 0.3593287467956543, "grad_norm": 67.92670440673828, "learning_rate": 2.129207286861638e-07, "logits/chosen": 0.15380798280239105, "logits/rejected": 0.1734359711408615, "logps/chosen": -383.62274169921875, "logps/ref_chosen": -65.16290283203125, "logps/ref_rejected": -87.18678283691406, "logps/rejected": -566.5956420898438, "loss": 0.9771, "margin_dpo/margin_mean": 160.94903564453125, "margin_dpo/margin_std": 202.40145874023438, "step": 405 }, { "KL/chosen_KL_mean": -313.0897216796875, "KL/mean": -410.04827880859375, "KL/rejected_KL_mean": -507.0068359375, "KL/std": 204.11318969726562, "epoch": 0.5961820851688693, "fcm_dpo/beta": 0.004310654476284981, "fcm_dpo/delta": -0.0407475009560585, "fcm_dpo/margin": 193.91712951660156, "fcm_dpo/q_t": 0.33301207423210144, "grad_norm": 39.647132873535156, "learning_rate": 2.1165208628032861e-07, "logits/chosen": 0.2193293273448944, "logits/rejected": 0.24135665595531464, "logps/chosen": -362.83050537109375, "logps/ref_chosen": -49.740814208984375, "logps/ref_rejected": -92.07862854003906, "logps/rejected": -599.08544921875, "loss": 0.8882, "margin_dpo/margin_mean": 193.91714477539062, "margin_dpo/margin_std": 212.67950439453125, "step": 406 }, { "KL/chosen_KL_mean": -339.6982421875, "KL/mean": -417.26922607421875, "KL/rejected_KL_mean": -494.8402099609375, "KL/std": 171.96035766601562, "epoch": 0.5976505139500734, "fcm_dpo/beta": 0.004385569132864475, "fcm_dpo/delta": 0.12469163537025452, "fcm_dpo/margin": 155.1419677734375, "fcm_dpo/q_t": 0.35749387741088867, "grad_norm": 62.01276779174805, "learning_rate": 2.1038445437768375e-07, "logits/chosen": 0.3068956434726715, "logits/rejected": 0.3434358835220337, "logps/chosen": -396.0289306640625, "logps/ref_chosen": -56.33069610595703, "logps/ref_rejected": -77.51209259033203, "logps/rejected": -572.352294921875, "loss": 0.9859, "margin_dpo/margin_mean": 155.1419677734375, "margin_dpo/margin_std": 195.72064208984375, "step": 407 }, { "KL/chosen_KL_mean": -355.1640625, "KL/mean": -427.469482421875, "KL/rejected_KL_mean": -499.77490234375, "KL/std": 156.67254638671875, "epoch": 0.5991189427312775, "fcm_dpo/beta": 0.00446933601051569, "fcm_dpo/delta": 0.16449061036109924, "fcm_dpo/margin": 144.61085510253906, "fcm_dpo/q_t": 0.36376452445983887, "grad_norm": 48.290897369384766, "learning_rate": 2.0911786638150872e-07, "logits/chosen": 0.24663367867469788, "logits/rejected": 0.28335240483283997, "logps/chosen": -424.953369140625, "logps/ref_chosen": -69.789306640625, "logps/ref_rejected": -90.09693908691406, "logps/rejected": -589.871826171875, "loss": 0.9679, "margin_dpo/margin_mean": 144.61085510253906, "margin_dpo/margin_std": 171.00955200195312, "step": 408 }, { "KL/chosen_KL_mean": -332.8118896484375, "KL/mean": -415.9732360839844, "KL/rejected_KL_mean": -499.1345520019531, "KL/std": 192.5561981201172, "epoch": 0.6005873715124816, "fcm_dpo/beta": 0.0045572854578495026, "fcm_dpo/delta": 0.045145727694034576, "fcm_dpo/margin": 166.3226776123047, "fcm_dpo/q_t": 0.34245866537094116, "grad_norm": 54.05333709716797, "learning_rate": 2.0785235566757517e-07, "logits/chosen": 0.17667712271213531, "logits/rejected": 0.1955818235874176, "logps/chosen": -400.12933349609375, "logps/ref_chosen": -67.31744384765625, "logps/ref_rejected": -84.904296875, "logps/rejected": -584.038818359375, "loss": 0.8977, "margin_dpo/margin_mean": 166.3226776123047, "margin_dpo/margin_std": 175.45404052734375, "step": 409 }, { "KL/chosen_KL_mean": -328.254150390625, "KL/mean": -412.47406005859375, "KL/rejected_KL_mean": -496.6939392089844, "KL/std": 183.95452880859375, "epoch": 0.6020558002936858, "fcm_dpo/beta": 0.004573145881295204, "fcm_dpo/delta": 0.032191164791584015, "fcm_dpo/margin": 168.43978881835938, "fcm_dpo/q_t": 0.33852922916412354, "grad_norm": 47.99551010131836, "learning_rate": 2.065879555832674e-07, "logits/chosen": 0.1765487790107727, "logits/rejected": 0.1874597668647766, "logps/chosen": -379.7195129394531, "logps/ref_chosen": -51.465354919433594, "logps/ref_rejected": -83.198974609375, "logps/rejected": -579.8929443359375, "loss": 0.9054, "margin_dpo/margin_mean": 168.43978881835938, "margin_dpo/margin_std": 182.5813751220703, "step": 410 }, { "KL/chosen_KL_mean": -345.7894287109375, "KL/mean": -429.1628112792969, "KL/rejected_KL_mean": -512.5361938476562, "KL/std": 178.01425170898438, "epoch": 0.6035242290748899, "fcm_dpo/beta": 0.004561232402920723, "fcm_dpo/delta": 0.040590718388557434, "fcm_dpo/margin": 166.74673461914062, "fcm_dpo/q_t": 0.3462197184562683, "grad_norm": 67.33753967285156, "learning_rate": 2.0532469944670343e-07, "logits/chosen": 0.30428582429885864, "logits/rejected": 0.32055914402008057, "logps/chosen": -398.0967102050781, "logps/ref_chosen": -52.30727005004883, "logps/ref_rejected": -80.69495391845703, "logps/rejected": -593.2311401367188, "loss": 0.9128, "margin_dpo/margin_mean": 166.7467498779297, "margin_dpo/margin_std": 185.79019165039062, "step": 411 }, { "KL/chosen_KL_mean": -347.623046875, "KL/mean": -430.669921875, "KL/rejected_KL_mean": -513.716796875, "KL/std": 168.90745544433594, "epoch": 0.604992657856094, "fcm_dpo/beta": 0.0046454789116978645, "fcm_dpo/delta": 0.03075917437672615, "fcm_dpo/margin": 166.09373474121094, "fcm_dpo/q_t": 0.34237197041511536, "grad_norm": 45.43087387084961, "learning_rate": 2.0406262054585738e-07, "logits/chosen": 0.2442624866962433, "logits/rejected": 0.2392767369747162, "logps/chosen": -400.7671813964844, "logps/ref_chosen": -53.144126892089844, "logps/ref_rejected": -100.0608139038086, "logps/rejected": -613.777587890625, "loss": 0.9166, "margin_dpo/margin_mean": 166.09373474121094, "margin_dpo/margin_std": 189.59783935546875, "step": 412 }, { "KL/chosen_KL_mean": -368.34149169921875, "KL/mean": -449.5134582519531, "KL/rejected_KL_mean": -530.6854248046875, "KL/std": 179.1047821044922, "epoch": 0.6064610866372981, "fcm_dpo/beta": 0.004703960847109556, "fcm_dpo/delta": 0.03838766738772392, "fcm_dpo/margin": 162.3439178466797, "fcm_dpo/q_t": 0.3409743309020996, "grad_norm": 68.5807113647461, "learning_rate": 2.0280175213768205e-07, "logits/chosen": 0.14463230967521667, "logits/rejected": 0.15965858101844788, "logps/chosen": -429.9234619140625, "logps/ref_chosen": -61.58196258544922, "logps/ref_rejected": -99.47340393066406, "logps/rejected": -630.1588134765625, "loss": 0.929, "margin_dpo/margin_mean": 162.3439178466797, "margin_dpo/margin_std": 190.46279907226562, "step": 413 }, { "KL/chosen_KL_mean": -328.48309326171875, "KL/mean": -433.293212890625, "KL/rejected_KL_mean": -538.1033935546875, "KL/std": 182.9401092529297, "epoch": 0.6079295154185022, "fcm_dpo/beta": 0.00460551492869854, "fcm_dpo/delta": -0.1832694113254547, "fcm_dpo/margin": 209.62030029296875, "fcm_dpo/q_t": 0.2981165647506714, "grad_norm": 48.85835266113281, "learning_rate": 2.0154212744723247e-07, "logits/chosen": 0.28938111662864685, "logits/rejected": 0.30956315994262695, "logps/chosen": -375.11456298828125, "logps/ref_chosen": -46.63148498535156, "logps/ref_rejected": -87.64653015136719, "logps/rejected": -625.7498779296875, "loss": 0.781, "margin_dpo/margin_mean": 209.62030029296875, "margin_dpo/margin_std": 176.873291015625, "step": 414 }, { "KL/chosen_KL_mean": -363.9873046875, "KL/mean": -444.49053955078125, "KL/rejected_KL_mean": -524.9938354492188, "KL/std": 190.98814392089844, "epoch": 0.6093979441997063, "fcm_dpo/beta": 0.00454542925581336, "fcm_dpo/delta": 0.07343503087759018, "fcm_dpo/margin": 161.0065155029297, "fcm_dpo/q_t": 0.3501279354095459, "grad_norm": 53.78864288330078, "learning_rate": 2.002837796667909e-07, "logits/chosen": 0.15784013271331787, "logits/rejected": 0.17198491096496582, "logps/chosen": -442.6055908203125, "logps/ref_chosen": -78.6182861328125, "logps/ref_rejected": -100.47752380371094, "logps/rejected": -625.4713134765625, "loss": 0.9377, "margin_dpo/margin_mean": 161.00653076171875, "margin_dpo/margin_std": 189.8529052734375, "step": 415 }, { "KL/chosen_KL_mean": -329.3747253417969, "KL/mean": -443.10009765625, "KL/rejected_KL_mean": -556.8255004882812, "KL/std": 199.22900390625, "epoch": 0.6108663729809104, "fcm_dpo/beta": 0.0044495644979178905, "fcm_dpo/delta": -0.23528949916362762, "fcm_dpo/margin": 227.45074462890625, "fcm_dpo/q_t": 0.28878968954086304, "grad_norm": 37.943359375, "learning_rate": 1.990267419549914e-07, "logits/chosen": 0.18898329138755798, "logits/rejected": 0.20634178817272186, "logps/chosen": -387.65386962890625, "logps/ref_chosen": -58.27912521362305, "logps/ref_rejected": -90.56871795654297, "logps/rejected": -647.3942260742188, "loss": 0.7443, "margin_dpo/margin_mean": 227.45074462890625, "margin_dpo/margin_std": 177.98248291015625, "step": 416 }, { "KL/chosen_KL_mean": -336.8597412109375, "KL/mean": -438.7206726074219, "KL/rejected_KL_mean": -540.5816650390625, "KL/std": 184.04443359375, "epoch": 0.6123348017621145, "fcm_dpo/beta": 0.004322909750044346, "fcm_dpo/delta": -0.08837594091892242, "fcm_dpo/margin": 203.72186279296875, "fcm_dpo/q_t": 0.31164246797561646, "grad_norm": 35.816017150878906, "learning_rate": 1.9777104743594686e-07, "logits/chosen": 0.3776714503765106, "logits/rejected": 0.4244968295097351, "logps/chosen": -387.0584716796875, "logps/ref_chosen": -50.1987190246582, "logps/ref_rejected": -68.15184020996094, "logps/rejected": -608.7334594726562, "loss": 0.7978, "margin_dpo/margin_mean": 203.72186279296875, "margin_dpo/margin_std": 167.75634765625, "step": 417 }, { "KL/chosen_KL_mean": -322.0848388671875, "KL/mean": -421.3946228027344, "KL/rejected_KL_mean": -520.7044067382812, "KL/std": 208.01580810546875, "epoch": 0.6138032305433186, "fcm_dpo/beta": 0.00430900976061821, "fcm_dpo/delta": -0.06341703236103058, "fcm_dpo/margin": 198.61956787109375, "fcm_dpo/q_t": 0.3312340974807739, "grad_norm": 37.52000427246094, "learning_rate": 1.965167291983757e-07, "logits/chosen": 0.09415233135223389, "logits/rejected": 0.1325725018978119, "logps/chosen": -404.0633239746094, "logps/ref_chosen": -81.97846984863281, "logps/ref_rejected": -104.69148254394531, "logps/rejected": -625.3958740234375, "loss": 0.8878, "margin_dpo/margin_mean": 198.6195526123047, "margin_dpo/margin_std": 221.0111083984375, "step": 418 }, { "KL/chosen_KL_mean": -281.81658935546875, "KL/mean": -387.3044738769531, "KL/rejected_KL_mean": -492.7923583984375, "KL/std": 190.3030548095703, "epoch": 0.6152716593245228, "fcm_dpo/beta": 0.004175344482064247, "fcm_dpo/delta": -0.08911103755235672, "fcm_dpo/margin": 210.97579956054688, "fcm_dpo/q_t": 0.3172077536582947, "grad_norm": 30.190216064453125, "learning_rate": 1.9526382029472988e-07, "logits/chosen": 0.056289106607437134, "logits/rejected": 0.07095717638731003, "logps/chosen": -334.7652282714844, "logps/ref_chosen": -52.948646545410156, "logps/ref_rejected": -91.58309936523438, "logps/rejected": -584.37548828125, "loss": 0.8322, "margin_dpo/margin_mean": 210.97579956054688, "margin_dpo/margin_std": 200.0166015625, "step": 419 }, { "KL/chosen_KL_mean": -297.31671142578125, "KL/mean": -375.4432678222656, "KL/rejected_KL_mean": -453.5698547363281, "KL/std": 171.12130737304688, "epoch": 0.6167400881057269, "fcm_dpo/beta": 0.004269109107553959, "fcm_dpo/delta": 0.14179980754852295, "fcm_dpo/margin": 156.25314331054688, "fcm_dpo/q_t": 0.3652932941913605, "grad_norm": 44.23408126831055, "learning_rate": 1.9401235374032425e-07, "logits/chosen": 0.11774572730064392, "logits/rejected": 0.1661440134048462, "logps/chosen": -375.08660888671875, "logps/ref_chosen": -77.7699203491211, "logps/ref_rejected": -69.31985473632812, "logps/rejected": -522.8897094726562, "loss": 0.9903, "margin_dpo/margin_mean": 156.25314331054688, "margin_dpo/margin_std": 211.12954711914062, "step": 420 }, { "KL/chosen_KL_mean": -253.75180053710938, "KL/mean": -329.0003662109375, "KL/rejected_KL_mean": -404.2489013671875, "KL/std": 165.86834716796875, "epoch": 0.618208516886931, "fcm_dpo/beta": 0.004399363417178392, "fcm_dpo/delta": 0.14674726128578186, "fcm_dpo/margin": 150.49708557128906, "fcm_dpo/q_t": 0.358406126499176, "grad_norm": 47.58359146118164, "learning_rate": 1.9276236251246653e-07, "logits/chosen": 0.04012005031108856, "logits/rejected": 0.05817845091223717, "logps/chosen": -307.51763916015625, "logps/ref_chosen": -53.765865325927734, "logps/ref_rejected": -89.28144836425781, "logps/rejected": -493.5303649902344, "loss": 0.9651, "margin_dpo/margin_mean": 150.49708557128906, "margin_dpo/margin_std": 175.0691680908203, "step": 421 }, { "KL/chosen_KL_mean": -280.9953308105469, "KL/mean": -360.83282470703125, "KL/rejected_KL_mean": -440.67034912109375, "KL/std": 170.18374633789062, "epoch": 0.6196769456681351, "fcm_dpo/beta": 0.004488244652748108, "fcm_dpo/delta": 0.08926088362932205, "fcm_dpo/margin": 159.67501831054688, "fcm_dpo/q_t": 0.348935067653656, "grad_norm": 43.716209411621094, "learning_rate": 1.9151387954958792e-07, "logits/chosen": 0.12674188613891602, "logits/rejected": 0.15268999338150024, "logps/chosen": -349.62908935546875, "logps/ref_chosen": -68.6337661743164, "logps/ref_rejected": -87.86351013183594, "logps/rejected": -528.5338745117188, "loss": 0.9463, "margin_dpo/margin_mean": 159.6750030517578, "margin_dpo/margin_std": 188.54898071289062, "step": 422 }, { "KL/chosen_KL_mean": -255.04405212402344, "KL/mean": -343.0213928222656, "KL/rejected_KL_mean": -430.998779296875, "KL/std": 158.25143432617188, "epoch": 0.6211453744493393, "fcm_dpo/beta": 0.004496719688177109, "fcm_dpo/delta": 0.009527413174510002, "fcm_dpo/margin": 175.95469665527344, "fcm_dpo/q_t": 0.3338441848754883, "grad_norm": 42.53166198730469, "learning_rate": 1.902669377503756e-07, "logits/chosen": 0.10014916211366653, "logits/rejected": 0.10393344610929489, "logps/chosen": -310.03436279296875, "logps/ref_chosen": -54.99030303955078, "logps/ref_rejected": -86.30654907226562, "logps/rejected": -517.3052978515625, "loss": 0.8742, "margin_dpo/margin_mean": 175.95469665527344, "margin_dpo/margin_std": 175.76312255859375, "step": 423 }, { "KL/chosen_KL_mean": -227.72738647460938, "KL/mean": -312.34954833984375, "KL/rejected_KL_mean": -396.9716796875, "KL/std": 165.03700256347656, "epoch": 0.6226138032305433, "fcm_dpo/beta": 0.004506932571530342, "fcm_dpo/delta": 0.03997078165411949, "fcm_dpo/margin": 169.24429321289062, "fcm_dpo/q_t": 0.34591084718704224, "grad_norm": 58.1711311340332, "learning_rate": 1.890215699729057e-07, "logits/chosen": 0.023241115733981133, "logits/rejected": 0.05266202241182327, "logps/chosen": -283.73931884765625, "logps/ref_chosen": -56.01192092895508, "logps/ref_rejected": -66.47896575927734, "logps/rejected": -463.45062255859375, "loss": 0.9243, "margin_dpo/margin_mean": 169.24429321289062, "margin_dpo/margin_std": 199.03924560546875, "step": 424 }, { "KL/chosen_KL_mean": -243.14434814453125, "KL/mean": -318.24798583984375, "KL/rejected_KL_mean": -393.3516540527344, "KL/std": 154.16336059570312, "epoch": 0.6240822320117474, "fcm_dpo/beta": 0.004639962688088417, "fcm_dpo/delta": 0.10944442451000214, "fcm_dpo/margin": 150.20730590820312, "fcm_dpo/q_t": 0.35073035955429077, "grad_norm": 32.66258239746094, "learning_rate": 1.8777780903377732e-07, "logits/chosen": 0.09410110861063004, "logits/rejected": 0.09336138516664505, "logps/chosen": -290.0133361816406, "logps/ref_chosen": -46.86899948120117, "logps/ref_rejected": -95.92545318603516, "logps/rejected": -489.277099609375, "loss": 0.9379, "margin_dpo/margin_mean": 150.20730590820312, "margin_dpo/margin_std": 164.84654235839844, "step": 425 }, { "KL/chosen_KL_mean": -237.71083068847656, "KL/mean": -332.9663391113281, "KL/rejected_KL_mean": -428.22186279296875, "KL/std": 173.30087280273438, "epoch": 0.6255506607929515, "fcm_dpo/beta": 0.0046137673780322075, "fcm_dpo/delta": -0.0870470330119133, "fcm_dpo/margin": 190.51104736328125, "fcm_dpo/q_t": 0.31865280866622925, "grad_norm": 66.36375427246094, "learning_rate": 1.8653568770724803e-07, "logits/chosen": 0.0031908173114061356, "logits/rejected": 0.057025354355573654, "logps/chosen": -314.29437255859375, "logps/ref_chosen": -76.58354187011719, "logps/ref_rejected": -81.26658630371094, "logps/rejected": -509.48846435546875, "loss": 0.841, "margin_dpo/margin_mean": 190.51101684570312, "margin_dpo/margin_std": 184.35989379882812, "step": 426 }, { "KL/chosen_KL_mean": -248.94723510742188, "KL/mean": -313.7269287109375, "KL/rejected_KL_mean": -378.50665283203125, "KL/std": 144.7581787109375, "epoch": 0.6270190895741556, "fcm_dpo/beta": 0.0046956404112279415, "fcm_dpo/delta": 0.20422431826591492, "fcm_dpo/margin": 129.55941772460938, "fcm_dpo/q_t": 0.3734450936317444, "grad_norm": 40.48114776611328, "learning_rate": 1.8529523872436977e-07, "logits/chosen": 0.058118224143981934, "logits/rejected": 0.08482979238033295, "logps/chosen": -313.8011169433594, "logps/ref_chosen": -64.8538818359375, "logps/ref_rejected": -78.5660171508789, "logps/rejected": -457.07269287109375, "loss": 0.9929, "margin_dpo/margin_mean": 129.55941772460938, "margin_dpo/margin_std": 163.17843627929688, "step": 427 }, { "KL/chosen_KL_mean": -307.9185791015625, "KL/mean": -392.6045837402344, "KL/rejected_KL_mean": -477.29058837890625, "KL/std": 186.76837158203125, "epoch": 0.6284875183553598, "fcm_dpo/beta": 0.004741638898849487, "fcm_dpo/delta": -0.003403865732252598, "fcm_dpo/margin": 169.37196350097656, "fcm_dpo/q_t": 0.3380899727344513, "grad_norm": 40.947059631347656, "learning_rate": 1.8405649477212697e-07, "logits/chosen": 0.03091556765139103, "logits/rejected": 0.043592870235443115, "logps/chosen": -370.5552673339844, "logps/ref_chosen": -62.63666534423828, "logps/ref_rejected": -103.28181457519531, "logps/rejected": -580.5723876953125, "loss": 0.9307, "margin_dpo/margin_mean": 169.37197875976562, "margin_dpo/margin_std": 204.71743774414062, "step": 428 }, { "KL/chosen_KL_mean": -319.4748840332031, "KL/mean": -398.7829284667969, "KL/rejected_KL_mean": -478.0909423828125, "KL/std": 163.31729125976562, "epoch": 0.6299559471365639, "fcm_dpo/beta": 0.004819995723664761, "fcm_dpo/delta": 0.03594332933425903, "fcm_dpo/margin": 158.6160430908203, "fcm_dpo/q_t": 0.3449682593345642, "grad_norm": 59.84564971923828, "learning_rate": 1.828194884925749e-07, "logits/chosen": 0.09432405978441238, "logits/rejected": 0.1432572454214096, "logps/chosen": -400.70892333984375, "logps/ref_chosen": -81.23401641845703, "logps/ref_rejected": -91.79493713378906, "logps/rejected": -569.8858642578125, "loss": 0.9474, "margin_dpo/margin_mean": 158.6160430908203, "margin_dpo/margin_std": 191.10890197753906, "step": 429 }, { "KL/chosen_KL_mean": -291.04022216796875, "KL/mean": -372.7721862792969, "KL/rejected_KL_mean": -454.504150390625, "KL/std": 185.45314025878906, "epoch": 0.631424375917768, "fcm_dpo/beta": 0.004818527959287167, "fcm_dpo/delta": 0.012564368546009064, "fcm_dpo/margin": 163.4639434814453, "fcm_dpo/q_t": 0.33801186084747314, "grad_norm": 33.16270446777344, "learning_rate": 1.8158425248197928e-07, "logits/chosen": 0.02384110540151596, "logits/rejected": 0.03353766351938248, "logps/chosen": -351.96051025390625, "logps/ref_chosen": -60.920326232910156, "logps/ref_rejected": -104.42280578613281, "logps/rejected": -558.927001953125, "loss": 0.8826, "margin_dpo/margin_mean": 163.4639434814453, "margin_dpo/margin_std": 169.9510955810547, "step": 430 }, { "KL/chosen_KL_mean": -307.9400329589844, "KL/mean": -417.70831298828125, "KL/rejected_KL_mean": -527.4765625, "KL/std": 187.70150756835938, "epoch": 0.6328928046989721, "fcm_dpo/beta": 0.004604096990078688, "fcm_dpo/delta": -0.23632526397705078, "fcm_dpo/margin": 219.53659057617188, "fcm_dpo/q_t": 0.2936067581176758, "grad_norm": 48.98324203491211, "learning_rate": 1.8035081928995788e-07, "logits/chosen": 0.19879499077796936, "logits/rejected": 0.2163088172674179, "logps/chosen": -365.28875732421875, "logps/ref_chosen": -57.34874725341797, "logps/ref_rejected": -92.84022521972656, "logps/rejected": -620.3167724609375, "loss": 0.7763, "margin_dpo/margin_mean": 219.53659057617188, "margin_dpo/margin_std": 192.73171997070312, "step": 431 }, { "KL/chosen_KL_mean": -305.25067138671875, "KL/mean": -402.177490234375, "KL/rejected_KL_mean": -499.10430908203125, "KL/std": 181.79977416992188, "epoch": 0.6343612334801763, "fcm_dpo/beta": 0.004561593756079674, "fcm_dpo/delta": -0.09420053660869598, "fcm_dpo/margin": 193.8536376953125, "fcm_dpo/q_t": 0.316600501537323, "grad_norm": 32.64216995239258, "learning_rate": 1.791192214186223e-07, "logits/chosen": 0.0449921078979969, "logits/rejected": 0.07461453974246979, "logps/chosen": -376.325439453125, "logps/ref_chosen": -71.07479095458984, "logps/ref_rejected": -98.57952880859375, "logps/rejected": -597.683837890625, "loss": 0.8207, "margin_dpo/margin_mean": 193.8536376953125, "margin_dpo/margin_std": 174.2122039794922, "step": 432 }, { "KL/chosen_KL_mean": -363.0186767578125, "KL/mean": -432.42047119140625, "KL/rejected_KL_mean": -501.82232666015625, "KL/std": 174.90367126464844, "epoch": 0.6358296622613803, "fcm_dpo/beta": 0.004609774332493544, "fcm_dpo/delta": 0.17070356011390686, "fcm_dpo/margin": 138.8036651611328, "fcm_dpo/q_t": 0.365825891494751, "grad_norm": 56.30537796020508, "learning_rate": 1.7788949132172193e-07, "logits/chosen": 0.16686058044433594, "logits/rejected": 0.1833561360836029, "logps/chosen": -421.2918701171875, "logps/ref_chosen": -58.273193359375, "logps/ref_rejected": -95.95089721679688, "logps/rejected": -597.773193359375, "loss": 1.0208, "margin_dpo/margin_mean": 138.8036651611328, "margin_dpo/margin_std": 193.2645721435547, "step": 433 }, { "KL/chosen_KL_mean": -309.8509521484375, "KL/mean": -392.6344299316406, "KL/rejected_KL_mean": -475.41790771484375, "KL/std": 162.32717895507812, "epoch": 0.6372980910425844, "fcm_dpo/beta": 0.00465200562030077, "fcm_dpo/delta": 0.032210834324359894, "fcm_dpo/margin": 165.56695556640625, "fcm_dpo/q_t": 0.34336385130882263, "grad_norm": 43.65876007080078, "learning_rate": 1.7666166140378853e-07, "logits/chosen": 0.15230190753936768, "logits/rejected": 0.17373520135879517, "logps/chosen": -371.82464599609375, "logps/ref_chosen": -61.97370147705078, "logps/ref_rejected": -78.49861145019531, "logps/rejected": -553.91650390625, "loss": 0.8959, "margin_dpo/margin_mean": 165.56695556640625, "margin_dpo/margin_std": 182.33963012695312, "step": 434 }, { "KL/chosen_KL_mean": -277.0976867675781, "KL/mean": -369.7054748535156, "KL/rejected_KL_mean": -462.3132629394531, "KL/std": 176.11752319335938, "epoch": 0.6387665198237885, "fcm_dpo/beta": 0.004619380459189415, "fcm_dpo/delta": -0.06102815270423889, "fcm_dpo/margin": 185.21556091308594, "fcm_dpo/q_t": 0.32273826003074646, "grad_norm": 27.393489837646484, "learning_rate": 1.7543576401928218e-07, "logits/chosen": 0.22029942274093628, "logits/rejected": 0.24756871163845062, "logps/chosen": -328.5997314453125, "logps/ref_chosen": -51.502052307128906, "logps/ref_rejected": -87.56689453125, "logps/rejected": -549.880126953125, "loss": 0.8636, "margin_dpo/margin_mean": 185.21554565429688, "margin_dpo/margin_std": 187.09512329101562, "step": 435 }, { "KL/chosen_KL_mean": -269.60888671875, "KL/mean": -361.96820068359375, "KL/rejected_KL_mean": -454.3275146484375, "KL/std": 168.82907104492188, "epoch": 0.6402349486049926, "fcm_dpo/beta": 0.004574541002511978, "fcm_dpo/delta": -0.049284037202596664, "fcm_dpo/margin": 184.71865844726562, "fcm_dpo/q_t": 0.3233397603034973, "grad_norm": 41.69789505004883, "learning_rate": 1.742118314717391e-07, "logits/chosen": 0.021279800683259964, "logits/rejected": 0.06457970291376114, "logps/chosen": -341.0125732421875, "logps/ref_chosen": -71.40371704101562, "logps/ref_rejected": -82.72775268554688, "logps/rejected": -537.0552978515625, "loss": 0.8396, "margin_dpo/margin_mean": 184.71865844726562, "margin_dpo/margin_std": 174.83840942382812, "step": 436 }, { "KL/chosen_KL_mean": -287.47174072265625, "KL/mean": -368.12969970703125, "KL/rejected_KL_mean": -448.78765869140625, "KL/std": 144.98049926757812, "epoch": 0.6417033773861968, "fcm_dpo/beta": 0.004624016582965851, "fcm_dpo/delta": 0.058035340160131454, "fcm_dpo/margin": 161.31593322753906, "fcm_dpo/q_t": 0.34285643696784973, "grad_norm": 31.177148818969727, "learning_rate": 1.7298989601292036e-07, "logits/chosen": 0.14124414324760437, "logits/rejected": 0.1843366026878357, "logps/chosen": -352.21600341796875, "logps/ref_chosen": -64.7442626953125, "logps/ref_rejected": -82.04356384277344, "logps/rejected": -530.8312377929688, "loss": 0.903, "margin_dpo/margin_mean": 161.31593322753906, "margin_dpo/margin_std": 168.4124755859375, "step": 437 }, { "KL/chosen_KL_mean": -271.91680908203125, "KL/mean": -367.9270935058594, "KL/rejected_KL_mean": -463.9373779296875, "KL/std": 161.41934204101562, "epoch": 0.6431718061674009, "fcm_dpo/beta": 0.004532171413302422, "fcm_dpo/delta": -0.07903608679771423, "fcm_dpo/margin": 192.02059936523438, "fcm_dpo/q_t": 0.315254807472229, "grad_norm": 40.19487762451172, "learning_rate": 1.7176998984196144e-07, "logits/chosen": 0.16977214813232422, "logits/rejected": 0.21760720014572144, "logps/chosen": -330.93548583984375, "logps/ref_chosen": -59.0186653137207, "logps/ref_rejected": -83.07682800292969, "logps/rejected": -547.0142211914062, "loss": 0.8096, "margin_dpo/margin_mean": 192.02059936523438, "margin_dpo/margin_std": 160.93798828125, "step": 438 }, { "KL/chosen_KL_mean": -286.5731201171875, "KL/mean": -363.4912109375, "KL/rejected_KL_mean": -440.4093322753906, "KL/std": 169.75950622558594, "epoch": 0.644640234948605, "fcm_dpo/beta": 0.00456035602837801, "fcm_dpo/delta": 0.10353709012269974, "fcm_dpo/margin": 153.836181640625, "fcm_dpo/q_t": 0.3557645082473755, "grad_norm": 33.6142463684082, "learning_rate": 1.7055214510452458e-07, "logits/chosen": 0.15101350843906403, "logits/rejected": 0.15643706917762756, "logps/chosen": -340.3572082519531, "logps/ref_chosen": -53.78407669067383, "logps/ref_rejected": -83.98545837402344, "logps/rejected": -524.394775390625, "loss": 0.9447, "margin_dpo/margin_mean": 153.836181640625, "margin_dpo/margin_std": 