{ "best_global_step": null, "best_metric": null, "best_model_checkpoint": null, "epoch": 0.999244142101285, "eval_steps": 200, "global_step": 661, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0015117157974300832, "fcm_dpo/beta": 0.10000000149011612, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.0013532638549804688, "fcm_dpo/q_t": 0.5000336766242981, "grad_norm": 28.21952247619629, "learning_rate": 0.0, "logits/chosen": 0.13337239623069763, "logits/rejected": 0.12492948770523071, "logps/chosen": -64.5841293334961, "logps/ref_chosen": -64.61280822753906, "logps/ref_rejected": -64.17195129394531, "logps/rejected": -64.14192199707031, "loss": 1.3866, "margin_dpo/margin_mean": -0.0013527870178222656, "margin_dpo/margin_std": 0.2561596930027008, "step": 1 }, { "epoch": 0.007558578987150416, "fcm_dpo/beta": 0.10000000894069672, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.6003847122192383e-05, "fcm_dpo/q_t": 0.4999997317790985, "grad_norm": 29.559865951538086, "learning_rate": 2.9850746268656714e-08, "logits/chosen": 0.0971517488360405, "logits/rejected": 0.0687209814786911, "logps/chosen": -65.35282897949219, "logps/ref_chosen": -65.34695434570312, "logps/ref_rejected": -79.315673828125, "logps/rejected": -79.32157135009766, "loss": 1.3866, "margin_dpo/margin_mean": 1.5869736671447754e-05, "margin_dpo/margin_std": 0.3168628215789795, "step": 5 }, { "epoch": 0.015117157974300832, "fcm_dpo/beta": 0.10000000894069672, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.012913530692458153, "fcm_dpo/q_t": 0.5003229975700378, "grad_norm": 29.694316864013672, "learning_rate": 6.71641791044776e-08, "logits/chosen": 0.10982687771320343, "logits/rejected": 0.07068048417568207, "logps/chosen": -56.66190719604492, "logps/ref_chosen": -56.65692901611328, "logps/ref_rejected": -80.12786865234375, "logps/rejected": -80.11993408203125, "loss": 1.3879, "margin_dpo/margin_mean": -0.012913751415908337, "margin_dpo/margin_std": 0.32729893922805786, "step": 10 }, { "epoch": 0.022675736961451247, "fcm_dpo/beta": 0.10000000894069672, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.008519763126969337, "fcm_dpo/q_t": 0.49978700280189514, "grad_norm": 33.27705001831055, "learning_rate": 1.044776119402985e-07, "logits/chosen": 0.08304646611213684, "logits/rejected": 0.05341915413737297, "logps/chosen": -60.09071731567383, "logps/ref_chosen": -60.09392166137695, "logps/ref_rejected": -78.99056243896484, "logps/rejected": -78.99589538574219, "loss": 1.3857, "margin_dpo/margin_mean": 0.008519655093550682, "margin_dpo/margin_std": 0.3333361744880676, "step": 15 }, { "epoch": 0.030234315948601664, "fcm_dpo/beta": 0.10000000894069672, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.03206338733434677, "fcm_dpo/q_t": 0.49919837713241577, "grad_norm": 29.28687858581543, "learning_rate": 1.4179104477611938e-07, "logits/chosen": 0.10171808302402496, "logits/rejected": 0.07373008131980896, "logps/chosen": -55.454063415527344, "logps/ref_chosen": -55.464561462402344, "logps/ref_rejected": -77.40013122558594, "logps/rejected": -77.42169189453125, "loss": 1.3833, "margin_dpo/margin_mean": 0.03206341713666916, "margin_dpo/margin_std": 0.2828105092048645, "step": 20 }, { "epoch": 0.03779289493575208, "fcm_dpo/beta": 0.10000000894069672, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": -0.018508916720747948, "fcm_dpo/q_t": 0.5004625916481018, "grad_norm": 29.7989559173584, "learning_rate": 1.7910447761194027e-07, "logits/chosen": 0.10036797821521759, "logits/rejected": 0.06887850165367126, "logps/chosen": -60.73334503173828, "logps/ref_chosen": -60.711814880371094, "logps/ref_rejected": -82.71756744384766, "logps/rejected": -82.72056579589844, "loss": 1.3884, "margin_dpo/margin_mean": -0.01850915513932705, "margin_dpo/margin_std": 0.28522950410842896, "step": 25 }, { "epoch": 0.045351473922902494, "fcm_dpo/beta": 0.10000000894069672, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.015936514362692833, "fcm_dpo/q_t": 0.4996017813682556, "grad_norm": 30.30830955505371, "learning_rate": 2.1641791044776117e-07, "logits/chosen": 0.10710334777832031, "logits/rejected": 0.082417331635952, "logps/chosen": -60.89277267456055, "logps/ref_chosen": -60.880210876464844, "logps/ref_rejected": -78.44148254394531, "logps/rejected": -78.46997833251953, "loss": 1.385, "margin_dpo/margin_mean": 0.01593649946153164, "margin_dpo/margin_std": 0.3193960189819336, "step": 30 }, { "epoch": 0.05291005291005291, "fcm_dpo/beta": 0.10000000894069672, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.03081861510872841, "fcm_dpo/q_t": 0.4992298483848572, "grad_norm": 27.896162033081055, "learning_rate": 2.537313432835821e-07, "logits/chosen": 0.07303156703710556, "logits/rejected": 0.04675190895795822, "logps/chosen": -62.2817268371582, "logps/ref_chosen": -62.248138427734375, "logps/ref_rejected": -79.56475830078125, "logps/rejected": -79.62915802001953, "loss": 1.3835, "margin_dpo/margin_mean": 0.030818644911050797, "margin_dpo/margin_std": 0.3220459818840027, "step": 35 }, { "epoch": 0.06046863189720333, "fcm_dpo/beta": 0.10000000894069672, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.05380158871412277, "fcm_dpo/q_t": 0.49865588545799255, "grad_norm": 31.392417907714844, "learning_rate": 2.9104477611940296e-07, "logits/chosen": 0.10881145298480988, "logits/rejected": 0.06401894986629486, "logps/chosen": -58.942840576171875, "logps/ref_chosen": -58.87812423706055, "logps/ref_rejected": -84.22982025146484, "logps/rejected": -84.34834289550781, "loss": 1.3813, "margin_dpo/margin_mean": 0.0538015179336071, "margin_dpo/margin_std": 0.36063262820243835, "step": 40 }, { "epoch": 0.06802721088435375, "fcm_dpo/beta": 0.10000000894069672, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.06139056012034416, "fcm_dpo/q_t": 0.4984658360481262, "grad_norm": 31.959802627563477, "learning_rate": 3.2835820895522385e-07, "logits/chosen": 0.06842530518770218, "logits/rejected": 0.041332852095365524, "logps/chosen": -66.02344512939453, "logps/ref_chosen": -65.88298034667969, "logps/ref_rejected": -83.87881469726562, "logps/rejected": -84.08067321777344, "loss": 1.3805, "margin_dpo/margin_mean": 0.06139082834124565, "margin_dpo/margin_std": 0.3803684115409851, "step": 45 }, { "epoch": 0.07558578987150416, "fcm_dpo/beta": 0.10000000894069672, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.11904527992010117, "fcm_dpo/q_t": 0.49702557921409607, "grad_norm": 28.045490264892578, "learning_rate": 3.6567164179104475e-07, "logits/chosen": 0.09588325768709183, "logits/rejected": 0.06175718456506729, "logps/chosen": -55.35811233520508, "logps/ref_chosen": -55.172386169433594, "logps/ref_rejected": -69.63300323486328, "logps/rejected": -69.9377670288086, "loss": 1.3749, "margin_dpo/margin_mean": 0.1190454512834549, "margin_dpo/margin_std": 0.4321363866329193, "step": 50 }, { "epoch": 0.08314436885865457, "fcm_dpo/beta": 0.10000000894069672, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.20462636649608612, "fcm_dpo/q_t": 0.4948909282684326, "grad_norm": 31.350805282592773, "learning_rate": 4.0298507462686564e-07, "logits/chosen": 0.07329290360212326, "logits/rejected": 0.03742387518286705, "logps/chosen": -57.520660400390625, "logps/ref_chosen": -57.193580627441406, "logps/ref_rejected": -79.69940948486328, "logps/rejected": -80.23112487792969, "loss": 1.3667, "margin_dpo/margin_mean": 0.2046263962984085, "margin_dpo/margin_std": 0.5350508093833923, "step": 55 }, { "epoch": 0.09070294784580499, "fcm_dpo/beta": 0.10000000894069672, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.21316950023174286, "fcm_dpo/q_t": 0.494695246219635, "grad_norm": 29.479215621948242, "learning_rate": 4.4029850746268654e-07, "logits/chosen": 0.11556963622570038, "logits/rejected": 0.08139903843402863, "logps/chosen": -60.59169387817383, "logps/ref_chosen": -60.068870544433594, "logps/ref_rejected": -74.41178894042969, "logps/rejected": -75.14778137207031, "loss": 1.3669, "margin_dpo/margin_mean": 0.2131694257259369, "margin_dpo/margin_std": 0.7954167127609253, "step": 60 }, { "epoch": 0.0982615268329554, "fcm_dpo/beta": 0.10000000894069672, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.2978232204914093, "fcm_dpo/q_t": 0.4925784468650818, "grad_norm": 30.698984146118164, "learning_rate": 4.776119402985074e-07, "logits/chosen": 0.142142653465271, "logits/rejected": 0.11199666559696198, "logps/chosen": -58.89265060424805, "logps/ref_chosen": -58.1558952331543, "logps/ref_rejected": -76.06512451171875, "logps/rejected": -77.09969329833984, "loss": 1.3592, "margin_dpo/margin_mean": 0.2978229522705078, "margin_dpo/margin_std": 0.9595780372619629, "step": 65 }, { "epoch": 0.10582010582010581, "fcm_dpo/beta": 0.10000000894069672, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.3882128596305847, "fcm_dpo/q_t": 0.4903312623500824, "grad_norm": 28.7775936126709, "learning_rate": 4.999860140229787e-07, "logits/chosen": 0.09354963898658752, "logits/rejected": 0.0587652213871479, "logps/chosen": -68.39994812011719, "logps/ref_chosen": -67.35506439208984, "logps/ref_rejected": -82.24962615966797, "logps/rejected": -83.68272399902344, "loss": 1.351, "margin_dpo/margin_mean": 0.38821321725845337, "margin_dpo/margin_std": 1.0907655954360962, "step": 70 }, { "epoch": 0.11337868480725624, "fcm_dpo/beta": 0.10000000894069672, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.5082494020462036, "fcm_dpo/q_t": 0.4873982071876526, "grad_norm": 26.24626922607422, "learning_rate": 4.998286897523808e-07, "logits/chosen": 0.11840841919183731, "logits/rejected": 0.08255789428949356, "logps/chosen": -58.29798126220703, "logps/ref_chosen": -56.86763381958008, "logps/ref_rejected": -72.56938934326172, "logps/rejected": -74.50798797607422, "loss": 1.3414, "margin_dpo/margin_mean": 0.5082495212554932, "margin_dpo/margin_std": 1.3911728858947754, "step": 75 }, { "epoch": 0.12093726379440665, "fcm_dpo/beta": 0.10000000894069672, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 0.6494876742362976, "fcm_dpo/q_t": 0.48396021127700806, "grad_norm": 28.429441452026367, "learning_rate": 4.994966691179711e-07, "logits/chosen": 0.15457963943481445, "logits/rejected": 0.10858809947967529, "logps/chosen": -59.45546340942383, "logps/ref_chosen": -57.687095642089844, "logps/ref_rejected": -78.06813049316406, "logps/rejected": -80.48597717285156, "loss": 1.3301, "margin_dpo/margin_mean": 