170.22177124023438, "step": 439 }, { "KL/chosen_KL_mean": -303.6018981933594, "KL/mean": -376.49786376953125, "KL/rejected_KL_mean": -449.39385986328125, "KL/std": 186.78781127929688, "epoch": 0.6461086637298091, "fcm_dpo/beta": 0.004747721366584301, "fcm_dpo/delta": 0.11485690623521805, "fcm_dpo/margin": 145.7919464111328, "fcm_dpo/q_t": 0.3601565361022949, "grad_norm": 40.849365234375, "learning_rate": 1.6933639389195134e-07, "logits/chosen": 0.10528746247291565, "logits/rejected": 0.14277926087379456, "logps/chosen": -382.16864013671875, "logps/ref_chosen": -78.56671905517578, "logps/ref_rejected": -96.49775695800781, "logps/rejected": -545.8916015625, "loss": 0.9642, "margin_dpo/margin_mean": 145.79196166992188, "margin_dpo/margin_std": 182.39181518554688, "step": 440 }, { "KL/chosen_KL_mean": -312.75555419921875, "KL/mean": -401.1040344238281, "KL/rejected_KL_mean": -489.45245361328125, "KL/std": 192.77627563476562, "epoch": 0.6475770925110133, "fcm_dpo/beta": 0.0047426121309399605, "fcm_dpo/delta": -0.04161972180008888, "fcm_dpo/margin": 176.6968994140625, "fcm_dpo/q_t": 0.33483031392097473, "grad_norm": 37.40399932861328, "learning_rate": 1.681227682404166e-07, "logits/chosen": 0.1145353838801384, "logits/rejected": 0.13276338577270508, "logps/chosen": -373.58001708984375, "logps/ref_chosen": -60.824440002441406, "logps/ref_rejected": -96.47080993652344, "logps/rejected": -585.9232788085938, "loss": 0.9214, "margin_dpo/margin_mean": 176.6968994140625, "margin_dpo/margin_std": 215.09902954101562, "step": 441 }, { "KL/chosen_KL_mean": -285.8372802734375, "KL/mean": -389.34747314453125, "KL/rejected_KL_mean": -492.857666015625, "KL/std": 188.40069580078125, "epoch": 0.6490455212922174, "fcm_dpo/beta": 0.004635802935808897, "fcm_dpo/delta": -0.17678695917129517, "fcm_dpo/margin": 207.02035522460938, "fcm_dpo/q_t": 0.30531585216522217, "grad_norm": 27.13920783996582, "learning_rate": 1.669113001300851e-07, "logits/chosen": 0.11083649098873138, "logits/rejected": 0.13127142190933228, "logps/chosen": -332.8485107421875, "logps/ref_chosen": -47.01121520996094, "logps/ref_rejected": -76.53926086425781, "logps/rejected": -569.3969116210938, "loss": 0.7884, "margin_dpo/margin_mean": 207.02035522460938, "margin_dpo/margin_std": 183.5563507080078, "step": 442 }, { "KL/chosen_KL_mean": -326.73773193359375, "KL/mean": -398.07366943359375, "KL/rejected_KL_mean": -469.4096374511719, "KL/std": 170.14450073242188, "epoch": 0.6505139500734214, "fcm_dpo/beta": 0.004615455400198698, "fcm_dpo/delta": 0.15127958357334137, "fcm_dpo/margin": 142.67193603515625, "fcm_dpo/q_t": 0.3653959035873413, "grad_norm": 64.4756851196289, "learning_rate": 1.6570202148426815e-07, "logits/chosen": 0.12177616357803345, "logits/rejected": 0.13952183723449707, "logps/chosen": -398.0107421875, "logps/ref_chosen": -71.27301788330078, "logps/ref_rejected": -86.679931640625, "logps/rejected": -556.089599609375, "loss": 1.0129, "margin_dpo/margin_mean": 142.6719207763672, "margin_dpo/margin_std": 198.99505615234375, "step": 443 }, { "KL/chosen_KL_mean": -347.0489501953125, "KL/mean": -446.3876953125, "KL/rejected_KL_mean": -545.7264404296875, "KL/std": 186.73049926757812, "epoch": 0.6519823788546255, "fcm_dpo/beta": 0.004596185404807329, "fcm_dpo/delta": -0.12484229356050491, "fcm_dpo/margin": 198.67745971679688, "fcm_dpo/q_t": 0.31873539090156555, "grad_norm": 42.87467575073242, "learning_rate": 1.6449496416858282e-07, "logits/chosen": 0.24206379055976868, "logits/rejected": 0.2568224370479584, "logps/chosen": -404.2626647949219, "logps/ref_chosen": -57.213706970214844, "logps/ref_rejected": -97.25489807128906, "logps/rejected": -642.9813232421875, "loss": 0.8479, "margin_dpo/margin_mean": 198.67747497558594, "margin_dpo/margin_std": 211.93258666992188, "step": 444 }, { "KL/chosen_KL_mean": -325.6207580566406, "KL/mean": -424.31805419921875, "KL/rejected_KL_mean": -523.0152587890625, "KL/std": 176.32916259765625, "epoch": 0.6534508076358296, "fcm_dpo/beta": 0.004501967690885067, "fcm_dpo/delta": -0.0975315123796463, "fcm_dpo/margin": 197.39456176757812, "fcm_dpo/q_t": 0.319426029920578, "grad_norm": 37.24625015258789, "learning_rate": 1.6329015999011182e-07, "logits/chosen": 0.16529598832130432, "logits/rejected": 0.19533869624137878, "logps/chosen": -392.9205322265625, "logps/ref_chosen": -67.29979705810547, "logps/ref_rejected": -92.68267059326172, "logps/rejected": -615.697998046875, "loss": 0.8327, "margin_dpo/margin_mean": 197.39456176757812, "margin_dpo/margin_std": 190.868896484375, "step": 445 }, { "KL/chosen_KL_mean": -309.30706787109375, "KL/mean": -408.1361999511719, "KL/rejected_KL_mean": -506.96533203125, "KL/std": 178.90072631835938, "epoch": 0.6549192364170338, "fcm_dpo/beta": 0.004462126176804304, "fcm_dpo/delta": -0.09048521518707275, "fcm_dpo/margin": 197.65823364257812, "fcm_dpo/q_t": 0.31490451097488403, "grad_norm": 40.9002799987793, "learning_rate": 1.6208764069656578e-07, "logits/chosen": 0.2029511034488678, "logits/rejected": 0.20954206585884094, "logps/chosen": -368.40557861328125, "logps/ref_chosen": -59.098487854003906, "logps/ref_rejected": -101.26419067382812, "logps/rejected": -608.2294921875, "loss": 0.8093, "margin_dpo/margin_mean": 197.65823364257812, "margin_dpo/margin_std": 169.7216796875, "step": 446 }, { "KL/chosen_KL_mean": -334.24090576171875, "KL/mean": -437.32513427734375, "KL/rejected_KL_mean": -540.4093627929688, "KL/std": 209.1260986328125, "epoch": 0.6563876651982379, "fcm_dpo/beta": 0.004316662438213825, "fcm_dpo/delta": -0.10000540316104889, "fcm_dpo/margin": 206.16847229003906, "fcm_dpo/q_t": 0.3189159333705902, "grad_norm": 46.66408157348633, "learning_rate": 1.608874379754465e-07, "logits/chosen": 0.19859518110752106, "logits/rejected": 0.20376473665237427, "logps/chosen": -390.31622314453125, "logps/ref_chosen": -56.07533264160156, "logps/ref_rejected": -98.69475555419922, "logps/rejected": -639.1041259765625, "loss": 0.8258, "margin_dpo/margin_mean": 206.16847229003906, "margin_dpo/margin_std": 204.78793334960938, "step": 447 }, { "KL/chosen_KL_mean": -352.8293762207031, "KL/mean": -461.22027587890625, "KL/rejected_KL_mean": -569.6112060546875, "KL/std": 188.87646484375, "epoch": 0.657856093979442, "fcm_dpo/beta": 0.004251912701874971, "fcm_dpo/delta": -0.1339014321565628, "fcm_dpo/margin": 216.7818145751953, "fcm_dpo/q_t": 0.30998510122299194, "grad_norm": 50.98678970336914, "learning_rate": 1.5968958345321177e-07, "logits/chosen": 0.1657218635082245, "logits/rejected": 0.1783699095249176, "logps/chosen": -412.8332214355469, "logps/ref_chosen": -60.00384521484375, "logps/ref_rejected": -102.26465606689453, "logps/rejected": -671.8758544921875, "loss": 0.7985, "margin_dpo/margin_mean": 216.78179931640625, "margin_dpo/margin_std": 192.1991424560547, "step": 448 }, { "KL/chosen_KL_mean": -352.67364501953125, "KL/mean": -451.06048583984375, "KL/rejected_KL_mean": -549.4473266601562, "KL/std": 202.05987548828125, "epoch": 0.6593245227606461, "fcm_dpo/beta": 0.004183411132544279, "fcm_dpo/delta": -0.025334613397717476, "fcm_dpo/margin": 196.7737274169922, "fcm_dpo/q_t": 0.33970075845718384, "grad_norm": 35.54926300048828, "learning_rate": 1.584941086944423e-07, "logits/chosen": 0.14113399386405945, "logits/rejected": 0.17219001054763794, "logps/chosen": -420.20025634765625, "logps/ref_chosen": -67.52661895751953, "logps/ref_rejected": -88.59690856933594, "logps/rejected": -638.0442504882812, "loss": 0.9209, "margin_dpo/margin_mean": 196.77371215820312, "margin_dpo/margin_std": 245.24896240234375, "step": 449 }, { "KL/chosen_KL_mean": -308.78033447265625, "KL/mean": -417.45819091796875, "KL/rejected_KL_mean": -526.1360473632812, "KL/std": 202.91213989257812, "epoch": 0.6607929515418502, "fcm_dpo/beta": 0.004141074605286121, "fcm_dpo/delta": -0.11008300632238388, "fcm_dpo/margin": 217.355712890625, "fcm_dpo/q_t": 0.30950620770454407, "grad_norm": 32.117591857910156, "learning_rate": 1.573010452010098e-07, "logits/chosen": 0.22345803678035736, "logits/rejected": 0.23786935210227966, "logps/chosen": -365.8884582519531, "logps/ref_chosen": -57.10811996459961, "logps/ref_rejected": -102.75494384765625, "logps/rejected": -628.8909912109375, "loss": 0.8082, "margin_dpo/margin_mean": 217.355712890625, "margin_dpo/margin_std": 193.65924072265625, "step": 450 }, { "KL/chosen_KL_mean": -373.90386962890625, "KL/mean": -466.65606689453125, "KL/rejected_KL_mean": -559.4083251953125, "KL/std": 207.1577606201172, "epoch": 0.6622613803230544, "fcm_dpo/beta": 0.004139560274779797, "fcm_dpo/delta": 0.033666130155324936, "fcm_dpo/margin": 185.50445556640625, "fcm_dpo/q_t": 0.33645984530448914, "grad_norm": 49.498046875, "learning_rate": 1.5611042441124687e-07, "logits/chosen": 0.3563879132270813, "logits/rejected": 0.3978639543056488, "logps/chosen": -432.3726806640625, "logps/ref_chosen": -58.46883010864258, "logps/ref_rejected": -72.92941284179688, "logps/rejected": -632.3377075195312, "loss": 0.9322, "margin_dpo/margin_mean": 185.50445556640625, "margin_dpo/margin_std": 218.109130859375, "step": 451 }, { "KL/chosen_KL_mean": -291.71258544921875, "KL/mean": -394.4192199707031, "KL/rejected_KL_mean": -497.1258544921875, "KL/std": 187.85025024414062, "epoch": 0.6637298091042585, "fcm_dpo/beta": 0.004082968458533287, "fcm_dpo/delta": -0.04251035302877426, "fcm_dpo/margin": 205.4132843017578, "fcm_dpo/q_t": 0.3228408098220825, "grad_norm": 27.555904388427734, "learning_rate": 1.549222776991186e-07, "logits/chosen": 0.2295912206172943, "logits/rejected": 0.2270086109638214, "logps/chosen": -342.1031494140625, "logps/ref_chosen": -50.39055252075195, "logps/ref_rejected": -97.77142333984375, "logps/rejected": -594.8972778320312, "loss": 0.8281, "margin_dpo/margin_mean": 205.4132843017578, "margin_dpo/margin_std": 183.20945739746094, "step": 452 }, { "KL/chosen_KL_mean": -329.4156799316406, "KL/mean": -421.95611572265625, "KL/rejected_KL_mean": -514.4965209960938, "KL/std": 187.1441650390625, "epoch": 0.6651982378854625, "fcm_dpo/beta": 0.0040916576981544495, "fcm_dpo/delta": 0.04588655009865761, "fcm_dpo/margin": 185.08084106445312, "fcm_dpo/q_t": 0.3440532088279724, "grad_norm": 33.83127975463867, "learning_rate": 1.5373663637339584e-07, "logits/chosen": 0.14756467938423157, "logits/rejected": 0.19183680415153503, "logps/chosen": -387.13055419921875, "logps/ref_chosen": -57.71485137939453, "logps/ref_rejected": -82.20741271972656, "logps/rejected": -596.7039794921875, "loss": 0.8993, "margin_dpo/margin_mean": 185.08082580566406, "margin_dpo/margin_std": 200.801025390625, "step": 453 }, { "KL/chosen_KL_mean": -362.6598815917969, "KL/mean": -471.7563171386719, "KL/rejected_KL_mean": -580.852783203125, "KL/std": 216.17813110351562, "epoch": 0.6666666666666666, "fcm_dpo/beta": 0.004048048052936792, "fcm_dpo/delta": -0.09220601618289948, "fcm_dpo/margin": 218.19287109375, "fcm_dpo/q_t": 0.32163387537002563, "grad_norm": 33.92574691772461, "learning_rate": 1.5255353167683017e-07, "logits/chosen": 0.24445435404777527, "logits/rejected": 0.27604761719703674, "logps/chosen": -423.60552978515625, "logps/ref_chosen": -60.945648193359375, "logps/ref_rejected": -84.95079040527344, "logps/rejected": -665.8035888671875, "loss": 0.8446, "margin_dpo/margin_mean": 218.19287109375, "margin_dpo/margin_std": 226.92523193359375, "step": 454 }, { "KL/chosen_KL_mean": -319.8667297363281, "KL/mean": -438.78204345703125, "KL/rejected_KL_mean": -557.6973876953125, "KL/std": 214.24429321289062, "epoch": 0.6681350954478708, "fcm_dpo/beta": 0.003983162343502045, "fcm_dpo/delta": -0.16275690495967865, "fcm_dpo/margin": 237.83067321777344, "fcm_dpo/q_t": 0.3083864152431488, "grad_norm": 38.621768951416016, "learning_rate": 1.5137299478533064e-07, "logits/chosen": 0.14638906717300415, "logits/rejected": 0.15352049469947815, "logps/chosen": -364.75341796875, "logps/ref_chosen": -44.88671112060547, "logps/ref_rejected": -115.30147552490234, "logps/rejected": -672.9989013671875, "loss": 0.8024, "margin_dpo/margin_mean": 237.83065795898438, "margin_dpo/margin_std": 217.93719482421875, "step": 455 }, { "KL/chosen_KL_mean": -313.37225341796875, "KL/mean": -429.8499755859375, "KL/rejected_KL_mean": -546.32763671875, "KL/std": 212.5858154296875, "epoch": 0.6696035242290749, "fcm_dpo/beta": 0.0038606026209890842, "fcm_dpo/delta": -0.10905548930168152, "fcm_dpo/margin": 232.95541381835938, "fcm_dpo/q_t": 0.31055253744125366, "grad_norm": 32.60857009887695, "learning_rate": 1.5019505680714232e-07, "logits/chosen": 0.1347317397594452, "logits/rejected": 0.13765022158622742, "logps/chosen": -370.4090270996094, "logps/ref_chosen": -57.036781311035156, "logps/ref_rejected": -105.21784210205078, "logps/rejected": -651.5455322265625, "loss": 0.7867, "margin_dpo/margin_mean": 232.9553985595703, "margin_dpo/margin_std": 192.2313232421875, "step": 456 }, { "KL/chosen_KL_mean": -303.17724609375, "KL/mean": -408.58001708984375, "KL/rejected_KL_mean": -513.9827880859375, "KL/std": 202.3392333984375, "epoch": 0.671071953010279, "fcm_dpo/beta": 0.0038118790835142136, "fcm_dpo/delta": -0.004409227520227432, "fcm_dpo/margin": 210.80557250976562, "fcm_dpo/q_t": 0.32770806550979614, "grad_norm": 39.375999450683594, "learning_rate": 1.4901974878202627e-07, "logits/chosen": 0.1512976586818695, "logits/rejected": 0.1942361295223236, "logps/chosen": -357.4197998046875, "logps/ref_chosen": -54.24253845214844, "logps/ref_rejected": -85.10956573486328, "logps/rejected": -599.0924072265625, "loss": 0.8532, "margin_dpo/margin_mean": 210.80555725097656, "margin_dpo/margin_std": 192.32322692871094, "step": 457 }, { "KL/chosen_KL_mean": -301.6228332519531, "KL/mean": -418.083251953125, "KL/rejected_KL_mean": -534.543701171875, "KL/std": 199.40939331054688, "epoch": 0.6725403817914831, "fcm_dpo/beta": 0.003761657979339361, "fcm_dpo/delta": -0.08457423001527786, "fcm_dpo/margin": 232.92083740234375, "fcm_dpo/q_t": 0.3179261386394501, "grad_norm": 29.335128784179688, "learning_rate": 1.4784710168044212e-07, "logits/chosen": 0.19247975945472717, "logits/rejected": 0.2139834612607956, "logps/chosen": -357.0317077636719, "logps/ref_chosen": -55.40888214111328, "logps/ref_rejected": -97.68325805664062, "logps/rejected": -632.2269287109375, "loss": 0.8221, "margin_dpo/margin_mean": 232.92083740234375, "margin_dpo/margin_std": 219.72933959960938, "step": 458 }, { "KL/chosen_KL_mean": -328.488037109375, "KL/mean": -435.84149169921875, "KL/rejected_KL_mean": -543.1949462890625, "KL/std": 195.81832885742188, "epoch": 0.6740088105726872, "fcm_dpo/beta": 0.0037542409263551235, "fcm_dpo/delta": -0.006693059578537941, "fcm_dpo/margin": 214.70687866210938, "fcm_dpo/q_t": 0.3288930654525757, "grad_norm": 31.528162002563477, "learning_rate": 1.466771464027316e-07, "logits/chosen": 0.19821825623512268, "logits/rejected": 0.21426156163215637, "logps/chosen": -375.0455322265625, "logps/ref_chosen": -46.55748748779297, "logps/ref_rejected": -86.16854095458984, "logps/rejected": -629.363525390625, "loss": 0.8524, "margin_dpo/margin_mean": 214.70687866210938, "margin_dpo/margin_std": 196.4297332763672, "step": 459 }, { "KL/chosen_KL_mean": -353.2541809082031, "KL/mean": -470.1562805175781, "KL/rejected_KL_mean": -587.058349609375, "KL/std": 202.15203857421875, "epoch": 0.6754772393538914, "fcm_dpo/beta": 0.0037221387028694153, "fcm_dpo/delta": -0.07688962668180466, "fcm_dpo/margin": 233.80418395996094, "fcm_dpo/q_t": 0.31593120098114014, "grad_norm": 36.09447479248047, "learning_rate": 1.4550991377830423e-07, "logits/chosen": 0.11949439346790314, "logits/rejected": 0.1151965856552124, "logps/chosen": -404.88909912109375, "logps/ref_chosen": -51.63489532470703, "logps/ref_rejected": -104.11935424804688, "logps/rejected": -691.177734375, "loss": 0.8139, "margin_dpo/margin_mean": 233.80418395996094, "margin_dpo/margin_std": 204.37997436523438, "step": 460 }, { "KL/chosen_KL_mean": -376.63531494140625, "KL/mean": -465.3780517578125, "KL/rejected_KL_mean": -554.1207885742188, "KL/std": 202.32070922851562, "epoch": 0.6769456681350955, "fcm_dpo/beta": 0.0037951446138322353, "fcm_dpo/delta": 0.13449496030807495, "fcm_dpo/margin": 177.4854736328125, "fcm_dpo/q_t": 0.3620460033416748, "grad_norm": 37.02994155883789, "learning_rate": 1.4434543456482518e-07, "logits/chosen": 0.14143106341362, "logits/rejected": 0.15454688668251038, "logps/chosen": -431.8172607421875, "logps/ref_chosen": -55.18195724487305, "logps/ref_rejected": -86.47689819335938, "logps/rejected": -640.59765625, "loss": 0.962, "margin_dpo/margin_mean": 177.48548889160156, "margin_dpo/margin_std": 217.33816528320312, "step": 461 }, { "KL/chosen_KL_mean": -361.41998291015625, "KL/mean": -443.12005615234375, "KL/rejected_KL_mean": -524.820068359375, "KL/std": 192.26971435546875, "epoch": 0.6784140969162996, "fcm_dpo/beta": 0.0038944468833506107, "fcm_dpo/delta": 0.1750825047492981, "fcm_dpo/margin": 163.40008544921875, "fcm_dpo/q_t": 0.36596518754959106, "grad_norm": 50.11738967895508, "learning_rate": 1.4318373944740484e-07, "logits/chosen": 0.20698338747024536, "logits/rejected": 0.23005270957946777, "logps/chosen": -431.3480224609375, "logps/ref_chosen": -69.92803192138672, "logps/ref_rejected": -78.84111022949219, "logps/rejected": -603.6611938476562, "loss": 0.9594, "margin_dpo/margin_mean": 163.40008544921875, "margin_dpo/margin_std": 186.82803344726562, "step": 462 }, { "KL/chosen_KL_mean": -377.38116455078125, "KL/mean": -470.04058837890625, "KL/rejected_KL_mean": -562.7000732421875, "KL/std": 201.7800750732422, "epoch": 0.6798825256975036, "fcm_dpo/beta": 0.004008530639111996, "fcm_dpo/delta": 0.05973725765943527, "fcm_dpo/margin": 185.31890869140625, "fcm_dpo/q_t": 0.3500801920890808, "grad_norm": 59.69301223754883, "learning_rate": 1.4202485903778976e-07, "logits/chosen": 0.18849530816078186, "logits/rejected": 0.20773401856422424, "logps/chosen": -432.655517578125, "logps/ref_chosen": -55.27437210083008, "logps/ref_rejected": -89.02497863769531, "logps/rejected": -651.7250366210938, "loss": 0.9511, "margin_dpo/margin_mean": 185.31890869140625, "margin_dpo/margin_std": 225.99717712402344, "step": 463 }, { "KL/chosen_KL_mean": -368.3976135253906, "KL/mean": -504.92755126953125, "KL/rejected_KL_mean": -641.4574584960938, "KL/std": 232.31454467773438, "epoch": 0.6813509544787077, "fcm_dpo/beta": 0.0038046562112867832, "fcm_dpo/delta": -0.2696046829223633, "fcm_dpo/margin": 273.0598449707031, "fcm_dpo/q_t": 0.28859981894493103, "grad_norm": 39.533355712890625, "learning_rate": 1.4086882387355658e-07, "logits/chosen": 0.23187042772769928, "logits/rejected": 0.23417802155017853, "logps/chosen": -419.3099365234375, "logps/ref_chosen": -50.91230010986328, "logps/ref_rejected": -102.4893798828125, "logps/rejected": -743.9468383789062, "loss": 0.7533, "margin_dpo/margin_mean": 273.0598449707031, "margin_dpo/margin_std": 236.38284301757812, "step": 464 }, { "KL/chosen_KL_mean": -363.36859130859375, "KL/mean": -487.6932067871094, "KL/rejected_KL_mean": -612.017822265625, "KL/std": 230.75079345703125, "epoch": 0.6828193832599119, "fcm_dpo/beta": 0.0037093700375407934, "fcm_dpo/delta": -0.1346028596162796, "fcm_dpo/margin": 248.6492156982422, "fcm_dpo/q_t": 0.30622804164886475, "grad_norm": 36.44754409790039, "learning_rate": 1.3971566441730714e-07, "logits/chosen": 0.21581992506980896, "logits/rejected": 0.22753633558750153, "logps/chosen": -423.4854431152344, "logps/ref_chosen": -60.116851806640625, "logps/ref_rejected": -113.94602966308594, "logps/rejected": -725.9638671875, "loss": 0.7884, "margin_dpo/margin_mean": 248.64923095703125, "margin_dpo/margin_std": 211.38388061523438, "step": 465 }, { "KL/chosen_KL_mean": -404.96728515625, "KL/mean": -510.61505126953125, "KL/rejected_KL_mean": -616.2628173828125, "KL/std": 214.00537109375, "epoch": 0.684287812041116, "fcm_dpo/beta": 0.00367511878721416, "fcm_dpo/delta": 0.025151047855615616, "fcm_dpo/margin": 211.29556274414062, "fcm_dpo/q_t": 0.33769696950912476, "grad_norm": 48.69867706298828, "learning_rate": 1.3856541105586545e-07, "logits/chosen": 0.19872266054153442, "logits/rejected": 0.20927634835243225, "logps/chosen": -457.88818359375, "logps/ref_chosen": -52.920921325683594, "logps/ref_rejected": -90.3154296875, "logps/rejected": -706.5782470703125, "loss": 0.8961, "margin_dpo/margin_mean": 211.29556274414062, "margin_dpo/margin_std": 224.05996704101562, "step": 466 }, { "KL/chosen_KL_mean": -474.77044677734375, "KL/mean": -590.2333984375, "KL/rejected_KL_mean": -705.6962890625, "KL/std": 236.89700317382812, "epoch": 0.6857562408223201, "fcm_dpo/beta": 0.0036290897987782955, "fcm_dpo/delta": -0.044904135167598724, "fcm_dpo/margin": 230.92593383789062, "fcm_dpo/q_t": 0.3338162899017334, "grad_norm": 70.50183868408203, "learning_rate": 1.3741809409947729e-07, "logits/chosen": 0.24021601676940918, "logits/rejected": 0.2654044032096863, "logps/chosen": -553.4862670898438, "logps/ref_chosen": -78.7158203125, "logps/ref_rejected": -102.86019897460938, "logps/rejected": -808.5565185546875, "loss": 0.9236, "margin_dpo/margin_mean": 230.92593383789062, "margin_dpo/margin_std": 273.190673828125, "step": 467 }, { "KL/chosen_KL_mean": -392.1413269042969, "KL/mean": -511.47100830078125, "KL/rejected_KL_mean": -630.8007202148438, "KL/std": 235.54798889160156, "epoch": 0.6872246696035242, "fcm_dpo/beta": 0.003602225799113512, "fcm_dpo/delta": -0.06676232814788818, "fcm_dpo/margin": 238.6593780517578, "fcm_dpo/q_t": 0.32230842113494873, "grad_norm": 40.31625747680664, "learning_rate": 1.362737437810114e-07, "logits/chosen": 0.16735252737998962, "logits/rejected": 0.20658326148986816, "logps/chosen": -462.0766906738281, "logps/ref_chosen": -69.93536376953125, "logps/ref_rejected": -101.02880859375, "logps/rejected": -731.8295288085938, "loss": 0.8509, "margin_dpo/margin_mean": 238.6593780517578, "margin_dpo/margin_std": 236.20262145996094, "step": 468 }, { "KL/chosen_KL_mean": -438.883056640625, "KL/mean": -564.942138671875, "KL/rejected_KL_mean": -691.0013427734375, "KL/std": 231.7757568359375, "epoch": 0.6886930983847284, "fcm_dpo/beta": 0.0035334881395101547, "fcm_dpo/delta": -0.10173042863607407, "fcm_dpo/margin": 252.11825561523438, "fcm_dpo/q_t": 0.3151817321777344, "grad_norm": 39.68192672729492, "learning_rate": 1.351323902551631e-07, "logits/chosen": 0.17181815207004547, "logits/rejected": 0.21328867971897125, "logps/chosen": -507.00775146484375, "logps/ref_chosen": -68.12469482421875, "logps/ref_rejected": -104.78640747070312, "logps/rejected": -795.7877197265625, "loss": 0.8161, "margin_dpo/margin_mean": 252.11825561523438, "margin_dpo/margin_std": 229.35256958007812, "step": 469 }, { "KL/chosen_KL_mean": -379.2088928222656, "KL/mean": -501.94903564453125, "KL/rejected_KL_mean": -624.6891479492188, "KL/std": 232.1168670654297, "epoch": 0.6901615271659325, "fcm_dpo/beta": 0.0035080176312476397, "fcm_dpo/delta": -0.06688985973596573, "fcm_dpo/margin": 245.4802703857422, "fcm_dpo/q_t": 0.31790733337402344, "grad_norm": 41.94991683959961, "learning_rate": 1.339940635976592e-07, "logits/chosen": 0.24481558799743652, "logits/rejected": 0.26332205533981323, "logps/chosen": -423.0008239746094, "logps/ref_chosen": -43.791927337646484, "logps/ref_rejected": -82.70285034179688, "logps/rejected": -707.3919677734375, "loss": 0.8406, "margin_dpo/margin_mean": 245.48028564453125, "margin_dpo/margin_std": 234.01971435546875, "step": 470 }, { "KL/chosen_KL_mean": -424.17352294921875, "KL/mean": -526.1630859375, "KL/rejected_KL_mean": -628.1526489257812, "KL/std": 224.7677764892578, "epoch": 0.6916299559471366, "fcm_dpo/beta": 0.0035170000046491623, "fcm_dpo/delta": 0.08863566070795059, "fcm_dpo/margin": 203.97914123535156, "fcm_dpo/q_t": 0.3468366861343384, "grad_norm": 40.956668853759766, "learning_rate": 1.3285879380446563e-07, "logits/chosen": 0.22680866718292236, "logits/rejected": 0.24914070963859558, "logps/chosen": -487.5130310058594, "logps/ref_chosen": -63.33952331542969, "logps/ref_rejected": -83.61048126220703, "logps/rejected": -711.7631225585938, "loss": 0.9042, "margin_dpo/margin_mean": 203.9791259765625, "margin_dpo/margin_std": 201.23226928710938, "step": 471 }, { "KL/chosen_KL_mean": -403.254150390625, "KL/mean": -534.051513671875, "KL/rejected_KL_mean": -664.848876953125, "KL/std": 286.0874328613281, "epoch": 0.6930983847283406, "fcm_dpo/beta": 0.0034922936465591192, "fcm_dpo/delta": -0.1249755322933197, "fcm_dpo/margin": 261.59478759765625, "fcm_dpo/q_t": 0.32117411494255066, "grad_norm": 30.76375389099121, "learning_rate": 1.317266107909975e-07, "logits/chosen": 0.0545220784842968, "logits/rejected": 0.09548470377922058, "logps/chosen": -486.92022705078125, "logps/ref_chosen": -83.66610717773438, "logps/ref_rejected": -117.20919799804688, "logps/rejected": -782.05810546875, "loss": 0.8339, "margin_dpo/margin_mean": 261.59478759765625, "margin_dpo/margin_std": 272.6410827636719, "step": 472 }, { "KL/chosen_KL_mean": -475.15594482421875, "KL/mean": -565.5804443359375, "KL/rejected_KL_mean": -656.0048828125, "KL/std": 236.00418090820312, "epoch": 0.6945668135095447, "fcm_dpo/beta": 0.0035597169771790504, "fcm_dpo/delta": 0.16679316759109497, "fcm_dpo/margin": 180.84896850585938, "fcm_dpo/q_t": 0.3708559274673462, "grad_norm": 72.96520233154297, "learning_rate": 1.3059754439133002e-07, "logits/chosen": 0.2742480933666229, "logits/rejected": 0.3015744984149933, "logps/chosen": -538.6528930664062, "logps/ref_chosen": -63.49696731567383, "logps/ref_rejected": -81.14657592773438, "logps/rejected": -737.1514892578125, "loss": 1.0307, "margin_dpo/margin_mean": 180.84896850585938, "margin_dpo/margin_std": 274.80029296875, "step": 473 }, { "KL/chosen_KL_mean": -417.2640075683594, "KL/mean": -509.0142822265625, "KL/rejected_KL_mean": -600.7645874023438, "KL/std": 221.151611328125, "epoch": 0.6960352422907489, "fcm_dpo/beta": 0.003691623918712139, "fcm_dpo/delta": 0.12826226651668549, "fcm_dpo/margin": 183.5005645751953, "fcm_dpo/q_t": 0.35935598611831665, "grad_norm": 61.98820114135742, "learning_rate": 