0.6494876146316528, "margin_dpo/margin_std": 1.6857645511627197, "step": 80 }, { "epoch": 0.12849584278155707, "fcm_dpo/beta": 0.10000000894069672, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.0505290031433105, "fcm_dpo/q_t": 0.47403591871261597, "grad_norm": 26.629859924316406, "learning_rate": 4.989901842900325e-07, "logits/chosen": 0.15848767757415771, "logits/rejected": 0.11160705238580704, "logps/chosen": -59.09556198120117, "logps/ref_chosen": -56.96040725708008, "logps/ref_rejected": -75.22166442871094, "logps/rejected": -78.40734100341797, "loss": 1.2925, "margin_dpo/margin_mean": 1.0505287647247314, "margin_dpo/margin_std": 1.8001108169555664, "step": 85 }, { "epoch": 0.1360544217687075, "fcm_dpo/beta": 0.10000000894069672, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.128682255744934, "fcm_dpo/q_t": 0.47230347990989685, "grad_norm": 29.16546058654785, "learning_rate": 4.983095894354857e-07, "logits/chosen": 0.19552312791347504, "logits/rejected": 0.1472882330417633, "logps/chosen": -60.3253059387207, "logps/ref_chosen": -57.41730499267578, "logps/ref_rejected": -80.87986755371094, "logps/rejected": -84.91656494140625, "loss": 1.2924, "margin_dpo/margin_mean": 1.1286823749542236, "margin_dpo/margin_std": 2.453193426132202, "step": 90 }, { "epoch": 0.1436130007558579, "fcm_dpo/beta": 0.10000000894069672, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.432558298110962, "fcm_dpo/q_t": 0.46547260880470276, "grad_norm": 29.277822494506836, "learning_rate": 4.974553604702332e-07, "logits/chosen": 0.20299391448497772, "logits/rejected": 0.15903373062610626, "logps/chosen": -57.57347869873047, "logps/ref_chosen": -54.08087158203125, "logps/ref_rejected": -76.15860748291016, "logps/rejected": -81.0837631225586, "loss": 1.2756, "margin_dpo/margin_mean": 1.432558298110962, "margin_dpo/margin_std": 3.3028881549835205, "step": 95 }, { "epoch": 0.15117157974300832, "fcm_dpo/beta": 0.10000000894069672, "fcm_dpo/delta": 0.0, "fcm_dpo/margin": 1.2184693813323975, "fcm_dpo/q_t": 0.47156065702438354, "grad_norm": 32.97772216796875, "learning_rate": 4.964280947263676e-07, "logits/chosen": 0.21789178252220154, "logits/rejected": 0.18518556654453278, "logps/chosen": -68.62821197509766, "logps/ref_chosen": -63.875038146972656, "logps/ref_rejected": -82.077880859375, "logps/rejected": -88.04951477050781, "loss": 1.322, "margin_dpo/margin_mean": 1.218469262123108, "margin_dpo/margin_std": 4.569883346557617, "step": 100 }, { "epoch": 0.15873015873015872, "fcm_dpo/beta": 0.10632991790771484, "fcm_dpo/delta": 0.10528752952814102, "fcm_dpo/margin": 1.7072112560272217, "fcm_dpo/q_t": 0.45956483483314514, "grad_norm": 34.242095947265625, "learning_rate": 4.952285105344791e-07, "logits/chosen": 0.23205998539924622, "logits/rejected": 0.18055513501167297, "logps/chosen": -67.48213195800781, "logps/ref_chosen": -62.572479248046875, "logps/ref_rejected": -80.93415069580078, "logps/rejected": -87.55101013183594, "loss": 1.2809, "margin_dpo/margin_mean": 1.7072112560272217, "margin_dpo/margin_std": 4.9097795486450195, "step": 105 }, { "epoch": 0.16628873771730915, "fcm_dpo/beta": 0.1157844066619873, "fcm_dpo/delta": 0.08823363482952118, "fcm_dpo/margin": 1.5210223197937012, "fcm_dpo/q_t": 0.46007877588272095, "grad_norm": 55.85200881958008, "learning_rate": 4.938574467213517e-07, "logits/chosen": 0.19681879878044128, "logits/rejected": 0.1736396998167038, "logps/chosen": -74.27507781982422, "logps/ref_chosen": -68.67534637451172, "logps/ref_rejected": -78.82028198242188, "logps/rejected": -85.94103240966797, "loss": 1.3155, "margin_dpo/margin_mean": 1.5210223197937012, "margin_dpo/margin_std": 5.531653881072998, "step": 110 }, { "epoch": 0.17384731670445955, "fcm_dpo/beta": 0.13739340007305145, "fcm_dpo/delta": 0.2502194941043854, "fcm_dpo/margin": 1.9611819982528687, "fcm_dpo/q_t": 0.441068172454834, "grad_norm": 44.24903106689453, "learning_rate": 4.923158620234019e-07, "logits/chosen": 0.23888036608695984, "logits/rejected": 0.18526621162891388, "logps/chosen": -63.93567657470703, "logps/ref_chosen": -58.65370559692383, "logps/ref_rejected": -81.89688873291016, "logps/rejected": -89.1400375366211, "loss": 1.2282, "margin_dpo/margin_mean": 1.9611822366714478, "margin_dpo/margin_std": 4.587324142456055, "step": 115 }, { "epoch": 0.18140589569160998, "fcm_dpo/beta": 0.1672566831111908, "fcm_dpo/delta": 0.12861795723438263, "fcm_dpo/margin": 2.2945079803466797, "fcm_dpo/q_t": 0.41543060541152954, "grad_norm": 44.8306884765625, "learning_rate": 4.906048344162676e-07, "logits/chosen": 0.24071356654167175, "logits/rejected": 0.19102147221565247, "logps/chosen": -61.27440643310547, "logps/ref_chosen": -56.16423797607422, "logps/ref_rejected": -75.87689971923828, "logps/rejected": -83.28157806396484, "loss": 1.1491, "margin_dpo/margin_mean": 2.2945079803466797, "margin_dpo/margin_std": 4.038813591003418, "step": 120 }, { "epoch": 0.1889644746787604, "fcm_dpo/beta": 0.18768858909606934, "fcm_dpo/delta": 0.08637391030788422, "fcm_dpo/margin": 2.755068302154541, "fcm_dpo/q_t": 0.39426344633102417, "grad_norm": 50.769744873046875, "learning_rate": 4.887255603610184e-07, "logits/chosen": 0.2529417872428894, "logits/rejected": 0.19630905985832214, "logps/chosen": -65.41935729980469, "logps/ref_chosen": -59.744285583496094, "logps/ref_rejected": -86.77314758300781, "logps/rejected": -95.20329284667969, "loss": 1.0755, "margin_dpo/margin_mean": 2.755068302154541, "margin_dpo/margin_std": 4.354895114898682, "step": 125 }, { "epoch": 0.1965230536659108, "fcm_dpo/beta": 0.2045096904039383, "fcm_dpo/delta": 0.08989689499139786, "fcm_dpo/margin": 2.5075488090515137, "fcm_dpo/q_t": 0.3945372402667999, "grad_norm": 59.820316314697266, "learning_rate": 4.866793539675126e-07, "logits/chosen": 0.17989113926887512, "logits/rejected": 0.14967137575149536, "logps/chosen": -70.37138366699219, "logps/ref_chosen": -64.15296936035156, "logps/ref_rejected": -75.17271423339844, "logps/rejected": -83.89866638183594, "loss": 1.1195, "margin_dpo/margin_mean": 2.5075488090515137, "margin_dpo/margin_std": 4.384979248046875, "step": 130 }, { "epoch": 0.20408163265306123, "fcm_dpo/beta": 0.20533373951911926, "fcm_dpo/delta": -0.03491649031639099, "fcm_dpo/margin": 3.0766525268554688, "fcm_dpo/q_t": 0.37100949883461, "grad_norm": 47.41992950439453, "learning_rate": 4.844676460754862e-07, "logits/chosen": 0.22673170268535614, "logits/rejected": 0.18738315999507904, "logps/chosen": -63.47684860229492, "logps/ref_chosen": -57.006690979003906, "logps/ref_rejected": -73.71768188476562, "logps/rejected": -83.26448822021484, "loss": 1.0509, "margin_dpo/margin_mean": 3.076652765274048, "margin_dpo/margin_std": 4.7008538246154785, "step": 135 }, { "epoch": 0.21164021164021163, "fcm_dpo/beta": 0.19947479665279388, "fcm_dpo/delta": -0.0736149474978447, "fcm_dpo/margin": 3.355564594268799, "fcm_dpo/q_t": 0.3690628409385681, "grad_norm": 58.970130920410156, "learning_rate": 4.820919832540181e-07, "logits/chosen": 0.20238320529460907, "logits/rejected": 0.15974575281143188, "logps/chosen": -71.1889419555664, "logps/ref_chosen": -63.36246871948242, "logps/ref_rejected": -79.62621307373047, "logps/rejected": -90.8082504272461, "loss": 1.0837, "margin_dpo/margin_mean": 3.355564594268799, "margin_dpo/margin_std": 5.568287372589111, "step": 140 }, { "epoch": 0.21919879062736206, "fcm_dpo/beta": 0.18338093161582947, "fcm_dpo/delta": -0.04409990459680557, "fcm_dpo/margin": 3.4895870685577393, "fcm_dpo/q_t": 0.3738616108894348, "grad_norm": 56.26732635498047, "learning_rate": 4.795540267200686e-07, "logits/chosen": 0.21315772831439972, "logits/rejected": 0.19285325706005096, "logps/chosen": -73.33685302734375, "logps/ref_chosen": -65.01470184326172, "logps/ref_rejected": -80.49073791503906, "logps/rejected": -92.30247497558594, "loss": 1.1029, "margin_dpo/margin_mean": 3.4895870685577393, "margin_dpo/margin_std": 6.0685014724731445, "step": 145 }, { "epoch": 0.22675736961451248, "fcm_dpo/beta": 0.17459313571453094, "fcm_dpo/delta": -0.12236909568309784, "fcm_dpo/margin": 4.1057610511779785, "fcm_dpo/q_t": 0.35746586322784424, "grad_norm": 39.50666046142578, "learning_rate": 4.768555511768486e-07, "logits/chosen": 0.2202187478542328, "logits/rejected": 0.17578254640102386, "logps/chosen": -67.75491333007812, "logps/ref_chosen": -59.19135284423828, "logps/ref_rejected": -74.0339126586914, "logps/rejected": -86.70323181152344, "loss": 1.0186, "margin_dpo/margin_mean": 4.105760097503662, "margin_dpo/margin_std": 5.879034519195557, "step": 150 }, { "epoch": 0.23431594860166288, "fcm_dpo/beta": 0.1496421992778778, "fcm_dpo/delta": -0.08264624327421188, "fcm_dpo/margin": 4.5022430419921875, "fcm_dpo/q_t": 0.36458662152290344, "grad_norm": 41.55534362792969, "learning_rate": 4.7399844357283393e-07, "logits/chosen": 0.23970897495746613, "logits/rejected": 0.20661358535289764, "logps/chosen": -69.90528869628906, "logps/ref_chosen": -60.93949508666992, "logps/ref_rejected": -74.51151275634766, "logps/rejected": -87.97954559326172, "loss": 1.04, "margin_dpo/margin_mean": 4.5022430419921875, "margin_dpo/margin_std": 6.7144455909729, "step": 155 }, { "epoch": 0.2418745275888133, "fcm_dpo/beta": 0.1405264288187027, "fcm_dpo/delta": -0.07398034632205963, "fcm_dpo/margin": 4.7481889724731445, "fcm_dpo/q_t": 0.3655697703361511, "grad_norm": 35.43960189819336, "learning_rate": 4.7098470178228755e-07, "logits/chosen": 0.22334730625152588, "logits/rejected": 0.186327263712883, "logps/chosen": -68.7593002319336, "logps/ref_chosen": -58.763816833496094, "logps/ref_rejected": -74.94743347167969, "logps/rejected": -89.69109344482422, "loss": 1.0324, "margin_dpo/margin_mean": 4.7481889724731445, "margin_dpo/margin_std": 6.990658760070801, "step": 160 }, { "epoch": 0.2494331065759637, "fcm_dpo/beta": 0.12699048221111298, "fcm_dpo/delta": -0.14860393106937408, "fcm_dpo/margin": 5.813238143920898, "fcm_dpo/q_t": 0.34892064332962036, "grad_norm": 32.44923400878906, "learning_rate": 4.678164332082175e-07, "logits/chosen": 0.26088446378707886, "logits/rejected": 0.20988254249095917, "logps/chosen": -67.06852722167969, "logps/ref_chosen": -55.70417022705078, "logps/ref_rejected": -76.59439849853516, "logps/rejected": -93.77200317382812, "loss": 0.9805, "margin_dpo/margin_mean": 5.813237190246582, "margin_dpo/margin_std": 7.511562347412109, "step": 165 }, { "epoch": 0.25699168556311414, "fcm_dpo/beta": 0.11718620359897614, "fcm_dpo/delta": -0.03449869900941849, "fcm_dpo/margin": 5.392855644226074, "fcm_dpo/q_t": 0.3719424307346344, "grad_norm": 28.188493728637695, "learning_rate": 4.6449585330874425e-07, "logits/chosen": 0.23041145503520966, "logits/rejected": 0.19097033143043518, "logps/chosen": -72.28368377685547, "logps/ref_chosen": -61.169105529785156, "logps/ref_rejected": -77.21674346923828, "logps/rejected": -93.72418212890625, "loss": 1.0359, "margin_dpo/margin_mean": 5.392855644226074, "margin_dpo/margin_std": 7.895319938659668, "step": 170 }, { "epoch": 0.26455026455026454, "fcm_dpo/beta": 0.1070084348320961, "fcm_dpo/delta": -0.1262674331665039, "fcm_dpo/margin": 6.6942877769470215, "fcm_dpo/q_t": 0.35919734835624695, "grad_norm": 29.46046257019043, "learning_rate": 4.6102528404790965e-07, "logits/chosen": 0.2935723066329956, "logits/rejected": 0.23458893597126007, "logps/chosen": -71.26114654541016, "logps/ref_chosen": -59.24176788330078, "logps/ref_rejected": -81.80384826660156, "logps/rejected": -100.51751708984375, "loss": 1.0369, "margin_dpo/margin_mean": 6.6942877769470215, "margin_dpo/margin_std": 9.981403350830078, "step": 175 }, { "epoch": 0.272108843537415, "fcm_dpo/beta": 0.09683612734079361, "fcm_dpo/delta": -0.05345671623945236, "fcm_dpo/margin": 6.68777322769165, "fcm_dpo/q_t": 0.3701228201389313, "grad_norm": 29.12032699584961, "learning_rate": 4.5740715227200897e-07, "logits/chosen": 0.25504690408706665, "logits/rejected": 0.2136983871459961, "logps/chosen": -75.70313262939453, "logps/ref_chosen": -63.24883270263672, "logps/ref_rejected": -79.00736236572266, "logps/rejected": -98.14942932128906, "loss": 1.0588, "margin_dpo/margin_mean": 6.687772274017334, "margin_dpo/margin_std": 10.323027610778809, "step": 180 }, { "epoch": 0.2796674225245654, "fcm_dpo/beta": 0.09016869217157364, "fcm_dpo/delta": -0.05902116373181343, "fcm_dpo/margin": 7.227444648742676, "fcm_dpo/q_t": 0.36848214268684387, "grad_norm": 25.792259216308594, "learning_rate": 4.5364398801258394e-07, "logits/chosen": 0.29077380895614624, "logits/rejected": 0.24018768966197968, "logps/chosen": -68.09630584716797, "logps/ref_chosen": -56.390625, "logps/ref_rejected": -76.81001281738281, "logps/rejected": -95.74314880371094, "loss": 1.0435, "margin_dpo/margin_mean": 7.227444648742676, "margin_dpo/margin_std": 10.630643844604492, "step": 185 }, { "epoch": 0.2872260015117158, "fcm_dpo/beta": 0.08473627269268036, "fcm_dpo/delta": -0.11705086380243301, "fcm_dpo/margin": 8.364612579345703, "fcm_dpo/q_t": 0.35912564396858215, "grad_norm": 29.976728439331055, "learning_rate": 4.4973842271726024e-07, "logits/chosen": 0.28480634093284607, "logits/rejected": 0.23106631636619568, "logps/chosen": -79.46668243408203, "logps/ref_chosen": -68.25389099121094, "logps/ref_rejected": -86.461181640625, "logps/rejected": -106.03858947753906, "loss": 1.0203, "margin_dpo/margin_mean": 8.364612579345703, "margin_dpo/margin_std": 12.093205451965332, "step": 190 }, { "epoch": 0.2947845804988662, "fcm_dpo/beta": 0.07972662150859833, "fcm_dpo/delta": -0.029689926654100418, "fcm_dpo/margin": 7.86168909072876, "fcm_dpo/q_t": 0.37187460064888, "grad_norm": 26.973175048828125, "learning_rate": 4.4569318740967043e-07, "logits/chosen": 0.300616979598999, "logits/rejected": 0.27948108315467834, "logps/chosen": -75.24478912353516, "logps/ref_chosen": -62.1484260559082, "logps/ref_rejected": -71.33458709716797, "logps/rejected": -92.2926254272461, "loss": 1.056, "margin_dpo/margin_mean": 7.86168909072876, "margin_dpo/margin_std": 11.976266860961914, "step": 195 }, { "epoch": 0.30234315948601664, "fcm_dpo/beta": 0.07483974099159241, "fcm_dpo/delta": -0.10010068118572235, "fcm_dpo/margin": 9.267468452453613, "fcm_dpo/q_t": 0.3587101101875305, "grad_norm": 22.538490295410156, "learning_rate": 4.415111107797445e-07, "logits/chosen": 0.3558996617794037, "logits/rejected": 0.2972090244293213, "logps/chosen": -69.53649139404297, "logps/ref_chosen": -56.950096130371094, "logps/ref_rejected": -78.66989135742188, "logps/rejected": -100.52375793457031, "loss": 1.0309, "margin_dpo/margin_mean": 9.267467498779297, "margin_dpo/margin_std": 13.319124221801758, "step": 200 }, { "epoch": 0.30234315948601664, "eval_fcm_dpo/beta": 0.0705723762512207, "eval_fcm_dpo/delta": -0.007554640062153339, "eval_fcm_dpo/margin": 8.440375328063965, "eval_fcm_dpo/q_t": 0.38098761439323425, "eval_logits/chosen": 0.3429543077945709, "eval_logits/rejected": 0.29528242349624634, "eval_logps/chosen": -88.24712371826172, "eval_logps/ref_chosen": -74.85946655273438, "eval_logps/ref_rejected": -79.54898834228516, "eval_logps/rejected": -101.37702178955078, "eval_loss": 0.5478904247283936, "eval_margin_dpo/margin_mean": 8.440375328063965, "eval_margin_dpo/margin_std": 14.14255428314209, "eval_runtime": 38.387, "eval_samples_per_second": 59.994, "eval_steps_per_second": 1.876, "step": 200 }, { "epoch": 0.30990173847316704, "fcm_dpo/beta": 0.06560994684696198, "fcm_dpo/delta": -0.028526384383440018, "fcm_dpo/margin": 9.536072731018066, "fcm_dpo/q_t": 0.3714202344417572, "grad_norm": 28.681243896484375, "learning_rate": 4.3719511720570814e-07, "logits/chosen": 0.35043153166770935, "logits/rejected": 0.2849184572696686, "logps/chosen": -72.63653564453125, "logps/ref_chosen": -57.99428176879883, "logps/ref_rejected": -83.5367431640625, "logps/rejected": -107.71504974365234, "loss": 1.0472, "margin_dpo/margin_mean": 9.536072731018066, "margin_dpo/margin_std": 14.235984802246094, "step": 205 }, { "epoch": 0.31746031746031744, "fcm_dpo/beta": 0.06596539914608002, "fcm_dpo/delta": -0.020397888496518135, "fcm_dpo/margin": 9.372965812683105, "fcm_dpo/q_t": 0.3780348598957062, "grad_norm": 22.18747901916504, "learning_rate": 4.327482247091679e-07, "logits/chosen": 0.3586636185646057, "logits/rejected": 0.3060414493083954, "logps/chosen": -80.08374786376953, "logps/ref_chosen": -63.77195358276367, "logps/ref_rejected": -82.56491088867188, "logps/rejected": -108.24967956542969, "loss": 1.0977, "margin_dpo/margin_mean": 9.372964859008789, "margin_dpo/margin_std": 15.807058334350586, "step": 210 }, { "epoch": 0.3250188964474679, "fcm_dpo/beta": 0.0596235916018486, "fcm_dpo/delta": -0.1393919438123703, "fcm_dpo/margin": 12.262106895446777, "fcm_dpo/q_t": 0.35417699813842773, "grad_norm": 21.598207473754883, "learning_rate": 4.281735428447157e-07, "logits/chosen": 0.3513588309288025, "logits/rejected": 0.28718453645706177, "logps/chosen": -74.48625183105469, "logps/ref_chosen": -60.27800750732422, "logps/ref_rejected": -83.91607666015625, "logps/rejected": -110.38641357421875, "loss": 0.9949, "margin_dpo/margin_mean": 12.262106895446777, "margin_dpo/margin_std": 16.72055435180664, "step": 215 }, { "epoch": 0.3325774754346183, "fcm_dpo/beta": 0.05317453667521477, "fcm_dpo/delta": -0.02484332025051117, "fcm_dpo/margin": 11.684633255004883, "fcm_dpo/q_t": 0.3738202452659607, "grad_norm": 20.055500030517578, "learning_rate": 4.234742705255272e-07, "logits/chosen": 0.356646329164505, "logits/rejected": 0.29619479179382324, "logps/chosen": -76.64110565185547, "logps/ref_chosen": -60.88572311401367, "logps/ref_rejected": -80.1805191040039, "logps/rejected": -107.62054443359375, "loss": 1.055, "margin_dpo/margin_mean": 11.684633255004883, "margin_dpo/margin_std": 17.75079917907715, "step": 220 }, { "epoch": 0.3401360544217687, "fcm_dpo/beta": 0.05095481872558594, "fcm_dpo/delta": -0.07625424116849899, "fcm_dpo/margin": 13.132394790649414, "fcm_dpo/q_t": 0.3655363917350769, "grad_norm": 22.153240203857422, "learning_rate": 4.186536937864752e-07, "logits/chosen": 0.43087267875671387, "logits/rejected": 0.3594560921192169, "logps/chosen": -76.80330657958984, "logps/ref_chosen": -61.02507781982422, "logps/ref_rejected": -91.92439270019531, "logps/rejected": -120.83500671386719, "loss": 1.0157, "margin_dpo/margin_mean": 13.132394790649414, "margin_dpo/margin_std": 18.37344741821289, "step": 225 }, { "epoch": 0.3476946334089191, "fcm_dpo/beta": 0.0496797040104866, "fcm_dpo/delta": -0.028627533465623856, "fcm_dpo/margin": 12.607888221740723, "fcm_dpo/q_t": 0.37395650148391724, "grad_norm": 20.35274314880371, "learning_rate": 4.137151834863213e-07, "logits/chosen": 0.469480037689209, "logits/rejected": 0.4215773642063141, "logps/chosen": -70.93717956542969, "logps/ref_chosen": -54.49797821044922, "logps/ref_rejected": -71.96363830566406, "logps/rejected": -101.01072692871094, "loss": 1.076, "margin_dpo/margin_mean": 12.607889175415039, "margin_dpo/margin_std": 20.359996795654297, "step": 230 }, { "epoch": 0.35525321239606955, "fcm_dpo/beta": 0.04417480155825615, "fcm_dpo/delta": -0.1501634567975998, "fcm_dpo/margin": 14.744340896606445, "fcm_dpo/q_t": 0.36673247814178467, "grad_norm": 18.327699661254883, "learning_rate": 4.08662192950594e-07, "logits/chosen": 0.4181364178657532, "logits/rejected": 0.3918091058731079, "logps/chosen": -82.67752075195312, "logps/ref_chosen": -63.250282287597656, "logps/ref_rejected": -73.09049987792969, "logps/rejected": -107.26206970214844, "loss": 1.0368, "margin_dpo/margin_mean": 14.744340896606445, "margin_dpo/margin_std": 20.939937591552734, "step": 235 }, { "epoch": 0.36281179138321995, "fcm_dpo/beta": 0.040509648621082306, "fcm_dpo/delta": -0.09375976026058197, "fcm_dpo/margin": 17.00118637084961, "fcm_dpo/q_t": 0.35856300592422485, "grad_norm": 15.43008041381836, "learning_rate": 4.0349825555680045e-07, "logits/chosen": 0.4488718509674072, "logits/rejected": 0.3973858654499054, "logps/chosen": -88.68736267089844, "logps/ref_chosen": -65.26150512695312, "logps/ref_rejected": -87.60311126708984, "logps/rejected": -128.0301513671875, "loss": 1.0096, "margin_dpo/margin_mean": 17.00118637084961, "margin_dpo/margin_std": 23.233928680419922, "step": 240 }, { "epoch": 0.37037037037037035, "fcm_dpo/beta": 0.03684638813138008, "fcm_dpo/delta": -0.12539520859718323, "fcm_dpo/margin": 19.555816650390625, "fcm_dpo/q_t": 0.3527432084083557, "grad_norm": 34.119014739990234, "learning_rate": 3.982269822636601e-07, "logits/chosen": 0.4494594633579254, "logits/rejected": 0.41574639081954956, "logps/chosen": -89.88227844238281, "logps/ref_chosen": -65.73170471191406, "logps/ref_rejected": -75.19642639160156, "logps/rejected": -118.90281677246094, "loss": 0.9743, "margin_dpo/margin_mean": 19.555816650390625, "margin_dpo/margin_std": 24.575748443603516, "step": 245 }, { "epoch": 