1.2947162435741277e-07, "logits/chosen": 0.18188732862472534, "logits/rejected": 0.20040690898895264, "logps/chosen": -469.8759460449219, "logps/ref_chosen": -52.6119384765625, "logps/ref_rejected": -90.08041381835938, "logps/rejected": -690.844970703125, "loss": 0.9928, "margin_dpo/margin_mean": 183.50054931640625, "margin_dpo/margin_std": 239.28359985351562, "step": 474 }, { "KL/chosen_KL_mean": -372.036865234375, "KL/mean": -496.2984619140625, "KL/rejected_KL_mean": -620.56005859375, "KL/std": 217.1239013671875, "epoch": 0.697503671071953, "fcm_dpo/beta": 0.003616760950535536, "fcm_dpo/delta": -0.10878843069076538, "fcm_dpo/margin": 248.52316284179688, "fcm_dpo/q_t": 0.313048779964447, "grad_norm": 60.8675651550293, "learning_rate": 1.2834888035828596e-07, "logits/chosen": 0.27965113520622253, "logits/rejected": 0.28925400972366333, "logps/chosen": -414.5320739746094, "logps/ref_chosen": -42.49519348144531, "logps/ref_rejected": -90.06294250488281, "logps/rejected": -710.623046875, "loss": 0.8097, "margin_dpo/margin_mean": 248.52317810058594, "margin_dpo/margin_std": 225.99652099609375, "step": 475 }, { "KL/chosen_KL_mean": -366.5478515625, "KL/mean": -471.92510986328125, "KL/rejected_KL_mean": -577.3023681640625, "KL/std": 207.46705627441406, "epoch": 0.6989720998531571, "fcm_dpo/beta": 0.003618250833824277, "fcm_dpo/delta": 0.040546517819166183, "fcm_dpo/margin": 210.7545166015625, "fcm_dpo/q_t": 0.33902594447135925, "grad_norm": 44.247074127197266, "learning_rate": 1.2722934197929802e-07, "logits/chosen": 0.23911496996879578, "logits/rejected": 0.2586338520050049, "logps/chosen": -409.49725341796875, "logps/ref_chosen": -42.94938278198242, "logps/ref_rejected": -73.71023559570312, "logps/rejected": -651.0125732421875, "loss": 0.878, "margin_dpo/margin_mean": 210.7545166015625, "margin_dpo/margin_std": 205.98703002929688, "step": 476 }, { "KL/chosen_KL_mean": -377.1195983886719, "KL/mean": -477.4259948730469, "KL/rejected_KL_mean": -577.7323608398438, "KL/std": 214.20526123046875, "epoch": 0.7004405286343612, "fcm_dpo/beta": 0.0036724847741425037, "fcm_dpo/delta": 0.06813560426235199, "fcm_dpo/margin": 200.6127471923828, "fcm_dpo/q_t": 0.34649232029914856, "grad_norm": 49.856998443603516, "learning_rate": 1.2611303872132631e-07, "logits/chosen": 0.13930974900722504, "logits/rejected": 0.1919388622045517, "logps/chosen": -447.8922119140625, "logps/ref_chosen": -70.77261352539062, "logps/ref_rejected": -76.13737487792969, "logps/rejected": -653.8697509765625, "loss": 0.9433, "margin_dpo/margin_mean": 200.61273193359375, "margin_dpo/margin_std": 242.2550048828125, "step": 477 }, { "KL/chosen_KL_mean": -317.3870544433594, "KL/mean": -434.1727294921875, "KL/rejected_KL_mean": -550.9583740234375, "KL/std": 205.1141357421875, "epoch": 0.7019089574155654, "fcm_dpo/beta": 0.0036650304682552814, "fcm_dpo/delta": -0.06150561571121216, "fcm_dpo/margin": 233.57131958007812, "fcm_dpo/q_t": 0.3226568102836609, "grad_norm": 36.77776336669922, "learning_rate": 1.2500000000000005e-07, "logits/chosen": 0.15008513629436493, "logits/rejected": 0.16725000739097595, "logps/chosen": -358.82757568359375, "logps/ref_chosen": -41.440513610839844, "logps/ref_rejected": -85.36196899414062, "logps/rejected": -636.3203125, "loss": 0.824, "margin_dpo/margin_mean": 233.57131958007812, "margin_dpo/margin_std": 212.3719024658203, "step": 478 }, { "KL/chosen_KL_mean": -379.9550476074219, "KL/mean": -474.99371337890625, "KL/rejected_KL_mean": -570.0324096679688, "KL/std": 183.4116668701172, "epoch": 0.7033773861967695, "fcm_dpo/beta": 0.003735472448170185, "fcm_dpo/delta": 0.09173765778541565, "fcm_dpo/margin": 190.0773468017578, "fcm_dpo/q_t": 0.34823429584503174, "grad_norm": 43.84068298339844, "learning_rate": 1.2389025514492456e-07, "logits/chosen": 0.1662050038576126, "logits/rejected": 0.1779286414384842, "logps/chosen": -433.86297607421875, "logps/ref_chosen": -53.907920837402344, "logps/ref_rejected": -95.1163330078125, "logps/rejected": -665.1487426757812, "loss": 0.9327, "margin_dpo/margin_mean": 190.07733154296875, "margin_dpo/margin_std": 199.5963897705078, "step": 479 }, { "KL/chosen_KL_mean": -391.65753173828125, "KL/mean": -476.282470703125, "KL/rejected_KL_mean": -560.9075317382812, "KL/std": 195.97769165039062, "epoch": 0.7048458149779736, "fcm_dpo/beta": 0.0038283944595605135, "fcm_dpo/delta": 0.15897266566753387, "fcm_dpo/margin": 169.2499542236328, "fcm_dpo/q_t": 0.36624574661254883, "grad_norm": 58.622314453125, "learning_rate": 1.227838333989088e-07, "logits/chosen": 0.20244859158992767, "logits/rejected": 0.23983803391456604, "logps/chosen": -450.3402099609375, "logps/ref_chosen": -58.682701110839844, "logps/ref_rejected": -82.93248748779297, "logps/rejected": -643.8399658203125, "loss": 0.9996, "margin_dpo/margin_mean": 169.2499542236328, "margin_dpo/margin_std": 217.9320068359375, "step": 480 }, { "KL/chosen_KL_mean": -344.9322509765625, "KL/mean": -453.0932312011719, "KL/rejected_KL_mean": -561.254150390625, "KL/std": 219.3949737548828, "epoch": 0.7063142437591777, "fcm_dpo/beta": 0.003801655489951372, "fcm_dpo/delta": -0.024488603696227074, "fcm_dpo/margin": 216.32192993164062, "fcm_dpo/q_t": 0.3276780843734741, "grad_norm": 44.4178352355957, "learning_rate": 1.2168076391719489e-07, "logits/chosen": 0.1561613380908966, "logits/rejected": 0.18021619319915771, "logps/chosen": -399.89654541015625, "logps/ref_chosen": -54.964271545410156, "logps/ref_rejected": -92.42044067382812, "logps/rejected": -653.6746215820312, "loss": 0.8579, "margin_dpo/margin_mean": 216.32192993164062, "margin_dpo/margin_std": 211.07113647460938, "step": 481 }, { "KL/chosen_KL_mean": -358.85565185546875, "KL/mean": -441.71246337890625, "KL/rejected_KL_mean": -524.5692749023438, "KL/std": 197.96920776367188, "epoch": 0.7077826725403817, "fcm_dpo/beta": 0.0038895574398338795, "fcm_dpo/delta": 0.16643041372299194, "fcm_dpo/margin": 165.71365356445312, "fcm_dpo/q_t": 0.3640594482421875, "grad_norm": 45.240779876708984, "learning_rate": 1.2058107576668938e-07, "logits/chosen": 0.04070214927196503, "logits/rejected": 0.056692514568567276, "logps/chosen": -426.40911865234375, "logps/ref_chosen": -67.553466796875, "logps/ref_rejected": -87.58953857421875, "logps/rejected": -612.1588134765625, "loss": 0.9918, "margin_dpo/margin_mean": 165.71365356445312, "margin_dpo/margin_std": 216.013916015625, "step": 482 }, { "KL/chosen_KL_mean": -339.46856689453125, "KL/mean": -464.3380126953125, "KL/rejected_KL_mean": -589.2074584960938, "KL/std": 215.98431396484375, "epoch": 0.7092511013215859, "fcm_dpo/beta": 0.003823434002697468, "fcm_dpo/delta": -0.17181670665740967, "fcm_dpo/margin": 249.7388916015625, "fcm_dpo/q_t": 0.3055006265640259, "grad_norm": 33.2778434753418, "learning_rate": 1.194847979251979e-07, "logits/chosen": 0.10603588819503784, "logits/rejected": 0.14284950494766235, "logps/chosen": -402.79840087890625, "logps/ref_chosen": -63.32981872558594, "logps/ref_rejected": -95.78697204589844, "logps/rejected": -684.994384765625, "loss": 0.7874, "margin_dpo/margin_mean": 249.7388916015625, "margin_dpo/margin_std": 225.68191528320312, "step": 483 }, { "KL/chosen_KL_mean": -308.18743896484375, "KL/mean": -417.3006591796875, "KL/rejected_KL_mean": -526.4138793945312, "KL/std": 197.73350524902344, "epoch": 0.71071953010279, "fcm_dpo/beta": 0.0038092981558293104, "fcm_dpo/delta": -0.035128235816955566, "fcm_dpo/margin": 218.22645568847656, "fcm_dpo/q_t": 0.3240478038787842, "grad_norm": 34.91592788696289, "learning_rate": 1.1839195928066101e-07, "logits/chosen": 0.17551764845848083, "logits/rejected": 0.20704582333564758, "logps/chosen": -367.3255615234375, "logps/ref_chosen": -59.13812255859375, "logps/ref_rejected": -84.37144470214844, "logps/rejected": -610.7853393554688, "loss": 0.8259, "margin_dpo/margin_mean": 218.2264404296875, "margin_dpo/margin_std": 186.5542449951172, "step": 484 }, { "KL/chosen_KL_mean": -285.511474609375, "KL/mean": -398.1817626953125, "KL/rejected_KL_mean": -510.85205078125, "KL/std": 202.4456024169922, "epoch": 0.7121879588839941, "fcm_dpo/beta": 0.003755971323698759, "fcm_dpo/delta": -0.05070215463638306, "fcm_dpo/margin": 225.34059143066406, "fcm_dpo/q_t": 0.32257890701293945, "grad_norm": 31.80185317993164, "learning_rate": 1.1730258863039347e-07, "logits/chosen": 0.06667602062225342, "logits/rejected": 0.08204781264066696, "logps/chosen": -344.36102294921875, "logps/ref_chosen": -58.849571228027344, "logps/ref_rejected": -103.36408233642578, "logps/rejected": -614.2161865234375, "loss": 0.842, "margin_dpo/margin_mean": 225.34059143066406, "margin_dpo/margin_std": 212.36538696289062, "step": 485 }, { "KL/chosen_KL_mean": -327.9944763183594, "KL/mean": -442.83380126953125, "KL/rejected_KL_mean": -557.673095703125, "KL/std": 211.04153442382812, "epoch": 0.7136563876651982, "fcm_dpo/beta": 0.003716058097779751, "fcm_dpo/delta": -0.05855964869260788, "fcm_dpo/margin": 229.6786651611328, "fcm_dpo/q_t": 0.3270586133003235, "grad_norm": 57.901912689208984, "learning_rate": 1.1621671468032493e-07, "logits/chosen": 0.14983290433883667, "logits/rejected": 0.17763462662696838, "logps/chosen": -383.254150390625, "logps/ref_chosen": -55.25966262817383, "logps/ref_rejected": -92.13936614990234, "logps/rejected": -649.8125, "loss": 0.8833, "margin_dpo/margin_mean": 229.6786651611328, "margin_dpo/margin_std": 250.62548828125, "step": 486 }, { "KL/chosen_KL_mean": -321.1956481933594, "KL/mean": -420.681640625, "KL/rejected_KL_mean": -520.1676025390625, "KL/std": 171.1717529296875, "epoch": 0.7151248164464024, "fcm_dpo/beta": 0.003711356781423092, "fcm_dpo/delta": 0.0663451999425888, "fcm_dpo/margin": 198.97198486328125, "fcm_dpo/q_t": 0.3368815779685974, "grad_norm": 44.26673126220703, "learning_rate": 1.1513436604424378e-07, "logits/chosen": 0.10643830895423889, "logits/rejected": 0.1315649002790451, "logps/chosen": -374.25897216796875, "logps/ref_chosen": -53.06330871582031, "logps/ref_rejected": -92.41883087158203, "logps/rejected": -612.58642578125, "loss": 0.8648, "margin_dpo/margin_mean": 198.97198486328125, "margin_dpo/margin_std": 166.13626098632812, "step": 487 }, { "KL/chosen_KL_mean": -299.201416015625, "KL/mean": -398.70916748046875, "KL/rejected_KL_mean": -498.2169189453125, "KL/std": 195.89431762695312, "epoch": 0.7165932452276065, "fcm_dpo/beta": 0.003781733103096485, "fcm_dpo/delta": 0.050800926983356476, "fcm_dpo/margin": 199.01553344726562, "fcm_dpo/q_t": 0.3352172374725342, "grad_norm": 42.39677047729492, "learning_rate": 1.1405557124304335e-07, "logits/chosen": 0.13744409382343292, "logits/rejected": 0.16995671391487122, "logps/chosen": -351.4295654296875, "logps/ref_chosen": -52.22815704345703, "logps/ref_rejected": -84.00656127929688, "logps/rejected": -582.2235107421875, "loss": 0.8553, "margin_dpo/margin_mean": 199.01553344726562, "margin_dpo/margin_std": 166.0457305908203, "step": 488 }, { "KL/chosen_KL_mean": -295.97509765625, "KL/mean": -381.76287841796875, "KL/rejected_KL_mean": -467.5506896972656, "KL/std": 193.10520935058594, "epoch": 0.7180616740088106, "fcm_dpo/beta": 0.0038628973998129368, "fcm_dpo/delta": 0.1469191014766693, "fcm_dpo/margin": 171.57559204101562, "fcm_dpo/q_t": 0.36090078949928284, "grad_norm": 44.343040466308594, "learning_rate": 1.1298035870396985e-07, "logits/chosen": 0.03998090326786041, "logits/rejected": 0.06467624753713608, "logps/chosen": -351.9647216796875, "logps/ref_chosen": -55.989627838134766, "logps/ref_rejected": -79.39812469482422, "logps/rejected": -546.9488525390625, "loss": 0.9379, "margin_dpo/margin_mean": 171.5756072998047, "margin_dpo/margin_std": 188.32510375976562, "step": 489 }, { "KL/chosen_KL_mean": -306.4906005859375, "KL/mean": -406.5208435058594, "KL/rejected_KL_mean": -506.55108642578125, "KL/std": 221.122802734375, "epoch": 0.7195301027900147, "fcm_dpo/beta": 0.003920567687600851, "fcm_dpo/delta": 0.01635659858584404, "fcm_dpo/margin": 200.06051635742188, "fcm_dpo/q_t": 0.3403034508228302, "grad_norm": 34.440757751464844, "learning_rate": 1.1190875675987355e-07, "logits/chosen": 0.07778964936733246, "logits/rejected": 0.08467268198728561, "logps/chosen": -358.85699462890625, "logps/ref_chosen": -52.36639404296875, "logps/ref_rejected": -110.4090576171875, "logps/rejected": -616.9601440429688, "loss": 0.9101, "margin_dpo/margin_mean": 200.06051635742188, "margin_dpo/margin_std": 225.04501342773438, "step": 490 }, { "KL/chosen_KL_mean": -340.3194580078125, "KL/mean": -423.6872253417969, "KL/rejected_KL_mean": -507.05499267578125, "KL/std": 196.24343872070312, "epoch": 0.7209985315712188, "fcm_dpo/beta": 0.003996577113866806, "fcm_dpo/delta": 0.14273640513420105, "fcm_dpo/margin": 166.73556518554688, "fcm_dpo/q_t": 0.36313116550445557, "grad_norm": 44.4641227722168, "learning_rate": 1.1084079364846241e-07, "logits/chosen": 0.1599426120519638, "logits/rejected": 0.18364211916923523, "logps/chosen": -400.4356994628906, "logps/ref_chosen": -60.11626434326172, "logps/ref_rejected": -73.27278900146484, "logps/rejected": -580.3277587890625, "loss": 0.9682, "margin_dpo/margin_mean": 166.73556518554688, "margin_dpo/margin_std": 207.41067504882812, "step": 491 }, { "KL/chosen_KL_mean": -371.9100341796875, "KL/mean": -440.6224060058594, "KL/rejected_KL_mean": -509.33477783203125, "KL/std": 191.6051025390625, "epoch": 0.7224669603524229, "fcm_dpo/beta": 0.0041659013368189335, "fcm_dpo/delta": 0.24118927121162415, "fcm_dpo/margin": 137.42471313476562, "fcm_dpo/q_t": 0.38607728481292725, "grad_norm": 58.81159591674805, "learning_rate": 1.097764975115576e-07, "logits/chosen": 0.24060788750648499, "logits/rejected": 0.2635948956012726, "logps/chosen": -425.90423583984375, "logps/ref_chosen": -53.994178771972656, "logps/ref_rejected": -72.65962219238281, "logps/rejected": -581.994384765625, "loss": 1.1029, "margin_dpo/margin_mean": 137.42471313476562, "margin_dpo/margin_std": 237.78509521484375, "step": 492 }, { "KL/chosen_KL_mean": -365.27728271484375, "KL/mean": -450.06695556640625, "KL/rejected_KL_mean": -534.8565673828125, "KL/std": 205.73464965820312, "epoch": 0.723935389133627, "fcm_dpo/beta": 0.0042484961450099945, "fcm_dpo/delta": 0.08563549071550369, "fcm_dpo/margin": 169.57925415039062, "fcm_dpo/q_t": 0.3462260067462921, "grad_norm": 44.68215560913086, "learning_rate": 1.0871589639435203e-07, "logits/chosen": 0.22852273285388947, "logits/rejected": 0.2710832357406616, "logps/chosen": -440.7745361328125, "logps/ref_chosen": -75.49723815917969, "logps/ref_rejected": -87.32301330566406, "logps/rejected": -622.1795654296875, "loss": 0.9105, "margin_dpo/margin_mean": 169.5792236328125, "margin_dpo/margin_std": 173.59259033203125, "step": 493 }, { "KL/chosen_KL_mean": -329.7471008300781, "KL/mean": -448.144775390625, "KL/rejected_KL_mean": -566.5425415039062, "KL/std": 210.83590698242188, "epoch": 0.7254038179148311, "fcm_dpo/beta": 0.004156805574893951, "fcm_dpo/delta": -0.20466670393943787, "fcm_dpo/margin": 236.79537963867188, "fcm_dpo/q_t": 0.2935987114906311, "grad_norm": 42.584938049316406, "learning_rate": 1.0765901824467166e-07, "logits/chosen": 0.22543635964393616, "logits/rejected": 0.2317037582397461, "logps/chosen": -371.10638427734375, "logps/ref_chosen": -41.35926818847656, "logps/ref_rejected": -86.09136962890625, "logps/rejected": -652.6339111328125, "loss": 0.7692, "margin_dpo/margin_mean": 236.79537963867188, "margin_dpo/margin_std": 198.07933044433594, "step": 494 }, { "KL/chosen_KL_mean": -370.0741882324219, "KL/mean": -475.6034851074219, "KL/rejected_KL_mean": -581.1327514648438, "KL/std": 207.8790283203125, "epoch": 0.7268722466960352, "fcm_dpo/beta": 0.004097241908311844, "fcm_dpo/delta": -0.07118328660726547, "fcm_dpo/margin": 211.05856323242188, "fcm_dpo/q_t": 0.3244919180870056, "grad_norm": 42.264686584472656, "learning_rate": 1.0660589091223854e-07, "logits/chosen": 0.1444653421640396, "logits/rejected": 0.16225489974021912, "logps/chosen": -433.6092529296875, "logps/ref_chosen": -63.53507995605469, "logps/ref_rejected": -91.42443084716797, "logps/rejected": -672.5571899414062, "loss": 0.8478, "margin_dpo/margin_mean": 211.05856323242188, "margin_dpo/margin_std": 213.89817810058594, "step": 495 }, { "KL/chosen_KL_mean": -429.86212158203125, "KL/mean": -512.4937744140625, "KL/rejected_KL_mean": -595.12548828125, "KL/std": 178.7728729248047, "epoch": 0.7283406754772394, "fcm_dpo/beta": 0.004128463100641966, "fcm_dpo/delta": 0.1264854371547699, "fcm_dpo/margin": 165.26336669921875, "fcm_dpo/q_t": 0.3561249375343323, "grad_norm": 46.71969223022461, "learning_rate": 1.0555654214793722e-07, "logits/chosen": 0.20871032774448395, "logits/rejected": 0.24310749769210815, "logps/chosen": -502.45404052734375, "logps/ref_chosen": -72.5919189453125, "logps/ref_rejected": -84.32933807373047, "logps/rejected": -679.454833984375, "loss": 0.9337, "margin_dpo/margin_mean": 165.26336669921875, "margin_dpo/margin_std": 178.31752014160156, "step": 496 }, { "KL/chosen_KL_mean": -427.16143798828125, "KL/mean": -507.173828125, "KL/rejected_KL_mean": -587.1861572265625, "KL/std": 211.1545867919922, "epoch": 0.7298091042584435, "fcm_dpo/beta": 0.004193710163235664, "fcm_dpo/delta": 0.13649700582027435, "fcm_dpo/margin": 160.02471923828125, "fcm_dpo/q_t": 0.3586384654045105, "grad_norm": 39.61761474609375, "learning_rate": 1.0451099960308374e-07, "logits/chosen": 0.22612471878528595, "logits/rejected": 0.2511892318725586, "logps/chosen": -485.75543212890625, "logps/ref_chosen": -58.59397506713867, "logps/ref_rejected": -76.28836822509766, "logps/rejected": -663.4745483398438, "loss": 0.9579, "margin_dpo/margin_mean": 160.02471923828125, "margin_dpo/margin_std": 175.66616821289062, "step": 497 }, { "KL/chosen_KL_mean": -437.52252197265625, "KL/mean": -525.6541137695312, "KL/rejected_KL_mean": -613.7857055664062, "KL/std": 198.28433227539062, "epoch": 0.7312775330396476, "fcm_dpo/beta": 0.00430191308259964, "fcm_dpo/delta": 0.04517921805381775, "fcm_dpo/margin": 176.26318359375, "fcm_dpo/q_t": 0.3466299772262573, "grad_norm": 46.89978790283203, "learning_rate": 1.0346929082869641e-07, "logits/chosen": 0.22263531386852264, "logits/rejected": 0.24923723936080933, "logps/chosen": -508.7281494140625, "logps/ref_chosen": -71.20565795898438, "logps/ref_rejected": -83.95803833007812, "logps/rejected": -697.7437744140625, "loss": 0.9419, "margin_dpo/margin_mean": 176.26318359375, "margin_dpo/margin_std": 214.70091247558594, "step": 498 }, { "KL/chosen_KL_mean": -423.108642578125, "KL/mean": -536.3960571289062, "KL/rejected_KL_mean": -649.6835327148438, "KL/std": 213.3042449951172, "epoch": 0.7327459618208517, "fcm_dpo/beta": 0.004202335141599178, "fcm_dpo/delta": -0.16854572296142578, "fcm_dpo/margin": 226.5749053955078, "fcm_dpo/q_t": 0.3053300976753235, "grad_norm": 44.01908493041992, "learning_rate": 1.0243144327477013e-07, "logits/chosen": 0.3921792805194855, "logits/rejected": 0.40327873826026917, "logps/chosen": -474.36383056640625, "logps/ref_chosen": -51.25519561767578, "logps/ref_rejected": -101.07870483398438, "logps/rejected": -750.76220703125, "loss": 0.8274, "margin_dpo/margin_mean": 226.5749053955078, "margin_dpo/margin_std": 223.81814575195312, "step": 499 }, { "KL/chosen_KL_mean": -462.30181884765625, "KL/mean": -562.2794189453125, "KL/rejected_KL_mean": -662.257080078125, "KL/std": 209.902587890625, "epoch": 0.7342143906020558, "fcm_dpo/beta": 0.004142479971051216, "fcm_dpo/delta": -0.031111618503928185, "fcm_dpo/margin": 199.95518493652344, "fcm_dpo/q_t": 0.3318169414997101, "grad_norm": 54.924827575683594, "learning_rate": 1.0139748428955333e-07, "logits/chosen": 0.2692311108112335, "logits/rejected": 0.2694999873638153, "logps/chosen": -519.3292846679688, "logps/ref_chosen": -57.027442932128906, "logps/ref_rejected": -93.93421173095703, "logps/rejected": -756.1912841796875, "loss": 0.8904, "margin_dpo/margin_mean": 199.95518493652344, "margin_dpo/margin_std": 222.5806884765625, "step": 500 }, { "KL/chosen_KL_mean": -428.00701904296875, "KL/mean": -532.467529296875, "KL/rejected_KL_mean": -636.9281005859375, "KL/std": 207.19921875, "epoch": 0.73568281938326, "fcm_dpo/beta": 0.0041295429691672325, "fcm_dpo/delta": -0.0689932331442833, "fcm_dpo/margin": 208.921142578125, "fcm_dpo/q_t": 0.3278210759162903, "grad_norm": 55.09239959716797, "learning_rate": 1.0036744111882672e-07, "logits/chosen": 0.2993459403514862, "logits/rejected": 0.3334452509880066, "logps/chosen": -482.36651611328125, "logps/ref_chosen": -54.359527587890625, "logps/ref_rejected": -80.15670013427734, "logps/rejected": -717.0848388671875, "loss": 0.8729, "margin_dpo/margin_mean": 208.921142578125, "margin_dpo/margin_std": 227.49624633789062, "step": 501 }, { "KL/chosen_KL_mean": -392.55389404296875, "KL/mean": -508.1488342285156, "KL/rejected_KL_mean": -623.7437744140625, "KL/std": 193.1407470703125, "epoch": 0.737151248164464, "fcm_dpo/beta": 0.00403004651889205, "fcm_dpo/delta": -0.14517813920974731, "fcm_dpo/margin": 231.1898956298828, "fcm_dpo/q_t": 0.29726549983024597, "grad_norm": 32.64988327026367, "learning_rate": 9.934134090518592e-08, "logits/chosen": 0.20329806208610535, "logits/rejected": 0.23956874012947083, "logps/chosen": -460.1544189453125, "logps/ref_chosen": -67.60050964355469, "logps/ref_rejected": -82.94876098632812, "logps/rejected": -706.6925048828125, "loss": 0.7467, "margin_dpo/margin_mean": 231.1898956298828, "margin_dpo/margin_std": 155.42315673828125, "step": 502 }, { "KL/chosen_KL_mean": -414.6917419433594, "KL/mean": -511.0799560546875, "KL/rejected_KL_mean": -607.4681396484375, "KL/std": 182.72433471679688, "epoch": 0.7386196769456681, "fcm_dpo/beta": 0.0039795683696866035, "fcm_dpo/delta": 0.03540978953242302, "fcm_dpo/margin": 192.7764129638672, "fcm_dpo/q_t": 0.335806667804718, "grad_norm": 36.449703216552734, "learning_rate": 9.831921068732571e-08, "logits/chosen": 0.30815067887306213, "logits/rejected": 0.33968621492385864, "logps/chosen": -469.7701416015625, "logps/ref_chosen": -55.078407287597656, "logps/ref_rejected": -82.50544738769531, "logps/rejected": -689.9736328125, "loss": 0.8623, "margin_dpo/margin_mean": 192.7764129638672, "margin_dpo/margin_std": 176.17752075195312, "step": 503 }, { "KL/chosen_KL_mean": -423.18780517578125, "KL/mean": -543.5272216796875, "KL/rejected_KL_mean": -663.86669921875, "KL/std": 212.20285034179688, "epoch": 0.7400881057268722, "fcm_dpo/beta": 0.00390539038926363, "fcm_dpo/delta": -0.154744952917099, "fcm_dpo/margin": 240.67892456054688, "fcm_dpo/q_t": 0.3052040934562683, "grad_norm": 37.20085906982422, "learning_rate": 9.730107739932805e-08, "logits/chosen": 0.29225870966911316, "logits/rejected": 0.29667824506759644, "logps/chosen": -483.1535339355469, "logps/ref_chosen": -59.96575164794922, "logps/ref_rejected": -103.76212310791016, "logps/rejected": -767.6288452148438, "loss": 0.7907, "margin_dpo/margin_mean": 240.67892456054688, "margin_dpo/margin_std": 207.40602111816406, "step": 504 }, { "KL/chosen_KL_mean": -443.1572570800781, "KL/mean": -517.259521484375, "KL/rejected_KL_mean": -591.3617553710938, "KL/std": 220.9525146484375, "epoch": 0.7415565345080763, "fcm_dpo/beta": 0.00401611253619194, "fcm_dpo/delta": 0.2171427607536316, "fcm_dpo/margin": 148.2044677734375, "fcm_dpo/q_t": 0.3818475008010864, "grad_norm": 64.7453384399414, "learning_rate": 9.628696786995188e-08, "logits/chosen": 0.24049341678619385, "logits/rejected": 0.2687991261482239, "logps/chosen": -519.3121337890625, "logps/ref_chosen": -76.1549072265625, "logps/ref_rejected": -88.58537292480469, "logps/rejected": -679.9471435546875, "loss": 1.0594, "margin_dpo/margin_mean": 148.20448303222656, "margin_dpo/margin_std": 231.49600219726562, "step": 505 }, { "KL/chosen_KL_mean": -386.2157287597656, "KL/mean": -494.36199951171875, "KL/rejected_KL_mean": -602.50830078125, "KL/std": 219.51278686523438, "epoch": 0.7430249632892805, "fcm_dpo/beta": 0.0039803218096494675, "fcm_dpo/delta": -0.06841699779033661, "fcm_dpo/margin": 216.2925567626953, "fcm_dpo/q_t": 0.3228600323200226, "grad_norm": 34.961204528808594, "learning_rate": 9.527690882192635e-08, "logits/chosen": 0.281488299369812, "logits/rejected": 0.30117908120155334, "logps/chosen": -435.17620849609375, "logps/ref_chosen": -48.96050262451172, "logps/ref_rejected": -78.41505432128906, "logps/rejected": -680.92333984375, "loss": 0.8502, "margin_dpo/margin_mean": 216.2925567626953, "margin_dpo/margin_std": 212.88848876953125, "step": 506 }, { "KL/chosen_KL_mean": -417.7148132324219, "KL/mean": -510.52874755859375, "KL/rejected_KL_mean": -603.3427124023438, "KL/std": 215.5636749267578, "epoch": 0.7444933920704846, "fcm_dpo/beta": 0.004029689822345972, "fcm_dpo/delta": 0.056185171008110046, "fcm_dpo/margin": 185.62789916992188, "fcm_dpo/q_t": 0.3467479944229126, "grad_norm": 33.984588623046875, "learning_rate": 9.427092687124691e-08, "logits/chosen": 0.32121342420578003, "logits/rejected": 0.34334227442741394, "logps/chosen": -484.51629638671875, "logps/ref_chosen": -66.80149841308594, "logps/ref_rejected": -95.37289428710938, "logps/rejected": -698.715576171875, "loss": 0.9057, "margin_dpo/margin_mean": 185.62791442871094, "margin_dpo/margin_std": 205.94578552246094, "step": 507 }, { "KL/chosen_KL_mean": -395.92645263671875, "KL/mean": -485.9989013671875, "KL/rejected_KL_mean": -576.0712890625, "KL/std": 222.60186767578125, "epoch": 0.7459618208516887, "fcm_dpo/beta": 0.0040952954441308975, "fcm_dpo/delta": 0.06677697598934174, "fcm_dpo/margin": 180.14486694335938, "fcm_dpo/q_t": 0.3528652787208557, "grad_norm": 45.727481842041016, "learning_rate": 9.326904852647344e-08, "logits/chosen": 0.2253168672323227, "logits/rejected": 0.23925551772117615, "logps/chosen": -467.22991943359375, "logps/ref_chosen": -71.303466796875, "logps/ref_rejected": -95.6275405883789, "logps/rejected": -671.6988525390625, "loss": 0.9525, "margin_dpo/margin_mean": 180.14486694335938, "margin_dpo/margin_std": 225.4059295654297, "step": 508 }, { "KL/chosen_KL_mean": -346.7314453125, "KL/mean": -436.817626953125, "KL/rejected_KL_mean": -526.90380859375, "KL/std": 187.9528045654297, "epoch": 0.7474302496328928, "fcm_dpo/beta": 0.0041460758075118065, "fcm_dpo/delta": 0.05677645653486252, "fcm_dpo/margin": 180.17236328125, "fcm_dpo/q_t": 0.34532153606414795, "grad_norm": 53.75666427612305, "learning_rate": 9.227130018803195e-08, "logits/chosen": 0.09170372784137726, "logits/rejected": 0.10787695646286011, "logps/chosen": -410.5503845214844, "logps/ref_chosen": -63.81895065307617, "logps/ref_rejected": -83.25643920898438, "logps/rejected": -610.1602783203125, "loss": 0.9047, "margin_dpo/margin_mean": 180.17236328125, "margin_dpo/margin_std": 193.12432861328125, "step": 509 }, { "KL/chosen_KL_mean": -383.3047180175781, "KL/mean": -495.38372802734375, "KL/rejected_KL_mean": -607.4627685546875, "KL/std": 199.19956970214844, "epoch": 0.748898678414097, "fcm_dpo/beta": 0.004071595147252083, "fcm_dpo/delta": -0.12390496581792831, "fcm_dpo/margin": 224.15802001953125, "fcm_dpo/q_t": 0.3062693178653717, "grad_norm": 41.95302200317383, "learning_rate": 9.127770814751932e-08, "logits/chosen": 0.1895131915807724, "logits/rejected": 0.1943446695804596, "logps/chosen": -435.18316650390625, "logps/ref_chosen": -51.878448486328125, "logps/ref_rejected": -102.7651596069336, "logps/rejected": -710.2279052734375, "loss": 0.7903, "margin_dpo/margin_mean": 224.1580352783203, "margin_dpo/margin_std": 186.15345764160156, "step": 510 }, { "KL/chosen_KL_mean": -357.55572509765625, "KL/mean": -462.718017578125, "KL/rejected_KL_mean": -567.88037109375, "KL/std": 215.96746826171875, "epoch": 0.750367107195301, "fcm_dpo/beta": 0.004031065851449966, "fcm_dpo/delta": -0.052543915808200836, "fcm_dpo/margin": 210.32461547851562, "fcm_dpo/q_t": 0.3308233618736267, "grad_norm": 35.133419036865234, "learning_rate": 9.028829858700973e-08, "logits/chosen": 0.2895551919937134, "logits/rejected": 0.3053359091281891, "logps/chosen": -417.7938537597656, "logps/ref_chosen": -60.23811721801758, "logps/ref_rejected": -92.85676574707031, "logps/rejected": -660.7371215820312, "loss": 0.8985, "margin_dpo/margin_mean": 210.32460021972656, "margin_dpo/margin_std": 240.78411865234375, "step": 511 }, { "KL/chosen_KL_mean": -330.9661865234375, "KL/mean": -445.86968994140625, "KL/rejected_KL_mean": -560.773193359375, "KL/std": 198.74737548828125, "epoch": 0.7518355359765051, "fcm_dpo/beta": 0.003930510953068733, "fcm_dpo/delta": -0.11358829587697983, "fcm_dpo/margin": 229.80703735351562, "fcm_dpo/q_t": 0.3084445595741272, "grad_norm": 35.95514678955078, "learning_rate": 8.930309757836516e-08, "logits/chosen": 0.2970372438430786, "logits/rejected": 0.3260389566421509, "logps/chosen": -385.8717041015625, "logps/ref_chosen": -54.905494689941406, "logps/ref_rejected": -81.87586975097656, "logps/rejected": -642.6491088867188, "loss": 0.7866, "margin_dpo/margin_mean": 229.80703735351562, "margin_dpo/margin_std": 188.13687133789062, "step": 512 }, { "KL/chosen_KL_mean": -364.7303466796875, "KL/mean": -452.98175048828125, "KL/rejected_KL_mean": -541.2332153320312, "KL/std": 184.21725463867188, "epoch": 0.7533039647577092, "fcm_dpo/beta": 0.0039430903270840645, "fcm_dpo/delta": 0.1113673523068428, "fcm_dpo/margin": 176.50286865234375, "fcm_dpo/q_t": 0.3536303639411926, "grad_norm": 54.2635498046875, "learning_rate": 8.832213108254863e-08, "logits/chosen": 0.24457916617393494, "logits/rejected": 0.28037741780281067, "logps/chosen": -429.64678955078125, "logps/ref_chosen": -64.91644287109375, "logps/ref_rejected": -76.06245422363281, "logps/rejected": -617.295654296875, "loss": 0.9551, "margin_dpo/margin_mean": 176.5028533935547, "margin_dpo/margin_std": 205.36679077148438, "step": 513 }, { "KL/chosen_KL_mean": -346.7003479003906, "KL/mean": -434.6519775390625, "KL/rejected_KL_mean": -522.6036376953125, "KL/std": 185.1610870361328, "epoch": 0.7547723935389133, "fcm_dpo/beta": 0.0040502483025193214, "fcm_dpo/delta": 0.0943315327167511, "fcm_dpo/margin": 175.90335083007812, "fcm_dpo/q_t": 0.35776999592781067, "grad_norm": 48.51445770263672, "learning_rate": 8.734542494893954e-08, "logits/chosen": 0.15230363607406616, "logits/rejected": 0.1823866069316864, "logps/chosen": -420.929931640625, "logps/ref_chosen": -74.22957611083984, "logps/ref_rejected": -78.945556640625, "logps/rejected": -601.5491943359375, "loss": 0.9599, "margin_dpo/margin_mean": 175.90333557128906, "margin_dpo/margin_std": 223.0550079345703, "step": 514 }, { "KL/chosen_KL_mean": -307.29766845703125, "KL/mean": -386.40625, "KL/rejected_KL_mean": -465.5148010253906, "KL/std": 185.33645629882812, "epoch": 0.7562408223201175, "fcm_dpo/beta": 0.00415542908012867, "fcm_dpo/delta": 0.15283825993537903, "fcm_dpo/margin": 158.21713256835938, "fcm_dpo/q_t": 0.36064812541007996, "grad_norm": 34.504180908203125, "learning_rate": 8.637300491465272e-08, "logits/chosen": 0.12813155353069305, "logits/rejected": 0.13620558381080627, "logps/chosen": -357.6992492675781, "logps/ref_chosen": -50.40156555175781, "logps/ref_rejected": -87.09774780273438, "logps/rejected": -552.612548828125, "loss": 0.9722, "margin_dpo/margin_mean": 158.21713256835938, "margin_dpo/margin_std": 191.54144287109375, "step": 515 }, { "KL/chosen_KL_mean": -318.70489501953125, "KL/mean": -427.0130920410156, "KL/rejected_KL_mean": -535.3212890625, "KL/std": 196.55706787109375, "epoch": 0.7577092511013216, "fcm_dpo/beta": 0.004153442569077015, "fcm_dpo/delta": -0.10953111201524734, "fcm_dpo/margin": 216.61639404296875, "fcm_dpo/q_t": 0.3109952211380005, "grad_norm": 40.440147399902344, "learning_rate": 8.540489660386064e-08, "logits/chosen": 0.19949766993522644, "logits/rejected": 0.20106914639472961, "logps/chosen": -383.3544921875, "logps/ref_chosen": -64.64956665039062, "logps/ref_rejected": -111.72237396240234, "logps/rejected": -647.043701171875, "loss": 0.7982, "margin_dpo/margin_mean": 216.61639404296875, "margin_dpo/margin_std": 184.9787139892578, "step": 516 }, { "KL/chosen_KL_mean": -329.6126708984375, "KL/mean": -439.53631591796875, "KL/rejected_KL_mean": -549.4599609375, "KL/std": 210.18545532226562, "epoch": 0.7591776798825257, "fcm_dpo/beta": 0.0040242038667202, "fcm_dpo/delta": -0.09458979964256287, "fcm_dpo/margin": 219.8473358154297, "fcm_dpo/q_t": 0.31907662749290466, "grad_norm": 38.87144088745117, "learning_rate": 8.444112552711752e-08, "logits/chosen": 0.13860376179218292, "logits/rejected": 0.1707230806350708, "logps/chosen": -390.5262145996094, "logps/ref_chosen": -60.913551330566406, "logps/ref_rejected": -89.08308410644531, "logps/rejected": -638.5430908203125, "loss": 0.8339, "margin_dpo/margin_mean": 219.8473358154297, "margin_dpo/margin_std": 214.79222106933594, "step": 517 }, { "KL/chosen_KL_mean": -329.9923095703125, "KL/mean": -424.657958984375, "KL/rejected_KL_mean": -519.3236083984375, "KL/std": 184.806884765625, "epoch": 0.7606461086637298, "fcm_dpo/beta": 0.00404348922893405, "fcm_dpo/delta": 0.03724416345357895, "fcm_dpo/margin": 189.33128356933594, "fcm_dpo/q_t": 0.3417090177536011, "grad_norm": 34.10972595214844, "learning_rate": 8.348171708068747e-08, "logits/chosen": 0.1555401086807251, "logits/rejected": 0.1683722883462906, "logps/chosen": -387.44818115234375, "logps/ref_chosen": -57.45589065551758, "logps/ref_rejected": -85.31269836425781, "logps/rejected": -604.63623046875, "loss": 0.9226, "margin_dpo/margin_mean": 189.331298828125, "margin_dpo/margin_std": 217.88580322265625, "step": 518 }, { "KL/chosen_KL_mean": -317.1443786621094, "KL/mean": -401.89031982421875, "KL/rejected_KL_mean": -486.63629150390625, "KL/std": 192.45364379882812, "epoch": 0.762114537444934, "fcm_dpo/beta": 0.004131737165153027, "fcm_dpo/delta": 0.10707257688045502, "fcm_dpo/margin": 169.49188232421875, "fcm_dpo/q_t": 0.35672521591186523, "grad_norm": 38.87397384643555, "learning_rate": 8.25266965458755e-08, "logits/chosen": 0.10519933700561523, "logits/rejected": 0.12509751319885254, "logps/chosen": -391.20770263671875, "logps/ref_chosen": -74.06331634521484, "logps/ref_rejected": -104.44416809082031, "logps/rejected": -591.0804443359375, "loss": 0.954, "margin_dpo/margin_mean": 169.49188232421875, "margin_dpo/margin_std": 207.0226287841797, "step": 519 }, { "KL/chosen_KL_mean": -316.388916015625, "KL/mean": -407.5377502441406, "KL/rejected_KL_mean": -498.68658447265625, "KL/std": 183.60214233398438, "epoch": 0.7635829662261381, "fcm_dpo/beta": 0.0041408720426261425, "fcm_dpo/delta": 0.047438718378543854, "fcm_dpo/margin": 182.29766845703125, "fcm_dpo/q_t": 0.3425585627555847, "grad_norm": 35.9736213684082, "learning_rate": 8.15760890883607e-08, "logits/chosen": 0.13645502924919128, "logits/rejected": 0.1466117650270462, "logps/chosen": -386.688720703125, "logps/ref_chosen": -70.2998275756836, "logps/ref_rejected": -99.98133850097656, "logps/rejected": -598.66796875, "loss": 0.8873, "margin_dpo/margin_mean": 182.29766845703125, "margin_dpo/margin_std": 177.412841796875, "step": 520 }, { "KL/chosen_KL_mean": -308.92608642578125, "KL/mean": -396.453125, "KL/rejected_KL_mean": -483.9801330566406, "KL/std": 180.9889678955078, "epoch": 0.7650513950073421, "fcm_dpo/beta": 0.004272553138434887, "fcm_dpo/delta": 0.0528116375207901, "fcm_dpo/margin": 175.0540771484375, "fcm_dpo/q_t": 0.3469804525375366, "grad_norm": 47.978233337402344, "learning_rate": 8.062991975753378e-08, "logits/chosen": 0.2017364799976349, "logits/rejected": 0.22750027477741241, "logps/chosen": -367.0690002441406, "logps/ref_chosen": -58.14292526245117, "logps/ref_rejected": -83.28060913085938, "logps/rejected": -567.2607421875, "loss": 0.9271, "margin_dpo/margin_mean": 175.05406188964844, "margin_dpo/margin_std": 195.4500732421875, "step": 521 }, { "KL/chosen_KL_mean": -332.21148681640625, "KL/mean": -413.32086181640625, "KL/rejected_KL_mean": -494.43017578125, "KL/std": 178.46389770507812, "epoch": 0.7665198237885462, "fcm_dpo/beta": 0.004303772002458572, "fcm_dpo/delta": 0.10948702692985535, "fcm_dpo/margin": 162.21871948242188, "fcm_dpo/q_t": 0.35133394598960876, "grad_norm": 61.93113327026367, "learning_rate": 7.968821348583643e-08, "logits/chosen": 0.21870219707489014, "logits/rejected": 0.2350131869316101, "logps/chosen": -378.7591552734375, "logps/ref_chosen": -46.54766845703125, "logps/ref_rejected": -66.01388549804688, "logps/rejected": -560.444091796875, "loss": 0.9312, "margin_dpo/margin_mean": 162.21871948242188, "margin_dpo/margin_std": 176.44223022460938, "step": 522 }, { "KL/chosen_KL_mean": -325.970703125, "KL/mean": -411.48406982421875, "KL/rejected_KL_mean": -496.9974670410156, "KL/std": 194.84500122070312, "epoch": 0.7679882525697503, "fcm_dpo/beta": 0.004355207085609436, "fcm_dpo/delta": 0.05959649384021759, "fcm_dpo/margin": 171.02676391601562, "fcm_dpo/q_t": 0.34175705909729004, "grad_norm": 45.9669189453125, "learning_rate": 7.875099508810484e-08, "logits/chosen": 0.1334128975868225, "logits/rejected": 0.1688709706068039, "logps/chosen": -387.74029541015625, "logps/ref_chosen": -61.76960372924805, "logps/ref_rejected": -83.76141357421875, "logps/rejected": -580.7589111328125, "loss": 0.9417, "margin_dpo/margin_mean": 171.02676391601562, "margin_dpo/margin_std": 201.88861083984375, "step": 523 }, { "KL/chosen_KL_mean": -332.4698791503906, "KL/mean": -422.786865234375, "KL/rejected_KL_mean": -513.1038818359375, "KL/std": 188.12332153320312, "epoch": 0.7694566813509545, "fcm_dpo/beta": 0.0043691061437129974, "fcm_dpo/delta": 0.011401128023862839, "fcm_dpo/margin": 180.63401794433594, "fcm_dpo/q_t": 0.3315002918243408, "grad_norm": 45.90804672241211, "learning_rate": 7.781828926091535e-08, "logits/chosen": 0.15166175365447998, "logits/rejected": 0.19428567588329315, "logps/chosen": -410.54193115234375, "logps/ref_chosen": -78.0720443725586, "logps/ref_rejected": -81.30198669433594, "logps/rejected": -594.4058837890625, "loss": 0.877, "margin_dpo/margin_mean": 180.63400268554688, "margin_dpo/margin_std": 177.1727294921875, "step": 524 }, { "KL/chosen_KL_mean": -329.3613586425781, "KL/mean": -443.7110290527344, "KL/rejected_KL_mean": -558.0606689453125, "KL/std": 204.1719970703125, "epoch": 0.7709251101321586, "fcm_dpo/beta": 0.004234597086906433, "fcm_dpo/delta": -0.18942071497440338, "fcm_dpo/margin": 228.69935607910156, "fcm_dpo/q_t": 0.2971675992012024, "grad_norm": 40.713829040527344, "learning_rate": 7.689012058193384e-08, "logits/chosen": 0.23026111721992493, "logits/rejected": 0.23547565937042236, "logps/chosen": -380.189208984375, "logps/ref_chosen": -50.827857971191406, "logps/ref_rejected": -100.05294036865234, "logps/rejected": -658.1136474609375, "loss": 0.7668, "margin_dpo/margin_mean": 228.69935607910156, "margin_dpo/margin_std": 187.14825439453125, "step": 525 }, { "KL/chosen_KL_mean": -326.16119384765625, "KL/mean": -438.5784606933594, "KL/rejected_KL_mean": -550.9957275390625, "KL/std": 197.4532470703125, "epoch": 0.7723935389133627, "fcm_dpo/beta": 0.004164829384535551, "fcm_dpo/delta": -0.1505172997713089, "fcm_dpo/margin": 224.83453369140625, "fcm_dpo/q_t": 0.3045281171798706, "grad_norm": 35.237789154052734, "learning_rate": 7.596651350926836e-08, "logits/chosen": 0.11118023097515106, "logits/rejected": 0.14706848561763763, "logps/chosen": -389.32843017578125, "logps/ref_chosen": -63.167236328125, "logps/ref_rejected": -86.30934143066406, "logps/rejected": -637.3050537109375, "loss": 0.7863, "margin_dpo/margin_mean": 224.83456420898438, "margin_dpo/margin_std": 190.66650390625, "step": 526 }, { "KL/chosen_KL_mean": -355.33746337890625, "KL/mean": -443.8697509765625, "KL/rejected_KL_mean": -532.4020385742188, "KL/std": 185.5819091796875, "epoch": 0.7738619676945668, "fcm_dpo/beta": 0.0041229259222745895, "fcm_dpo/delta": 0.07519938796758652, "fcm_dpo/margin": 177.0645751953125, "fcm_dpo/q_t": 0.3446522355079651, "grad_norm": 32.963722229003906, "learning_rate": 7.504749238082414e-08, "logits/chosen": 0.27513420581817627, "logits/rejected": 0.3138313889503479, "logps/chosen": -426.46612548828125, "logps/ref_chosen": -71.12867736816406, "logps/ref_rejected": -78.3425521850586, "logps/rejected": -610.74462890625, "loss": 0.8913, "margin_dpo/margin_mean": 177.0645751953125, "margin_dpo/margin_std": 170.20639038085938, "step": 527 }, { "KL/chosen_KL_mean": -355.3000183105469, "KL/mean": -449.8873291015625, "KL/rejected_KL_mean": -544.474609375, "KL/std": 202.4794921875, "epoch": 0.775330396475771, "fcm_dpo/beta": 0.004161626100540161, "fcm_dpo/delta": 0.013718700036406517, "fcm_dpo/margin": 189.17462158203125, "fcm_dpo/q_t": 0.34284335374832153, "grad_norm": 35.8046760559082, "learning_rate": 7.413308141366254e-08, "logits/chosen": 0.22551241517066956, "logits/rejected": 0.2533000111579895, "logps/chosen": -423.38946533203125, "logps/ref_chosen": -68.0894546508789, "logps/ref_rejected": -93.91006469726562, "logps/rejected": -638.3847045898438, "loss": 0.9099, "margin_dpo/margin_mean": 189.17462158203125, "margin_dpo/margin_std": 217.96946716308594, "step": 528 }, { "KL/chosen_KL_mean": -412.2044372558594, "KL/mean": -486.0581970214844, "KL/rejected_KL_mean": -559.9119262695312, "KL/std": 171.0134735107422, "epoch": 0.7767988252569751, "fcm_dpo/beta": 0.004288380034267902, "fcm_dpo/delta": 0.1777651607990265, "fcm_dpo/margin": 147.70750427246094, "fcm_dpo/q_t": 0.3686566948890686, "grad_norm": 50.290096282958984, "learning_rate": 7.322330470336313e-08, "logits/chosen": 0.32002973556518555, "logits/rejected": 0.32282739877700806, "logps/chosen": -467.7794189453125, "logps/ref_chosen": -55.57495880126953, "logps/ref_rejected": -89.20909118652344, "logps/rejected": -649.1210327148438, "loss": 1.0074, "margin_dpo/margin_mean": 147.70748901367188, "margin_dpo/margin_std": 195.08078002929688, "step": 529 }, { "KL/chosen_KL_mean": -339.9566650390625, "KL/mean": -456.3387451171875, "KL/rejected_KL_mean": -572.7208251953125, "KL/std": 200.80197143554688, "epoch": 0.7782672540381792, "fcm_dpo/beta": 0.004182647448033094, "fcm_dpo/delta": -0.19384393095970154, "fcm_dpo/margin": 232.76409912109375, "fcm_dpo/q_t": 0.3022667169570923, "grad_norm": 34.06055450439453, "learning_rate": 7.231818622338822e-08, "logits/chosen": 0.2138541340827942, "logits/rejected": 0.2278527319431305, "logps/chosen": -387.55810546875, "logps/ref_chosen": -47.601417541503906, "logps/ref_rejected": -87.2845230102539, "logps/rejected": -660.00537109375, "loss": 0.8102, "margin_dpo/margin_mean": 232.7640838623047, "margin_dpo/margin_std": 222.93911743164062, "step": 530 }, { "KL/chosen_KL_mean": -359.77294921875, "KL/mean": -463.82427978515625, "KL/rejected_KL_mean": -567.8756103515625, "KL/std": 231.09474182128906, "epoch": 0.7797356828193832, "fcm_dpo/beta": 0.004124460741877556, "fcm_dpo/delta": -0.06376082450151443, "fcm_dpo/margin": 208.1026153564453, "fcm_dpo/q_t": 0.3246777057647705, "grad_norm": 45.138309478759766, "learning_rate": 7.141774982445147e-08, "logits/chosen": 0.2456667125225067, "logits/rejected": 0.27481505274772644, "logps/chosen": -415.01904296875, "logps/ref_chosen": -55.246063232421875, "logps/ref_rejected": -70.60598754882812, "logps/rejected": -638.4815673828125, "loss": 0.847, "margin_dpo/margin_mean": 208.10260009765625, "margin_dpo/margin_std": 207.6148681640625, "step": 531 }, { "KL/chosen_KL_mean": -383.8913269042969, "KL/mean": -493.2384033203125, "KL/rejected_KL_mean": -602.5855102539062, "KL/std": 211.597900390625, "epoch": 0.7812041116005873, "fcm_dpo/beta": 0.0039962041191756725, "fcm_dpo/delta": -0.08703669160604477, "fcm_dpo/margin": 218.69418334960938, "fcm_dpo/q_t": 0.32373446226119995, "grad_norm": 54.648067474365234, "learning_rate": 7.052201923388953e-08, "logits/chosen": 0.2216191440820694, "logits/rejected": 0.25669944286346436, "logps/chosen": -454.1773376464844, "logps/ref_chosen": -70.28601837158203, "logps/ref_rejected": -86.5913314819336, "logps/rejected": -689.1768798828125, "loss": 0.8813, "margin_dpo/margin_mean": 218.69418334960938, "margin_dpo/margin_std": 233.58529663085938, "step": 532 }, { "KL/chosen_KL_mean": -381.64031982421875, "KL/mean": -468.139892578125, "KL/rejected_KL_mean": -554.6394653320312, "KL/std": 193.63104248046875, "epoch": 0.7826725403817915, "fcm_dpo/beta": 0.00412764260545373, "fcm_dpo/delta": 0.09021273255348206, "fcm_dpo/margin": 172.99917602539062, "fcm_dpo/q_t": 0.3548873960971832, "grad_norm": 38.81024932861328, "learning_rate": 6.963101805503646e-08, "logits/chosen": 0.34045305848121643, "logits/rejected": 0.37642914056777954, "logps/chosen": -446.49542236328125, "logps/ref_chosen": -64.8551025390625, "logps/ref_rejected": -76.58805847167969, "logps/rejected": -631.2275390625, "loss": 0.9619, "margin_dpo/margin_mean": 172.99917602539062, "margin_dpo/margin_std": 212.29209899902344, "step": 533 }, { "KL/chosen_KL_mean": -384.55780029296875, "KL/mean": -476.9858093261719, "KL/rejected_KL_mean": -569.413818359375, "KL/std": 202.78414916992188, "epoch": 0.7841409691629956, "fcm_dpo/beta": 0.004138133954256773, "fcm_dpo/delta": 0.03789476677775383, "fcm_dpo/margin": 184.8560791015625, "fcm_dpo/q_t": 0.3369339108467102, "grad_norm": 42.43622970581055, "learning_rate": 6.874476976660184e-08, "logits/chosen": 0.2626418173313141, "logits/rejected": 0.2821497619152069, "logps/chosen": -444.67718505859375, "logps/ref_chosen": -60.119388580322266, "logps/ref_rejected": -78.54347229003906, "logps/rejected": -647.957275390625, "loss": 0.8917, "margin_dpo/margin_mean": 184.8560791015625, "margin_dpo/margin_std": 190.45816040039062, "step": 534 }, { "KL/chosen_KL_mean": -347.71710205078125, "KL/mean": -462.66558837890625, "KL/rejected_KL_mean": -577.6140747070312, "KL/std": 206.4551544189453, "epoch": 0.7856093979441997, "fcm_dpo/beta": 0.004087292123585939, "fcm_dpo/delta": -0.15449239313602448, "fcm_dpo/margin": 229.89695739746094, "fcm_dpo/q_t": 0.3038337826728821, "grad_norm": 39.14911651611328, "learning_rate": 6.786329772205246e-08, "logits/chosen": 0.2347976267337799, "logits/rejected": 0.2542041838169098, "logps/chosen": -402.0473327636719, "logps/ref_chosen": -54.330238342285156, "logps/ref_rejected": -96.30763244628906, "logps/rejected": -673.9216918945312, "loss": 0.7864, "margin_dpo/margin_mean": 229.89694213867188, "margin_dpo/margin_std": 190.98971557617188, "step": 535 }, { "KL/chosen_KL_mean": -333.5712890625, "KL/mean": -463.6264953613281, "KL/rejected_KL_mean": -593.681640625, "KL/std": 235.50570678710938, "epoch": 0.7870778267254038, "fcm_dpo/beta": 0.0038786856457591057, "fcm_dpo/delta": -0.23251280188560486, "fcm_dpo/margin": 260.1103515625, "fcm_dpo/q_t": 0.29672110080718994, "grad_norm": 32.708805084228516, "learning_rate": 6.698662514899638e-08, "logits/chosen": 0.28418147563934326, "logits/rejected": 0.30755770206451416, "logps/chosen": -380.65185546875, "logps/ref_chosen": -47.08053207397461, "logps/ref_rejected": -89.09783935546875, "logps/rejected": -682.7794799804688, "loss": 0.76, "margin_dpo/margin_mean": 260.1103515625, "margin_dpo/margin_std": 234.55047607421875, "step": 536 }, { "KL/chosen_KL_mean": -354.51861572265625, "KL/mean": -460.71697998046875, "KL/rejected_KL_mean": -566.9154052734375, "KL/std": 200.9781494140625, "epoch": 0.788546255506608, "fcm_dpo/beta": 0.003838079050183296, "fcm_dpo/delta": -0.01674051769077778, "fcm_dpo/margin": 212.39678955078125, "fcm_dpo/q_t": 0.3316698670387268, "grad_norm": 39.343170166015625, "learning_rate": 6.611477514857114e-08, "logits/chosen": 0.33055514097213745, "logits/rejected": 0.3605949878692627, "logps/chosen": -412.2660827636719, "logps/ref_chosen": -57.747467041015625, "logps/ref_rejected": -70.43838500976562, "logps/rejected": -637.353759765625, "loss": 0.8782, "margin_dpo/margin_mean": 212.39678955078125, "margin_dpo/margin_std": 221.20095825195312, "step": 537 }, { "KL/chosen_KL_mean": -386.23681640625, "KL/mean": -490.32037353515625, "KL/rejected_KL_mean": -594.4039306640625, "KL/std": 197.8646697998047, "epoch": 0.7900146842878121, "fcm_dpo/beta": 0.003815419040620327, "fcm_dpo/delta": 0.006140880286693573, "fcm_dpo/margin": 208.16705322265625, "fcm_dpo/q_t": 0.3316478431224823, "grad_norm": 46.62859344482422, "learning_rate": 6.524777069483525e-08, "logits/chosen": 0.1873122751712799, "logits/rejected": 0.21275225281715393, "logps/chosen": -452.65277099609375, "logps/ref_chosen": -66.41594696044922, "logps/ref_rejected": -84.22808837890625, "logps/rejected": -678.6319580078125, "loss": 0.863, "margin_dpo/margin_mean": 208.16705322265625, "margin_dpo/margin_std": 199.23483276367188, "step": 538 }, { "KL/chosen_KL_mean": -381.7893981933594, "KL/mean": -473.711669921875, "KL/rejected_KL_mean": -565.6339721679688, "KL/std": 190.73289489746094, "epoch": 0.7914831130690162, "fcm_dpo/beta": 0.003870626911520958, "fcm_dpo/delta": 0.09528693556785583, "fcm_dpo/margin": 183.84457397460938, "fcm_dpo/q_t": 0.3515859842300415, "grad_norm": 36.88890838623047, "learning_rate": 6.438563463416221e-08, "logits/chosen": 0.2974720001220703, "logits/rejected": 0.335232138633728, "logps/chosen": -440.2822570800781, "logps/ref_chosen": -58.492855072021484, "logps/ref_rejected": -91.85395050048828, "logps/rejected": -657.4879150390625, "loss": 0.9384, "margin_dpo/margin_mean": 183.84457397460938, "margin_dpo/margin_std": 208.98944091796875, "step": 539 }, { "KL/chosen_KL_mean": -350.2906494140625, "KL/mean": -476.7803955078125, "KL/rejected_KL_mean": -603.2701416015625, "KL/std": 221.31947326660156, "epoch": 0.7929515418502202, "fcm_dpo/beta": 0.003814155701547861, "fcm_dpo/delta": -0.18212105333805084, "fcm_dpo/margin": 252.97952270507812, "fcm_dpo/q_t": 0.30182313919067383, "grad_norm": 34.22950744628906, "learning_rate": 6.352838968463919e-08, "logits/chosen": 0.20235419273376465, "logits/rejected": 0.2219226211309433, "logps/chosen": -413.77313232421875, "logps/ref_chosen": -63.482513427734375, "logps/ref_rejected": -116.42999267578125, "logps/rejected": -719.7001342773438, "loss": 0.7934, "margin_dpo/margin_mean": 252.97952270507812, "margin_dpo/margin_std": 223.89288330078125, "step": 540 }, { "KL/chosen_KL_mean": -401.22979736328125, "KL/mean": -487.3087158203125, "KL/rejected_KL_mean": -573.3875732421875, "KL/std": 208.10055541992188, "epoch": 0.7944199706314243, "fcm_dpo/beta": 0.0038143517449498177, "fcm_dpo/delta": 0.1531994491815567, "fcm_dpo/margin": 172.1578369140625, "fcm_dpo/q_t": 0.3660222291946411, "grad_norm": 36.54274368286133, "learning_rate": 6.267605843546767e-08, "logits/chosen": 0.24667702615261078, "logits/rejected": 0.2629375755786896, "logps/chosen": -479.5101623535156, "logps/ref_chosen": -78.28036499023438, "logps/ref_rejected": -103.273681640625, "logps/rejected": -676.6612548828125, "loss": 0.9814, "margin_dpo/margin_mean": 172.1578369140625, "margin_dpo/margin_std": 218.95196533203125, "step": 541 }, { "KL/chosen_KL_mean": -390.4757080078125, "KL/mean": -500.078369140625, "KL/rejected_KL_mean": -609.6810302734375, "KL/std": 207.84117126464844, "epoch": 0.7958883994126285, "fcm_dpo/beta": 0.003827635897323489, "fcm_dpo/delta": -0.043748270720243454, "fcm_dpo/margin": 219.20529174804688, "fcm_dpo/q_t": 0.3255137801170349, "grad_norm": 54.06191635131836, "learning_rate": 6.182866334636888e-08, "logits/chosen": 0.33256542682647705, "logits/rejected": 0.34963661432266235, "logps/chosen": -447.960693359375, "logps/ref_chosen": -57.48497009277344, "logps/ref_rejected": -96.47506713867188, "logps/rejected": -706.1560668945312, "loss": 0.8597, "margin_dpo/margin_mean": 219.20529174804688, "margin_dpo/margin_std": 215.65493774414062, "step": 542 }, { "KL/chosen_KL_mean": -401.6968994140625, "KL/mean": -475.1959533691406, "KL/rejected_KL_mean": -548.6950073242188, "KL/std": 224.97775268554688, "epoch": 0.7973568281938326, "fcm_dpo/beta": 0.003976443782448769, "fcm_dpo/delta": 0.22884975373744965, "fcm_dpo/margin": 146.99813842773438, "fcm_dpo/q_t": 0.38513779640197754, "grad_norm": 41.66843032836914, "learning_rate": 6.098622674699147e-08, "logits/chosen": 0.216482013463974, "logits/rejected": 