0.3779289493575208, "fcm_dpo/beta": 0.033488351851701736, "fcm_dpo/delta": -0.029287120327353477, "fcm_dpo/margin": 17.215007781982422, "fcm_dpo/q_t": 0.383605420589447, "grad_norm": 19.24226188659668, "learning_rate": 3.9285205908608934e-07, "logits/chosen": 0.5213881731033325, "logits/rejected": 0.5142917633056641, "logps/chosen": -104.99214172363281, "logps/ref_chosen": -70.71224212646484, "logps/ref_rejected": -76.12723541259766, "logps/rejected": -127.62215423583984, "loss": 1.1187, "margin_dpo/margin_mean": 17.215011596679688, "margin_dpo/margin_std": 29.802139282226562, "step": 250 }, { "epoch": 0.3854875283446712, "fcm_dpo/beta": 0.033009327948093414, "fcm_dpo/delta": -0.005891552660614252, "fcm_dpo/margin": 16.929611206054688, "fcm_dpo/q_t": 0.3879493176937103, "grad_norm": 22.598548889160156, "learning_rate": 3.873772445177015e-07, "logits/chosen": 0.5174540281295776, "logits/rejected": 0.4729521870613098, "logps/chosen": -98.29107666015625, "logps/ref_chosen": -61.767662048339844, "logps/ref_rejected": -77.38813018798828, "logps/rejected": -130.8411407470703, "loss": 1.1283, "margin_dpo/margin_mean": 16.929611206054688, "margin_dpo/margin_std": 30.323169708251953, "step": 255 }, { "epoch": 0.3930461073318216, "fcm_dpo/beta": 0.030517011880874634, "fcm_dpo/delta": -0.11403815448284149, "fcm_dpo/margin": 23.142879486083984, "fcm_dpo/q_t": 0.358590304851532, "grad_norm": 20.97575569152832, "learning_rate": 3.818063669026256e-07, "logits/chosen": 0.5084746479988098, "logits/rejected": 0.43368783593177795, "logps/chosen": -100.66755676269531, "logps/ref_chosen": -61.57584762573242, "logps/ref_rejected": -91.87513732910156, "logps/rejected": -154.1096954345703, "loss": 1.0132, "margin_dpo/margin_mean": 23.142879486083984, "margin_dpo/margin_std": 32.55268096923828, "step": 260 }, { "epoch": 0.40060468631897206, "fcm_dpo/beta": 0.029083365574479103, "fcm_dpo/delta": -0.01140249241143465, "fcm_dpo/margin": 20.991168975830078, "fcm_dpo/q_t": 0.37674590945243835, "grad_norm": 18.634904861450195, "learning_rate": 3.7614332175848027e-07, "logits/chosen": 0.47952961921691895, "logits/rejected": 0.42895132303237915, "logps/chosen": -102.08247375488281, "logps/ref_chosen": -65.75422668457031, "logps/ref_rejected": -77.9569320678711, "logps/rejected": -135.27633666992188, "loss": 1.0721, "margin_dpo/margin_mean": 20.991168975830078, "margin_dpo/margin_std": 33.38361740112305, "step": 265 }, { "epoch": 0.40816326530612246, "fcm_dpo/beta": 0.02850642241537571, "fcm_dpo/delta": -0.009921222925186157, "fcm_dpo/margin": 21.323711395263672, "fcm_dpo/q_t": 0.3749794363975525, "grad_norm": 17.992469787597656, "learning_rate": 3.7039206905237656e-07, "logits/chosen": 0.5241806507110596, "logits/rejected": 0.4762350022792816, "logps/chosen": -94.45357513427734, "logps/ref_chosen": -62.27649688720703, "logps/ref_rejected": -76.56950378417969, "logps/rejected": -130.07029724121094, "loss": 1.0539, "margin_dpo/margin_mean": 21.323711395263672, "margin_dpo/margin_std": 31.970006942749023, "step": 270 }, { "epoch": 0.41572184429327286, "fcm_dpo/beta": 0.02760036289691925, "fcm_dpo/delta": -0.06361333280801773, "fcm_dpo/margin": 23.88420295715332, "fcm_dpo/q_t": 0.368137389421463, "grad_norm": 17.147911071777344, "learning_rate": 3.645566304318526e-07, "logits/chosen": 0.5007555484771729, "logits/rejected": 0.45732545852661133, "logps/chosen": -96.15426635742188, "logps/ref_chosen": -61.854393005371094, "logps/ref_rejected": -77.22246551513672, "logps/rejected": -135.4065399169922, "loss": 1.0454, "margin_dpo/margin_mean": 23.88420295715332, "margin_dpo/margin_std": 35.828495025634766, "step": 275 }, { "epoch": 0.42328042328042326, "fcm_dpo/beta": 0.02634030021727085, "fcm_dpo/delta": -0.08169054239988327, "fcm_dpo/margin": 25.740819931030273, "fcm_dpo/q_t": 0.3601287007331848, "grad_norm": 15.523616790771484, "learning_rate": 3.586410864126781e-07, "logits/chosen": 0.5158032178878784, "logits/rejected": 0.4866085648536682, "logps/chosen": -96.00532531738281, "logps/ref_chosen": -61.29896926879883, "logps/ref_rejected": -73.35762023925781, "logps/rejected": -133.80477905273438, "loss": 0.9899, "margin_dpo/margin_mean": 25.740819931030273, "margin_dpo/margin_std": 32.85602951049805, "step": 280 }, { "epoch": 0.4308390022675737, "fcm_dpo/beta": 0.02367565594613552, "fcm_dpo/delta": -0.0362996831536293, "fcm_dpo/margin": 26.715778350830078, "fcm_dpo/q_t": 0.3702591061592102, "grad_norm": 19.280351638793945, "learning_rate": 3.5264957352549375e-07, "logits/chosen": 0.5398426651954651, "logits/rejected": 0.4950627386569977, "logps/chosen": -106.26362609863281, "logps/ref_chosen": -63.435462951660156, "logps/ref_rejected": -79.73661804199219, "logps/rejected": -149.28054809570312, "loss": 1.0277, "margin_dpo/margin_mean": 26.715778350830078, "margin_dpo/margin_std": 37.812416076660156, "step": 285 }, { "epoch": 0.4383975812547241, "fcm_dpo/beta": 0.02178608998656273, "fcm_dpo/delta": -0.09696368873119354, "fcm_dpo/margin": 31.538326263427734, "fcm_dpo/q_t": 0.35935452580451965, "grad_norm": 20.562551498413086, "learning_rate": 3.465862814232821e-07, "logits/chosen": 0.5788189172744751, "logits/rejected": 0.5170078277587891, "logps/chosen": -110.50838470458984, "logps/ref_chosen": -57.696876525878906, "logps/ref_rejected": -79.78132629394531, "logps/rejected": -164.13116455078125, "loss": 0.9951, "margin_dpo/margin_mean": 31.538326263427734, "margin_dpo/margin_std": 42.166587829589844, "step": 290 }, { "epoch": 0.4459561602418745, "fcm_dpo/beta": 0.02080090157687664, "fcm_dpo/delta": -0.02840259112417698, "fcm_dpo/margin": 30.041790008544922, "fcm_dpo/q_t": 0.37336617708206177, "grad_norm": 20.723968505859375, "learning_rate": 3.4045544995169125e-07, "logits/chosen": 0.6665645837783813, "logits/rejected": 0.5920735597610474, "logps/chosen": -118.08138275146484, "logps/ref_chosen": -55.430633544921875, "logps/ref_rejected": -78.1390151977539, "logps/rejected": -170.83157348632812, "loss": 1.0454, "margin_dpo/margin_mean": 30.041790008544922, "margin_dpo/margin_std": 44.420555114746094, "step": 295 }, { "epoch": 0.45351473922902497, "fcm_dpo/beta": 0.020768912509083748, "fcm_dpo/delta": 0.014924813993275166, "fcm_dpo/margin": 28.187641143798828, "fcm_dpo/q_t": 0.381052166223526, "grad_norm": 17.88654327392578, "learning_rate": 3.3426136618426043e-07, "logits/chosen": 0.6102429628372192, "logits/rejected": 0.556972861289978, "logps/chosen": -131.4066925048828, "logps/ref_chosen": -61.207069396972656, "logps/ref_rejected": -75.23294067382812, "logps/rejected": -173.62020874023438, "loss": 1.0841, "margin_dpo/margin_mean": 28.187641143798828, "margin_dpo/margin_std": 45.52865982055664, "step": 300 }, { "epoch": 0.46107331821617537, "fcm_dpo/beta": 0.02134864404797554, "fcm_dpo/delta": -0.011247384361922741, "fcm_dpo/margin": 28.5620059967041, "fcm_dpo/q_t": 0.3759697377681732, "grad_norm": 18.560199737548828, "learning_rate": 3.280083614246217e-07, "logits/chosen": 0.588545024394989, "logits/rejected": 0.5514200925827026, "logps/chosen": -128.7273406982422, "logps/ref_chosen": -63.06663131713867, "logps/ref_rejected": -78.45845031738281, "logps/rejected": -172.68116760253906, "loss": 1.0782, "margin_dpo/margin_mean": 28.5620059967041, "margin_dpo/margin_std": 45.59451675415039, "step": 305 }, { "epoch": 0.46863189720332576, "fcm_dpo/beta": 0.022000327706336975, "fcm_dpo/delta": 0.11391662061214447, "fcm_dpo/margin": 22.363637924194336, "fcm_dpo/q_t": 0.40100401639938354, "grad_norm": 20.21632194519043, "learning_rate": 3.2170080817777257e-07, "logits/chosen": 0.6432641744613647, "logits/rejected": 0.5922056436538696, "logps/chosen": -123.6684799194336, "logps/ref_chosen": -63.60908889770508, "logps/ref_rejected": -74.06394958496094, "logps/rejected": -156.48696899414062, "loss": 1.1536, "margin_dpo/margin_mean": 22.363637924194336, "margin_dpo/margin_std": 42.26737976074219, "step": 310 }, { "epoch": 0.47619047619047616, "fcm_dpo/beta": 0.022742778062820435, "fcm_dpo/delta": -0.050774771720170975, "fcm_dpo/margin": 28.4006290435791, "fcm_dpo/q_t": 0.3682178854942322, "grad_norm": 15.559822082519531, "learning_rate": 3.1534311709253723e-07, "logits/chosen": 0.5904372334480286, "logits/rejected": 0.5356141328811646, "logps/chosen": -112.68599700927734, "logps/ref_chosen": -62.31493377685547, "logps/ref_rejected": -75.07472229003906, "logps/rejected": -153.84640502929688, "loss": 1.0239, "margin_dpo/margin_mean": 28.4006290435791, "margin_dpo/margin_std": 40.05788040161133, "step": 315 }, { "epoch": 0.4837490551776266, "fcm_dpo/beta": 0.02097228541970253, "fcm_dpo/delta": -0.10140693187713623, "fcm_dpo/margin": 31.283824920654297, "fcm_dpo/q_t": 0.3660925030708313, "grad_norm": 17.410728454589844, "learning_rate": 3.0893973387735683e-07, "logits/chosen": 0.5705456137657166, "logits/rejected": 0.5071772336959839, "logps/chosen": -102.21686553955078, "logps/ref_chosen": -55.336036682128906, "logps/ref_rejected": -80.05536651611328, "logps/rejected": -158.2200164794922, "loss": 1.0213, "margin_dpo/margin_mean": 31.283824920654297, "margin_dpo/margin_std": 43.42279052734375, "step": 320 }, { "epoch": 0.491307634164777, "fcm_dpo/beta": 0.019703637808561325, "fcm_dpo/delta": -0.03309765085577965, "fcm_dpo/margin": 31.972396850585938, "fcm_dpo/q_t": 0.36921393871307373, "grad_norm": 15.720254898071289, "learning_rate": 3.0249513619156206e-07, "logits/chosen": 0.5923582315444946, "logits/rejected": 0.5292683243751526, "logps/chosen": -112.33642578125, "logps/ref_chosen": -57.90629959106445, "logps/ref_rejected": -74.2243881225586, "logps/rejected": -160.62692260742188, "loss": 1.0317, "margin_dpo/margin_mean": 31.972393035888672, "margin_dpo/margin_std": 45.57465744018555, "step": 325 }, { "epoch": 0.4988662131519274, "fcm_dpo/beta": 0.02025691233575344, "fcm_dpo/delta": 0.007672114763408899, "fcm_dpo/margin": 29.19550132751465, "fcm_dpo/q_t": 0.381369411945343, "grad_norm": 16.641584396362305, "learning_rate": 2.9601383051430505e-07, "logits/chosen": 0.5546280145645142, "logits/rejected": 0.516182005405426, "logps/chosen": -129.70407104492188, "logps/ref_chosen": -65.17555236816406, "logps/ref_rejected": -78.53681182861328, "logps/rejected": -172.26084899902344, "loss": 1.093, "margin_dpo/margin_mean": 29.19550132751465, "margin_dpo/margin_std": 48.34088134765625, "step": 330 }, { "epoch": 0.5064247921390779, "fcm_dpo/beta": 