0.2214020937681198, "logps/chosen": -462.31439208984375, "logps/ref_chosen": -60.61750793457031, "logps/ref_rejected": -105.59896850585938, "logps/rejected": -654.2939453125, "loss": 1.0618, "margin_dpo/margin_mean": 146.99813842773438, "margin_dpo/margin_std": 232.58026123046875, "step": 543 }, { "KL/chosen_KL_mean": -414.082275390625, "KL/mean": -527.2254028320312, "KL/rejected_KL_mean": -640.3685302734375, "KL/std": 213.15147399902344, "epoch": 0.7988252569750367, "fcm_dpo/beta": 0.00395975261926651, "fcm_dpo/delta": -0.1055314838886261, "fcm_dpo/margin": 226.28622436523438, "fcm_dpo/q_t": 0.3122187852859497, "grad_norm": 40.56058883666992, "learning_rate": 6.01487708363232e-08, "logits/chosen": 0.2870156764984131, "logits/rejected": 0.3034914433956146, "logps/chosen": -473.724609375, "logps/ref_chosen": -59.642303466796875, "logps/ref_rejected": -100.95469665527344, "logps/rejected": -741.3232421875, "loss": 0.8012, "margin_dpo/margin_mean": 226.2862091064453, "margin_dpo/margin_std": 201.3153076171875, "step": 544 }, { "KL/chosen_KL_mean": -384.6066589355469, "KL/mean": -493.9320068359375, "KL/rejected_KL_mean": -603.25732421875, "KL/std": 203.6180877685547, "epoch": 0.8002936857562408, "fcm_dpo/beta": 0.0039124698378145695, "fcm_dpo/delta": -0.060617730021476746, "fcm_dpo/margin": 218.65066528320312, "fcm_dpo/q_t": 0.32227370142936707, "grad_norm": 41.63042068481445, "learning_rate": 5.9316317682106294e-08, "logits/chosen": 0.24290139973163605, "logits/rejected": 0.2708423137664795, "logps/chosen": -452.2552490234375, "logps/ref_chosen": -67.64859771728516, "logps/ref_rejected": -95.90800476074219, "logps/rejected": -699.1653442382812, "loss": 0.8485, "margin_dpo/margin_mean": 218.65066528320312, "margin_dpo/margin_std": 211.84780883789062, "step": 545 }, { "KL/chosen_KL_mean": -358.6064453125, "KL/mean": -456.1077575683594, "KL/rejected_KL_mean": -553.6090698242188, "KL/std": 187.7569580078125, "epoch": 0.801762114537445, "fcm_dpo/beta": 0.0038997926749289036, "fcm_dpo/delta": 0.04258295148611069, "fcm_dpo/margin": 195.0026397705078, "fcm_dpo/q_t": 0.334590345621109, "grad_norm": 35.64754867553711, "learning_rate": 5.848888922025552e-08, "logits/chosen": 0.24750055372714996, "logits/rejected": 0.25986167788505554, "logps/chosen": -409.3506774902344, "logps/ref_chosen": -50.744232177734375, "logps/ref_rejected": -81.86622619628906, "logps/rejected": -635.475341796875, "loss": 0.8731, "margin_dpo/margin_mean": 195.0026397705078, "margin_dpo/margin_std": 179.05099487304688, "step": 546 }, { "KL/chosen_KL_mean": -405.56817626953125, "KL/mean": -502.1980285644531, "KL/rejected_KL_mean": -598.827880859375, "KL/std": 210.65838623046875, "epoch": 0.8032305433186491, "fcm_dpo/beta": 0.00398046150803566, "fcm_dpo/delta": 0.031187400221824646, "fcm_dpo/margin": 193.25973510742188, "fcm_dpo/q_t": 0.3425300419330597, "grad_norm": 67.2081069946289, "learning_rate": 5.7666507254280265e-08, "logits/chosen": 0.15230880677700043, "logits/rejected": 0.18068407475948334, "logps/chosen": -479.2558898925781, "logps/ref_chosen": -73.6877212524414, "logps/ref_rejected": -90.76136779785156, "logps/rejected": -689.5892333984375, "loss": 0.9202, "margin_dpo/margin_mean": 193.25973510742188, "margin_dpo/margin_std": 217.03179931640625, "step": 547 }, { "KL/chosen_KL_mean": -402.8782958984375, "KL/mean": -503.20465087890625, "KL/rejected_KL_mean": -603.531005859375, "KL/std": 230.92819213867188, "epoch": 0.8046989720998532, "fcm_dpo/beta": 0.00395785178989172, "fcm_dpo/delta": 0.006317374296486378, "fcm_dpo/margin": 200.65269470214844, "fcm_dpo/q_t": 0.337555468082428, "grad_norm": 41.28041076660156, "learning_rate": 5.684919345471029e-08, "logits/chosen": 0.2283911406993866, "logits/rejected": 0.25433415174484253, "logps/chosen": -468.1246337890625, "logps/ref_chosen": -65.24634552001953, "logps/ref_rejected": -94.11807250976562, "logps/rejected": -697.6490478515625, "loss": 0.8981, "margin_dpo/margin_mean": 200.65267944335938, "margin_dpo/margin_std": 227.65338134765625, "step": 548 }, { "KL/chosen_KL_mean": -392.39306640625, "KL/mean": -486.2154846191406, "KL/rejected_KL_mean": -580.037841796875, "KL/std": 183.3447723388672, "epoch": 0.8061674008810573, "fcm_dpo/beta": 0.003991181496530771, "fcm_dpo/delta": 0.05513293296098709, "fcm_dpo/margin": 187.64480590820312, "fcm_dpo/q_t": 0.34387335181236267, "grad_norm": 46.40816879272461, "learning_rate": 5.603696935852426e-08, "logits/chosen": 0.38207992911338806, "logits/rejected": 0.4072762727737427, "logps/chosen": -441.6054382324219, "logps/ref_chosen": -49.21235656738281, "logps/ref_rejected": -73.91031646728516, "logps/rejected": -653.9481811523438, "loss": 0.9046, "margin_dpo/margin_mean": 187.64480590820312, "margin_dpo/margin_std": 201.8494415283203, "step": 549 }, { "KL/chosen_KL_mean": -383.96148681640625, "KL/mean": -468.65374755859375, "KL/rejected_KL_mean": -553.3460083007812, "KL/std": 199.1809539794922, "epoch": 0.8076358296622613, "fcm_dpo/beta": 0.00405759084969759, "fcm_dpo/delta": 0.12119770795106888, "fcm_dpo/margin": 169.38455200195312, "fcm_dpo/q_t": 0.35117456316947937, "grad_norm": 39.510982513427734, "learning_rate": 5.5229856368582376e-08, "logits/chosen": 0.21509867906570435, "logits/rejected": 0.2270488440990448, "logps/chosen": -440.7684326171875, "logps/ref_chosen": -56.80695343017578, "logps/ref_rejected": -95.12580871582031, "logps/rejected": -648.4718017578125, "loss": 0.9164, "margin_dpo/margin_mean": 169.38455200195312, "margin_dpo/margin_std": 168.8013458251953, "step": 550 }, { "KL/chosen_KL_mean": -375.353515625, "KL/mean": -500.36590576171875, "KL/rejected_KL_mean": -625.3782958984375, "KL/std": 219.81362915039062, "epoch": 0.8091042584434655, "fcm_dpo/beta": 0.0039620669558644295, "fcm_dpo/delta": -0.21213532984256744, "fcm_dpo/margin": 250.02468872070312, "fcm_dpo/q_t": 0.29334428906440735, "grad_norm": 48.182064056396484, "learning_rate": 5.4427875753062734e-08, "logits/chosen": 0.23227638006210327, "logits/rejected": 0.24997326731681824, "logps/chosen": -434.4598693847656, "logps/ref_chosen": -59.10633087158203, "logps/ref_rejected": -111.67280578613281, "logps/rejected": -737.051025390625, "loss": 0.7635, "margin_dpo/margin_mean": 250.02468872070312, "margin_dpo/margin_std": 206.69105529785156, "step": 551 }, { "KL/chosen_KL_mean": -364.43963623046875, "KL/mean": -506.6065979003906, "KL/rejected_KL_mean": -648.7735595703125, "KL/std": 241.85960388183594, "epoch": 0.8105726872246696, "fcm_dpo/beta": 0.0037087902892380953, "fcm_dpo/delta": -0.29273247718811035, "fcm_dpo/margin": 284.33392333984375, "fcm_dpo/q_t": 0.2889665365219116, "grad_norm": 31.593692779541016, "learning_rate": 5.363104864490034e-08, "logits/chosen": 0.28846144676208496, "logits/rejected": 0.3311532139778137, "logps/chosen": -426.79425048828125, "logps/ref_chosen": -62.35459899902344, "logps/ref_rejected": -104.56210327148438, "logps/rejected": -753.335693359375, "loss": 0.7566, "margin_dpo/margin_mean": 284.33392333984375, "margin_dpo/margin_std": 249.8412322998047, "step": 552 }, { "KL/chosen_KL_mean": -394.02691650390625, "KL/mean": -487.235107421875, "KL/rejected_KL_mean": -580.4432983398438, "KL/std": 202.1296844482422, "epoch": 0.8120411160058737, "fcm_dpo/beta": 0.0037544944789260626, "fcm_dpo/delta": 0.10777129977941513, "fcm_dpo/margin": 186.4163818359375, "fcm_dpo/q_t": 0.3546206057071686, "grad_norm": 40.13036346435547, "learning_rate": 5.2839396041230415e-08, "logits/chosen": 0.2658536434173584, "logits/rejected": 0.2780281603336334, "logps/chosen": -462.28570556640625, "logps/ref_chosen": -68.25881958007812, "logps/ref_rejected": -98.0971450805664, "logps/rejected": -678.5404663085938, "loss": 0.9369, "margin_dpo/margin_mean": 186.4163818359375, "margin_dpo/margin_std": 208.78411865234375, "step": 553 }, { "KL/chosen_KL_mean": -389.50079345703125, "KL/mean": -510.76385498046875, "KL/rejected_KL_mean": -632.02685546875, "KL/std": 229.28463745117188, "epoch": 0.8135095447870778, "fcm_dpo/beta": 0.0037528942339122295, "fcm_dpo/delta": -0.12198401987552643, "fcm_dpo/margin": 242.52606201171875, "fcm_dpo/q_t": 0.31189385056495667, "grad_norm": 38.72057342529297, "learning_rate": 5.205293880283551e-08, "logits/chosen": 0.222869411110878, "logits/rejected": 0.27173179388046265, "logps/chosen": -457.448486328125, "logps/ref_chosen": -67.94767761230469, "logps/ref_rejected": -89.78272247314453, "logps/rejected": -721.8095703125, "loss": 0.838, "margin_dpo/margin_mean": 242.52606201171875, "margin_dpo/margin_std": 234.7236328125, "step": 554 }, { "KL/chosen_KL_mean": -394.62530517578125, "KL/mean": -514.3497924804688, "KL/rejected_KL_mean": -634.0742797851562, "KL/std": 226.64907836914062, "epoch": 0.8149779735682819, "fcm_dpo/beta": 0.0036291368305683136, "fcm_dpo/delta": -0.07678112387657166, "fcm_dpo/margin": 239.44900512695312, "fcm_dpo/q_t": 0.32688435912132263, "grad_norm": 41.123992919921875, "learning_rate": 5.127169765359515e-08, "logits/chosen": 0.24900421500205994, "logits/rejected": 0.2614419162273407, "logps/chosen": -447.9557800292969, "logps/ref_chosen": -53.33049011230469, "logps/ref_rejected": -108.47937774658203, "logps/rejected": -742.5537109375, "loss": 0.8651, "margin_dpo/margin_mean": 239.44900512695312, "margin_dpo/margin_std": 255.90615844726562, "step": 555 }, { "KL/chosen_KL_mean": -399.20941162109375, "KL/mean": -492.0675048828125, "KL/rejected_KL_mean": -584.925537109375, "KL/std": 213.84945678710938, "epoch": 0.8164464023494861, "fcm_dpo/beta": 0.0037066666409373283, "fcm_dpo/delta": 0.11965786665678024, "fcm_dpo/margin": 185.71615600585938, "fcm_dpo/q_t": 0.3593837022781372, "grad_norm": 39.725162506103516, "learning_rate": 5.049569317994012e-08, "logits/chosen": 0.331043541431427, "logits/rejected": 0.3502947688102722, "logps/chosen": -457.8538818359375, "logps/ref_chosen": -58.64447021484375, "logps/ref_rejected": -101.34040832519531, "logps/rejected": -686.2659912109375, "loss": 0.9503, "margin_dpo/margin_mean": 185.71615600585938, "margin_dpo/margin_std": 223.22207641601562, "step": 556 }, { "KL/chosen_KL_mean": -410.8165588378906, "KL/mean": -513.6693115234375, "KL/rejected_KL_mean": -616.5220947265625, "KL/std": 224.26837158203125, "epoch": 0.8179148311306902, "fcm_dpo/beta": 0.003730146214365959, "fcm_dpo/delta": 0.035248152911663055, "fcm_dpo/margin": 205.70550537109375, "fcm_dpo/q_t": 0.3421016335487366, "grad_norm": 39.09975051879883, "learning_rate": 4.9724945830310144e-08, "logits/chosen": 0.2737029194831848, "logits/rejected": 0.29351887106895447, "logps/chosen": -478.6572265625, "logps/ref_chosen": -67.84066009521484, "logps/ref_rejected": -109.93965911865234, "logps/rejected": -726.4617309570312, "loss": 0.9037, "margin_dpo/margin_mean": 205.70550537109375, "margin_dpo/margin_std": 222.4257354736328, "step": 557 }, { "KL/chosen_KL_mean": -386.15411376953125, "KL/mean": -531.5560913085938, "KL/rejected_KL_mean": -676.9581298828125, "KL/std": 238.99404907226562, "epoch": 0.8193832599118943, "fcm_dpo/beta": 0.0035815369337797165, "fcm_dpo/delta": -0.2720809578895569, "fcm_dpo/margin": 290.80401611328125, "fcm_dpo/q_t": 0.2847321629524231, "grad_norm": 30.991840362548828, "learning_rate": 4.8959475914614554e-08, "logits/chosen": 0.3357253074645996, "logits/rejected": 0.3656435012817383, "logps/chosen": -448.5223388671875, "logps/ref_chosen": -62.36824035644531, "logps/ref_rejected": -102.16102600097656, "logps/rejected": -779.119140625, "loss": 0.7381, "margin_dpo/margin_mean": 290.8039855957031, "margin_dpo/margin_std": 235.93875122070312, "step": 558 }, { "KL/chosen_KL_mean": -433.37945556640625, "KL/mean": -544.3544921875, "KL/rejected_KL_mean": -655.32958984375, "KL/std": 226.74810791015625, "epoch": 0.8208516886930984, "fcm_dpo/beta": 0.0035474528558552265, "fcm_dpo/delta": 0.013534316793084145, "fcm_dpo/margin": 221.95016479492188, "fcm_dpo/q_t": 0.3346388638019562, "grad_norm": 42.02165603637695, "learning_rate": 4.8199303603697614e-08, "logits/chosen": 0.40246227383613586, "logits/rejected": 0.45247191190719604, "logps/chosen": -494.13177490234375, "logps/ref_chosen": -60.752323150634766, "logps/ref_rejected": -93.44229125976562, "logps/rejected": -748.7718505859375, "loss": 0.8745, "margin_dpo/margin_mean": 221.95016479492188, "margin_dpo/margin_std": 219.32046508789062, "step": 559 }, { "KL/chosen_KL_mean": -384.0647277832031, "KL/mean": -466.8363342285156, "KL/rejected_KL_mean": -549.60791015625, "KL/std": 207.88125610351562, "epoch": 0.8223201174743024, "fcm_dpo/beta": 0.0036635929718613625, "fcm_dpo/delta": 0.20640337467193604, "fcm_dpo/margin": 165.54315185546875, "fcm_dpo/q_t": 0.3670238256454468, "grad_norm": 40.4533805847168, "learning_rate": 4.7444448928806615e-08, "logits/chosen": 0.19835129380226135, "logits/rejected": 0.23476439714431763, "logps/chosen": -442.1685485839844, "logps/ref_chosen": -58.10382080078125, "logps/ref_rejected": -79.99122619628906, "logps/rejected": -629.59912109375, "loss": 0.9562, "margin_dpo/margin_mean": 165.54315185546875, "margin_dpo/margin_std": 169.37548828125, "step": 560 }, { "KL/chosen_KL_mean": -452.8956298828125, "KL/mean": -540.2266845703125, "KL/rejected_KL_mean": -627.5577392578125, "KL/std": 207.67529296875, "epoch": 0.8237885462555066, "fcm_dpo/beta": 0.0038133093621581793, "fcm_dpo/delta": 0.14196228981018066, "fcm_dpo/margin": 174.6620635986328, "fcm_dpo/q_t": 0.3567246198654175, "grad_norm": 46.923065185546875, "learning_rate": 4.669493178106432e-08, "logits/chosen": 0.39849230647087097, "logits/rejected": 0.41054990887641907, "logps/chosen": -503.8085021972656, "logps/ref_chosen": -50.912879943847656, "logps/ref_rejected": -99.06856536865234, "logps/rejected": -726.6262817382812, "loss": 0.9425, "margin_dpo/margin_mean": 174.6620635986328, "margin_dpo/margin_std": 190.61865234375, "step": 561 }, { "KL/chosen_KL_mean": -404.7358703613281, "KL/mean": -527.0230712890625, "KL/rejected_KL_mean": -649.310302734375, "KL/std": 235.49441528320312, "epoch": 0.8252569750367107, "fcm_dpo/beta": 0.003720635548233986, "fcm_dpo/delta": -0.12324625253677368, "fcm_dpo/margin": 244.5744171142578, "fcm_dpo/q_t": 0.31406521797180176, "grad_norm": 39.70796585083008, "learning_rate": 4.5950771910944596e-08, "logits/chosen": 0.339399516582489, "logits/rejected": 0.3695017397403717, "logps/chosen": -464.20025634765625, "logps/ref_chosen": -59.46440124511719, "logps/ref_rejected": -96.54266357421875, "logps/rejected": -745.8529663085938, "loss": 0.812, "margin_dpo/margin_mean": 244.57443237304688, "margin_dpo/margin_std": 231.22760009765625, "step": 562 }, { "KL/chosen_KL_mean": -439.4453430175781, "KL/mean": -530.522216796875, "KL/rejected_KL_mean": -621.59912109375, "KL/std": 201.02435302734375, "epoch": 0.8267254038179148, "fcm_dpo/beta": 0.003821668680757284, "fcm_dpo/delta": 0.10885676741600037, "fcm_dpo/margin": 182.15380859375, "fcm_dpo/q_t": 0.35326603055000305, "grad_norm": 45.26503372192383, "learning_rate": 4.521198892775202e-08, "logits/chosen": 0.4035786986351013, "logits/rejected": 0.41392847895622253, "logps/chosen": -500.05352783203125, "logps/ref_chosen": -60.60819625854492, "logps/ref_rejected": -94.56770324707031, "logps/rejected": -716.1668701171875, "loss": 0.9582, "margin_dpo/margin_mean": 182.15380859375, "margin_dpo/margin_std": 211.79025268554688, "step": 563 }, { "KL/chosen_KL_mean": -413.4257507324219, "KL/mean": -526.880126953125, "KL/rejected_KL_mean": -640.33447265625, "KL/std": 224.15956115722656, "epoch": 0.8281938325991189, "fcm_dpo/beta": 0.0037801191210746765, "fcm_dpo/delta": -0.0631168857216835, "fcm_dpo/margin": 226.90872192382812, "fcm_dpo/q_t": 0.31227725744247437, "grad_norm": 34.521514892578125, "learning_rate": 4.447860229910544e-08, "logits/chosen": 0.3029387593269348, "logits/rejected": 0.3524191975593567, "logps/chosen": -487.6941223144531, "logps/ref_chosen": -74.26837921142578, "logps/ref_rejected": -93.23818969726562, "logps/rejected": -733.5726318359375, "loss": 0.7874, "margin_dpo/margin_mean": 226.90870666503906, "margin_dpo/margin_std": 165.97833251953125, "step": 564 }, { "KL/chosen_KL_mean": -429.74359130859375, "KL/mean": -546.124267578125, "KL/rejected_KL_mean": -662.5048828125, "KL/std": 252.56912231445312, "epoch": 0.8296622613803231, "fcm_dpo/beta": 0.0037043162155896425, "fcm_dpo/delta": -0.06889684498310089, "fcm_dpo/margin": 232.76126098632812, "fcm_dpo/q_t": 0.33259573578834534, "grad_norm": 33.25062942504883, "learning_rate": 4.375063135042445e-08, "logits/chosen": 0.2388344705104828, "logits/rejected": 0.26381534337997437, "logps/chosen": -498.7635192871094, "logps/ref_chosen": -69.0199203491211, "logps/ref_rejected": -85.7789306640625, "logps/rejected": -748.2838134765625, "loss": 0.8844, "margin_dpo/margin_mean": 232.76126098632812, "margin_dpo/margin_std": 266.616455078125, "step": 565 }, { "KL/chosen_KL_mean": -427.2159423828125, "KL/mean": -536.328857421875, "KL/rejected_KL_mean": -645.4417724609375, "KL/std": 245.62149047851562, "epoch": 0.8311306901615272, "fcm_dpo/beta": 0.003733118064701557, "fcm_dpo/delta": -0.017540939152240753, "fcm_dpo/margin": 218.225830078125, "fcm_dpo/q_t": 0.3383902311325073, "grad_norm": 40.47699737548828, "learning_rate": 4.3028095264420525e-08, "logits/chosen": 0.3286312520503998, "logits/rejected": 0.3403201997280121, "logps/chosen": -493.76123046875, "logps/ref_chosen": -66.5453109741211, "logps/ref_rejected": -103.86932373046875, "logps/rejected": -749.31103515625, "loss": 0.9307, "margin_dpo/margin_mean": 218.22581481933594, "margin_dpo/margin_std": 258.7934265136719, "step": 566 }, { "KL/chosen_KL_mean": -430.94976806640625, "KL/mean": -536.2676391601562, "KL/rejected_KL_mean": -641.58544921875, "KL/std": 210.7244873046875, "epoch": 0.8325991189427313, "fcm_dpo/beta": 0.0036982346791774035, "fcm_dpo/delta": 0.022707507014274597, "fcm_dpo/margin": 210.63571166992188, "fcm_dpo/q_t": 0.33780205249786377, "grad_norm": 41.74602508544922, "learning_rate": 4.231101308059165e-08, "logits/chosen": 0.4672779440879822, "logits/rejected": 0.4847617745399475, "logps/chosen": -483.80804443359375, "logps/ref_chosen": -52.85829544067383, "logps/ref_rejected": -85.37095642089844, "logps/rejected": -726.9564208984375, "loss": 0.889, "margin_dpo/margin_mean": 210.63571166992188, "margin_dpo/margin_std": 218.34823608398438, "step": 567 }, { "KL/chosen_KL_mean": -407.42236328125, "KL/mean": -529.507080078125, "KL/rejected_KL_mean": -651.591796875, "KL/std": 220.82595825195312, "epoch": 0.8340675477239354, "fcm_dpo/beta": 0.0036355347838252783, "fcm_dpo/delta": -0.09775268286466599, "fcm_dpo/margin": 244.16946411132812, "fcm_dpo/q_t": 0.3122249245643616, "grad_norm": 37.94232940673828, "learning_rate": 4.1599403694720145e-08, "logits/chosen": 0.4218721389770508, "logits/rejected": 0.4371316432952881, "logps/chosen": -452.6147155761719, "logps/ref_chosen": -45.1923828125, "logps/ref_rejected": -89.09236907958984, "logps/rejected": -740.6842041015625, "loss": 0.8192, "margin_dpo/margin_mean": 244.16946411132812, "margin_dpo/margin_std": 215.54330444335938, "step": 568 }, { "KL/chosen_KL_mean": -464.3712158203125, "KL/mean": -566.7379760742188, "KL/rejected_KL_mean": -669.104736328125, "KL/std": 238.54409790039062, "epoch": 0.8355359765051396, "fcm_dpo/beta": 0.0036928458139300346, "fcm_dpo/delta": 0.04605846479535103, "fcm_dpo/margin": 204.73341369628906, "fcm_dpo/q_t": 0.34364962577819824, "grad_norm": 44.00316619873047, "learning_rate": 4.089328585837512e-08, "logits/chosen": 0.38823315501213074, "logits/rejected": 0.41456738114356995, "logps/chosen": -528.091796875, "logps/ref_chosen": -63.72056198120117, "logps/ref_rejected": -79.10325622558594, "logps/rejected": -748.2079467773438, "loss": 0.9298, "margin_dpo/margin_mean": 204.73342895507812, "margin_dpo/margin_std": 236.36395263671875, "step": 569 }, { "KL/chosen_KL_mean": -430.80474853515625, "KL/mean": -545.889404296875, "KL/rejected_KL_mean": -660.97412109375, "KL/std": 216.6941680908203, "epoch": 0.8370044052863436, "fcm_dpo/beta": 0.003635650733485818, "fcm_dpo/delta": -0.04079482704401016, "fcm_dpo/margin": 230.1693572998047, "fcm_dpo/q_t": 0.32262638211250305, "grad_norm": 45.467594146728516, "learning_rate": 4.019267817841834e-08, "logits/chosen": 0.40148618817329407, "logits/rejected": 0.43486684560775757, "logps/chosen": -492.4193115234375, "logps/ref_chosen": -61.61454391479492, "logps/ref_rejected": -82.14186096191406, "logps/rejected": -743.115966796875, "loss": 0.8298, "margin_dpo/margin_mean": 230.16937255859375, "margin_dpo/margin_std": 200.99362182617188, "step": 570 }, { "KL/chosen_KL_mean": -439.533447265625, "KL/mean": -559.2398681640625, "KL/rejected_KL_mean": -678.9462890625, "KL/std": 225.64910888671875, "epoch": 0.8384728340675477, "fcm_dpo/beta": 0.003588983789086342, "fcm_dpo/delta": -0.06584499031305313, "fcm_dpo/margin": 239.4127960205078, "fcm_dpo/q_t": 0.3223009705543518, "grad_norm": 41.48381805419922, "learning_rate": 3.9497599116513705e-08, "logits/chosen": 0.4067568778991699, "logits/rejected": 0.4149426817893982, "logps/chosen": -492.5875244140625, "logps/ref_chosen": -53.05406188964844, "logps/ref_rejected": -91.33682250976562, "logps/rejected": -770.2830810546875, "loss": 0.8685, "margin_dpo/margin_mean": 239.4127960205078, "margin_dpo/margin_std": 250.37957763671875, "step": 571 }, { "KL/chosen_KL_mean": -467.0171203613281, "KL/mean": -577.5743408203125, "KL/rejected_KL_mean": -688.1314086914062, "KL/std": 266.3624572753906, "epoch": 0.8399412628487518, "fcm_dpo/beta": 0.003590911626815796, "fcm_dpo/delta": 0.006393725052475929, "fcm_dpo/margin": 221.11422729492188, "fcm_dpo/q_t": 0.34955984354019165, "grad_norm": 40.90983200073242, "learning_rate": 3.880806698864086e-08, "logits/chosen": 0.4752965569496155, "logits/rejected": 0.4910871684551239, "logps/chosen": -515.4764404296875, "logps/ref_chosen": -48.45928955078125, "logps/ref_rejected": -83.55703735351562, "logps/rejected": -771.6884155273438, "loss": 0.9482, "margin_dpo/margin_mean": 221.11422729492188, "margin_dpo/margin_std": 291.041748046875, "step": 572 }, { "KL/chosen_KL_mean": -441.55621337890625, "KL/mean": -551.5286254882812, "KL/rejected_KL_mean": -661.5009765625, "KL/std": 213.51548767089844, "epoch": 0.8414096916299559, "fcm_dpo/beta": 0.0036352514289319515, "fcm_dpo/delta": -0.0013968870043754578, "fcm_dpo/margin": 219.9447784423828, "fcm_dpo/q_t": 0.3341352939605713, "grad_norm": 35.48988342285156, "learning_rate": 3.812409996461275e-08, "logits/chosen": 0.468715637922287, "logits/rejected": 0.4946897029876709, "logps/chosen": -493.1788330078125, "logps/ref_chosen": -51.62262725830078, "logps/ref_rejected": -85.32499694824219, "logps/rejected": -746.8259887695312, "loss": 0.8768, "margin_dpo/margin_mean": 219.94479370117188, "margin_dpo/margin_std": 218.58367919921875, "step": 573 }, { "KL/chosen_KL_mean": -424.98175048828125, "KL/mean": -547.060546875, "KL/rejected_KL_mean": -669.1392822265625, "KL/std": 229.44229125976562, "epoch": 0.8428781204111601, "fcm_dpo/beta": 0.0035688234493136406, "fcm_dpo/delta": -0.07810860127210617, "fcm_dpo/margin": 244.15753173828125, "fcm_dpo/q_t": 0.3205257058143616, "grad_norm": 35.287593841552734, "learning_rate": 3.74457160675965e-08, "logits/chosen": 0.33639222383499146, "logits/rejected": 0.35209882259368896, "logps/chosen": -476.0262145996094, "logps/ref_chosen": -51.04446029663086, "logps/ref_rejected": -92.80640411376953, "logps/rejected": -761.9456787109375, "loss": 0.8361, "margin_dpo/margin_mean": 244.15753173828125, "margin_dpo/margin_std": 235.37338256835938, "step": 574 }, { "KL/chosen_KL_mean": -473.6583557128906, "KL/mean": -580.267578125, "KL/rejected_KL_mean": -686.8767700195312, "KL/std": 237.57046508789062, "epoch": 0.8443465491923642, "fcm_dpo/beta": 0.0035233558155596256, "fcm_dpo/delta": 0.04960758984088898, "fcm_dpo/margin": 213.21844482421875, "fcm_dpo/q_t": 0.3446178436279297, "grad_norm": 44.265228271484375, "learning_rate": 3.677293317363864e-08, "logits/chosen": 0.35401681065559387, "logits/rejected": 0.3639416992664337, "logps/chosen": -545.448486328125, "logps/ref_chosen": -71.7901382446289, "logps/ref_rejected": -95.38619995117188, "logps/rejected": -782.262939453125, "loss": 0.9677, "margin_dpo/margin_mean": 213.2184295654297, "margin_dpo/margin_std": 257.09466552734375, "step": 575 }, { "KL/chosen_KL_mean": -453.58380126953125, "KL/mean": -559.41845703125, "KL/rejected_KL_mean": -665.2530517578125, "KL/std": 233.61068725585938, "epoch": 0.8458149779735683, "fcm_dpo/beta": 0.003570101223886013, "fcm_dpo/delta": 0.04643898457288742, "fcm_dpo/margin": 211.66921997070312, "fcm_dpo/q_t": 0.3475932478904724, "grad_norm": 49.50276565551758, "learning_rate": 3.6105769011194224e-08, "logits/chosen": 0.4114004373550415, "logits/rejected": 0.4138645529747009, "logps/chosen": -507.8467712402344, "logps/ref_chosen": -54.262962341308594, "logps/ref_rejected": -100.75428009033203, "logps/rejected": -766.00732421875, "loss": 0.9458, "margin_dpo/margin_mean": 211.66921997070312, "margin_dpo/margin_std": 250.59912109375, "step": 576 }, { "KL/chosen_KL_mean": -411.8608703613281, "KL/mean": -526.9452514648438, "KL/rejected_KL_mean": -642.0296630859375, "KL/std": 223.98553466796875, "epoch": 0.8472834067547724, "fcm_dpo/beta": 0.0035911118611693382, "fcm_dpo/delta": -0.029081393033266068, "fcm_dpo/margin": 230.16879272460938, "fcm_dpo/q_t": 0.3222889304161072, "grad_norm": 39.094970703125, "learning_rate": 3.5444241160659304e-08, "logits/chosen": 0.36829674243927, "logits/rejected": 0.39512234926223755, "logps/chosen": -473.77056884765625, "logps/ref_chosen": -61.909706115722656, "logps/ref_rejected": -84.07069396972656, "logps/rejected": -726.100341796875, "loss": 0.8506, "margin_dpo/margin_mean": 230.16879272460938, "margin_dpo/margin_std": 211.40219116210938, "step": 577 }, { "KL/chosen_KL_mean": -424.4698486328125, "KL/mean": -533.7640380859375, "KL/rejected_KL_mean": -643.0581665039062, "KL/std": 227.083251953125, "epoch": 0.8487518355359766, "fcm_dpo/beta": 0.0035750193055719137, "fcm_dpo/delta": 0.019029080867767334, "fcm_dpo/margin": 