0.01816193014383316, "fcm_dpo/delta": -0.08961853384971619, "fcm_dpo/margin": 37.44696807861328, "fcm_dpo/q_t": 0.36152511835098267, "grad_norm": 21.882186889648438, "learning_rate": 2.895003489933375e-07, "logits/chosen": 0.6239126324653625, "logits/rejected": 0.5766108632087708, "logps/chosen": -128.90843200683594, "logps/ref_chosen": -62.62797927856445, "logps/ref_rejected": -79.9095458984375, "logps/rejected": -183.636962890625, "loss": 1.0065, "margin_dpo/margin_mean": 37.44696807861328, "margin_dpo/margin_std": 51.33563232421875, "step": 335 }, { "epoch": 0.5139833711262283, "fcm_dpo/beta": 0.018349742516875267, "fcm_dpo/delta": 0.032100338488817215, "fcm_dpo/margin": 31.00588035583496, "fcm_dpo/q_t": 0.38639387488365173, "grad_norm": 17.62742042541504, "learning_rate": 2.8295924627584004e-07, "logits/chosen": 0.5883735418319702, "logits/rejected": 0.5494518280029297, "logps/chosen": -136.67315673828125, "logps/ref_chosen": -61.1064567565918, "logps/ref_rejected": -76.71846008300781, "logps/rejected": -183.29104614257812, "loss": 1.0985, "margin_dpo/margin_mean": 31.00588035583496, "margin_dpo/margin_std": 52.399314880371094, "step": 340 }, { "epoch": 0.5215419501133787, "fcm_dpo/beta": 0.01758972555398941, "fcm_dpo/delta": -0.06735783815383911, "fcm_dpo/margin": 35.31150436401367, "fcm_dpo/q_t": 0.37285932898521423, "grad_norm": 16.583221435546875, "learning_rate": 2.7639509632351927e-07, "logits/chosen": 0.616736888885498, "logits/rejected": 0.5733811855316162, "logps/chosen": -128.89810180664062, "logps/ref_chosen": -60.12370681762695, "logps/ref_rejected": -78.58574676513672, "logps/rejected": -182.67166137695312, "loss": 1.0437, "margin_dpo/margin_mean": 35.311500549316406, "margin_dpo/margin_std": 51.527427673339844, "step": 345 }, { "epoch": 0.5291005291005291, "fcm_dpo/beta": 0.016854315996170044, "fcm_dpo/delta": -0.06805837154388428, "fcm_dpo/margin": 39.38313674926758, "fcm_dpo/q_t": 0.3659011721611023, "grad_norm": 17.990488052368164, "learning_rate": 2.698124892141971e-07, "logits/chosen": 0.6850234866142273, "logits/rejected": 0.6155174970626831, "logps/chosen": -119.6955337524414, "logps/ref_chosen": -55.104461669921875, "logps/ref_rejected": -80.63292694091797, "logps/rejected": -184.60714721679688, "loss": 1.013, "margin_dpo/margin_mean": 39.38313674926758, "margin_dpo/margin_std": 54.5691032409668, "step": 350 }, { "epoch": 0.5366591080876795, "fcm_dpo/beta": 0.015808746218681335, "fcm_dpo/delta": -0.05140886455774307, "fcm_dpo/margin": 40.9948616027832, "fcm_dpo/q_t": 0.364651083946228, "grad_norm": 17.801734924316406, "learning_rate": 2.632160279321328e-07, "logits/chosen": 0.646989643573761, "logits/rejected": 0.5661012530326843, "logps/chosen": -116.357666015625, "logps/ref_chosen": -54.87224197387695, "logps/ref_rejected": -77.01316833496094, "logps/rejected": -179.4934844970703, "loss": 1.0215, "margin_dpo/margin_mean": 40.9948616027832, "margin_dpo/margin_std": 57.23894500732422, "step": 355 }, { "epoch": 0.54421768707483, "fcm_dpo/beta": 0.015229565091431141, "fcm_dpo/delta": -0.004581466782838106, "fcm_dpo/margin": 33.99236297607422, "fcm_dpo/q_t": 0.39209577441215515, "grad_norm": 15.181863784790039, "learning_rate": 2.5661032514931834e-07, "logits/chosen": 0.5896913409233093, "logits/rejected": 0.5458609461784363, "logps/chosen": -131.3365478515625, "logps/ref_chosen": -60.75285720825195, "logps/ref_rejected": -75.21507263183594, "logps/rejected": -179.79110717773438, "loss": 1.1095, "margin_dpo/margin_mean": 33.99236297607422, "margin_dpo/margin_std": 56.61328887939453, "step": 360 }, { "epoch": 0.5517762660619804, "fcm_dpo/beta": 0.01466338336467743, "fcm_dpo/delta": -0.07188267260789871, "fcm_dpo/margin": 45.488773345947266, "fcm_dpo/q_t": 0.3608033359050751, "grad_norm": 13.636979103088379, "learning_rate": 2.5e-07, "logits/chosen": 0.6325476765632629, "logits/rejected": 0.5525530576705933, "logps/chosen": -128.69789123535156, "logps/ref_chosen": -58.56513595581055, "logps/ref_rejected": -84.06403350830078, "logps/rejected": -199.68553161621094, "loss": 0.9882, "margin_dpo/margin_mean": 45.488773345947266, "margin_dpo/margin_std": 58.70988082885742, "step": 365 }, { "epoch": 0.5593348450491308, "fcm_dpo/beta": 0.014749327674508095, "fcm_dpo/delta": 0.050389669835567474, "fcm_dpo/margin": 37.40172576904297, "fcm_dpo/q_t": 0.3838474154472351, "grad_norm": 15.615440368652344, "learning_rate": 2.4338967485068164e-07, "logits/chosen": 0.598479151725769, "logits/rejected": 0.5479581356048584, "logps/chosen": -129.9058837890625, "logps/ref_chosen": -59.443138122558594, "logps/ref_rejected": -75.80937194824219, "logps/rejected": -183.67385864257812, "loss": 1.082, "margin_dpo/margin_mean": 37.40172576904297, "margin_dpo/margin_std": 59.58678436279297, "step": 370 }, { "epoch": 0.5668934240362812, "fcm_dpo/beta": 0.014894585125148296, "fcm_dpo/delta": -0.010102972388267517, "fcm_dpo/margin": 38.15407180786133, "fcm_dpo/q_t": 0.3826829791069031, "grad_norm": 17.60791015625, "learning_rate": 2.3678397206786715e-07, "logits/chosen": 0.641754150390625, "logits/rejected": 0.5882676839828491, "logps/chosen": -131.53549194335938, "logps/ref_chosen": -58.59185028076172, "logps/ref_rejected": -73.7529525756836, "logps/rejected": -184.85067749023438, "loss": 1.0957, "margin_dpo/margin_mean": 38.15407180786133, "margin_dpo/margin_std": 62.604209899902344, "step": 375 }, { "epoch": 0.5744520030234316, "fcm_dpo/beta": 0.014794014394283295, "fcm_dpo/delta": -0.023977603763341904, "fcm_dpo/margin": 42.04156494140625, "fcm_dpo/q_t": 0.3720186650753021, "grad_norm": 14.933080673217773, "learning_rate": 2.3018751078580283e-07, "logits/chosen": 0.6510765552520752, "logits/rejected": 0.5984392166137695, "logps/chosen": -130.0701904296875, "logps/ref_chosen": -58.93424606323242, "logps/ref_rejected": -76.27055358886719, "logps/rejected": -189.4480438232422, "loss": 1.0408, "margin_dpo/margin_mean": 42.04156494140625, "margin_dpo/margin_std": 61.29801559448242, "step": 380 }, { "epoch": 0.582010582010582, "fcm_dpo/beta": 0.014798015356063843, "fcm_dpo/delta": 0.008013037964701653, "fcm_dpo/margin": 36.892799377441406, "fcm_dpo/q_t": 0.3861129879951477, "grad_norm": 14.123330116271973, "learning_rate": 2.2360490367648084e-07, "logits/chosen": 0.6028137803077698, "logits/rejected": 0.5752580761909485, "logps/chosen": -141.14915466308594, "logps/ref_chosen": -66.42684173583984, "logps/ref_rejected": -76.96304321289062, "logps/rejected": -188.5781707763672, "loss": 1.0783, "margin_dpo/margin_mean": 36.892799377441406, "margin_dpo/margin_std": 57.62492752075195, "step": 385 }, { "epoch": 0.5895691609977324, "fcm_dpo/beta": 0.014910424128174782, "fcm_dpo/delta": 0.034819699823856354, "fcm_dpo/margin": 37.98401641845703, "fcm_dpo/q_t": 0.3820473253726959, "grad_norm": 14.307579040527344, "learning_rate": 2.170407537241599e-07, "logits/chosen": 0.6292208433151245, "logits/rejected": 0.5711486339569092, "logps/chosen": -135.46531677246094, "logps/ref_chosen": -60.984214782714844, "logps/ref_rejected": -79.54056549072266, "logps/rejected": -192.00570678710938, "loss": 1.0604, "margin_dpo/margin_mean": 37.98401641845703, "margin_dpo/margin_std": 57.268310546875, "step": 390 }, { "epoch": 0.5971277399848829, "fcm_dpo/beta": 0.014500841498374939, "fcm_dpo/delta": -0.03668956086039543, "fcm_dpo/margin": 43.64439010620117, "fcm_dpo/q_t": 0.3682101368904114, "grad_norm": 18.736286163330078, "learning_rate": 2.104996510066625e-07, "logits/chosen": 0.6559888124465942, "logits/rejected": 0.5947400331497192, "logps/chosen": -128.33447265625, "logps/ref_chosen": -58.30937957763672, "logps/ref_rejected": -80.09587097167969, "logps/rejected": -193.765380859375, "loss": 1.0149, "margin_dpo/margin_mean": 43.64439010620117, "margin_dpo/margin_std": 59.09160232543945, "step": 395 }, { "epoch": 0.6046863189720333, "fcm_dpo/beta": 0.014643101021647453, "fcm_dpo/delta": 0.00806088000535965, "fcm_dpo/margin": 40.27904510498047, "fcm_dpo/q_t": 0.37856990098953247, "grad_norm": 23.970535278320312, "learning_rate": 2.0398616948569493e-07, "logits/chosen": 0.6093511581420898, "logits/rejected": 0.5225298404693604, "logps/chosen": -140.3079833984375, "logps/ref_chosen": -61.39867401123047, "logps/ref_rejected": -89.0177993774414, "logps/rejected": -208.20614624023438, "loss": 1.058, "margin_dpo/margin_mean": 40.27904510498047, "margin_dpo/margin_std": 59.58887481689453, "step": 400 }, { "epoch": 0.6046863189720333, "eval_fcm_dpo/beta": 0.015623129904270172, "eval_fcm_dpo/delta": 0.00515873683616519, "eval_fcm_dpo/margin": 37.30812072753906, "eval_fcm_dpo/q_t": 0.3823583722114563, "eval_logits/chosen": 0.6411985158920288, "eval_logits/rejected": 0.5883597731590271, "eval_logps/chosen": -151.68788146972656, "eval_logps/ref_chosen": -74.85946655273438, "eval_logps/ref_rejected": -79.54898834228516, "eval_logps/rejected": -193.68551635742188, "eval_loss": 0.5441513061523438, "eval_margin_dpo/margin_mean": 37.30812454223633, "eval_margin_dpo/margin_std": 61.142723083496094, "eval_runtime": 38.3592, "eval_samples_per_second": 60.038, "eval_steps_per_second": 1.877, "step": 400 }, { "epoch": 0.6122448979591837, "fcm_dpo/beta": 0.014417248778045177, "fcm_dpo/delta": -0.06634485721588135, "fcm_dpo/margin": 45.74829864501953, "fcm_dpo/q_t": 0.365024596452713, "grad_norm": 14.021449089050293, "learning_rate": 1.975048638084379e-07, "logits/chosen": 0.6662113070487976, "logits/rejected": 0.5925288796424866, "logps/chosen": -130.4202880859375, "logps/ref_chosen": -55.953521728515625, "logps/ref_rejected": -77.67539978027344, "logps/rejected": -197.8904571533203, "loss": 1.0027, "margin_dpo/margin_mean": 45.74829864501953, "margin_dpo/margin_std": 60.512489318847656, "step": 405 }, { "epoch": 0.6198034769463341, "fcm_dpo/beta": 0.013882627710700035, "fcm_dpo/delta": -0.022442538291215897, "fcm_dpo/margin": 44.60148239135742, "fcm_dpo/q_t": 0.36986368894577026, "grad_norm": 13.81552505493164, "learning_rate": 1.9106026612264315e-07, "logits/chosen": 0.6225928068161011, "logits/rejected": 0.5533093214035034, "logps/chosen": -144.36842346191406, "logps/ref_chosen": -63.40419387817383, "logps/ref_rejected": -80.85710144042969, "logps/rejected": -206.4228057861328, "loss": 1.016, "margin_dpo/margin_mean": 44.601478576660156, "margin_dpo/margin_std": 59.1302375793457, "step": 410 }, { "epoch": 0.6273620559334845, "fcm_dpo/beta": 0.013873199000954628, "fcm_dpo/delta": 0.0005700096371583641, "fcm_dpo/margin": 43.196556091308594, "fcm_dpo/q_t": 0.37501293420791626, "grad_norm": 14.810785293579102, "learning_rate": 1.846568829074628e-07, "logits/chosen": 0.6255594491958618, "logits/rejected": 0.5735424757003784, "logps/chosen": -148.52664184570312, "logps/ref_chosen": -57.6942024230957, "logps/ref_rejected": -71.74036407470703, "logps/rejected": -205.7693634033203, "loss": 1.0421, "margin_dpo/margin_mean": 43.19655227661133, "margin_dpo/margin_std": 62.83301544189453, "step": 415 }, { "epoch": 0.6349206349206349, "fcm_dpo/beta": 0.013408740982413292, "fcm_dpo/delta": -0.017916694283485413, "fcm_dpo/margin": 39.38436508178711, "fcm_dpo/q_t": 0.3907091021537781, "grad_norm": 16.092344284057617, "learning_rate": 1.782991918222275e-07, "logits/chosen": 0.6533733606338501, "logits/rejected": 0.6073211431503296, "logps/chosen": -156.04934692382812, "logps/ref_chosen": -59.169517517089844, "logps/ref_rejected": -69.47721099853516, "logps/rejected": -205.7414093017578, "loss": 1.1125, "margin_dpo/margin_mean": 39.38436508178711, "margin_dpo/margin_std": 66.23831176757812, "step": 420 }, { "epoch": 0.6424792139077853, "fcm_dpo/beta": 0.013325588777661324, "fcm_dpo/delta": -0.042791374027729034, "fcm_dpo/margin": 47.987789154052734, "fcm_dpo/q_t": 0.36810848116874695, "grad_norm": 16.099363327026367, "learning_rate": 1.7199163857537824e-07, "logits/chosen": 0.639710545539856, "logits/rejected": 0.5940824747085571, "logps/chosen": -149.20211791992188, "logps/ref_chosen": -58.09320831298828, "logps/ref_rejected": -73.98226165771484, "logps/rejected": -213.0789794921875, "loss": 1.0223, "margin_dpo/margin_mean": 47.987789154052734, "margin_dpo/margin_std": 66.80766296386719, "step": 425 }, { "epoch": 0.6500377928949358, "fcm_dpo/beta": 0.013703095726668835, "fcm_dpo/delta": 0.015814904123544693, "fcm_dpo/margin": 42.556365966796875, "fcm_dpo/q_t": 0.3806520104408264, "grad_norm": 15.242733001708984, "learning_rate": 1.6573863381573954e-07, "logits/chosen": 0.5981285572052002, "logits/rejected": 0.5805391073226929, "logps/chosen": -165.95809936523438, "logps/ref_chosen": -62.7039909362793, "logps/ref_rejected": -74.52284240722656, "logps/rejected": -220.3333282470703, "loss": 1.0933, "margin_dpo/margin_mean": 42.556365966796875, "margin_dpo/margin_std": 70.32002258300781, "step": 430 }, { "epoch": 0.6575963718820862, "fcm_dpo/beta": 0.013115659356117249, "fcm_dpo/delta": -0.03992351144552231, "fcm_dpo/margin": 48.57158660888672, "fcm_dpo/q_t": 0.3666061460971832, "grad_norm": 17.644805908203125, "learning_rate": 1.5954455004830878e-07, "logits/chosen": 0.7035324573516846, "logits/rejected": 0.6597825884819031, "logps/chosen": -146.27212524414062, "logps/ref_chosen": -56.12516403198242, "logps/ref_rejected": -74.36073303222656, "logps/rejected": -213.0792999267578, "loss": 1.025, "margin_dpo/margin_mean": 48.57158660888672, "margin_dpo/margin_std": 67.80796813964844, "step": 435 }, { "epoch": 0.6651549508692366, "fcm_dpo/beta": 0.012779777869582176, "fcm_dpo/delta": -0.02733975648880005, "fcm_dpo/margin": 48.86384201049805, "fcm_dpo/q_t": 0.3719489872455597, "grad_norm": 22.34282684326172, "learning_rate": 1.534137185767178e-07, "logits/chosen": 0.6284548044204712, "logits/rejected": 0.548454225063324, "logps/chosen": -148.30421447753906, "logps/ref_chosen": -55.67548751831055, "logps/ref_rejected": -76.62055206298828, "logps/rejected": -218.11312866210938, "loss": 1.0416, "margin_dpo/margin_mean": 48.86384201049805, "margin_dpo/margin_std": 70.98736572265625, "step": 440 }, { "epoch": 0.672713529856387, "fcm_dpo/beta": 0.011895755305886269, "fcm_dpo/delta": -0.0644448846578598, "fcm_dpo/margin": 51.74351119995117, "fcm_dpo/q_t": 0.3690551221370697, "grad_norm": 16.176162719726562, "learning_rate": 1.473504264745062e-07, "logits/chosen": 0.6225286722183228, "logits/rejected": 0.5559119582176208, "logps/chosen": -155.0991973876953, "logps/ref_chosen": -59.903411865234375, "logps/ref_rejected": -82.02873229980469, "logps/rejected": -228.96804809570312, "loss": 1.0151, "margin_dpo/margin_mean": 51.74351119995117, "margin_dpo/margin_std": 68.27033996582031, "step": 445 }, { "epoch": 0.6802721088435374, "fcm_dpo/beta": 0.011266094632446766, "fcm_dpo/delta": -0.007983550429344177, "fcm_dpo/margin": 50.21624755859375, "fcm_dpo/q_t": 0.3788827955722809, "grad_norm": 16.09433937072754, "learning_rate": 1.4135891358732205e-07, "logits/chosen": 0.643526554107666, "logits/rejected": 0.5692092776298523, "logps/chosen": -150.8534698486328, "logps/ref_chosen": -55.83526611328125, "logps/ref_rejected": -79.63658142089844, "logps/rejected": -224.8710479736328, "loss": 1.0399, "margin_dpo/margin_mean": 50.21624755859375, "margin_dpo/margin_std": 67.95020294189453, "step": 450 }, { "epoch": 0.6878306878306878, "fcm_dpo/beta": 0.01194181852042675, "fcm_dpo/delta": 0.029494991526007652, "fcm_dpo/margin": 47.85871505737305, "fcm_dpo/q_t": 0.3804067373275757, "grad_norm": 14.716300964355469, "learning_rate": 1.354433695681474e-07, "logits/chosen": 0.586762547492981, "logits/rejected": 0.5374394655227661, "logps/chosen": -152.26715087890625, "logps/ref_chosen": -60.59226608276367, "logps/ref_rejected": -73.37936401367188, "logps/rejected": -212.9129638671875, "loss": 1.0577, "margin_dpo/margin_mean": 47.85871505737305, "margin_dpo/margin_std": 71.9669189453125, "step": 455 }, { "epoch": 0.6953892668178382, "fcm_dpo/beta": 0.011895488947629929, "fcm_dpo/delta": -0.04346683993935585, "fcm_dpo/margin": 53.85084915161133, "fcm_dpo/q_t": 0.36482182145118713, "grad_norm": 14.025132179260254, "learning_rate": 1.2960793094762345e-07, "logits/chosen": 0.6461644172668457, "logits/rejected": 0.5591436624526978, "logps/chosen": -149.276611328125, "logps/ref_chosen": -56.21283721923828, "logps/ref_rejected": -83.02075958251953, "logps/rejected": -229.93539428710938, "loss": 0.9967, "margin_dpo/margin_mean": 53.85084915161133, "margin_dpo/margin_std": 68.7022476196289, "step": 460 }, { "epoch": 0.7029478458049887, "fcm_dpo/beta": 0.01112604420632124, "fcm_dpo/delta": -0.010392392054200172, "fcm_dpo/margin": 54.66693115234375, "fcm_dpo/q_t": 0.3700573742389679, "grad_norm": 17.04851722717285, "learning_rate": 1.238566782415197e-07, "logits/chosen": 0.6725102066993713, "logits/rejected": 0.6229192614555359, "logps/chosen": -153.4409942626953, "logps/ref_chosen": -59.0674934387207, "logps/ref_rejected": -74.53498840332031, "logps/rejected": -223.5753936767578, "loss": 1.0133, "margin_dpo/margin_mean": 54.66693115234375, "margin_dpo/margin_std": 70.4762191772461, "step": 465 }, { "epoch": 0.7105064247921391, "fcm_dpo/beta": 0.011645983904600143, "fcm_dpo/delta": 0.009741531684994698, "fcm_dpo/margin": 50.60006332397461, "fcm_dpo/q_t": 0.37510764598846436, "grad_norm": 14.018816947937012, "learning_rate": 1.1819363309737438e-07, "logits/chosen": 0.6600295305252075, "logits/rejected": 0.6102507710456848, "logps/chosen": -155.27413940429688, "logps/ref_chosen": -58.3397331237793, "logps/ref_rejected": -74.33660125732422, "logps/rejected": -221.87109375, "loss": 1.0271, "margin_dpo/margin_mean": 50.600067138671875, "margin_dpo/margin_std": 68.35984802246094, "step": 470 }, { "epoch": 0.7180650037792895, "fcm_dpo/beta": 0.01119951345026493, "fcm_dpo/delta": -0.040131840854883194, "fcm_dpo/margin": 56.87284469604492, "fcm_dpo/q_t": 0.3679283559322357, "grad_norm": 11.378620147705078, "learning_rate": 1.126227554822985e-07, "logits/chosen": 0.66913902759552, "logits/rejected": 0.5974291563034058, "logps/chosen": -144.87710571289062, "logps/ref_chosen": -54.60407638549805, "logps/ref_rejected": -79.94635009765625, "logps/rejected": -227.0922088623047, "loss": 1.0247, "margin_dpo/margin_mean": 56.87284469604492, "margin_dpo/margin_std": 79.54670715332031, "step": 475 }, { "epoch": 0.7256235827664399, "fcm_dpo/beta": 0.011514843441545963, "fcm_dpo/delta": 0.07255177199840546, "fcm_dpo/margin": 46.03236389160156, "fcm_dpo/q_t": 0.3892512619495392, "grad_norm": 18.64552879333496, "learning_rate": 1.0714794091391072e-07, "logits/chosen": 0.6036539077758789, "logits/rejected": 0.582791268825531, "logps/chosen": -167.0009002685547, "logps/ref_chosen": -63.0672492980957, "logps/ref_rejected": -68.59602355957031, "logps/rejected": -218.5620574951172, "loss": 1.1136, "margin_dpo/margin_mean": 46.0323600769043, "margin_dpo/margin_std": 79.20867919921875, "step": 480 }, { "epoch": 0.7331821617535903, "fcm_dpo/beta": 0.011711874976754189, "fcm_dpo/delta": 0.006459618918597698, "fcm_dpo/margin": 50.647342681884766, "fcm_dpo/q_t": 0.37608593702316284, "grad_norm": 16.108354568481445, "learning_rate": 1.0177301773633992e-07, "logits/chosen": 0.6584078073501587, "logits/rejected": 0.5911266207695007, "logps/chosen": -159.75131225585938, "logps/ref_chosen": -58.75799560546875, "logps/ref_rejected": -79.72233581542969, "logps/rejected": -231.3629608154297, "loss": 1.0454, "margin_dpo/margin_mean": 50.647342681884766, "margin_dpo/margin_std": 73.55042266845703, "step": 485 }, { "epoch": 0.7407407407407407, "fcm_dpo/beta": 0.012160698883235455, "fcm_dpo/delta": 0.057119082659482956, "fcm_dpo/margin": 44.87133026123047, "fcm_dpo/q_t": 0.3892955183982849, "grad_norm": 16.322307586669922, "learning_rate": 9.650174444319956e-08, "logits/chosen": 0.685616135597229, "logits/rejected": 0.6174665689468384, "logps/chosen": -169.23971557617188, "logps/ref_chosen": -61.394195556640625, "logps/ref_rejected": -81.1914291381836, "logps/rejected": -233.9082489013672, "loss": 1.1167, "margin_dpo/margin_mean": 44.871334075927734, "margin_dpo/margin_std": 78.93081665039062, "step": 490 }, { "epoch": 0.7482993197278912, "fcm_dpo/beta": 0.012708587571978569, "fcm_dpo/delta": 0.024855732917785645, "fcm_dpo/margin": 45.255348205566406, "fcm_dpo/q_t": 0.3839653432369232, "grad_norm": 21.597793579101562, "learning_rate": 9.133780704940594e-08, "logits/chosen": 0.6221252679824829, "logits/rejected": 0.5517234206199646, "logps/chosen": -163.41339111328125, "logps/ref_chosen": -59.85382843017578, "logps/ref_rejected": -80.63748931884766, "logps/rejected": -229.452392578125, "loss": 1.1054, "margin_dpo/margin_mean": 45.255348205566406, "margin_dpo/margin_std": 77.06855773925781, "step": 495 }, { "epoch": 0.7558578987150416, "fcm_dpo/beta": 0.012341910973191261, "fcm_dpo/delta": -0.06186755746603012, "fcm_dpo/margin": 53.241127014160156, "fcm_dpo/q_t": 