218.58831787109375, "fcm_dpo/q_t": 0.3325645327568054, "grad_norm": 35.93294143676758, "learning_rate": 3.478836705390808e-08, "logits/chosen": 0.313589870929718, "logits/rejected": 0.33140918612480164, "logps/chosen": -473.7335205078125, "logps/ref_chosen": -49.26368713378906, "logps/ref_rejected": -83.4362564086914, "logps/rejected": -726.494384765625, "loss": 0.859, "margin_dpo/margin_mean": 218.58831787109375, "margin_dpo/margin_std": 195.58734130859375, "step": 578 }, { "KL/chosen_KL_mean": -435.37188720703125, "KL/mean": -529.50390625, "KL/rejected_KL_mean": -623.6358642578125, "KL/std": 218.8971405029297, "epoch": 0.8502202643171806, "fcm_dpo/beta": 0.0036643778439611197, "fcm_dpo/delta": 0.11842280626296997, "fcm_dpo/margin": 188.2639617919922, "fcm_dpo/q_t": 0.3539060950279236, "grad_norm": 43.979103088378906, "learning_rate": 3.41381639738331e-08, "logits/chosen": 0.35747650265693665, "logits/rejected": 0.36031123995780945, "logps/chosen": -494.2576904296875, "logps/ref_chosen": -58.88581848144531, "logps/ref_rejected": -94.78762817382812, "logps/rejected": -718.4234619140625, "loss": 0.9431, "margin_dpo/margin_mean": 188.26394653320312, "margin_dpo/margin_std": 212.978271484375, "step": 579 }, { "KL/chosen_KL_mean": -360.24114990234375, "KL/mean": -488.2108154296875, "KL/rejected_KL_mean": -616.1805419921875, "KL/std": 242.11322021484375, "epoch": 0.8516886930983847, "fcm_dpo/beta": 0.0035925463307648897, "fcm_dpo/delta": -0.13383767008781433, "fcm_dpo/margin": 255.93939208984375, "fcm_dpo/q_t": 0.3108557462692261, "grad_norm": 30.20220375061035, "learning_rate": 3.349364905389032e-08, "logits/chosen": 0.23727968335151672, "logits/rejected": 0.25300416350364685, "logps/chosen": -408.9479675292969, "logps/ref_chosen": -48.70683670043945, "logps/ref_rejected": -81.7583999633789, "logps/rejected": -697.93896484375, "loss": 0.8221, "margin_dpo/margin_mean": 255.9394073486328, "margin_dpo/margin_std": 249.52294921875, "step": 580 }, { "KL/chosen_KL_mean": -450.68023681640625, "KL/mean": -544.3995361328125, "KL/rejected_KL_mean": -638.118896484375, "KL/std": 226.23956298828125, "epoch": 0.8531571218795888, "fcm_dpo/beta": 0.003664699150249362, "fcm_dpo/delta": 0.12125371396541595, "fcm_dpo/margin": 187.4386444091797, "fcm_dpo/q_t": 0.35620471835136414, "grad_norm": 39.6314582824707, "learning_rate": 3.285483927764726e-08, "logits/chosen": 0.4068590998649597, "logits/rejected": 0.4267165958881378, "logps/chosen": -512.902587890625, "logps/ref_chosen": -62.22235107421875, "logps/ref_rejected": -91.73568725585938, "logps/rejected": -729.8546142578125, "loss": 0.9458, "margin_dpo/margin_mean": 187.43865966796875, "margin_dpo/margin_std": 216.36355590820312, "step": 581 }, { "KL/chosen_KL_mean": -409.5931396484375, "KL/mean": -532.977294921875, "KL/rejected_KL_mean": -656.3615112304688, "KL/std": 232.1181182861328, "epoch": 0.8546255506607929, "fcm_dpo/beta": 0.0036610260140150785, "fcm_dpo/delta": -0.11489932984113693, "fcm_dpo/margin": 246.76834106445312, "fcm_dpo/q_t": 0.3156302571296692, "grad_norm": 32.382896423339844, "learning_rate": 3.222175147833556e-08, "logits/chosen": 0.37507086992263794, "logits/rejected": 0.3699257969856262, "logps/chosen": -467.82183837890625, "logps/ref_chosen": -58.228660583496094, "logps/ref_rejected": -110.06959533691406, "logps/rejected": -766.4310913085938, "loss": 0.8497, "margin_dpo/margin_mean": 246.76832580566406, "margin_dpo/margin_std": 242.81654357910156, "step": 582 }, { "KL/chosen_KL_mean": -448.87646484375, "KL/mean": -538.7371215820312, "KL/rejected_KL_mean": -628.5977783203125, "KL/std": 230.92794799804688, "epoch": 0.856093979441997, "fcm_dpo/beta": 0.0036516801919788122, "fcm_dpo/delta": 0.15361306071281433, "fcm_dpo/margin": 179.72134399414062, "fcm_dpo/q_t": 0.3652193546295166, "grad_norm": 43.001060485839844, "learning_rate": 3.159440233840763e-08, "logits/chosen": 0.3155054450035095, "logits/rejected": 0.3352203071117401, "logps/chosen": -505.73931884765625, "logps/ref_chosen": -56.86286163330078, "logps/ref_rejected": -88.4039306640625, "logps/rejected": -717.001708984375, "loss": 0.9814, "margin_dpo/margin_mean": 179.72134399414062, "margin_dpo/margin_std": 224.05186462402344, "step": 583 }, { "KL/chosen_KL_mean": -398.5378112792969, "KL/mean": -541.896240234375, "KL/rejected_KL_mean": -685.254638671875, "KL/std": 245.92782592773438, "epoch": 0.8575624082232012, "fcm_dpo/beta": 0.0035745359491556883, "fcm_dpo/delta": -0.2507776916027069, "fcm_dpo/margin": 286.71685791015625, "fcm_dpo/q_t": 0.2880155146121979, "grad_norm": 34.256526947021484, "learning_rate": 3.0972808389096635e-08, "logits/chosen": 0.39042460918426514, "logits/rejected": 0.42286020517349243, "logps/chosen": -455.4384765625, "logps/ref_chosen": -56.90068054199219, "logps/ref_rejected": -97.63606262207031, "logps/rejected": -782.8907470703125, "loss": 0.7289, "margin_dpo/margin_mean": 286.71685791015625, "margin_dpo/margin_std": 226.08956909179688, "step": 584 }, { "KL/chosen_KL_mean": -460.065185546875, "KL/mean": -570.1961669921875, "KL/rejected_KL_mean": -680.3271484375, "KL/std": 234.60592651367188, "epoch": 0.8590308370044053, "fcm_dpo/beta": 0.0035448740236461163, "fcm_dpo/delta": 0.02077949233353138, "fcm_dpo/margin": 220.26193237304688, "fcm_dpo/q_t": 0.3401145339012146, "grad_norm": 36.705047607421875, "learning_rate": 3.035698600998121e-08, "logits/chosen": 0.3827165961265564, "logits/rejected": 0.4175301790237427, "logps/chosen": -521.0391845703125, "logps/ref_chosen": -60.973968505859375, "logps/ref_rejected": -84.16952514648438, "logps/rejected": -764.4967041015625, "loss": 0.9029, "margin_dpo/margin_mean": 220.26193237304688, "margin_dpo/margin_std": 237.73251342773438, "step": 585 }, { "KL/chosen_KL_mean": -461.75762939453125, "KL/mean": -558.2662353515625, "KL/rejected_KL_mean": -654.7747802734375, "KL/std": 208.19444274902344, "epoch": 0.8604992657856094, "fcm_dpo/beta": 0.0036112640518695116, "fcm_dpo/delta": 0.11033973097801208, "fcm_dpo/margin": 193.01708984375, "fcm_dpo/q_t": 0.3578340709209442, "grad_norm": 36.1845588684082, "learning_rate": 2.974695142855388e-08, "logits/chosen": 0.41748034954071045, "logits/rejected": 0.4262012243270874, "logps/chosen": -518.61328125, "logps/ref_chosen": -56.85559844970703, "logps/ref_rejected": -91.80261993408203, "logps/rejected": -746.577392578125, "loss": 0.971, "margin_dpo/margin_mean": 193.01708984375, "margin_dpo/margin_std": 244.68115234375, "step": 586 }, { "KL/chosen_KL_mean": -383.22210693359375, "KL/mean": -492.18524169921875, "KL/rejected_KL_mean": -601.1483154296875, "KL/std": 240.5768585205078, "epoch": 0.8619676945668135, "fcm_dpo/beta": 0.003636622568592429, "fcm_dpo/delta": 0.008001847192645073, "fcm_dpo/margin": 217.92623901367188, "fcm_dpo/q_t": 0.33623430132865906, "grad_norm": 48.267826080322266, "learning_rate": 2.9142720719793122e-08, "logits/chosen": 0.463498592376709, "logits/rejected": 0.46634483337402344, "logps/chosen": -427.9136962890625, "logps/ref_chosen": -44.69159698486328, "logps/ref_rejected": -82.62385559082031, "logps/rejected": -683.772216796875, "loss": 0.8911, "margin_dpo/margin_mean": 217.9262237548828, "margin_dpo/margin_std": 232.04367065429688, "step": 587 }, { "KL/chosen_KL_mean": -446.12030029296875, "KL/mean": -551.8760986328125, "KL/rejected_KL_mean": -657.6319580078125, "KL/std": 213.5641632080078, "epoch": 0.8634361233480177, "fcm_dpo/beta": 0.003631887724623084, "fcm_dpo/delta": 0.03419031202793121, "fcm_dpo/margin": 211.51162719726562, "fcm_dpo/q_t": 0.34210073947906494, "grad_norm": 43.596290588378906, "learning_rate": 2.8544309805740018e-08, "logits/chosen": 0.37401342391967773, "logits/rejected": 0.3757128119468689, "logps/chosen": -496.415283203125, "logps/ref_chosen": -50.29494857788086, "logps/ref_rejected": -107.36988067626953, "logps/rejected": -765.0018310546875, "loss": 0.899, "margin_dpo/margin_mean": 211.51162719726562, "margin_dpo/margin_std": 227.99298095703125, "step": 588 }, { "KL/chosen_KL_mean": -417.7206726074219, "KL/mean": -556.0487670898438, "KL/rejected_KL_mean": -694.3768920898438, "KL/std": 251.08468627929688, "epoch": 0.8649045521292217, "fcm_dpo/beta": 0.003547674510627985, "fcm_dpo/delta": -0.20126084983348846, "fcm_dpo/margin": 276.6562194824219, "fcm_dpo/q_t": 0.29785215854644775, "grad_norm": 36.69344711303711, "learning_rate": 2.7951734455078786e-08, "logits/chosen": 0.36783045530319214, "logits/rejected": 0.3880786597728729, "logps/chosen": -477.65057373046875, "logps/ref_chosen": -59.929908752441406, "logps/ref_rejected": -111.65534973144531, "logps/rejected": -806.0322265625, "loss": 0.7685, "margin_dpo/margin_mean": 276.6562194824219, "margin_dpo/margin_std": 234.46910095214844, "step": 589 }, { "KL/chosen_KL_mean": -402.7647705078125, "KL/mean": -522.7323608398438, "KL/rejected_KL_mean": -642.699951171875, "KL/std": 220.0315399169922, "epoch": 0.8663729809104258, "fcm_dpo/beta": 0.003481434192508459, "fcm_dpo/delta": -0.03877948969602585, "fcm_dpo/margin": 239.93521118164062, "fcm_dpo/q_t": 0.32213735580444336, "grad_norm": 31.813249588012695, "learning_rate": 2.736501028272095e-08, "logits/chosen": 0.2651674449443817, "logits/rejected": 0.26474422216415405, "logps/chosen": -458.5745849609375, "logps/ref_chosen": -55.80979537963867, "logps/ref_rejected": -106.06282043457031, "logps/rejected": -748.7627563476562, "loss": 0.8322, "margin_dpo/margin_mean": 239.93521118164062, "margin_dpo/margin_std": 213.3310546875, "step": 590 }, { "KL/chosen_KL_mean": -433.9034729003906, "KL/mean": -554.28173828125, "KL/rejected_KL_mean": -674.659912109375, "KL/std": 218.40057373046875, "epoch": 0.8678414096916299, "fcm_dpo/beta": 0.0034645451232790947, "fcm_dpo/delta": -0.0372135229408741, "fcm_dpo/margin": 240.75643920898438, "fcm_dpo/q_t": 0.32035619020462036, "grad_norm": 36.15827560424805, "learning_rate": 2.678415274939408e-08, "logits/chosen": 0.47813642024993896, "logits/rejected": 0.5184578895568848, "logps/chosen": -490.14410400390625, "logps/ref_chosen": -56.24061965942383, "logps/ref_rejected": -83.78629302978516, "logps/rejected": -758.4462280273438, "loss": 0.8311, "margin_dpo/margin_mean": 240.75643920898438, "margin_dpo/margin_std": 211.91455078125, "step": 591 }, { "KL/chosen_KL_mean": -464.30401611328125, "KL/mean": -573.9903564453125, "KL/rejected_KL_mean": -683.6767578125, "KL/std": 229.73587036132812, "epoch": 0.869309838472834, "fcm_dpo/beta": 0.003464614972472191, "fcm_dpo/delta": 0.043035559356212616, "fcm_dpo/margin": 219.3727569580078, "fcm_dpo/q_t": 0.3421498239040375, "grad_norm": 41.69553756713867, "learning_rate": 2.6209177161234442e-08, "logits/chosen": 0.5036705732345581, "logits/rejected": 0.5235416889190674, "logps/chosen": -512.2442626953125, "logps/ref_chosen": -47.94025421142578, "logps/ref_rejected": -75.73287963867188, "logps/rejected": -759.40966796875, "loss": 0.9679, "margin_dpo/margin_mean": 219.37274169921875, "margin_dpo/margin_std": 278.5716552734375, "step": 592 }, { "KL/chosen_KL_mean": -416.8182373046875, "KL/mean": -500.542236328125, "KL/rejected_KL_mean": -584.2662353515625, "KL/std": 228.785400390625, "epoch": 0.8707782672540382, "fcm_dpo/beta": 0.0035965959541499615, "fcm_dpo/delta": 0.21053200960159302, "fcm_dpo/margin": 167.44802856445312, "fcm_dpo/q_t": 0.3748427629470825, "grad_norm": 50.23680114746094, "learning_rate": 2.564009866938349e-08, "logits/chosen": 0.31660762429237366, "logits/rejected": 0.3367197811603546, "logps/chosen": -465.50897216796875, "logps/ref_chosen": -48.690757751464844, "logps/ref_rejected": -60.90800094604492, "logps/rejected": -645.1742553710938, "loss": 1.0208, "margin_dpo/margin_mean": 167.44802856445312, "margin_dpo/margin_std": 228.37478637695312, "step": 593 }, { "KL/chosen_KL_mean": -397.3114013671875, "KL/mean": -511.14520263671875, "KL/rejected_KL_mean": -624.97900390625, "KL/std": 272.0648193359375, "epoch": 0.8722466960352423, "fcm_dpo/beta": 0.0036704763770103455, "fcm_dpo/delta": -0.04310472309589386, "fcm_dpo/margin": 227.66758728027344, "fcm_dpo/q_t": 0.3338211178779602, "grad_norm": 39.49517059326172, "learning_rate": 2.5076932269588708e-08, "logits/chosen": 0.3315795361995697, "logits/rejected": 0.3542073965072632, "logps/chosen": -452.2463073730469, "logps/ref_chosen": -54.93488693237305, "logps/ref_rejected": -86.09967803955078, "logps/rejected": -711.0787353515625, "loss": 0.9104, "margin_dpo/margin_mean": 227.6676025390625, "margin_dpo/margin_std": 253.65826416015625, "step": 594 }, { "KL/chosen_KL_mean": -421.3223876953125, "KL/mean": -522.7353515625, "KL/rejected_KL_mean": -624.1483154296875, "KL/std": 207.23240661621094, "epoch": 0.8737151248164464, "fcm_dpo/beta": 0.003634433727711439, "fcm_dpo/delta": 0.0678904801607132, "fcm_dpo/margin": 202.82586669921875, "fcm_dpo/q_t": 0.34680286049842834, "grad_norm": 36.4913444519043, "learning_rate": 2.451969280180849e-08, "logits/chosen": 0.37026911973953247, "logits/rejected": 0.39209526777267456, "logps/chosen": -470.7427978515625, "logps/ref_chosen": -49.4204216003418, "logps/ref_rejected": -80.62731170654297, "logps/rejected": -704.7755737304688, "loss": 0.9031, "margin_dpo/margin_mean": 202.82586669921875, "margin_dpo/margin_std": 217.4996337890625, "step": 595 }, { "KL/chosen_KL_mean": -426.61724853515625, "KL/mean": -521.0478515625, "KL/rejected_KL_mean": -615.478515625, "KL/std": 226.96453857421875, "epoch": 0.8751835535976505, "fcm_dpo/beta": 0.003726873081177473, "fcm_dpo/delta": 0.10264364629983902, "fcm_dpo/margin": 188.86123657226562, "fcm_dpo/q_t": 0.3537423014640808, "grad_norm": 52.47248077392578, "learning_rate": 2.396839494982103e-08, "logits/chosen": 0.35756534337997437, "logits/rejected": 0.38035476207733154, "logps/chosen": -486.408935546875, "logps/ref_chosen": -59.791683197021484, "logps/ref_rejected": -80.09111785888672, "logps/rejected": -695.569580078125, "loss": 0.9556, "margin_dpo/margin_mean": 188.8612518310547, "margin_dpo/margin_std": 228.7149200439453, "step": 596 }, { "KL/chosen_KL_mean": -430.0118713378906, "KL/mean": -549.8414916992188, "KL/rejected_KL_mean": -669.671142578125, "KL/std": 246.92037963867188, "epoch": 0.8766519823788547, "fcm_dpo/beta": 0.003649506950750947, "fcm_dpo/delta": -0.0845845490694046, "fcm_dpo/margin": 239.65924072265625, "fcm_dpo/q_t": 0.3215242922306061, "grad_norm": 36.12924575805664, "learning_rate": 2.3423053240837514e-08, "logits/chosen": 0.2778972089290619, "logits/rejected": 0.2754002809524536, "logps/chosen": -487.27264404296875, "logps/ref_chosen": -57.26078796386719, "logps/ref_rejected": -100.6937255859375, "logps/rejected": -770.3648681640625, "loss": 0.8523, "margin_dpo/margin_mean": 239.65924072265625, "margin_dpo/margin_std": 241.89382934570312, "step": 597 }, { "KL/chosen_KL_mean": -445.42364501953125, "KL/mean": -549.445068359375, "KL/rejected_KL_mean": -653.4664916992188, "KL/std": 218.0029754638672, "epoch": 0.8781204111600588, "fcm_dpo/beta": 0.0036473223008215427, "fcm_dpo/delta": 0.042003557085990906, "fcm_dpo/margin": 208.04281616210938, "fcm_dpo/q_t": 0.33893561363220215, "grad_norm": 56.39809036254883, "learning_rate": 2.2883682045119062e-08, "logits/chosen": 0.4461902976036072, "logits/rejected": 0.46808603405952454, "logps/chosen": -497.9421691894531, "logps/ref_chosen": -52.51850509643555, "logps/ref_rejected": -89.44385528564453, "logps/rejected": -742.9103393554688, "loss": 0.9324, "margin_dpo/margin_mean": 208.04281616210938, "margin_dpo/margin_std": 228.42031860351562, "step": 598 }, { "KL/chosen_KL_mean": -409.59747314453125, "KL/mean": -512.0952758789062, "KL/rejected_KL_mean": -614.5930786132812, "KL/std": 232.806884765625, "epoch": 0.8795888399412628, "fcm_dpo/beta": 0.0037043686024844646, "fcm_dpo/delta": 0.043617166578769684, "fcm_dpo/margin": 204.99563598632812, "fcm_dpo/q_t": 0.33468934893608093, "grad_norm": 44.148353576660156, "learning_rate": 2.2350295575598367e-08, "logits/chosen": 0.3408563733100891, "logits/rejected": 0.3492530882358551, "logps/chosen": -459.400146484375, "logps/ref_chosen": -49.802677154541016, "logps/ref_rejected": -82.978515625, "logps/rejected": -697.5715942382812, "loss": 0.8645, "margin_dpo/margin_mean": 204.99563598632812, "margin_dpo/margin_std": 180.45458984375, "step": 599 }, { "KL/chosen_KL_mean": -465.63177490234375, "KL/mean": -566.748779296875, "KL/rejected_KL_mean": -667.86572265625, "KL/std": 217.93067932128906, "epoch": 0.8810572687224669, "fcm_dpo/beta": 0.0037527671083807945, "fcm_dpo/delta": 0.044459305703639984, "fcm_dpo/margin": 202.23391723632812, "fcm_dpo/q_t": 0.3429974317550659, "grad_norm": 52.01202392578125, "learning_rate": 2.1822907887504932e-08, "logits/chosen": 0.39914214611053467, "logits/rejected": 0.41855812072753906, "logps/chosen": -532.066650390625, "logps/ref_chosen": -66.43487548828125, "logps/ref_rejected": -85.45649719238281, "logps/rejected": -753.3222045898438, "loss": 0.9215, "margin_dpo/margin_mean": 202.2339324951172, "margin_dpo/margin_std": 229.5970458984375, "step": 600 }, { "KL/chosen_KL_mean": -471.54205322265625, "KL/mean": -581.1104736328125, "KL/rejected_KL_mean": -690.678955078125, "KL/std": 239.7049102783203, "epoch": 0.882525697503671, "fcm_dpo/beta": 0.0037345641758292913, "fcm_dpo/delta": -0.02068949118256569, "fcm_dpo/margin": 219.13687133789062, "fcm_dpo/q_t": 0.33221566677093506, "grad_norm": 43.92999267578125, "learning_rate": 2.1301532877994742e-08, "logits/chosen": 0.35876715183258057, "logits/rejected": 0.3796877861022949, "logps/chosen": -530.6756591796875, "logps/ref_chosen": -59.13361358642578, "logps/ref_rejected": -94.69093322753906, "logps/rejected": -785.369873046875, "loss": 0.8863, "margin_dpo/margin_mean": 219.13687133789062, "margin_dpo/margin_std": 236.77896118164062, "step": 601 }, { "KL/chosen_KL_mean": -375.32904052734375, "KL/mean": -511.0220947265625, "KL/rejected_KL_mean": -646.715087890625, "KL/std": 236.4164581298828, "epoch": 0.8839941262848752, "fcm_dpo/beta": 0.0036473888903856277, "fcm_dpo/delta": -0.2102523148059845, "fcm_dpo/margin": 271.3861083984375, "fcm_dpo/q_t": 0.2962424159049988, "grad_norm": 41.782928466796875, "learning_rate": 2.0786184285784298e-08, "logits/chosen": 0.37781381607055664, "logits/rejected": 0.3983898162841797, "logps/chosen": -423.92254638671875, "logps/ref_chosen": -48.59352111816406, "logps/ref_rejected": -87.6685562133789, "logps/rejected": -734.3836669921875, "loss": 0.7675, "margin_dpo/margin_mean": 271.3861083984375, "margin_dpo/margin_std": 233.6563262939453, "step": 602 }, { "KL/chosen_KL_mean": -410.86407470703125, "KL/mean": -521.83203125, "KL/rejected_KL_mean": -632.7999877929688, "KL/std": 244.81134033203125, "epoch": 0.8854625550660793, "fcm_dpo/beta": 0.0036042386200278997, "fcm_dpo/delta": -2.483464777469635e-05, "fcm_dpo/margin": 221.93588256835938, "fcm_dpo/q_t": 0.33811676502227783, "grad_norm": 44.7195930480957, "learning_rate": 2.0276875690788204e-08, "logits/chosen": 0.28510865569114685, "logits/rejected": 0.3141903281211853, "logps/chosen": -481.2786865234375, "logps/ref_chosen": -70.41461944580078, "logps/ref_rejected": -100.32559967041016, "logps/rejected": -733.1256103515625, "loss": 0.8952, "margin_dpo/margin_mean": 221.93589782714844, "margin_dpo/margin_std": 245.03948974609375, "step": 603 }, { "KL/chosen_KL_mean": -377.646240234375, "KL/mean": -506.044677734375, "KL/rejected_KL_mean": -634.4429931640625, "KL/std": 218.4007568359375, "epoch": 0.8869309838472834, "fcm_dpo/beta": 0.0035292627289891243, "fcm_dpo/delta": -0.11708834767341614, "fcm_dpo/margin": 256.7967834472656, "fcm_dpo/q_t": 0.30936452746391296, "grad_norm": 39.38010787963867, "learning_rate": 1.977362051376158e-08, "logits/chosen": 0.3584352433681488, "logits/rejected": 0.3700353503227234, "logps/chosen": -424.1043395996094, "logps/ref_chosen": -46.45808029174805, "logps/ref_rejected": -91.8544921875, "logps/rejected": -726.2974853515625, "loss": 0.824, "margin_dpo/margin_mean": 256.7967834472656, "margin_dpo/margin_std": 237.95303344726562, "step": 604 }, { "KL/chosen_KL_mean": -433.8162841796875, "KL/mean": -544.268310546875, "KL/rejected_KL_mean": -654.7203369140625, "KL/std": 223.2996063232422, "epoch": 0.8883994126284875, "fcm_dpo/beta": 0.0035091633908450603, "fcm_dpo/delta": 0.026538338512182236, "fcm_dpo/margin": 220.90402221679688, "fcm_dpo/q_t": 0.34067851305007935, "grad_norm": 29.942068099975586, "learning_rate": 1.9276432015946446e-08, "logits/chosen": 0.30465924739837646, "logits/rejected": 0.3171878755092621, "logps/chosen": -500.06561279296875, "logps/ref_chosen": -66.24933624267578, "logps/ref_rejected": -102.30496978759766, "logps/rejected": -757.0252685546875, "loss": 0.8932, "margin_dpo/margin_mean": 220.90403747558594, "margin_dpo/margin_std": 245.03860473632812, "step": 605 }, { "KL/chosen_KL_mean": -403.8315734863281, "KL/mean": -526.64990234375, "KL/rejected_KL_mean": -649.4683837890625, "KL/std": 218.93609619140625, "epoch": 0.8898678414096917, "fcm_dpo/beta": 0.0035280329175293446, "fcm_dpo/delta": -0.07430367916822433, "fcm_dpo/margin": 245.6367645263672, "fcm_dpo/q_t": 0.3186294436454773, "grad_norm": 45.23887252807617, "learning_rate": 1.8785323298722093e-08, "logits/chosen": 0.354251503944397, "logits/rejected": 0.3615856170654297, "logps/chosen": -458.65069580078125, "logps/ref_chosen": -54.819122314453125, "logps/ref_rejected": -98.37146759033203, "logps/rejected": -747.83984375, "loss": 0.8256, "margin_dpo/margin_mean": 245.6367645263672, "margin_dpo/margin_std": 220.42266845703125, "step": 606 }, { "KL/chosen_KL_mean": -425.9730224609375, "KL/mean": -522.0699462890625, "KL/rejected_KL_mean": -618.1668090820312, "KL/std": 217.41937255859375, "epoch": 0.8913362701908958, "fcm_dpo/beta": 0.003552454523742199, "fcm_dpo/delta": 0.1254272311925888, "fcm_dpo/margin": 192.19375610351562, "fcm_dpo/q_t": 0.35280275344848633, "grad_norm": 35.883914947509766, "learning_rate": 1.8300307303259904e-08, "logits/chosen": 0.3167957663536072, "logits/rejected": 0.3372369408607483, "logps/chosen": -484.05706787109375, "logps/ref_chosen": -58.08403778076172, "logps/ref_rejected": -79.777099609375, "logps/rejected": -697.9439086914062, "loss": 0.9203, "margin_dpo/margin_mean": 192.19375610351562, "margin_dpo/margin_std": 190.5302734375, "step": 607 }, { "KL/chosen_KL_mean": -415.585693359375, "KL/mean": -521.9092407226562, "KL/rejected_KL_mean": -628.2327880859375, "KL/std": 218.6051025390625, "epoch": 0.8928046989720999, "fcm_dpo/beta": 0.003595003392547369, "fcm_dpo/delta": 0.03830421715974808, "fcm_dpo/margin": 212.6471405029297, "fcm_dpo/q_t": 0.3348754942417145, "grad_norm": 34.83126449584961, "learning_rate": 1.7821396810182437e-08, "logits/chosen": 0.30306464433670044, "logits/rejected": 0.33365947008132935, "logps/chosen": -473.0364990234375, "logps/ref_chosen": -57.450836181640625, "logps/ref_rejected": -94.77339172363281, "logps/rejected": -723.0062255859375, "loss": 0.8616, "margin_dpo/margin_mean": 212.6471405029297, "margin_dpo/margin_std": 190.2625732421875, "step": 608 }, { "KL/chosen_KL_mean": -408.29541015625, "KL/mean": -530.133056640625, "KL/rejected_KL_mean": -651.970703125, "KL/std": 240.07913208007812, "epoch": 0.8942731277533039, "fcm_dpo/beta": 0.0035578757524490356, "fcm_dpo/delta": -0.07329061627388, "fcm_dpo/margin": 243.67532348632812, "fcm_dpo/q_t": 0.31681567430496216, "grad_norm": 38.81436538696289, "learning_rate": 1.7348604439226617e-08, "logits/chosen": 0.3799295425415039, "logits/rejected": 0.4069666266441345, "logps/chosen": -467.10076904296875, "logps/ref_chosen": -58.805355072021484, "logps/ref_rejected": -88.81600952148438, "logps/rejected": -740.7867431640625, "loss": 0.8195, "margin_dpo/margin_mean": 243.67532348632812, "margin_dpo/margin_std": 222.36453247070312, "step": 609 }, { "KL/chosen_KL_mean": -399.6932373046875, "KL/mean": -503.40771484375, "KL/rejected_KL_mean": -607.1221923828125, "KL/std": 210.8145751953125, "epoch": 0.895741556534508, "fcm_dpo/beta": 0.0035813814029097557, "fcm_dpo/delta": 0.06151764839887619, "fcm_dpo/margin": 207.42897033691406, "fcm_dpo/q_t": 0.3418347239494324, "grad_norm": 38.237361907958984, "learning_rate": 1.6881942648911074e-08, "logits/chosen": 0.3074260950088501, "logits/rejected": 0.3430824875831604, "logps/chosen": -465.3882751464844, "logps/ref_chosen": -65.69503784179688, "logps/ref_rejected": -83.40538787841797, "logps/rejected": -690.527587890625, "loss": 0.8948, "margin_dpo/margin_mean": 207.42898559570312, "margin_dpo/margin_std": 206.7069549560547, "step": 610 }, { "KL/chosen_KL_mean": -426.5433044433594, "KL/mean": -542.793701171875, "KL/rejected_KL_mean": -659.0440673828125, "KL/std": 228.12405395507812, "epoch": 0.8972099853157122, "fcm_dpo/beta": 0.003594658337533474, "fcm_dpo/delta": -0.04010578989982605, "fcm_dpo/margin": 232.50071716308594, "fcm_dpo/q_t": 0.3268967270851135, "grad_norm": 36.26604461669922, "learning_rate": 1.6421423736208e-08, "logits/chosen": 0.3895799517631531, "logits/rejected": 0.4121781587600708, "logps/chosen": -479.14276123046875, "logps/ref_chosen": -52.59946823120117, "logps/ref_rejected": -86.33099365234375, "logps/rejected": -745.375, "loss": 0.8602, "margin_dpo/margin_mean": 232.50071716308594, "margin_dpo/margin_std": 227.5194854736328, "step": 611 }, { "KL/chosen_KL_mean": -422.2787780761719, "KL/mean": -545.01708984375, "KL/rejected_KL_mean": -667.7554321289062, "KL/std": 208.62973022460938, "epoch": 0.8986784140969163, "fcm_dpo/beta": 0.003525904379785061, "fcm_dpo/delta": -0.07167483866214752, "fcm_dpo/margin": 245.47662353515625, "fcm_dpo/q_t": 0.31289637088775635, "grad_norm": 41.34052658081055, "learning_rate": 1.5967059836219042e-08, "logits/chosen": 0.3888121247291565, "logits/rejected": 0.4153617024421692, "logps/chosen": -481.6025085449219, "logps/ref_chosen": -59.32372283935547, "logps/ref_rejected": -88.31239318847656, "logps/rejected": -756.0678100585938, "loss": 0.7975, "margin_dpo/margin_mean": 245.47662353515625, "margin_dpo/margin_std": 193.85101318359375, "step": 612 }, { "KL/chosen_KL_mean": -387.6116638183594, "KL/mean": -502.537353515625, "KL/rejected_KL_mean": -617.4630126953125, "KL/std": 222.54290771484375, "epoch": 0.9001468428781204, "fcm_dpo/beta": 0.0035072462633252144, "fcm_dpo/delta": -0.006700858939439058, "fcm_dpo/margin": 229.85130310058594, "fcm_dpo/q_t": 0.3251153230667114, "grad_norm": 40.078956604003906, "learning_rate": 1.551886292185553e-08, "logits/chosen": 0.2332531213760376, "logits/rejected": 0.23660428822040558, "logps/chosen": -447.34161376953125, "logps/ref_chosen": -59.72996520996094, "logps/ref_rejected": -105.10752868652344, "logps/rejected": -722.570556640625, "loss": 0.8226, "margin_dpo/margin_mean": 229.85128784179688, "margin_dpo/margin_std": 184.74441528320312, "step": 613 }, { "KL/chosen_KL_mean": -423.49017333984375, "KL/mean": -545.447021484375, "KL/rejected_KL_mean": -667.4038696289062, "KL/std": 228.83831787109375, "epoch": 0.9016152716593245, "fcm_dpo/beta": 0.003461125772446394, "fcm_dpo/delta": -0.04895541071891785, "fcm_dpo/margin": 243.9136962890625, "fcm_dpo/q_t": 0.3211820125579834, "grad_norm": 47.83527755737305, "learning_rate": 1.507684480352292e-08, "logits/chosen": 0.43596264719963074, "logits/rejected": 0.44089680910110474, "logps/chosen": -476.4291687011719, "logps/ref_chosen": -52.93898010253906, "logps/ref_rejected": -104.67938232421875, "logps/rejected": -772.083251953125, "loss": 0.8376, "margin_dpo/margin_mean": 243.91368103027344, "margin_dpo/margin_std": 221.93887329101562, "step": 614 }, { "KL/chosen_KL_mean": -399.7521667480469, "KL/mean": -496.95867919921875, "KL/rejected_KL_mean": -594.1651611328125, "KL/std": 232.88909912109375, "epoch": 0.9030837004405287, "fcm_dpo/beta": 0.0035465422552078962, "fcm_dpo/delta": 0.11760629713535309, "fcm_dpo/margin": 194.41302490234375, "fcm_dpo/q_t": 0.3538597524166107, "grad_norm": 35.42436599731445, "learning_rate": 1.4641017128809801e-08, "logits/chosen": 0.23408935964107513, "logits/rejected": 0.25060975551605225, "logps/chosen": -465.5694580078125, "logps/ref_chosen": -65.81727600097656, "logps/ref_rejected": -95.17749786376953, "logps/rejected": -689.3426513671875, "loss": 0.9149, "margin_dpo/margin_mean": 194.41302490234375, "margin_dpo/margin_std": 198.14556884765625, "step": 615 }, { "KL/chosen_KL_mean": -471.95989990234375, "KL/mean": -557.6803588867188, "KL/rejected_KL_mean": -643.40087890625, "KL/std": 202.3949432373047, "epoch": 0.9045521292217328, "fcm_dpo/beta": 0.0036708577536046505, "fcm_dpo/delta": 0.18005023896694183, "fcm_dpo/margin": 171.44093322753906, "fcm_dpo/q_t": 0.36842331290245056, "grad_norm": 45.34204864501953, "learning_rate": 1.4211391382180637e-08, "logits/chosen": 0.34676775336265564, "logits/rejected": 0.37802496552467346, "logps/chosen": -537.0927734375, "logps/ref_chosen": -65.13285827636719, "logps/ref_rejected": -74.70050048828125, "logps/rejected": -718.101318359375, "loss": 0.9901, "margin_dpo/margin_mean": 171.44094848632812, "margin_dpo/margin_std": 215.03761291503906, "step": 616 }, { "KL/chosen_KL_mean": -433.5853271484375, "KL/mean": -518.7471923828125, "KL/rejected_KL_mean": -603.9089965820312, "KL/std": 193.24517822265625, "epoch": 0.9060205580029369, "fcm_dpo/beta": 0.0037929839454591274, "fcm_dpo/delta": 0.16294321417808533, "fcm_dpo/margin": 170.32363891601562, "fcm_dpo/q_t": 0.3635261654853821, "grad_norm": 44.73681640625, "learning_rate": 1.378797888467345e-08, "logits/chosen": 0.30464887619018555, "logits/rejected": 0.329974889755249, "logps/chosen": -496.59088134765625, "logps/ref_chosen": -63.005550384521484, "logps/ref_rejected": -64.234130859375, "logps/rejected": -668.1431274414062, "loss": 0.9621, "margin_dpo/margin_mean": 170.32363891601562, "margin_dpo/margin_std": 197.97344970703125, "step": 617 }, { "KL/chosen_KL_mean": -440.02435302734375, "KL/mean": -560.755859375, "KL/rejected_KL_mean": -681.4873046875, "KL/std": 222.37545776367188, "epoch": 0.9074889867841409, "fcm_dpo/beta": 0.003762473352253437, "fcm_dpo/delta": -0.1195143535733223, "fcm_dpo/margin": 241.4630126953125, "fcm_dpo/q_t": 0.31400012969970703, "grad_norm": 43.647186279296875, "learning_rate": 1.3370790793601371e-08, "logits/chosen": 0.3098023533821106, "logits/rejected": 0.35203438997268677, "logps/chosen": -507.1257019042969, "logps/ref_chosen": -67.10134887695312, "logps/ref_rejected": -92.15340423583984, "logps/rejected": -773.6407470703125, "loss": 0.8398, "margin_dpo/margin_mean": 241.46299743652344, "margin_dpo/margin_std": 234.85427856445312, "step": 618 }, { "KL/chosen_KL_mean": -433.3082580566406, "KL/mean": -527.0829467773438, "KL/rejected_KL_mean": -620.8576049804688, "KL/std": 218.11892700195312, "epoch": 0.908957415565345, "fcm_dpo/beta": 0.003781287930905819, "fcm_dpo/delta": 0.09707266837358475, "fcm_dpo/margin": 187.5493621826172, "fcm_dpo/q_t": 0.35968726873397827, "grad_norm": 39.4860725402832, "learning_rate": 1.2959838102258535e-08, "logits/chosen": 0.35403043031692505, "logits/rejected": 0.3776950538158417, "logps/chosen": -489.2864990234375, "logps/ref_chosen": -55.978233337402344, "logps/ref_rejected": -93.1854019165039, "logps/rejected": -714.04296875, "loss": 0.9901, "margin_dpo/margin_mean": 187.5493621826172, "margin_dpo/margin_std": 255.41448974609375, "step": 619 }, { "KL/chosen_KL_mean": -406.3300476074219, "KL/mean": -509.80792236328125, "KL/rejected_KL_mean": -613.2857666015625, "KL/std": 210.70660400390625, "epoch": 0.9104258443465492, "fcm_dpo/beta": 0.0037673478946089745, "fcm_dpo/delta": 0.020875804126262665, "fcm_dpo/margin": 206.95571899414062, "fcm_dpo/q_t": 0.3405061960220337, "grad_norm": 50.14612579345703, "learning_rate": 1.2555131639630567e-08, "logits/chosen": 0.3730146288871765, "logits/rejected": 0.40417569875717163, "logps/chosen": -466.1275634765625, "logps/ref_chosen": -59.79750061035156, "logps/ref_rejected": -78.41075134277344, "logps/rejected": -691.696533203125, "loss": 0.9001, "margin_dpo/margin_mean": 206.95571899414062, "margin_dpo/margin_std": 227.1011505126953, "step": 620 }, { "KL/chosen_KL_mean": -416.3944091796875, "KL/mean": -542.18994140625, "KL/rejected_KL_mean": -667.9854736328125, "KL/std": 217.998291015625, "epoch": 0.9118942731277533, "fcm_dpo/beta": 0.003742808010429144, "fcm_dpo/delta": -0.15637120604515076, "fcm_dpo/margin": 251.5911102294922, "fcm_dpo/q_t": 0.3048107326030731, "grad_norm": 43.15443801879883, "learning_rate": 1.2156682070109086e-08, "logits/chosen": 0.47899624705314636, "logits/rejected": 0.49514925479888916, "logps/chosen": -470.3281555175781, "logps/ref_chosen": -53.93375778198242, "logps/ref_rejected": -88.36951446533203, "logps/rejected": -756.35498046875, "loss": 0.7993, "margin_dpo/margin_mean": 251.5911102294922, "margin_dpo/margin_std": 224.11578369140625, "step": 621 }, { "KL/chosen_KL_mean": -408.5030212402344, "KL/mean": -518.6573486328125, "KL/rejected_KL_mean": -628.8116455078125, "KL/std": 197.64805603027344, "epoch": 0.9133627019089574, "fcm_dpo/beta": 0.0036497246474027634, "fcm_dpo/delta": -0.006030656397342682, "fcm_dpo/margin": 220.30859375, "fcm_dpo/q_t": 0.33241087198257446, "grad_norm": 36.527130126953125, "learning_rate": 1.1764499893210878e-08, "logits/chosen": 0.2730959951877594, "logits/rejected": 0.28664201498031616, "logps/chosen": -468.7888488769531, "logps/ref_chosen": -60.28582000732422, "logps/ref_rejected": -85.51873779296875, "logps/rejected": -714.330322265625, "loss": 0.8572, "margin_dpo/margin_mean": 220.30859375, "margin_dpo/margin_std": 212.906005859375, "step": 622 }, { "KL/chosen_KL_mean": -440.00592041015625, "KL/mean": -529.0611572265625, "KL/rejected_KL_mean": -618.116455078125, "KL/std": 196.42318725585938, "epoch": 0.9148311306901615, "fcm_dpo/beta": 0.0037519452162086964, "fcm_dpo/delta": 0.14135098457336426, "fcm_dpo/margin": 178.11048889160156, "fcm_dpo/q_t": 0.36387038230895996, "grad_norm": 47.815452575683594, "learning_rate": 1.1378595443300998e-08, "logits/chosen": 0.45722734928131104, "logits/rejected": 0.4867405891418457, "logps/chosen": -504.16290283203125, "logps/ref_chosen": -64.1569595336914, "logps/ref_rejected": -85.08304595947266, "logps/rejected": -703.199462890625, "loss": 0.9863, "margin_dpo/margin_mean": 178.11048889160156, "margin_dpo/margin_std": 237.95982360839844, "step": 623 }, { "KL/chosen_KL_mean": -433.9539794921875, "KL/mean": -552.2816162109375, "KL/rejected_KL_mean": -670.609375, "KL/std": 225.03269958496094, "epoch": 0.9162995594713657, "fcm_dpo/beta": 0.0037298626266419888, "fcm_dpo/delta": -0.0908697247505188, "fcm_dpo/margin": 236.65536499023438, "fcm_dpo/q_t": 0.31607311964035034, "grad_norm": 42.4611930847168, "learning_rate": 1.0998978889320582e-08, "logits/chosen": 0.272026389837265, "logits/rejected": 0.31344348192214966, "logps/chosen": -505.8725891113281, "logps/ref_chosen": -71.91862487792969, "logps/ref_rejected": -97.13203430175781, "logps/rejected": -767.7413330078125, "loss": 0.8393, "margin_dpo/margin_mean": 236.65538024902344, "margin_dpo/margin_std": 227.3984375, "step": 624 }, { "KL/chosen_KL_mean": -404.8118896484375, "KL/mean": -517.1644287109375, "KL/rejected_KL_mean": -629.5169067382812, "KL/std": 228.5667266845703, "epoch": 0.9177679882525698, "fcm_dpo/beta": 0.003692234866321087, "fcm_dpo/delta": -0.032496195286512375, "fcm_dpo/margin": 224.70501708984375, "fcm_dpo/q_t": 0.3303322196006775, "grad_norm": 46.905677795410156, "learning_rate": 1.0625660234518913e-08, "logits/chosen": 0.34732866287231445, "logits/rejected": 0.3889002799987793, "logps/chosen": -463.1539611816406, "logps/ref_chosen": -58.342071533203125, "logps/ref_rejected": -86.09038543701172, "logps/rejected": -715.6072998046875, "loss": 0.8505, "margin_dpo/margin_mean": 224.7050323486328, "margin_dpo/margin_std": 226.28468322753906, "step": 625 }, { "KL/chosen_KL_mean": -436.82305908203125, "KL/mean": -520.375732421875, "KL/rejected_KL_mean": -603.928466796875, "KL/std": 223.38800048828125, "epoch": 0.9192364170337739, "fcm_dpo/beta": 0.003800017526373267, "fcm_dpo/delta": 0.1756470501422882, "fcm_dpo/margin": 167.10537719726562, "fcm_dpo/q_t": 0.36899447441101074, "grad_norm": 52.31256866455078, "learning_rate": 1.0258649316189721e-08, "logits/chosen": 0.21514081954956055, "logits/rejected": 0.23527026176452637, "logps/chosen": -511.9356689453125, "logps/ref_chosen": -75.11260986328125, "logps/ref_rejected": -99.188720703125, "logps/rejected": -703.1171875, "loss": 0.9835, "margin_dpo/margin_mean": 167.10537719726562, "margin_dpo/margin_std": 206.45814514160156, "step": 626 }, { "KL/chosen_KL_mean": -373.7752380371094, "KL/mean": -493.3340759277344, "KL/rejected_KL_mean": -612.8929443359375, "KL/std": 255.04714965820312, "epoch": 0.920704845814978, "fcm_dpo/beta": 0.003763319458812475, "fcm_dpo/delta": -0.10977232456207275, "fcm_dpo/margin": 239.11770629882812, "fcm_dpo/q_t": 0.3324761986732483, "grad_norm": 54.27702331542969, "learning_rate": 9.897955805412e-09, "logits/chosen": 0.2599024772644043, "logits/rejected": 0.2572460472583771, "logps/chosen": -421.51837158203125, "logps/ref_chosen": -47.74314880371094, "logps/ref_rejected": -106.75448608398438, "logps/rejected": -719.6474609375, "loss": 0.9026, "margin_dpo/margin_mean": 239.11770629882812, "margin_dpo/margin_std": 296.06634521484375, "step": 627 }, { "KL/chosen_KL_mean": -424.67303466796875, "KL/mean": -548.8368530273438, "KL/rejected_KL_mean": -673.0006713867188, "KL/std": 218.74835205078125, "epoch": 0.922173274596182, "fcm_dpo/beta": 0.0036470850463956594, "fcm_dpo/delta": -0.11861401796340942, "fcm_dpo/margin": 248.32763671875, "fcm_dpo/q_t": 0.3126278221607208, "grad_norm": 31.617103576660156, "learning_rate": 9.543589206795238e-09, "logits/chosen": 0.3635599613189697, "logits/rejected": 0.3939652144908905, "logps/chosen": -484.8559875488281, "logps/ref_chosen": -60.182945251464844, "logps/ref_rejected": -101.55467224121094, "logps/rejected": -774.5553588867188, "loss": 0.8208, "margin_dpo/margin_mean": 248.32762145996094, "margin_dpo/margin_std": 228.94168090820312, "step": 628 }, { "KL/chosen_KL_mean": -443.83966064453125, "KL/mean": -550.1680908203125, "KL/rejected_KL_mean": -656.4964599609375, "KL/std": 213.83009338378906, "epoch": 0.9236417033773862, "fcm_dpo/beta": 0.00366116501390934, "fcm_dpo/delta": 0.023219341412186623, "fcm_dpo/margin": 212.65676879882812, "fcm_dpo/q_t": 0.3320544362068176, "grad_norm": 44.99949645996094, "learning_rate": 9.19555885822887e-09, "logits/chosen": 0.3668147921562195, "logits/rejected": 0.39162763953208923, "logps/chosen": -508.0531921386719, "logps/ref_chosen": -64.21354675292969, "logps/ref_rejected": -91.65367126464844, "logps/rejected": -748.150146484375, "loss": 0.8466, "margin_dpo/margin_mean": 212.65676879882812, "margin_dpo/margin_std": 182.10867309570312, "step": 629 }, { "KL/chosen_KL_mean": -376.22467041015625, "KL/mean": -463.4542236328125, "KL/rejected_KL_mean": -550.683837890625, "KL/std": 224.61175537109375, "epoch": 0.9251101321585903, "fcm_dpo/beta": 0.003765605855733156, "fcm_dpo/delta": 0.15256844460964203, "fcm_dpo/margin": 174.45912170410156, "fcm_dpo/q_t": 0.36428147554397583, "grad_norm": 54.26115417480469, "learning_rate": 8.85387393063622e-09, "logits/chosen": 0.2349189817905426, "logits/rejected": 0.25930139422416687, "logps/chosen": -435.51568603515625, "logps/ref_chosen": -59.29100036621094, "logps/ref_rejected": -83.59829711914062, "logps/rejected": -634.2821044921875, "loss": 0.9658, "margin_dpo/margin_mean": 174.45912170410156, "margin_dpo/margin_std": 213.33807373046875, "step": 630 }, { "KL/chosen_KL_mean": -453.9466552734375, "KL/mean": -548.2926635742188, "KL/rejected_KL_mean": -642.638671875, "KL/std": 223.95266723632812, "epoch": 0.9265785609397944, "fcm_dpo/beta": 0.003838051576167345, "fcm_dpo/delta": 0.08150237053632736, "fcm_dpo/margin": 188.6920623779297, "fcm_dpo/q_t": 0.3542540371417999, "grad_norm": 41.3115348815918, "learning_rate": 8.518543427732949e-09, "logits/chosen": 0.4632750153541565, "logits/rejected": 0.49574044346809387, "logps/chosen": -513.4002685546875, "logps/ref_chosen": -59.45360565185547, "logps/ref_rejected": -80.95156860351562, "logps/rejected": -723.5902709960938, "loss": 0.984, "margin_dpo/margin_mean": 188.69207763671875, "margin_dpo/margin_std": 250.87442016601562, "step": 631 }, { "KL/chosen_KL_mean": -445.6044616699219, "KL/mean": -548.7745361328125, "KL/rejected_KL_mean": -651.944580078125, "KL/std": 197.17782592773438, "epoch": 0.9280469897209985, "fcm_dpo/beta": 0.0038299935404211283, "fcm_dpo/delta": 0.00984569638967514, "fcm_dpo/margin": 206.340087890625, "fcm_dpo/q_t": 0.33502843976020813, "grad_norm": 51.954689025878906, "learning_rate": 8.189576185789637e-09, "logits/chosen": 0.4514876902103424, "logits/rejected": 0.48781538009643555, "logps/chosen": -506.95599365234375, "logps/ref_chosen": -61.35155487060547, "logps/ref_rejected": -86.16017150878906, "logps/rejected": -738.104736328125, "loss": 0.9268, "margin_dpo/margin_mean": 206.340087890625, "margin_dpo/margin_std": 238.1063232421875, "step": 632 }, { "KL/chosen_KL_mean": -445.6147155761719, "KL/mean": -528.477783203125, "KL/rejected_KL_mean": -611.3408203125, "KL/std": 218.96871948242188, "epoch": 0.9295154185022027, "fcm_dpo/beta": 0.003975407220423222, "fcm_dpo/delta": 0.14825844764709473, "fcm_dpo/margin": 165.72608947753906, "fcm_dpo/q_t": 0.3641470968723297, "grad_norm": 43.61115646362305, "learning_rate": 7.866980873399015e-09, "logits/chosen": 0.39295196533203125, "logits/rejected": 0.39339303970336914, "logps/chosen": -502.89288330078125, "logps/ref_chosen": -57.27816390991211, "logps/ref_rejected": -91.58395385742188, "logps/rejected": -702.9247436523438, "loss": 1.0067, "margin_dpo/margin_mean": 165.72608947753906, "margin_dpo/margin_std": 223.81736755371094, "step": 633 }, { "KL/chosen_KL_mean": -447.18731689453125, "KL/mean": -530.3380126953125, "KL/rejected_KL_mean": -613.48876953125, "KL/std": 233.11648559570312, "epoch": 0.9309838472834068, "fcm_dpo/beta": 0.004039407707750797, "fcm_dpo/delta": 0.13770422339439392, "fcm_dpo/margin": 166.3013916015625, "fcm_dpo/q_t": 0.36527395248413086, "grad_norm": 50.8237190246582, "learning_rate": 7.550765991247654e-09, "logits/chosen": 0.2625778913497925, "logits/rejected": 0.26742881536483765, "logps/chosen": -513.8062744140625, "logps/ref_chosen": -66.61896514892578, "logps/ref_rejected": -107.12564849853516, "logps/rejected": -720.6143798828125, "loss": 0.9847, "margin_dpo/margin_mean": 166.3013916015625, "margin_dpo/margin_std": 215.0738525390625, "step": 634 }, { "KL/chosen_KL_mean": -420.28643798828125, "KL/mean": -511.7974548339844, "KL/rejected_KL_mean": -603.3084716796875, "KL/std": 232.0856475830078, "epoch": 0.9324522760646109, "fcm_dpo/beta": 0.004115741699934006, "fcm_dpo/delta": 0.05048708617687225, "fcm_dpo/margin": 183.02200317382812, "fcm_dpo/q_t": 0.3472629189491272, "grad_norm": 42.469337463378906, "learning_rate": 7.240939871891699e-09, "logits/chosen": 0.23344659805297852, "logits/rejected": 0.2846734821796417, "logps/chosen": -494.241943359375, "logps/ref_chosen": -73.95551300048828, "logps/ref_rejected": -82.50045776367188, "logps/rejected": -685.8089599609375, "loss": 0.9132, "margin_dpo/margin_mean": 183.0220184326172, "margin_dpo/margin_std": 207.57760620117188, "step": 635 }, { "KL/chosen_KL_mean": -409.06329345703125, "KL/mean": -514.981689453125, "KL/rejected_KL_mean": -620.900146484375, "KL/std": 233.29795837402344, "epoch": 0.933920704845815, "fcm_dpo/beta": 0.004116167314350605, "fcm_dpo/delta": -0.08010595291852951, "fcm_dpo/margin": 211.83688354492188, "fcm_dpo/q_t": 0.31917688250541687, "grad_norm": 41.35598373413086, "learning_rate": 6.937510679537628e-09, "logits/chosen": 0.2601292133331299, "logits/rejected": 0.2906838655471802, "logps/chosen": -468.69219970703125, "logps/ref_chosen": -59.628910064697266, "logps/ref_rejected": -81.97883605957031, "logps/rejected": -702.8790283203125, "loss": 0.8391, "margin_dpo/margin_mean": 211.83688354492188, "margin_dpo/margin_std": 202.292724609375, "step": 636 }, { "KL/chosen_KL_mean": -432.2290344238281, "KL/mean": -548.3233642578125, "KL/rejected_KL_mean": -664.4176635742188, "KL/std": 238.66481018066406, "epoch": 0.9353891336270191, "fcm_dpo/beta": 0.003974968567490578, "fcm_dpo/delta": -0.13554571568965912, "fcm_dpo/margin": 232.18862915039062, "fcm_dpo/q_t": 0.31292521953582764, "grad_norm": 50.888275146484375, "learning_rate": 6.640486409826785e-09, "logits/chosen": 0.45192140340805054, "logits/rejected": 0.47448259592056274, "logps/chosen": -481.8817138671875, "logps/ref_chosen": -49.652687072753906, "logps/ref_rejected": -98.40513610839844, "logps/rejected": -762.8228149414062, "loss": 0.8169, "margin_dpo/margin_mean": 232.1886444091797, "margin_dpo/margin_std": 223.49786376953125, "step": 637 }, { "KL/chosen_KL_mean": -392.0146179199219, "KL/mean": -504.70819091796875, "KL/rejected_KL_mean": -617.4017333984375, "KL/std": 214.28329467773438, "epoch": 0.9368575624082232, "fcm_dpo/beta": 0.003946408629417419, "fcm_dpo/delta": -0.10087276995182037, "fcm_dpo/margin": 225.3870849609375, "fcm_dpo/q_t": 0.312773197889328, "grad_norm": 45.85356140136719, "learning_rate": 6.349874889624962e-09, "logits/chosen": 0.3099361062049866, "logits/rejected": 0.3372785747051239, "logps/chosen": -450.1712646484375, "logps/ref_chosen": -58.156639099121094, "logps/ref_rejected": -79.3014907836914, "logps/rejected": -696.7032470703125, "loss": 0.7993, "margin_dpo/margin_mean": 225.3870849609375, "margin_dpo/margin_std": 178.96417236328125, "step": 638 }, { "KL/chosen_KL_mean": -464.5148620605469, "KL/mean": -549.45556640625, "KL/rejected_KL_mean": -634.396240234375, "KL/std": 196.27197265625, "epoch": 0.9383259911894273, "fcm_dpo/beta": 0.003963670693337917, "fcm_dpo/delta": 0.13499195873737335, "fcm_dpo/margin": 169.88137817382812, "fcm_dpo/q_t": 0.3594561815261841, "grad_norm": 48.99491500854492, "learning_rate": 6.065683776815933e-09, "logits/chosen": 0.2308313548564911, "logits/rejected": 0.26190468668937683, "logps/chosen": -536.8380126953125, "logps/ref_chosen": -72.32319641113281, "logps/ref_rejected": -74.2749252319336, "logps/rejected": -708.671142578125, "loss": 0.9754, "margin_dpo/margin_mean": 169.88137817382812, "margin_dpo/margin_std": 211.46803283691406, "step": 639 }, { "KL/chosen_KL_mean": -420.45050048828125, "KL/mean": -549.830078125, "KL/rejected_KL_mean": -679.209716796875, "KL/std": 223.74356079101562, "epoch": 0.9397944199706314, "fcm_dpo/beta": 0.0038788788951933384, "fcm_dpo/delta": -0.22599047422409058, "fcm_dpo/margin": 258.75921630859375, "fcm_dpo/q_t": 0.30272677540779114, "grad_norm": 39.78026580810547, "learning_rate": 5.7879205600998296e-09, "logits/chosen": 0.29272258281707764, "logits/rejected": 0.326360285282135, "logps/chosen": -476.58489990234375, "logps/ref_chosen": -56.13436508178711, "logps/ref_rejected": -108.60014343261719, "logps/rejected": -787.8098754882812, "loss": 0.7921, "margin_dpo/margin_mean": 258.75921630859375, "margin_dpo/margin_std": 251.56118774414062, "step": 640 }, { "KL/chosen_KL_mean": -481.4560241699219, "KL/mean": -571.785400390625, "KL/rejected_KL_mean": -662.1146850585938, "KL/std": 199.70594787597656, "epoch": 0.9412628487518355, "fcm_dpo/beta": 0.0038441240321844816, "fcm_dpo/delta": 0.11339370161294937, "fcm_dpo/margin": 180.65867614746094, "fcm_dpo/q_t": 0.3584289252758026, "grad_norm": 49.26630401611328, "learning_rate": 5.516592558795746e-09, "logits/chosen": 0.3979315161705017, "logits/rejected": 0.42757904529571533, "logps/chosen": -546.452880859375, "logps/ref_chosen": -64.99689483642578, "logps/ref_rejected": -86.99232482910156, "logps/rejected": -749.1070556640625, "loss": 0.9607, "margin_dpo/margin_mean": 180.65866088867188, "margin_dpo/margin_std": 226.99839782714844, "step": 641 }, { "KL/chosen_KL_mean": -422.64208984375, "KL/mean": -533.4561157226562, "KL/rejected_KL_mean": -644.2701416015625, "KL/std": 238.713623046875, "epoch": 0.9427312775330396, "fcm_dpo/beta": 0.0038909269496798515, "fcm_dpo/delta": -0.06998465955257416, "fcm_dpo/margin": 221.62806701660156, "fcm_dpo/q_t": 0.33400386571884155, "grad_norm": 46.33919906616211, "learning_rate": 5.251706922648868e-09, "logits/chosen": 0.2569606304168701, "logits/rejected": 0.2733567953109741, "logps/chosen": -488.33135986328125, "logps/ref_chosen": -65.68924713134766, "logps/ref_rejected": -110.24205017089844, "logps/rejected": -754.51220703125, "loss": 0.9007, "margin_dpo/margin_mean": 221.6280517578125, "margin_dpo/margin_std": 262.16290283203125, "step": 642 }, { "KL/chosen_KL_mean": -417.83306884765625, "KL/mean": -518.9586181640625, "KL/rejected_KL_mean": -620.084228515625, "KL/std": 239.10421752929688, "epoch": 0.9441997063142438, "fcm_dpo/beta": 0.0038129761815071106, "fcm_dpo/delta": 0.029478274285793304, "fcm_dpo/margin": 202.25112915039062, "fcm_dpo/q_t": 0.33728426694869995, "grad_norm": 35.140663146972656, "learning_rate": 4.993270631642038e-09, "logits/chosen": 0.4179457128047943, "logits/rejected": 0.4211399555206299, "logps/chosen": -469.7830810546875, "logps/ref_chosen": -51.94999694824219, "logps/ref_rejected": -87.46833801269531, "logps/rejected": -707.5525512695312, "loss": 0.8858, "margin_dpo/margin_mean": 202.25112915039062, "margin_dpo/margin_std": 198.2943115234375, "step": 643 }, { "KL/chosen_KL_mean": -431.06512451171875, "KL/mean": -527.8743896484375, "KL/rejected_KL_mean": -624.68359375, "KL/std": 257.38726806640625, "epoch": 0.9456681350954479, "fcm_dpo/beta": 0.00388428196310997, "fcm_dpo/delta": 0.05185030400753021, "fcm_dpo/margin": 193.61854553222656, "fcm_dpo/q_t": 0.3501347303390503, "grad_norm": 50.662540435791016, "learning_rate": 4.741290495811873e-09, "logits/chosen": 0.3143874704837799, "logits/rejected": 0.33662909269332886, "logps/chosen": -490.0827941894531, "logps/ref_chosen": -59.017662048339844, "logps/ref_rejected": -87.13668823242188, "logps/rejected": -711.8203125, "loss": 0.9535, "margin_dpo/margin_mean": 193.6185302734375, "margin_dpo/margin_std": 244.5896759033203, "step": 644 }, { "KL/chosen_KL_mean": -429.000244140625, "KL/mean": -505.9991760253906, "KL/rejected_KL_mean": -582.9981079101562, "KL/std": 221.11386108398438, "epoch": 0.947136563876652, "fcm_dpo/beta": 0.0039842319674789906, "fcm_dpo/delta": 0.19895383715629578, "fcm_dpo/margin": 153.99783325195312, "fcm_dpo/q_t": 0.3828478455543518, "grad_norm": 52.093482971191406, "learning_rate": 4.495773155069299e-09, "logits/chosen": 0.27324581146240234, "logits/rejected": 0.2675897479057312, "logps/chosen": -484.87628173828125, "logps/ref_chosen": -55.87602233886719, "logps/ref_rejected": -97.78080749511719, "logps/rejected": -680.7789306640625, "loss": 1.0637, "margin_dpo/margin_mean": 153.99783325195312, "margin_dpo/margin_std": 245.86737060546875, "step": 645 }, { "KL/chosen_KL_mean": -431.00897216796875, "KL/mean": -530.1956176757812, "KL/rejected_KL_mean": -629.3822021484375, "KL/std": 180.67453002929688, "epoch": 0.9486049926578561, "fcm_dpo/beta": 0.004018571227788925, "fcm_dpo/delta": 0.001294448971748352, "fcm_dpo/margin": 198.37326049804688, "fcm_dpo/q_t": 0.33232998847961426, "grad_norm": 46.75684356689453, "learning_rate": 4.256725079024553e-09, "logits/chosen": 0.405788779258728, "logits/rejected": 0.43029361963272095, "logps/chosen": -492.2847595214844, "logps/ref_chosen": -61.275787353515625, "logps/ref_rejected": -77.50580596923828, "logps/rejected": -706.8880615234375, "loss": 0.8649, "margin_dpo/margin_mean": 198.37326049804688, "margin_dpo/margin_std": 189.44442749023438, "step": 646 }, { "KL/chosen_KL_mean": -395.1177673339844, "KL/mean": -494.07708740234375, "KL/rejected_KL_mean": -593.036376953125, "KL/std": 234.684326171875, "epoch": 0.9500734214390602, "fcm_dpo/beta": 0.0040655904449522495, "fcm_dpo/delta": -0.005251521244645119, "fcm_dpo/margin": 197.91859436035156, "fcm_dpo/q_t": 0.3365859389305115, "grad_norm": 46.11881637573242, "learning_rate": 4.024152566816791e-09, "logits/chosen": 0.3098648488521576, "logits/rejected": 