0.3657708168029785, "grad_norm": 20.970151901245117, "learning_rate": 8.628481651367875e-08, "logits/chosen": 0.6411687135696411, "logits/rejected": 0.5881538391113281, "logps/chosen": -168.67385864257812, "logps/ref_chosen": -66.17753601074219, "logps/ref_rejected": -83.75955200195312, "logps/rejected": -239.49697875976562, "loss": 1.0368, "margin_dpo/margin_mean": 53.241127014160156, "margin_dpo/margin_std": 78.86491394042969, "step": 500 }, { "epoch": 0.763416477702192, "fcm_dpo/beta": 0.012122569605708122, "fcm_dpo/delta": -0.005300647113472223, "fcm_dpo/margin": 49.808876037597656, "fcm_dpo/q_t": 0.3728008270263672, "grad_norm": 17.493913650512695, "learning_rate": 8.134630621352483e-08, "logits/chosen": 0.629217267036438, "logits/rejected": 0.6063399314880371, "logps/chosen": -159.32550048828125, "logps/ref_chosen": -62.11005401611328, "logps/ref_rejected": -74.64705657958984, "logps/rejected": -221.67138671875, "loss": 1.0325, "margin_dpo/margin_mean": 49.80887222290039, "margin_dpo/margin_std": 68.94776153564453, "step": 505 }, { "epoch": 0.7709750566893424, "fcm_dpo/beta": 0.012009757570922375, "fcm_dpo/delta": -0.01045228075236082, "fcm_dpo/margin": 50.718772888183594, "fcm_dpo/q_t": 0.3756389617919922, "grad_norm": 16.645915985107422, "learning_rate": 7.652572947447272e-08, "logits/chosen": 0.6274048686027527, "logits/rejected": 0.5604012608528137, "logps/chosen": -164.4465789794922, "logps/ref_chosen": -64.42265319824219, "logps/ref_rejected": -87.00096130371094, "logps/rejected": -237.74365234375, "loss": 1.0613, "margin_dpo/margin_mean": 50.718772888183594, "margin_dpo/margin_std": 78.1541976928711, "step": 510 }, { "epoch": 0.7785336356764928, "fcm_dpo/beta": 0.011663762852549553, "fcm_dpo/delta": -0.026412179693579674, "fcm_dpo/margin": 53.50532150268555, "fcm_dpo/q_t": 0.3693719506263733, "grad_norm": 17.784894943237305, "learning_rate": 7.182645715528435e-08, "logits/chosen": 0.6600545644760132, "logits/rejected": 0.6091629862785339, "logps/chosen": -156.29165649414062, "logps/ref_chosen": -58.284393310546875, "logps/ref_rejected": -79.09356689453125, "logps/rejected": -230.6061553955078, "loss": 1.0248, "margin_dpo/margin_mean": 53.50532150268555, "margin_dpo/margin_std": 74.59442901611328, "step": 515 }, { "epoch": 0.7860922146636432, "fcm_dpo/beta": 0.01241991762071848, "fcm_dpo/delta": 0.12249796092510223, "fcm_dpo/margin": 38.903263092041016, "fcm_dpo/q_t": 0.40102332830429077, "grad_norm": 20.327253341674805, "learning_rate": 6.725177529083209e-08, "logits/chosen": 0.6678498983383179, "logits/rejected": 0.623926043510437, "logps/chosen": -163.60305786132812, "logps/ref_chosen": -61.03638458251953, "logps/ref_rejected": -72.15824890136719, "logps/rejected": -213.628173828125, "loss": 1.1589, "margin_dpo/margin_mean": 38.903263092041016, "margin_dpo/margin_std": 75.42561340332031, "step": 520 }, { "epoch": 0.7936507936507936, "fcm_dpo/beta": 0.012936798855662346, "fcm_dpo/delta": 0.010262085124850273, "fcm_dpo/margin": 45.59959411621094, "fcm_dpo/q_t": 0.37764960527420044, "grad_norm": 21.973684310913086, "learning_rate": 6.280488279429185e-08, "logits/chosen": 0.5573867559432983, "logits/rejected": 0.5065245032310486, "logps/chosen": -167.7998046875, "logps/ref_chosen": -68.02732849121094, "logps/ref_rejected": -85.41429901123047, "logps/rejected": -230.786376953125, "loss": 1.0859, "margin_dpo/margin_mean": 45.59960174560547, "margin_dpo/margin_std": 74.41303253173828, "step": 525 }, { "epoch": 0.8012093726379441, "fcm_dpo/beta": 0.013767753727734089, "fcm_dpo/delta": 0.014420712366700172, "fcm_dpo/margin": 42.34447479248047, "fcm_dpo/q_t": 0.3815748989582062, "grad_norm": 23.141691207885742, "learning_rate": 5.848888922025552e-08, "logits/chosen": 0.6335071325302124, "logits/rejected": 0.5626540184020996, "logps/chosen": -162.35891723632812, "logps/ref_chosen": -58.67436599731445, "logps/ref_rejected": -79.38807678222656, "logps/rejected": -225.41708374023438, "loss": 1.111, "margin_dpo/margin_mean": 42.34447479248047, "margin_dpo/margin_std": 71.12373352050781, "step": 530 }, { "epoch": 0.8087679516250945, "fcm_dpo/beta": 0.013212913647294044, "fcm_dpo/delta": -0.020001692697405815, "fcm_dpo/margin": 46.78357696533203, "fcm_dpo/q_t": 0.3747491240501404, "grad_norm": 20.625534057617188, "learning_rate": 5.430681259032957e-08, "logits/chosen": 0.624588131904602, "logits/rejected": 0.5618466138839722, "logps/chosen": -157.4933319091797, "logps/ref_chosen": -57.640098571777344, "logps/ref_rejected": -77.25399780273438, "logps/rejected": -223.8908233642578, "loss": 1.0721, "margin_dpo/margin_mean": 46.78357696533203, "margin_dpo/margin_std": 74.61866760253906, "step": 535 }, { "epoch": 0.8163265306122449, "fcm_dpo/beta": 0.012540744617581367, "fcm_dpo/delta": -0.0733686089515686, "fcm_dpo/margin": 53.17271041870117, "fcm_dpo/q_t": 0.3620644211769104, "grad_norm": 17.126371383666992, "learning_rate": 5.026157728273966e-08, "logits/chosen": 0.6427196860313416, "logits/rejected": 0.5600071549415588, "logps/chosen": -163.82118225097656, "logps/ref_chosen": -60.17341995239258, "logps/ref_rejected": -85.50316619873047, "logps/rejected": -242.32363891601562, "loss": 0.989, "margin_dpo/margin_mean": 53.17271041870117, "margin_dpo/margin_std": 68.73289489746094, "step": 540 }, { "epoch": 0.8238851095993953, "fcm_dpo/beta": 0.011654900386929512, "fcm_dpo/delta": -0.05292552709579468, "fcm_dpo/margin": 55.58978271484375, "fcm_dpo/q_t": 0.3632128834724426, "grad_norm": 15.992572784423828, "learning_rate": 4.635601198741607e-08, "logits/chosen": 0.6579868793487549, "logits/rejected": 0.6033874750137329, "logps/chosen": -153.53955078125, "logps/ref_chosen": -56.985809326171875, "logps/ref_rejected": -73.21353912353516, "logps/rejected": -225.3570556640625, "loss": 1.0015, "margin_dpo/margin_mean": 55.58978271484375, "margin_dpo/margin_std": 72.88048553466797, "step": 545 }, { "epoch": 0.8314436885865457, "fcm_dpo/beta": 0.011842542327940464, "fcm_dpo/delta": 0.04592089354991913, "fcm_dpo/margin": 43.93280792236328, "fcm_dpo/q_t": 0.39010632038116455, "grad_norm": 18.3736629486084, "learning_rate": 4.259284772799099e-08, "logits/chosen": 0.6481934189796448, "logits/rejected": 0.5963156223297119, "logps/chosen": -161.9949188232422, "logps/ref_chosen": -59.600929260253906, "logps/ref_rejected": -75.24870300292969, "logps/rejected": -221.5755157470703, "loss": 1.1012, "margin_dpo/margin_mean": 43.93280792236328, "margin_dpo/margin_std": 72.24934387207031, "step": 550 }, { "epoch": 0.8390022675736961, "fcm_dpo/beta": 0.012416016310453415, "fcm_dpo/delta": 0.019648931920528412, "fcm_dpo/margin": 46.74137496948242, "fcm_dpo/q_t": 0.3777768015861511, "grad_norm": 18.58342170715332, "learning_rate": 3.89747159520904e-08, "logits/chosen": 0.62873375415802, "logits/rejected": 0.576077401638031, "logps/chosen": -169.8580322265625, "logps/ref_chosen": -63.578895568847656, "logps/ref_rejected": -78.87867736816406, "logps/rejected": -231.899169921875, "loss": 1.0686, "margin_dpo/margin_mean": 46.74137496948242, "margin_dpo/margin_std": 72.33360290527344, "step": 555 }, { "epoch": 0.8465608465608465, "fcm_dpo/beta": 0.012055947445333004, "fcm_dpo/delta": -0.05154518038034439, "fcm_dpo/margin": 53.696868896484375, "fcm_dpo/q_t": 0.3679019510746002, "grad_norm": 16.838607788085938, "learning_rate": 3.550414669125573e-08, "logits/chosen": 0.6742048263549805, "logits/rejected": 0.6106564402580261, "logps/chosen": -165.71080017089844, "logps/ref_chosen": -58.651512145996094, "logps/ref_rejected": -78.67181396484375, "logps/rejected": -239.427978515625, "loss": 1.0164, "margin_dpo/margin_mean": 53.696868896484375, "margin_dpo/margin_std": 73.68684387207031, "step": 560 }, { "epoch": 0.854119425547997, "fcm_dpo/beta": 0.011463714763522148, "fcm_dpo/delta": -0.03745627775788307, "fcm_dpo/margin": 55.323936462402344, "fcm_dpo/q_t": 0.3676683008670807, "grad_norm": 15.158285140991211, "learning_rate": 3.218356679178252e-08, "logits/chosen": 0.6555038690567017, "logits/rejected": 0.6098966002464294, "logps/chosen": -162.03512573242188, "logps/ref_chosen": -60.3114128112793, "logps/ref_rejected": -78.25270080566406, "logps/rejected": -235.3003692626953, "loss": 1.0158, "margin_dpo/margin_mean": 55.323936462402344, "margin_dpo/margin_std": 75.325927734375, "step": 565 }, { "epoch": 0.8616780045351474, "fcm_dpo/beta": 0.011363612487912178, "fcm_dpo/delta": -0.02258976735174656, "fcm_dpo/margin": 54.51459503173828, "fcm_dpo/q_t": 0.3701397478580475, "grad_norm": 13.35988998413086, "learning_rate": 2.9015298217712453e-08, "logits/chosen": 0.635320782661438, "logits/rejected": 0.5687986612319946, "logps/chosen": -158.89346313476562, "logps/ref_chosen": -57.752410888671875, "logps/ref_rejected": -76.99858093261719, "logps/rejected": -232.6542205810547, "loss": 1.0243, "margin_dpo/margin_mean": 54.51459884643555, "margin_dpo/margin_std": 73.74528503417969, "step": 570 }, { "epoch": 0.8692365835222978, "fcm_dpo/beta": 0.01154360268265009, "fcm_dpo/delta": 0.05699888616800308, "fcm_dpo/margin": 47.14950180053711, "fcm_dpo/q_t": 0.3850955665111542, "grad_norm": 16.961835861206055, "learning_rate": 2.600155642716606e-08, "logits/chosen": 0.6573041677474976, "logits/rejected": 0.6165328025817871, "logps/chosen": -169.44454956054688, "logps/ref_chosen": -63.61958694458008, "logps/ref_rejected": -79.51353454589844, "logps/rejected": -232.48800659179688, "loss": 1.0768, "margin_dpo/margin_mean": 47.14950180053711, "margin_dpo/margin_std": 72.31434631347656, "step": 575 }, { "epoch": 0.8767951625094482, "fcm_dpo/beta": 0.011755215004086494, "fcm_dpo/delta": 0.009137720800936222, "fcm_dpo/margin": 50.234230041503906, "fcm_dpo/q_t": 0.3770310878753662, "grad_norm": 17.833290100097656, "learning_rate": 2.3144448823151392e-08, "logits/chosen": 0.6418476700782776, "logits/rejected": 0.592076301574707, "logps/chosen": -157.65115356445312, "logps/ref_chosen": -57.3541145324707, "logps/ref_rejected": -73.14434051513672, "logps/rejected": -223.6756134033203, "loss": 1.056, "margin_dpo/margin_mean": 50.234230041503906, "margin_dpo/margin_std": 74.81739807128906, "step": 580 }, { "epoch": 0.8843537414965986, "fcm_dpo/beta": 0.011965381912887096, "fcm_dpo/delta": 0.03466043993830681, "fcm_dpo/margin": 47.38968276977539, "fcm_dpo/q_t": 0.3831913471221924, "grad_norm": 16.188016891479492, "learning_rate": 2.044597327993153e-08, "logits/chosen": 0.692867636680603, "logits/rejected": 0.6268293261528015, "logps/chosen": -160.28024291992188, "logps/ref_chosen": -56.0127067565918, "logps/ref_rejected": -77.16522216796875, "logps/rejected": -228.8224334716797, "loss": 1.0892, "margin_dpo/margin_mean": 47.38968276977539, "margin_dpo/margin_std": 77.33354187011719, "step": 585 }, { "epoch": 0.891912320483749, "fcm_dpo/beta": 0.01196371577680111, "fcm_dpo/delta": -0.05104018375277519, "fcm_dpo/margin": 54.1452751159668, "fcm_dpo/q_t": 0.36301979422569275, "grad_norm": 15.689112663269043, "learning_rate": 1.7908016745981856e-08, "logits/chosen": 0.6593881249427795, "logits/rejected": 0.6252022981643677, "logps/chosen": -164.38351440429688, "logps/ref_chosen": -60.5894660949707, "logps/ref_rejected": -74.34771728515625, "logps/rejected": -232.2870635986328, "loss": 1.0103, "margin_dpo/margin_mean": 54.1452751159668, "margin_dpo/margin_std": 73.66951751708984, "step": 590 }, { "epoch": 0.8994708994708994, "fcm_dpo/beta": 0.011334089562296867, "fcm_dpo/delta": -0.034378211945295334, "fcm_dpo/margin": 55.64763259887695, "fcm_dpo/q_t": 0.3687277138233185, "grad_norm": 15.465320587158203, "learning_rate": 1.553235392451377e-08, "logits/chosen": 0.6757582426071167, "logits/rejected": 0.5952944159507751, "logps/chosen": -151.85220336914062, "logps/ref_chosen": -54.77838897705078, "logps/ref_rejected": -78.102783203125, "logps/rejected": -230.82424926757812, "loss": 1.0279, "margin_dpo/margin_mean": 55.64763259887695, "margin_dpo/margin_std": 77.98216247558594, "step": 595 }, { "epoch": 0.9070294784580499, "fcm_dpo/beta": 0.012342174537479877, "fcm_dpo/delta": 0.13277825713157654, "fcm_dpo/margin": 38.14574432373047, "fcm_dpo/q_t": 0.4049660563468933, "grad_norm": 20.803640365600586, "learning_rate": 1.3320646032487393e-08, "logits/chosen": 0.668152928352356, "logits/rejected": 0.6297565698623657, "logps/chosen": -166.8023681640625, "logps/ref_chosen": -58.45500564575195, "logps/ref_rejected": -70.7367172241211, "logps/rejected": -217.22982788085938, "loss": 1.1526, "margin_dpo/margin_mean": 38.14574432373047, "margin_dpo/margin_std": 72.89844512939453, "step": 600 }, { "epoch": 0.9070294784580499, "eval_fcm_dpo/beta": 0.012666304595768452, "eval_fcm_dpo/delta": -0.008464168757200241, "eval_fcm_dpo/margin": 46.96513366699219, "eval_fcm_dpo/q_t": 0.3796946406364441, "eval_logits/chosen": 0.6470546722412109, "eval_logits/rejected": 0.5976328253746033, "eval_logps/chosen": -176.9747314453125, "eval_logps/ref_chosen": -74.85946655273438, "eval_logps/ref_rejected": -79.54898834228516, "eval_logps/rejected": -228.62939453125, "eval_loss": 0.5427461266517639, "eval_margin_dpo/margin_mean": 46.96513366699219, "eval_margin_dpo/margin_std": 76.42948150634766, "eval_runtime": 38.2644, "eval_samples_per_second": 60.187, "eval_steps_per_second": 1.882, "step": 600 }, { "epoch": 0.9145880574452003, "fcm_dpo/beta": 0.01174403727054596, "fcm_dpo/delta": -0.022130563855171204, "fcm_dpo/margin": 52.78753662109375, "fcm_dpo/q_t": 0.37097370624542236, "grad_norm": 17.305164337158203, "learning_rate": 1.1274439638981532e-08, "logits/chosen": 0.651776909828186, "logits/rejected": 0.6054097414016724, "logps/chosen": -154.97470092773438, "logps/ref_chosen": -59.87483596801758, "logps/ref_rejected": -75.75318908691406, "logps/rejected": -223.6405792236328, "loss": 1.0357, "margin_dpo/margin_mean": 52.78753662109375, "margin_dpo/margin_std": 76.32707214355469, "step": 605 }, { "epoch": 0.9221466364323507, "fcm_dpo/beta": 0.011598466895520687, "fcm_dpo/delta": -0.03721737116575241, "fcm_dpo/margin": 54.65764236450195, "fcm_dpo/q_t": 0.3674970269203186, "grad_norm": 14.85545825958252, "learning_rate": 9.395165583732379e-09, "logits/chosen": 0.6490997672080994, "logits/rejected": 0.6031283140182495, "logps/chosen": -160.9161376953125, "logps/ref_chosen": -60.35883712768555, "logps/ref_rejected": -81.3543930053711, "logps/rejected": -236.5693359375, "loss": 1.0079, "margin_dpo/margin_mean": 54.65764236450195, "margin_dpo/margin_std": 72.39617919921875, "step": 610 }, { "epoch": 0.9297052154195011, "fcm_dpo/beta": 0.011334126815199852, "fcm_dpo/delta": -0.02821548655629158, "fcm_dpo/margin": 55.235130310058594, "fcm_dpo/q_t": 0.36660271883010864, "grad_norm": 16.21509552001953, "learning_rate": 7.684137976598088e-09, "logits/chosen": 0.6232366561889648, "logits/rejected": 0.5847585797309875, "logps/chosen": -159.66122436523438, "logps/ref_chosen": -59.17219161987305, "logps/ref_rejected": -79.92167663574219, "logps/rejected": -235.64584350585938, "loss": 1.0121, "margin_dpo/margin_mean": 55.235130310058594, "margin_dpo/margin_std": 73.58882141113281, "step": 615 }, { "epoch": 0.9372637944066515, "fcm_dpo/beta": 0.011340516619384289, "fcm_dpo/delta": 0.013897893019020557, "fcm_dpo/margin": 48.57066345214844, "fcm_dpo/q_t": 0.38371890783309937, "grad_norm": 18.236835479736328, "learning_rate": 6.142553278648238e-09, "logits/chosen": 0.6844276189804077, "logits/rejected": 0.6169918775558472, "logps/chosen": -159.17074584960938, "logps/ref_chosen": -58.052696228027344, "logps/ref_rejected": -78.37252807617188, "logps/rejected": -228.0612335205078, "loss": 1.0546, "margin_dpo/margin_mean": 48.57066345214844, "margin_dpo/margin_std": 69.87725830078125, "step": 620 }, { "epoch": 0.9448223733938019, "fcm_dpo/beta": 0.011483758687973022, "fcm_dpo/delta": 0.0366099514067173, "fcm_dpo/margin": 49.20580291748047, "fcm_dpo/q_t": 0.38207700848579407, "grad_norm": 15.810871124267578, "learning_rate": 4.7714894655209174e-09, "logits/chosen": 0.6805665493011475, "logits/rejected": 0.5953696966171265, "logps/chosen": -160.55422973632812, "logps/ref_chosen": -56.957862854003906, "logps/ref_rejected": -82.68255615234375, "logps/rejected": -235.48471069335938, "loss": 1.0589, "margin_dpo/margin_mean": 49.20580291748047, "margin_dpo/margin_std": 73.78382873535156, "step": 625 }, { "epoch": 0.9523809523809523, "fcm_dpo/beta": 0.011226166971027851, "fcm_dpo/delta": -0.060714829713106155, "fcm_dpo/margin": 58.395973205566406, "fcm_dpo/q_t": 0.3628186285495758, "grad_norm": 12.203248023986816, "learning_rate": 3.5719052736323806e-09, "logits/chosen": 0.6524108648300171, "logits/rejected": 0.5797510743141174, "logps/chosen": -154.64805603027344, "logps/ref_chosen": -56.71510696411133, "logps/ref_rejected": -82.94544219970703, "logps/rejected": -239.2743682861328, "loss": 1.0058, "margin_dpo/margin_mean": 58.395965576171875, "margin_dpo/margin_std": 77.0825424194336, "step": 630 }, { "epoch": 0.9599395313681028, "fcm_dpo/beta": 0.010768270120024681, "fcm_dpo/delta": 0.03129608556628227, "fcm_dpo/margin": 49.29106521606445, "fcm_dpo/q_t": 0.38956043124198914, "grad_norm": 15.734051704406738, "learning_rate": 2.5446395297668287e-09, "logits/chosen": 0.6613369584083557, "logits/rejected": 0.6100578904151917, "logps/chosen": -160.97386169433594, "logps/ref_chosen": -59.33793258666992, "logps/ref_rejected": -75.01703643798828, "logps/rejected": -225.9440155029297, "loss": 1.0885, "margin_dpo/margin_mean": 49.29106521606445, "margin_dpo/margin_std": 76.83003234863281, "step": 635 }, { "epoch": 0.9674981103552532, "fcm_dpo/beta": 0.011434816755354404, "fcm_dpo/delta": 0.013064498081803322, "fcm_dpo/margin": 51.36224365234375, "fcm_dpo/q_t": 0.3775227963924408, "grad_norm": 15.347606658935547, "learning_rate": 1.690410564514244e-09, "logits/chosen": 0.6572569012641907, "logits/rejected": 0.5778030753135681, "logps/chosen": -165.37149047851562, "logps/ref_chosen": -58.1605339050293, "logps/ref_rejected": -79.85365295410156, "logps/rejected": -238.42684936523438, "loss": 1.0496, "margin_dpo/margin_mean": 51.36224365234375, "margin_dpo/margin_std": 76.05950927734375, "step": 640 }, { "epoch": 0.9750566893424036, "fcm_dpo/beta": 0.011937716044485569, "fcm_dpo/delta": 0.06634578108787537, "fcm_dpo/margin": 44.95128631591797, "fcm_dpo/q_t": 0.3872304856777191, "grad_norm": 16.9814453125, "learning_rate": 1.0098157099674987e-09, "logits/chosen": 0.6469696760177612, "logits/rejected": 0.6165892481803894, "logps/chosen": -168.5662384033203, "logps/ref_chosen": -63.45180130004883, "logps/ref_rejected": -74.18285369873047, "logps/rejected": -224.2485809326172, "loss": 1.0905, "margin_dpo/margin_mean": 44.95128631591797, "margin_dpo/margin_std": 72.61418914794922, "step": 645 }, { "epoch": 0.982615268329554, "fcm_dpo/beta": 0.011954518966376781, "fcm_dpo/delta": -0.033706896007061005, "fcm_dpo/margin": 52.76036834716797, "fcm_dpo/q_t": 0.36990886926651, "grad_norm": 17.623476028442383, "learning_rate": 5.033308820289184e-10, "logits/chosen": 0.6857269406318665, "logits/rejected": 0.6130499243736267, "logps/chosen": -168.08609008789062, "logps/ref_chosen": -59.75496292114258, "logps/ref_rejected": -84.31481170654297, "logps/rejected": -245.4062957763672, "loss": 1.0462, "margin_dpo/margin_mean": 52.76036834716797, "margin_dpo/margin_std": 78.340087890625, "step": 650 }, { "epoch": 0.9901738473167044, "fcm_dpo/beta": 0.012021342292428017, "fcm_dpo/delta": -0.0002914905489888042, "fcm_dpo/margin": 49.76955032348633, "fcm_dpo/q_t": 0.37682992219924927, "grad_norm": 14.043972969055176, "learning_rate": 1.7131024761923852e-10, "logits/chosen": 0.6599605679512024, "logits/rejected": 0.5831471681594849, "logps/chosen": -160.75221252441406, "logps/ref_chosen": -57.817848205566406, "logps/ref_rejected": -79.81755065917969, "logps/rejected": -232.521484375, "loss": 1.0505, "margin_dpo/margin_mean": 49.76955032348633, "margin_dpo/margin_std": 72.32919311523438, "step": 655 }, { "epoch": 0.9977324263038548, "fcm_dpo/beta": 0.01167194452136755, "fcm_dpo/delta": -0.019878769293427467, "fcm_dpo/margin": 52.92683029174805, "fcm_dpo/q_t": 0.370964378118515, "grad_norm": 16.102828979492188, "learning_rate": 1.3985977021235829e-11, "logits/chosen": 0.7206791639328003, "logits/rejected": 0.6511392593383789, "logps/chosen": -165.01748657226562, "logps/ref_chosen": -59.12651443481445, "logps/ref_rejected": -79.42085266113281, "logps/rejected": -238.2386474609375, "loss": 1.0408, "margin_dpo/margin_mean": 52.92683029174805, "margin_dpo/margin_std": 77.05123138427734, "step": 660 }, { "epoch": 0.999244142101285, "step": 661, "total_flos": 0.0, "train_loss": 1.103111317586971, "train_runtime": 1749.744, "train_samples_per_second": 24.196, "train_steps_per_second": 0.378 } ], "logging_steps": 5, "max_steps": 661, "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 }