0.32036569714546204, "logps/chosen": -449.9701843261719, "logps/ref_chosen": -54.8524169921875, "logps/ref_rejected": -93.5194091796875, "logps/rejected": -686.5557861328125, "loss": 0.9117, "margin_dpo/margin_mean": 197.9185791015625, "margin_dpo/margin_std": 225.00868225097656, "step": 647 }, { "KL/chosen_KL_mean": -425.3743896484375, "KL/mean": -558.19775390625, "KL/rejected_KL_mean": -691.0211181640625, "KL/std": 252.74656677246094, "epoch": 0.9515418502202643, "fcm_dpo/beta": 0.003904239274561405, "fcm_dpo/delta": -0.26420071721076965, "fcm_dpo/margin": 265.6467590332031, "fcm_dpo/q_t": 0.30838528275489807, "grad_norm": 40.06119918823242, "learning_rate": 3.798061746947995e-09, "logits/chosen": 0.40732091665267944, "logits/rejected": 0.4274319112300873, "logps/chosen": -479.54583740234375, "logps/ref_chosen": -54.17146682739258, "logps/ref_rejected": -98.7127914428711, "logps/rejected": -789.7339477539062, "loss": 0.8385, "margin_dpo/margin_mean": 265.64678955078125, "margin_dpo/margin_std": 304.36572265625, "step": 648 }, { "KL/chosen_KL_mean": -435.12518310546875, "KL/mean": -517.2813110351562, "KL/rejected_KL_mean": -599.4374389648438, "KL/std": 208.06198120117188, "epoch": 0.9530102790014684, "fcm_dpo/beta": 0.0039540440775454044, "fcm_dpo/delta": 0.15951567888259888, "fcm_dpo/margin": 164.31224060058594, "fcm_dpo/q_t": 0.3723710775375366, "grad_norm": 54.13444137573242, "learning_rate": 3.5784585771215235e-09, "logits/chosen": 0.3933088779449463, "logits/rejected": 0.4235169291496277, "logps/chosen": -497.60552978515625, "logps/ref_chosen": -62.480350494384766, "logps/ref_rejected": -80.07717895507812, "logps/rejected": -679.5146484375, "loss": 1.0082, "margin_dpo/margin_mean": 164.312255859375, "margin_dpo/margin_std": 231.4293212890625, "step": 649 }, { "KL/chosen_KL_mean": -433.6247253417969, "KL/mean": -552.1237182617188, "KL/rejected_KL_mean": -670.6226806640625, "KL/std": 234.92752075195312, "epoch": 0.9544787077826725, "fcm_dpo/beta": 0.003866554470732808, "fcm_dpo/delta": -0.12990686297416687, "fcm_dpo/margin": 236.99801635742188, "fcm_dpo/q_t": 0.31889447569847107, "grad_norm": 50.74138641357422, "learning_rate": 3.3653488440851253e-09, "logits/chosen": 0.3859984278678894, "logits/rejected": 0.4003593921661377, "logps/chosen": -489.717529296875, "logps/ref_chosen": -56.09281921386719, "logps/ref_rejected": -98.26483917236328, "logps/rejected": -768.8875732421875, "loss": 0.8666, "margin_dpo/margin_mean": 236.99801635742188, "margin_dpo/margin_std": 258.2767639160156, "step": 650 }, { "KL/chosen_KL_mean": -375.603271484375, "KL/mean": -498.3697814941406, "KL/rejected_KL_mean": -621.1362915039062, "KL/std": 243.66693115234375, "epoch": 0.9559471365638766, "fcm_dpo/beta": 0.0037730569019913673, "fcm_dpo/delta": -0.14011479914188385, "fcm_dpo/margin": 245.53298950195312, "fcm_dpo/q_t": 0.3070628046989441, "grad_norm": 38.228172302246094, "learning_rate": 3.158738163478475e-09, "logits/chosen": 0.3418081998825073, "logits/rejected": 0.34055066108703613, "logps/chosen": -419.02874755859375, "logps/ref_chosen": -43.42544937133789, "logps/ref_rejected": -99.95791625976562, "logps/rejected": -721.0941772460938, "loss": 0.7932, "margin_dpo/margin_mean": 245.53298950195312, "margin_dpo/margin_std": 212.47959899902344, "step": 651 }, { "KL/chosen_KL_mean": -399.320556640625, "KL/mean": -506.40570068359375, "KL/rejected_KL_mean": -613.4908447265625, "KL/std": 232.77484130859375, "epoch": 0.9574155653450808, "fcm_dpo/beta": 0.0037672333419322968, "fcm_dpo/delta": -0.007625289261341095, "fcm_dpo/margin": 214.17034912109375, "fcm_dpo/q_t": 0.3373221755027771, "grad_norm": 51.426910400390625, "learning_rate": 2.9586319796851555e-09, "logits/chosen": 0.3437264561653137, "logits/rejected": 0.36428600549697876, "logps/chosen": -461.8973388671875, "logps/ref_chosen": -62.57680892944336, "logps/ref_rejected": -111.76779174804688, "logps/rejected": -725.2586669921875, "loss": 0.8841, "margin_dpo/margin_mean": 214.17034912109375, "margin_dpo/margin_std": 232.0792236328125, "step": 652 }, { "KL/chosen_KL_mean": -456.14044189453125, "KL/mean": -569.4014892578125, "KL/rejected_KL_mean": -682.6625366210938, "KL/std": 250.85052490234375, "epoch": 0.9588839941262849, "fcm_dpo/beta": 0.0037269643507897854, "fcm_dpo/delta": -0.048316843807697296, "fcm_dpo/margin": 226.52212524414062, "fcm_dpo/q_t": 0.33045560121536255, "grad_norm": 45.9043083190918, "learning_rate": 2.7650355656892166e-09, "logits/chosen": 0.39898934960365295, "logits/rejected": 0.40452104806900024, "logps/chosen": -517.25341796875, "logps/ref_chosen": -61.11295700073242, "logps/ref_rejected": -103.24960327148438, "logps/rejected": -785.912109375, "loss": 0.8724, "margin_dpo/margin_mean": 226.52212524414062, "margin_dpo/margin_std": 243.38613891601562, "step": 653 }, { "KL/chosen_KL_mean": -455.92962646484375, "KL/mean": -552.9722900390625, "KL/rejected_KL_mean": -650.0150146484375, "KL/std": 228.19635009765625, "epoch": 0.960352422907489, "fcm_dpo/beta": 0.00373679306358099, "fcm_dpo/delta": 0.08018621057271957, "fcm_dpo/margin": 194.08538818359375, "fcm_dpo/q_t": 0.3584059476852417, "grad_norm": 45.61049270629883, "learning_rate": 2.577954022936174e-09, "logits/chosen": 0.3774251341819763, "logits/rejected": 0.3957637548446655, "logps/chosen": -517.6577758789062, "logps/ref_chosen": -61.7281379699707, "logps/ref_rejected": -98.7738037109375, "logps/rejected": -748.788818359375, "loss": 0.9826, "margin_dpo/margin_mean": 194.08538818359375, "margin_dpo/margin_std": 264.8636474609375, "step": 654 }, { "KL/chosen_KL_mean": -419.35821533203125, "KL/mean": -520.3919677734375, "KL/rejected_KL_mean": -621.4256591796875, "KL/std": 227.86204528808594, "epoch": 0.9618208516886931, "fcm_dpo/beta": 0.0037737099919468164, "fcm_dpo/delta": 0.03982290253043175, "fcm_dpo/margin": 202.06744384765625, "fcm_dpo/q_t": 0.35060518980026245, "grad_norm": 53.30228042602539, "learning_rate": 2.397392281198729e-09, "logits/chosen": 0.43709027767181396, "logits/rejected": 0.43980515003204346, "logps/chosen": -468.9350280761719, "logps/ref_chosen": -49.576812744140625, "logps/ref_rejected": -98.29183197021484, "logps/rejected": -719.717529296875, "loss": 0.944, "margin_dpo/margin_mean": 202.0674591064453, "margin_dpo/margin_std": 251.4522705078125, "step": 655 }, { "KL/chosen_KL_mean": -409.84423828125, "KL/mean": -561.7591552734375, "KL/rejected_KL_mean": -713.674072265625, "KL/std": 255.30087280273438, "epoch": 0.9632892804698973, "fcm_dpo/beta": 0.0036293421871960163, "fcm_dpo/delta": -0.3396731913089752, "fcm_dpo/margin": 303.82989501953125, "fcm_dpo/q_t": 0.2718742787837982, "grad_norm": 53.86192321777344, "learning_rate": 2.223355098446622e-09, "logits/chosen": 0.35994166135787964, "logits/rejected": 0.3799811601638794, "logps/chosen": -462.3936767578125, "logps/ref_chosen": -52.54943084716797, "logps/ref_rejected": -113.67464447021484, "logps/rejected": -827.3487548828125, "loss": 0.6877, "margin_dpo/margin_mean": 303.82989501953125, "margin_dpo/margin_std": 221.92645263671875, "step": 656 }, { "KL/chosen_KL_mean": -405.48822021484375, "KL/mean": -541.8997192382812, "KL/rejected_KL_mean": -678.311279296875, "KL/std": 256.29046630859375, "epoch": 0.9647577092511013, "fcm_dpo/beta": 0.003453074023127556, "fcm_dpo/delta": -0.15873435139656067, "fcm_dpo/margin": 272.82305908203125, "fcm_dpo/q_t": 0.30371803045272827, "grad_norm": 35.55537796020508, "learning_rate": 2.055847060721566e-09, "logits/chosen": 0.4056025743484497, "logits/rejected": 0.41361987590789795, "logps/chosen": -452.1887512207031, "logps/ref_chosen": -46.700538635253906, "logps/ref_rejected": -97.91487121582031, "logps/rejected": -776.2261352539062, "loss": 0.776, "margin_dpo/margin_mean": 272.82305908203125, "margin_dpo/margin_std": 231.46688842773438, "step": 657 }, { "KL/chosen_KL_mean": -450.57080078125, "KL/mean": -557.7039794921875, "KL/rejected_KL_mean": -664.8372192382812, "KL/std": 209.4986572265625, "epoch": 0.9662261380323054, "fcm_dpo/beta": 0.0034792963415384293, "fcm_dpo/delta": 0.05883553624153137, "fcm_dpo/margin": 214.26651000976562, "fcm_dpo/q_t": 0.3380606174468994, "grad_norm": 33.780548095703125, "learning_rate": 1.8948725820160662e-09, "logits/chosen": 0.4116886258125305, "logits/rejected": 0.43388843536376953, "logps/chosen": -511.52899169921875, "logps/ref_chosen": -60.95820999145508, "logps/ref_rejected": -95.93949127197266, "logps/rejected": -760.7767333984375, "loss": 0.8741, "margin_dpo/margin_mean": 214.26651000976562, "margin_dpo/margin_std": 196.41368103027344, "step": 658 }, { "KL/chosen_KL_mean": -411.34381103515625, "KL/mean": -527.1859741210938, "KL/rejected_KL_mean": -643.0281372070312, "KL/std": 235.82553100585938, "epoch": 0.9676945668135095, "fcm_dpo/beta": 0.0034661423414945602, "fcm_dpo/delta": -0.003968711942434311, "fcm_dpo/margin": 231.68438720703125, "fcm_dpo/q_t": 0.3321956992149353, "grad_norm": 31.871097564697266, "learning_rate": 1.7404359041573723e-09, "logits/chosen": 0.22371786832809448, "logits/rejected": 0.26776695251464844, "logps/chosen": -488.0867919921875, "logps/ref_chosen": -76.74298095703125, "logps/ref_rejected": -87.4709701538086, "logps/rejected": -730.4991455078125, "loss": 0.8695, "margin_dpo/margin_mean": 231.68435668945312, "margin_dpo/margin_std": 228.46832275390625, "step": 659 }, { "KL/chosen_KL_mean": -431.14703369140625, "KL/mean": -555.50244140625, "KL/rejected_KL_mean": -679.8578491210938, "KL/std": 207.2142333984375, "epoch": 0.9691629955947136, "fcm_dpo/beta": 0.0034529143013060093, "fcm_dpo/delta": -0.0642768144607544, "fcm_dpo/margin": 248.7108154296875, "fcm_dpo/q_t": 0.315839558839798, "grad_norm": 30.965335845947266, "learning_rate": 1.592541096695571e-09, "logits/chosen": 0.4715263247489929, "logits/rejected": 0.5011571645736694, "logps/chosen": -490.1949462890625, "logps/ref_chosen": -59.04788589477539, "logps/ref_rejected": -75.96005249023438, "logps/rejected": -755.81787109375, "loss": 0.8116, "margin_dpo/margin_mean": 248.71083068847656, "margin_dpo/margin_std": 211.64910888671875, "step": 660 }, { "KL/chosen_KL_mean": -400.5611877441406, "KL/mean": -514.7567749023438, "KL/rejected_KL_mean": -628.952392578125, "KL/std": 235.23724365234375, "epoch": 0.9706314243759178, "fcm_dpo/beta": 0.00347909121774137, "fcm_dpo/delta": 0.00342472642660141, "fcm_dpo/margin": 228.3911895751953, "fcm_dpo/q_t": 0.3365651071071625, "grad_norm": 35.08213424682617, "learning_rate": 1.4511920567963908e-09, "logits/chosen": 0.344051718711853, "logits/rejected": 0.3751180171966553, "logps/chosen": -451.23516845703125, "logps/ref_chosen": -50.673973083496094, "logps/ref_rejected": -86.00569152832031, "logps/rejected": -714.9581298828125, "loss": 0.893, "margin_dpo/margin_mean": 228.39120483398438, "margin_dpo/margin_std": 240.8126678466797, "step": 661 }, { "KL/chosen_KL_mean": -426.64874267578125, "KL/mean": -528.4208984375, "KL/rejected_KL_mean": -630.193115234375, "KL/std": 217.40945434570312, "epoch": 0.9720998531571219, "fcm_dpo/beta": 0.0034924065694212914, "fcm_dpo/delta": 0.09593937546014786, "fcm_dpo/margin": 203.54437255859375, "fcm_dpo/q_t": 0.3504369258880615, "grad_norm": 31.930891036987305, "learning_rate": 1.3163925091384532e-09, "logits/chosen": 0.2393924593925476, "logits/rejected": 0.2724754810333252, "logps/chosen": -495.9097900390625, "logps/ref_chosen": -69.26106262207031, "logps/ref_rejected": -89.05593872070312, "logps/rejected": -719.2490234375, "loss": 0.9342, "margin_dpo/margin_mean": 203.54437255859375, "margin_dpo/margin_std": 228.84689331054688, "step": 662 }, { "KL/chosen_KL_mean": -405.4971923828125, "KL/mean": -521.6410522460938, "KL/rejected_KL_mean": -637.784912109375, "KL/std": 248.77371215820312, "epoch": 0.973568281938326, "fcm_dpo/beta": 0.0034925497602671385, "fcm_dpo/delta": -0.012477856129407883, "fcm_dpo/margin": 232.28768920898438, "fcm_dpo/q_t": 0.33259835839271545, "grad_norm": 34.18339538574219, "learning_rate": 1.1881460058152382e-09, "logits/chosen": 0.26970383524894714, "logits/rejected": 0.28010663390159607, "logps/chosen": -470.3760986328125, "logps/ref_chosen": -64.87890625, "logps/ref_rejected": -113.92536926269531, "logps/rejected": -751.7102661132812, "loss": 0.8659, "margin_dpo/margin_mean": 232.28768920898438, "margin_dpo/margin_std": 239.78091430664062, "step": 663 }, { "KL/chosen_KL_mean": -440.453369140625, "KL/mean": -567.512939453125, "KL/rejected_KL_mean": -694.572509765625, "KL/std": 230.0228729248047, "epoch": 0.9750367107195301, "fcm_dpo/beta": 0.003445860929787159, "fcm_dpo/delta": -0.08343236148357391, "fcm_dpo/margin": 254.11920166015625, "fcm_dpo/q_t": 0.3171752691268921, "grad_norm": 51.455936431884766, "learning_rate": 1.066455926241383e-09, "logits/chosen": 0.371050089597702, "logits/rejected": 0.3953966796398163, "logps/chosen": -501.3418273925781, "logps/ref_chosen": -60.88847351074219, "logps/ref_rejected": -105.521728515625, "logps/rejected": -800.09423828125, "loss": 0.8572, "margin_dpo/margin_mean": 254.11920166015625, "margin_dpo/margin_std": 262.84967041015625, "step": 664 }, { "KL/chosen_KL_mean": -427.70672607421875, "KL/mean": -538.66796875, "KL/rejected_KL_mean": -649.629150390625, "KL/std": 203.9910888671875, "epoch": 0.9765051395007343, "fcm_dpo/beta": 0.003454443998634815, "fcm_dpo/delta": 0.03614869341254234, "fcm_dpo/margin": 221.92245483398438, "fcm_dpo/q_t": 0.3373722434043884, "grad_norm": 51.79203796386719, "learning_rate": 9.513254770636137e-10, "logits/chosen": 0.5427942276000977, "logits/rejected": 0.5816524028778076, "logps/chosen": -488.2708435058594, "logps/ref_chosen": -60.56413269042969, "logps/ref_rejected": -84.80882263183594, "logps/rejected": -734.43798828125, "loss": 0.8702, "margin_dpo/margin_mean": 221.9224395751953, "margin_dpo/margin_std": 207.96560668945312, "step": 665 }, { "KL/chosen_KL_mean": -430.86529541015625, "KL/mean": -541.441650390625, "KL/rejected_KL_mean": -652.01806640625, "KL/std": 219.16168212890625, "epoch": 0.9779735682819384, "fcm_dpo/beta": 0.0034718806855380535, "fcm_dpo/delta": 0.03468765318393707, "fcm_dpo/margin": 221.15277099609375, "fcm_dpo/q_t": 0.33385133743286133, "grad_norm": 41.26150894165039, "learning_rate": 8.427576920763956e-10, "logits/chosen": 0.2883135974407196, "logits/rejected": 0.31811410188674927, "logps/chosen": -495.2852783203125, "logps/ref_chosen": -64.41996002197266, "logps/ref_rejected": -95.8916244506836, "logps/rejected": -747.90966796875, "loss": 0.8674, "margin_dpo/margin_mean": 221.15277099609375, "margin_dpo/margin_std": 200.27261352539062, "step": 666 }, { "KL/chosen_KL_mean": -439.869384765625, "KL/mean": -552.4962158203125, "KL/rejected_KL_mean": -665.123046875, "KL/std": 228.28411865234375, "epoch": 0.9794419970631424, "fcm_dpo/beta": 0.003500557504594326, "fcm_dpo/delta": 0.012448148801922798, "fcm_dpo/margin": 225.253662109375, "fcm_dpo/q_t": 0.328310489654541, "grad_norm": 37.877227783203125, "learning_rate": 7.407554321417764e-10, "logits/chosen": 0.3518103361129761, "logits/rejected": 0.4001040458679199, "logps/chosen": -509.14642333984375, "logps/ref_chosen": -69.27702331542969, "logps/ref_rejected": -87.83549499511719, "logps/rejected": -752.95849609375, "loss": 0.8385, "margin_dpo/margin_mean": 225.253662109375, "margin_dpo/margin_std": 187.47291564941406, "step": 667 }, { "KL/chosen_KL_mean": -464.80975341796875, "KL/mean": -560.3359375, "KL/rejected_KL_mean": -655.8621826171875, "KL/std": 241.27587890625, "epoch": 0.9809104258443465, "fcm_dpo/beta": 0.0035862771328538656, "fcm_dpo/delta": 0.12170104682445526, "fcm_dpo/margin": 191.05247497558594, "fcm_dpo/q_t": 0.3601748049259186, "grad_norm": 59.456485748291016, "learning_rate": 6.453213851142225e-10, "logits/chosen": 0.3165743947029114, "logits/rejected": 0.3272818326950073, "logps/chosen": -537.4137573242188, "logps/ref_chosen": -72.60400390625, "logps/ref_rejected": -103.73905944824219, "logps/rejected": -759.601318359375, "loss": 0.969, "margin_dpo/margin_mean": 191.05245971679688, "margin_dpo/margin_std": 241.36276245117188, "step": 668 }, { "KL/chosen_KL_mean": -378.2247314453125, "KL/mean": -504.9932861328125, "KL/rejected_KL_mean": -631.7618408203125, "KL/std": 232.52206420898438, "epoch": 0.9823788546255506, "fcm_dpo/beta": 0.0035265935584902763, "fcm_dpo/delta": -0.10357140004634857, "fcm_dpo/margin": 253.537109375, "fcm_dpo/q_t": 0.31180116534233093, "grad_norm": 32.91140365600586, "learning_rate": 5.564580657695939e-10, "logits/chosen": 0.37182021141052246, "logits/rejected": 0.4036600589752197, "logps/chosen": -424.34112548828125, "logps/ref_chosen": -46.116416931152344, "logps/ref_rejected": -77.92434692382812, "logps/rejected": -709.6861572265625, "loss": 0.823, "margin_dpo/margin_mean": 253.537109375, "margin_dpo/margin_std": 236.66348266601562, "step": 669 }, { "KL/chosen_KL_mean": -376.5809020996094, "KL/mean": -510.88238525390625, "KL/rejected_KL_mean": -645.1838989257812, "KL/std": 223.031494140625, "epoch": 0.9838472834067548, "fcm_dpo/beta": 0.003458178834989667, "fcm_dpo/delta": -0.141901895403862, "fcm_dpo/margin": 268.60302734375, "fcm_dpo/q_t": 0.3073880076408386, "grad_norm": 31.727798461914062, "learning_rate": 4.741678157389739e-10, "logits/chosen": 0.3487994074821472, "logits/rejected": 0.36743998527526855, "logps/chosen": -438.9266357421875, "logps/ref_chosen": -62.34575271606445, "logps/ref_rejected": -96.9405517578125, "logps/rejected": -742.1244506835938, "loss": 0.799, "margin_dpo/margin_mean": 268.60302734375, "margin_dpo/margin_std": 237.01303100585938, "step": 670 }, { "KL/chosen_KL_mean": -434.69549560546875, "KL/mean": -543.616943359375, "KL/rejected_KL_mean": -652.538330078125, "KL/std": 201.36114501953125, "epoch": 0.9853157121879589, "fcm_dpo/beta": 0.0034685127902776003, "fcm_dpo/delta": 0.04600398987531662, "fcm_dpo/margin": 217.8428192138672, "fcm_dpo/q_t": 0.34018510580062866, "grad_norm": 39.71592712402344, "learning_rate": 3.9845280344705245e-10, "logits/chosen": 0.40200304985046387, "logits/rejected": 0.421750009059906, "logps/chosen": -482.69561767578125, "logps/ref_chosen": -48.00010681152344, "logps/ref_rejected": -83.81932067871094, "logps/rejected": -736.357666015625, "loss": 0.9002, "margin_dpo/margin_mean": 217.84280395507812, "margin_dpo/margin_std": 222.2292022705078, "step": 671 }, { "KL/chosen_KL_mean": -462.762939453125, "KL/mean": -562.1483764648438, "KL/rejected_KL_mean": -661.5338134765625, "KL/std": 223.07664489746094, "epoch": 0.986784140969163, "fcm_dpo/beta": 0.003505786880850792, "fcm_dpo/delta": 0.11078579723834991, "fcm_dpo/margin": 198.7708740234375, "fcm_dpo/q_t": 0.3557296097278595, "grad_norm": 42.18230438232422, "learning_rate": 3.293150240547549e-10, "logits/chosen": 0.40828007459640503, "logits/rejected": 0.43640410900115967, "logps/chosen": -521.34619140625, "logps/ref_chosen": -58.58328628540039, "logps/ref_rejected": -93.14015197753906, "logps/rejected": -754.6739501953125, "loss": 0.9532, "margin_dpo/margin_mean": 198.7708740234375, "margin_dpo/margin_std": 234.2972412109375, "step": 672 }, { "KL/chosen_KL_mean": -421.04833984375, "KL/mean": -542.5941772460938, "KL/rejected_KL_mean": -664.1400756835938, "KL/std": 201.49720764160156, "epoch": 0.9882525697503671, "fcm_dpo/beta": 0.0034946207888424397, "fcm_dpo/delta": -0.054121218621730804, "fcm_dpo/margin": 243.0917510986328, "fcm_dpo/q_t": 0.31589680910110474, "grad_norm": 34.77309799194336, "learning_rate": 2.6675629940689504e-10, "logits/chosen": 0.45647066831588745, "logits/rejected": 0.4740258753299713, "logps/chosen": -467.77154541015625, "logps/ref_chosen": -46.72320556640625, "logps/ref_rejected": -85.29623413085938, "logps/rejected": -749.436279296875, "loss": 0.8136, "margin_dpo/margin_mean": 243.0917510986328, "margin_dpo/margin_std": 200.75131225585938, "step": 673 }, { "KL/chosen_KL_mean": -399.45220947265625, "KL/mean": -528.6614990234375, "KL/rejected_KL_mean": -657.8707885742188, "KL/std": 229.89468383789062, "epoch": 0.9897209985315712, "fcm_dpo/beta": 0.0034479242749512196, "fcm_dpo/delta": -0.09983845800161362, "fcm_dpo/margin": 258.4185791015625, "fcm_dpo/q_t": 0.3155099153518677, "grad_norm": 30.887535095214844, "learning_rate": 2.1077827798404725e-10, "logits/chosen": 0.422026664018631, "logits/rejected": 0.44703778624534607, "logps/chosen": -444.897705078125, "logps/ref_chosen": -45.445526123046875, "logps/ref_rejected": -70.04593658447266, "logps/rejected": -727.916748046875, "loss": 0.8132, "margin_dpo/margin_mean": 258.4185791015625, "margin_dpo/margin_std": 237.52337646484375, "step": 674 }, { "KL/chosen_KL_mean": -390.8414306640625, "KL/mean": -529.8683471679688, "KL/rejected_KL_mean": -668.895263671875, "KL/std": 247.90399169921875, "epoch": 0.9911894273127754, "fcm_dpo/beta": 0.003326475154608488, "fcm_dpo/delta": -0.13932178914546967, "fcm_dpo/margin": 278.0538330078125, "fcm_dpo/q_t": 0.3059813976287842, "grad_norm": 27.469951629638672, "learning_rate": 1.6138243485910863e-10, "logits/chosen": 0.3832669258117676, "logits/rejected": 0.4059588313102722, "logps/chosen": -435.0177001953125, "logps/ref_chosen": -44.17628479003906, "logps/ref_rejected": -74.09197998046875, "logps/rejected": -742.9872436523438, "loss": 0.7809, "margin_dpo/margin_mean": 278.0538635253906, "margin_dpo/margin_std": 232.23904418945312, "step": 675 }, { "KL/chosen_KL_mean": -423.63348388671875, "KL/mean": -543.1539306640625, "KL/rejected_KL_mean": -662.6744384765625, "KL/std": 235.30764770507812, "epoch": 0.9926578560939795, "fcm_dpo/beta": 0.0033134541008621454, "fcm_dpo/delta": 0.008331548422574997, "fcm_dpo/margin": 239.04092407226562, "fcm_dpo/q_t": 0.32824352383613586, "grad_norm": 35.013919830322266, "learning_rate": 1.1857007165852472e-10, "logits/chosen": 0.2349829077720642, "logits/rejected": 0.2648879289627075, "logps/chosen": -495.03204345703125, "logps/ref_chosen": -71.39852905273438, "logps/ref_rejected": -88.3587646484375, "logps/rejected": -751.033203125, "loss": 0.8388, "margin_dpo/margin_mean": 239.04090881347656, "margin_dpo/margin_std": 200.05181884765625, "step": 676 }, { "KL/chosen_KL_mean": -425.65728759765625, "KL/mean": -543.5009155273438, "KL/rejected_KL_mean": -661.3445434570312, "KL/std": 198.14552307128906, "epoch": 0.9941262848751835, "fcm_dpo/beta": 0.0033442298881709576, "fcm_dpo/delta": 0.012598922476172447, "fcm_dpo/margin": 235.68731689453125, "fcm_dpo/q_t": 0.3280433714389801, "grad_norm": 32.42127990722656, "learning_rate": 8.23423165278725e-11, "logits/chosen": 0.4454345703125, "logits/rejected": 0.48518845438957214, "logps/chosen": -482.1846923828125, "logps/ref_chosen": -56.527435302734375, "logps/ref_rejected": -78.22654724121094, "logps/rejected": -739.5711059570312, "loss": 0.8385, "margin_dpo/margin_mean": 235.68731689453125, "margin_dpo/margin_std": 192.02017211914062, "step": 677 }, { "KL/chosen_KL_mean": -405.45452880859375, "KL/mean": -529.14453125, "KL/rejected_KL_mean": -652.8345947265625, "KL/std": 241.40565490722656, "epoch": 0.9955947136563876, "fcm_dpo/beta": 0.003331228159368038, "fcm_dpo/delta": -0.026290910318493843, "fcm_dpo/margin": 247.3800811767578, "fcm_dpo/q_t": 0.32946473360061646, "grad_norm": 31.699295043945312, "learning_rate": 5.270012410216185e-11, "logits/chosen": 0.3945927619934082, "logits/rejected": 0.4203825891017914, "logps/chosen": -451.5890197753906, "logps/ref_chosen": -46.13447570800781, "logps/ref_rejected": -80.60462951660156, "logps/rejected": -733.439208984375, "loss": 0.8689, "margin_dpo/margin_mean": 247.3800811767578, "margin_dpo/margin_std": 250.53334045410156, "step": 678 }, { "KL/chosen_KL_mean": -416.02294921875, "KL/mean": -518.5464477539062, "KL/rejected_KL_mean": -621.0698852539062, "KL/std": 217.73243713378906, "epoch": 0.9970631424375918, "fcm_dpo/beta": 0.0033760610967874527, "fcm_dpo/delta": 0.11577928066253662, "fcm_dpo/margin": 205.04698181152344, "fcm_dpo/q_t": 0.3537669777870178, "grad_norm": 43.124732971191406, "learning_rate": 2.9644275480772416e-11, "logits/chosen": 0.3546045124530792, "logits/rejected": 0.3747688829898834, "logps/chosen": -466.31787109375, "logps/ref_chosen": -50.294921875, "logps/ref_rejected": -76.59813690185547, "logps/rejected": -697.6680297851562, "loss": 0.9293, "margin_dpo/margin_mean": 205.04696655273438, "margin_dpo/margin_std": 223.11904907226562, "step": 679 }, { "KL/chosen_KL_mean": -422.438232421875, "KL/mean": -542.1629028320312, "KL/rejected_KL_mean": -661.8876342773438, "KL/std": 242.85482788085938, "epoch": 0.9985315712187959, "fcm_dpo/beta": 0.00338840764015913, "fcm_dpo/delta": -0.012371369637548923, "fcm_dpo/margin": 239.4494171142578, "fcm_dpo/q_t": 0.33219584822654724, "grad_norm": 35.54345703125, "learning_rate": 1.31753782067201e-11, "logits/chosen": 0.28841865062713623, "logits/rejected": 0.3221368193626404, "logps/chosen": -499.35394287109375, "logps/ref_chosen": -76.91569519042969, "logps/ref_rejected": -112.384765625, "logps/rejected": -774.2723388671875, "loss": 0.8816, "margin_dpo/margin_mean": 239.44940185546875, "margin_dpo/margin_std": 254.77374267578125, "step": 680 }, { "KL/chosen_KL_mean": -436.93084716796875, "KL/mean": -549.1124267578125, "KL/rejected_KL_mean": -661.2940673828125, "KL/std": 232.23263549804688, "epoch": 1.0, "fcm_dpo/beta": 0.003436029888689518, "fcm_dpo/delta": 0.029315777122974396, "fcm_dpo/margin": 224.3632354736328, "fcm_dpo/q_t": 0.33981072902679443, "grad_norm": 44.08296585083008, "learning_rate": 3.2938662507808745e-12, "logits/chosen": 0.33995503187179565, "logits/rejected": 0.36489707231521606, "logps/chosen": -497.8881530761719, "logps/ref_chosen": -60.957279205322266, "logps/ref_rejected": -88.55797576904297, "logps/rejected": -749.85205078125, "loss": 0.8963, "margin_dpo/margin_mean": 224.36322021484375, "margin_dpo/margin_std": 229.5245361328125, "step": 681 }, { "epoch": 1.0, "step": 681, "total_flos": 0.0, "train_loss": 0.9160819085158965, "train_runtime": 1729.8503, "train_samples_per_second": 25.203, "train_steps_per_second": 0.394 } ], "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 }